Power Prediction of Ultra-Short-Term Photovoltaic Power Generation Based on Multi-Feature Fusion

Pengfa Zang; Keqi Wang; Zhongwei Zhang; Zhongyang Zhao; Longchao Yao; Weiguo Weng; Xuecheng Wu; Chenghang Zheng; Xiang Gao

doi:10.3788/LOP241387

Laser & Optoelectronics Progress, Volume. 62, Issue 5, 0525001(2025)

Power Prediction of Ultra-Short-Term Photovoltaic Power Generation Based on Multi-Feature Fusion

Pengfa Zang^1、*, Keqi Wang¹, Zhongwei Zhang², Zhongyang Zhao¹, Longchao Yao¹, Weiguo Weng³, Xuecheng Wu¹, Chenghang Zheng^1,3, and Xiang Gao^1,3

Author Affiliations

¹State Key Laboratory of Clean Energy Utilization, Zhejiang University, Hangzhou 310027, Zhejiang , China

²Dongfang Electric Yangtze River Delta (Hangzhou) Innovation Research Institute Co., Ltd., Hangzhou 310019, Zhejiang , China

³Jiaxing Research Institute, Zhejiang University, Jiaxing 314001, Zhejiang , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(16)

Fig. 1. Variation of the conversion parameters of the irradiance absorbed by photovoltaic modules. (a) Different types of irradiance absorbed by photovoltaic modules; (b) relative position of sun and earth; (c) relative position change between sun and photovoltaic module

Download full size

Fig. 2. Flow chart of photovoltaic power generation power prediction model based on multi-feature extraction

Download full size

Fig. 3. Four-year time series curves for each variable

Download full size

Fig. 4. Heat map of correlation coefficient between photovoltaic power generation power and environmental factors

Download full size

Fig. 5. Comparison of multi-model prediction results in spring scenarios

Download full size

Fig. 6. Comparison of multi-model prediction results in summer scenarios

Download full size

Fig. 7. Comparison of multi-model prediction results in autumn scenarios

Download full size

Fig. 8. Comparison of multi-model prediction results for winter scenarios

Download full size

Fig. 9. Comparison of multi-model prediction results under different weather conditions

Download full size

Fig. 10. Comparison of the model in this paper with the predictions of commercial software

Download full size

Table 1. Spring scene multi-model can predict outcome evaluation index
View table
Table 1. Spring scene multi-model can predict outcome evaluation index
Model Accuracy /% R² RMSE /kW MAE /kW
LSTM 92.51 0.9264 8.248 4.119
1 feature 93.98 0.9403 7.594 3.763
2 features 94.27 0.9359 7.429 3.601
4 features 94.64 0.9509 6.891 3.365

Table 2. Evaluation index of multi-model prediction results of summer scenes
View table
Table 2. Evaluation index of multi-model prediction results of summer scenes
Model Accuracy /% R² RMSE /kW MAE /kW
LSTM 90.02 0.8490 12.398 6.386
1 feature 90.26 0.8490 12.399 6.308
2 features 91.30 0.8921 10.540 5.523
4 features 91.45 0.8991 10.298 5.210

Table 3. Evaluation index of multi-model prediction results in autumn scenarios
View table
Table 3. Evaluation index of multi-model prediction results in autumn scenarios
Model Accuracy /% R² RMSE /kW MAE /kW
LSTM 93.08 0.9106 8.311 3.860
1 feature 93.02 0.9089 8.389 3.842
2 features 94.61 0.9420 6.695 3.040
4 features 94.60 0.9438 6.589 3.215

Table 4. Evaluation index of multi-model prediction results of winter scenes
View table
Table 4. Evaluation index of multi-model prediction results of winter scenes
Model Accuracy /% R² RMSE /kW MAE /kW
LSTM 95.40 0.9556 5.328 2.430
1 feature 95.44 0.9542 5.436 2.316
2 features 96.32 0.9693 4.449 1.917
4 features 96.88 0.9736 4.231 1.735

Table 5. Evaluation index of multi-model prediction results under different weather conditions

View table

Table 5. Evaluation index of multi-model prediction results under different weather conditions

Weather type	Model	Accuracy /%	R²	RMSE /kW	MAE /kW
Sunny	LSTM	94.97	0.9674	6.027	3.020
	1 feature	95.80	0.9794	4.812	2.348
	2 features	96.89	0.9880	3.673	2.005
	4 features	97.13	0.9896	3.420	1.934
Cloudy	LSTM	91.60	0.9080	10.671	5.373
	1 feature	92.67	0.9313	9.219	4.257
	2 features	92.67	0.9339	9.045	4.384
	4 features	92.96	0.9375	8.797	4.573
Rainy	LSTM	92.42	0.8503	9.792	4.675
	1 feature	92.63	0.8589	9.508	4.512
	2 features	92.76	0.8678	9.200	4.426
	4 features	93.14	0.8779	8.844	4.424
Cloudy to rainy	LSTM	92.03	0.8954	10.228	4.848
	1 feature	92.27	0.9029	9.853	4.588
	2 features	92.97	0.9154	9.197	4.363
	4 features	93.11	0.9182	9.045	4.332
Sunny to cloudy	LSTM	91.24	0.8529	10.922	5.297
	1 feature	91.16	0.8520	10.957	5.283
	2 features	91.81	0.8764	10.064	4.709
	4 features	92.16	0.8751	10.013	4.990
Sunny to rainy	LSTM	93.62	0.9378	7.988	3.661
	1 feature	94.21	0.9509	7.097	3.307
	2 features	93.58	0.9410	7.776	3.656
	4 features	94.33	0.9531	6.948	3.217

Table 6. Evaluation indicators for model in this paper and commercial software prediction results
View table
Table 6. Evaluation indicators for model in this paper and commercial software prediction results
Model Accuracy /% R² RMSE /MW MAE /MW
4 features model 92.51 0.9371 4.976 2.03
Business software 90.38 0.8843 6.611 2.54

Tools

Get Citation

Copy Citation Text

Pengfa Zang, Keqi Wang, Zhongwei Zhang, Zhongyang Zhao, Longchao Yao, Weiguo Weng, Xuecheng Wu, Chenghang Zheng, Xiang Gao. Power Prediction of Ultra-Short-Term Photovoltaic Power Generation Based on Multi-Feature Fusion[J]. Laser & Optoelectronics Progress, 2025, 62(5): 0525001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites