Evidential Reasoning for Enhanced Node Selection in Power Network Reduction: a Complex Network Perspective

An effective node selection criterion (NSC) is paramount for the efficient reduction of networks in power systems. Unlike prior studies that primarily focus on structural attributes, this paper introduces a comprehensive NSC that encompasses both structural and electrical characteristics. This NSC is formulated using the evidential reasoning approach. Within this framework, quasi-one-hot encoding plays a pivotal role in establishing the evaluation grades for various criteria and attributes. Subsequent integration of these criteria is achieved through multi-evidence reasoning. Ultimately, a utility evaluation refines this into a definitive, quantitative NSC. The evidential reasoning can accommodate multiple criteria and address inherent uncertainties in the evaluation process. By merging the proposed NSC with existing model reduction techniques, more accurate reduced models can be derived. Case studies based on a standard power grid corroborate the applicability and effectiveness of the proposed NSC.

I. Introduction

Network reduction is crucial to the analysis of the interconnected power gird. The goal of the network reduction or equivalent is to represent the original power grid with a smaller equivalent model, which can enhance the computation speed, alleviate the memory requirement, and improve the analysis efficiency [1]. The applications of network reduction are notably widespread, including but not limited to the development of novel power grid planning tools, static security analysis, dynamic stability analysis, reactive power optimization, and assessments for integrating renewable energy sources. The critical and indispensable initial step in network reduction is the strategic selection of nodes to be pruned. Past research on node selection has often been case-dependent, lacking in quantitative standards, offering limited transferability, and usually focusing on a singular metric. This paper introduces a quantified, electric characteristic-aware, multi-index approach to node selection for reduction. This solution aims to tackle the aforementioned challenges by providing a quantifiable guideline for network reduction efforts.