基于信号稀疏化欠定求解的居民用户非侵入式负荷分解算法

Non-Intrusive Residential Load Decomposition Algorithm Solving Underdetermined Equations Based on Signal Sparsification

非侵入式负荷监测是实现智能用电的关键技术,负荷分解是实现该技术的重要环节。立足于非侵入采集模式下电流信号的欠定求解,研究了一种负荷分解算法。利用居民用户电器启动时间存在时间差的特点,将求解模型建立为单位时间间隔内仅从采集信号中分离两路信号,一路为新投入运行负荷的独立电流,一路为上一时刻的混合电流。该模型使每个投入运行的负荷均可独立分解,同时保证了欠定维度小,从而可有效求解。求解过程结合负荷电流信号的频域稀疏性,将欠定方程转化为最优化约束问题,通过两步迭代收缩阈值算法从混合信号中恢复出两路电流信号,并通过相似系数判断分离是否有效。利用实测用电数据验证了算法的有效性,通过正迭代方法得到当前参与运行的用电负荷,并根据相似系数确定负荷类型,实现了负荷辨识。

Non-intrusive residential load monitoring is a key technology for smart power. Load decomposition is an important part of realization of the technology. This paper studies a load decomposition algorithm resolving problem of morbid equation with current signals based on non-intrusive acquisition mode. Starting time points for different electrical appliances are different. Taking this as background, the solving model is set up to separate two signals from acquisition signals in unit time interval. One is mixed signal produced by previous device and another is independent running current signal produced by newly added load. With this model, each load put into service can be decomposed independently. It can ensure small underdetermined dimensionality and be solved effectively. In solution process, combined with sparsification of the load current signals in frequency domain, morbid equation is transformed into an optimization constraint. And the two signals are recovered from the mixed signal with two step iterative shrinkage threshold algorithm. Then similarity coefficient is calculated to estimate whether signal separation is successful. In order to verify effectiveness of the algorithm, real sampling load data are used. With positive iteration method, present power load is obtained. According to similarity coefficient, load types can be determined.