Abstract:Non-intrusive load monitoring technology can guide users to reasonably arrange electricity consumption and reduce electricity consumption expenditure for households. Meanwhile, power departments can understand the law and trend of load electricity consumption and improve power planning by using household electricity consumption data. Most of the existing studies use advanced intelligent algorithms to learn load characteristics. In view of the shortcomings of existing algorithms in identifying features, a BP neural network load identification method based on steady-state waveform decomposition is proposed. This algorithm mainly use the stably waveform superposition to extract the harmonic features of the decomposed current waveform. Combined with the neural network training, the weights and thresholds are obtained, and the embedded load is recognized. The method has been successfully implemented on embedded devices and the desired results have been obtained.