Price Spike Classification and Regression Using A Hybrid Oversampling Method

Price spikes are types of electricity prices that are much higher than normal electricity prices. However, due to low occurrence, price spikes are hard to be predicted by normal price forecast methods, which are designed to model patterns from the dataset with massive normal price cases. To facilitate price spike classification and regression, in this paper, a hybrid oversampling method is proposed to increase the number of price spike cases. As electricity prices are time series data, to enable that the synthetic price spike cases have the similar structure as the real ones, the enhanced structure preserving oversampling technique is applied to conserve the temporal relationships. Also, the synthetic minority oversampling method for regression is utilized to ensure the authenticity of the oversampled price spike values. Numerous case studies are used to evaluate the effectiveness of the proposed hybrid oversampling method, and the promising results demonstrate the capability of the proposed method to improve price spike classification and regression accuracy.