Exploring the Benefits and Limitations of Digital Twin Technology in Building Energy

: Buildings consume a signiﬁcant amount of energy throughout their lifecycle; Thus, sustainable energy management is crucial for all buildings, and controlling energy consumption has become increasingly important for achieving sustainable construction. Digital twin (DT) technology, which lies at the core of Industry 4.0, has gained widespread adoption in various ﬁelds, including building energy analysis. With the ability to monitor, optimize


Introduction
The main question and objective of this study is to deeply explore the application potential of digital twin technology in building energy management. Especially in the context of Building Information Modeling (BIM) and Digital Twins (DT), how to improve energy efficiency and achieve sustainable building energy management by monitoring, optimizing, and predicting building energy consumption in real time. This is also the literature gap that this study aims to address. The motivation behind the research is to drive improvements in energy efficiency and sustainability in buildings. Advances in building energy analytics and machine learning technologies are expected to improve forecasting accuracy, optimize energy efficiency, and improve the management processes to achieve sustainable development goals.
The application and importance of digital twin technology in the field of building energy management have been widely recognized. This innovative technology provides managers with an in-depth understanding of building energy consumption patterns by monitoring the energy usage of building facilities in real time, such as lighting systems, power consumption, and HVAC usage, to achieve energy use improvement and improvement optimization.
The innovation of this study lies in the in-depth exploration of the application potential of digital twin technology in building energy management. The unique contribution of this study to the field of investigation is that it not only helps us to better understand and manage energy use in buildings, driving energy efficiency, and environmental goals but also significantly improves the economics of buildings. Therefore, digital twin technology should be fully utilized and developed to achieve higher efficiency and environmental protection in building energy management.

Definition and Development History of Digital Twin
Digital twin technology connects a physical entity with its virtual model in realtime, thanks to its comprehensive and sophisticated capabilities. Using this technology, various functions can be realized, including simulation, integration, testing, monitoring, and maintenance. Using accurate and detailed virtual models, they monitor device status, predict their future behavior, and even explore a wide range of operational strategies. This allows for a deeper understanding of real-world systems and optimization. Digital twin technology was first used by NASA in the Apollo space program in 2002. As a result of these innovations, NASA built replicas and models of the spacecraft systems on the ground (e.g., a digital twin) to efficiently manage and maintain spacecrafts far out in space. This innovation served as the basis for the development of digital twin technology. After Michael Grieves of the University of Michigan made a public presentation introducing the idea of the digital twin in 2003 [1].
Since the beginning of the 21st century, information technology has rapidly developed, especially with the emergence of a new generation of information technologies, including the Internet of Things (IoT), cloud computing, big data analytics, artificial intelligence (AI), and many others. Digitalization has significantly accelerated [2]. As a result of this development, digital twin technology has increasingly been used in various fields. Several fields have been transformed by integrating the physical and virtual worlds, making digitization a driving force of innovation. Various applications of digital twins have been found in areas such as smart cities, healthcare, agriculture, transportation, automotive, aerospace, manufacturing, energy, electricity, and other fields. Manufacturing companies can use digital twins to improve product quality, reduce production costs, increase delivery speed, and predict bottlenecks during the production process to accelerate the launch of new products [3]. An urban planner may use digital twin technology to simulate the infrastructure and operations of a city to provide planning and decision-making advice. Digital twins can be used in the energy sector to simulate and analyze energy demand, for example, to test various lighting designs in various application scenarios through computer simulation to choose the most appropriate one [4].
A digital twin technology in architecture allows real-time connectivity between an actual building and its virtual counterpart. Further to this subject, an effective assessment of unanticipated or unpredictable aspects of a building can be accomplished. Furthermore, it contributes significantly to streamlining building workflows, reducing maintenance costs, improving building user engagement, and integrating a variety of building information technologies into the building operations [5]. Digital twin technology has significantly improved the construction project implementation process, including reducing the time to completion and lowering the budget. Many emerging fields will benefit from this, including the design and construction of smart buildings, where using this technology can reduce the cost and time of building a building, enhance collaboration and productivity, improve the efficiency of utilizing existing buildings, and improve safety and sustainability [6]. In addition, digital twin technology can help achieve some previously unreachable building goals. Additionally, digital twin technology can help meet previously unattainable goals in the construction industry. As an example, this technology can help develop more effective maintenance strategies by providing real-time information on the status of systems or equipment. The construction industry could achieve this by performing preventive maintenance and improving facility management efficiency. In urban planning, digital twin technology may provide researchers and planners with more information regarding the growth of cities or help predict the effectiveness of using urban facilities. Though digital twin technology offers significant potential, several challenges remain to overcome, including data processing on a large scale, data security and privacy, cost control, technical limitations, and seamless integration of the physical and digital worlds. Several additional challenges and complexity arise in the architectural field, particularly in data collection and processing, real-time monitoring, data updates, data access, personalization of user interface design, cybersecurity, and integration with intelligent solutions [7]. Professionals in related industries are actively engaged in research and development to overcome these challenges. As a result of these initiatives, several industry leaders have made significant investments in digital twin technology. At the same time, the increased attention and increasing number of publications have led to a strong impetus for advancing the technology [8]. The application of digital twin technology has achieved significant growth in recent years, a trend closely related to the rapid development of Industry 4.0. Along with providing a wide array of application scenarios, the development of Industry 4.0 also provides the necessary support for the technology's development. With the continuous development of Industry 4.0, digital twin technology has more significant potential for solving current problems and advancing the industry further.

The Importance and Applications of Building Energy Management
The application of digital twin technology has contributed positively to this. The industrial process sector accounts for 35.2% of global CO 2 equivalent emissions, and 69% of these are directly related to industrial energy use, highlighting the need for immediate action to reduce industrial energy consumption and emissions. Digital twin technology has great potential to provide the best physical solution to support plant operations and asset management to effectively address the energy emissions challenge [9].
Digital twin technology has become increasingly crucial for managing buildings' energy consumption in recent years. This innovative technology allows managers to monitor the energy usage of building facilities, including lighting, plug loads (already accounting for 33% of total energy use in commercial buildings, [10] and HVAC systems. In this way, they can understand and optimize building energy consumption patterns in detail. It is worth mentioning that the commercial and office sector accounts for 71% of total global energy consumption, of which at least 18% is used for lighting, accounting for 20% of global energy consumption [11].
As a result, the annual energy use of an office building is influenced by many factors, including location, equipment uses, hours of operation, HVAC systems, and lighting technology, which makes improving the energy efficiency of lighting systems, while ensuring lighting quality and illumination, a key area for green building energy efficiency research [12].
In addition, digital twin technology can simulate the operating state of devices and systems to optimize their operation in real time, thus improving energy efficiency. This feature makes the digital twin powerful technical support for energy-saving efforts. In addition, the predictive maintenance function of digital twins can anticipate and deal with possible equipment problems in advance, avoiding energy waste due to equipment failure and thus saving maintenance and repair costs.
Digital twin technology can simulate the energy performance of different design options during the planning, design, and construction phases of a building, providing designers with a basis for making more energy-efficient decisions. It can also generate comprehensive energy usage reports and forecasts, which provide an important reference for developing energy management strategies. For example, during the construction phase, the digital twin can simulate the characteristics of utilities to enhance the operation, maintenance, and lifetime of facilities, and combine real-time data such as building information and building automation systems to accurately reflect the actual performance and energy consumption of the building [13].
Combined with an intelligent building management system, digital twin (DT) technology automates a building's energy use, thereby achieving energy savings and reducing carbon emissions. At the same time, DT technology helps reduce operating costs by improving energy efficiency and reducing maintenance costs, which in turn improves the economic efficiency of buildings. In addition, the advantages of DT technology are reflected in ensuring low-carbon production, mitigating the impact of economic activities on carbon emissions, testing energy strategies online, monitoring abnormal energy consumption, and warning of high-emission behaviour. Together, these benefits and applications highlight the key role of DT technology in driving environmental goals and improving energy efficiency [14].
Overall, the role of the digital twin in building energy management is indispensable. It not only helps us understand and manage the energy use of buildings in greater depth, driving energy efficiency, and environmental goals but also significantly improves the economic efficiency of buildings. Therefore, we should fully utilize and develop digital twin technology to achieve more efficient and environmentally friendly building energy management.

Research Objectives
The objective of this study is to explore in depth the potential of digital twin technologies in building energy management, especially in the context of Building Information Modelling (BIM) and Digital Twins (DT), to improve energy efficiency and achieve sustainability in building energy management by monitoring, optimizing, and predicting building energy consumption in real time. The study explores in depth the application of Building Information Modelling (BIM) and Digital Twins (DT) technologies in energy management, understands their role and impact in the energy sector, predicts their future trends and challenges, and explores their implications and guidance for future research and practice. Overall, this study expects to drive energy efficiency improvements and building sustainability through an in-depth study of the potential of digital twin technologies to improve energy management in the building sector. With advances in building energy analytics and machine learning technologies, the expectation is to improve prediction accuracy, optimize energy efficiency and improve management processes to achieve sustainability goals.

Structure of the Literature Review
The structure of this literature review follows a logical sequence. First, Section 1 briefly outlines the topic and purpose of the review as well as its research value and previews the main contents and structure of the review. Next, Section 2 details the research methods of the review, including the process of literature search and selection, as well as the methods and criteria for literature analysis.
Section 3 on Building Information Modeling (BIM) and Digital Twins (DT) details the development of these two concepts, their main functions, and their applications in the building and energy sectors. The section on energy and its digital transformation then delves into the trends in the energy sector, especially the digital transformation, and the impact of this transformation on energy management and energy efficiency.
In the section on digital twin technology applied to the energy sector, the application of digital twin technology in the energy sector is presented in detail, including specific application cases and effects, as well as the impact of this technology on the development of the energy sector. In the section on future directions and challenges, the future directions of digital twin technology in the energy sector are predicted, as well as possible challenges and issues, including technical, economic, policy, and social challenges.
In the section on implications for future research and practice, the possible implications of the studies in this review for future research and practice are explored, including implications for future research directions and guidance for the application of digital twin technology in practice. Finally, in Section 7 and the recommendations section, the full text is summarized with observations on the application of digital twin technology in the energy sector and recommendations for future research and practice.

Methodology of the Literature Review
This section will elaborate on the methodology of this literature review. In this study, a comprehensive and in-depth look was taken at 189 papers on recent examples of digital twins using different energy simulation software and emerging machine learning algorithms. There are two main reasons for choosing the method. On the one hand, by studying these papers in depth, we can have a more comprehensive understanding of the latest research trends and development directions; On the other hand, this is a general method that has been used in many review studies. The systematic approach is divided into several steps: identification, screening and classification, conformity assessment, and analysis.
In the identification phase, the keywords, "Digital Twin", "BIM", "BEM", and "Machine" were searched extensively in Google Scholar and the University of Sydney Library's bibliography.
For the screening and classification stage, after accumulating numerous papers, an initial screening was conducted based on the titles and abstracts of the articles to ensure that the papers were relevant to our research objectives. During this process, our focus was narrowed gradually to eliminate papers that were not relevant to our research objectives while we classified the literature. As the title and abstract are the most important parts to reflect the theme and content of an article, screening by title and abstract can effectively narrow the scope of research and improve research efficiency. Then, the literature found was put into three main categories:

1.
Those exploring background knowledge on the topic.

2.
Those related to established software applied to building energy management.

3.
Those related to emerging machine learning models being applied to building energy management.
During the conformity assessment phase, an in-depth study was conducted on the selected articles to assess the quality and relevance to further narrow the study's scope. Factors considered in the assessment include the date of publication (with a significant emphasis on recent papers), the quality of the research methodology, the use of innovative software or machine learning algorithms, and the impact of the research in the field.
Finally, a thorough analysis and summary of the final selected literature was conducted in the analysis phase. Papers are categorized according to the type of energy simulation software and machine learning algorithms used in the papers. An insight was provided into applying these techniques in the digital twins and BIM field. At the same time, we assess their strengths and weaknesses and possible directions for future development. When assessing strengths and weaknesses, we refer to previous studies and evaluations of these technologies to ensure the comprehensiveness and accuracy of our assessment.

Background of BIM and Digital Twin
In the construction field, the application of digital twin modeling can be categorized into three stages: The first stage involves the widespread adoption of conventional Building Information Modelling (BIM) methods, which primarily facilitate information exchange among stakeholders to manage the lifecycle of structures effectively. The second stage entails the integration of digital twins with real-time monitoring and simulation based on 3D BIM models. The third stage, digital twin technology, combines more advanced technologies such as deep learning and AI to drive digitalization further and enable realtime decision-making in building management.
This section will provide an overview of the evolution of BIM and digital twin technologies, as well as an exploration of their current interactions.

Building Information Modelling (BIM)
Building Information Modelling (BIM) was first proposed in the 1970s as a digital representation of the physical and functional characteristics, including various building data and information, and used throughout the life cycle of the buildings [15]. The initial application of BIM in new construction is mainly about design, visualization, coordination, and throughout project management phases [16]. Since BIM includes all the information of the structure, it is multidimensional from its nature, and each subset containing different information can be called a dimension. BIM can be mainly divided into eight dimensions, which are object model, time, cost, operation, sustainability, and safety [17]. In simple terms, BIM can be both a technology and a process; it provides a visual platform for stakeholders to better collaborate and communicate throughout a project. For traditional project collaboration, project documents, and technical drawings are manually passed between stakeholders, and subjective interpretation can easily lead to misinterpretation during information transformation, negatively affecting collaboration efficiency.
However, the application of BIM enables stakeholders to communicate and share information through a database with a visualized model. The information-sharing process allows real-time remote communication, supported by cloud computing, significantly increasing collaboration flexibility [18]. Graphical models enable stakeholders to reach a consensus on decisions early in the project life cycle, significantly reducing project costs and improving project quality [19]. For example, converting 2D technical drawings into 3D for sharing between design and construction teams can quickly gain a common understanding of expected deliverables.
Digital transformation drives the energy sector through the Fourth Industrial Revolution, and BIM is expected to be the core of the "digital twin" for actual buildings [20,21]. The application of various digital technologies such as smart meters, advanced control systems, artificial intelligence, and deep learning algorithms positively impact the management of global energy systems, and these new methods of combining different technologies and software with BIM are called Integrated Building Information Modelling (iBIM) [22] BIM has been proven to comprehensively shape digital twins when combined with methods and tools such as Augmented Reality (AR), the Internet of Things (IoT), and Big Data [23,24].
Building Information Modelling (BIM) has also emerged as a potential solution to improve energy efficiency. Digital representations of construction processes through BIM platforms can facilitate information exchange and interoperability in digital formats. The application of digitalization and BIM can significantly improve the overall efficiency of the energy system, as well as the collaboration, flexibility, and information update throughout the building life cycle. Combining BIM with Building Energy Modelling (BEM) can save energy and costs [25].

Digital Twin
The development of this technology can be traced back to the 1960s when NASA devised "digital twins" to assess failures in the Apollo missions. The same physical spacecraft was built to simulate and study the different conditions the spacecraft would face in space. Later, David Gelernter proposed the idea of Digital Twin technology in 1991. Subsequently, Michael Grieves proposed the concept of Digital Twin Software in 2002; this was the first time to apply software to the manufacturing industry. In 2010, NASA defined the concept of the Digital Twin in detail, becoming an essential part of the aerospace field [4]. In 2002, NASA's Greaves and Vickers introduced the concept of digital twins for the product lifecycle management (PLM) [26]. Broadly defined as "a digital twin is a dynamic, self-evolving digital/virtual model that represents the state of a real-life object and its physical twin. A digital twin is created by exchanging real-time data and preserving historical data realized".
However, for a long time, the 'digital twin' has been just a concept without any suitable technical means to support its application. With the advent of Industry 4.0, digital technologies are developing rapidly and promoting digital transformation in various industries. The transformation also promotes the development and application of digital twin technology. Digital twins are widely used in manufacturing, construction, medical care, urban planning, energy, agriculture, and other fields to optimize production, design, maintenance, and monitoring, improve efficiency and sustainability, and manage and optimize entities, processes and systems to achieve Increased efficiency, reduced costs and increased sustainability [27].
For the construction industry, a digital twin is defined as "a real-time representation of a fully or partially completed and developed building or structure to represent the state and characteristics of the building or structure it reflects" [28]. Thus, it enables seamless synchronization and monitoring of energy systems through computerized and virtual world simulations based on data, information, and consumer behavior. 3D modelling, digital prototyping, and system simulation are all products based on the digital twin concept [28].
As mentioned before, the digital twin (DT) aims to gain insight and predict the performance of a physical product, process, or infrastructure through a virtual model. A digital twin consists of a physical system, a virtual model, and the data network between the two [4]. Digital twins are used to improve physical entities' performance by using virtual models' computational tools. Real-time data from physical entities is used to inform the development of virtual model parameters, boundary conditions, and dynamics. This results in a more realistic representation of the actual entity being modelled [26].
The digital twin is a digital version of a physical entity with the principle of establishing a digital model by collecting data on physical entities, processes or systems, and interacting and collaborating with physical entities in real time to achieve optimization and improvement. Therefore, to make digital twin become practical, there is a need for an integrated analysis system [11].
Digital Twin can predict potential problems in real time by monitoring physical entities and providing feedback to aid decision-making and improvement, optimize design management processes, reduce costs, and promote sustainability [29]. Due to these characteristics, digital twins were introduced into the construction industry and combined with BIM based on CAD, SolidWorks, Revit, etc., to create an integrated construction management process covering the entire building life cycle [29].
Digital twin technology is at the heart of building energy efficiency. Electricity, heating, and air conditioning (HVAC systems) account for a large percentage of total building expenditures. Through the application of digital twin technology, it is possible to simulate the energy performance of buildings and predict their energy consumption patterns to control and reduce energy consumption and improve the energy efficiency of buildings. This will positively impact reducing energy waste and lowering the total cost of ownership of the building. Managing the costs, benefits and risks associated with the energy transition is essential to achieving sustainability. Within the Sustainable Development Goals (SDGs) framework, the development of sustainable, affordable, reliable, and innovative energy systems and services has become an important issue. Digitalization and digital twin technologies can help drive energy sector sustainability and the transformation and upgrading of energy systems.

From BIM to Digital Twin
As mentioned before, Industry 4.0 technologies have significantly impacted the construction industry. Building Information Modelling (BIM) has been widely adopted and continuously improved recently. Integrating building component information into building models through parametric modelling extends the 3D model of real-world projects. The process and framework of BIM are used to establish a clear project vision in the early stages of the project design [30].
As for three critical areas of technology, safety and management, BIM has been proven to assist in realizing digital twins (DT), significantly improving the prediction, management and monitoring of project quality and performance [31]. The integration of BIM and DT is based on the BIM model to create a virtual replica of a building or facility, combined with other software to fully simulate the physical environment of the building entity to create a Digital Twin [32].
In simple terms, in construction, a digital twin can be considered an iterative and upgraded version that includes BIM. Digital Twin is dynamic, and its real-time nature allows the project to be optimized and adjusted during the construction and implementation phase, realizing the synchronization of the real and virtual models in the platform. This integration combines the advantages of BIM and digital twins, enabling stakeholders to increase process visibility, efficient information transfer and communication based on real-time data synchronized to the digital twin. DTs provide insight into the lifetime and characteristics of individual products and can optimize their sustainability, helping to improve future product generations [33].
While current federated technologies are still limited by data transparency, concurrent viewing and editing of a single federated model, and controlled coordination of information access, [31] applications of BIM and digital twins are expected to play a vital role in smart building management.

Energy and Its Digital Transformation
Energy demand is fundamental to economic and social progress. As a major fuel supplier to the entire economy, the energy sector is crucial in expanding manufacturing and services [34]. The Sustainable Development Goals (SDGs) aim to develop sustainable, affordable, reliable, and innovative energy systems and services [35]. The energy transition is breaking down the constraints of energy models, and managing the costs, benefits, and risks involved in transitioning to a more energy-efficient energy system is an essential part of achieving sustainability.

Development before Digitalization
Since "sustainable development" became a hot topic in the 1980s, people began to feel an aversion to early modern society's revolutionary industrial growth model [36]. The concept of sustainable development has gradually been advocated in various fields, and people are paying more and more attention to controlling greenhouse gas emissions and saving energy. Building energy consumption has become the focus of sustainable development. Research shows that for different types of buildings, from traditional buildings to passive and low-energy buildings, energy consumption dominates the entire building life cycle, and its share is increasing [37]. Furthermore, energy consumption in daily life is also very important. In developed countries, residential and commercial buildings can account for 20% to 40% of total energy consumption [38]. Before the 20th century, people were mainly committed to upgrading building materials and energy systems to save energy, such as looking for high-efficiency insulation materials, energy-saving windows, lighting systems, etc., and HVAC systems were developed over this period.

Digitalization in Building Energy Management
At the end of the 20th century, the third industrial revolution was characterized by the widespread use of computers and the Internet, laying the foundation for new building energy management technologies [21]. Computer simulation optimization has been introduced into energy management, and various energy simulation programs have gradually emerged. Building Energy Modelling (BEM) has become increasingly popular as the application of energy simulation programs which has proven effective in improving energy efficiency. The study used computer simulations to optimize integrated lighting, passive solar heating and cooling, and energy efficiency strategies in building design. The results showed a 50% to 70% reduction in building energy consumption compared to code requirements [39]. Different energy simulation programs were incorporated during that time to meet different user needs. Most programs are equipped with capabilities for building modelling, renewable energy systems, electrical systems and equipment, HVAC systems and equipment, environmental emissions, economic evaluation, and user interface [40].
We are currently living through the Fourth Industrial Revolution, with the combination of BIM and Building Energy Modelling (BEM) being expected to collaborate with other advanced technologies, such as the Internet of Things (IoT), to improve the efficiency of energy simulation further. Combining these methods for real-time simulation is a future priority for Industry 5.0. Sensors, big data processing techniques, and deep machine learning are expected to be included for better simulations and energy forecasts [23,41,42]. Digital transformation is driving the energy industry through the Fourth Industrial Revolution [21], and the application of various digital management of energy systems has had a positive impact, such as smart meters, advanced control systems, artificial intelligence, and deep learning algorithms. Building Information Modelling (BIM) has also emerged as a potential solution to improve energy efficiency. Digital representations of construction processes through BIM platforms can facilitate information exchange and interoperability in digital formats.
The Architecture, Engineering, Construction and Operations (AECO) sector is one of the primary sources of global energy consumption and its negative impact on the environment cannot be ignored. Energy use in residential and commercial buildings has continued to increase by up to 40% [37]. Therefore, reducing the energy demand for Sustainability is crucial. Section 3.3 explains that digital twin technology can potentially improve buildings' energy efficiency. It is generally accepted that incorporating digital twins into the entire energy management lifecycle of building design can positively impact the design process and final design [34].
Recent research on the application of digital twins in building energy mainly falls into two categories: one focuses on using sensors and BIM software to visualize better and control energy consumption. The other category is more focused on using artificial intelligence to predict future energy demand to optimize building energy system management efficiency. Optimizing electrical, heating, and cooling loads (HVAC systems) is critical to building energy management. From the results of these two types of research, there are three main ways to apply the digital twin method to building energy management optimization: First, the digital twin can monitor and analyze real-time data from sensors installed in the building to determine whether energy consumption can lower area. Second, the digital twin tests and evaluates the impact of different energy management strategies on the building's energy consumption. Third, digital twins help managers make informed decisions about energy use and identify energy-saving opportunities. Digital twins can also help reduce maintenance costs by predicting equipment failures, scheduling maintenance activities accordingly, and helping improve safety by simulating hazardous scenarios and testing contingency plans [43].
Digital twin technology can model buildings and analyze their energy status to determine factors that affect energy efficiency. Then, take corresponding measures to reduce energy consumption and improve the sustainability of the building. For example, digital twin technology can optimize a building's lighting and air conditioning systems to reduce electricity consumption and carbon emissions. In addition, it can optimize the materials and design of building facades and windows, thereby improving the building's thermal insulation and ventilation, reducing energy consumption and negative environmental impact. Overall, digital twins provide building energy with data and analytics that support decision-making and implementation to improve energy efficiency and enable more energy-efficient operations.

Digital Twin Technology for Energy Sector Development
The use and management of energy in building design and operation is an important part of current global environmental issues. Building energy analysis software is being used more frequently currently to improve energy efficiency and reduce environmental impact. With their powerful simulation and analysis capabilities, these tools provide more accurate energy use data and predictions for the building design, construction, and operation phases. This chapter is a review of a range of software tools and machine learning modules that are widely used in building energy analysis and explores how they have been applied to different types of buildings and different energy issues by analyzing various real-life examples. Table 1 shows 68 applications in the field of building energy that have been linked to the use of energy analysis software or techniques in this section and indicates the year of publication, the software used and the application content. The widespread use of software such as EnergyPlus and Insight can be seen in many studies in recent years. The characteristics and scope of application of this software vary, but their common goal is to understand and optimize the energy consumption of buildings. There are many different types of software used to carry out simulations and analyses in the study and evaluation of energy efficiency in buildings. With their unique characteristics and capabilities, each of these tools has advantages when dealing with a variety of specific tasks. Several trends and patterns are evident based on the data collected. In terms of frequency of software use, EnergyPlus is the most used tool, with features covering a wide range of areas from energy consumption and thermal performance analysis to complex environmental simulations. Simulation of the temperature domain within buildings using Energy Plus included Zhong et al. [75] analyzing the energy impact of excessive cooling in four public buildings within a university in Hong Kong. Temperature measurements were collected by deploying 577 sensors and compared to comfort theory. Simultaneous building energy simulations using EnergyPlus revealed that a 2.5-degree deviation in the number of degrees of subcooling leads to a significant increase in annual energy consumption, with excessive cooling leading to unnecessary energy waste and low thermal satisfaction. Araújo et al. [63] investigated different switching in offices located in Lisbon (Portugal) and Copenhagen (Denmark). Temperature range and the effect of thermochromic coated glass on energy use in Lisbon (Portugal) and Copenhagen (Denmark offices). Simulations were carried out to investigate the effect of different temperature ranges of thermochromic-coated glass transmittance on office energy use and thus to adjust the properties of thermochromic glass to reduce energy consumption. The results show that by adjusting the properties of thermochromic glass, it is possible to provide optimum performance for specific climatic conditions. Simulations using Energy Plus in the field of building materials and building design included modelling and energy analysis of a residential building in Egypt by An-Naggar et al. [46] demonstrates that the use of insulation can reduce building energy consumption. The research methodology included an analysis of different wall and roof materials. The results focus on understanding the thermal environment of buildings and provide a basis for improving the thermal performance and energy efficiency of buildings. Design and optimization of energy consumption in low-rise buildings in composite climates for buildings in the Indian region by Verma et al. [89] The research methodology included the use of different passive design strategies such as phase change materials (PCM), cold roofs, green roofs etc. and energy modelling and analysis using EnergyPlus. Al-Yasiri et al. [91] carried out building modelling and energy analysis of buildings in Iraqi cities. The extent to which different building insulation materials improve energy efficiency in summer is also being investigated. The research methodology focuses on the thermal performance of phase change materials and expanded polystyrene in buildings through simulation analysis. It is concluded that by integrating phase change materials (PCM) and expanded polystyrene (EPS) in the roofs and walls of buildings, the energy efficiency of buildings can be significantly improved during the harsh summer months. Kashani et al. conducted a study of bio-based phase change materials (BioPCM) in building walls to improve thermal performance and energy efficiency in a building in Tehran, Iran. The research methodology consisted mainly of conducting building energy simulations in which four different types of materials (BioPCM) were used. The effectiveness of biomaterials in improving the thermal performance and energy efficiency of buildings was assessed by comparing the energy consumption of buildings with and without the use of biomaterials. The results showed that the use of biomaterials was effective in reducing the energy consumption of the building. Elnabawi et al. [98] studied the effect of roof retrofitting techniques on the indoor energy consumption of buildings. The article used EnergyPlus to analyze and compare the energy before and after the roof retrofitting of a residential building in Bahrain. Jiang et al. [104] studied a three-star green office building on the thermal performance and carbon sequestration of the building. The results of the study showed that the case building with a green roof and green façade could save energy during the cooling season. The energy-saving effects were simulated with the green area index and height domains included. Ahmed et al. [53] used DesignBuilder to conduct a study on energy retrofitting of residential buildings in the Eastern Province of Saudi Arabia. The study included villas and apartment buildings. The study methodology involved implementing and evaluating various energy efficiency measures and using conducting energy, economic, and environmental analyses. The results of the study show that the annual energy consumption of villas and apartment buildings can be effectively reduced through the implementation of energy retrofit programs. The study concluded that although energy retrofitting can significantly reduce energy consumption in dwellings, it is not economically viable in the current situation. Sadeghifam et al. [48] used EnergyPlus software to evaluate energy-efficient designs for tropical residential buildings, focusing on the choice of building materials. The study was conducted on a typical single-story row house in Kuala Lumpur. The study methodology included BIM energy performance simulations and assessed the impact of noise factors. The results show that an appropriate selection of ceiling, window, wall, roof, and floor components can significantly improve energy efficiency.

Software for Energy Prediction-Actual Software Cases
Simulation of the field of service equipment in a building using Energy Plus includes an in-depth analysis of the energy performance of a service building in the district of Aveiro, Portugal by Silva et al. [51] The article analyzes in detail the building's characteristics, equipment, and HVAC systems. It reveals that lighting and air conditioning are the main sources of energy consumption. The importance of considering different factors when analyzing energy consumption is emphasized, including lighting, air conditioning, and building-specific usage. Kamel et al. [79] conducted a thermal analysis and building energy modelling (BEM) study of 3D-printed homes in two climate zones in the USA. The study methodology included modelling the 3D printed homes using Building Information Modelling (BIM) followed by an energy analysis using EnergyPlus. The results of the study showed that a properly designed all-electric 3D-printed residential building can reduce energy use intensity and CO 2 emissions. In addition, common 3D-printed wall systems do not meet the energy requirements of cold climate zones. Esmaeilzadeh et al. [95] investigated ways to create a zero-emission zone using a combination of control methods, including the application of control algorithms, environmental footprint considerations, solar energy utilization, and the modification of HVAC systems, retrofitting of HVAC systems, and the use of hybrid energy resources. EnergyPlus was used to validate the combination of control methods, apply optimized controllers to simulations, and compare the effects of different controllers on the efficiency of the building's hybrid HVAC system. Cooling system performance under different weather conditions. An evaluation of the use of the new EnergyPlus plug-in Spawn is discussed. The research methodology focuses on numerical simulations to compare the performance of CO 2 cooling systems under different weather conditions. It is concluded that the implementation of clean refrigerants such as CO 2 is a promising solution to mitigate global warming. Singh et al. [106] carried out an energy performance analysis of two models of Air Conditioning Systems for a building of 511 m2 under composite weather conditions. Panagiotakis et al. [52] conducted a study on energy efficiency and energy use efficiency (PUE) of a data center in an educational building in Heraklion, Crete, Greece. The results show that PUE can be effectively reduced, and energy efficiency improved by interventions in the building envelope and by increasing the temperature of the data center. However, the article also emphazises that PUE should not be the only indicator of data center energy efficiency and that increasing the operational efficiency of IT equipment does not necessarily reduce PUE unless power and cooling equipment is also optimized to accommodate new IT loads. Liang Zhao et al. [69] DesignBuilder simulated the ventilation design of a public toilet and its building energy consumption and indoor comfort. It was found that when the vent height was 400 mm and the air exchange frequency was once every 12 h, it was effective in reducing the concentration of air pollutants inside public toilets. This finding reveals that good ventilation design can improve air quality and save energy, which is important for improving the overall energy efficiency of a building.
Related areas of standardized analysis, comparison, validation, and prediction of building energy simulation results using Energy Plus include the analysis and modeling of energy consumption of a residential building in Lebanon by Yathreb et al. [44] The creation of a 'baseline model' reflecting existing residential buildings will provide a basis for studying specific building technologies and measuring progress towards the goal of zero-energy buildings. The building owner or tenant will understand how the baseline model of their building's energy performance may be used to develop energy source code for building and building equipment standards. Kunwar et al. [85] modeled and calibrated a specific supermarket building to assess the impact of proposed energy conservation measures (ECMs) and control strategies. The results demonstrate the use of all energy conservation measures to achieve energy savings and net cost savings. Zou et al. [88] conducted a comprehensive analysis of the energy resilience performance of urban residential buildings in hot and humid zones in China, considering factors such as climate change, typical buildings, and occupant behavior. The methodology included energy analysis and building simulation using EnergyPlus software and future energy forecasting. The results show that the future energy demand of the residential sector in the hot and humid zones of China will increase over time. Yang et al. [93] studied the carbon emissions and energy consumption of buildings in Changsha, China and used EnergyPlus to obtain building energy intensity to develop a model to describe the annual carbon emissions of the city-level building stock. The results show that the scenarios set out in this study do not achieve carbon neutrality in the building sector by 2060 and that additional carbon reduction plans are needed to accelerate the process of carbon reduction in society. Di Turi et al. [101] conducted a building energy analysis of non-residential buildings in Italy to demonstrate the energy and economic viability of an economic assessment of new office buildings in Italy. Tayari et al. [103] conducted a study of the Validation of heat gains in traditional courtyard houses in the Kerman region of Iran using DesignBuilder simulation software, demonstrating the difference in heat gains in traditional courtyard houses in Kerman and the effectiveness of DesignBuilder simulation software. Zhao et al. [60] conducted a study of near-zero energy retrofitting of an existing building using BIM technology and energy modelling. Energy models of the buildings were created using 3D laser scanning technology and energy simulations were carried out using DesignBuilder software to evaluate retrofit options. The study was carried out on a residential building in Dezhou, Shandong Province, China. The results show that electricity generation can be increased by adjusting the angle of the solar PV modules. The use of BIM and 3D laser scanning technology can be used to effectively evaluate retrofit options for existing buildings with almost zero energy consumption and to improve energy efficiency by adjusting the angle of the solar PV modules. The discussion pointed out that while China is making progress in new almost-zero-energy buildings, a more comprehensive assessment strategy is still needed for the almost-zero-energy retrofitting of existing buildings.
In some of the newer studies, EnergyPlus is being used in conjunction with programming software to perform energy simulations and forecasts. When EnergyPlus is used as a core tool for building energy modelling, the decision-making tools that accompany it can greatly enhance energy analysis and optimization capabilities. Such decision tools can include multi-criteria decision analysis tools, life cycle cost analysis tools and machine learning tools.
Areas of relevance to the use of EnergyPlus simulation results with multi-criteria decision analysis tools to evaluate building energy strategies include Pinto et al. [81]'s energy simulation of an office building in north-western Italy with decision tools to support investment in intelligent shading devices for the office building. The article uses EnergyPlus as an energy dynamic simulation tool and Visual PROMETHEE as a multi-criteria decision analysis tool for building modelling BIM and energy analysis. The results show that external blinds in combination with control strategies can reduce indoor air conditioning demand. The use of control strategies can have a positive impact on buildings in terms of energy savings, environmental impact, thermal comfort and visual comfort. Wu et al. [83] compared the impact of photovoltaic electrochromic windows on the building lighting environment and energy savings in Chengdu, China when combined with different control strategies. The study methodology consisted mainly of constructing a test setup for zoned PV electrochromic windows with different control strategies using Energy Plus to simulate the building's interior lighting environment and energy performance. The results show that different control strategies can reduce indoor building energy consumption in specific application scenarios. In future research, we will integrate four control strategies for different types of buildings at specific times to achieve an optimal combination of control strategies considering the indoor lighting environment and building energy consumption. Ramon et al. [110] modelled the dynamics of operational energy use in a life cycle assessment (LCA) of a Belgian office building. The building is a commercial office building. The dynamic modelling approach developed in this paper uses EnergyPlus software to combine dynamic energy use in the operational phase with natural gas and electricity scenarios, in conjunction with climate data for the Belgian region. The final environmental cost of electricity increases while the environmental cost of natural gas decreases.
Using EnergyPlus simulation results combined with time series or geo-information analysis, predictions are made using historical and real-time simulation data to aid understanding of multiple building energy use patterns, advance predictions of likely energy demand, and more accurate energy strategy decisions. This area includes a study by Yu et al. [73] of how the interaction between ten typical buildings located in a high-density urban area of Hong Kong affects the energy consumption of buildings. . . The inter-building effect (IBE) of building energy consumption in a high-density urban office environment in Hong Kong was analyzed using EnergyPlus. GIS (Geographic Information Systems) was used to create and analyze models of buildings and their surroundings, and the completed geometric models were exported to EnergyPlus for dynamic building energy simulation. The results showed that inter-building effects (IBE) have a significant impact on building energy consumption, which is related to factors such as building height, shape, orientation, window-to-wall ratio, window U-value, and solar heat gain coefficient. Tian et al. [80] used time-series energy data from 240 office buildings in Shanghai to identify unnecessary building energy consumption and thus improve energy efficiency. Three methods of extracting runtimes from time series data are proposed in conjunction with simulations using EnergyPlus, and the use of cumulative histogram methods to extract building operating characteristics can improve the accuracy of energy predictions and identify unnecessary energy consumption. For optimization of thermal comfort, the methodology included detailed simulations (considering all possible sky conditions during the year) and multiobjective optimization (considering indoor visual and thermal comfort as well as visual connectivity to the exterior) and was validated by field measurements to find the optimum for local Sydney homes. The results show that by optimizing the Anidolic daylighting system, daylight ingress can be optimized without the need for expensive active tracking systems, thus achieving a reasonable compromise between visual and thermal comfort and energy consumption. Liu et al. [58] studied specific building networks in seven Chinese cities, i.e., the impact of shading between buildings on the energy demand of buildings, and the use of models and simulations to make predictions and assessments of this impact to assess the impact of shading from surrounding buildings on the heating/cooling energy demand of different community forms. A parametric building model was modelled using Grasshopper's graphical algorithm editor and then multi-threaded simulations were carried out using EnergyPlus. The results show that building shading has a deeper energy impact in summer and a weaker one in winter. The SVR model built in an urban environment allows for more accurate and faster prediction of building heating/cooling energy demand under the influence of shading. In the future, it will become increasingly important to capture the important shading that can influence the accuracy of building energy modelling predictions, particularly in highly urbanized areas.
Simulation results from EnergyPlus are used in conjunction with programming tools to compare and select suitable energy systems. This area includes the optimization of energy demand for a multi-story hotel structure in the Tehran region of Iran by Nateghi et al. [74] Using data from the Tehran region entered into EnergyPlus to calculate hotel energy consumption, the optimization of building parameters using JEPlus + EA software and nondominated ranking genetic algorithms (NSGA), and economic calculations using Homer Pro software. By selecting the optimal building variables in line with the climatic conditions, the energy demand of the building can be significantly reduced. Furthermore, despite the high cost of renewable energy systems, with the significant reduction in the cost of electricity production, long-term savings can be invested in the installation of photovoltaic solar systems, making the buildings more sustainable. Song et al. [72] investigated a methodological framework for predicting building energy consumption in light temporary residential buildings built during COVID-19, i.e., the impact of wall design on inherent energy consumption and operational energy consumption. Rapid energy consumption predictions for a light building in Nanjing, China were achieved using BIM, the Eco-invent database and EnergyPlus software. Conclusions were obtained regarding the impact of material thickness and façade materials, and an automated method for calculating energy consumption data using a combination of EnergyPlus, SPSS, and programming software was identified. Fabrizio Ascione et al. [84] used MATLAB and EnergyPlus for building modelling and energy analysis to compare comfort performance and cost control for a near-zero energy building in southern Italy. The research methods were model predictive control (MPC) and multi-objective genetic algorithms. The researcher used a framework of simulation and optimization for model predictive control of space cooling systems. The optimization problem was solved by running a genetic algorithm (a variant of NSGA II) in a MATLAB environment. The objective function was evaluated by coupling MATLAB and EnergyPlus. The results show that the approach of combining model predictive control with different thermal comfort models can result in some cost savings while maintaining similar comfort performance.
Gupta et al. [94] used EnergyPlus with an optimization algorithm to optimize the design of a conventional house in northeast India based on building energy simulation results. Fong et al. [111] used EnergyPlus and genetic algorithms for energy simulation and optimization of a residential building in Hong Kong under extreme weather conditions. The study introduced a new modelling strategy to investigate and compare actual building performance under different cooling supply scenarios applied to a high-rise residential building under near-extreme weather conditions. The conclusions show that this combined operation can assist in the use of fan strategies. As well as certain design factors can provide opportunities to reduce total cooling energy in near-extreme weather conditions. Jia et al. [55] used EnergyPlus and PMFserv for building modeling and energy analysis. PMFserv was used to develop models of occupant behavior that take into account physical comfort and psychological states. The research methodology involved developing and validating an occupant behavior model based on existing buildings and coupling it with a conventional building energy model. The coupling of the models was achieved by using BCVTB as a bridging procedure. It is concluded that the coupled simulation experiment integrates the agent-based resident behavior model with the building energy model in EnergyPlus, providing a comprehensive assessment of building performance through a multi-perspective analysis. Key findings include that occupant behavior influences energy use throughout the building from the room level that thermal comfort is not the only driver, and that occupant behavior leads to multifaceted changes in the built environment. Naderi et al. [57] used EnergyPlus and jEPlus + EA software for building energy modeling and optimization, with the NSGA-II algorithm used to perform the optimization. The study focused on buildings in Iran and used a multi-objective simulationbased optimization approach to control blind specifications to reduce energy consumption and thermal visual discomfort. The results show that different control strategies and building specifications significantly affect energy consumption and comfort. For example, in Tehran, the optimal strategy included setting the first point at 17.8 • C, using internal shading positions and slat reflectivity of 0.46, with an annual electricity consumption of 2340 kWh, DGI of 0.9, and PPD of 6.9%. The study concludes that the use of a controlled blinds specification can significantly reduce energy consumption and improve thermal and visual comfort in buildings, with the optimum specification varying by location and climate. The article discusses research related to shading systems and lighting performance control and highlights the application of automated shading and lighting systems to improve energy performance while exploring the impact of different control strategies and building codes on energy performance and comfort and highlighting the importance of selecting the best shading features and physical characteristics according to climatic conditions. Jafari et al. [61] used EnergyPlus to perform building energy simulations while building a digital twin of the asset using techniques such as big data streaming and artificial neural networks (ANN). These methods enabled the authors to evaluate the effectiveness of the new HVAC setpoint controls under various weather and climate conditions. The results show that digital twin technology can be effective in optimizing building energy efficiency and asset performance. It is concluded that the integration of existing building digital twin and optimization technologies can provide a framework for optimizing building energy efficiency. This extends the application of digital twin technology from the design phase to the building maintenance and operations program. Y. Zhao et al. [71] used digital twin technology and machine learning algorithms in combination with building modelling BIM and energy analysis software (e.g., Revit, Insight, EnergyPlus, Autodesk, etc.) to build a digital twin model for the building operations and maintenance process. The feasibility of integrating the digital twin with machine learning was demonstrated by verifying the results of artificial neural network predictions. The results show that the model can achieve intelligent prediction and diagnosis of building O & M status with reliability and efficiency. However, further work is needed for practical application research to apply the theoretical approach to real projects. In summary, the application of digital twin technology and machine learning algorithms provides new ideas for intelligent building O & M, but its implementation is still in its infancy and there is still room for future development.
The use of other tools has also shown advantages within the field of building energy. Using Green Building Studio, Changsaar et al. [62] conducted an energy analysis of each appliance in an approximately 1450 sq ft building at the University Technology Malaysia. By collecting data on electrical devices, energy usage intensity, and annual electricity consumption, the total percentage consumption of each type of appliance was obtained and the building's energy performance was redesigned based on the appliance configuration and the recycling system was redesigned according to the energy performance of the building. This allowed annual electricity use to be reduced from 23% to 17.2% and overall energy costs to be reduced by 9%. The benefits of integrating BIM and GBS for building energy performance measurement were identified. Tahmasebinia et al. [64,65] modelled the Abercrombie Building (ABS) building at the University of Sydney and used Green Building Studio (GBS) to generate an energy model. Regression methods were used for building energy performance estimation. The results present several linear regression models that focus on wall construction, plug load efficiency, infiltration, roof construction and lighting efficiency. Rstudio and Excel were then used to process the data results and perform regression analysis. The results show that the accuracy of the multi-linear models using simplified window-to-wall ratios and EUI energy use intensities for different room shapes remains to be seen, but the simulated energy is generally lower than realistic values. By integrating mathematical optimization, Building Information Modelling (BIM) and Life Cycle Assessment (LCA), the lifetime costs and energy consumption of buildings can be significantly reduced. Energy performance is mainly influenced by the building's evaporative cooler and the building envelope.
Using Insight software, including Kaewunruen et al. [47] used Revit to apply house models of two buildings in the UK in order to study net zero energy buildings NZEB, and simulated the annual running costs of the thermal performance of the buildings through Insight, demonstrating the way in which digital technologies such as Building Information Modelling (BIM) can be applied in existing built environments to achieve the NZEB (Near Zero Emission Buildings) goal. Yang et al. [97] conducted a Building Information Modelling (BIM) and energy analysis of a large public building in Xinghua, Jiangsu Province, China, in order to investigate the comfort and carbon emissions of a large public building design. It was concluded that an interactive design of outdoor ambient paving and indoor ventilation atrium skylight orientation can effectively balance comfort and carbon reduction in large public buildings, providing an important reference for low-carbon development and green building design strategies. Saranathan Pragati et al. conducted a study on the energy performance of green roofs and green walls in a building in Singapore in a tropical climate. The study methodology included simulating the energy performance of the building's green roof and green walls using Autodesk Revit and Insight software. The results show that green roofs and green walls can significantly reduce the energy consumption of buildings, especially in tropical climates. Utkucu et al. [54] used tools such as Revit, Dynamo, Project Fractal, and Insight 360 for building BIM modeling and energy analysis. The research methods include building façade optimization, energy performance prediction, and optimization, the impact of natural ventilation on the indoor environment and comfort, and design quality studies. The results show that the use of 3D, energy, and CFD models of buildings allows for interoperability and data exchange and increases efficiency. This study presents a methodology for examining the energy performance of buildings, focusing on building façade optimization, model-based energy performance prediction, and optimization, the impact of natural ventilation on air quality and comfort, and a design quality study with simulations included in the field of energy consumption. Tanriverdi et al. [99] conducted an energy analysis of an office building in London to study the impact of HVAC system retrofitting on the building. For the impact of energy consumption, the detailed analysis of the HVAC system retrofit includes the impact of different retrofit strategies on total energy consumption, operational carbon emissions, and operating costs. Section 7 states that energy efficiency improvements are not always feasible due to design, application, and cost and that energy retrofitting of HVAC systems is an effective way to improve energy, carbon, and cost. This study has established a methodology for HVAC energy retrofitting in existing office buildings to reduce energy, carbon, and costs.
In addition to traditional energy modelling software, some studies have incorporated other software and techniques. Chen et al. modeled and compared real-time energy consumption between buildings of different years of construction in the central Olympic area of Beijing. The eQUEST software was used to simulate and analyze building energy consumption. Nineteen influencing factors were selected for analysis, including household income, household conditions, building conditions, lifestyle, household appliances, and energy efficiency concepts. The results showed that the height of the building, floor area per person, number of inhabitants, number of air conditioners and computers, and air conditioning cooling methods were the most important factors influencing energy consumption. Yeweon Kim et al. [86] conducted Zero-Energy Building Certification for selected buildings in Korea using an energy simulation program (ECO2). For the study methodology, the main approach involved collecting and analyzing seasonal temperature data from hydrothermal sources, constructing a database of hydrothermal sources, and then using the ECO2 program to calculate geothermal/hydrothermal energy self-sufficiency. Liu et al. [77] investigated how urban form affects the energy consumption and solar potential of buildings, using the example of a residential area in Jianhu, China. The article uses the Grasshopper platform for building modeling and energy analysis. Three optimization objectives were identified: minimizing building energy consumption, maximizing solar potential, and maximizing daylight hours. To perform the performance simulations, the article uses the Ladybug Tools plugin and performs multi-objective optimization in Wallace X. Based on the findings of this study, a design strategy for energy-efficient urban block forms is proposed. It was established that urban form has a significant impact on building energy consumption and solar potential, but few studies have presented a comprehensive framework to discuss both. Another issue that needs to be addressed is how the numerical relationships expressed by the data can be applied to or assist in the design of energy-efficient blocks. Based on this background, this paper proposes an energy efficiency-oriented multiobjective urban form optimization workflow based on the Grasshopper platform, with a case study in Jianhu County, Jiangsu Province. In addition, the quantitative relationship between urban form and building energy consumption as well as solar potential is explored. Krarti et al. [87] investigate the optimization of insulated sunshades for residential building energy performance. The article presents a full building simulation and dynamic envelope system modelling of flat dwelling units in different climate zones in the USA, followed by building modelling and energy analysis using DOE-2 software to determine the optimal placement of dynamic sliding sunshades, and various operational strategies for dynamic sunshades were also considered. The results show that dynamic sunshades have significant energy efficiency potential in all US climatic and design conditions. The influence of various design and operating conditions on the optimal performance of dynamic sunshades is discussed. Q. Jiang et al. [78] conducted an energy consumption analysis of energy use behavior within a typical university in northern China. The research methodology consisted mainly of energy use behavior analysis, and simulation calculations of building energy consumption using the DeST energy consumption model, followed by strategies to change energy use behavior to reduce energy consumption. Mansó Borràs et al. [112] investigate the use of City Energy Analyst (CEA) for energy modelling and analysis of urban buildings to assess the potential of energy communities. The main tasks include collecting the necessary data on building characteristics, modeling the building using City Energy Analyst (CEA), generating energy demand and PV documentation for the building, defining different energy community configurations and energy sharing algorithms, as well as their main defined parameters, and finally simulating different energy community case scenarios and evaluating them through the introduction of Key Performance Indicators (KPIs). The performance of the KPIs introduced. It is an effective tool for assessing the energy situation and costs. Wang et al. [82] used a statistical model to investigate the impact of building energy consumption on the built environment around an urban park in Nanjing, China. The study methodology included the creation of grid groups to examine different built environment layouts around urban parks in terms of 3D building characteristics, land cover, and road networks. A spatial regression model was then used to examine the model including the relationship between the built environment and the domain of building energy consumption, with an emphasis on analyzing and comparing the positive and differential effects of the built environment on building energy consumption. Agile heating and cooling energy demand modelling for residential buildings using technologies such as 3D GIS was investigated by Prades-Gil et al. [90] The main research areas and themes of the article include conducting urban energy planning, studying the heating and cooling energy demand of buildings, and considering retrofitting and climate change. The research methodology includes the use of GIS technology to collect geometric information on buildings, solar radiation analysis, and energy demand analysis using the degree-day method. In addition, the model incorporates European standard methodologies and has been validated with several case studies. The results show that the model is a useful energy planning tool for predicting the energy demand of residential buildings and regions, as well as the impact of climate change on their energy demand and the consequences of countermeasures such as retrofitting. Ma et al. [92] investigated the pattern of window-towall ratio (WWR) on energy demand in residential buildings in different regions of China. The research methodology focuses on theoretical analysis and simulation calculations. A mathematical model of the building energy balance was developed to investigate how the window-to-wall ratio (WWR) affects the energy demand of a building. Simulations were then carried out using DeST software to verify the accuracy of this theoretical analysis method. The results show that the window-to-wall ratio (WWR) has a significant impact on the annual cooling and heating loads. The cooling load grows faster than the thermal load. The window-to-wall ratio (WWR) is an inherent geometric characteristic of the building itself, and this theoretical approach allows an accurate assessment of how the window-towall ratio (WWR) affects the building's energy demand. In addition, simulations where the absolute reduction in energy demand is closely domain with local climatic conditions are included, while the rate of change in heating and cooling loads varies little between cities. Jiménez Torres et al. [96] assessed the impact of climate change on building energy demand in Mexico by collecting climatic data and then calculating and analyzing this data. Meteonorm software was used to collect climatic data including ambient temperature, wind speed and direction, relative humidity, and solar radiation to study the effects of climate change on building energy demand in Mexico, particularly for single-family dwellings based on representative concentration pathways.
Quddus Tushar et al. [59] quantified, compared, and improved building design options through an integrated approach of BIM and energy simulation, aiming to reduce the carbon footprint and energy consumption of a residential building in Melbourne. The study used Autodesk Revit as the BIM tool, FirstRate5 as the energy assessment tool and sensitivity analysis with the @Risk Palisade decision tool to optimize operational energy efficiency. The study methodology included the use of BIM, energy rating tools, life cycle assessment to quantify, compare, and improve building design options, and sensitivity analysis to optimize operational energy efficiency. The results show that the use of passive design strategies such as orientation, shading, sealing, glazing, and insulation can significantly reduce the need for artificial heating and cooling systems, accounting for 40% of the total energy use in residential buildings. Insulation contributes 4% of the carbon emissions and 7% of the primary energy demand, although it only accounts for 1% of the total mass. Therefore, an integrated workflow of BIM and energy simulation software allows for indepth analysis of building solutions and identification of strategies to improve energy efficiency and environmental impact. The researchers [70] used digital twin technology to study carbon emissions and energy consumption in the Malaysian city of Bertam. The research methodology involved using digital twin technology to create a model of Bertam City and carry out simulations of energy performance. The simulations took into account digital twin data on factors such as location, solar radiation, building run times, glazing ratios, building materials, and building occupancy. The energy consumption and carbon emissions of the digital twin city were estimated based on four planning scenarios. It was concluded that with the optimization work, iCD estimated energy savings and carbon reductions of 8.6% for Bertam City. This study provides a benchmark for potential carbon emission reductions and energy savings. Further optimization work could achieve a potential energy saving and carbon reduction of 15.7% by converting the existing HVAC system to one with better energy performance. In addition, the installation of solar PV panels in selected residential areas could result in a further 24.95% energy savings and carbon emission reduction. Combining these optimization measures, the total potential energy savings and carbon reductions for the city of Bertram could reach 49.25%. Using a bottom-up approach, Turki Alajmi et al. [56] analyzed the energy consumption of various devices in detail by aggregating this data to predict the energy consumption of the entire building complex. The results of the study show that air-conditioning loads are the highest in residential buildings in Kuwait and are expected to grow exponentially from 2022 onwards, reaching a projected 60 TWh by 2040. Air-conditioning loads are the main component of energy consumption in residential buildings in Kuwait and are expected to grow exponentially. These results provide an opportunity to develop more effective energy policies as well as future energy efficiency initiatives. Seo et al. [67] used an assessment framework based on digital twin technology to conduct building modeling and energy analysis. The research methodology consisted mainly of creating operating schedules and generating operating schedules for lighting systems based on lecture schedules. The results of the study showed that the difference in energy savings between using the PIR system and the manager's manual control of the lighting was only 2%. In addition, LED lighting in most classrooms was found to be over-engineered and when illumination levels were adjusted to the appropriate level, energy savings of 46% could be achieved. Therefore, lighting control strategies can have a significant impact on energy consumption in educational buildings.
The increase in the number and range of applications of building energy simulation software has also led to the importance of examining the variability and mutual consistency of simulation results between these tools. Each software has its own unique algorithms, modelling approach, and treatment of different building and environmental parameters, which can lead to significant differences in the results obtained when modelling the same problem. Rodrigues et al. [63] developed a BIM model using Revit and GBS (Green Building Studio) to carry out an energy analysis of a two-story service building in Lisbon, Portugal. To validate the results of the GBS energy analysis, the EC0.AP software and EnergyPlus outputs were used for comparison. In terms of total annual energy consumption, the results obtained from Autodesk Revit's Energy Analysis were approximately 61% higher than the results from ECO.AP Energy Simulator. In Section 7, the article points out that GBS is primarily applicable in the early stages of the building lifecycle, where the energy simulation results can support decisions aimed at improving the building's energy performance during the operational phase. The input limitations of GBS, mainly in the customization of the HVAC system, affect the building's performance and energy performance assessment under actual operating conditions. In the discussion section, the article discusses the differences between the results obtained by the two approaches, mainly due to the limitations in the customization of HVAC systems in Revit and the fact that the 12 HVAC solutions available in Revit are based on the Commercial Building Energy Consumption Survey (CBECS) in the US and Australia, which makes them unrepresentative of the Portuguese situation. In addition, the different energy simulation engines used by the two methods (GBS uses DOE-2 and ECO.AP uses EnergyPlus) may also lead to differences in the results.
Gennaro et al. [76] used a variety of building energy simulation tools (EnergyPlus, TRNSYS, IDA ICE, and IES-VE building energy simulation tools) to model a naturally ventilated single-story double-skin façade, followed by validation and inter-software comparison. The results show that there are differences in the performance of the various building energy simulation tools in predicting, for example, chamber air temperature and surface temperature. For example, TRNSYS performed well in predicting the solar radiation transmitted through the double-skin façade, while IDA-ICE was less accurate in predicting high temperatures. Section 7 points out that there are some challenges with the building energy simulation tools in predicting the performance of naturally ventilated single-story double-skin façades, such as predicting chamber air temperatures and surface temperatures.
In addition, detailed studies should be conducted for specific opening types to model the flow more accurately through DSF chambers. The discussion section of the article focuses on the evaluation of the performance of four building energy simulation tools (EnergyPlus, TRNSYS, IDA-ICE and IES-VE) in predicting the thermal behavior of naturally ventilated single-story double-skin curtain walls. The article notes that it was not possible to determine which tool performed best for all the quantities analyzed (particularly the cavity air temperature, which was the least accurate parameter of all the software because of the underestimation of daytime peaks.) IES-VE was the most accurate in predicting supply air and thermal buffer patterns when shading devices were deployed, while EnergyPlus was the most accurate in predicting outdoor air curtain patterns. TRNSYS performed best in predicting surface temperature and solar radiation through the double-skinned curtain wall. In addition, the study explores the challenges designers may face when modelling naturally ventilated double-skinned curtain walls using whole-building simulation tools. In addition, the study clarifies which challenges have a greater impact on the performance of building energy simulation tools in order to improve their reliability.
Del Ama Gonzalo et al. [105] used multiple building energy simulation tools for modeling and energy analysis of office units in Boston, USA and Madrid, Spain to assess the cross-validation of energy simulation tools software tools and the energy efficiency of the buildings. A variety of software tools were used in the article for the simulation of energy management systems, including TRNSYS, IDA ICE, and EnergyPlus. The results showed that the mean percentage error values showed good agreement between the programs, with deviations between the results of the different software and the mean ranging from 0.1% to 5.3%. However, the hourly energy demand analysis showed normalized root mean square error values between 35% and 50%, which is well above the acceptable standard. It is recommended that multiple modelling tools be used in the early design stages and that detailed hourly analyses be conducted to predict the energy demand of buildings more accurately. In the study by Li et al. [42], they focused on energy consumption during the operational phase of the building life cycle. Using Autodesk Revit and EnergyPlus as BIM and energy simulation tools, they propose an effective technical framework to address the interoperability between the two tools. They used this framework to simulate the energy consumption of a Canadian home and carried out sensitivity and parametric analyses. The results of the analysis compared to monitoring data and the results of another simulation tool, HOT2000, predicted a total energy consumption difference of no more than 8.0%. Dimitrios Gourlis and his team investigated the practical application of BIM and BaMa in the manufacture of building energy models and digital twins for industrial facilities. They first used BIM technology to create and simplify building models, followed by BaMa for the construction and simulation of building energy models. The experimental results proved that the reliability of BaMa's hybrid simulations was confirmed by comparative tests with EnergyPlus and by comparison with monitoring data from real facilities. As a result, they concluded that BIM and BaMa can be effectively applied to the generation and simulation of building energy models and digital twins for industrial facilities. However, this process requires a moderate simplification of the BIM model to meet the specific requirements of the energy model. Table 2 shows the summary of 75 application literature related to building energy management that uses machine learning technology, indicating the year of publication, the machine learning technology used and the application content. According to the application direction of the literature, the research topics can be roughly divided into energy prediction, energy efficiency improvement, energy model calibration and evaluation, energy system control, etc. It is worth noting that many of these studies focus on the establishment of forecasting models, such as "short-term energy consumption forecasting", "heating and cooling load forecasting", "building energy use forecasting", and so on. Among the paper reviewed, the most widely used algorithms are Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Deep Neural Network (DNN). These algorithms can predict overall or specific energy demand and even further predict environmental impact or related parameters such as electricity prices, carbon emissions, etc. The establishment and application of these predictive models cover various types of buildings, including residences, offices, commercial buildings, educational buildings, and hotels.   In the study of modern building energy management, machine learning algorithms are widely used to improve the accuracy of energy demand forecasting. In the single-building domain, Sean Kapp [171] performed well using support vector machines (SVM) to predict energy consumption in industrial buildings. Qingyao Qiao [172] proposes a hybrid machine learning approach that considers human behavior and the interaction of building energy management systems to predict building energy consumption successfully. Tekler [170] presents a minimal sensing strategy for occupancy prediction using a comprehensive set of sensor data and deep learning methods, emphasizing the importance of indoor CO 2 levels and Wi-Fi-connected devices in predicting occupancy of different space types. The modelling model developed by Yan Zhang [175] using the Light Gradient Boosting Machine combined with the Shapley Additive exPlanation algorithm can predict the energy use and greenhouse gas emissions of residential buildings and identify the most influential variables. Yan's [108] multi-layer perceptron neural network can effectively predict the multi-energy load of buildings and found that the optimization algorithm impacts the prediction effect. Muhammad Faiq's team [182] used long short-term memory (LSTM) models and weather data to predict the energy consumption of institutional buildings, and the results were better than support vector regression (SVR) and Gaussian process regression (GPR). Ogwumike [146]'s research found that data segmentation and multimodel combination can improve the accuracy of energy use forecasts, and the forecast accuracy varies with the forecast range.

Algorithms and Deep Learning for Energy Efficiency
In building population energy, Sapnken et al. [173] studied the feasibility of machine learning (ML) for predicting building energy demand during the design phase. They found that deep neural network (DNN) is the most effective ML model and is independent of building populations or specific climate region's data impact. Anh-Duc Pham et al. [136] used the random forest model to predict the short-term energy consumption of multiple buildings and found that the random forest model showed good accuracy in prediction and had satisfactory generalization ability, better than M5P, and random tree models. Anand et al. [147] found that the prediction accuracy of the deep neural network model is slightly higher than that of the traditional model. However, the calculation speed of the gradient boosting model is faster. In addition, the relationship between occupancy and energy consumption may vary over time and space, and multilevel modeling is required. Liu et al. [142] applied linear regression and gradient boosting machine models to predict and analyze building energy consumption analysis building data with weather conditions provided by ASHRAE, and the results were also satisfactory.
In the field of building consumption, Sehovac et al. [131] adopted the Sequence-to-Sequence (S2S) deep learning algorithms based on Gated Recurrent Unit (GRU) and Long Short-Term Memory (LSTM) as a machine learning model. The results show that the GRU S2S model outperforms the LSTM S2S, RNN S2S, and deep neural network models in terms of short-term, medium-term, and long-term prediction lengths. Liu et al. [77] used artificial neural networks (ANN) and building network analysis to predict energy use in multiple buildings. This method uses the input layer vector, the hidden layer vector, the output layer vector, and the weight vector between the layers to make predictions. X.J. Luo [144] used three machine learning techniques (artificial neural network, support vector regression, and long short-term memory neural network) to predict the building's energy load and BIPV electricity production. The results showed that the ANN-based predictive model produced the smallest mean absolute percentage error, while the SVM-based model took the shortest computation time.
In urban architecture, Robinson et al. [117] studied the energy consumption of commercial buildings in specific cities, such as in the United States in New York City and Atlanta. They found that the gradient-boosting regression model performed best. Seyedzadeh et al. [128] used machine learning models to predict building energy consumption load. Gradient Boosted Regression Trees (GBRT) are proven to provide the most accurate predictions, Support Vector Machines (SVM) are suitable for simple datasets, and multilayer neural networks (NN) are suitable for complex datasets. NN can estimate energy needs faster. Wei et al. [118], has used six methods to estimate household gas and electricity usage in London, including full linear, Lasso, MARS, SVM, bagging MARS, and boosting. They found that all nonparametric models performed better than linear models. SVM models performed best in gas and electricity forecasting, bagging MARS second only to SVM in gas forecasting, and Lasso models with similar predictive power to fully linear models. Culaba et al. [138] uses machine learning techniques to cluster and predict energy consumption in mixed-use buildings. They employed a k-means algorithm for clustering and a support vector machine for prediction, which showed improved prediction accuracy compared to previous models. Mohammadiziazi et al. [139] used random forest and extreme gradient boosting models to successfully predict commercial buildings' annual energy use intensity (EUI) in the United States.
In terms of residential buildings, on the one hand, Feng et al. [152] used the XGBoost model and public data to predict the air-conditioning energy consumption of residential buildings in three different climate zones in the United States. The results showed that the model performed well (the overall R2 value was about 97% and the cross-validated R2 value 92%) and can help homeowners understand building performance and support energy efficiency decisions. On the other hand, Lan Wang et al. [149] quantified the impact of driving factors on residential buildings in the United States by combining machine learning and Monte Carlo methods, revealing that the relationship between total energy end use and total building area, number of rooms, number of windows, and indoor heating temperature setting Correlations of factors such as location, age of residents, and complex nonlinear effects of heating and cooling degree days also found that harsh climates may offset the effect of insulation level.
For public buildings, Ding [153] used six machine learning algorithms and recursive feature elimination to predict and analyze the importance of energy consumption features of public buildings. He found that the contribution rate of the first ten features exceeded 80%. When the number of features continued to increase, the marginal utility decreased significantly, highlighting data sets critical impact on model performance. At the same time, Yue [180] used six meta-modelling techniques, including RDPG, A3C, LSTM, CNN, ANN and SVR, to predict the hour-level multi-performance vector of the gymnasium and found that the RDPG model prediction is the most accurate. The LSTM has high efficiency, which is suitable for emphasizing time and accuracy users, while the ANN model is easy to use and robust, suitable for time-limited users.
In terms of early building energy consumption prediction, Singh's team [148] collected training data through incremental and enrichment methods. They found that enrichment methods are better at filling the design space and reducing generalization errors while emphasizing the analysis of sample distribution to improve the importance of generalization ability. On the other hand, Veiga [158] used machine learning and EnergyPlus simulations to estimate the energy use intensity of bank branches and found that lighting power density and weather variables are vital factors, and the support vector machine (SVM) model has the highest accuracy, which can be applied to future benchmark tests and other types of buildings.
Machine learning techniques are also widely used to optimize prediction accuracy for building cooling and heating load simulation. Alyakoob et al. [176] used microclimate data and machine learning to predict cooling load. They found that microclimate variables such as average air temperature and absolute humidity had a significant impact. At the same time, Tsanas and Xifara and Pham et al. [113,136] used statistical machine learning and multiple machine learning methods to quantify residential buildings' energy performance and predict office building cooling load, and both achieved minor prediction errors. Moreover, the artificial neural network model with a bagging ensemble outperformed other models in prediction accuracy. In addition, Yang [156] introduced a machine-learning model based on dynamic feedback, which significantly reduces the computational load and maintains good energy and thermal comfort performance. In addition, Ceballos-Fuentealba et al. [132] used JAVA programming and parameter optimization strategies to predict and optimize the application of building energy efficiency with high computational efficiency. Chen [168] adopted Broad Learning System (BLS) technology for building modelling and energy analysis. This model has higher prediction accuracy and fast real-time update capabilities compared to traditional models. The Build2Vec spatial model developed by Abdelrahman et al. [169], combined with building information modelling (BIM) and other data, has also improved prediction accuracy.
In building energy consumption prediction, Amasyali and El-Gohary [150] found that four algorithms (CART, EBT, ANN, and DNN) performed well, among which the DNN model with four hidden layers was the best. Research by Gao's team [127] shows that ran-dom forest models best predict heating loads in energy-efficient buildings. Zhou et al. [154] compared 15 different machine learning models and found that the Gaussian process regression model and support vector machine performed well in prediction accuracy, model stability, and calculation speed. Ikeda and Nagai [155] proposed a method of mixing meta-heuristics and machine learning, which can optimize the daily operation plan of building energy systems and reduce daily operating costs by more than 13.4%. Studies by Guo et al. [121] and Gao et al. [127] have shown that machine learning is effective. Guo et al. used various machine learning methods, especially the extreme learning machine model performed best under feature set 4, and the optimal number of nodes was 11. On the other hand, Gao et al. [127] found that wall area, total height, orientation, and glaze area are the main features affecting building energy consumption, and the prediction performance of RF, lazy K-star, RDT, and AMT models was excellent. Furthermore, feature reduction can improve model performance, and the most accurate MLP models provide an optimized structure for predicting HL and CL variables. J.-M.L. et al. and Xinyue Li et al. [96,181] predicted the energy consumption of air conditioners in buildings by using a variety of machine learning algorithms, such as SVM, RF, MLP, DNN, RNN, LSTM, and GRU, and found that the performance of RF and SVM particularly prominent. Research by Rahman and Smith [120] demonstrates that machine learning predictions of fuel consumption in commercial buildings are also feasible in different climates and building types. De Araujo Passos et al. [179] used surrogate models and MPC methods to optimize the energy retrofit for specific buildings and realize the maximum passive operation of the HVAC system. In the application of digital twin technology, studies by Anders Clausen, H. Hosamo Hosamo, Yasmin Fathy, and S. H. Khajavi [134,160,163,185] have shown that digital twin technology can significantly improve the energy efficiency of buildings through building modelling, energy analysis, and multi-objective optimization and efficiency and reduce energy consumption.
For other specific energy consumption predictions, Hosseini et al. [135] used general circulation models to assess the impact of climate change on building energy performance. Li and Dong [119] developed a Markov model based on change point analysis to improve occupancy prediction accuracy for commercial buildings. López-Pérez and Flores-Prieto [177] used artificial neural networks and adaptive neuro-fuzzy inference system models to significantly improve thermal comfort and save energy in educational buildings in tropical climates. Martinez-Soto et al. [161] conducted architectural modelling and energy analysis on Chilean houses, which were successfully applied in different climate regions. Using data-driven methods and digital twin technology, Hodavand et al. [184] enabled real-time data collection and analysis to improve resident comfort and ensure sustainable operations.
In energy efficiency analysis and improvement, Omar Mata et al. [178] estimates the human metabolic rate by depth vision sensor and RNN to adjust the set point of HVAC equipment in real-time to achieve effective energy saving. Nutkiewicz et al. [116] combine data-driven machine learning models and physics-based energy simulations to predict and understand the impact of urban environments on energy consumption more accurately. Rätz et al. [125] used multiple machine learning methods for data-driven modeling of building energy systems, enabling prediction, potential analysis, and optimization of control strategies. Fan et al. [122] developed a building energy performance model based on interpretable machine learning, which helps to understand the reasoning mechanism of the predictive model and balance model complexity and interpretability through the evaluation of "trust". Haosen Qin et al. [174] used reinforcement learning and deep Qlearning to optimize the control of HVAC systems, improving energy efficiency and thermal comfort. Dian Zhuang's team [186] presented a data-driven predictive control approach that uses time-series prediction and reinforcement learning to optimize HVAC operations in IoT smart buildings, resulting in energy savings and improved thermal comfort.
Choid and Kim [109] used the MOB algorithm and recursive partition method to study the impact of energy variables on energy consumption, revealing the correlation between building energy consumption and specific design and system operation. This provides an essential basis for the energy-saving design of American office buildings. At the same time, Tan's team [164] used the YOLOv4 algorithm and digital twin technology, combined with environmental perception and keyframe similarity mechanism, to achieve a significant improvement in the energy efficiency of indoor lighting, reducing operation and maintenance time and costs. In addition, Wei Tian's team [157] used ten machine learning algorithms to deeply explore the energy characteristics of urban buildings through model adjustment, variable importance analysis, and local spatial analysis.
For calibration and evaluation of energy models, Naganathan's team [115] proposed a new algorithm for optimizing energy loss through semi-supervised machine learning and clustering algorithms for building energy modelling. The team also developed a website to visualize the automated SSEM model. Zhu et al. [143] combined approximate Bayesian computation and machine learning for fast and reliable calibration of building energy models created with the EnergyPlus program. Zhou et al. [154] used extreme learning machine (ELM), multiple linear regression, support vector regression, and other methods to develop prediction models. They proposed a prediction time step strategy based on building thermal response time. Feature clusters perform best.
Sofia Agostinelli et al. [162] use digital twins and AI to improve the cyber-physical building energy management system and develop an intelligent energy grid management system. The results show that DT-based real-time monitoring can reduce the gap between energy performance and actual building performance. Abigail Francisco's team used algorithms such as decision trees, support vector machines, and digital twins to optimize real-time energy pricing and provide more specific insights into efficiency. You et al. [166] combined digital twin technology and deep neural network to improve the efficiency of multi-vector and integrated energy systems and reduce operating costs. Research by Huang et al. [167] emphasizes the value of digital twins in planning, design, and operation optimization. The research of Haidar Hosamo Hsamo's team [185] aims to improve the comfort of building users through digital twin technology, combined with probabilistic models and provides adequate support for decision making.

BEM Software
Energy simulation software is widely used nowadays, such as EnergyPlus, but they have a series of challenges and limitations in its use. First, a common limiting factor in research is model uncertainty due to its internal complexity, resource constraints during simulation, and the lack of components [84,104]. Furthermore, this simulation software is relatively complicated to operate. It is not friendly to new users, which can easily lead to their misuse, thus further increasing the uncertainty and error of the model [102,105]. COVID-19 has also had a profound impact on global energy consumption. Differences between actual energy consumption and modelled energy results are becoming increasingly significant [187]. At the same time, although increasing the number of simulation runs can improve the accuracy to a certain extent if the number of iterations is increased too high, it may negatively affect the accuracy of the final data analysis results and increase the error rate [64].
In addition, such simulation software may ignore some critical factors in the simulation process. For example, in-depth studies of the calorific value of building structures [64], human actions such as improper temperature set points, and improper HVAC system design and operation [75]. In using tools such as Revit, for example, the HVAC system cannot be created, and an existing HVAC system needs to be selected from the software library [51].
New building design challenges, such as unconventional wall designs and properties and complex building material properties, may also increase the difficulty of simulation, leading to inaccurate simulation results [79]. Finally, in the energy investment decision problem, the diversity of assessment tools makes it possible to analyze all sustainability aspects. However, the diversity of methods only sometimes leads to consistent assessment results [81].
In general, although the use of energy simulation tools provides essential support for building design and energy decision-making, in practical applications, due to various factors, data needs to be cleaned and processed more carefully to reduce simulation errors [90].
Several vital areas need to be focused on for the future directions of energy simulation software. First, the energy performance analysis of the building life cycle can be explored in more depth, including building material parameters and the role of different cooling loads in the energy performance analysis [62]. This may require varying the thermal properties of building materials to study the impact of this parameter on building energy consumption [64].
Second, a more detailed study of the layout and configuration of buildings is also a meaningful direction. This may include looking at random layouts of buildings and considering factors such as shading devices for different buildings [58]. At the urban scale, the use of 3D GIS for urban energy planning, the study of building heating and cooling energy requirements, and the consideration of building retrofits and climate change are future research priorities [90].
Determining the weight of the objective function according to the city's climatic conditions also requires further study [74]. As for the energy system, it can be considered to promote the use of renewable energy systems, conduct more energy consumption analysis, and in practical applications, develop a thermal energy storage system to decouple the availability, demand of thermal energy, and improve the control function to improve Efficiency of the system in cities with different climates [74,102].
Finally, new research and development also require continuous testing and validation to ensure new methods' validity [105]. In general, the future development of energy simulation software will pay more attention to a comprehensive analysis of building life cycles, more detailed architectural and urban planning research, more diverse energy system analysis, and continuous method testing and improvement.

BEM Algorithms
We are in the early stages of exploring machine learning applications for energy simulation and are facing several challenges, mainly in data and model selection, prediction accuracy, interpretability, computational complexity, and interdisciplinary collaboration.
Data is the foundation of machine learning models; thus, it is quality and availability that directly impact model performance [174,188]. For example, on the one hand, lack of detailed data and poor data accuracy can affect the accuracy of prediction results [151]. On the other hand, to accurately predict energy use, models require a large amount of historical data [151,182], which may only sometimes be available, especially for resourcepoor areas [177]. In addition, the absence and handling of data, and the lack of a specific energy code and standard to refer to, are also current issues [109].
Regarding model selection and prediction accuracy, appropriate features and construction are essential for the accurate estimation of machine learning models for the energy use [171] and how to select the suitable machine learning algorithms and parameters for a given dataset and problem. How to select suitable machine learning algorithms and parameters for a given dataset and problem. In the case of building energy simulation, the diversity of building characteristics (e.g., geographical location, building type, age of the building, etc.) can also affect the performance of the model [175,182], making the computational accuracy of the model a challenge.
In addition, machine learning models' interpretability and computational complexity are very important issues. The decision process and results of the model may be more brutal to understand [94,174,189]. Challenges in this area also include how to deal with and understand the impact of human behavior on building energy prediction, a complex issue requiring more research to address [172,180]. The computational time required to run simulations with large areas and high-resolution data can be a significant drawback for the computational complexity [176]. Furthermore, optimizing, and tuning models require significant computational power, while manual modelling requires expert knowledge and significant working time [125].
Furthermore, interdisciplinary collaboration and integrated analysis are critical to successfully addressing the above challenges. For example, building energy efficiency requires consideration of multiple factors such as building design parameters, weather conditions, equipment systems, and occupant behaviors [180,181], and this requires interdisciplinary collaboration and integrated analysis [177]. In addition, climate change will exacerbate the challenges of modelling and forecasting building energy use, and more research and practice is needed to refine and validate methods for this emerging field [177].
Based on the literature covered in this survey, several possible future research directions exist for applying machine learning in building energy simulation. The first is that the generalization capability and accuracy of the models can be explored in depth. Models need to be continuously validated and improved to increase their reliability [172,173]. For example, the model's performance can be improved by integrating more industrial data to model complex non-linear relationships [171]. Also, the scope of application of the models needs to be further expanded, and these ML models will be expected to be applied to a broader range of building types and climatic conditions, as well as to more prominent scale buildings [178].
In terms of data, massive datasets are needed for model training and testing [173], while data quality and availability are also expected to improve [174]. Future research could also focus more on empirical data such as building samples, occupancy rates, and building use patterns to support the application of models [182]. Integrated sensor technologies can also be further explored to provide valuable insights using machine learning algorithms for the energy simulation [176].
In terms of machine learning methods themselves, we could explore additional machine learning algorithms and agent models [172], determine parameter settings for machine learning models [126], optimize or 'tune' models for optimal prediction accuracy and consistency [128], and further development and refinement of different machine learning workflows, such as the DUE-S workflow [116]. At the same time, improving the explanatory power of models is an important research direction. This may involve using more explanatory machine learning models (e.g., decision trees or linear models) or developing new techniques to explain more complex models, such as deep learning models. In addition, integrating model interpretation and reliability into model design is an important issue [171,174]. This may help us understand how the model makes predictions and provide deeper insights, such as identifying the most important features or discovering hidden patterns and relationships. This understanding may also lead us to improve the model or adapt our strategies and methods to optimize the energy efficiency of the building.

Conclusions
The contribution of this article is a review of the latest literature on energy consumption throughout the life cycle of buildings and sustainable building energy management. The numerous recent literature reviews identify the continued importance of controlling energy consumption for future sustainability impacts in terms of buildings. The literature reviews numerous applications of the use of Digital Twins (DT) technology in various areas of building energy analysis or prediction, the use of building information models in conjunction with energy analysis software for building energy optimization, and prediction and the ability of programming tools to monitor building energy in real time. It is demonstrated that current DT technologies can improve the efficiency of building energy management and reduce energy costs, thereby achieving building sustainability goals. In particular, this article focuses on the latest applications of DT and BIM technologies for building energy optimization, describing the use of traditional energy analysis software to simulate building energy performance, including thermal loads, energy consumption, internal temperatures, and other parameters with high accuracy and the use of building energy simulation results with multi-decision tools to evaluate the building materials, building design, and internal building services. There is also a growing interest in energy analysis software with programming tools for energy management, indoor environmental monitoring, and building energy efficiency analysis. However, these methods and tools also face challenges, such as model uncertainty, data quality issues, model complexity, and the need for interdisciplinary collaboration, which limit their performance and prediction accuracy in practical applications.
Despite these limitations and challenges, future research directions have emerged: more comprehensive analysis of the building life cycle, detailed architectural and urban planning studies, promotion of the use of renewable energy systems, diversified energy system analysis, and Use of machine learning more effectively to improve model generalization and accuracy. For energy analysis software, the complexity and uncertainty of building models and the diversity of energy efficiency assessment methods can lead to errors in simulation results. In addition, the availability and validity of input data are particularly important for energy simulation software and machine learning algorithms, which determine the applicability of the models. In addition, there are further issues with the application of machine learning, such as the interpretability and complexity of the models, the requirements for computing power, and the impact of environmental changes on the predictive effectiveness of the models, and more effective use of machine learning to improve model generalization and accuracy.
Numerous reviews of the literature have highlighted future research directions combining machine learning techniques with DT techniques, summarizing that the field of building energy analysis together with machine learning techniques can improve the accuracy of energy prediction, optimize energy efficiency, and improve management processes. And, this hybrid technology is currently the focus of much of the literature and has the potential to facilitate the transition to clean energy and ultimately the achievement of sustainable development goals. For future research and development of building energy simulation, continuous verification and improvement of models, improvement of data quality, strengthening of interdisciplinary cooperation, and enhancement of model interpretability will play a key role.