Forecasting short-term electricity prices in a deregulated electricity market is challenging due to the inherent uncertainty and volatility of the prices, often exacerbated by unexpected events in generation and unpredictable price spikes coupled with unclear price patterns. As a response, this research introduces a novel hybrid Deep Learning model that employs both a Convolutional Neural Network (CNN) and a Gated Recurrent Unit (GRU) to forecast one-step, two-step, and three-step ahead electricity prices. The proposed architecture consists of three consecutive CNN-GRU models, each modelling the input at a different time granularity. By down-sampling input data using pooling layers at the beginning of two model streams, the model simultaneously captures differing frequencies of price patterns. Additionally, the forecasting models consider external variables, including previous prices, electricity load, generation, import and export, and weather data, to assess their potential to enhance model efficiency. This approach is designed to tackle the intrinsic challenges of electricity price forecasting by leveraging the strengths of CNNs for spatial pattern recognition and GRUs for capturing temporal dependencies, thus providing a more complete view of the price trends. Three studies for different weeks of 2022 were carried out in the Ontario electricity market to assess the model. The results indicate that the proposed model reduces the forecasting error by 63.3% in the first experiment, 41.8% in the second, and 28.2% in the third, on average. The proposed model outperforms several baseline models, including statistical time-series (Auto-regressive (AR), Auto-regressive Integrated Moving Average (ARIMA), and vector autoregression (VAR)), Machine Learning (Linear Regression (LR), Support Vector Regression (SVR), k-Nearest Neighbors (KNN), and Decision Tree (DT)), and Deep Learning (Multi-Layer Perceptron (MLP), Long Short-Term Memory (LSTM), GRU, and hybrid LSTM-GRU) models.
Read full abstract