A Hybrid Machine Learning Approach for Price Forecasting in Electricity Market with Smart Bidding Strategies and Wind Energy Influence

Abstract

The price for renewable power has seen a dip in recent years with increased advanced approaches in the electricity market. For the benefit of both the producers and the consumers of electricity, a forecasting technique to predict the price for the next day is required to handle competitive market participation facilitating the prosumers. Therefore, the study’s main objective is to model a day-ahead electricity price and apply a forecasting tool based on time series analysis, promising accuracy, and minimal average forecast errors. Day-ahead real-time price data of the Nordic pool market forecast study for a period between January 2019 and March 2021 is considered for the study. The impact of wind energy production on day-ahead electricity Nord Pool electricity market for improved participation in the market is forecasted based on time series. Machine learning techniques, namely Support Vector Regression (SVR), Generalized Autoregressive Conditional Heteroskedastic (GARCH), Autoregressive Integrated Moving Average (ARIMA), Multiple Linear Regression (MLR), and hybrid of SVR-GARCH are analyzed in detail to find the best forecasting approach. The proposed forecasting technique on next-day electricity price series influenced by wind power production is validated using performance indices, namely, the Mean Absolute Percentage Error and the Root Mean Squared Error considered for checking the reliability of the proposed new model. A similar study on the Indian market was also conducted to validate the robustness of the model. The electricity price forecasting model is developed and analyzed using a machine learning technique in Python (Jupyter Notebook).