Knowledge-Inspired Data-Aided Robust Power Flow in Distribution Networks With ZIP Loads and High DER Penetration

Characterized by increasing penetration of distributed energy resources, active distribution networks necessitate developing uncertainty-adaptive power flow (PF) algorithms to cover broad operating conditions. Despite the success of data-driven methods in improving such adaptivity, the efficacy of these methods relies heavily on large, precise, and outlier-free datasets, which limits their materialization in practical grids. To address these dual issues, this paper proposes a knowledge-inspired data-aided robust PF algorithm in unbalanced distribution systems with ZIP load models and high penetration of distributed energy resources. The proposed method first uses Taylor expansion to derive an explicitly analytical linear solution for the PF calculation. A data-driven support vector regression-based method is further proposed to mitigate the approximation loss of the linearized PF model, which might surge in widening voltage variations. Inspired by physical knowledge of distribution system operation, the proposed method can adapt to a wide range of operating conditions without retraining and thus can be applied to passive/active distribution networks. Case studies in the IEEE 13- and 123-bus unbalanced feeders illustrate that the proposed algorithm exhibits superior computation efficiency and guaranteed accuracy, under variable penetration levels and lightweight datasets.