Optimal strategy to select load identification features by using a particle resampling algorithm

Abstract

This paper proposes a robust strategy to select the load identification features, which is based on particle resampling to promote the performance for the successive load identification. Firstly, the sliding window incorporated with the bilateral cumulative sum control chart (CUSUM) method is utilized to obtain the load event. Then, the minimum inner-class variance, using the time-serial data, is introduced to judge the happened time as precise as possible, thus marking the changing point of the state of load for the following feature extraction. Due to the fluctuating data of current and voltage sampled by the monitoring device, the particle resampling method, containing the importance principle, is applied to find the steady and effectiveness point, ensuring that the obtained features have the desired fit with its actual features. The fitness measurement is then carried out by using the 2-D fuzzy theory. Finally, the proposed method was tested on the real household measurements in the labs. The result demonstrates an improvement in obtaining the desired load features when applied to the real household for the following load identification.