Non-Intrusive Load Monitoring and Identification Based on Maximum Likelihood Method

Abstract

Aiming at the problem of low identification accuracy for non-intrusive load identification technology at low sampling rate, a novel load identification algorithm based on maximum likelihood method is proposed for load monitoring and identification. Firstly, the event detection algorithm is proposed based on the load operation to judge load switching event. Then, the possibility of the load is determined by the initial identification of integer programming in order to reduce the dimension of the load identification. Secondly, the probability of each load is calculated by the maximum likelihood classification and the probability density function of the operation power, the change of operation power, operation time and closing time, the maximum probability is the result of load identification. Finally, through the simulation experiments of artificial load and load identification for different users, the average results of the identification accuracy are more than 80%, the reasonableness and immediacy of the algorithm are proved to meet the requirement of load identification at low sampling rate.