基于近邻传播聚类和遗传优化的非侵入式负荷分解方法

A Non-Intrusive Load Decomposition Method Based on Affinity Propagation and Genetic Algorithm Optimization

负荷监测和分解的研究具有广阔前景,将用电信息细化到内部用电细节,对智能电网中的双向互动、需求侧管理等具有较高应用价值,可为电力公司、用户等带来效益。针对现有非侵入式负荷监测和分解(NILMD)方法缺乏对用户用电特性的关注、多工作状态负荷处理的问题,细分用户用电模式,在对负荷按工/休日进行近邻传播(AP)聚类分析的基础上,增加设备工作状态合理性判断并结合遗传优化实现从总功率中分解出不同负荷设备及其对应工作状态。算例结果表明,所提方法可有效地辨识负荷类型和工作情况。该方法以稳态功率作为负荷特征,对一般的采样设备友好,数据获取较为简易,减少了硬件成本。

Research on load monitoring and decomposition has broad prospects.Electricity information can be refined to internal load composition details,which has high application value for the two-way interaction and demand-side management in the smart grid,achieving benefits to electric power companies and power consumers.However,the existing non-intrusive load monitoring and disaggregation(NILMD)methods lack the concern about electricity consumption characteristics and solution to multi-state electrical appliances.Thus,this paper considers the difference between load patterns,and then divides the loads of weekdays and weekends separately into certain clusters based on the affinity propagation(AP)method.After that,combining with the conditional judgment to working state rationality,this paper realizes the decomposition of different electrical appliances and the corresponding state-changes from the total load signal according to multi-feature genetic iteration.The case study shows that the proposed method can effectively identify various electrical appliances.The proposed method uses steady state power signatures,reduces hardware cost to some extent,and is friendly to the general sampling equipment.