Electric Power ›› 2025, Vol. 58 ›› Issue (3): 183-192.DOI: 10.11930/j.issn.1004-9649.202410048

• New Energy and Energy Storage • Previous Articles     Next Articles

A Joint Scenario Generation Method for Wind-Solar Meteorological Resources Based on Improved Generative Adversarial Network

Tonghai JIANG1(), Feng WANG1(), Ziqi LIU1(), Shuaijie SHAN2()   

  1. 1. CGN (Shandong) New Energy Investment Co., Ltd., Jinan 250014, China
    2. School of Electrical Engineering, Shandong University, Jinan 250061, China
  • Received:2024-10-15 Accepted:2025-01-13 Online:2025-03-23 Published:2025-03-28
  • Supported by:
    This work is supported by Science and Technology Project of SGCC (No.4000-202355381A-2-3-XG) and National Natural Science Foundation of China (No.52177095).

Abstract:

With the large-scale development of renewable energy, the volatility and uncertainty of wind and solar meteorological resources have a great impact on the safe and stable operation of the power system. To support the power system planning and decision-making, and overcome the uncertainty of meteorological resources in predicting new energy power, this paper proposes a wind-solar meteorological resource joint scenario generation method based on the deep convolutional generative adversarial networks (DCGAN). Firstly, based on the spatial distribution pattern of wind and photovoltaic power stations, the spatial distribution of new energy stations are associated with meteorological factors. Secondly, the DCGAN technology is used to simulate the long time series scenarios of wind speed-irradiance meteorological elements, thus achieving the joint scenario generation of wind- solar resources in a regional scope; Next, the K-medoids method is used to reduce the redundance scenarios and keep the typical scenarios at the cluster center, subsequently realizing the direct evaluation of the scenario generation effects. And then, the meteorological elements are converted into wind power and photovoltaic output, thus achieving the indirect evaluation of typical scenarios. Finally, the effectiveness of the scenario generation algorithm is tested through analyzing the direct and indirect evaluation results of the meteorological element generation scenarios. The case study results show that the typical scenarios generated based on the DCGAN model can effectively reflect the spatial distribution of wind-solar resources in the region, and the proposed method has good applicability in generating meteorological element scenarios.

Key words: renewable energy, joint scenario generation, scenario reduction, deep learning