Electricity Price Forecasting for Norwegian Day-Ahead Market using Hybrid AI Models

With the deregulation of the power market, competitive auction-based market platforms were formed. Besides increasingly decentralized and decarbonized characteristics of the interconnected European power market and system, the prices are becoming more vulnerable to external instabilities, resulting in the recent events of high and unstable prices. This situation increases the need for improved predictability for market participants, and thus for well-functioning forecasting tools. Digitalization technologies such as artificial intelligence (AI) and machine learning (ML) offer a large spectrum of solutions for the power sector. Energy forecasting is one of those high potential areas where AI and ML can be utilized to provide added value to the entire energy value chain. Hence, in this paper, artificial neural network (ANN) and long short-term memory (LSTM) network are implemented along with a hybrid solution of clustering and modifications such as including an additional hidden or dropout layer. The hourly Norwegian day-ahead electricity area prices of 2021 are predicted, and the results reveal the ANN model performs better for certain zones, while LSTM improves the forecasting accuracy in other zones, and how the additional modifications contribute less to improve the results even further.