A CL-MDT Method of Multi-energy Load Forecasting in Integrated Energy System

Source-load scheduling is based on multi-energy-load forecasting. The integrated energy system mainly includes three types of energy: electricity, cooling and heating. Studying the correlation among electricity, cooling and heating may improve the accuracy of multi-energy-load forecasting. This paper considers the correlation of three energy sources, fully analyzes the correlation, and applies the correlation of the three in the forecasting model. This paper proposes a CL-MDT (CNN-LSTM-Multi-Decoder-Transformer) model for multi-energy-load forecasting. The model is based on the Transformer, and the Multi-Head Attention part in the Encoder is replaced by a 2dimensional 3*3 CNN (Convolutional Neural Network) feature extraction module for feature extraction of data. And a 1dimensional CNN feature extraction module and LSTM structure are added to the Decoder. The structure of single Encoder and multiple Decoders is used in this paper to realize the application of the correlation of the three in the forecasting model. Finally, the model is tested on public datasets and the forecasting results of CL-MDT are compared with that of LSTM model for multi-energy-load joint forecasting. The results show that the CL-MDT model proposed in this paper has better forecasting accuracy.