Electrical Price Prediction using Machine Learning Algorithms

Abstract

Electricity is a derived energy source generated from renewable sources such as biomass, wind energy, and solar energy and nonrenewable sources like coal, fossil fuel, and natural gas. Electricity cannot be stored in large amounts and the power system stability requires a constant balance between production and consumption. At the same time, electrical energy demand depends on meteorology, climate, and the intensity of business and everyday activities such as on-peak, off-peak hours, and weekdays and weekends. These unique characteristics lead to price oscillations and fluctuations not observed in any other market, exhibiting daily, weekly, and often annual seasonality and abrupt, short-lived, and generally unanticipated price spikes. For power plants operating in competitive regions, an accurate prediction of electricity prices under conditions for the days ahead is essential. The correct electricity price forecasting would help in better management of the resources required for the generation of electricity. Therefore, in this chapter, electricity price forecasting is done using machine learning algorithms with an open-source dataset. A web tool, Jupyter notebook, is used for the prediction and visualization of electricity prices better than the already forecasted values in the dataset. Electricity price forecasting is a time-series problem. Hence, univariate models such as autoregressive (AR) moving average (MA), autoregressive moving average (ARMA), autoregressive integrated moving average (ARIMA), and long short-term memory (LSTM) are used for price forecasting. Multivariate forecasting models such as deep neural networks (DNN), convolutional neural networks (CNN), CNN-LSTM, and CNN-LSTM-DNN are also applied to predict the electricity prices. Subsequently, the results obtained are compared using mean absolute error (MAE) and mean absolute percentage error (MAPE) metrics to determine the accuracy of values obtained.