基于SAGA-FCM算法的非侵入式负荷监测方法

Non-intrusive load monitoring method based on SAGA-FCM algorithm

针对现有的非侵入式负荷监测(NILM)方法对小功率设备识别准确率不够,以及监测数据量过大时,准确率下降严重等问题,提出一种新颖的非侵入负荷监测方法。该方法以模糊C均值聚类算法(FCM)为基础,采用差量特征提取法提取任意时刻的特征变化值,引入模拟退火算法(SA)和遗传算法(GA)对聚类过程进行优化,实现了多类型电器负荷的聚类识别。实验数据表明,随着监测数据量的增加,该方法最终目标函数始终小且稳定,具有较好的稳定性和可靠性,适用于NILM大数据监测环境,采用谐波特征后识别准确率有一定的提升。

A novel non-intrusive load monitoring(NILM)method is proposed to solve the problems of low accuracy of lowpower electrical apparatus identification and serious decline of recognition accuracy when the monitoring data volume is excessive. On the basis of the fuzzy C-means algorithm(FCM),the delta feature extraction method is adopted in the method to extract the feature variation value at any time and the simulated annealing algorithm(SA) and genetic algorithm(GA) are introduced to optimize the clustering process. Thus,clustering identification of multiple types of electrical loads is achieved. The experimental data show that the final objective function of this method is always small and stable even if the increase of monitoring data volume, the method has better stability and reliability and is suitable for NILM big data monitoring environment,and the recognition accuracy of this method is improved after harmonic characteristics are adopted.