Energy market predictions with hybrid neural network 1D-CNN-BiGRU

Abstract

Electricity price forecasting is important for managing supply and demand, planning investments in energy projects, ensuring energy security and efficient use of resources. This paper presents a hybrid neural network of two types of neural networks: the convolutional neural network (CNN) and the recurrent neural network (RNN) for energy market data analysis to forecast electricity prices. CNN is used to extract features from the raw data by applying a convolution operation on the temporal axis with different filters. At the same time, bidirectional gated recurrent units (BiGRU) of RNN are used for subsequent analysis of the temporal dependence of the extracted features, allowing the historical data to be considered. BiGRU is particularly useful for applications where the current input depends on past and future contexts. Thus, the ID-CNN-BiGRU method allows effective time series analysis and prediction of future values and can be widely used in the tasks of forecasting electricity prices, stocks, traffic, and other time series. The results indicate that the presented model is promising for use in a highly dynamic energy market.