Research on load forecasting of integrated energy system based on multi-dimensional factor uncertainty

Abstract

Load forecasting is the basis of power network planning and power market transaction. Aiming at the characteristics of load driven by multi-dimensional factors and strong uncertainty, a long-term load probability prediction model based on non-parametric combined regression was proposed. Through Granger causality analysis, the multi-dimensional variables driving the long-term development of load are preliminarily screened. In order to improve the prediction accuracy, nonparametric combination regression modeling was carried out for the selected variable set based on the stepwise average combination to realize the optimal combination model and integrate the dynamic driving of each variable to the long-term load. Based on the random rate of change, the uncertainty modeling of multi-dimensional variables contained in the optimal combination model was carried out and applied to the probability prediction of long-term load to obtain the loci values of 10%, 50% and 90% of long-term load. Non-parametric combined regression model not only has high accuracy, but also can realize long-term load probability prediction combined with multi-dimensional variable uncertainty modeling.