Identification of nonlinear load power using wavelet neural networks

In this paper, a Dmeyer wavelet function is utilized to decompose the mixed load current waveforms and extract the parameter values which represent the nonlinear loads. A three layer BP neural network model is established, which is trained using Levenberg-Marquardt (LM) algorithm which shows fast convergence and strong stability. The results indicate that the proposed method of the wavelet BP neural network load identification has good practicability and reliability. The scientific management of university student's apartments is realized. This study is very important for eliminating the campuses fire hidden trouble.