DEEP LEARNING APPROACH TO FORECASTING ELECTRICITY PRICE FROM LOAD DATA

Vladimir BABUSHKIN, Gheorghe CĂPĂȚÂNĂ State University of Moldova

Abstract

The accurate forecasting of electricity price and load is essential for maintaining a stable interplay between demand and supply in the dynamic electricity market. In this work we propose a deep Convolutional Neural Network-based model for day-ahead electricity price forecasting from historical price/load data and predicted load values. The model was tested on the data for New York and New South Wales and demonstrated high prediction accuracy for both datasets. Keywords: Deep Learning, Machine Learning, Long-Short Term Memory Networks, Convolutional Neural Networks.