Comprehensive analysis of admittance matrix estimation considering different noise models

Abstract

The previous knowledge of the admittance matrix represents an important issue in power system analysis, specifically regarding load flow, voltage stability, and protection setting. Some parameter estimation techniques in technical literature determine the admittance matrix of electric power grids, leading to notable advances in measurement and monitoring. This paper proposes a robust approach to determine the admittance matrix using deep learning techniques. Throughout the paper, results demonstrate that the proposed approach handles Gaussian and non-Gaussian noise reliably, outperforming other works in the technical literature. This paper also evaluates the proposed method in several scenarios, including different numbers of samples and varying noise level, as well as loads with non-linear variations. The proposed method has low computational complexity because it considers only a few features while estimating admittance parameters. Results demonstrate that the proposed approach sustains accuracy and robustness, even when subjected to high noise levels in the measurements. This paper evaluates the proposed approach by estimating the parameters of the IEEE 14-bus and 57-bus systems and presents the performance of all parameters for the 14-bus system.