Improving Frequency Stability Assessment through K-Nearest Neighbors and Machine Learning Techniques
Bwandakassy Elenga Baningobera, Irina Oļeiņikova · 2023
This paper proposes a method that can effectively impute the missing measurement values in order to address the issue of missing data using a combination of machine learning and physical constraints. The high-resolution data about the state of the power system can be provided by synchrophasor measurements obtained from Phasor Measurement Units (PMUs). However, due to various factors such as communication failures, some PMU measurements may be lost, which can lead to inaccurate modeling and control of the system. The paper proposes a deep learning-based imputation framework that takes advantage of the temporal and spatial correlations in the PMU measurements to overcome the challenge and fill in the missing values. Physical constraints such as the power flow equations are incorporated in the framework to ensure that the imputed data is consistent with the underlying physics of the system. The paper presents experimental results on a real-world power system dataset, which demonstrates that the proposed method can effectively impute missing measurements data and improve the accuracy of power system models. The proposed approach is successfully implemented and validated on the Nordic 44-bus test system for evaluation.