Forecasting Electricity Price Spikes in Competitive Markets Using a Hybrid Deep Learning Framework with SHAP and Grey Wolf Optimization

Accurate electricity price forecasting in competitive markets is crucial for both producers and consumers. Given the complexity and extreme volatility of electricity prices, predicting the occurrence and magnitude of price spikes is a significant challenge. This paper presents a novel hybrid strategy based on deep learning and the Grey Wolf Optimization (GWO) algorithm for forecasting both the occurrence and magnitude of electricity price spikes. In this approach, an initial feature analysis is conducted using the SHAP (SHapley Additive exPlanations) importance index to eliminate less influential features. Subsequently, a Convolutional Neural Network (CNN) is employed to extract complex features and identify spike patterns in time series data. The Recursive Feature Elimination (RFE) algorithm is applied to optimize input features. Finally, GWO is utilized to optimize the CNN weights for accurate spike magnitude prediction. The proposed method is evaluated using real-world electricity market data, and the results demonstrate its high accuracy in forecasting both the occurrence and magnitude of electricity price spikes.