Non-Intrusive Load Monitoring Modeling and Analysis Based on Optimized K-Nearest Neighbor Algorithm

Aiming at the problem that the recognition rate of the K-Nearest Neighbor (KNN) algorithm is not high in the face of similar power fingerprints in non-intrusive load recognition, an improved KNN algorithm model based on the combination of transient steady-state features is proposed. The feature library is constructed by fusing transient and steady-state features, and the SVM (Support Vector Machine) algorithm is used to quantify the feature weights, combined with the KNN improvement method based on distance weights, which effectively improves the discrimination ability of similar samples. Experimental results show that the proposed improved KNN algorithm is better than the traditional method in load recognition accuracy, and provides a feasible solution for lightweight load identification in smart meters.