遗传算法优化的BP神经网络在居民负荷分类中的应用

Residential Load Classification based on Genetic Algorithm Optimized BP Neural Network

非侵入式负荷监测的性能取决于所选电器特征的唯一性,特征选择是非侵入式负荷监技术的重要步骤。本文提出一种基于遗传算法优化的BP神经网络的居民负荷分类方法,采用遗传算法对BP神经网络的初始连接权值和阈值进行优化,利用电器消耗类型特征实现负荷分类。算例分析结果表明,本文的负荷分类方法对常用电器识别分类准确性高,具有负荷特征提取简单的优点,对非侵入式负荷监测技术的低频特征负荷分解研究具有有效的指导和参考作用,提出优化的BP神经网络算法较原始算法表现出更优的分类性能。

The performance of non-intrusive load monitoring depends on the uniqueness of the selected appliance characteristics,and feature selection is an important step in non-intrusive load monitoring techniques.This paper uses Genetic Algorithm(GA)to optimize the BP neural network and apply it to appliances identification.By comparing the classification performance of typical low frequency features in common electrical appliances,the classification performance of various low frequency feature subsets of electrical appliances is verified.The superiority of the proposed classification method is verified by comparison with BP neural network.