Day-ahead interval scheduling strategy of power systems based on improved adaptive diffusion kernel density estimation

Abstract

The increasing penetration of renewable energy causes high uncertainty, which further complicates the optimal scheduling operation of power systems. The uncertainty of renewable energy output is represented by empirical prediction intervals in the traditional interval optimization scheduling model, but the interval range is not precise enough. To address this problem, a novel improved interval optimization method is proposed in this paper. First, the improved adaptive diffusion kernel density estimation (IADKDE) is used to obtain more accurate intervals for the renewable energy output. Furthermore, a data-driven adaptive optimal bandwidth selection is adopted to select the optimal bandwidth instead of normal reference rules in IADKDE. In addition, the day-ahead scheduling optimization model is developed by IADKDE considering the driving requirements of electric vehicles (EVs) owners. The proposed model is described in details and solved by interval linear programming method. Finally, the effectiveness and accuracy of the proposed method are validated, and the comparative analysis with interval optimization by empirical prediction intervals, extreme learning machine(ELM) and stochastic optimization is given. It is demonstrated that the proposed method can obtain more accurate interval ranges of uncertain variables and has strong applicability.