Scholarlink

<p>Wind energy, as one of the renewable energy sources (RESs), is becoming increasingly popular. However, its operating volatility and randomness pose significant challenges to the stable operation and economic dispatching of the power grid. To characterize the uncertainty of wind power, existing scenario forecasting models often ignore the temporal correlation between data. In this paper, a model combining generative adversarial network (GAN) and long short-term memory (LSTM) is proposed to character the time series properties of wind power and generate forecast scenarios that conform to the distribution of measured data. The LSTM network is used as the generator and the discriminator in the GAN. LSTM can capture the temporal correlation of time series, and GAN can generate real samples. The proposed method combines these two characteristics and generates realistic forecast scenarios. The feasibility and superiority of the proposed method are verified using wind power data from the National Renewable Energy Laboratory (NREL). The simulation results show that the proposed method can correctly capture the characteristics of wind power and can easily generate the required forecast scenarios.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Wind power scenario forecasting based on combination of generative adversarial network and long short-term memory network

  • 期刊论文

  • 发表日期：

    2025年5月4日

查看全文
  • Fangjun Jin,
  • Congmei Jiang,
  • XianKui Wen

摘要

Wind energy, as one of the renewable energy sources (RESs), is becoming increasingly popular. However, its operating volatility and randomness pose significant challenges to the stable operation and economic dispatching of the power grid. To characterize the uncertainty of wind power, existing scenario forecasting models often ignore the temporal correlation between data. In this paper, a model combining generative adversarial network (GAN) and long short-term memory (LSTM) is proposed to character the time series properties of wind power and generate forecast scenarios that conform to the distribution of measured data. The LSTM network is used as the generator and the discriminator in the GAN. LSTM can capture the temporal correlation of time series, and GAN can generate real samples. The proposed method combines these two characteristics and generates realistic forecast scenarios. The feasibility and superiority of the proposed method are verified using wind power data from the National Renewable Energy Laboratory (NREL). The simulation results show that the proposed method can correctly capture the characteristics of wind power and can easily generate the required forecast scenarios.