Short-term Load Forecast of Integrated Energy System in Multi-station Fusion Scenario Based on Robust Model

Abstract

Integrated Energy System (IES) is the development direction of energy consumption in the future. Through the coordinated design and planning of energy systems such as electricity, cooling, and heat, energy utilization can be effectively improved and the development of renewable energy units can be promoted. The purpose of this paper is to predict the short-term load of an integrated energy system in a multi-station fusion scenario based on a robust model. Convex Quadratic Loss Functions to Suppress Negative Effects of Outliers Non-convex Quadratic Loss Functions limit the maximum loss penalty for outliers. First, the loss function is expressed as the difference of two quadratic functions, and the corresponding robust model is built. Second, the optimization problem corresponding to the robust model is transformed into a system of linear equations using the CCCP technique and KKT conditions. Finally, the prediction accuracy of RLS-SVR model, SVR model and BP neural network model is compared. The results show that for cooling load prediction, the average relative error of the prediction model constructed by RLS-SVR is 1.05% and 1.12%, which is lower than other models.