Probabilistic Feasible Region Characterization of Active Distribution Networks Driven by Data-Model Fusion

Abstract

With the large-scale intergration of distributed energy resources (DERs) into distribution networks, the traditional paradigm of regarding the distribution networks as a static parameter load has become increasingly obsolete. However, this shift has highlighted the necessity of characterizing the equivalent model of active distribution network (ADN) amidst profound uncertainty. Therefore, the concept of probabilistic feasible region (PFR) considering the stochastic characteristics and temporal-coupling characteristics of DERs is proposed in this paper, enabling ADN to provide equivalent models with different confidence levels for power system operators (PSOs). Based on chance constraints programming and feasible region projection theory, we theoretically derive the characterization method of PFR, which can be characterized as a constraint set formed by the extreme points of the dual space of the ADN optimization model under different confidence interval. To overcome the inefficiency of analytical methods, an intelligent method driven by data-model fusion is proposed to accurately and efficiently characterize PFR. Furthermore, a fused neural network algorithm is employed to map the relationship between operational data and security constraints, and the loss function is improved according to the results of the theoretical algorithms to correct the model, which avoids the problem of violating security constraints due to overgeneralization. Case studies based on a modified IEEE 33-bus distribution system validate the effectiveness and computational efficiency of the proposed method.