Electric Power ›› 2023, Vol. 56 ›› Issue (9): 187-195.DOI: 10.11930/j.issn.1004-9649.202212013

• New Energy • Previous Articles     Next Articles

Short-Term Photovoltaic Power Prediction Based on Standard Clear Sky Set Defined by No Climbing Event

GUO Hongwu1, CHE Jianfeng2, YAN Yixun3, WANG Lijie4   

  1. 1. Chifeng Power Supply Company, State Grid Inner Mongolia East Electric Power Co.. Ltd., Chifeng 024000, China;
    2. China Electric Power Research Institute, Beijing 100192, China;
    3. China Railway Publishing House Company Limited, Beijing 100054, China;
    4. School of Automation, Beijing Information Science and Technology University, Beijing 100192, China
  • Received:2022-12-02 Revised:2023-05-24 Accepted:2023-03-02 Online:2023-09-23 Published:2023-09-28
  • Supported by:
    This work is supported by Science and Technology Project of State Grid Inner Mongolia East Electric Power Co., Ltd. (No.SGMDCF00YWJS2000769).

Abstract: Photovoltaic power output is influenced by season, weather conditions, and other factors with randomness and uncertainty. It is difficult to predict the power output under bad weather with strong volatility. Therefore, this paper proposed a short-term photovoltaic power prediction model based on the standard clear sky set with no climbing event definition. The sample points with no climbing events in one day were extracted by the climbing definition and defined as a standard clear sky set, and the difference between them and the historical actual power was made. The obtained difference was used as the output target variable, and the numerical weather forecast was used as the input variable. The long-short term memory (LSTM) model was used to model and forecast the difference. Finally, the predicted photovoltaic output power was obtained indirectly by making the difference between the standard clear sky set and the predicted difference. Through the simulation of a photovoltaic power station and comparison of calculation examples, the short-term photovoltaic power prediction accuracy of the proposed model was improved by 2%~4%. In severe weather, this method can reduce mean absolute error (MAE) and root-mean-square error (RMSE) by about 3%, which verifies the prediction performance and effectiveness of the proposed model.

Key words: short-term photovoltaic power prediction, non-climbing sample extraction, standard clear sky set, long-short term memory model