A non-intrusive method of industrial load disaggregation based on load operating states and improved grey wolf algorithm

Abstract

As for the internal load of electricity users, it can be determined by non-intrusive load monitoring (NILM), which is conducive to peak-load shift in the power grid. At present, the research on NILM focuses mainly on residential appliances, while industrial loads that consume large amounts of electricity may have more effective regulation capabilities in the face of large fluctuations in grid power brought about by a great deal of new energy connected to the grid, however, there is still little research conducted on the non-intrusive monitoring of industrial loads. In this paper, our consideration is given to the fact that industrial loads, compared to residential loads, operate steadily in the same state for a long time. Then, a non-intrusive industrial load disaggregation method is proposed on the basis of load operating states and improved grey wolf algorithm. On the one hand, this method takes into account the different times of load operation in different states, with the expanded label method proposed to improve the initialization of the population. On the other hand, median filtering is performed to process the results of first-generation load disaggregation in two times. Moreover, a strategy is proposed for multiple disaggregation of a single power according to this method. In addition, the search mechanism of the grey wolf algorithm is improved through the beta distribution of the three-point estimate and the genetic algorithm. According to the experimental results, the proposed method is effective in improving the accuracy of load disaggregation, the impact of high-power load on low power load disaggregation is mitigated to a certain extent, and the accuracy of each load reaches a minimum of 90%.