Moving Average-Based Performance Enhancement of Sample Convolution and Interactive Learning for Short-Term Load Forecasting

Abstract

Efficient and accurate short-term load forecasting (STLF) is significance in modern electricity markets. However, accurate short-term load forecasting is challenging due to the non-stationary power load patterns. In this work, we propose a short-term load forecasting framework based on maximal information coefficient (MIC), moving average filter (MAF) and sample convolution and interactive learning (SCINet), Firstly, MIC is used for feature selection. Secondly, the filtered input features are decomposed using MAF individually. Finally, the data are used in an advanced SCINet for short-term load forecasting. The performance of the proposed method is evaluated using datasets from two different regions of the US electricity market. In addition, we compare the prediction results with support vector regression machines (SVR), long short-term memory networks (LSTM), temporal convolutional networks (TCN), light gradient boosting machine (LightGBM), artificial neural network (ANN), random forest (RF), and sample convolution and interaction networks (SCINet). The proposed model achieves accurate prediction results among all the machine learning models used in this paper.