Dynamic Evolution of Energy Efficiency in the Building Sector: A Changepoint Detection and Text Processing-Based Bibliometric Analysis

Abstract

Energy efficiency in buildings is a vital subject within sustainable construction and climate change mitigation, yet comprehensive bibliometric analyses mapping the complete evolution of this domain remain limited. This study provides a comprehensive four-decade analysis (1981–2025) of building energy efficiency research using data from the Web of Science database, employing VOSviewer (1.6.20), Bibliometrix (4.3.0), and custom Python (3.12.3) scripts with automated terminology normalization through TF-IDF vectorization (n-grams 2–3) and cosine similarity algorithms (threshold = 0.75). Two critical methodological innovations distinguish this investigation: first, Pruned Exact Linear Time changepoint detection statistically validated 2011 as the field’s statistically validated transition point (Mann–Whitney U test, p < 0.000001, effect size = 2.48), replacing arbitrary decade-based periodization; second, computational keyword harmonization enabled precise thematic evolution mapping across inconsistent terminology. The analysis reveals marked increase in research post-2011, with median annual output increasing from 15 articles (1981–2011) to 840.5 articles (2012–2024), and China emerging as the preeminent research center with 2978 publications. Thematic evolution analysis demonstrates fundamental transformation from seven specialized research themes (i.e., behavior, heat-transfer, simulation, impact, performance, consumption, optimization) in the foundational period to dramatic consolidation into two dominant themes (i.e., performance and simulation) in the contemporary period, reflecting maturation from fragmented, component-focused investigations toward holistic, integrated frameworks. International collaboration network analysis identifies four distinct geographic clusters with China, United States, United Kingdom, and Italy serving as central hubs. These findings provide actionable intelligence for researchers, policymakers, and industry stakeholders, while the computationally enhanced framework offers a replicable methodology for bibliometric analysis in other rapidly evolving interdisciplinary domains.

1. Introduction

Climate change necessitates a fundamental transformation in energy systems and technology, influencing multiple sectors and driving the need for innovative solutions [1]. The building industry is responsible for approximately 27% of global energy usage and contributes around 17% of total carbon dioxide (CO₂) emissions [2].

The high energy consumption in the building sector is mainly due to the significant dependence on heating and cooling systems, as well as the widespread use of various electrical devices [3].

Energy efficiency plays a fundamental role in the design and operation of environmentally friendly and sustainable buildings [4]. In the context of buildings, energy efficiency can be approached from various perspectives, with two key areas drawing significant focus from architects and engineers: the creation and development of energy-efficient structures, particularly concerning new construction projects and the retrofitting and enhancement of energy performance in existing buildings [5].

Building energy efficiency strategies have evolved considerably over time, progressing from passive design approaches in early construction (i.e., thermal mass, natural ventilation, building orientation) through 20th-century active systems (i.e., heating, ventilation and air conditioning, heat pumps, advanced glazing) to contemporary integrated solutions combining smart building technologies, renewable energy systems, and advanced materials such as phase change materials and dynamic insulation [6,7,8]. This evolution reflects growing recognition that energy efficiency must address not only technical performance but also economic viability, environmental sustainability, and occupant comfort [9].

Improving energy efficiency in both new and existing buildings offers a rapid approach to mitigating the environmental, economic, social, and other challenges within the construction sector [10]. Sustainable construction policies encompass a range of approaches designed to promote eco-friendly practices and optimize resource use throughout the building process [11].

Over the past few years, energy efficiency has gained significant attention as a key strategy for mitigating carbon emissions, optimizing energy utilization, enhancing building performance, and contributing to the creation of sustainable built environments [12].

Building energy management strategies focus on optimizing energy usage through various techniques like predictive modeling, optimization algorithms, and machine learning. Common approaches include particle swarm optimization, genetic algorithms, and demand-side management to improve efficiency while ensuring comfort. Additionally, fault detection and diagnosis techniques help identify system issues in real time. These strategies, supported by advanced software, enhance building performance by reducing energy consumption and promoting sustainability [13].

The significance of technological transfer in attaining energy efficiency in buildings and the development of Zero Energy Buildings (ZEB) resides in its capacity to connect research innovations with practical applications, guaranteeing the effective implementation of advanced solutions in the construction industry. The Triple Helix concept, which fosters collaboration among academics, industry, and government, has produced significant outcomes, promoting sustainable practices and expediting the use of energy-efficient technologies in the built environment [14].

Bibliometric analysis has emerged as an increasingly important approach for mapping research landscapes in building energy efficiency, offering systematic frameworks for identifying influential publications, tracking thematic shifts, revealing collaboration patterns, and mapping emerging frontiers. However, existing bibliometric investigations demonstrate significant methodological and scope limitations that constrain comprehensive understanding of the field’s evolution. The majority of studies focus on narrow research subdomains, including zero-energy buildings [15], specific technologies such as machine learning applications [16], or specialized topics like biomimicry [17]. Temporal scope represents a critical limitation, with most investigations covering periods of 10–15 years, insufficient for capturing the complete transformation from foundational 1970s–1980s innovations through contemporary integrated approaches. For instance, studies by Wang et al. [18] and Varshabi et al. [17] examined only 2006–2018 and 2010–2021 respectively, thus not encompassing foundational research periods essential for understanding technological evolution. Previous bibliometric investigations of building energy efficiency face significant methodological limitations. Most studies suffer from one or more constraints: narrow subdomain focus (e.g., examining only zero-energy buildings or specific technologies), restricted temporal coverage (typically 10–15 years), single-database dependence, or manual keyword categorization that cannot capture semantic variants of identical concepts. Critically, no prior study has employed computational terminology normalization techniques such as TF-IDF vectorization and cosine similarity to address the field’s fundamental challenge: evolving nomenclature where identical concepts appear as multiple terminological variants (e.g., “net zero energy building,” “zero energy building,” “nearly zero energy building”). This computational approach is essential for comprehensive, long-term bibliometric analysis because building energy efficiency research has developed inconsistent terminology over four decades, making manual categorization both imprecise and non-replicable across large datasets.

To outline the methodological and conceptual distinctions of existing research,

Table 1. Specific evaluation of recent bibliometric analyses in energy efficiency field.

Study Title/First Author (Year)	Topic Focus	Time Span	DTB	Tools	Limitations (Number of Articles)	Ref.
Energy efficiency in buildings: analysis of scientific literature and identification of data analysis techniques from a bibliometric study/Cristino et al. (2018)	Energy efficiency in buildings	1980–2016	Scopus	Excel, VoSviewer and Minitab	Limited number of articles evaluated (513); no changepoints	[19]
Scientometric of Nearly Zero Energy Building Research: A Systematic Review from the Perspective of Co-Citation Analysis/Wang et al. (2019)	Nearly zero-energy building research	2006–2018	Web of Science	HistCite and CiteSpace softwares	Citation lag, reliance on Web of Science only, narrow search term, exclusion of related terms and limited number of articles evaluated (704)	[18]
Bibliometric analysis of zero energy building research, challenges and solutions/ Agbodjan et al. (2022)	Zero-energy buildings	1976–2020	Scopus	VOSviewer and Biblioshiny	Narrow search term, limited number of articles evaluated (1051) and no macro-evolution	[20]
Biomimicry for Energy-Efficient Building Design: A Bibliometric Analysis/Varshabi et al. (2022)	Biomimicry for energy-efficient building design	2010–2021	Web of Science	VOSviewer and SankeyMATIC software	Short analysis period, reliance on a single database, narrow search terms, insufficient biomimicry research volume, possible omission of relevant studies and limited number of articles evaluated (53)	[17]
A bibliometric review of net zero energy building research 1995–2022/Omrany et al. (2022)	Net-zero energy buildings (NZEBs)	1995–2022	Web of Science	VOSviewer software and SciMAT tool	Reliance on a single database, fragmented research field and volatile thematic evolution	[15]
Systematic literature review and bibliometric analysis of energy efficiency/Tripathy et al. (2024)	General energy efficiency (multi-sector)	2001–2022	Scopus and Web of Science	PRISMA-2020 framework, Bibliometrix in R (RStudio) and VOSViewer software	Narrow search term, limited number of articles evaluated (619) and short analysis period	[21]
Applications and Trends of Machine Learning in Building Energy Optimization: A Bibliometric Analysis/Liu and Chen (2025)	Machine-learning applications in building energy optimization	2020–2024	Web of Science	SALSA framework, VOSviewer and Bibliometrix (Biblioshin)	Limited number of articles evaluated (496), short analysis period and uneven research output	[16]

DTB, database; Ref, references.

summarizes previous bibliometric investigations in building energy efficiency and related subdomains, focusing on their temporal scope, analytical depth, and database selection.

This research addresses these limitations by providing a comprehensive bibliometric analysis of building energy efficiency research spanning four decades (1981–2025). This extended temporal scope is essential for capturing the field’s complete maturation trajectory: from foundational research establishing core thermal physics principles in the 1980s–1990s, through the emergence of integrated design approaches following the Kyoto Protocol (1997) [22], to the dramatic acceleration coinciding with the Paris Agreement (2015) [23] and subsequent net-zero commitments by major economies. Understanding this four-decade evolution is critical for identifying how research priorities have responded to, and anticipated, major policy shifts, and for revealing the knowledge accumulation patterns that enabled contemporary integrated building performance frameworks. While this study does not directly implement building energy efficiency measures, it provides evidence-based guidance for research strategy and policy development, by identifying productive research clusters, emerging technologies, and underexplored areas requiring targeted investment.

The bibliometric analysis introduces two critical methodological innovations that enhance analytical rigor and reproducibility. First, this study employs Pruned Exact Linear Time (PELT) changepoint detection to objectively identify temporal transitions through statistical validation, replacing the arbitrary decade-based periodization commonly employed in bibliometric investigations. Second, automated keyword harmonization through TF-IDF vectorization (n-grams 2–3) and cosine similarity algorithms (threshold = 0.75) addresses the semantic inconsistency challenge inherent in evolving technical nomenclature, a computational approach not previously applied in building energy efficiency bibliometrics. This study aims to systematically map the temporal evolution of research priorities from foundational thermal physics through contemporary smart building integration, examine geographic distribution and international collaboration networks across countries, identify institutional leadership patterns and research productivity dynamics. It also traces the thematic evolution from component-focused investigations to holistic building performance optimization and reveals citation impact patterns that highlight influential breakthroughs and emerging research frontiers. The analysis integrates publication trend assessment, keyword co-occurrence mapping, collaboration network visualization, and citation impact evaluation to provide actionable intelligence for researchers, policymakers, and industry stakeholders navigating the transition toward net-zero building portfolios.

2. Materials and Methods

The identification of key documents in the field was facilitated by the construction of a search algorithm, which incorporated the Boolean operators OR and AND to expand the search area and select relevant documents. For minimizing the number of false positive results included in this analysis, techniques such as exact phrase matching, which involves enclosing the word or phrase in quotation marks (e.g., “keyword”), and the wildcard operator *, which allows matching all variations of a word regardless of its ending (e.g., efficient* will identify terms such as efficiency and efficient), were used. These strategies have been integrated to ensure an accurate and rigorous selection of relevant literature, minimizing interference and improving the quality of the results obtained. The initial search query employing wildcard operators yielded 77,120 results from the Web of Science Core Collection; however, preliminary analysis revealed substantial numbers of false positive results due to broad pattern matching. To enhance precision, the query was refined through exact phrase matching by enclosing key terms in quotation marks (e.g., “energy efficien*” rather than energy efficien*), reducing the dataset to 18,449 documents while eliminating irrelevant publications that contained the search terms in non-building energy contexts. Subsequently, systematic filters were applied to restrict the dataset to articles and review papers written in English, ensuring compatibility with the computational text processing framework and resulting in a final analytical dataset of 14,253 documents (Figure 1). This multi-stage refinement process prioritized relevance and analytical precision over comprehensive coverage, focusing the investigation on peer-reviewed research directly addressing building energy efficiency. This systematic approach improved algorithm performance and ensured rigorous data selection. Data for this investigation were retrieved exclusively from the Web of Science Core Collection. The decision to rely on a single database was driven by the need for consistent data architecture and quality standards throughout the analysis. Crucially, the database provides uniformly structured metadata fields, which proved essential for the computational methods employed in this study.

The analytical pipeline integrating VOSviewer (version 1.6.20) for network visualization, Bibliometrix 4.3.0 package in R. for statistical analysis, and custom Python 3.12.3 algorithms for computational keyword harmonization depends critically on standardized data formats. Consistent field structures across records enabled accurate processing of co-authorship networks, precise temporal trend analysis, and reliable thematic evolution mapping without the data integration challenges and systematic inconsistencies that arise when merging multiple bibliographic sources. Beyond these immediate technical considerations, the single-database approach proves particularly critical for the changepoint detection methodology employed in this study. The PELT algorithm applied to identify temporal transitions in publication activity requires uninterrupted, consistently measured time series data to accurately detect statistical breakpoints. Introducing data from multiple sources would create artifacts in the temporal signal due to varying indexing delays, different inclusion criteria across databases, and inconsistent publication date recording conventions. Similarly, the thematic evolution analysis depends on uniform keyword assignment practices maintained by Web of Science indexers, whereas merging databases with disparate controlled vocabulary systems would introduce noise into the co-occurrence networks and potentially obscure genuine thematic transitions. These methodological requirements demonstrate that for advanced computational bibliometric analyses, database consistency constitutes a prerequisite for analytical validity rather than merely a convenience factor.

The choice of software programs used is of paramount importance for the execution of a bibliometric analysis. The present study primarily employed VOSviewer (version 1.6.20), a widely utilized software program for bibliometric analysis, which facilitates the generation of bibliometric maps with ease. The Bibliometrix library (version 4.3.0) was employed for detailed analysis of bibliometric data, in conjunction with the Biblioshiny web interface, which significantly facilitates the use of the library. Microsoft Excel was utilized for the cleaning of bibliometric data and the creation of graphs. The Bibliometrix library, via the Biblioshiny interface, was executed using RStudio (version 4.4.2), an Integrated Development Environment for the R language. This combination of software tools enabled a rigorous bibliometric analysis and clear visualization of the results [24,25].

The annual number of publications was extracted using the Bibliometrix library, and the resulting data were visualized to depict trends in scientific output over time. Graphical representations of publication trends were generated using Microsoft Excel, facilitating a clear understanding of temporal patterns. The Mean Total Citations per Year (MeanTCperYear) metric was employed as a normalized measure of citation impact. This metric was calculated for each publication by dividing the total number of citations received by the number of years since publication. These values were computed automatically using the Biblioshiny web interface, ensuring consistency and accuracy. Country-level and institution-level scientific production metrics were derived from data exported via VOSviewer (version 1.6.20), allowing for an in-depth analysis of geographical and institutional contributions. Temporal trends in scientific output by country and institution were analyzed using the Biblioshiny interface, providing valuable insights into the evolution of research activity across regions and organizations.

In the visualization of international research collaboration networks, each nation’s academic productivity is indicated by the diameter of its corresponding circle with more prolific countries represented by more expansive circular elements. The degree of academic partnership between nations manifests in the thickness of the connective elements joining their respective circles, where more substantial lines signify more extensive collaborative engagement. The network’s structural organization employs a color-coding methodology to delineate distinct collaborative communities. These chromatic groupings highlight the natural clustering of nations that maintain particularly robust research relationships, as countries within shared color designations typically demonstrate heightened levels of academic interchange and cooperative scholarly endeavors.

Temporal research mapping provides visual representations that capture the developmental patterns of scientific themes across successive periods. These graphs enable researchers to observe the transformation of research priorities and to monitor the emergence or decline of specific topics within disciplines. The Bibliometrix platform generates these graphs through comprehensive co-word analysis and clustering methodologies, thereby revealing the progressive development of scientific knowledge across temporal intervals.

To identify key transition points in publication trends objectively, the PELT algorithm was applied using the Python ruptures library (version 1.1.0) with a Radial Basis Function (RBF) model. This algorithm efficiently detects significant shifts in time series data by identifying changes in statistical patterns. The analysis examined annual publication volumes from 1981 to 2024 to detect major transitions in research activity, with data handling performed using pandas (2.0.0) and numpy (1.21.0). To ensure reliability, the algorithm was tested with five different penalty parameter values (1, 2, 3, 4, 5), which control detection sensitivity, lower penalties identify more potential breakpoints while higher penalties require stronger evidence for change detection. Only breakpoints consistently identified across multiple tests were retained, specifically, breakpoints appearing in at least three out of five analyses were classified as robust and used for temporal periodization. The detected periods were then statistically validated using Mann–Whitney U tests implemented through scipy (1.9.0) to confirm significant differences between consecutive time periods. This data-driven approach provides objective evidence for defining distinct research phases, replacing arbitrary decade-based divisions with statistically supported temporal boundaries.

Trending research topics are analyzed to identify and visualize how scientific themes change over time. This methodological framework enables researchers to observe the increasing prominence of emerging research domains and the diminishing focus on established areas. Through the Bibliometrix platform’s capability to analyze keyword frequency distributions across temporal periods, researchers can conduct robust assessments of evolving research landscapes.

The map of the journal collaboration network shows each node as a source, with the size of the node proportional to the number of documents published by that source. The color of each node shows the average year of publication of documents from that specific source, offering valuable insights into the trends over time. The lines that connect two nodes show how documents from one source are mentioned in the other, and the thickness of the lines shows how important the connection is. If a line is thick, it shows that there are more mentions of one document in another.

The utilization of a singular language in bibliometric analysis constitutes the inaugural step in the standardization process. Subsequent to this, the generation of thesaurus documents is imperative, as these facilitate the unification of terms that may be synonymous or manifest in diverse forms. To illustrate this point, the standardization of country names necessitates thesaurus generation for harmonizing disparate variants of the same name or to incorporate regions such as Wales, which is a constituent of the United Kingdom (UK). The generation of a thesaurus for keywords is a more complex undertaking, given the vast number of terms requiring analysis. The implementation of a normalization script written in Python 3.12.3 was developed, with a focus on a specialized thesaurus generator designed to enhance the analysis of keyword trends and the study of thematic evolution (Figure 2). The generator employs character-level TF-IDF vectorization with n-grams (2–3-character sequences) and cosine similarity metrics (threshold = 0.75), implemented using scikit-learn (1.1.0) with supporting data processing through numpy (1.21.0), in combination with comprehensive text preprocessing, to identify and cluster semantically and lexically related terms. A cosine similarity threshold of 0.75 was selected based on empirical testing: lower thresholds (e.g., 0.6–0.7) produced clusters with excessive noise and unrelated terms, while higher thresholds (>0.8) fragmented semantically coherent groups. A value of 0.75 provided the best balance between precision and recall. While semantic embedding approaches using pre-trained language models (e.g., BERT) were evaluated, they proved unsuitable due to the lack of domain-specific training on building energy efficiency terminology. The character-level n-gram approach was selected for its superior ability to capture lexical variation, subtle morphological patterns, and spelling inconsistencies frequently encountered in highly specialized technical nomenclature across diverse datasets and contexts, ensuring improved consistency and overall model robustness.

3. Results

3.1. Literature Overview

During the initial period of 1981–1996, the mean citation impact was found to be comparatively low, with an average of below 0.5 citations per year. This finding suggests that papers published during this time received an average of fewer citations each year after their initial publication. The subsequent period of 1997–2007 witnessed a significant increase in citation impact, with values ranging between 1.0 and 3.0 mean total citations per year. The year 2000 demonstrated a notable peak of 2.36 mean total citations per year for papers published in that year. From 2008 to 2021, a significant increase in citation impact was observed, with MeanTCperYear values consistently above 3.5. Notably, papers published between 2011 and 2020 exhibited particularly high levels of influence, receiving an average of 4–4.6 citations per year. Conversely, recent years have exhibited a decline in MeanTCperYear, which may be attributed to the reduced time these documents have had to accrue citations. The number of papers published in the initial 25-year period (1981–2005) was minimal, with less than 100 papers being published, suggesting a paucity of interest in the field during that period. The subsequent period (2005–2016) witnessed a consistent rise to approximately 500 papers, indicating a growing interest among researchers. The period 2016–2024 marked a significant increase, culminating in 2024 with a total of 2303 papers published. Figure 3 illustrates the trends in the researched area between 1981–2025, highlighting the dynamics of scientific output and the growth of academic interest over time.

3.2. Country Scientific Production

Of the 131 countries that published papers in this field, only 38 (29.01%) had at least 100 papers published. China is distinguished as the clear leader in terms of volume of research, with 2978 papers published, significantly more than the United States (USA) in second place (1478 papers) and the UK in third place (1421 papers. This result reflects China’s strong investment in this research domain. The distribution reveals a notable disparity between the top four countries (China, USA, UK and Italy, each with over 1300 papers) and the remaining countries included in the study. With regard to research impact, the UK (33.24 citations/paper) and USA (33.15 citations/paper) have the highest citation impact per publication, indicating that their research tends to be more influential at the individual level per published paper.

Table 2. Top 10 countries by scientific output and citation impact on energy efficiency of buildings (1981–2025).

Country	No. of Documents	Total Citations	Average Citations/ Document
China	2978	82,112	27.57
USA	1478	49,001	33.15
UK	1421	47,230	33.24
Italy	1339	37,042	27.66
Spain	844	19,041	22.56
Australia	672	19,688	29.3
India	570	11,980	21.02
Canada	513	16,164	31.51
South Korea	466	9713	20.84
Turkey	433	7695	17.77

highlights the scientific output and citation impact of the top 10 countries in the field of energy efficiency of buildings between 1981 and 2025.

The scientific output reveals distinct periods of research activity. The period between 1981 and 2000 is characterized by the USA pioneering role in the field, with consistent publications since 1981, suggesting the potential origin of the field in US research institutions. By 2000, the field had begun to attract researchers from other countries, albeit in limited numbers. In the subsequent period, a substantial increase in publications from other countries is observed, with the USA maintaining its dominance in the field. The period between 2010 and 2020 is characterized by a significant surge in the number of papers published, particularly from China. By the end of 2020, China had published 3167 papers, compared to 2150 from the USA. The subsequent period (2021–2025) witnessed China sustaining its upward trajectory in terms of publication output, while other countries also exhibited an increase, though at a comparatively lower rate. Figure 4 presents the scientific output over time of the most productive countries.

3.3. Institution Scientific Production

Figure 5 shows the most productive affiliations in energy efficient construction. The top five most productive institutions are all located in China, with a particularly strong representation of Hong Kong institutions. Hong Kong Polytechnic University leads the ranking with 458 articles, followed by the City University of Hong Kong with 266 articles. Together, these two Hong Kong institutions contribute a total of 724 articles, highlighting Hong Kong’s important role as a research center in this field.

Figure 6 illustrates the progression of scientific output by the most prolific institutions over time. During the period spanning from 1981 to 1999, these institutions demonstrated a complete absence of research output in this domain, with activity primarily cantered in the USA.

After 2000, however, the City University of Hong Kong initiated a pioneering role among these institutions, commencing in 2000 with two publications and subsequently stabilizing at a rate of four publications per annum. In 2002, Hong Kong Polytechnic University joined the effort, highlighting the leading role of Hong Kong institutions in this area of research. Between 2005 and 2010, Hong Kong universities began to publish an increasing number of papers, while mainland Chinese institutions also entered the field. The subsequent period, 2011–2025, was characterized by accelerated growth. Hong Kong Polytechnic University demonstrated the most consistent and robust development throughout this period, maintaining its leading position through a steady expansion of scientific output. Southeast University showed the most dramatic transformation, moving from minimal output in the early years to major contributor status in the later years, while the other universities showed a gradual increase in the number of published papers.

3.4. Most Influential Papers in the Evaluated Field

As demonstrated in

Table 3. Top 10 most influential documents in the field of energy efficiency in buildings (1981–2025).

Paper/Source	TC	TC/Year	Normalized TC	DOI
Bocken N.M.P., 2014, J CLEAN PROD	1967	163.92	36.79	10.1016/j.jclepro.2013.11.039
Lund H., 2014, ENERGY	1522	126.83	28.46	10.1016/j.energy.2014.02.089
Swan L.G., 2009, RENEW SUST ENERG REV	1320	77.65	20.69	10.1016/j.rser.2008.09.033
Nejat P., 2015, RENEW SUST ENERG REV	1273	115.73	25.98	10.1016/j.rser.2014.11.066
Omer A.M., 2008, RENEW SUST ENERG REV	1248	69.33	17.24	10.1016/j.rser.2007.05.001
Crawley D.B., 2008, BUILD ENVIRON	1040	57.78	14.37	10.1016/j.buildenv.2006.10.027
Yang L., 2014, APPL ENERGY	956	79.67	17.88	10.1016/j.apenergy.2013.10.062
Dittenber D.B., 2012, COMPOS PT A-APPL SCI MANUF	953	68.07	16.32	10.1016/j.compositesa.2011.11.019
Sadineni S.B., 2011, RENEW SUST ENERG REV	888	59.20	13.83	10.1016/j.rser.2011.07.014
Cao X., 2016, ENERGY BUILD	868	86.80	20.38	10.1016/j.enbuild.2016.06.089

TC, total citations.

, which details the most influential papers in this area, the most cited paper by Bocken et al. (2014) focuses on archetypes of sustainable business models, accumulating 1967 citations. Its high impact (163.92 citations per year) suggests the growing importance of sustainability in business model innovation, and the prominence of this work indicates that the field places a strong emphasis on the practical implementation of sustainability principles in business contexts. A further notable observation is that articles published since 2014 tend to have a higher number of citations, reflecting a growing interest in energy efficiency and sustainability, likely influenced by intensifying global concerns about climate change and environmental policy.

3.5. Scientific Mapping on Country Collaboration

Figure 7 presents a network map of co-authorship among countries, focusing exclusively on those with a minimum of 40 published research papers. The network analysis reveals four distinct clusters that highlight regional and global patterns of collaboration. The red cluster, which encompasses 30 countries, is predominantly composed of European nations, with Italy serving as the central node. This cluster reflects strong regional cohesion and intense collaborative efforts within Europe. The green cluster, which includes 26 countries, is dominated by three major research hubs: China, USA and the UK. These countries act as central pillars of the global research landscape, facilitating extensive international partnerships and underscoring their roles as influential leaders in the field. In contrast, the yellow and blue clusters represent smaller groups of five countries each, positioned in proximity to the red cluster. These mini clusters indicate robust collaborations with European nations while also highlighting shared linguistic and cultural ties that strengthen these partnerships. The blue cluster, particularly, demonstrates frequent and consistent collaborations with the red cluster, further emphasizing common affinities. The network, when considered as a whole, demonstrates a global structure of research cooperation, with the major hubs, China, the USA, the UK, and Italy playing a central role in fostering international connections, while the smaller clusters contribute regional and cultural nuances to the broader research ecosystem.

3.6. Thematic Evolution

Prior to examining thematic evolution, a PELT changepoint detection analysis was conducted to objectively identify significant transitions in publication activity. The analysis identified 2011 as the most robust breakpoint, detected consistently across all sensitivity tests (Figure 8, left panel). This transition shows strong statistical significance (Mann–Whitney U test, p < 0.000001, effect size = 2.48), with median annual output increasing from 15 articles (1981–2011) to 840.5 articles (2011–2024). The growth rate analysis (Figure 8, right panel) reveals volatile patterns during 2001–2010, with annual fluctuations exceeding 200%, reflecting the field’s transitional phase. Post-2011, growth rates stabilized with markedly reduced volatility, indicating sustained and predictable field maturation. These data-driven breakpoints validate the thematic evolution periods employed in this study, with 2011 marking the critical transition from foundational research to the contemporary era of integrated, application-focused investigations.

The thematic evolution analysis, structured according to these PELT-validated periods, reveals a major transformation in research organization (Figure 9). The foundational period (1981–2011) was characterized by seven distinct research themes behavior, heat-transfer, simulation, impact, performance, consumption, and optimization reflecting exploratory, specialized investigations across multiple technical dimensions. In contrast, the contemporary period (2012–2025) demonstrates dramatic consolidation into two dominant themes: performance and simulation, with performance emerging as the clearly predominant focus. The Sankey diagram illustrates that all seven foundational themes contribute flows to these two contemporary themes, indicating integration rather than abandonment of earlier research directions. Flow weight analysis reveals distinct integration patterns: consumption (weighted inclusion index = 0.85), optimization (0.70), and impact (0.57) predominantly transitioned into performance-focused research, while behavior (1.00) completely transitioned to simulation-driven approaches. Heat-transfer research exhibited bifurcation, with stronger contributions to simulation (0.40) than performance (0.10), reflecting the dual role of thermal analysis in both predictive modeling and performance assessment. The performance and simulation themes themselves demonstrated substantial continuity (0.75 and 0.76 respectively), indicating that these domains matured from foundational concepts rather than emerging de novo in the contemporary period.

This evolution reflects the field’s maturation from fragmented, component-focused investigations toward holistic, integrated frameworks centered on comprehensive performance assessment and advanced simulation methodologies.

3.7. Trend Evolution

Figure 10 illustrates the trending topics in the field of energy efficiency in buildings over different time periods. A review of the initial research topics (2006–2010) indicates a predominant focus on foundational systems and regional studies, with electric heating systems and assessment schemes being of particular note. During this period, pinch technology emerged as a method for optimizing energy systems, reflecting an early interest in energy optimization techniques. The early 2010s (2011–2015) witnessed a substantial expansion in the research landscape, with environmental assessment methods gaining prominence, signifying a growing awareness of sustainability metrics. By the mid-2010s (2015–2018), research began to adopt more comprehensive analytical approaches, with life-cycle analysis and life-cycle energy studies gaining increasing significance. This transition is evidenced by a substantial increase in publications dedicated to life-cycle energy, with 53 publications focusing on this subject during the period under review. The most recent period (2019–2024) has witnessed a significant expansion across several key areas. The core themes of buildings (1537 publications), consumption (1222), and performance (2637) have dominated the research landscape, while design (1629) and optimization (1144) have also featured prominently. These trends reflect a mature field increasingly focused on practical improvements and applications. New technological frontiers have emerged, particularly in materials science, where PCMs have garnered significant attention (271 publications), along with concrete (248). Artificial intelligence has begun to play a role in building energy efficiency, with notable developments observed in 2023–2024. The utilization of advanced materials, such as graphene oxide, in conjunction with the adoption of Building Information Modelling (BIM), signifies the advent of a novel era of innovation, underscoring the field’s perpetual evolution and its prioritization of state-of-the-art technologies.

3.8. Keywords Co-Occurrence Network-Map

As illustrated in Figure 11, a keyword co-occurrence network map comprising 85 keywords, each occurring a minimum of 30 times in the analyzed corpus, is presented. The visualization reveals five distinct clusters, each representing interconnected thematic areas in building energy efficiency and sustainability research. The largest cluster, depicted in red, encompasses 57 keywords and primarily focuses on policy implementation and consumption patterns. This cluster is centered around energy efficiency and consumption metrics, incorporating terminology related to decision-making processes, carbon emissions assessment, renovation strategies, and climate change mitigation approaches. The second-largest cluster, which is represented by the green color and 45 keywords, pertains to the technical and physical aspects of building performance. This cluster encompasses terminology related to the assessment of thermal performance, the mechanisms of heat transfer, the characteristics of the building envelope, and insulation specifications. The blue cluster, containing 29 keywords, emphasizes building energy and thermal performance metrics, thus indicating a strong focus on quantifiable performance indicators and efficiency measurements in building systems, as suggested by the interconnections within this cluster. The yellow cluster, comprising 26 keywords, concentrates on energy consumption optimization and methodological approaches for achieving enhanced efficiency, demonstrating strong connections to both analytical methods and practical implementation strategies. The purple cluster, though smaller with 10 keywords, focuses on building thermal comfort parameters, and its proximity to both the yellow and green clusters indicates the significance of optimizing comfort-related aspects in conjunction with technical and efficiency considerations. This spatial relationship in the visualization suggests an integrated approach to building performance optimization that considers both occupant comfort and energy efficiency.

4. Discussion

Over time, research in building energy efficiency has evolved from focusing on individual components and systems to adopting more integrated and holistic approaches.

The level of scientific interest displayed in the field of building energy efficiency has undergone significant evolution over time. Distinct periods of growth and development can be identified in this area of research. In recent decades, the field has undergone considerable expansion, reflecting the mounting global emphasis on sustainable construction and energy conservation in buildings.

The PELT changepoint analysis identified 2011 as a statistically significant transition point in publication activity. This breakpoint reflects the convergence of major policy initiatives that created unprecedented research demand: the European Union’s Energy Performance of Buildings Directive recast (2010/31/EU), introducing mandatory Nearly Zero-Energy Building targets [26], China’s 12th Five-Year Plan (2011–2015) established targets for energy intensity reduction [27] and the United States adopted ASHRAE Standard 90.1-2010, which demonstrated 18.2% energy savings over previous standards [28].

The foundational period (1981–2011) was characterized by pioneering research primarily led by the USA, suggesting its foundational role in establishing this field of study. Within this extended formative phase, observable sub-phases emerged: the initial years (1981–2000) showed relatively modest publication output with limited international participation, while the subsequent decade (2001–2010) witnessed considerable transformation with the emergence of novel research centers, particularly within the Asian region. This period marked the advent of Hong Kong institutions into the field, with the City University of Hong Kong and the Hong Kong Polytechnic University attaining distinction as substantial contributors, signifying mounting global recognition of building energy efficiency as a pivotal research domain. The contemporary period (2012–2025) represents the most dramatic growth phase, characterized by China’s emergence as the dominant force in this research domain, with the 2011 breakpoint marking a fundamental shift from exploratory investigations to mature, application-focused research. China’s remarkable ascent is evidenced by its publication output of 2978 papers, substantially surpassing the USA (1478 papers) and the UK (1421 papers). This surge in Chinese research output coincides with the country’s increased focus on sustainable development and environmental protection measures.

The apparent paradox of China’s high publication volume (2978 papers) coupled with lower average citation impact compared to the UK (33.24 citations/paper) and USA (33.15 citations/paper) can be primarily attributed to citation lag effects rather than research quality differences. As demonstrated by the PELT analysis, China’s research productivity surge occurred predominantly after 2011, with the country transitioning from minimal output in the foundational period to dominant contributor status in the contemporary period. Consequently, the majority of Chinese publications are temporally recent and have had substantially shorter citation windows compared to UK and USA publications, which maintain more balanced temporal distribution across both foundational and contemporary periods. Citations accumulate gradually over years following publication, with peak citation rates typically occurring 3–5 years post-publication. Thus, China’s concentration of output in the 2012–2025 period inherently constrains average citation metrics when calculated across the entire corpus. Additional contributing factors may include the UK and USA’s established roles as central hubs in international collaboration networks (Figure 7), which tend to enhance citation visibility, and the longer research tradition in these countries enabling publication in higher-impact specialized journals.

At the institutional level, the pattern of growth mirrors the country-level trends, with Chinese institutions, particularly those in Hong Kong, assuming leadership positions. The Hong Kong Polytechnic University’s output of 458 articles and the City University of Hong Kong’s contribution of 266 articles underscore the region’s emergence as a major research hub in building energy efficiency.

The co-authorship network map reveals distinct geographical clustering and collaboration patterns that illuminate the global research ecosystem in building energy efficiency. The visualization demonstrates three primary clusters that reflect both regional strengths and international partnerships. It is evident that while regional research hubs maintain their importance, successful research in this field increasingly relies on international collaboration. This global approach enables the sharing of diverse perspectives, methodologies, and solutions, which is particularly valuable given the varying climatic, technological, and regulatory contexts across different regions. The observed collaboration patterns indicate that future research endeavors should concentrate on the consolidation of cross-regional partnerships, whilst concurrently preserving the unique expertise cultivated within each distinct cluster. This balanced strategy would serve to address both the local building energy challenges and the global sustainability objectives. The European cluster’s cohesion reflects structural policy frameworks that mandate cross-border collaboration. EU Framework Programmes (FP7, Horizon) [29,30] require collaborative projects to include multiple organizations from different EU Member States or Associated Countries, creating policy-driven incentives for multi-country research networks. The Concerted Action for the Energy Performance of Buildings Directive (CA EPBD), launched in 2005 and continuing through multiple phases, established systematic platforms for knowledge exchange among all EU Member States and Norway specifically for implementing building energy efficiency directives. This combination of mandatory collaboration requirements and sector-specific coordination mechanisms transformed building energy efficiency research from isolated national efforts into integrated European networks.

Different research trends have been identified during the period under study, denoting the significant evolution demonstrated by this field. In the period before 2000 when interest in this field was still relatively low research during this phase focused primarily on individual components and systems, as exemplified by Athienitis et al.’s (1997) work on PCM in passive solar applications [31]. This period also saw the emergence of innovative approaches to building envelope design, such as dynamic insulation (Taylor and Imbabi, 1998) [32]. Notably, the economic dimensions of energy-efficient construction began to receive scholarly attention, with Bartlett and Howard (2000) providing critical analysis that challenged prevailing cost assumptions about green buildings [33]. The early 2000s witnessed a paradigm shift toward environmentally conscious building design. Bell and Lowe’s (2000) seminal study demonstrated the feasibility of achieving 50% emissions reduction through housing modernization using existing technologies [34]. This period also saw the emergence of comprehensive evaluation methodologies, exemplified by Swan and Ugursal’s (2009) development of sophisticated building energy performance modeling frameworks [3]. The integration of environmental considerations into building design marked a significant departure from purely efficiency-focused approaches. Recent documents in this field highlight the maturation of the research topic, characterized by the advancement of integrated building energy solutions. Marszal et al.’s (2011) seminal work on ZEBs established a foundational framework for comprehensive energy-neutral design [35]. More recently, Wang et al.’s (2022) critical analysis of PCM applications [36] and Hafez et al.’s (2023) systematic review of energy efficiency strategies [37] underscore the field’s current focus on holistic sustainability approaches.

To better situate the evolution of the field within its broader global context, a timeline of major policy and technology milestones is provided below. Figure 12 presents a chronological timeline of key developments that have shaped building energy efficiency research and help contextualize the 2011 breakpoint.

The 2011 breakpoint coincides with a rare confluence of policy and technology shifts that reshaped building-energy research trajectories: the launch of ISO 50001 established a global energy-management standard applicable to building portfolios and facility operations, creating a shared framework for measurement and continual improvement [38].

In Europe, the recast EPBD (2010/31/EU) had just mandated nearly zero-energy buildings for the coming decade, catalyzing methodological work and national transpositions from 2010 onward [26].

In the United States, the DOE 2011 determination on ASHRAE 90.1-2010 ratcheted commercial-building code stringency and signaled measurable savings versus the 2007 edition, prompting code adoption and compliance research. This regulatory advancement established that newly constructed commercial buildings adhering to the updated standard would demonstrate quantifiable reductions in energy consumption when benchmarked against structures built under previous code requirements [39].

At the same time, the IEA’s 2011 update of its 25 Energy Efficiency Policy Recommendations renewed high-level commitments on building efficiency and appliances, reinforcing a global policy push. This policy refresh documented substantial progress in cross-sectoral implementation, revealing that numerous countries had transitioned from planning stages to active deployment of stringency measures initially proposed in 2009. The evaluation highlighted a coordinated international effort toward enhanced building codes, with nations such as Canada, Korea, Luxembourg, the Netherlands, and the United Kingdom introducing strengthened requirements for new construction. Additionally, the systematic expansion of minimum energy performance standards for appliances and equipment demonstrated a maturing regulatory landscape. The IEA’s renewed emphasis on compliance monitoring, enforcement mechanisms, and standardized measurement protocols signaled a strategic shift from aspirational targets to accountable implementation, thereby catalyzing a synchronized global movement toward verifiable energy efficiency achievements in the built environment and consumer product sectors [40].

On the technology side, the PV market inflected sharply in 2010–2011, with documented price declines and an 82% annual growth rate of cumulative installed capacity across IEA-PVPS countries and ≥1 GW annual additions in six nations, conditions that accelerated work on nZEB/BIPV and integrated building-energy systems [41,42].

Crucially for the post-2011 publication surge and China’s prominence, the 12th Five-Year Plan (2011–2015) set a binding −16% energy-intensity target [43] and was followed by the 2013 Green Building Action Plan (targeting ~1 billion m² of new green buildings) [44], which together expanded funding, demonstration, and regulatory activity.

Consolidation continued immediately after the breakpoint with the EU Energy Efficiency Directive (2012/27/EU), which introduced binding measures (audits, renovation obligations, public-sector leadership) [45], and then with the Paris Agreement (2015), which anchored national commitments and finance, both amplifying the post-2011 research wave rather than creating it [23].

It is imperative to prioritize energy efficiency in the construction of sustainable structures by employing a variety of strategies that are designed to reduce environmental footprints. These strategies include the integration of building-mounted photovoltaic systems and solar panels, the utilization of materials that conserve resources, and the exploration of cutting-edge energy technologies. These strategies also involve enhancing illumination solutions, conducting energy simulations, assessing thermal transfer properties, optimizing building envelopes, incorporating wind turbines into skyscrapers, instituting energy conservation strategies, and employing multi-objective optimization techniques [2].

The field has evolved from the provision of focused technical solutions to the delivery of comprehensive approaches that incorporate multiple sustainability dimensions. This progression is indicative of two interrelated phenomena: firstly, the advancement of technology, and secondly, the growing awareness of the role of building energy efficiency in addressing global environmental challenges. The future direction of research will continue to integrate smart technologies and to place greater emphasis on resilient, adaptive building systems.

While this bibliometric investigation provides comprehensive insights into building energy efficiency research evolution, several methodological limitations should be acknowledged. The exclusive use of Web of Science Core Collection represents a deliberate choice that prioritizes analytical precision over comprehensive coverage. This single-database approach ensures consistent metadata structures, standardized citation linkages, and uniform field coding all necessary for reliable network analysis, temporal trends, and computational keyword normalization.

The restriction to publications written in English represents a deliberate methodological necessity for the quantitative computational analysis employed in this study. The computational terminology normalization framework utilizing TF-IDF vectorization and cosine similarity algorithms operates on character-level and word-level patterns that are fundamentally language-specific and cannot validly compare terms across different linguistic systems. Attempting multilingual analysis would require either separate processing pipelines for each language and thus fragmenting the dataset and preventing integrated thematic evolution analysis, or automated translation systems that introduce semantic drift and compromise keyword clustering precision. Notably, the dataset demonstrates substantial representation from non-English speaking regions, with China alone contributing 2978 papers and dominating the contemporary period, indicating that researchers from diverse linguistic backgrounds actively publish in English to achieve international scholarly visibility.

The PELT-based temporal periodization, while statistically robust, identifies breakpoints in publication volume rather than thematic transitions. As a result, these statistically detected periods may not perfectly align with major policy initiatives or technological breakthroughs that shaped the field conceptually. Additionally, the keyword co-occurrence analysis depends on author-assigned terms and database indexing conventions, which may not fully capture underlying research themes or interdisciplinary connections that emerge implicitly across studies. Future investigations could benefit from hybrid approaches that integrate multiple databases and languages while maintaining the analytical rigor required for reliable bibliometric network construction and computational text analysis.

5. Conclusions

Energy efficiency research in buildings has evolved through several distinct phases of growth and diversification. The rise in worldwide collaboration, coupled with an intensified focus on climate change mitigation and energy optimization, has advanced the field toward more sophisticated, integrated solutions. Consequently, the emphasis is now transitioning to innovative materials, optimization of energy systems, and the creation of ZEBs.

This bibliometric analysis establishes a computationally enhanced framework with direct implications for academic research strategy and methodological transferability. The collaboration network analysis identifies four distinct geographic clusters, providing researchers with evidence-based guidance for optimizing international partnerships and maximizing citation impact through strategic co-authorship positioning. Methodologically, the integration of PELT changepoint detection with TF-IDF-based terminology normalization addresses two persistent challenges in bibliometric research: objective temporal periodization and automated keyword harmonization across evolving technical nomenclature. This approach is directly adaptable to other rapidly evolving interdisciplinary domains. The framework demonstrates that computational text processing techniques can enhance traditional bibliometric analysis without requiring specialized language models, making it accessible for emerging research fields.

Future research in energy efficiency in buildings is anticipated to focus on artificial intelligence integration, energy storage systems, and sustainable building materials. The advent of PCM and innovative technologies such as graphene oxide, in conjunction with Building Information Modeling, is expected to significantly influence future advancements. Cross-regional cooperation will be increasingly vital as global sustainability objectives become more interrelated, necessitating solutions that reconcile local requirements with international standards. Furthermore, with a heightened emphasis on carbon neutrality, forthcoming research will probably investigate more extensive frameworks for assessing environmental effect, incorporating policy recommendations, and addressing urbanization difficulties. From a bibliometric perspective, future investigations could extend this computational framework through comparative analysis across multiple databases (e.g., Scopus, regional repositories) to validate temporal patterns, integration of natural language processing for full-text analysis to capture implicit thematic connections beyond keywords, and tracking real-world research impact through policy citations and patent references to assess practical implementation influence.