Pub Date : 2025-09-18DOI: 10.1016/j.adapen.2025.100242
Dorothea Kistinger , Maurizio Titz , Philipp C. Böttcher , Michael T. Schaub , Sandra Venghaus , Dirk Witthaut
Effective governance of energy system transformation away from fossil resources requires a quantitative understanding of the diffusion of green technologies and its key influencing factors. In this article, we propose a novel machine learning approach to diffusion research focusing on actual decisions and spatial aspects complementing research on intentions and temporal dynamics. We develop machine learning models that predict regional differences in the accumulated peak power of household-scale photovoltaic systems and the share of battery electric vehicles from a large set of demographic, geographic, political, and socio-economic features. Tools from explainable artificial intelligence enable a consistent identification of the key influencing factors and quantify their impact. Focusing on data from German municipal associations, we identify common themes and differences in the adoption of green technologies. Specifically, the adoption of battery electric vehicles is strongly associated with income and election results, while the adoption of photovoltaic systems correlates with the prevalence of large dwellings and levels of global solar radiation.
{"title":"Revealing drivers of green technology adoption through explainable Artificial Intelligence","authors":"Dorothea Kistinger , Maurizio Titz , Philipp C. Böttcher , Michael T. Schaub , Sandra Venghaus , Dirk Witthaut","doi":"10.1016/j.adapen.2025.100242","DOIUrl":"10.1016/j.adapen.2025.100242","url":null,"abstract":"<div><div>Effective governance of energy system transformation away from fossil resources requires a quantitative understanding of the diffusion of green technologies and its key influencing factors. In this article, we propose a novel machine learning approach to diffusion research focusing on actual decisions and spatial aspects complementing research on intentions and temporal dynamics. We develop machine learning models that predict regional differences in the accumulated peak power of household-scale photovoltaic systems and the share of battery electric vehicles from a large set of demographic, geographic, political, and socio-economic features. Tools from explainable artificial intelligence enable a consistent identification of the key influencing factors and quantify their impact. Focusing on data from German municipal associations, we identify common themes and differences in the adoption of green technologies. Specifically, the adoption of battery electric vehicles is strongly associated with income and election results, while the adoption of photovoltaic systems correlates with the prevalence of large dwellings and levels of global solar radiation.</div></div>","PeriodicalId":34615,"journal":{"name":"Advances in Applied Energy","volume":"20 ","pages":"Article 100242"},"PeriodicalIF":13.8,"publicationDate":"2025-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145160159","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-12DOI: 10.1016/j.adapen.2025.100243
Apoorv Lal , Fengqi You
The rapid growth of artificial intelligence (AI) infrastructure deployment presents significant challenges for global energy systems and climate goals. While previous reviews address the sustainability of traditional data centers, Green AI approaches centered on model-level improvements or the application of AI in advancing sustainability across sectors, the energy and climate consequences of deploying AI infrastructure itself remain underexplored in prior literature. This paper reviews existing analyses on AI infrastructure’s energy and climate implications and proposes quantitative scenario-based frameworks, highlighting key research challenges at the intersection of AI-driven energy demand, region-specific clean energy strategies and their economic competitiveness, strategic levers in energy sourcing decisions, and policy dynamics. Additionally, this work identifies future research directions for aligning AI infrastructure growth with clean energy transitions through targeted mitigation opportunities across spatial and temporal horizons. First, the ambitious investment pathways for AI infrastructure development in the US underscore the need for spatially resolved scenario frameworks that reflect regional differences in deployment patterns and clean energy integration, along with the associated cost trajectories, to guide federal and state regulators. Second, the global expansion of AI infrastructure emphasizes the need for comprehensive frameworks that assess country-specific electricity demand shares, renewable transition pathways, and the influence of geopolitical restrictions, offering actionable insights for climate-conscious strategies. Finally, to prevent reinforcing fossil fuel dependency, particularly under disruptive growth scenarios, energy pathways incorporating nuclear power, renewables, energy storage, and varying grid reliance are explored as part of broader clean energy transitions, especially in regions facing energy security challenges.
{"title":"Advances and challenges in energy and climate alignment of AI infrastructure expansion","authors":"Apoorv Lal , Fengqi You","doi":"10.1016/j.adapen.2025.100243","DOIUrl":"10.1016/j.adapen.2025.100243","url":null,"abstract":"<div><div>The rapid growth of artificial intelligence (AI) infrastructure deployment presents significant challenges for global energy systems and climate goals. While previous reviews address the sustainability of traditional data centers, Green AI approaches centered on model-level improvements or the application of AI in advancing sustainability across sectors, the energy and climate consequences of deploying AI infrastructure itself remain underexplored in prior literature. This paper reviews existing analyses on AI infrastructure’s energy and climate implications and proposes quantitative scenario-based frameworks, highlighting key research challenges at the intersection of AI-driven energy demand, region-specific clean energy strategies and their economic competitiveness, strategic levers in energy sourcing decisions, and policy dynamics. Additionally, this work identifies future research directions for aligning AI infrastructure growth with clean energy transitions through targeted mitigation opportunities across spatial and temporal horizons. First, the ambitious investment pathways for AI infrastructure development in the US underscore the need for spatially resolved scenario frameworks that reflect regional differences in deployment patterns and clean energy integration, along with the associated cost trajectories, to guide federal and state regulators. Second, the global expansion of AI infrastructure emphasizes the need for comprehensive frameworks that assess country-specific electricity demand shares, renewable transition pathways, and the influence of geopolitical restrictions, offering actionable insights for climate-conscious strategies. Finally, to prevent reinforcing fossil fuel dependency, particularly under disruptive growth scenarios, energy pathways incorporating nuclear power, renewables, energy storage, and varying grid reliance are explored as part of broader clean energy transitions, especially in regions facing energy security challenges.</div></div>","PeriodicalId":34615,"journal":{"name":"Advances in Applied Energy","volume":"20 ","pages":"Article 100243"},"PeriodicalIF":13.8,"publicationDate":"2025-09-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145097826","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-10DOI: 10.1016/j.adapen.2025.100241
Ze Li , Jianheng Chen , Wenqi Wang , Yang Fu , Xin Li , Aiqiang Pan , Yiying Zhou , Shimelis Admassie , Chi Yan Tso
Radiative cooling (RC) coatings have emerged as a promising strategy to mitigate the urban heat island effect and improve energy performance in residential buildings. However, their effect varies significantly across different climate zones and urban configurations, underscoring the need for targeted deployment strategies. In this study, an ensemble learning framework was developed by integrating the urban canopy model with the building energy model to predict the energy performance of RC coatings on residential buildings throughout China. A dataset of 5080 cases was generated, and CatBoost demonstrated excellent predictive accuracy (R2 = 0.948–0.989). SHapley Additive exPlanations analysis identified longwave radiation and building geometry as the most influential factors affecting RC coating energy performance. The trained prediction model was further applied to evaluate six representative cities across diverse climate zones, for community-level evaluation. Additionally, national-scale predictions were conducted by the framework, using simulations of 111 cities, showing RC coatings are most effective in climate zones with hot summer and warm winter, with maximum annual electricity savings of approximately 50 MWh and maximum carbon emission reductions of around 20 kg·m-2 per year in a hypothetical residential neighborhood. In contrast, their benefits are more limited in cold climate zones due to increased heating demand. These findings provide an effective framework for optimizing RC coating deployment strategies under varying climatic conditions. Furthermore, the framework holds the potential to expand these analyses globally, enabling the evaluation of RC coatings across diverse building types and regions to support worldwide energy and carbon reduction goals.
{"title":"Ensemble learning framework for radiative cooling coatings in China’s buildings","authors":"Ze Li , Jianheng Chen , Wenqi Wang , Yang Fu , Xin Li , Aiqiang Pan , Yiying Zhou , Shimelis Admassie , Chi Yan Tso","doi":"10.1016/j.adapen.2025.100241","DOIUrl":"10.1016/j.adapen.2025.100241","url":null,"abstract":"<div><div>Radiative cooling (RC) coatings have emerged as a promising strategy to mitigate the urban heat island effect and improve energy performance in residential buildings. However, their effect varies significantly across different climate zones and urban configurations, underscoring the need for targeted deployment strategies. In this study, an ensemble learning framework was developed by integrating the urban canopy model with the building energy model to predict the energy performance of RC coatings on residential buildings throughout China. A dataset of 5080 cases was generated, and CatBoost demonstrated excellent predictive accuracy (R<sup>2</sup> = 0.948–0.989). SHapley Additive exPlanations analysis identified longwave radiation and building geometry as the most influential factors affecting RC coating energy performance. The trained prediction model was further applied to evaluate six representative cities across diverse climate zones, for community-level evaluation. Additionally, national-scale predictions were conducted by the framework, using simulations of 111 cities, showing RC coatings are most effective in climate zones with hot summer and warm winter, with maximum annual electricity savings of approximately 50 MWh and maximum carbon emission reductions of around 20 kg·m<sup>-2</sup> per year in a hypothetical residential neighborhood. In contrast, their benefits are more limited in cold climate zones due to increased heating demand. These findings provide an effective framework for optimizing RC coating deployment strategies under varying climatic conditions. Furthermore, the framework holds the potential to expand these analyses globally, enabling the evaluation of RC coatings across diverse building types and regions to support worldwide energy and carbon reduction goals.</div></div>","PeriodicalId":34615,"journal":{"name":"Advances in Applied Energy","volume":"20 ","pages":"Article 100241"},"PeriodicalIF":13.8,"publicationDate":"2025-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145057216","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.1016/j.adapen.2025.100239
Jiaqi Li , Hongbin Xie , Jingyuan Zhang , Lianxin Li , Ge Song , Hongdi Fu , Panxi Chen , Chenyang Liu , Liyu Zhang , Zhuoran Shi , Qing Yu , Xuan Song , Haoran Zhang
Building-Integrated Photovoltaic (BIPV), as an emerging clean energy solution, plays a crucial role in energy saving, emission reduction, and grid load regulation. However, due to the uncertainty of dynamic environments and the complexity of multiple sensitive parameters, traditional scheduling methods fail to achieve optimal results. Considering that reinforcement learning, as an advanced research approach, demonstrates great potential in decision-making for high-dimensional problems and stability in dynamic environments, integrating reinforcement learning with BIPV is a feasible solution to address scheduling challenges in BIPV systems. However, there is still a lack of comprehensive analysis and systematic understanding of reinforcement learning applications in the BIPV field, which, to some extent, limits its further development in the BIPV domain. To this end, this review conducts an in-depth analysis of the effectiveness of reinforcement learning in BIPV applications from the perspective of the system construction life cycle. By considering the algorithm modeling life cycle of reinforcement learning, it comprehensively examines the potential issues in its application to BIPV, highlighting the challenges faced by existing research and future applications. Additionally, this paper integrates cutting-edge reinforcement learning knowledge, summarizes and categorizes its potential applications in BIPV, providing reference guidance for future research directions. Through this systematic review of reinforcement learning applications in the BIPV field, this study aims to offer valuable insights for subsequent research.
{"title":"A systematic review of reinforcement learning in Building-Integrated Photovoltaic (BIPV) optimization","authors":"Jiaqi Li , Hongbin Xie , Jingyuan Zhang , Lianxin Li , Ge Song , Hongdi Fu , Panxi Chen , Chenyang Liu , Liyu Zhang , Zhuoran Shi , Qing Yu , Xuan Song , Haoran Zhang","doi":"10.1016/j.adapen.2025.100239","DOIUrl":"10.1016/j.adapen.2025.100239","url":null,"abstract":"<div><div>Building-Integrated Photovoltaic (BIPV), as an emerging clean energy solution, plays a crucial role in energy saving, emission reduction, and grid load regulation. However, due to the uncertainty of dynamic environments and the complexity of multiple sensitive parameters, traditional scheduling methods fail to achieve optimal results. Considering that reinforcement learning, as an advanced research approach, demonstrates great potential in decision-making for high-dimensional problems and stability in dynamic environments, integrating reinforcement learning with BIPV is a feasible solution to address scheduling challenges in BIPV systems. However, there is still a lack of comprehensive analysis and systematic understanding of reinforcement learning applications in the BIPV field, which, to some extent, limits its further development in the BIPV domain. To this end, this review conducts an in-depth analysis of the effectiveness of reinforcement learning in BIPV applications from the perspective of the system construction life cycle. By considering the algorithm modeling life cycle of reinforcement learning, it comprehensively examines the potential issues in its application to BIPV, highlighting the challenges faced by existing research and future applications. Additionally, this paper integrates cutting-edge reinforcement learning knowledge, summarizes and categorizes its potential applications in BIPV, providing reference guidance for future research directions. Through this systematic review of reinforcement learning applications in the BIPV field, this study aims to offer valuable insights for subsequent research.</div></div>","PeriodicalId":34615,"journal":{"name":"Advances in Applied Energy","volume":"19 ","pages":"Article 100239"},"PeriodicalIF":13.8,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145044367","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The widespread adoption of electric vehicles (EVs) creates opportunities to use EV charging load as a flexible resource to improve grid operation. In urban areas, EV users typically follow trip chains in their daily travel, offering temporal and spatial flexibility in EV charging. Specifically, charging at slow-charging spots at trip destinations is temporally flexible when the parking duration exceeds the required charging time. In contrast, charging at fast charging stations (FCSs) during trips is spatially flexible, with route and FCS choice influenced by traffic congestion, FCS charging prices, and user perception. In this paper, we propose a bi-criterion stochastic dynamic user equilibrium (SDUE) model with trip chain demand, which captures route and FCS choice of EV users and derives fast and slow charging loads. The model accounts for user response to traffic congestion and FCS charging prices, along with the randomness in user perception of trip utility. A quantitative evaluation is also presented on the spatial flexibility of fast charging driven by price incentives, and the temporal flexibility of slow charging enabled by long parking durations. A case study in Sioux Falls is conducted to evaluate the flexibility potential of EV charging, revealing that reduced randomness in user perception enhances the spatial flexibility potential of fast charging. Additionally, the temporal flexibility potential of slow charging varies across location types, such as home, work, and other locations, depending on arrival times and parking durations. This research provides key insights for optimizing grid management and enhancing EV integration into power systems.
{"title":"Flexibility potential of electric vehicle charging: A trip chain analysis under bi-criterion stochastic dynamic user equilibrium","authors":"Shuyi Tang, Yunfei Mu, Hongjie Jia, Xiaolong Jin, Xiaodan Yu","doi":"10.1016/j.adapen.2025.100240","DOIUrl":"10.1016/j.adapen.2025.100240","url":null,"abstract":"<div><div>The widespread adoption of electric vehicles (EVs) creates opportunities to use EV charging load as a flexible resource to improve grid operation. In urban areas, EV users typically follow trip chains in their daily travel, offering temporal and spatial flexibility in EV charging. Specifically, charging at slow-charging spots at trip destinations is temporally flexible when the parking duration exceeds the required charging time. In contrast, charging at fast charging stations (FCSs) during trips is spatially flexible, with route and FCS choice influenced by traffic congestion, FCS charging prices, and user perception. In this paper, we propose a bi-criterion stochastic dynamic user equilibrium (SDUE) model with trip chain demand, which captures route and FCS choice of EV users and derives fast and slow charging loads. The model accounts for user response to traffic congestion and FCS charging prices, along with the randomness in user perception of trip utility. A quantitative evaluation is also presented on the spatial flexibility of fast charging driven by price incentives, and the temporal flexibility of slow charging enabled by long parking durations. A case study in Sioux Falls is conducted to evaluate the flexibility potential of EV charging, revealing that reduced randomness in user perception enhances the spatial flexibility potential of fast charging. Additionally, the temporal flexibility potential of slow charging varies across location types, such as home, work, and other locations, depending on arrival times and parking durations. This research provides key insights for optimizing grid management and enhancing EV integration into power systems.</div></div>","PeriodicalId":34615,"journal":{"name":"Advances in Applied Energy","volume":"19 ","pages":"Article 100240"},"PeriodicalIF":13.8,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144921123","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.1016/j.adapen.2025.100228
Zhang Bai , Wenjie Hao , Qi Li , Rujing Yan , Bin Ding , Weiming Shao , Long Gao , Tieliu Jiang , Yongsheng Wang , Caifeng Wen
Wind-powered water electrolysis for hydrogen production is a sustainable and environmentally friendly energy technology. However, the inherent intermittency and variability of wind power, significantly damage the stability and efficiency of the hydrogen production system. To enhance the operational flexibility and system efficiency, a novel wind-hydrogen production system is proposed, which integrates a new coordination of the conventional alkaline electrolyzers (AEL) and proton exchange membrane electrolyzers (PEMEL), for optimizing the dynamic operation of the system under fluctuating wind power. The developed approach employs variational mode decomposition to classify wind power fluctuations into different frequency components, which are then allocated to suitable type of electrolyzers. The configurations of the developed system are optimized using the non-dominated sorting genetic algorithm, and the operating scenarios are dynamically analyzed through clustering techniques. Compared to the AEL-only system, the proposed system demonstrates significant enhancements, with energy efficiency and internal rate of return increased by 5.78 % and 10.65 %, respectively. Meanwhile, the coordinated operation extends the continuous operating time of the AEL by 7.08 %. The proposed approach enhances the economic viability and operational stability of wind-powered hydrogen production, providing a valuable reference for industrial green hydrogen applications.
{"title":"Enhancing flexibility in wind-powered hydrogen production systems through coordinated electrolyzer operation","authors":"Zhang Bai , Wenjie Hao , Qi Li , Rujing Yan , Bin Ding , Weiming Shao , Long Gao , Tieliu Jiang , Yongsheng Wang , Caifeng Wen","doi":"10.1016/j.adapen.2025.100228","DOIUrl":"10.1016/j.adapen.2025.100228","url":null,"abstract":"<div><div>Wind-powered water electrolysis for hydrogen production is a sustainable and environmentally friendly energy technology. However, the inherent intermittency and variability of wind power, significantly damage the stability and efficiency of the hydrogen production system. To enhance the operational flexibility and system efficiency, a novel wind-hydrogen production system is proposed, which integrates a new coordination of the conventional alkaline electrolyzers (AEL) and proton exchange membrane electrolyzers (PEMEL), for optimizing the dynamic operation of the system under fluctuating wind power. The developed approach employs variational mode decomposition to classify wind power fluctuations into different frequency components, which are then allocated to suitable type of electrolyzers. The configurations of the developed system are optimized using the non-dominated sorting genetic algorithm, and the operating scenarios are dynamically analyzed through clustering techniques. Compared to the AEL-only system, the proposed system demonstrates significant enhancements, with energy efficiency and internal rate of return increased by 5.78 % and 10.65 %, respectively. Meanwhile, the coordinated operation extends the continuous operating time of the AEL by 7.08 %. The proposed approach enhances the economic viability and operational stability of wind-powered hydrogen production, providing a valuable reference for industrial green hydrogen applications.</div></div>","PeriodicalId":34615,"journal":{"name":"Advances in Applied Energy","volume":"19 ","pages":"Article 100228"},"PeriodicalIF":13.8,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145105004","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-11DOI: 10.1016/j.adapen.2025.100237
Zixin Jiang, Xuezheng Wang, Bing Dong
Physics-informed machine learning (PIML) provides a promising solution for building energy modeling and can be used as a virtual environment to enable reinforcement learning (RL) agents to interact and learn. However, how to integrate physics priors efficiently, evaluate the effectiveness of physics constraints, balance model accuracy and physics consistency, and enable real-world implementation remain open challenges. To address these gaps, this study introduces a Physics-Informed Modularized Neural Network (PI-ModNN), which integrates physics priors through a physics-informed model structure, loss functions, and hard constraints. A new evaluation matrix called “temperature response violation” is developed to quantify the physical consistency of data-driven building dynamic models under varying control inputs and training data sizes. Additionally, a physics prior evaluation framework based on “rule importance” is proposed to quantify the contribution of each individual physical priors, offering guidance on selecting appropriate PIML techniques. The results indicate that incorporating physical priors does not always improve model performance; inappropriate physical priors could decrease model accuracy and consistency. However, hard constraints effectively enforce model consistency. Furthermore, we present a general workflow for developing control-oriented PIML models and integrating them with deep reinforcement learning (DRL). Following this framework, a case study of implementation DRL in an office space for three months demonstrates potential energy savings of 31.4%. Finally, we provide a general guideline for integrating data-driven models with advanced building control through a four-step evaluation framework, paving the way for reliable and scalable implementation of advanced building controls.
{"title":"Physics-informed modularized neural network for advanced building control by deep reinforcement learning","authors":"Zixin Jiang, Xuezheng Wang, Bing Dong","doi":"10.1016/j.adapen.2025.100237","DOIUrl":"10.1016/j.adapen.2025.100237","url":null,"abstract":"<div><div>Physics-informed machine learning (PIML) provides a promising solution for building energy modeling and can be used as a virtual environment to enable reinforcement learning (RL) agents to interact and learn. However, how to integrate physics priors efficiently, evaluate the effectiveness of physics constraints, balance model accuracy and physics consistency, and enable real-world implementation remain open challenges. To address these gaps, this study introduces a Physics-Informed Modularized Neural Network (PI-ModNN), which integrates physics priors through a physics-informed model structure, loss functions, and hard constraints. A new evaluation matrix called “temperature response violation” is developed to quantify the physical consistency of data-driven building dynamic models under varying control inputs and training data sizes. Additionally, a physics prior evaluation framework based on “rule importance” is proposed to quantify the contribution of each individual physical priors, offering guidance on selecting appropriate PIML techniques. The results indicate that incorporating physical priors does not always improve model performance; inappropriate physical priors could decrease model accuracy and consistency. However, hard constraints effectively enforce model consistency. Furthermore, we present a general workflow for developing control-oriented PIML models and integrating them with deep reinforcement learning (DRL). Following this framework, a case study of implementation DRL in an office space for three months demonstrates potential energy savings of 31.4%. Finally, we provide a general guideline for integrating data-driven models with advanced building control through a four-step evaluation framework, paving the way for reliable and scalable implementation of advanced building controls.</div></div>","PeriodicalId":34615,"journal":{"name":"Advances in Applied Energy","volume":"19 ","pages":"Article 100237"},"PeriodicalIF":13.8,"publicationDate":"2025-08-11","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144864331","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-08DOI: 10.1016/j.adapen.2025.100236
Han Li , William O’Brien , Vivian Loftness , Erica Cochran Hameen , Tianzhen Hong
Residential buildings consume a significant portion (17 % in 2023) of the global primary energy. Smart thermostat has become a proven technology in the residential building sector that offers insights into energy efficiency, HVAC system operation, and indoor thermal comfort of occupants. Although there are an increasing number of studies using the available large scale smart thermostat dataset, there lacks a holistic review of the existing literature to understand what applications have been conducted and what outcomes have been offered. This paper reviews 57 articles published between January 2015 and March 2025 using the open access ecobee Donate Your Data (DYD) dataset, where >200,000 customers participated in the voluntary data donation program. Articles are analyzed by major application areas including occupant behavior and IEQ assessment, energy performance evaluation, HVAC operations and controls, and building thermal dynamics. Two major limitations of the DYD dataset are the lack of measured energy use of HVAC systems and the coarse city-level building location information and limits applications requiring energy use data and introduces errors in ignoring the urban microclimate effects influencing a home’s operation and performance. Gaps and challenges of using the ecobee thermostat dataset for research were analyzed. Future efforts should focus on improving data collection and fusing other datasets with the ecobee DYD dataset to unlock new applications and improve analytics accuracy. Furthermore, AI emerges as a powerful tool to help clean up, integrate, and analyze the thermostat dataset, create and calibrate energy models, as well as inferring residential building operation and performance at scale.
{"title":"A critical review of use cases and insights from a large dataset of smart thermostats","authors":"Han Li , William O’Brien , Vivian Loftness , Erica Cochran Hameen , Tianzhen Hong","doi":"10.1016/j.adapen.2025.100236","DOIUrl":"10.1016/j.adapen.2025.100236","url":null,"abstract":"<div><div>Residential buildings consume a significant portion (17 % in 2023) of the global primary energy. Smart thermostat has become a proven technology in the residential building sector that offers insights into energy efficiency, HVAC system operation, and indoor thermal comfort of occupants. Although there are an increasing number of studies using the available large scale smart thermostat dataset, there lacks a holistic review of the existing literature to understand what applications have been conducted and what outcomes have been offered. This paper reviews 57 articles published between January 2015 and March 2025 using the open access ecobee Donate Your Data (DYD) dataset, where >200,000 customers participated in the voluntary data donation program. Articles are analyzed by major application areas including occupant behavior and IEQ assessment, energy performance evaluation, HVAC operations and controls, and building thermal dynamics. Two major limitations of the DYD dataset are the lack of measured energy use of HVAC systems and the coarse city-level building location information and limits applications requiring energy use data and introduces errors in ignoring the urban microclimate effects influencing a home’s operation and performance. Gaps and challenges of using the ecobee thermostat dataset for research were analyzed. Future efforts should focus on improving data collection and fusing other datasets with the ecobee DYD dataset to unlock new applications and improve analytics accuracy. Furthermore, AI emerges as a powerful tool to help clean up, integrate, and analyze the thermostat dataset, create and calibrate energy models, as well as inferring residential building operation and performance at scale.</div></div>","PeriodicalId":34615,"journal":{"name":"Advances in Applied Energy","volume":"19 ","pages":"Article 100236"},"PeriodicalIF":13.8,"publicationDate":"2025-08-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144842059","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-08-06DOI: 10.1016/j.adapen.2025.100235
Francesco De Marco, Jacob Mannhardt, Alfredo Oneto, Giovanni Sansavini
As the energy mix increasingly relies on weather-dependent renewable sources, energy systems become more vulnerable to climate variability and extremes. However, current planning approaches struggle to incorporate climate uncertainty in the design phase while maintaining computational tractability. We address this challenge by developing a framework that combines system-informed scenario reduction and stochastic optimization to design climate-resilient energy systems. Our method reduces data complexity by identifying representative climate scenarios that capture stress events through system response. Remarkably, five distinct patterns of multi-day energy shortages emerge across Europe, each characterized by different combinations of renewable resource availability and demand profiles. Stochastic optimization then incorporates these representative climate scenarios with their associated probabilities to design energy systems that are resilient across the full spectrum of climate variability. Results show that climate-resilient designs consistently outperform conventional single-climate designs, achieving lower costs (on average 14.8 bn EUR) for equivalent resilience levels. We identify two trade-off regions with different marginal costs of resilience: a low-resilience and a high-resilience region where marginal costs increase fivefold. Despite higher costs, trade-offs between the cost of resilience investments against energy not supplied justify pursuing the high levels of resilience. Combinations of onshore wind and hydrogen storage emerge as effective mitigation against multi-day events of energy shortage. This framework provides energy planners and policymakers with quantifiable insights into resilience investment strategies and technology selection for future climate-aware energy planning.
{"title":"Climate-resilient energy systems planning via system-informed identification of stressful events","authors":"Francesco De Marco, Jacob Mannhardt, Alfredo Oneto, Giovanni Sansavini","doi":"10.1016/j.adapen.2025.100235","DOIUrl":"10.1016/j.adapen.2025.100235","url":null,"abstract":"<div><div>As the energy mix increasingly relies on weather-dependent renewable sources, energy systems become more vulnerable to climate variability and extremes. However, current planning approaches struggle to incorporate climate uncertainty in the design phase while maintaining computational tractability. We address this challenge by developing a framework that combines system-informed scenario reduction and stochastic optimization to design climate-resilient energy systems. Our method reduces data complexity by identifying representative climate scenarios that capture stress events through system response. Remarkably, five distinct patterns of multi-day energy shortages emerge across Europe, each characterized by different combinations of renewable resource availability and demand profiles. Stochastic optimization then incorporates these representative climate scenarios with their associated probabilities to design energy systems that are resilient across the full spectrum of climate variability. Results show that climate-resilient designs consistently outperform conventional single-climate designs, achieving lower costs (on average 14.8 bn EUR) for equivalent resilience levels. We identify two trade-off regions with different marginal costs of resilience: a low-resilience and a high-resilience region where marginal costs increase fivefold. Despite higher costs, trade-offs between the cost of resilience investments against energy not supplied justify pursuing the high levels of resilience. Combinations of onshore wind and hydrogen storage emerge as effective mitigation against multi-day events of energy shortage. This framework provides energy planners and policymakers with quantifiable insights into resilience investment strategies and technology selection for future climate-aware energy planning.</div></div>","PeriodicalId":34615,"journal":{"name":"Advances in Applied Energy","volume":"19 ","pages":"Article 100235"},"PeriodicalIF":13.8,"publicationDate":"2025-08-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144864330","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-07-24DOI: 10.1016/j.adapen.2025.100233
Jonathan Vieth, Jan Westphal, Arne Speerforck
District heating networks play a critical role in the transition of the heating supply of buildings to renewable sources. The transition from coal-fired or gas-fired generation units to heat pumps requires new planning methods for district heating networks, since the efficiency of a heat pump is affected strongly by the supply temperature of the district heating network. Therefore, a co-planning approach including the operation of the district heating network in the planning process is required. This paper presents a novel co-planning approach consisting of two steps. First, an optimal district heating network topology is generated from real geo-referenced data. To determine the optimal topology, a new algorithm designed specifically for district heating networks is presented. Next, a simulation model is automatically generated from the respective topology. An optimization is used for the co-planning approach to select an optimal generation unit, find the optimal supply temperature, and dimension the pipes of the district heating network. In contrast to conventional district heating network planning procedures, the optimization includes a full-year dynamic simulation of the district heating network. The result of the planning process is a full y parameterized district heating network with a matching supply temperature. Furthermore, the use of simulation models allows the results to be reused for sensitivity analyses. This is illustrated by examining the selection of generation units under different price scenarios.
{"title":"District heating network topology optimization and optimal co-planning using dynamic simulations","authors":"Jonathan Vieth, Jan Westphal, Arne Speerforck","doi":"10.1016/j.adapen.2025.100233","DOIUrl":"10.1016/j.adapen.2025.100233","url":null,"abstract":"<div><div>District heating networks play a critical role in the transition of the heating supply of buildings to renewable sources. The transition from coal-fired or gas-fired generation units to heat pumps requires new planning methods for district heating networks, since the efficiency of a heat pump is affected strongly by the supply temperature of the district heating network. Therefore, a co-planning approach including the operation of the district heating network in the planning process is required. This paper presents a novel co-planning approach consisting of two steps. First, an optimal district heating network topology is generated from real geo-referenced data. To determine the optimal topology, a new algorithm designed specifically for district heating networks is presented. Next, a simulation model is automatically generated from the respective topology. An optimization is used for the co-planning approach to select an optimal generation unit, find the optimal supply temperature, and dimension the pipes of the district heating network. In contrast to conventional district heating network planning procedures, the optimization includes a full-year dynamic simulation of the district heating network. The result of the planning process is a full y parameterized district heating network with a matching supply temperature. Furthermore, the use of simulation models allows the results to be reused for sensitivity analyses. This is illustrated by examining the selection of generation units under different <span><math><msub><mrow><mi>CO</mi></mrow><mrow><mn>2</mn></mrow></msub></math></span> price scenarios.</div></div>","PeriodicalId":34615,"journal":{"name":"Advances in Applied Energy","volume":"19 ","pages":"Article 100233"},"PeriodicalIF":13.0,"publicationDate":"2025-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144711755","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号