MRA-YOLOv8: A Network Enhancing Feature Extraction Ability for Photovoltaic Cell Defects DOI Creative Commons

Nannan Wang,

Siqi Huang,

Xiangpeng Liu

et al.

Sensors, Journal Year: 2025, Volume and Issue: 25(5), P. 1542 - 1542

Published: March 2, 2025

To address the challenges posed by complex backgrounds and low occurrence in photovoltaic cell images captured industrial sensors, we propose a novel defect detection method: MRA-YOLOv8. First, multi-branch coordinate attention network (MBCANet) is introduced into backbone. The (CANet) incorporated to mitigate noise impact of background information on task, multiple branches are employed enhance model’s feature extraction capability. Second, integrate multi-path module, ResBlock, neck. This module provides finer-grained multi-scale features, improving from enhancing robustness. Finally, implement alpha-minimum point distance-based IoU (AMPDIoU) head. loss function enhances accuracy robustness small object integrating minimum (MPDIoU) Alpha-IoU methods. results demonstrate that MRA-YOLOv8 outperforms other mainstream methods performance. On electroluminescence anomaly (PVEL-AD) dataset, proposed method achieves mAP50 91.7%, representing an improvement 3.1% over YOLOv8 16.1% transformer (DETR). SPDI our 69.3%, showing 2.1% 6.6% DETR. also exhibits great deployment potential. It can be effectively integrated with drone-based inspection systems, allowing for efficient accurate PV plant inspections. Moreover, tackle issue data imbalance, generating synthetic via generative adversarial networks (GANs), which supplement limited samples improve generalization ability.

Language: Английский

Calibration of deterministic NWP forecasts and its impact on verification DOI Creative Commons
Martin János Mayer, Dazhi Yang

International Journal of Forecasting, Journal Year: 2022, Volume and Issue: 39(2), P. 981 - 991

Published: June 14, 2022

Deterministic forecasts (as opposed to ensemble or probabilistic forecasts) issued by numerical weather prediction (NWP) models require post-processing. Such corrective procedure can be viewed as a form of calibration. It is well known that, based on different objective functions, e.g., minimizing the mean square error absolute error, calibrated have impacts verification. In this regard, paper investigates how calibration directive affect various aspects forecast quality outlined in Murphy–Winkler distribution-oriented verification framework. argued that correlation coefficient best measure for potential performance NWP when linear involved, because (1) it not affected calibration, (2) used compute skill score linearly forecasts, and (3) avoid deficiency using squared rank forecasts. Since no single metric fully represent all quality, forecasters need understand trade-offs between strategies. To echo increasing bridge atmospheric sciences, renewable energy engineering, power system move toward grand goal carbon neutrality, first provides brief introduction solar forecasting, then revolves its discussion around forecasting case study, such readers journal gain further understanding subject thus potentially contribute it.

Language: Английский

Citations

36
Probabilistic photovoltaic power forecasting using a calibrated ensemble of model chains DOI Creative Commons
Martin János Mayer, Dazhi Yang

Renewable and Sustainable Energy Reviews, Journal Year: 2022, Volume and Issue: 168, P. 112821 - 112821

Published: Aug. 12, 2022

Physical model chain is a step-by-step modeling framework for the conversion of irradiance to photovoltaic (PV) power. When fed with forecasts, it provides corresponding PV power forecasts. Despite its advantages, forecasting chains has yet receive attention that deserves. In several recent works, however, idea model-chain-based solar been formally modernized, though was restricted deterministic forecasting. this work, extended probability space, in that, calibrated ensemble used generate probabilistic Using two-year data from eight plants Hungary, alongside professional weather forecasts issued by Hungarian Meteorological Services, empirically shown raw model-chain tend be underdispered, but adequate post-processing able improve calibration and reduce continuous ranked score ensembles 20%. Given fact uncertainty quantification cardinal importance grid integration, extension thought beneficial.

Language: Английский

Citations

32
Verifying operational intra-day solar forecasts from ECMWF and NOAA DOI Creative Commons
Dazhi Yang, Meng Wan, Jamie M. Bright

et al.

Solar Energy, Journal Year: 2022, Volume and Issue: 236, P. 743 - 755

Published: March 28, 2022

Language: Английский

Citations

31
Retrieval of sub-kilometer resolution solar irradiance from Fengyun-4A satellite using a region-adapted Heliosat-2 method DOI
Chunlin Huang, Hongrong Shi, Dazhi Yang

et al.

Solar Energy, Journal Year: 2023, Volume and Issue: 264, P. 112038 - 112038

Published: Sept. 30, 2023

Language: Английский

Citations

16
Predictability and forecast skill of solar irradiance over the contiguous United States DOI
Bai Liu, Dazhi Yang, Martin János Mayer

et al.

Renewable and Sustainable Energy Reviews, Journal Year: 2023, Volume and Issue: 182, P. 113359 - 113359

Published: May 19, 2023

Language: Английский

Citations

15
Deep graph gated recurrent unit network-based spatial–temporal multi-task learning for intelligent information fusion of multiple sites with application in short-term spatial–temporal probabilistic forecast of photovoltaic power DOI
Mingliang Bai, Z. C. Zhou, Jingjing Li

et al.

Expert Systems with Applications, Journal Year: 2023, Volume and Issue: 240, P. 122072 - 122072

Published: Nov. 10, 2023

Language: Английский

Citations

15
Global and direct solar irradiance estimation using deep learning and selected spectral satellite images DOI
Shanlin Chen, Chengxi Li, Yuying Xie

et al.

Applied Energy, Journal Year: 2023, Volume and Issue: 352, P. 121979 - 121979

Published: Sept. 29, 2023

Language: Английский

Citations

14
Improved satellite-based intra-day solar forecasting with a chain of deep learning models DOI
Shanlin Chen, Chengxi Li, Roland B. Stull

et al.

Energy Conversion and Management, Journal Year: 2024, Volume and Issue: 313, P. 118598 - 118598

Published: May 30, 2024

Language: Английский

Citations

5
Changes in photovoltaic potential over China in a warmer future DOI Creative Commons
Jintao Zhang, Qinglong You, Safi Ullah

et al.

Environmental Research Letters, Journal Year: 2022, Volume and Issue: 17(11), P. 114032 - 114032

Published: Nov. 1, 2022

Abstract Solar photovoltaic (PV) technology offers a promising path for addressing energy demand and mitigating climate change. However, climatic conditions relevant to the productivity of solar power may be changed in warmer future. Here, we quantify impact change on PV potential over China based high-resolution projections within framework Coordinated Regional Climate Downscaling Experiment (CORDEX). Our results indicate that annual mean ( POT ) would decrease by several percent relative reference period (1986–2005) under climate, with most pronounced Tibetan Plateau, which is currently solar-rich region China. beyond changes state, demonstrate substantial increase events extreme low outputs (i.e. those falling below 10th percentile probability distribution daily anomalies). For instance, frequency projected reach nearly three times level Plateau scenario 3 °C global warming (similar late-century current mitigation policies). Future are dominated surface irradiance, while condition further exacerbates . highlight both inputs should considered simultaneously when assessing impacts outputs.

Language: Английский

Citations

21
Implications of Future Price Trends and Interannual Resource Uncertainty on Firm Solar Power Delivery With Photovoltaic Overbuilding and Battery Storage DOI
Guoming Yang, Dazhi Yang, Chao Lyu

et al.

IEEE Transactions on Sustainable Energy, Journal Year: 2023, Volume and Issue: 14(4), P. 2036 - 2048

Published: May 8, 2023

Generation from solar is inherently variable. Through a strategic combination of excessive capacity expansion (i.e., overbuilding) and battery storage, the variable generation can be cost-effectively firmed up, in that, it able to meet required target with absolute certainty. Firming up implies additional cost, which quantified through firm kWh premium. This paper proposes new model for optimization premium either mixed-integer linear program or bilinear program, depending on whether generic detailed used. The (bi)linear-program formulation greatly reduces complexity original iterative approach. Additionally, since function photovoltaic prices, we show how future price change affect economics power delivery true grid parity eventually achieved. Lastly, sensitivity modeling uncertainty inter-annual resource analyzed.

Language: Английский

Citations

12