Pub Date : 2025-11-03DOI: 10.1109/JPHOTOV.2025.3619980
Daolei Wang;Zhi Huang;Lei Peng;Peng Yan;Shaokai Zheng;Jiawen Liu;Yang Long
To address the challenges of blurred target boundaries, uneven feature distribution, and high computational cost in complex photovoltaic (PV) environments, this article proposes a lightweight uncrewed aerial vehicle (UAV)-based infrared hotspot detection model—real-time detection transformer (RTDETR)-CELite. Utilizing RTDETR-R18, the model incorporates the CSP_ELGCA_CGLU module, which integrates local-global attention and gated channel enhancement. This improves the perception of key regions while reducing computational complexity. In addition, a ConvEdgeFusion module is designed to combine shallow edge structures with multiscale semantic features. This improvement improves the model’s ability to accurately depict hot spot boundaries and their distribution areas, thereby significantly reducing false positives and false negatives. Experimental results show that RTDETR-CELite significantly reduces the model scale without affecting detection performance. Compared to the original RTDETR-R18, mAP50 improves from 82.04% to 84.06%, and mAP50:95 improves from 62.01% to 62.56%. The number of parameters decreases by 31.6% to 13.6M, computational cost drops by 17.4% to 47.1 GFLOPs, and inference speed increases to 300.5 FPS. These results indicate that RTDETR-CELite strikes an effective compromise between precision and computational efficiency, rendering it highly applicable to UAV-based or edge-device deployment for timely identification of PV hotspots, and showcasing promising potential in practical scenarios.
{"title":"RTDETR-CELite: Lightweight Remote Sensing PV Defect Detection via Edge-Aware and Cross-Channel Feature Fusion","authors":"Daolei Wang;Zhi Huang;Lei Peng;Peng Yan;Shaokai Zheng;Jiawen Liu;Yang Long","doi":"10.1109/JPHOTOV.2025.3619980","DOIUrl":"https://doi.org/10.1109/JPHOTOV.2025.3619980","url":null,"abstract":"To address the challenges of blurred target boundaries, uneven feature distribution, and high computational cost in complex photovoltaic (PV) environments, this article proposes a lightweight uncrewed aerial vehicle (UAV)-based infrared hotspot detection model—real-time detection transformer (RTDETR)-CELite. Utilizing RTDETR-R18, the model incorporates the CSP_ELGCA_CGLU module, which integrates local-global attention and gated channel enhancement. This improves the perception of key regions while reducing computational complexity. In addition, a ConvEdgeFusion module is designed to combine shallow edge structures with multiscale semantic features. This improvement improves the model’s ability to accurately depict hot spot boundaries and their distribution areas, thereby significantly reducing false positives and false negatives. Experimental results show that RTDETR-CELite significantly reduces the model scale without affecting detection performance. Compared to the original RTDETR-R18, mAP50 improves from 82.04% to 84.06%, and mAP50:95 improves from 62.01% to 62.56%. The number of parameters decreases by 31.6% to 13.6M, computational cost drops by 17.4% to 47.1 GFLOPs, and inference speed increases to 300.5 FPS. These results indicate that RTDETR-CELite strikes an effective compromise between precision and computational efficiency, rendering it highly applicable to UAV-based or edge-device deployment for timely identification of PV hotspots, and showcasing promising potential in practical scenarios.","PeriodicalId":445,"journal":{"name":"IEEE Journal of Photovoltaics","volume":"16 1","pages":"160-175"},"PeriodicalIF":2.6,"publicationDate":"2025-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145802390","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-31DOI: 10.1109/JPHOTOV.2025.3620568
Wilkin Wöhler;Johannes M. Greulich;Andreas W. Bett
We derive an analytical description of leakage currents in an ohmic system of two conductive layers, with current in- and outflow at two line contacts on the first layer, and current flow in the second layer induced over a resistive interface. Examples of such interfaces in the photovoltaic context include the tunnel interface of TOPCon solar cells, the high-low junction of silicon heterojunction (SHJ) solar cells, and p-n junctions for low current densities. Experimentally the modeled leakage currents are observed in measurements of transfer length method (TLM) samples of SHJ solar cells due to the finite shunt resistivity of the p-n junction. Using the new model, we find that for a typical TLM-setup with a contacting distance of $l_{text{c}}=text{1 cm}$, apparent sheet resistance reductions of 0.3, 2.6, and 9.6 $Omega$ for a top layer of $R_{1}=text{100};{Omega }$ occur for interface resistivities $rho _{text{c}}$ of 100, 10, and 1 $text{k}Omega text{cm}^{2}$, respectively. Evaluating the measurement example by the commonly used linear regression, a twice higher contact resistivity is found in comparison to a numerical least square fit of the new model. Similar results are obtained in a synthetic data study using the solar cell simulation software Quokka3, with contact resistivity deviations of up to $text{10 m} Omega text{cm}^{2}$ for the linear regression evaluation. By evaluating the same data with the new analytical model, the original simulation parameters of contact resistivity and sheet resistance are recovered with relative deviations below 0.2%.
{"title":"Analytical Model of Leakage Currents in Contact Resistivity Measurements on Silicon Solar Cells","authors":"Wilkin Wöhler;Johannes M. Greulich;Andreas W. Bett","doi":"10.1109/JPHOTOV.2025.3620568","DOIUrl":"https://doi.org/10.1109/JPHOTOV.2025.3620568","url":null,"abstract":"We derive an analytical description of leakage currents in an ohmic system of two conductive layers, with current in- and outflow at two line contacts on the first layer, and current flow in the second layer induced over a resistive interface. Examples of such interfaces in the photovoltaic context include the tunnel interface of TOPCon solar cells, the high-low junction of silicon heterojunction (SHJ) solar cells, and p-n junctions for low current densities. Experimentally the modeled leakage currents are observed in measurements of transfer length method (TLM) samples of SHJ solar cells due to the finite shunt resistivity of the p-n junction. Using the new model, we find that for a typical TLM-setup with a contacting distance of <inline-formula><tex-math>$l_{text{c}}=text{1 cm}$</tex-math></inline-formula>, apparent sheet resistance reductions of 0.3, 2.6, and 9.6 <inline-formula><tex-math>$Omega$</tex-math></inline-formula> for a top layer of <inline-formula><tex-math>$R_{1}=text{100};{Omega }$</tex-math></inline-formula> occur for interface resistivities <inline-formula><tex-math>$rho _{text{c}}$</tex-math></inline-formula> of 100, 10, and 1 <inline-formula><tex-math>$text{k}Omega text{cm}^{2}$</tex-math></inline-formula>, respectively. Evaluating the measurement example by the commonly used linear regression, a twice higher contact resistivity is found in comparison to a numerical least square fit of the new model. Similar results are obtained in a synthetic data study using the solar cell simulation software Quokka3, with contact resistivity deviations of up to <inline-formula><tex-math>$text{10 m} Omega text{cm}^{2}$</tex-math></inline-formula> for the linear regression evaluation. By evaluating the same data with the new analytical model, the original simulation parameters of contact resistivity and sheet resistance are recovered with relative deviations below 0.2%.","PeriodicalId":445,"journal":{"name":"IEEE Journal of Photovoltaics","volume":"16 1","pages":"120-127"},"PeriodicalIF":2.6,"publicationDate":"2025-10-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145802344","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-28DOI: 10.1109/JPHOTOV.2025.3626191
{"title":"2025 Index IEEE Journal of Photovoltaics","authors":"","doi":"10.1109/JPHOTOV.2025.3626191","DOIUrl":"https://doi.org/10.1109/JPHOTOV.2025.3626191","url":null,"abstract":"","PeriodicalId":445,"journal":{"name":"IEEE Journal of Photovoltaics","volume":"15 6","pages":"995-1026"},"PeriodicalIF":2.6,"publicationDate":"2025-10-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11219677","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145405266","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-22DOI: 10.1109/JPHOTOV.2025.3611428
Mirra M. Rasmussen;J. Diego Zubieta Sempertegui;Nicholas Moser-Mancewicz;Jonathan L. Bryan;Natasha E. Hjerrild;Kristopher O. Davis;Mariana I. Bertoni;Laura S. Bruckman;Ina T. Martin
Advanced Si photovoltaic architectures incorporate different materials and processing pathways that influence degradation modes. Ultraviolet-induced degradation (UVID) is an understudied degradation mode for advanced cell architectures and is of increasing concern to industry due to growing adoption of UV-transparent encapsulation and bifacial technologies. In order to adopt new and evolving technologies confidently, novel component materials and processing techniques must be evaluated and designed for long-term stability, in addition to the conventional design focus on efficiency. In this work, a study protocol framework is presented for the rapid screening of unencapsulated devices against UVID. Unencapsulated passivated emitter rear contact (PERC) and tunnel oxide passivated contact (TOPCon) devices were aged under different UV irradiance intensities and measured via conventional nondestructive electrical characterization methods to assess performance degradation. Based on the results, protocol efficacy and recommendations for further study are discussed. This work is part of a broader effort to develop rapid screening processes that cut across architectures and exposure conditions to aid module manufacturers in vetting new materials choices for long-term stability.
{"title":"High-Intensity UV Exposure for the Rapid Screening of Silicon Photovoltaic Architectures","authors":"Mirra M. Rasmussen;J. Diego Zubieta Sempertegui;Nicholas Moser-Mancewicz;Jonathan L. Bryan;Natasha E. Hjerrild;Kristopher O. Davis;Mariana I. Bertoni;Laura S. Bruckman;Ina T. Martin","doi":"10.1109/JPHOTOV.2025.3611428","DOIUrl":"https://doi.org/10.1109/JPHOTOV.2025.3611428","url":null,"abstract":"Advanced Si photovoltaic architectures incorporate different materials and processing pathways that influence degradation modes. Ultraviolet-induced degradation (UVID) is an understudied degradation mode for advanced cell architectures and is of increasing concern to industry due to growing adoption of UV-transparent encapsulation and bifacial technologies. In order to adopt new and evolving technologies confidently, novel component materials and processing techniques must be evaluated and designed for long-term stability, in addition to the conventional design focus on efficiency. In this work, a study protocol framework is presented for the rapid screening of unencapsulated devices against UVID. Unencapsulated passivated emitter rear contact (PERC) and tunnel oxide passivated contact (TOPCon) devices were aged under different UV irradiance intensities and measured via conventional nondestructive electrical characterization methods to assess performance degradation. Based on the results, protocol efficacy and recommendations for further study are discussed. This work is part of a broader effort to develop rapid screening processes that cut across architectures and exposure conditions to aid module manufacturers in vetting new materials choices for long-term stability.","PeriodicalId":445,"journal":{"name":"IEEE Journal of Photovoltaics","volume":"16 1","pages":"142-149"},"PeriodicalIF":2.6,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145802368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-22DOI: 10.1109/JPHOTOV.2025.3621371
{"title":"Call for Papers for a Special Issue of IEEE Transactions on Electron Devices on “Reliability of Advanced Nodes”","authors":"","doi":"10.1109/JPHOTOV.2025.3621371","DOIUrl":"https://doi.org/10.1109/JPHOTOV.2025.3621371","url":null,"abstract":"","PeriodicalId":445,"journal":{"name":"IEEE Journal of Photovoltaics","volume":"15 6","pages":"993-994"},"PeriodicalIF":2.6,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11214299","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145339711","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-22DOI: 10.1109/JPHOTOV.2025.3621367
{"title":"IEEE Journal of Photovoltaics Information for Authors","authors":"","doi":"10.1109/JPHOTOV.2025.3621367","DOIUrl":"https://doi.org/10.1109/JPHOTOV.2025.3621367","url":null,"abstract":"","PeriodicalId":445,"journal":{"name":"IEEE Journal of Photovoltaics","volume":"15 6","pages":"C3-C3"},"PeriodicalIF":2.6,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11214302","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145339692","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-22DOI: 10.1109/JPHOTOV.2025.3621369
{"title":"Call for Papers for a Special Issue of IEEE Transactions on Electron Devices on “Ultrawide Band Gap Semiconductor Device for RF, Power and Optoelectronic Applications”","authors":"","doi":"10.1109/JPHOTOV.2025.3621369","DOIUrl":"https://doi.org/10.1109/JPHOTOV.2025.3621369","url":null,"abstract":"","PeriodicalId":445,"journal":{"name":"IEEE Journal of Photovoltaics","volume":"15 6","pages":"991-992"},"PeriodicalIF":2.6,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11214309","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145339687","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-22DOI: 10.1109/JPHOTOV.2025.3616635
Gabriel Lopez;Dmitrii Bogdanov;Rasul Satymov;Christian Breyer
Energy transition pathways for large continental areas are largely understood to be achievable using a diverse set of onshore renewable energy technologies. Previous research for the integrated United States and Canada energy–industry system indicated that solar photovoltaics (PVs) may dominate the primary energy structure, complemented by onshore wind power. However, societal constraints may require increased supply diversity, and onshore renewable energy may not be sufficient for densely populated regions, especially on the east coast of the United States. The LUT Energy System Transition Model was applied to investigate the role of floating offshore solar PV coupled with offshore wind and wave power when onshore solar PV is limited. The results indicate that, when onshore solar PV is limited to 60% of electricity generation, 434 GW of floating offshore solar PV may be installed by 2050 as part of a hybrid power plant sharing the same grid connection as offshore wind power, which reaches 414 GW of installed capacity, contributing 607 and 1576 TWh to the electricity supply, respectively. In total, 7.4 TW of solar PV capacity is installed by 2050, complemented by 1.4 TW of onshore wind power. Increased supply diversity still leads to a 42% reduction in the levelized cost of electricity, reaching 32.7 €/MWh in 2050. Compared with cost-optimal conditions, the levelized cost of final energy and nonenergy use in 2050 increases by 28% to 52.7 €/MWh. Nevertheless, such increased costs may be justifiable to meet societal constraints, and a diverse power-to-X economy structure for the United States and Canada may still be technoeconomically viable.
{"title":"Floating Offshore Solar Photovoltaics for Land-Constrained and Diverse Renewable Supply Conditions in the United States and Canada","authors":"Gabriel Lopez;Dmitrii Bogdanov;Rasul Satymov;Christian Breyer","doi":"10.1109/JPHOTOV.2025.3616635","DOIUrl":"https://doi.org/10.1109/JPHOTOV.2025.3616635","url":null,"abstract":"Energy transition pathways for large continental areas are largely understood to be achievable using a diverse set of onshore renewable energy technologies. Previous research for the integrated United States and Canada energy–industry system indicated that solar photovoltaics (PVs) may dominate the primary energy structure, complemented by onshore wind power. However, societal constraints may require increased supply diversity, and onshore renewable energy may not be sufficient for densely populated regions, especially on the east coast of the United States. The LUT Energy System Transition Model was applied to investigate the role of floating offshore solar PV coupled with offshore wind and wave power when onshore solar PV is limited. The results indicate that, when onshore solar PV is limited to 60% of electricity generation, 434 GW of floating offshore solar PV may be installed by 2050 as part of a hybrid power plant sharing the same grid connection as offshore wind power, which reaches 414 GW of installed capacity, contributing 607 and 1576 TWh to the electricity supply, respectively. In total, 7.4 TW of solar PV capacity is installed by 2050, complemented by 1.4 TW of onshore wind power. Increased supply diversity still leads to a 42% reduction in the levelized cost of electricity, reaching 32.7 €/MWh in 2050. Compared with cost-optimal conditions, the levelized cost of final energy and nonenergy use in 2050 increases by 28% to 52.7 €/MWh. Nevertheless, such increased costs may be justifiable to meet societal constraints, and a diverse power-to-X economy structure for the United States and Canada may still be technoeconomically viable.","PeriodicalId":445,"journal":{"name":"IEEE Journal of Photovoltaics","volume":"16 1","pages":"60-68"},"PeriodicalIF":2.6,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11214224","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145802327","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-22DOI: 10.1109/JPHOTOV.2025.3620446
{"title":"Golden List of Reviewers","authors":"","doi":"10.1109/JPHOTOV.2025.3620446","DOIUrl":"https://doi.org/10.1109/JPHOTOV.2025.3620446","url":null,"abstract":"","PeriodicalId":445,"journal":{"name":"IEEE Journal of Photovoltaics","volume":"15 6","pages":"988-990"},"PeriodicalIF":2.6,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11214300","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145339715","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-22DOI: 10.1109/JPHOTOV.2025.3608474
Tajrian Mollick;Md Jobayer;Md. Samrat Hossin;Shahidul Islam Khan;A. S. Nazmul Huda;Saifur Rahman Sabuj
Solar energy adoption is rapidly growing as a sustainable option, with solar panels used on residential buildings, commercial properties, and large-scale farms. However, the unpredictable nature of solar power can lead to suboptimal energy generation from photovoltaic (PV) panels. Despite the high effectiveness of deep learning (DL) models in forecasting PV power, they often struggle with the perception of being “closed boxes” that lack clear explanations for their prediction results, which fail to highlight the key features for PV prediction. To address the critical issue of full transparency, this study explores a well-known DL model named lightweight deep neural network (LWDNN) in PV power forecasting, along with the application of explainable artificial intelligence (XAI) tools like Shapley Additive exPlanations (SHAP) and local interpretable model-agnostic explanations (LIME). Real-time data collected from a grid-connected solar PV system located in Dhaka were utilized to perform the prediction. By enabling XAI model interpretation, we identified feature contributions and explained individual predictions, reducing training computational demands without compromising accuracy. The reliability of the LWDNN model is assessed using both complete and reduced feature sets through performance metrics such as root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R2). The test results show that the proposed LWDNN model based on SHAP analysis outperforms conventional schemes by achieving RMSE = 6.180 kW, MAE = 1.939 kW, and R2 = 0.988. Finally, the model was implemented on a Raspberry Pi for low-power solar forecasting, demonstrating the feasibility of edge deployment.
{"title":"An Interpretable Deep Learning Model for Solar Power Generation Forecasting in a Grid-Connected Hybrid Solar System","authors":"Tajrian Mollick;Md Jobayer;Md. Samrat Hossin;Shahidul Islam Khan;A. S. Nazmul Huda;Saifur Rahman Sabuj","doi":"10.1109/JPHOTOV.2025.3608474","DOIUrl":"https://doi.org/10.1109/JPHOTOV.2025.3608474","url":null,"abstract":"Solar energy adoption is rapidly growing as a sustainable option, with solar panels used on residential buildings, commercial properties, and large-scale farms. However, the unpredictable nature of solar power can lead to suboptimal energy generation from photovoltaic (PV) panels. Despite the high effectiveness of deep learning (DL) models in forecasting PV power, they often struggle with the perception of being “closed boxes” that lack clear explanations for their prediction results, which fail to highlight the key features for PV prediction. To address the critical issue of full transparency, this study explores a well-known DL model named lightweight deep neural network (LWDNN) in PV power forecasting, along with the application of explainable artificial intelligence (XAI) tools like Shapley Additive exPlanations (SHAP) and local interpretable model-agnostic explanations (LIME). Real-time data collected from a grid-connected solar PV system located in Dhaka were utilized to perform the prediction. By enabling XAI model interpretation, we identified feature contributions and explained individual predictions, reducing training computational demands without compromising accuracy. The reliability of the LWDNN model is assessed using both complete and reduced feature sets through performance metrics such as root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R<sup>2</sup>). The test results show that the proposed LWDNN model based on SHAP analysis outperforms conventional schemes by achieving RMSE = 6.180 kW, MAE = 1.939 kW, and R<sup>2</sup> = 0.988. Finally, the model was implemented on a Raspberry Pi for low-power solar forecasting, demonstrating the feasibility of edge deployment.","PeriodicalId":445,"journal":{"name":"IEEE Journal of Photovoltaics","volume":"15 6","pages":"941-954"},"PeriodicalIF":2.6,"publicationDate":"2025-10-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145339706","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：