Forecasting of Day-Ahead Electricity Price Using Long Short-Term Memory-Based Deep Learning Method

Abstract

Short-term price forecasting (STPF) is a crucial issue for market participants in the day-ahead electricity market. Producers need to forecast electricity prices in the short term to decide to supply bids and make trading strategies. Large consumers and electricity suppliers should protect themselves from electricity price volatility, so they should have a powerful forecasting tool for future electricity price values. By considering the importance of STPF, we propose a new model to predict electricity price in this study. This model consists of long short-term memory deep learning architecture, discrete wavelet transform and average true range. The proposed approach analyzes the input time series with different scales and processes each by a different model. The final forecasting is obtained by superimposing the forecasting results of all the components. The performance analysis is conducted on the data taken from one of the U.S. state hubs from April 1, 2013, to December 1, 2014. The forecasting model utilizes historical price, load, and wind data as input features. The performance analysis shows that the proposed model predicts one-day ahead electricity price more accurately than the compared methods. © 2022 Elsevier B.V., All rights reserved.