Outlier-robust hybrid electricity price forecasting model for electricity market management

Abstract

Electricity market management is of great importance for cleaner production in the development of society. However, despite this significance, electricity price forecasting remains a challenging task. Hybrid models are widely employed for forecasting electricity price, which has the characteristics of being non-stationarity, random, and non-linear. Despite their success, current hybrid models require improvement. In particular, data preprocessing, artificial intelligence optimization, feature selection, and basic forecasting engine selection should be considered. In this study, in addition to these issues, we consider the negative influence of outliers on the modeling of electricity price. In particular, a novel outlier-robust hybrid model is developed for forecasting electricity price, which combines a basic forecasting engine called outlier-robust extreme learning machine model and three new algorithms. Specifically, a new optimizer called chaotic sine cosine algorithm is developed to obtain the ideal parameters for phase space reconstruction, and then a novel feature selection method is developed to construct the optimal features in the modeling of electricity price. Moreover, an effective data preprocessing method is proposed for effective forecasting by capturing electricity price features. Subsequently, experiments based on electricity price data from the electricity markets of Australia and Singapore demonstrate that the proposed model is superior to other benchmark models. Further, the model can be a reliable forecasting method not only in electricity market management, but also in modeling time series with complex nonlinear characteristics and outliers.