基于特征分解的虚拟电厂运行负荷预测方法研究

Research on Operating Load Forecasting Method for Virtual Power Plants Based on Feature Decomposition

摘要:

Virtual power plants (VPPs) integrate various distributed energy resources to create an efficient energy management system, which can swiftly respond to the demands of power grid, reduce operational costs and enhance the overall reliability and stability of power supply. However, the management of power supply in VPPs is challenged by the volatility, stochasticity, seasonality and substantial peak-valley variations inherent in electricity load dynamics. To address these challenges, an operating load forecasting method for virtual power plants based on feature decomposition is proposed, which provides essential information support for both power dispatching and demand response within VPPs. Firstly, seasonal-trend decomposition using LOESS (STL) is employed to capture trend and seasonal variation characteristics in virtual power plant load data, thereby extracting historical load resource features. Secondly, a load forecasting model based on long short-term memory (LSTM) network is developed by integrating historical load resource data with external influencing factors. Finally, experimental verification is carried out using operational data from virtual power plants. The results show that the determination coefficient R2 of the proposed method is 0.98, which can accurately predict the power plant load.

Abstract:

虚拟电厂; 分布式能源资源; 负荷预测; 长短期记忆网络; 特征分解