Towards Sustainable Mosques: A Review of AI and ML Approaches for Energy Use Forecasting

Abstract

This thorough study looks at the use of machine learning (ML) techniques to forecast energy usage in buildings, with an emphasis on mosques. As energy use has a greater impact on both the environment and the economy, it is becoming increasingly important to optimize energy usage in buildings, especially for religious organizations such as mosques. The study goes into a variety of ML methods and models, including neural networks, regression models, decision trees, and clustering algorithms, each customized to a distinct difficulty in energy management. The paper evaluates the efficacy of several ML techniques, noting their merits, shortcomings, and potential applications. Additionally, it investigates the impact of climate, mosque design, occupancy patterns, and geographical variables on energy use. To achieve accurate energy consumption projections, rigorous data collecting, pre-processing, and model validation procedures are required. The paper also discusses important data sources and methodologies for mosque-specific energy analysis. Furthermore, the study emphasizes the practical benefits of applying ML in energy prediction, such as cost savings, increased environmental sustainability, and better resource allocation. This study’s ramifications extend beyond mosques, providing useful insights into energy management in buildings in general. By summarizing the current state of ML applications in mosque energy prediction, this study is an important resource for researchers, decision-makers, and energy management practitioners, paving the way for future advancements and the adoption of more sustainable energy practices in religious institutions.

1. Introduction

Building energy efficiency is a major topic under the Sustainable Development Goals (SDGs) and global energy initiatives [1]. In a 2022 review, the United Nations Intergovernmental Panel on Climate Change stressed the potential of smart technology to mitigate climate change and advance the SDGs [2]. The use of intelligent construction equipment, such as the IoT, AI, and ML, gained traction in solving energy usage and efficiency issues. IoT-enabled technic plays a pivotal role in management of energy consumption in buildings. AI applications are leveraged to forecast energy usage patterns and identify opportunities for conservation. ML, with its computations driven by data, is useful in improving energy use and devising novel energy-saving measures [3]. Globally, the construction sector stands as a significant consumer of primary energy resources [1]. Alarming statistics from the World Green Building Council underline the environmental impact of the building sector, accounting for nearly 36% of the world′s energy consumption, 39% of total carbon emissions, and half of all resource utilization. Projections suggest this environmental footprint could double by 2060 [2].

Various policies and standards promote energy efficiency in construction, including building codes, incentive programs, and benchmarking requirements [3,4]. These measures help reduce energy consumption, lower operational costs, and enhance indoor comfort over a building′s lifecycle. Zero-carbon-ready standards and transparency policies further support sustainability by incorporating lifecycle emissions and resilience metrics [5]. Tools such as the EPA’s ENERGY STAR Portfolio Manager assist municipalities in identifying and improving underperforming buildings [6]. In the broader movement toward sustainable and intelligent buildings, mosques present a unique opportunity to act as culturally significant drivers of energy innovation. Beyond their primary function as places of worship, mosques are embedded in the daily lives of millions and are often centrally located within communities, giving them high visibility and symbolic value. Leveraging this position, the integration of AI-based energy management in mosques offers not only technical benefits, but also a platform for promoting sustainability within broader society [7]. Mosques operate on distinct usage patterns dictated by religious practices for frequent yet short occupancy periods spread across the day, with higher peaks during congregational prayers and special religious occasions. This dynamic nature requires flexible and responsive energy systems, an area where AI excels. Rather than relying solely on predictive models of long-term usage, AI tools can adapt in near real-time to behavioral and environmental changes, optimizing systems such as heating, cooling, lighting, and water usage with high precision.

Furthermore, the transformation of mosques into smart, low-impact buildings holds powerful educational and symbolic potential. As trusted and frequented spaces, mosques can exemplify how religious architecture aligns with environmental ethics, encouraging local communities to adopt similar technologies in homes, schools, and businesses. This alignment between tradition and innovation reinforces the role of mosques as not just spiritual centers, but as active participants in ecological stewardship and sustainable urban development.

To this end, smart green buildings are supposed to be more advantageous in terms of energy and emissions saving; as well as reduced maintenance and operation costs [8]. In a rapidly evolving landscape, smart technologies such as artificial intelligence (AI) and ML are crucial [9]. They optimize energy usage, identify conservation opportunities, and contribute to overall sustainability goals in building operations [10]. Artificial intelligence (AI) can, for instance, be used to enhance building design to maximize thermal and lighting efficiency, as well as to identify and respond promptly to system irregularities. Aside from predicting energy consumption, AI can also improve energy management [11].

Furthermore, AI can be used to monitor building energy and water consumption to provide real-time feedback and optimize the energy use and costs, as well as to detect failures in the energy systems [12,13]. AI could also lead to the creation of environmentally friendly and more effective materials, along with the growth of RES, resulting in more effective and environmentally friendly power systems [14]. Some alternative works deal with many attractive materials to perform insulation and energy saving in smart buildings [14,15]. The actual trend is mainly related to the use of PCM materials due to their high storage density and ability to enhance thermal comfort [16]. Using in the Moroccan climate, PCM materials can be an appropriate solution and can profitably reduce energy demand and CO₂ emissions [17]. For example, vanadium dioxide (VO2) was selected for smart buildings because of its ability to regulate the infrared reflectivity according to the ambient temperature [18,19].

Another alternative is the use of materials such as carbon nanotubes (CNTs), which could store energy and function as insulation. CNTs have the potential to save energy and money while improving thermal comfort in buildings [20]. For example, a study by the University of California, Berkeley found that CNTs could save up to 50% on energy use in smart buildings [21]. Other materials used for insulation and energy efficiency in smart buildings include phase change fabrics (PCF) [22], aerogel insulation [23], aerogel-based thermal insulation systems [24], aerogel blankets, and insulating paints [25]. These materials provide excellent insulation performance, allowing significant energy savings over traditional insulation materials. Additionally, they are lightweight and non-toxic, making them safe and easy to install.

On the other hand, many authors highlight the need to integrate the smart HEMS ‘home energy management system’ to improve the integration of renewable energy and energy efficiency in the Moroccan building sector [26,27]. HEMSs typically involve a combination of hardware, such as smart thermostats and energy monitors, and software, such as mobile apps, to provide real-time energy usage data and allow for remote control of various devices in the home. They can also communicate with other devices, such as solar panels and battery storage systems, to optimize energy usage and reduce reliance on the grid [28]. The same emphasis was placed on energy performance modeling utilizing simulation tools such as Energy Plus [29] and TRNSYS [30] to optimize building energy use. These simulation programs employ mathematical models to forecast an establishment′s energy use, considering aspects such as the climate, population trends, and building architecture. In addition, there has been a growing interest in the development of innovative solutions and optimization strategies for the construction sector, such as the development of AI-based control systems [31], the use of ML-based building automation [32], and the integration of intelligent control systems [29]. Recently, research on AI has been developed, such as multiple linear regression (MLR), decision tree (DT), ANN, RNN, and embedded systems [30]. These approaches are used to investigate the non-linearity of building energy needs with the collected meteorological data [33]. In the realm of building energy management, ANNs have been used to anticipate a building’s energy consumption, and their usefulness has been demonstrated when compared to traditional approaches [34]. ANNs can generate highly accurate predictions, as they are able to learn from the data given to them and adapt to the changes in the environment. Additionally, ANNs are used to optimize the energy usage of a building, such as controlling air conditioning and heating systems [35]. RNNs have also been used to predict energy utilization in buildings [36]. RNNs can capture the temporal dependencies between the energy utilization of a building and the meteorological data. Furthermore, RNNs are capable of making predictions based on data from previous time steps, making them more effective in predicting energy use than traditional methods [37]. Furthermore, AI techniques such as support vector machines (SVMs) and deep learning (DT) have been used to analyze a building′s energy usage and determine the most critical elements influencing energy consumption [38]. These strategies are helpful in finding the elements that are crucial for the prediction of energy consumption and in generating models that can effectively anticipate a building′s energy usage. Therefore, AI techniques have proven to be effective in predicting and optimizing the energy usage of a building. However, ANNs are the most adopted models for building energy forecasting [39]. Data are fundamental for predictive models, as ML algorithms rely on large, diverse datasets for training and accuracy [34]. Enhancing big data methods is critical for smart energy management and informed urban planning [40]. Building energy forecasting requires inputs from various sources, such as sensors, meters, weather data, and occupancy patterns, to provide reliable insights [41]. The rise of smart technologies and building management systems significantly increased data availability [42]. Energy modeling approaches are typically categorized as physics-based (white-box), data-driven (black-box), or hybrid (grey-box) models [43]. Among these, data-driven models gained popularity due to their speed and high prediction accuracy [38]. They are widely used for profiling, forecasting, and performance evaluation [44], relying on relationships between independent variables such as time series data, weather, and building characteristics and energy consumption [45,46]. Despite their strengths, data-driven methods face challenges related to data quality, generalizability, and the need for technical expertise [41]. These approaches are divided into traditional techniques (e.g., MLR, scheduling graphs), which struggle with nonlinear patterns [47,48], and AI-based methods (e.g., ANN, SVM, and deep learning), which handle complex and large datasets more effectively, especially for long-term predictions [49,50]. In this context, mosques pose unique challenges due to their intermittent occupancy and unpredictable energy use, often overlooked in mainstream energy efficiency programs. Accurate load forecasting for such buildings may rely on statistical, physical, or AI-based models, with AI offering promising adaptability for these distinct patterns [51].

Methods involving statistics are further classified as time series models and ML methods. In one study [52], 19 regression algorithms were used to provide short-term electricity demand projections for the Memorial University the campus in Canada. These techniques fall into five categories: linear regression (LR), regression trees, support vector regression (SVR), Gaussian process regression (GPR), and ensemble approaches.

The university′ energy meter provided hourly electrical demand data, while meteorological data came from Newfoundland and the government of Canada. Weather factors considered were dry bulb temperature, humidity, precipitation, wind chill, wind direction, wind speed, and visibility. The analysis found that the GPR method had the greatest prediction accuracy, with a root mean square error (RMSE) of 301.76 kW and a correlation coefficient (R2) of 0.97. However, this study did not investigate the best feature set. In another study [53], SVR, artificial neural networks (ANN), and GPR were used to forecast short-term electrical requirements for the UK power system over a 30 min and 24 h period.

The RMSE ratios for accurate predictions over a 30 min prediction horizon were determined as 159.71 MW, 291.39 MW, and 179.71 MW for SVR, ANN, and GPR. The research found that the SVR technique was the most effective strategy for load prediction when compared to other algorithms. During the COVID-19 pandemic era, Ref. [54] built short-term power load forecast models for four countries: the United States, Germany, France, and the United Kingdom.

Mosques, as central institutions in Muslim communities worldwide, are increasingly being recognized not only for their spiritual and cultural roles, but also for their potential to lead in sustainable development. Traditionally seen as places of worship and community gathering, mosques are now integrating renewable energy systems, eco-friendly materials, and artificial intelligence (AI) technologies, particularly in energy management. This transformation reflects a growing commitment to ecological responsibility and intelligent infrastructure.

The use of AI in mosque energy systems marks a shift from reactive to predictive energy management. By analyzing occupancy patterns driven by prayer times, Friday gatherings, and seasonal events such as Ramadan, AI models can forecast demand with greater precision than traditional systems. Techniques such as long short-term memory (LSTM) networks, recurrent neural networks (RNNs), and extreme gradient boosting (XGBoost) effectively capture these temporal and behavioral dynamics, enabling optimized control of HVAC, lighting, and ablution systems. Smart sensors further enhance efficiency by detecting usage anomalies and reducing water and energy waste. These efforts align closely with several United Nations Sustainable Development Goals (SDGs), including SDG 7 (Affordable and Clean Energy), SDG 11 (Sustainable Cities and Communities), SDG 12 (Responsible Consumption and Production), and SDG 13 (Climate Action). Due to their visibility and social importance, mosques have a unique capacity to influence public behavior and promote sustainable practices within broader communities. Nonetheless, there are challenges to address. The lack of high-resolution, mosque-specific datasets hinders accurate model training, while cultural sensitivity, implementation costs, and data privacy remain ongoing concerns. Future research should prioritize context-aware, scalable, and interpretable AI models that can operate effectively in diverse mosque environments. In this context, mosques are emerging as powerful symbols of the convergence between tradition and innovation. By embracing AI-enhanced sustainability, they serve not only as places of worship, but also as beacons for ecological and technological transformation, guiding communities toward a more resilient and equitable future. Figure 1 below illustrates the conceptual framework of AI integration into mosque energy systems, highlighting data inputs, algorithmic processing, and responsive system control loops.

2. Research Methodology

This study adopts a structured and interdisciplinary research methodology to investigate the energy consumption characteristics of mosques and explore the potential role of artificial intelligence (AI) and machine learning (ML) in enhancing energy management. The primary aim is to understand the unique operational and architectural aspects of mosques that influence their energy behavior, and to assess how emerging technologies can be leveraged to support energy efficiency in alignment with global sustainability targets, particularly the United Nations Sustainable Development Goals (SDG 7 and SDG 11).

The research process followed a four-stage path. In the initial phase, a comprehensive literature review was conducted using major academic databases such as Scopus, Web of Science, IEEE Xplore, ScienceDirect, and Google Scholar. Keywords such as “mosque energy use,” “AI in building energy,” “sustainable mosque architecture,” and “machine learning forecasting” guided the search. Peer-reviewed publications from 2010 to 2024 were prioritized, with a particular focus on studies directly addressing mosque buildings, their energy usage patterns, and the application of AI/ML for energy optimization. Over 150 relevant sources were retrieved, with approximately 90 selected for in-depth analysis based on relevance and academic rigor.

In the second phase, the selected studies were categorized thematically to capture the multidimensional nature of the topic. The themes included mosque architectural typologies, energy usage patterns (especially HVAC and lighting systems), AI-based predictive modeling, water and thermal management in ablution spaces, and comparative assessments with other building types such as schools, offices, and residential structures. This thematic classification allowed for a comprehensive understanding of both the technical and cultural dimensions influencing mosque energy profiles.

The third phase of the research involved constructing an analytical framework to evaluate and synthesize the collected data. A strengths, weaknesses, opportunities, and threats analysis (SWOT) was developed to critically assess the integration of AI and machine learning in mosque energy management. In parallel, a comparative matrix was formulated to highlight the distinctive energy demands of mosques in contrast to conventional buildings, emphasizing the challenges posed by intermittent occupancy, religious scheduling, and architectural openness. Additionally, several regional case studies from countries including Malaysia, Saudi Arabia, Morocco, Turkey, and Indonesia were analyzed to provide contextual depth and demonstrate successful energy-saving interventions in diverse climatic and cultural settings.

The final phase of the methodology focused on synthesizing the findings and aligning them with broader sustainability objectives. Key results were interpreted in the context of SDG 7 (Affordable and Clean Energy) and SDG 11 (Sustainable Cities and Communities), emphasizing the need for context-specific, culturally sensitive, and technologically advanced solutions for mosque buildings. The study concludes by proposing future research pathways and implementation strategies that promote interdisciplinary collaboration among architecture, environmental engineering, and computational intelligence to advance sustainability in Islamic religious infrastructure.

This methodological approach ensures both depth and breadth by combining technical insights with socio-cultural considerations, ultimately contributing to the development of smarter, more efficient, and environmentally responsible mosque buildings (Figure 2).

3. Current Status and Challenges in Mosque Energy Management

3.1. Geometric Elegance and Symbolism

Mosques′ architectural designs are influenced by the regional traditions of the period and location where they were erected, resulting in a diverse range of styles, layouts, and decorations [55]. However, many architectural traits emerge in mosques around the globe due to their shared role as a venue for collective prayer [56]. Some of these features include the following:

-

Sahn (courtyard): A courtyard is a common feature of mosque architecture. It is an open space that is often surrounded by arcades and provides a place for worshippers to gather before and after prayer.

-

A minaret is an elevated structure belonging to a mosque that announces the call to pray. Minarets come in a variety of shapes, from the iconic spiral minaret of Samarra to the tall, pencil minarets of Ottoman Turkey.

-

The qibla wall is the mosque′s holy portion, facing Mecca. The qibla wall frequently includes a mihrab, a niche in the wall that symbolizes the direction of Mecca.

-

Dome: Some mosque designs include numerous domes in their construction, whereas some only have one. In mosques with a single dome, it is always placed over the qibla wall.

-

Decoration: Mosques often feature decorative elements such as calligraphic friezes or cartouches with prominent inscriptions, often from the Qur′an.

In addition to traditional mosque architecture, there are also modern mosques that blend contemporary design with age-old traditions (Figure 3). For instance, the Henning Larsen Architects designed a contemporary mosque in Copenhagen that displays a sculptural cluster of domes in a felicitous mix of Nordic and Islamic styles. Overall, the architectural designs of mosques around the world are a testament to the creativity and ingenuity of human beings, as well as the importance of religion and culture in shaping our built environment [57].

Figure 3. Various classifications of mosques based on design schools [58]; (a) conservative; (b) vernacular (c) adaptive; (d) modern; (e) postmodern; and (f) high-tech.

Figure 3. Various classifications of mosques based on design schools [58]; (a) conservative; (b) vernacular (c) adaptive; (d) modern; (e) postmodern; and (f) high-tech.

3.2. Distinctive Energy Use Characteristics of Mosques

Mosques possess distinct operational and architectural features that create unique energy consumption patterns, setting them apart from conventional building typologies such as offices, schools, or residential buildings. These differences necessitate the development of tailored energy modeling and optimization strategies to address the specific demands of religious facilities [59].

One of the most critical distinguishing features of mosques is their intermittent occupancy pattern, characterized by sharp peaks during the five daily prayer times, with pronounced surges during Friday Jumu′ah prayers and Islamic holy events such as Ramadan and Eid. Unlike buildings with predictable daytime occupancy, mosques exhibit short but intense periods of activity, followed by extended periods of inactivity. This usage behavior poses significant challenges for conventional energy management systems. A recent study demonstrated that HVAC energy consumption in mosques spikes during congregational prayers, accounting for nearly one-third of the weekly load, even though the building remains idle for most of the day [60]. Additionally, the temporal variability of operations driven by the Islamic lunar calendar further complicates energy modeling. Prayer times change daily and seasonally, affecting the schedules for lighting, heating, and cooling. Traditional energy scheduling models fail to capture these temporal dynamics, making them inadequate for mosque environments. To address this, a time-sensitive model was developed that incorporates prayer schedules and seasonal variations, leading to a 25% improvement in HVAC load forecasting accuracy [61].

From an architectural standpoint, mosques typically feature large, open prayer halls, high ceilings, domes, and minimal partitioning, which profoundly affect thermal comfort and airflow. These features contrast with conventional office or residential buildings that benefit from compartmentalized zoning and easier thermal regulation. CFD simulations conducted in Saudi mosques revealed that traditional HVAC zoning strategies are less effective due to the open architectural layout. As a result, sensor-based adaptive systems are recommended to manage indoor thermal conditions more efficiently [62]. Moreover, the geometric complexity of domed structures requires specialized thermal analysis, as these architectural elements can create unique heat transfer patterns that differ significantly from conventional flat-roof buildings [51]. Cultural and ritual practices shape mosques’ unique energy use. Ablution facilities require significant water and heating energy, sometimes accounting for up to 20% of total consumption in colder seasons. Additionally, the need for silence and comfort during prayers influences ventilation and acoustic controls [52]. Comparing mosques to other building types highlights their unique energy consumption patterns. While offices and schools have regular occupancy and predictable energy peaks during work or class hours, mosques face brief yet intense peak demands that necessitate rapid activation of HVAC and lighting systems, often at irregular times [53]. Additionally, the open architectural design and lack of internal partitions limit the effectiveness of standard HVAC zoning strategies used in other buildings.

Recent studies increasingly focused on the unique energy characteristics of mosques by applying machine learning (ML) and artificial intelligence (AI) techniques. Models such as LSTM and XGBoost have been used to forecast energy demand, integrating factors such as daily prayer schedules, occupancy patterns, and weather conditions. These AI-driven approaches demonstrated significantly better performance than linear regression methods, especially in predicting peak loads during Jumu′ah prayers and Ramadan evenings [54]. This data-driven approach supports energy efficiency goals outlined in Sustainable Development Goal 7 (Affordable and Clean Energy) and SDG 11 (Sustainable Cities and Communities).

Several region-specific case studies highlighted the effectiveness of tailored energy efficiency strategies for mosques. In Saudi Arabia, building envelope retrofits, such as low-E glazing and roof insulation, has been identified as a key measure to reduce solar heat gain in arid climates [55]. Similarly, studies in Malaysia have shown that passive cooling and night purge ventilation strategies, adapted to the open-plan design of mosques, can reduce cooling energy consumption by up to 46% [63].

Energy efficiency strategies such as adaptive HVAC control, daylight optimization, and smart ablution systems offer effective ways to enhance sustainability in mosques without disrupting comfort or religious functions. Adaptive HVAC adjusts to real-time occupancy, minimizing waste during low use. Natural lighting via domes and clerestory windows reduces artificial light needs, and smart ablution systems with sensors limit water and energy waste.

Overall, mosques require energy solutions that are context-specific, culturally respectful, and tech-driven. AI and machine learning provide promising tools for improving energy forecasting and management, tailored to the unique spatial and temporal dynamics of mosque use. Ongoing interdisciplinary research will be vital to support energy sustainability in mosques, which is in line with the UN Sustainable Development Goals.

3.3. Engineering and Environmental Considerations in Mosque Design

The design of a mosque is a complex endeavor that extends far beyond its architectural aesthetics. It encompasses a wide range of engineering and environmental considerations that are vital for ensuring both the functionality and sustainability of these sacred spaces. Two key aspects in this regard are acoustic engineering and environmental adaptation.

(a)

Acoustic Engineering in Mosque Design

Sound propagation: Sound within the prayer hall is a central aspect of mosque design. From the melodic call to prayer to the Imam′s sermon, advanced acoustical design is applied to guarantee that the congregation can hear clearly. This involves architectural elements such as domes, arches, and mihrabs, which contribute to the auditory experience within the mosque.

-

Scientific precision: The study of mosque architecture incorporates structural engineering and the selection of construction materials. Architects and engineers must consider geological conditions and local resources to ensure structural stability and longevity [64]. In regions prone to seismic activity, historical mosques incorporate engineering elements such as timber supports to withstand earthquakes. Material science advancements also facilitate the preservation of historical structures through restoration techniques and materials that respect the original design [65].

-

Sound reflection and absorption: Careful consideration is given to sound reflection and absorption in mosque interiors. Materials with varying acoustic properties are chosen to create an ideal acoustic environment [66,67]. Soft materials such as carpets and textiles are employed to absorb sound, while hard surfaces such as domes and walls are designed to reflect and amplify it. Advanced computer modeling and simulations are utilized in the design phase to optimize sound propagation within the mosque [68]. This guarantees superior sound quality for congregants.

(b)

Environmental Adaptation in Mosque Design

Contemporary mosque design has undergone a significant transformation, placing an emphasis on sustainability, climate responsiveness, and energy efficiency. Architects are now creating sanctuaries that breathe with the environment, responding to local climate conditions with an eco-conscious mindset. This includes the use of renewable materials and resource-efficient construction processes to minimize environmental impact [69]. The importance of sustainability in mosque design is reflected in the use of sustainable materials, such as reclaimed stone and sustainable wood, as well as the incorporation of renewable energy sources such as solar panels [70]. Notable examples of sustainable mosque design include the Great Mosque of Massy in France, which features double-wall insulation, an underfloor heating system, and LED lights, and the Badriya Jum’a Masjid in India, which is the world′s first zero-energy eco-friendly green mosque. One prominent feature of modern mosque design is the seamless integration of technology. Advanced building materials, such as smart glass for efficient lighting and energy-efficient HVAC systems, are employed to reduce energy consumption [71]. Furthermore, the use of modern construction techniques ensures structural longevity and, consequently, a reduced need for reconstruction, which is both cost-effective and ecologically sound [51]. The significance of community engagement is another pivotal aspect. Sustainable mosques serve as centers for community gatherings, education, and social services, contributing to the social and economic well-being of their neighborhoods [72]. These mosques often provide vocational training and serve as platforms for environmental awareness campaigns, extending their influence beyond their prayer halls [70]. For example, The Masjid al-Haram expansion project in Mecca is a prime example of the holistic approach to mosque design, which incorporates sustainable design principles and advanced technology (Figure 4). The project aims to accommodate a growing number of worshippers while minimizing the environmental footprint. The expansion project encompasses advanced energy-efficient lighting, cooling systems, and renewable energy sources. The use of modern construction techniques ensures structural longevity and reduces the need for reconstruction, which is both cost-effective and ecologically sound [73]. The enlargement project provides a one-of-a-kind opportunity for the long-term development of an iconic edifice of outstanding magnitude. The project expands radially outwards for 684 m, giving more praying places and crucial amenities to millions of pilgrims who visit Makkah periodically for Hajj and Umrah. The rebuilt Haram is a massive feat of art and architecture, with a large volume, huge height, and tall, evenly placed columns that emphasize its elegance and create an obvious line of vision [74]. The Masjid al-Haram expansion project is an excellent example of how sustainable mosque design can contribute to the social and economic well-being of their neighborhoods, providing vocational training and serving as platforms for environmental awareness campaigns.

In light of the global climate crisis and the growing interest in sustainable practices, contemporary mosque design not only meets the spiritual and physical needs of the community, but also respects the ecological context. These architectural marvels, embracing climate responsiveness, sustainability, and energy efficiency, redefined the role of religious spaces in modern society. They have become symbols of faith, sustainability, and community engagement, promising enduring sanctuaries for generations to come.

4. AI Models for Energy Management in Architectural Building

With the increase in demand for energy consumption, forecasting is becoming a necessity to meet future power demand and match generation. Confronted with various factors, electricity production and demand are dynamically changing. Thus, to better predict this balance, electricity suppliers use historical power and weather pieces of data related to temperature, humidity, and radiations [75], although there are many uncertainties about the management of energy demand in buildings [46]. The behavior of users and appliances ultimately determines the real level of power and energy demand to be met [76]. However, the capability for keeping energy has boundaries [77]. Various tactics and methodologies have been implemented in mosques to handle uncertainty in energy usage and management. These methodologies address distinct timeframes for short-term, medium-term, and long-term forecasting requirements. Some of the techniques include upgrading the mosque′s envelope to conserve energy, installing solar panels to lower energy expenses, optimizing HVAC system functioning, and employing data-driven models for load forecasting [77,78,79]. Furthermore, a study developed a unique energy control strategy for a microgrid powered by renewable energy sources, considering uncertainty and demand response [80]. These techniques seek to improve energy efficiency and encourage the use of RES in mosques.

However, several approaches and methods have been developed depending on the short-, medium-, and long-term forecast needs [81]. Hence, before presenting the energy demand forecasting models, it is stressed that forecasting can be divided into four categories depending on the time horizon [82].

-

Very short-term load forecasting: analyze with a time horizon of less than one hour (e.g., every 15 min).

-

Short-term load forecasting, which is a few hours to a day in advance. This horizon is necessary to anticipate energy-intensive loads during peak hours and to manage local daily demand.

-

Medium-term load forecasting, which is usually a few weeks. This horizon is used to plan maintenance and equipment in the power generation industry, as well as to consider expected sales and budgets to be spent.

-

Long-term load forecasting, which is more than one year. This type of predictions is applied by the electricity industries to manage energy reserves planning to increase their capacity to generate, transmit, and distribute electrical energy.

In general, the objective of forecasting models is to simulate the temporal evolution of the variable to be predicted over a given horizon. For this, there are two families of models; the first one uses only the past observations of the variable to be explain and the second is to use all the explanatory variables.

-

Autoregressive or univariate autoregressive (AR) models: Autoregressive models exploit the temporal dependencies between the data of a time series without studying the dependencies with the explanatory variables [74].

-

Models with exogenous variables autoregressive with eternal inputs (ARX): for this case, forecasts of exogenous variables provided by meteorological services are used instead of local measurements [83]. Thus, the availability of future exogenous data is fundamental. Given the emergence of IT systems, it is becoming increasingly easy to access high-precision weather forecasts.

Indeed, the forecasting models must consider the fact that the exogenic parameters are uncertain, and consumers′ behavior is often indeterminate. In view of the multitude of methodologies developed and tested over the last decades, this is particularly true. It is interesting to provide a classification to determine the degree of precision and adaptation of each method [84]. This categorization will not be perfect since the boundary between two levels of modeling is generally not well marked [76]. The most commonly used forecasting methods can be classified into the following: (i) physical white-box models; (ii) black-box models; and (iii) grey-box models [77]. Table 1 summarizes published building energy prediction research for white-, grey-, and black-box models.

5. Physical White-Box Models (WBPs)

WBP modeling is based on physics laws and describes the system explicitly. They are derived from the phenomena that govern the process with the help of theoretical knowledge. Therefore, they have a predictive and explanatory value [78]. These models are generally based on a zonal discretization of the volume of air and a resolution of the energy balance equations within the zones and at the surface of the building walls [79].

So far, WBPs are the most widely used for two main advantages. First is for their flexibility, i.e., the analysis can be manipulated under several scenarios without any need to re-modeling [80]. For their robustness, they describe the household appliances in detail using mathematical equations; therefore, more accurate results are generated [85]. In addition, they allow a complete sensitivity analysis and a good understanding of the building physics phenomena [86]. However, they are the most complex to carry out, as they contain very few approximations and require approximations and many parameters. Therefore, they require long computation times and powerful computers.

The physical models have been implemented in several building simulation software packages. The most commonly used simulation tools are Energy Plus, ESP-R, and TRNSYS [87]. These software tools are based on the physical properties, design, and other data of buildings to simulate the energy consumption of buildings on different time scales (year, month, day, or hour) [88]. In addition, the simulation can be widely performed for different locations in the world with weather files of the selected cities [89]. These models are applied to predict both the thermal and electrical needs [90]. They are considered to be the only methods capable of modeling the power demand without the need for historical energy demand. However, their effectiveness is limited due to the enormous time and cost required for the user to obtain and define all the parameters and characteristics of the building [91].

5.1. Grey-Box Models (GbXM)

The GbXM, or semi-physical theories, have an intermediate approach that combines physical modeling to replicate the building′s thermal behavior with black-box models. The grey-box models are physically intelligible and correspond to measurements [92]. They were introduced in the early 1990s for specific applications in automatic system control and optimization studies [93]. This category of models consists of modeling the building by an equivalent electrical circuit. In this case, these models involve knowledge of thermal parameters, such as the capacity and thermal resistance of different orders [94]. These grey-box models have been applied in the implementation of control systems, forecasting to improve the electrical demand in the building using energy-saving strategies [93]. Indeed, they can be applied in the estimation of electrical demand due to the heating and cooling system as well as in the evaluation of the indoor temperature. Grey-box models can be considered in cases where a physical model of the building is available, such as of the building, especially when modeling existing buildings, where it is usually difficult to reconstruct a detailed physical model, in which situation, grey-box models can be applied [95]. They also allow us to describe the thermal parameters and to keep a physical meaning of the parameters of the building model. In fact, they require fewer computational resources than white-box models, as they are more flexible [96]. However, some disadvantages are specific to these models, such as the need for a double competence for the user in building thermals and the knowledge of black-box models.

5.2. Black-Box Models (BbXM)

BbXM are in fact called data-driven models. These models do not consider the internal structure [97]. They simply allow a numerical resolution of the problem without providing a physical interpretation. The resolution method is based on empirical relationships that link the input and output parameters [98]. These relationships are the products of a regression analysis that requires experimental measurements. In addition, an amazingly simple model is obtained, and accurate results are obtained in a calculation time. Their precision depends on the quantity and quality of the available data. They can also handle non-linear variations in energy consumption by using adaptive learning to new pieces of data [99].

BbXM models have shown their limits in terms of the need for experimental measurements, which require significant resources [100]. Moreover, the optimization study is not based on a physical understanding of the phenomena. It is therefore only practical in the case of a prediction for a possible pre-dimensioning. The disadvantage of this type of model is its lack of flexibility; the algorithm used requires input data and returns a set of output variables [101]. Even if black-box models do not give any physical interpretation of the results of the studied building, they are very present in the modeling and forecasting of energy [92,102]. They make use of statistical learning, which represents a field of study in artificial intelligence.

5.3. Black-Box for Enhanced Energy Management in Mosquees

To further optimize energy usage and promote sustainability in mosques, ML models have been incorporated into mosque energy management. These models include time series analysis, long short-term memory (LSTM) networks, Gaussian process regression (GPR), XGBoost, and recurrent neural networks (RNNs) [103,104]. These models offer a data-driven approach to capture trends and patterns, leading to more precise load forecasting and efficient resource management. The choice of the appropriate model depends on the specific energy management task at hand and the available data.

Recent studies have shown that mosques consume an excessive amount of energy, particularly due to lighting and HVAC systems. However, various studies proposed design strategies to improve energy efficiency and promote sustainability in mosques [105,106]. These strategies include the use of sensor technology for HVAC activation, operational thermal zoning, and the use of energy-efficient ECMs and WCMs.

Combining energy optimization with ML and smart water management can benefit from the usage of machine learning methods. ML may be used to increase building energy efficiency by analyzing data from several sources and refining an internal model which might be explored for patterns and anomalies. It can be employed as well for energy consumption forecasting, which analyzes a building′s past energy consumption data to identify trends and anticipate future energy usage. ML may also be used in smart water management to identify leaks, anticipate consumption trends, and fine-tune water-saving devices in toilets and ablution areas [107]. It may also be utilized to optimize available information and data in order to make better decisions, improve service delivery, and save costs [108]. In conclusion, adopting diverse methods and techniques, as well as ML models, can improve energy efficiency, promote sustainability, and minimize energy consumption in mosques. The advancement of machine learning approaches in this sector is ongoing, with more refined and complex models promising to improve energy efficiency and sustainability in mosque structures (Table 2).

Table 1. Survey review on building energy prediction models.

Ref.	Activity Sector	Methods and Tools	Inputs	Predictions Outputs	Conclusion
	WBM
[109]	Residential	TRNSYS	Data from 104 homes and the climate	T hermal p erformance of the s olar DHW systems	In most cases, larger buildings with more than forty-eight apartments (70–90%) are feasible.
[110]	Residential	TRNSYS	Weather conditions and building geometry	Heating and cooling demand	The predicted mean vote (PMV) varies between −0.7 and 0.7; and the predicted percentage people dissatisfied is around 15%.
[111]	Commercial and residential	Energy Plus	Indoor temperature, air flow rate, load data, and weather conditions	Energy consumption of HVAC systems	The on/off regulator enhances HVAC precision by approximately 19% in terms of NMBE.
[112]	Residential district	Energy Plus, IDA ICE, TRNSYS	Weather data and building geometry	Heat performance	For urban-scale energy analysis, Energy Plus is more advantageous.
[113]	All types	Energy Plus, IDA ICE, TRNSYS	Outdoor environment, geometric and technical data	Building thermal behavior	Without phase change materials, all tools are exactly accurate. IDA ICE is recommended in the presence of PCM.
[114]	Residential	IES VE	building typology, thermal setpoint, and building properties	Energy consumption	Implementation of heat recovery ventilation reduces space heating demand by 50 to 65%.
	BbGM
[115]	residential buildings	Bayesian method and RC model	House characteristics, shading, and weather data	Energy consumption	With only 5–10 days of data, the models that were chosen produced Heat Transfer Coefficient estimates that were around 15%, notably during summertime.
[116]	residential buildings	RC model (3R2C)	Thermal properties, building data	Thermal performance	Deterministic models are more accurate than stochastic myorelaxants. Causal shift has a beneficial effect on the performance of the stochastic grey-box model.
[117]	smart buildings	Multi-zone RC model	Indoor conditions, weather data	Thermal energy consumption	Phase change materials and solar technologies can improve flexibility of building and proper control strategy
[118]	school building	Nonlinear model	Heat load, indoor climate, and weather data	Thermal comfort	The solar irradiation gain is an important disturbing factor for building climate control.
	BbXM
[119]	commercial	MLR models	Building proprieties	CO2 emissions	The RE in estimating reductions in emissions is around −13.0%, whereas the error in assessing decrease rate is 0.5%.
[120]	Healthcare	MLR	T, RH, Vwind and direction, Gh, and floor surface	Electricity and PV energy generation	MLRs are excellent for forecasting future weather parameters, energy consumption, and PV generation.
[121]	Hotel	SVM with Radial basis kernel	Weather parameters and operating parameters of air-conditioning system	Energy consumption	The RBF kernel function is most suitable for the support vector machine. The MSE value of the final model prediction was 2.22% and R2 was 0.94.
[122]	Public building	SVM with Gaussian kernel	Characteristics relating to ancient energy data, and weather and time-cycle factors.	Energy consomption	The SVM approach is appropriate for limited sample sizes and complex issue forecasts. In the case of daily building energy consumption predictions, the average RE has a prediction impact of 5.03%.
[123]	Office building	SVM, RBF, sigmoid, DSS, DFS)	Climate conditions, building characteristics	Cooling load prediction	SVR with distorted sample space (DSS) is more accurate with R=0.92 and a low RAE of 37.44%.
[124]	Office building	ANN	Energy and environmental data	Heating energy demand	The best artificial neural networks had low RMSE, MAE of less than 5%, and a determination coefficient of about one.
[125]	Office building	ANN and GA	Wall U-values, machinery load level, lighting density, penetration rate, and people count	Forecasting and adjustment of energy use	The optimization approach reduces energy use and costs by 35%.
[126]	Public building	Random forest	Relevant with building creation, he ating, co oling, and employment characteristics	Energy cost prediction	Random forest-Boruta with 450 trees on sixteen variables explained a lower SMAPE of 22.35% on the test sample.
[127]	Office building	XGBoost	Weather data and building parameters	Heat performance	XGBoost algorithm showed the highest F-1 score (0.75)
[128]	Educational	RNN	Climate, occupancy patterns, and working schedules	Building energy consumption	The suggested adaptable LSTM NN outperforms the GA-optimized DNN model by reducing MAE from 23.87% to 21.74% and increasing R2 from 3.39% to 2.45%.

Table 1. Survey review on building energy prediction models.

Ref.	Activity Sector	Methods and Tools	Inputs	Predictions Outputs	Conclusion
	WBM
[109]	Residential	TRNSYS	Data from 104 homes and the climate	T hermal p erformance of the s olar DHW systems	In most cases, larger buildings with more than forty-eight apartments (70–90%) are feasible.
[110]	Residential	TRNSYS	Weather conditions and building geometry	Heating and cooling demand	The predicted mean vote (PMV) varies between −0.7 and 0.7; and the predicted percentage people dissatisfied is around 15%.
[111]	Commercial and residential	Energy Plus	Indoor temperature, air flow rate, load data, and weather conditions	Energy consumption of HVAC systems	The on/off regulator enhances HVAC precision by approximately 19% in terms of NMBE.
[112]	Residential district	Energy Plus, IDA ICE, TRNSYS	Weather data and building geometry	Heat performance	For urban-scale energy analysis, Energy Plus is more advantageous.
[113]	All types	Energy Plus, IDA ICE, TRNSYS	Outdoor environment, geometric and technical data	Building thermal behavior	Without phase change materials, all tools are exactly accurate. IDA ICE is recommended in the presence of PCM.
[114]	Residential	IES VE	building typology, thermal setpoint, and building properties	Energy consumption	Implementation of heat recovery ventilation reduces space heating demand by 50 to 65%.
	BbGM
[115]	residential buildings	Bayesian method and RC model	House characteristics, shading, and weather data	Energy consumption	With only 5–10 days of data, the models that were chosen produced Heat Transfer Coefficient estimates that were around 15%, notably during summertime.
[116]	residential buildings	RC model (3R2C)	Thermal properties, building data	Thermal performance	Deterministic models are more accurate than stochastic myorelaxants. Causal shift has a beneficial effect on the performance of the stochastic grey-box model.
[117]	smart buildings	Multi-zone RC model	Indoor conditions, weather data	Thermal energy consumption	Phase change materials and solar technologies can improve flexibility of building and proper control strategy
[118]	school building	Nonlinear model	Heat load, indoor climate, and weather data	Thermal comfort	The solar irradiation gain is an important disturbing factor for building climate control.
	BbXM
[119]	commercial	MLR models	Building proprieties	CO2 emissions	The RE in estimating reductions in emissions is around −13.0%, whereas the error in assessing decrease rate is 0.5%.
[120]	Healthcare	MLR	T, RH, Vwind and direction, Gh, and floor surface	Electricity and PV energy generation	MLRs are excellent for forecasting future weather parameters, energy consumption, and PV generation.
[121]	Hotel	SVM with Radial basis kernel	Weather parameters and operating parameters of air-conditioning system	Energy consumption	The RBF kernel function is most suitable for the support vector machine. The MSE value of the final model prediction was 2.22% and R2 was 0.94.
[122]	Public building	SVM with Gaussian kernel	Characteristics relating to ancient energy data, and weather and time-cycle factors.	Energy consomption	The SVM approach is appropriate for limited sample sizes and complex issue forecasts. In the case of daily building energy consumption predictions, the average RE has a prediction impact of 5.03%.
[123]	Office building	SVM, RBF, sigmoid, DSS, DFS)	Climate conditions, building characteristics	Cooling load prediction	SVR with distorted sample space (DSS) is more accurate with R=0.92 and a low RAE of 37.44%.
[124]	Office building	ANN	Energy and environmental data	Heating energy demand	The best artificial neural networks had low RMSE, MAE of less than 5%, and a determination coefficient of about one.
[125]	Office building	ANN and GA	Wall U-values, machinery load level, lighting density, penetration rate, and people count	Forecasting and adjustment of energy use	The optimization approach reduces energy use and costs by 35%.
[126]	Public building	Random forest	Relevant with building creation, he ating, co oling, and employment characteristics	Energy cost prediction	Random forest-Boruta with 450 trees on sixteen variables explained a lower SMAPE of 22.35% on the test sample.
[127]	Office building	XGBoost	Weather data and building parameters	Heat performance	XGBoost algorithm showed the highest F-1 score (0.75)
[128]	Educational	RNN	Climate, occupancy patterns, and working schedules	Building energy consumption	The suggested adaptable LSTM NN outperforms the GA-optimized DNN model by reducing MAE from 23.87% to 21.74% and increasing R2 from 3.39% to 2.45%.

Table 2. Survey review on mosque smart energy management.

Ref.	Case Study	Area (m2)	Ceiling (m)	Methodology	Simulation Tool	Conclusion
					Thermal Comfort
[129]	Malaysia	75 × 180	5.5	Simulations and field experiment	CFD software	suitable Qiblah wall design can reduce indoor wall temperatures by up to 4 degrees Celsius, improving thermal comfort by 40% during daytime and up to 80% during night-time prayers
[130]	Turkey	300	-	Simulations field experiment	Design Builder-EnergyPlus	The study showed a significant reduction in discomfort hours during the winter period, with the lowest discomfort in October 2014 (meeting the 80% comfort limit) and the highest in January 2015 (446 h), marking a substantial 54.74% reduction compared to the baseline model.
[131]	Penang, Malaysia	640	6	Simulations	Predicted mean vote model	Predicted mean vote (PMV) model calculations often underestimated comfort temperatures, with worshippers largely experiencing neutral conditions. The 4.6 °C difference between PMV (25.9 °C) and actual mean vote (AMV) (30.4 °C) suggests that adherence to ASHRAE Standard-55 could have significant energy implications for the organization.
[106]	Penang, Malaysia	506	5.5	simulations	Visual DOE building energy simulation program	Effective HVAC strategies and op erational zoning in mosques can decrease annual cooling energy consumption by up to 23%, and combining insulation with appropriately sized HVAC systems for intermittent operation during prayers results in a significant 46% reduction, with the potential for an additional 30% energy savings.
					Enery performance
[132]	Saudi Arabia	-	-	ECMWF data, with the decision tree	DT, SVM, KNN, ANN	With the decision tree model as the top performer (8.916% nRMSE and 6.038% nMAE). The model’s deployability and potential for substantial energy savings are noteworthy, as it can predict future loads independently of past consumption data.
[133]	Morocco	505	7.5	Simulation	TRNSYS	The solar radiation gain is an important disturbing factor for building climate control.
[134]	Iran	45	5.4	Simulation	CFD (OpenFOAM)	CFD simulations show Emad-o-Dowlah Mosque’s windcatcher performance outperforms traditional ventilation, improving comfort by 101% and demonstrating its broader climatic applicability.
[59]	Turkey	-	-	Simulation	DesignBuilder- EnergyPlus	Mosques should prioritize energy efficiency by enhancing the building envelope to mitigate outdoor climate effects, with the study showing a 33% energy savings through insulation improvements and glazing considerations.
					Air distribution
[135]	Irak	400	6.5	Numerical calculation	Ansys-Fluent	CFD simulations affirm the windcatcher’s effectiveness at Emad-o-Dowlah Mosque, exceeding natural ventilation by 101% and aligning with modern windcatchers, indicating its broader applicability across climatic conditions when equipped with nighttime shutdown technology.
[136]	Egypt	266	16	Numerical simulations	ANN, MATLAB	The study highlights the need for architectural modifications in the mosque, emphasizes the value of the ANN model for precise predictions, and recommends reopening upper windows and air traps to enhance thermal comfort through improved ventilation and air circulation.
[137]	Malysia	1050	12	Numerical simulations	CFD, PMV	Using field measurements and CFD simulations, revealing that PMV and PPD values inside the main prayer hall currently exceed comfort standards, but installing four 56-cm diameter exhaust fans on the west-side wall can enhance thermal comfort significantly, reducing PMV and PPD values by over 60%.
[138]	Indonisia	-	-	Numerical simulations	MIT CoolVent, Energy plus, CFD	The air flow network model properly forecasts flow rates and temperatures with RSME values of 0.23 m3/s and 1.26 °C, respectively, whereas the CFD model effectively replicates velocity vectors and interior air temperatures with an RSME of 1.2 °C. These methods are used to assess a night purge system for a thermally large mosque, which results in a 74% decrease in internal temperatures over 32 °C and a 53% drop in degrees between 29 and 32 °C. The inclusion of low-energy exhaust ventilators further helps by lowering the highest temperatures by 59%, allowing for warmth conditioning as necessary.

Table 2. Survey review on mosque smart energy management.

Ref.	Case Study	Area (m2)	Ceiling (m)	Methodology	Simulation Tool	Conclusion
					Thermal Comfort
[129]	Malaysia	75 × 180	5.5	Simulations and field experiment	CFD software	suitable Qiblah wall design can reduce indoor wall temperatures by up to 4 degrees Celsius, improving thermal comfort by 40% during daytime and up to 80% during night-time prayers
[130]	Turkey	300	-	Simulations field experiment	Design Builder-EnergyPlus	The study showed a significant reduction in discomfort hours during the winter period, with the lowest discomfort in October 2014 (meeting the 80% comfort limit) and the highest in January 2015 (446 h), marking a substantial 54.74% reduction compared to the baseline model.
[131]	Penang, Malaysia	640	6	Simulations	Predicted mean vote model	Predicted mean vote (PMV) model calculations often underestimated comfort temperatures, with worshippers largely experiencing neutral conditions. The 4.6 °C difference between PMV (25.9 °C) and actual mean vote (AMV) (30.4 °C) suggests that adherence to ASHRAE Standard-55 could have significant energy implications for the organization.
[106]	Penang, Malaysia	506	5.5	simulations	Visual DOE building energy simulation program	Effective HVAC strategies and op erational zoning in mosques can decrease annual cooling energy consumption by up to 23%, and combining insulation with appropriately sized HVAC systems for intermittent operation during prayers results in a significant 46% reduction, with the potential for an additional 30% energy savings.
					Enery performance
[132]	Saudi Arabia	-	-	ECMWF data, with the decision tree	DT, SVM, KNN, ANN	With the decision tree model as the top performer (8.916% nRMSE and 6.038% nMAE). The model’s deployability and potential for substantial energy savings are noteworthy, as it can predict future loads independently of past consumption data.
[133]	Morocco	505	7.5	Simulation	TRNSYS	The solar radiation gain is an important disturbing factor for building climate control.
[134]	Iran	45	5.4	Simulation	CFD (OpenFOAM)	CFD simulations show Emad-o-Dowlah Mosque’s windcatcher performance outperforms traditional ventilation, improving comfort by 101% and demonstrating its broader climatic applicability.
[59]	Turkey	-	-	Simulation	DesignBuilder- EnergyPlus	Mosques should prioritize energy efficiency by enhancing the building envelope to mitigate outdoor climate effects, with the study showing a 33% energy savings through insulation improvements and glazing considerations.
					Air distribution
[135]	Irak	400	6.5	Numerical calculation	Ansys-Fluent	CFD simulations affirm the windcatcher’s effectiveness at Emad-o-Dowlah Mosque, exceeding natural ventilation by 101% and aligning with modern windcatchers, indicating its broader applicability across climatic conditions when equipped with nighttime shutdown technology.
[136]	Egypt	266	16	Numerical simulations	ANN, MATLAB	The study highlights the need for architectural modifications in the mosque, emphasizes the value of the ANN model for precise predictions, and recommends reopening upper windows and air traps to enhance thermal comfort through improved ventilation and air circulation.
[137]	Malysia	1050	12	Numerical simulations	CFD, PMV	Using field measurements and CFD simulations, revealing that PMV and PPD values inside the main prayer hall currently exceed comfort standards, but installing four 56-cm diameter exhaust fans on the west-side wall can enhance thermal comfort significantly, reducing PMV and PPD values by over 60%.
[138]	Indonisia	-	-	Numerical simulations	MIT CoolVent, Energy plus, CFD	The air flow network model properly forecasts flow rates and temperatures with RSME values of 0.23 m3/s and 1.26 °C, respectively, whereas the CFD model effectively replicates velocity vectors and interior air temperatures with an RSME of 1.2 °C. These methods are used to assess a night purge system for a thermally large mosque, which results in a 74% decrease in internal temperatures over 32 °C and a 53% drop in degrees between 29 and 32 °C. The inclusion of low-energy exhaust ventilators further helps by lowering the highest temperatures by 59%, allowing for warmth conditioning as necessary.

6. Comprehensive SWOT Analysis for Sustainable Development

The use of AI in mosque energy management has a lot of potential for meeting the SDGs. In the current work, a comprehensive strengths, weaknesses, opportunities, and threats (SWOT) analysis of incorporating AI for optimizing energy usage in mosques, and delves into how these AI-driven solutions align with the United Nations′ Sustainable Development Goals. Thus, energy management in religious buildings, particularly mosques, is an important aspect of sustainability. Mosques, as communal places of worship, often consume significant amounts of energy due to the need for adequate lighting, heating, and cooling, among other requirements. The introduction of AI into the energy management sphere of these buildings offers novel opportunities and challenges, with implications for achieving several SDGs.

(a)

Strengths

-

Energy efficiency: Alignment with SDG 7 (cheap and Clean Energy): Improving energy efficiency in mosques is inextricably related to SDG 7, which emphasizes cheap and clean energy for everybody. AI-driven EMS may drastically cut energy use and operational expenses, hence increasing access to inexpensive and clean energy.

-

Environmental impact: Alignment with SDG 13 (Climate Action): Lowering energy use reduces carbon footprint, which aligns with SDG 13′s goal of battling climate change and its effects.

-

Smart automation: Alignment with SDG 9 (Industry, Innovation, and Infrastructure): Incorporating AI reflects advancements in industry and innovation, as it promotes infrastructure development that is both intelligent and efficient, thereby supporting SDG 9.

-

Cost savings: Alignment with SDG 1 (No Poverty): The cost savings achieved through efficient energy management can be redirected to initiatives that alleviate poverty and improve the well-being of underprivileged members of the community.

-

Predictive maintenance: Alignment with SDG 11 (Sustainable Cities and Communities): Ensuring that mosques have dependable access to heating, lighting, and cooling enhances the overall sustainability and resilience of urban communities, aligning with SDG 11.

(b)

Weaknesses

-

Implementation costs: Challenge to SDG 10 (Reduced Inequalities). High initial setup costs can be prohibitive and may exacerbate inequalities in accessing the benefits of AI-driven energy management.

-

Technical expertise: Challenge to SDG 4 (Quality Education). The need for specialized technical expertise presents a barrier to achieving SDG 4, which promotes inclusive and equitable quality education.

-

Data privacy and security: Threat to Sustainable Development Goal 16 (Peace, Justice, and Strong Institutions). Ensuring data privacy and security can be difficult since it necessitates solid governance mechanisms that promote peace, justice, and strong institutions, as specified in SDG 16.

-

Maintenance: Challenge to SDG 8 (Decent Work and Economic Growth). Regular maintenance and training may strain the workforce and affect economic growth, thereby posing challenges to SDG 8.

-

Dependency on technology: Challenge to SDG 3 (Good Health and Well-being). Overreliance on AI systems, especially during worship, may impact the physical and mental well-being of congregants if technology fails.

(c)

Opportunities

-

Sustainability initiatives: Alignment with SDG 12 (Responsible Consumption and Production). Mosques can position themselves as responsible consumers of energy, aligning with SDG 12′s objectives.

-

Community engagement: Alignment with SDG 17 (Partnerships for the Goals). Promoting energy efficiency through AI can foster community engagement and partnerships, as outlined in SDG 17.

-

Energy resilience: Alignment with SDG 6 (Clean Water and Sanitation). Reliable energy sources contribute to ensuring access to clean water, which is essential for meeting the objectives of SDG 6.

-

Research and development: Alignment with SDG 9 (Industry, Innovation, and Infrastructure). The evolving field of AI for energy management presents opportunities for research and development in line with SDG 9.

-

Commercial partnerships: Alignment with SDG 17 (Partnerships for the Goals). Mosques can explore partnerships with AI companies and energy providers, aligning with the importance of partnerships in achieving SDGs.

(d)

Threats

-

Resistance to change: Challenge to SDG 4 (Quality Education). Resistance to AI adoption may hinder opportunities for improved quality education (SDG 4) in utilizing advanced technology.

-

Regulatory compliance: Challenge to SDG 16 (Peace, Justice, and Strong Institutions). Meeting evolving regulatory standards may pose challenges to the development of peace, justice, and strong institutions (SDG 16).

-

High Initial costs: Challenge to SDG 10 (Reduced Inequalities). Initial setup costs may increase disparities in energy management and energy access, thus challenging SDG 10.

-

Technological obsolescence: Challenge to SDG 9 (Industry, Innovation, and Infrastructure). The rapid advancement of AI could pose challenges in terms of technological obsolescence, affecting the progress of SDG 9.

-

Competition for resources: Challenge to SDG 1 (No Poverty). Limited funds and resources might compete with energy management initiatives, potentially affecting the achievement of SDG 1.

Incorporating AI in mosque energy management provides a robust platform for advancing sustainability. The SWOT analysis demonstrates the complex interplay of strengths, weaknesses, opportunities, and threats associated with AI-driven energy management in mosques. By aligning these AI-driven solutions with SDGs, we can ensure that the adoption of advanced technology contributes to the broader global agenda of achieving sustainable development. The success of these initiatives hinges on effective planning, community engagement, and leveraging partnerships to navigate potential challenges and capitalize on opportunities (Figure 5).

7. Mosques as Catalysts for Sustainable and Intelligent Transitions

Mosques, as pivotal religious and socio-cultural institutions for the global Muslim population, are increasingly becoming platforms for advancing sustainable development through the integration of renewable energy systems, eco-friendly building materials, and artificial intelligence (AI)-based energy management solutions. These developments are not only improving energy efficiency and reducing carbon footprints, but also reinforcing the global shift toward intelligent and environmentally responsive infrastructure. Such transformation aligns with multiple United Nations Sustainable Development Goals (SDGs), particularly SDG 7 (Affordable and Clean Energy), SDG 11 (Sustainable Cities and Communities), SDG 12 (Responsible Consumption and Production), and SDG 13 (Climate Action) [107,139]. As highly visible communal landmarks, mosques have the potential to serve as catalysts for public awareness and behavior change regarding energy consumption and climate responsibility. However, the successful implementation of AI-driven systems in mosques requires a nuanced understanding of their unique operational characteristics, especially their intermittent energy demand linked to daily prayers, weekly gatherings, and seasonal religious events such as Ramadan. These factors introduce temporal complexity and irregular load patterns that challenge conventional energy forecasting methods. AI techniques offer a variety of modeling approaches tailored to these complexities. Recurrent neural networks (RNNs), particularly long short-term memory (LSTM) models, are capable of capturing sequential dependencies in time series data, making them suitable for predicting fluctuations associated with structured but variable prayer schedules [108]. Artificial neural networks (ANNs) are also widely applied due to their ability to learn complex nonlinear relationships between environmental variables (e.g., temperature, humidity), occupancy profiles, and energy usage, though their requirement for large datasets and limited interpretability can constrain their practical use in resource-constrained settings. Support vector machines (SVMs), by contrast, provide robust performance in data-scarce scenarios and offer good generalization, though they are less effective in capturing long-term temporal dependencies common in mosque energy patterns. Decision trees and ensemble methods, such as random forests and gradient boosting, offer the advantage of interpretability and feature importance analysis, enabling stakeholders to understand key factors driving energy use. However, their performance typically declines when dealing with sequential and highly nonlinear patterns unless they are enhanced with temporal features or integrated into hybrid models [140]. Hybrid or grey-box models, which combine physics-based building simulation (e.g., thermal energy balance) with data-driven AI learning, represent a particularly promising avenue for mosque energy prediction. These models leverage domain knowledge such as building geometry, construction materials, HVAC system behavior, and religious occupancy patterns, while retaining the adaptability of machine learning techniques. As such, they offer improved prediction accuracy, reduced reliance on large datasets, and enhanced model transparency, key considerations for deployment in culturally and technically diverse mosque settings [141]. Ultimately, selecting appropriate AI models for mosque energy management must be grounded in their ability to adapt to intermittent demand, balance data requirements, and provide actionable insights. This tailored approach not only improves operational efficiency, but also reinforces the broader role of mosques as agents of ecological and intelligent transformation.

8. Limitations and Future Research Directions

While this study outlines the theoretical suitability of various AI/ML models for mosque energy forecasting, empirical validation remains limited due to a lack of dedicated, high-resolution datasets capturing mosque-specific usage patterns. Many existing models do not yet incorporate critical contextual features such as real-time occupancy detection, seasonal prayer fluctuations, or behavioral patterns rooted in religious practice.

Future research should prioritize the development of custom datasets that reflect the operational reality of mosques, including granular temporal resolution, indoor and outdoor environmental parameters, and sociocultural usage markers. Additionally, the integration of context-aware features, such as religious calendars, prayer schedules, and event timelines into AI models could substantially enhance predictive performance. Advanced methods, such as transfer learning, which leverages models trained on data-rich environments for application in data-poor mosques, and federated learning, which ensures privacy while pooling knowledge across different mosques, offer innovative solutions to current data limitations. Furthermore, model interpretability, scalability, and computational efficiency should be prioritized to facilitate broader, cost-effective adoption. Finally, successful implementation will require interdisciplinary collaboration between AI researchers, mosque administrators, urban planners, and policymakers to ensure solutions are technically sound, culturally appropriate, and socially inclusive. Addressing socio-economic barriers, such as initial investment costs and the availability of technical expertise, will also be essential for scaling AI-powered energy management across diverse mosque typologies worldwide.

9. Conclusions

The integration of artificial intelligence (AI) in mosque energy management systems presents a transformative opportunity to enhance sustainability and align with various United Nations Sustainable Development Goals (SDGs). This research has undertaken a comprehensive SWOT analysis to evaluate the strengths, weaknesses, opportunities, and threats associated with the application of AI in optimizing energy usage within mosques. The adoption of AI technologies in mosques can significantly optimize energy consumption, leading to substantial reductions in operational costs and environmental impact. By leveraging ML models, predictive maintenance, and data-driven insights, mosques can achieve higher energy efficiency and contribute to SDG 7 (Affordable and Clean Energy) and SDG 13 (Climate Action). These AI-powered advancements help mosques to better control their energy use, reduce waste, and create a more sustainable atmosphere. Despite the potential benefits, implementing AI in management of mosque energy presents significant obstacles. The initial costs for acquiring and integrating AI systems can be prohibitively high for many mosques.

Additionally, the successful deployment and maintenance of these technologies require skilled personnel, necessitating significant investment in training and education. Overcoming these barriers will require a committed, long-term approach to resource allocation and capacity building. The opportunities presented by AI in mosque energy management are considerable. AI technologies have the potential to improve operations, enhance resource allocation, and promote sustainable practices in mosques. Mosques may help to achieve SDGs 9 (Industry, Innovation, and Infrastructure), 11 (Sustainable Cities and Communities), and 12 (Responsible Consumption and Production) by addressing these possibilities. The adoption of AI can also catalyze further innovation in energy management, driving the development of new solutions and advancements in the field. However, there are significant threats that need to be managed.

The deployment of AI in mosque energy management holds substantial promise for enhancing sustainability and aligning with global SDGs. Mosques can serve as pioneers in adopting advanced technologies to reduce their carbon footprint, cut energy costs, and promote clean, affordable energy. To realize these benefits, a collaborative effort among religious leaders, technology providers, and government entities is essential. By investing in AI technologies, training skilled personnel, and addressing potential threats, mosques can significantly contribute to a more sustainable future. This study emphasizes the need for overcoming early hurdles and controlling risks in order to fully realize the promise of AI. By doing so, mosques can not only improve their own energy efficiency, but also set a powerful example for other institutions, driving broader societal progress towards sustainability and responsible resource consumption.