Calculation of line loss rate in transformer district based on improved k-means clustering algorithm and BP neural network

Abstract

Incomplete meter configuration, the difficulty of collecting operation data and vast numbers of components and nodes in low voltage transformer district lead to rather complicated work of calculation of line loss rate. To solve existing problems, a novel method of line loss rate calculation in transformer district was presented and realized by programming, which was combined improved K-Means clustering algorithm with BP neural network model optimized by Levenberg-Marquardt(LM) algorithm. Samples were classified by improved K-Means clustering algorithm according to electric characteristics. Thus, the numerical dispersion of line loss rate in transformer district was solved. On this basis, each class was trained by BP neural network optimized by LM algorithm. Variation of transformer district line loss rate was obtained by using BP neural network to map relation between line loss rate and electric characteristic parameters. 601 transformer districts in a region as an example, simulation and calculation were performed to verify the accuracy of the proposed method. The results show that the method has the advantages of fast convergence and high accuracy, compared to standard BP neural network.