Adaptive Distributionally Robust Unit Commitment Based on Non-parametric Statisticas

Aiming at the large-scale random power supply being integrated into the power grid, this paper builds an adaptive distributionally robust unit commitment model (ADRUC) to solve the optimal scheduling problem of the power system under uncertain operating conditions. First, the ambiguity set is built based on confidence bands for cumulative distribution function (CDF) from non-parametric statistics, and the interval of acceptable wind power output is obtained, then a polyhedral uncertainty set is constructed considering both the range and temporal domains. Then, in view of the flexibility and economy of fully adjustable robust optimization (FARO), the primal problem is divided into day-ahead UC master problem and the sub-problem of economic dispatch under the worst-case scenario. The strong duality theorem and the Big-M method are employed to transform the sub-problem into a MILP problem, the column and constraint generation (C & CG) algorithm is iterated repeatedly to obtain the target solution. Finally, the ADUC model is analyzed and validated by a ten-machine system.