Applied Energy最新文献_第2页

Application of transparent and semi-transparent photovoltaics in building windows: a review 透明与半透明光伏在建筑窗户中的应用综述

IF 11 1区工程技术 Q1 ENERGY & FUELS

Applied Energy

Pub Date : 2025-12-19 DOI: 10.1016/j.apenergy.2025.127264

Alibakhsh Kasaeian , Nastaran Zirak , Seyedmohammad Ghaziasgar , Mojtaba Akbari , Niloufar Fadaei , Kian Khazanedari , Nava Zarkhah , Sheida Khosravi Shahmirzadi , Fathollah Pourfayaz

This review paper investigates transparent and semi-transparent photovoltaics in building windows, such as the lightweight and flexible amorphous silicon solar cells, enabling easy integration into window designs. Next, we analyze PSCs, which have caught the attention of researchers for their impressive efficiency improvements and customizable optical features. Furthermore, we explore organic and polymeric solar cells, which provide benefits like lower production expenses and a wide range of visual options. These cells can be tailored to achieve different levels of transparency, making them appropriate for a range of architectural styles. We also investigate luminescent solar concentrators, which make use of specific luminescent materials to catch sunlight and guide it toward the edges of solar cells, improving energy collection while maintaining clear views. Dye-sensitized solar cells are known for their straightforward manufacturing processes and performance in dim conditions, while quantum dot solar cells are recognized for their ability to achieve high efficiencies using tailored light absorption strategies. Finally, we concentrate on cadmium telluride semi-transparent solar cells, emphasizing their affordability and proven track record in thin-film technology. Herein, we attempt to provide an in-depth evaluation with regards to the current state of semi-transparent photovoltaic technologies, and focus on their structure and performance. Furthermore, we discuss the challenges and benefits of the integration of these technologies into building designs. Semi-transparent photovoltaic windows show increasing promise as a sustainable solution to the growing global energy demand.

本文综述了透明和半透明光伏电池在建筑窗户中的应用，如轻质和柔性的非晶硅太阳能电池，使其易于集成到窗户设计中。接下来，我们分析了psc，它们因其令人印象深刻的效率提高和可定制的光学特性而引起了研究人员的注意。此外，我们还探索有机和聚合物太阳能电池，它们具有降低生产成本和广泛的视觉选择等优点。这些单元可以进行定制，以达到不同的透明度水平，使它们适合各种建筑风格。我们还研究了发光太阳能聚光器，它利用特定的发光材料捕捉阳光并将其引导到太阳能电池的边缘，在保持清晰视野的同时改善能量收集。染料敏化太阳能电池以其简单的制造工艺和在昏暗条件下的性能而闻名，而量子点太阳能电池则因其使用量身定制的光吸收策略实现高效率的能力而受到认可。最后，我们专注于碲化镉半透明太阳能电池，强调其可负担性和在薄膜技术方面的良好记录。在此，我们试图对半透明光伏技术的现状进行深入评估，并重点关注其结构和性能。此外，我们还讨论了将这些技术集成到建筑设计中的挑战和好处。半透明光伏窗作为一种可持续的解决方案，越来越有希望满足日益增长的全球能源需求。

{"title":"Application of transparent and semi-transparent photovoltaics in building windows: a review","authors":"Alibakhsh Kasaeian , Nastaran Zirak , Seyedmohammad Ghaziasgar , Mojtaba Akbari , Niloufar Fadaei , Kian Khazanedari , Nava Zarkhah , Sheida Khosravi Shahmirzadi , Fathollah Pourfayaz","doi":"10.1016/j.apenergy.2025.127264","DOIUrl":"10.1016/j.apenergy.2025.127264","url":null,"abstract":"<div><div>This review paper investigates transparent and semi-transparent photovoltaics in building windows, such as the lightweight and flexible amorphous silicon solar cells, enabling easy integration into window designs. Next, we analyze PSCs, which have caught the attention of researchers for their impressive efficiency improvements and customizable optical features. Furthermore, we explore organic and polymeric solar cells, which provide benefits like lower production expenses and a wide range of visual options. These cells can be tailored to achieve different levels of transparency, making them appropriate for a range of architectural styles. We also investigate luminescent solar concentrators, which make use of specific luminescent materials to catch sunlight and guide it toward the edges of solar cells, improving energy collection while maintaining clear views. Dye-sensitized solar cells are known for their straightforward manufacturing processes and performance in dim conditions, while quantum dot solar cells are recognized for their ability to achieve high efficiencies using tailored light absorption strategies. Finally, we concentrate on cadmium telluride semi-transparent solar cells, emphasizing their affordability and proven track record in thin-film technology. Herein, we attempt to provide an in-depth evaluation with regards to the current state of semi-transparent photovoltaic technologies, and focus on their structure and performance. Furthermore, we discuss the challenges and benefits of the integration of these technologies into building designs. Semi-transparent photovoltaic windows show increasing promise as a sustainable solution to the growing global energy demand.</div></div>","PeriodicalId":246,"journal":{"name":"Applied Energy","volume":"405 ","pages":"Article 127264"},"PeriodicalIF":11.0,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145788232","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

An integrated power, energy, and thermal management strategy using cascaded Control for off-road autonomous hybrid vehicles 采用级联控制的综合动力、能源和热管理策略，用于越野自动混合动力汽车

IF 11 1区工程技术 Q1 ENERGY & FUELS

Applied Energy

Pub Date : 2025-12-19 DOI: 10.1016/j.apenergy.2025.127269

Anirudh Sundar , Atharva Ghate , Qilun Zhu , Robert Prucka , Miriam Figueroa-Santos , Morgan Barron

Off-road hybrid vehicles operating in unstructured and thermally harsh environments require supervisory controllers capable of managing highly transient power demands while ensuring energy efficiency and component longevity. This paper presents a cascaded control strategy for integrated power, energy, and thermal management (IPETM) that addresses this multi-timescale challenge. The real-time framework combines a long-horizon optimizer that jointly minimizes fuel consumption and battery degradation with a fast compensatory controller that mitigates transient power fluctuations and uncertainties. Hardware-in-the-loop (HIL) testing using the actual vehicle control hardware and a high-fidelity virtual vehicle model demonstrates that the proposed approach maintains constraint-compliant operation under model and preview uncertainties. Benchmarking across nominal and extreme temperature off-road driving conditions shows that the IPETM strategy substantially reduces capacity degradation by ∼ 70 % relative to a real-time, long-horizon benchmark, without a significant increase in fuel consumption. Moreover, it achieves performance comparable to a synthesized short-update long-horizon controller that is computationally infeasible for real-time implementation and requires detailed future-demand previews. Sensitivity studies on control weights and update frequency further establish practical configuration guidelines. Overall, the results demonstrate that the proposed IPETM framework bridges the gap between real-time implementable and ideal optimization-based controllers, providing a computationally tractable and robust solution for integrated power, energy, and thermal management in autonomous hybrid off-road vehicles.

在非结构化和热恶劣环境中运行的越野车需要能够管理高瞬时功率需求的监控控制器，同时确保能源效率和组件寿命。本文提出了一种用于综合电力、能源和热管理（IPETM）的级联控制策略，以解决这一多时间尺度的挑战。实时框架结合了长期优化器和快速补偿控制器，前者可最大限度地减少燃料消耗和电池退化，后者可减轻瞬态功率波动和不确定性。利用实际车辆控制硬件和高保真虚拟车辆模型进行的硬件在环（HIL）测试表明，该方法在模型和预览不确定的情况下保持约束遵从性。在标称温度和极端温度的越野驾驶条件下进行的基准测试表明，相对于实时、长期基准测试，IPETM策略在不显著增加燃油消耗的情况下，大大减少了约70%的容量退化。此外，它的性能可与综合的短更新长视界控制器相媲美，后者在计算上不适合实时实现，并且需要详细的未来需求预览。对控制权值和更新频率的敏感性研究进一步建立了实用的配置指南。总体而言，结果表明，所提出的IPETM框架弥合了实时可实现控制器和理想优化控制器之间的差距，为自动混合动力越野车的集成电源、能源和热管理提供了计算上易于处理和强大的解决方案。

{"title":"An integrated power, energy, and thermal management strategy using cascaded Control for off-road autonomous hybrid vehicles","authors":"Anirudh Sundar , Atharva Ghate , Qilun Zhu , Robert Prucka , Miriam Figueroa-Santos , Morgan Barron","doi":"10.1016/j.apenergy.2025.127269","DOIUrl":"10.1016/j.apenergy.2025.127269","url":null,"abstract":"<div><div>Off-road hybrid vehicles operating in unstructured and thermally harsh environments require supervisory controllers capable of managing highly transient power demands while ensuring energy efficiency and component longevity. This paper presents a cascaded control strategy for integrated power, energy, and thermal management (IPETM) that addresses this multi-timescale challenge. The real-time framework combines a long-horizon optimizer that jointly minimizes fuel consumption and battery degradation with a fast compensatory controller that mitigates transient power fluctuations and uncertainties. Hardware-in-the-loop (HIL) testing using the actual vehicle control hardware and a high-fidelity virtual vehicle model demonstrates that the proposed approach maintains constraint-compliant operation under model and preview uncertainties. Benchmarking across nominal and extreme temperature off-road driving conditions shows that the IPETM strategy substantially reduces capacity degradation by ∼ 70 % relative to a real-time, long-horizon benchmark, without a significant increase in fuel consumption. Moreover, it achieves performance comparable to a synthesized short-update long-horizon controller that is computationally infeasible for real-time implementation and requires detailed future-demand previews. Sensitivity studies on control weights and update frequency further establish practical configuration guidelines. Overall, the results demonstrate that the proposed IPETM framework bridges the gap between real-time implementable and ideal optimization-based controllers, providing a computationally tractable and robust solution for integrated power, energy, and thermal management in autonomous hybrid off-road vehicles.</div></div>","PeriodicalId":246,"journal":{"name":"Applied Energy","volume":"406 ","pages":"Article 127269"},"PeriodicalIF":11.0,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145789781","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

SMART-EV: a stochastic macroscopic DRL-based method for enhancing distribution network resilience via EV coordination 智能电动汽车：一种基于随机宏观drl的电动汽车协调增强配电网弹性的方法

IF 11 1区工程技术 Q1 ENERGY & FUELS

Applied Energy

Pub Date : 2025-12-19 DOI: 10.1016/j.apenergy.2025.127218

Xianglong Lian , Lei Qiu , Weiming Liu , Yiqing Jiang , Chenkai Song , Lijun Liu , Wenhu Tang

The increasing penetration of electric vehicles (EVs) provides unprecedented flexibility for enhancing the resilience of distribution networks under extreme events. However, the inherent stochasticity in EV mobility and the need for real-time coordination with power system operations present significant challenges for disaster restoration. To address these challenges, this paper proposes a novel stochastic macroscopic adaptive resilience-trained EV (SMART-EV) method. Firstly, a stochastic macroscopic mobility model is developed to describe the random movement of EVs in a two-dimensional plane, enabling more realistic yet tractable modeling of large-scale fleet dynamics. Subsequently, the proposed SMART-EV achieves adaptive learning-driven physical systems by integrating deep reinforcement learning (DRL)-based decision-making of EV mobility and physical-layer grid operation. Finally, an improved deep deterministic policy gradient (DDPG) algorithm is introduced, which explicitly optimizes grid resilience objectives, accelerates convergence, and enhances training stability through auxiliary Nash-style (AuxiNash) multi-objective regularization. Case studies in IEEE 33-bus system demonstrate that the proposed SMART-EV method achieves in a 17.9 % increase in the resilience index and substantial reductions in computational effort.

电动汽车（ev）的日益普及为提高配电网在极端事件下的应变能力提供了前所未有的灵活性。然而，电动汽车移动性固有的随机性以及与电力系统运行实时协调的需求给灾难恢复带来了重大挑战。为了解决这些问题，本文提出了一种新的随机宏观自适应弹性训练EV （SMART-EV）方法。首先，建立了一个随机宏观机动模型，在二维平面上描述电动汽车的随机运动，使大规模车队动力学建模更加真实和易于处理。随后，本文提出的SMART-EV通过将基于深度强化学习（DRL）的电动汽车移动决策与物理层网格运行相结合，实现自适应学习驱动的物理系统。最后，介绍了一种改进的深度确定性策略梯度（DDPG）算法，该算法通过辅助纳什风格（AuxiNash）多目标正则化，明确优化网格弹性目标，加速收敛，增强训练稳定性。在IEEE 33总线系统中的案例研究表明，提出的SMART-EV方法实现了17.9%的弹性指数提高和计算量的大幅减少。

{"title":"SMART-EV: a stochastic macroscopic DRL-based method for enhancing distribution network resilience via EV coordination","authors":"Xianglong Lian , Lei Qiu , Weiming Liu , Yiqing Jiang , Chenkai Song , Lijun Liu , Wenhu Tang","doi":"10.1016/j.apenergy.2025.127218","DOIUrl":"10.1016/j.apenergy.2025.127218","url":null,"abstract":"<div><div>The increasing penetration of electric vehicles (EVs) provides unprecedented flexibility for enhancing the resilience of distribution networks under extreme events. However, the inherent stochasticity in EV mobility and the need for real-time coordination with power system operations present significant challenges for disaster restoration. To address these challenges, this paper proposes a novel stochastic macroscopic adaptive resilience-trained EV (SMART-EV) method. Firstly, a stochastic macroscopic mobility model is developed to describe the random movement of EVs in a two-dimensional plane, enabling more realistic yet tractable modeling of large-scale fleet dynamics. Subsequently, the proposed SMART-EV achieves adaptive learning-driven physical systems by integrating deep reinforcement learning (DRL)-based decision-making of EV mobility and physical-layer grid operation. Finally, an improved deep deterministic policy gradient (DDPG) algorithm is introduced, which explicitly optimizes grid resilience objectives, accelerates convergence, and enhances training stability through auxiliary Nash-style (AuxiNash) multi-objective regularization. Case studies in IEEE 33-bus system demonstrate that the proposed SMART-EV method achieves in a 17.9 % increase in the resilience index and substantial reductions in computational effort.</div></div>","PeriodicalId":246,"journal":{"name":"Applied Energy","volume":"406 ","pages":"Article 127218"},"PeriodicalIF":11.0,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145789782","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Generative AI impact assessment through a life cycle analysis of multiple data center typologies 通过对多个数据中心类型的生命周期分析进行生成式人工智能影响评估

IF 11 1区工程技术 Q1 ENERGY & FUELS

Applied Energy

Pub Date : 2025-12-19 DOI: 10.1016/j.apenergy.2025.127288

Alexandre d'Orgeval , Stuart Sheehan , Quentin Avenas , Edi Assoumou , Valentina Sessa

Data centers are energy-intensive infrastructures that generate, manage, and store information for our interconnected society. Models based on Artificial Intelligence (AI), such as ChatGPT, are increasingly accessible, leading to significant energy consumption and associated carbon emissions.

Assessing the environmental footprint of Generative AI (GenAI) is essential for evaluating its sustainability and promoting responsible AI development. In this work, a comprehensive environmental assessment of GenAI systems was performed – which includes both training and inference phases – using a life cycle assessment (LCA) approach. Prior studies have primarily focused on server-level assessments or energy consumption analyses. In contrast, this work considers the full lifecycle of data centers and evaluates environmental impacts across complete architectural configurations, offering a broader and more integrated perspective. Finally, multiple data center architectures are compared, from edge systems to AI dedicated infrastructures.

Two simulation-based use cases are presented: (1) A 20-year simulation comparing different data center architectures across three indicators – total emissions, emissions per year, and emissions per installed IT MW. For a subset of these architectures, emissions per Floating-Point Operations Per Second (FLOPS) are also included to assess performance efficiency – considering that FLOPS estimations can only be done on GPU based data center architectures; (2) A focused simulation comparing the environmental footprint of three large language models – GPT-4o, LLaMA 3.1405B, and DeepSeek V3 – to quantify trade-offs between benchmark performance and environmental impact. By expanding the scope of assessment and incorporating varied use cases, this work aims to inform strategies for minimizing the environmental costs of GenAI while advancing sustainable AI development.

数据中心是能源密集型的基础设施，为我们互联的社会生成、管理和存储信息。基于人工智能（AI）的模型，如ChatGPT，越来越容易获得，导致大量的能源消耗和相关的碳排放。评估生成式人工智能（GenAI）的环境足迹对于评估其可持续性和促进负责任的人工智能发展至关重要。在这项工作中，使用生命周期评估（LCA）方法对GenAI系统进行了全面的环境评估，包括训练和推理阶段。以前的研究主要集中在服务器级评估或能耗分析上。相比之下，这项工作考虑了数据中心的整个生命周期，并评估了整个体系结构配置对环境的影响，提供了更广泛、更集成的视角。最后，比较了多个数据中心架构，从边缘系统到人工智能专用基础设施。提出了两个基于模拟的用例：(1)20年的模拟，通过三个指标（总排放量、每年排放量和每安装IT兆瓦的排放量）比较不同的数据中心架构。对于这些架构的一个子集，还包括每秒浮点运算（FLOPS）的排放量来评估性能效率-考虑到FLOPS估计只能在基于GPU的数据中心架构上完成；(2)对gpt - 40、LLaMA 3.1405B和DeepSeek V3三种大型语言模型的环境足迹进行了重点仿真，以量化基准性能和环境影响之间的权衡。通过扩大评估范围和纳入各种用例，这项工作旨在为在推进可持续人工智能发展的同时最大限度地降低GenAI的环境成本的战略提供信息。

{"title":"Generative AI impact assessment through a life cycle analysis of multiple data center typologies","authors":"Alexandre d'Orgeval , Stuart Sheehan , Quentin Avenas , Edi Assoumou , Valentina Sessa","doi":"10.1016/j.apenergy.2025.127288","DOIUrl":"10.1016/j.apenergy.2025.127288","url":null,"abstract":"<div><div>Data centers are energy-intensive infrastructures that generate, manage, and store information for our interconnected society. Models based on Artificial Intelligence (AI), such as ChatGPT, are increasingly accessible, leading to significant energy consumption and associated carbon emissions.</div><div>Assessing the environmental footprint of Generative AI (GenAI) is essential for evaluating its sustainability and promoting responsible AI development. In this work, a comprehensive environmental assessment of GenAI systems was performed – which includes both training and inference phases – using a life cycle assessment (LCA) approach. Prior studies have primarily focused on server-level assessments or energy consumption analyses. In contrast, this work considers the full lifecycle of data centers and evaluates environmental impacts across complete architectural configurations, offering a broader and more integrated perspective. Finally, multiple data center architectures are compared, from edge systems to AI dedicated infrastructures.</div><div>Two simulation-based use cases are presented: (1) A 20-year simulation comparing different data center architectures across three indicators – total emissions, emissions per year, and emissions per installed IT MW. For a subset of these architectures, emissions per Floating-Point Operations Per Second (FLOPS) are also included to assess performance efficiency – considering that FLOPS estimations can only be done on GPU based data center architectures; (2) A focused simulation comparing the environmental footprint of three large language models – GPT-4o, LLaMA 3.1405B, and DeepSeek V3 – to quantify trade-offs between benchmark performance and environmental impact. By expanding the scope of assessment and incorporating varied use cases, this work aims to inform strategies for minimizing the environmental costs of GenAI while advancing sustainable AI development.</div></div>","PeriodicalId":246,"journal":{"name":"Applied Energy","volume":"406 ","pages":"Article 127288"},"PeriodicalIF":11.0,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145789780","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Transforming adverse effect into beneficial outcome: A design strategy leveraging non-uniform flow for enhanced performance of PEM fuel cells 将不利影响转化为有利结果：利用非均匀流动来提高PEM燃料电池性能的设计策略

IF 11 1区工程技术 Q1 ENERGY & FUELS

Applied Energy

Pub Date : 2025-12-19 DOI: 10.1016/j.apenergy.2025.127224

Bin Wang , Weitong Pan , Xinming Tian , Longfei Tang , Xueli Chen , Fuchen Wang

Non-uniform flow in Proton Exchange Membrane (PEM) fuel cells with parallel flow fields can enhance performance, whereas a targeted regulation strategy has not been established. Hence, a systematic numerical investigation is conducted in this work. First, the impacts of channel dimensions on flow uniformity and cell performance are examined, accompanied by a sensitivity analysis. The core lies in the transverse cross-rib flow in the porous medium induced by non-uniform flow distribution. Building on this, a design strategy combining channel refinement and active area scale-up is proposed to facilitate the capability accumulation and release of this flow pattern, with the expectation of mitigating the inherently restricted reactant transport beneath ribs and thereby strengthening the electrochemical reaction. Second, this strategy is applied in the Manifold configuration. The current density improves from 1.49 % below to 13.49 % above that in the Uniform configuration. Flow non-uniformity rises, and the transverse cross-rib flow intensity increases markedly from 1.99 to 1078.85. The adverse effect of non-uniform flow is successfully transformed into a beneficial outcome. Third, the applicability of the proposed strategy is validated under different operational and geometric conditions, with the Manifold configuration consistently outperforming the Uniform configuration by more than 10 %. Furthermore, system efficiency and SWOT analyses are conducted to assess its prospects.

平行流场下质子交换膜（PEM）燃料电池的非均匀流动可以提高电池性能，但目前尚未建立有针对性的调节策略。因此，在这项工作中进行了系统的数值研究。首先，研究了通道尺寸对流动均匀性和电池性能的影响，并进行了灵敏度分析。其核心在于多孔介质中由非均匀流动引起的横向横肋流动。在此基础上，提出了一种结合通道优化和活性区域放大的设计策略，以促进该流型的能力积累和释放，期望减轻肋下固有的限制性反应物传输，从而加强电化学反应。其次，该策略应用于歧管配置。电流密度由均匀结构下的1.49%提高到均匀结构下的13.49%。流动不均匀性提高，横向横肋流动强度从1.99显著增加到1078.85。不均匀流动的不利影响成功地转化为有利的结果。第三，在不同的操作和几何条件下验证了所提出策略的适用性，流形配置始终优于均匀配置10%以上。并通过系统效率分析和SWOT分析对其前景进行了评估。

{"title":"Transforming adverse effect into beneficial outcome: A design strategy leveraging non-uniform flow for enhanced performance of PEM fuel cells","authors":"Bin Wang , Weitong Pan , Xinming Tian , Longfei Tang , Xueli Chen , Fuchen Wang","doi":"10.1016/j.apenergy.2025.127224","DOIUrl":"10.1016/j.apenergy.2025.127224","url":null,"abstract":"<div><div>Non-uniform flow in Proton Exchange Membrane (PEM) fuel cells with parallel flow fields can enhance performance, whereas a targeted regulation strategy has not been established. Hence, a systematic numerical investigation is conducted in this work. First, the impacts of channel dimensions on flow uniformity and cell performance are examined, accompanied by a sensitivity analysis. The core lies in the transverse cross-rib flow in the porous medium induced by non-uniform flow distribution. Building on this, a design strategy combining channel refinement and active area scale-up is proposed to facilitate the capability accumulation and release of this flow pattern, with the expectation of mitigating the inherently restricted reactant transport beneath ribs and thereby strengthening the electrochemical reaction. Second, this strategy is applied in the Manifold configuration. The current density improves from 1.49 % below to 13.49 % above that in the Uniform configuration. Flow non-uniformity rises, and the transverse cross-rib flow intensity increases markedly from 1.99 to 1078.85. The adverse effect of non-uniform flow is successfully transformed into a beneficial outcome. Third, the applicability of the proposed strategy is validated under different operational and geometric conditions, with the Manifold configuration consistently outperforming the Uniform configuration by more than 10 %. Furthermore, system efficiency and SWOT analyses are conducted to assess its prospects.</div></div>","PeriodicalId":246,"journal":{"name":"Applied Energy","volume":"406 ","pages":"Article 127224"},"PeriodicalIF":11.0,"publicationDate":"2025-12-19","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145789414","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Physics-informed transfer learning by embedding physics into activation functions: an application in battery health management 通过将物理嵌入激活函数中的物理信息迁移学习：在电池健康管理中的应用

IF 11 1区工程技术 Q1 ENERGY & FUELS

Applied Energy

Pub Date : 2025-12-18 DOI: 10.1016/j.apenergy.2025.127161

Hung Le , Weikun Deng , Khanh T.P. Nguyen , Kamal Medjaher , Christian Gogu , Dazhong Wu

Accurate prediction of battery remaining useful life (RUL) is critical for ensuring battery safety and reliability. Although physics-informed (PI) machine learning models embed degradation mechanisms to improve accuracy, they often require system-specific knowledge, limiting cross-domain generalization. Transfer learning (TL) enables adaptation across datasets, however, it does not leverage prior physical knowledge or established physics-based models. To combine their strengths, we introduce a dual-branch physics-informed transfer learning framework, where the data-driven branch is pre-trained on the Stanford-MIT-Toyota dataset (source domain) and fine-tuned on the XJTU dataset (target domain), while the PI branch incorporates a solid electrolyte interphase degradation mechanism via an Arrhenius-based activation function. Both branches are co-trained on the target domain to demonstrate the improvement in prediction accuracy. The physics-informed transfer learning (PITL) framework consistently improves prediction accuracy across all tested charge-discharge protocols, achieving the lowest mean absolute percentage error (MAPE) of 9.09 %, compared with 10.53 % for TL model and 14.58 % for the baseline model trained from scratch. The PITL model also achieves the lowest MAPE of 7.03 % in the case of individual-battery prediction. A comparative study shows that replacing the standard activation functions with the Arrhenius-based activation function improves generalization and predictive performance by embedding physics into transfer learning.

准确预测电池剩余使用寿命（RUL）对于保证电池的安全性和可靠性至关重要。虽然物理信息（PI）机器学习模型嵌入退化机制以提高准确性，但它们通常需要系统特定的知识，限制了跨领域的泛化。迁移学习（TL）能够跨数据集进行适应，但是，它不利用先前的物理知识或已建立的基于物理的模型。为了结合它们的优势，我们引入了一个双分支物理迁移学习框架，其中数据驱动分支在斯坦福-麻省理工学院-丰田数据集（源域）上进行预训练，并在XJTU数据集（目标域）上进行微调，而PI分支通过基于arrhenius的激活函数结合了固体电解质间相降解机制。两个分支在目标域上进行了共同训练，以证明预测精度的提高。物理迁移学习（PITL）框架在所有测试的充放电协议中持续提高预测精度，实现最低的平均绝对百分比误差（MAPE）为9.09%，而TL模型为10.53%，从头开始训练的基线模型为14.58%。在单个电池预测的情况下，PITL模型也达到了7.03%的最低MAPE。一项比较研究表明，用基于arrhenius的激活函数代替标准激活函数，通过将物理嵌入到迁移学习中，提高了泛化和预测性能。

{"title":"Physics-informed transfer learning by embedding physics into activation functions: an application in battery health management","authors":"Hung Le , Weikun Deng , Khanh T.P. Nguyen , Kamal Medjaher , Christian Gogu , Dazhong Wu","doi":"10.1016/j.apenergy.2025.127161","DOIUrl":"10.1016/j.apenergy.2025.127161","url":null,"abstract":"<div><div>Accurate prediction of battery remaining useful life (RUL) is critical for ensuring battery safety and reliability. Although physics-informed (PI) machine learning models embed degradation mechanisms to improve accuracy, they often require system-specific knowledge, limiting cross-domain generalization. Transfer learning (TL) enables adaptation across datasets, however, it does not leverage prior physical knowledge or established physics-based models. To combine their strengths, we introduce a dual-branch physics-informed transfer learning framework, where the data-driven branch is pre-trained on the Stanford-MIT-Toyota dataset (source domain) and fine-tuned on the XJTU dataset (target domain), while the PI branch incorporates a solid electrolyte interphase degradation mechanism via an Arrhenius-based activation function. Both branches are co-trained on the target domain to demonstrate the improvement in prediction accuracy. The physics-informed transfer learning (PITL) framework consistently improves prediction accuracy across all tested charge-discharge protocols, achieving the lowest mean absolute percentage error (MAPE) of 9.09 %, compared with 10.53 % for TL model and 14.58 % for the baseline model trained from scratch. The PITL model also achieves the lowest MAPE of 7.03 % in the case of individual-battery prediction. A comparative study shows that replacing the standard activation functions with the Arrhenius-based activation function improves generalization and predictive performance by embedding physics into transfer learning.</div></div>","PeriodicalId":246,"journal":{"name":"Applied Energy","volume":"406 ","pages":"Article 127161"},"PeriodicalIF":11.0,"publicationDate":"2025-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145765877","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Attention-enhanced residual networks for real-time multi-label power quality disturbance classification with fast iterative filtering 基于快速迭代滤波的关注增强残差网络实时多标签电能质量扰动分类

IF 11 1区工程技术 Q1 ENERGY & FUELS

Applied Energy

Pub Date : 2025-12-18 DOI: 10.1016/j.apenergy.2025.127233

Han Shao , Rui Henriques , Hugo Morais , Elisabetta Tedeschi

The growing integration of renewable energy resources into modern power grids heightens their sensitivity to various types and degrees of perturbations, highlighting the relevance of real-time and comprehensive monitoring frameworks for power quality disturbances (PQDs). Notwithstanding, traditional approaches often struggle to identify complex PQD signals, particularly in noisy environments. This work proposes a lightweight classification framework that leverages fast iterative filtering (FIF), a multi-scale 1D residual network (ResNet), and efficient channel attention (ECA) for superior predictive performance. First, FIF with an adaptive mask length is employed to decompose PQD signals, effectively capturing dynamic perturbation characteristics. The B-spline filter is integrated to generate a non-negative and compact window, mitigating mode mixing while improving spectral resolution. Second, a deep neural network with parallel ResNet blocks for leveraging multi-scale receptive fields is designed to enhance feature discrimination by capturing both low-level and high-level signal patterns. Third, ECA modules are incorporated to adaptively reweight feature channels, minimizing redundancy and emphasizing disturbance-related patterns. Lastly, a multi-label-aware architecture is introduced to handle noisy and overlapping PQDs. Extensive experiments on synthetic datasets show that the proposed framework achieves superior accuracy, robustness, and real-time performance compared with state-of-the-art methods. Validation on two real-world PQD datasets further demonstrates its effectiveness, reaching accuracies of 97.94 % and 98.78 % with average inference times of 6.12 ms and 5.34 ms per sample, respectively. To support further research, the benchmark datasets and trained models are made publicly available.

可再生能源日益融入现代电网，提高了其对各种类型和程度的扰动的敏感性，突出了电能质量扰动（PQDs）的实时和全面监测框架的相关性。尽管如此，传统的方法往往难以识别复杂的PQD信号，特别是在嘈杂的环境中。本研究提出了一种轻量级分类框架，该框架利用快速迭代滤波（FIF）、多尺度一维残差网络（ResNet）和高效通道关注（ECA）来实现卓越的预测性能。首先，采用自适应掩模长度的FIF对PQD信号进行分解，有效捕获动态摄动特征；b样条滤波器集成，以产生非负和紧凑的窗口，减轻模式混合，同时提高光谱分辨率。其次，设计了一个具有并行ResNet块的深度神经网络，用于利用多尺度接受域，通过捕获低水平和高水平信号模式来增强特征识别。第三，ECA模块被纳入自适应地重新加权特征通道，最小化冗余并强调与干扰相关的模式。最后，介绍了一种多标签感知结构来处理噪声和重叠的pqd。在合成数据集上的大量实验表明，与最先进的方法相比，所提出的框架具有更高的准确性、鲁棒性和实时性。在两个真实PQD数据集上的验证进一步证明了其有效性，准确率分别达到97.94%和98.78%，平均推理时间分别为每个样本6.12 ms和5.34 ms。为了支持进一步的研究，基准数据集和训练模型都是公开的。

{"title":"Attention-enhanced residual networks for real-time multi-label power quality disturbance classification with fast iterative filtering","authors":"Han Shao , Rui Henriques , Hugo Morais , Elisabetta Tedeschi","doi":"10.1016/j.apenergy.2025.127233","DOIUrl":"10.1016/j.apenergy.2025.127233","url":null,"abstract":"<div><div>The growing integration of renewable energy resources into modern power grids heightens their sensitivity to various types and degrees of perturbations, highlighting the relevance of real-time and comprehensive monitoring frameworks for power quality disturbances (PQDs). Notwithstanding, traditional approaches often struggle to identify complex PQD signals, particularly in noisy environments. This work proposes a lightweight classification framework that leverages fast iterative filtering (FIF), a multi-scale 1D residual network (ResNet), and efficient channel attention (ECA) for superior predictive performance. First, FIF with an adaptive mask length is employed to decompose PQD signals, effectively capturing dynamic perturbation characteristics. The B-spline filter is integrated to generate a non-negative and compact window, mitigating mode mixing while improving spectral resolution. Second, a deep neural network with parallel ResNet blocks for leveraging multi-scale receptive fields is designed to enhance feature discrimination by capturing both low-level and high-level signal patterns. Third, ECA modules are incorporated to adaptively reweight feature channels, minimizing redundancy and emphasizing disturbance-related patterns. Lastly, a multi-label-aware architecture is introduced to handle noisy and overlapping PQDs. Extensive experiments on synthetic datasets show that the proposed framework achieves superior accuracy, robustness, and real-time performance compared with state-of-the-art methods. Validation on two real-world PQD datasets further demonstrates its effectiveness, reaching accuracies of 97.94 % and 98.78 % with average inference times of 6.12 ms and 5.34 ms per sample, respectively. To support further research, the benchmark datasets and trained models are made publicly available.</div></div>","PeriodicalId":246,"journal":{"name":"Applied Energy","volume":"406 ","pages":"Article 127233"},"PeriodicalIF":11.0,"publicationDate":"2025-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145789415","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

New distributionally robust optimization framework and algorithm for energy hub scheduling integrating demand response programs under ambiguous prices and loads 电价和负荷模糊情况下能源枢纽调度的分布式鲁棒优化框架和算法

IF 11 1区工程技术 Q1 ENERGY & FUELS

Applied Energy

Pub Date : 2025-12-17 DOI: 10.1016/j.apenergy.2025.127211

Shihai Liu , Ruofei Yang , Ying Liu

With the continuous rising of global energy demand, the highly efficient energy hub scheduling (EHS) is extremely crucial to enhance system resilience and improve energy efficiency. This paper studies an EHS problem with multi-energy inputs and multi-energy outputs. The energy hub combines storage solutions for electricity and hydrogen, while also integrating demand response programs (DRPs) for both electricity and heat consumption. In reality, energy prices and loads fluctuate in real time, and the estimation of accurate probability distributions is a huge challenge. Thus, we model the partial availability of distribution information for electricity, hydrogen, and natural gas prices, as well as electricity, natural gas, and heat loads. A globalized distributionally robust energy hub scheduling (GDR-EHS) model is introduced to enhance the stability and adaptability of energy supply by creating outer and inner ambiguity sets. Lagrange duality and strong duality theory are used to obtain a computationally tractable formulation, then the proposed model is reformulated into a mixed integer linear programming (MILP) model. To enhance computational efficiency, we design an accelerated Branch and Cut (B&C) algorithm with Gomory cuts. Finally, the numerical results indicate that incorporating energy storage systems leads to a 4.39 % reduction in costs, while the addition of DRPs further lowers costs by 4.35 %. Furthermore, compared to Gurobi and the standard B&C algorithm, the accelerated B&C algorithm enhances computational efficiency by 50.1 % and 57.5 %.

随着全球能源需求的不断增长，高效的能源枢纽调度（EHS）对于增强系统弹性和提高能源效率至关重要。本文研究了一个具有多能量输入和多能量输出的EHS问题。能源中心结合了电力和氢气的存储解决方案，同时还集成了电力和热量消耗的需求响应程序（DRPs）。在现实中，能源价格和负荷是实时波动的，准确估计概率分布是一个巨大的挑战。因此，我们对电力、氢气和天然气价格以及电力、天然气和热负荷的部分可用性分布信息进行建模。提出了一种全球化分布式鲁棒能源枢纽调度（GDR-EHS）模型，通过创建外部和内部模糊集来提高能源供应的稳定性和适应性。利用拉格朗日对偶和强对偶理论得到了一个计算易于处理的模型，然后将该模型重新表述为混合整数线性规划（MILP）模型。为了提高计算效率，我们设计了一种带有Gomory切割的加速分支与切割（B&；C）算法。最后，数值结果表明，加入储能系统导致成本降低4.39%，而加入drp进一步降低了4.35%的成本。此外，与Gurobi和标准的B&；C算法相比，加速的B&；C算法的计算效率分别提高了50.1%和57.5%。

{"title":"New distributionally robust optimization framework and algorithm for energy hub scheduling integrating demand response programs under ambiguous prices and loads","authors":"Shihai Liu , Ruofei Yang , Ying Liu","doi":"10.1016/j.apenergy.2025.127211","DOIUrl":"10.1016/j.apenergy.2025.127211","url":null,"abstract":"<div><div>With the continuous rising of global energy demand, the highly efficient energy hub scheduling (EHS) is extremely crucial to enhance system resilience and improve energy efficiency. This paper studies an EHS problem with multi-energy inputs and multi-energy outputs. The energy hub combines storage solutions for electricity and hydrogen, while also integrating demand response programs (DRPs) for both electricity and heat consumption. In reality, energy prices and loads fluctuate in real time, and the estimation of accurate probability distributions is a huge challenge. Thus, we model the partial availability of distribution information for electricity, hydrogen, and natural gas prices, as well as electricity, natural gas, and heat loads. A globalized distributionally robust energy hub scheduling (GDR-EHS) model is introduced to enhance the stability and adaptability of energy supply by creating outer and inner ambiguity sets. Lagrange duality and strong duality theory are used to obtain a computationally tractable formulation, then the proposed model is reformulated into a mixed integer linear programming (MILP) model. To enhance computational efficiency, we design an accelerated Branch and Cut (B&C) algorithm with Gomory cuts. Finally, the numerical results indicate that incorporating energy storage systems leads to a 4.39 % reduction in costs, while the addition of DRPs further lowers costs by 4.35 %. Furthermore, compared to Gurobi and the standard B&C algorithm, the accelerated B&C algorithm enhances computational efficiency by 50.1 % and 57.5 %.</div></div>","PeriodicalId":246,"journal":{"name":"Applied Energy","volume":"405 ","pages":"Article 127211"},"PeriodicalIF":11.0,"publicationDate":"2025-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145788237","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Assessment of offshore power potential in Zhoushan archipelago using a 45-year wind field product 利用45年风场产品评价舟山群岛海上电力潜力

IF 11 1区工程技术 Q1 ENERGY & FUELS

Applied Energy

Pub Date : 2025-12-17 DOI: 10.1016/j.apenergy.2025.127275

Honghuan Zhi , Han Liu , Liqin Zhou , Yifan Zhou , Feixiang Li , Xiaoran Wei , Yefei Bai

The growing demand for sustainable offshore wind energy has heightened the need for accurate wind resource assessment and feasible turbine deployment. Conventional multi-factor assessments fall short of identifying wind energy hotspots in multiple islands environment. This study proposes a three-stage wind resource assessment framework aimed to capture inter-island variations of wind resources across Zhoushan Archipelago. A long-term wind energy assessment at 100 m height is first conducted by using the optimal product through inter-comparison of seven mainstream global wind datasets against observations from ten evenly distributed sites. An Islands-Oriented Multi-Criterion assessment method, integrating energy, risk-stability, and environmental-cost factors, is then proposed to classify wind resources. Unconstrained offshore wind turbine zones are identified in the end through GIS-based spatial planning. Zhoushan Archipelago demonstrates substantial wind energy potential, with most areas exhibiting annual wind power density of about 420–550 W/m² but notable spatiotemporal variability, particularly low stability in central regions. The maximum energy potential concentrates in southeastern open seas (500–550 W/m²), with values gradually decreasing westward and exhibiting moderate levels near Hangzhou Bay (360–440 W/m²). The optimal turbine deployment zones are identified around Yushan Island, followed by Hangzhou Bay and east of Shanghai. Unconstrained suitable areas yield maximum technical potential of 29.48 TWh/yr. These findings provide actionable insights for wind energy planning in Zhoushan and other multi-island regions globally, emphasizing spatial prioritization and resource optimization.

对可持续海上风能的需求日益增长，提高了对准确的风力资源评估和可行的涡轮机部署的需求。传统的多因素评估无法识别多岛环境中的风能热点。本研究提出了一个三阶段风资源评价框架，旨在捕捉舟山群岛风资源的岛屿间变化。首先，通过将7个主流全球风数据集与10个均匀分布站点的观测结果进行相互比较，利用最优产品进行了100 m高度的长期风能评估。在此基础上，提出了一种综合能源、风险稳定性和环境成本等因素的面向岛屿的多准则评价方法。最后通过基于gis的空间规划来确定无约束的海上风力发电机区域。舟山群岛具有丰富的风能潜力，大部分地区年风电密度在420 ~ 550 W/m2之间，但存在明显的时空变异性，特别是中部地区的稳定性较低。最大能量潜力集中在东南海域（500 ~ 550 W/m2），向西逐渐减小，在杭州湾附近表现中等水平（360 ~ 440 W/m2）。确定了最佳水轮机部署区域为玉山岛周边，其次为杭州湾和上海以东。不受约束的合适区域的最大技术潜力为29.48太瓦时/年。研究结果为舟山及全球多岛地区的风能规划提供了可操作的见解，强调了空间优先和资源优化。

{"title":"Assessment of offshore power potential in Zhoushan archipelago using a 45-year wind field product","authors":"Honghuan Zhi , Han Liu , Liqin Zhou , Yifan Zhou , Feixiang Li , Xiaoran Wei , Yefei Bai","doi":"10.1016/j.apenergy.2025.127275","DOIUrl":"10.1016/j.apenergy.2025.127275","url":null,"abstract":"<div><div>The growing demand for sustainable offshore wind energy has heightened the need for accurate wind resource assessment and feasible turbine deployment. Conventional multi-factor assessments fall short of identifying wind energy hotspots in multiple islands environment. This study proposes a three-stage wind resource assessment framework aimed to capture inter-island variations of wind resources across Zhoushan Archipelago. A long-term wind energy assessment at 100 m height is first conducted by using the optimal product through inter-comparison of seven mainstream global wind datasets against observations from ten evenly distributed sites. An Islands-Oriented Multi-Criterion assessment method, integrating energy, risk-stability, and environmental-cost factors, is then proposed to classify wind resources. Unconstrained offshore wind turbine zones are identified in the end through GIS-based spatial planning. Zhoushan Archipelago demonstrates substantial wind energy potential, with most areas exhibiting annual wind power density of about 420–550 W/m<sup>2</sup> but notable spatiotemporal variability, particularly low stability in central regions. The maximum energy potential concentrates in southeastern open seas (500–550 W/m<sup>2</sup>), with values gradually decreasing westward and exhibiting moderate levels near Hangzhou Bay (360–440 W/m<sup>2</sup>). The optimal turbine deployment zones are identified around Yushan Island, followed by Hangzhou Bay and east of Shanghai. Unconstrained suitable areas yield maximum technical potential of 29.48 TWh/yr. These findings provide actionable insights for wind energy planning in Zhoushan and other multi-island regions globally, emphasizing spatial prioritization and resource optimization.</div></div>","PeriodicalId":246,"journal":{"name":"Applied Energy","volume":"406 ","pages":"Article 127275"},"PeriodicalIF":11.0,"publicationDate":"2025-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145765873","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Missing data-aware robust electrical load forecasting based on hierarchical downsampling-upsampling spatiotemporal graph network 基于分层下采样-上采样时空图网络的缺数据感知鲁棒电力负荷预测

IF 11 1区工程技术 Q1 ENERGY & FUELS

Applied Energy

Pub Date : 2025-12-17 DOI: 10.1016/j.apenergy.2025.127209

Pengfei Zhao , Zhirong Shen , Di Cao , Zhiping Lin , Zhe Chen , Weihao Hu

In real-world power systems, missing data is a frequent and inevitable issue due to sensor malfunctions and data transmission errors. This poses huge challenges for electrical load forecasting tasks since incomplete data can disrupt temporal patterns and obscure spatial dependencies. Traditional forecasting methods usually address this by imputing missing values before prediction, but such preprocessing can introduce bias and lead to error accumulation. To this end, this paper proposes a novel end-to-end missing data-oriented load forecasting framework based on the hierarchical downsampling-upsampling spatiotemporal graph network (HDU-STGNN). The proposed method can directly predict future load demand from partially observed inputs without requiring any data imputation process. Specifically, a temporal downsampling module is first developed to summarize observed portions of the input sequence into coarse-to-fine representations. This allows the model to capture long- and short-term load patterns while reducing sensitivity to irregularly missing entries. Then, a spatial coarsening and upsampling mechanism is proposed for effective information propagation across distant nodes. This allows the model to recover spatial dependencies even when local observations are unavailable. Finally, a multi-resolution attention fusion layer is utilized to adaptively reweight spatiotemporal features, which helps the model focus on reliable signals and suppress noise caused by incomplete data. Extensive experiments on real-world load datasets, covering three aggregation levels and missing rates up to 90 %, demonstrate the robustness of HDU-STGNN under varying degrees of data sparsity. In particular, the proposed model achieves an average reduction of 6.7 % in MAE and 9.9 % in MAPE compared with the strongest baseline across all settings, with improvements reaching up to 17.4 % under high missingness conditions.

在现实世界的电力系统中，由于传感器故障和数据传输错误，数据丢失是一个经常发生且不可避免的问题。这给电力负荷预测任务带来了巨大的挑战，因为不完整的数据会破坏时间模式并模糊空间依赖性。传统的预测方法通常通过在预测前输入缺失值来解决这一问题，但这种预处理会引入偏差并导致误差累积。为此，本文提出了一种基于分层下采样-上采样时空图网络（HDU-STGNN）的端到端缺失数据负载预测框架。该方法可以根据部分观测输入直接预测未来负荷需求，无需任何数据输入。具体来说，首先开发了一个时间下采样模块，将输入序列的观察部分总结为粗到细的表示。这允许模型捕捉长期和短期负载模式，同时降低对不规则缺失条目的敏感性。然后，提出了一种空间粗化和上采样机制，以实现信息在远距离节点间的有效传播。这使得模型即使在局部观测不可用时也能恢复空间依赖性。最后，利用多分辨率注意力融合层对时空特征进行自适应重加权，使模型更关注可靠信号，抑制数据不完整带来的噪声。在真实负载数据集上的大量实验，涵盖了三个聚合级别和高达90%的缺失率，证明了HDU-STGNN在不同程度的数据稀疏度下的鲁棒性。特别是，与所有设置中最强基线相比，所提出的模型在MAE和MAPE中平均降低了6.7%和9.9%，在高缺失条件下的改进可达17.4%。

{"title":"Missing data-aware robust electrical load forecasting based on hierarchical downsampling-upsampling spatiotemporal graph network","authors":"Pengfei Zhao , Zhirong Shen , Di Cao , Zhiping Lin , Zhe Chen , Weihao Hu","doi":"10.1016/j.apenergy.2025.127209","DOIUrl":"10.1016/j.apenergy.2025.127209","url":null,"abstract":"<div><div>In real-world power systems, missing data is a frequent and inevitable issue due to sensor malfunctions and data transmission errors. This poses huge challenges for electrical load forecasting tasks since incomplete data can disrupt temporal patterns and obscure spatial dependencies. Traditional forecasting methods usually address this by imputing missing values before prediction, but such preprocessing can introduce bias and lead to error accumulation. To this end, this paper proposes a novel end-to-end missing data-oriented load forecasting framework based on the hierarchical downsampling-upsampling spatiotemporal graph network (HDU-STGNN). The proposed method can directly predict future load demand from partially observed inputs without requiring any data imputation process. Specifically, a temporal downsampling module is first developed to summarize observed portions of the input sequence into coarse-to-fine representations. This allows the model to capture long- and short-term load patterns while reducing sensitivity to irregularly missing entries. Then, a spatial coarsening and upsampling mechanism is proposed for effective information propagation across distant nodes. This allows the model to recover spatial dependencies even when local observations are unavailable. Finally, a multi-resolution attention fusion layer is utilized to adaptively reweight spatiotemporal features, which helps the model focus on reliable signals and suppress noise caused by incomplete data. Extensive experiments on real-world load datasets, covering three aggregation levels and missing rates up to 90 %, demonstrate the robustness of HDU-STGNN under varying degrees of data sparsity. In particular, the proposed model achieves an average reduction of 6.7 % in MAE and 9.9 % in MAPE compared with the strongest baseline across all settings, with improvements reaching up to 17.4 % under high missingness conditions.</div></div>","PeriodicalId":246,"journal":{"name":"Applied Energy","volume":"405 ","pages":"Article 127209"},"PeriodicalIF":11.0,"publicationDate":"2025-12-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145788236","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0