A CNN and LSTM-based multi-task learning architecture for short and medium-term electricity load forecasting

Abstract

Electricity load forecasting is the forecast of power load in the future period based on historical load and its related factors. It is of great importance for power system planning, operation, and decision making. There is still an opportunity to improve the precision of forecasting using machine learning and deep learning models, according to previous research findings. In this research a deep learning framework based on multi-task learning (MTL), convolutional neural networks (CNN), and long short-term memory (LSTM) is suggested. By weighing the pertinent training data for the main task and the auxiliary task, the proposed hybrid deep learning network model MTMV-CNN-LSTM addresses the issues of too much repetitive data and poor convolution effect and significantly enhances the generalizability of the model. It uses the CNN layer to extract features from the input data and the LSTM layer for sequence learning. To forecast the short and medium-term electrical load, we created two different trials. The results show that our proposed method is a highly competitive new model with high priority compared to other models for both 10-day short-term and 3-month medium-term power forecasting tasks.