Research on Response Characteristics of the Load

Abstract

Electric load forecasting plays an important role in the safe dispatch of power systems and improving the economy of power system operation. With the development of demand response, electric vehicles, electric heating, and some industrial loads become excellent response resources and require more accurate load forecasting methods. Electric load parameters are influenced by multi-dimensional factors. In order to fully exploit the time-series features in electric load data and improve the accuracy of electric load forecasting, this paper proposes a method for load forecasting based on LightGBM feature selection and improved Transformer model. Using the electricity load data from the Tianjin Smart Energy Service Platform as the data set, the input features of the load are firstly divided into temporal features, historical features, weather features and input load features, and then the features are multi-optimally filtered using the LightGBM model to select the features with higher relevance, and finally the improved Transformer model is used for load forecasting. The method is compared with commonly used forecasting models using actual arithmetic examples, and the results of the arithmetic examples prove that the method can improve the forecasting accuracy of electricity load data.