Topology Guided Network Clustering and Allocation of Edge Estimation Points for Monitoring of Distribution Networks

Managing extensive measurement sets in the power system for state estimation (SE) leads to communication burden and delays. To address the above challenge, this paper explores the concept of edge computing, offering robust support for real-time SE. The method involves dividing the large-scale network into clusters, ensuring balanced cluster sizes in the three-phase distribution network, thereby reducing SE runtime. Additionally, edge estimation points (EEPs) are strategically allocated to minimize distances to the data sources and other EEPs, effectively reducing communication delays. Building upon this approach, a constrained multi-objective optimization problem is formulated with four objectives in the network, incorporating limited measurements from microphasor measurement units ( $\mu $ PMUs) and remote terminal units (RTUs). The paper introduces a modified Non-Dominated Sorting Genetic Algorithm II (NSGA-II), which combines guided radial mutation and trait crossover and utilizes partition topologies to expedite optimization. It reduces mathematical formulation complexity through gene regulation and real value encoding for improved efficiency. A comprehensive analysis of the IEEE 123 and 37-bus systems has been conducted, highlighting the effectiveness of the proposed formulations in addressing the specified objectives and adeptly managing associated trade-offs. The algorithm produces objective-aligned clusters with efficient convergence and diverse optimal solutions, resulting in enhanced SE runtime.