Outlier-immune Data-driven Linear Power Flow Model Construction via Mixed-Integer Programming

The common approaches to construct a data-driven linear power flow (DD-LPF) model cannot completely eliminate the adverse impacts of outliers in a training dataset. In this letter, a novel outlier-immune DD-LPF model construction method via mixed-integer programming is presented for automatically and optimally identifying outliers to form a more accurate LPF model. Two acceleration solution strategies are further suggested to reduce the computational time. Case studies demonstrate the superior accuracy and comparable computational time of the proposed method when compared to three common approaches.