Research on non-intrusive load disaggregation method based on multi-model combination

Abstract

Non-Intrusive Load Monitoring (NILM) technology is used to obtain the detail of household electricity consumption by analyzing total electricity consumption data without installation of any sensors. It is of great significance to power demand-side management. A new load disaggregation method based on combinatorial optimization is proposed in this paper. Firstly, the particle swarm optimization (PSO) algorithm is improved by applying time probability to the fitness function. Meanwhile, the iterative process of PSO is constrained to avoid falling into local optimal solution. The improvement is verified by experiments on the Reference Energy Disaggregation Data Set (REDD). And then, on the basis of a single model and with the idea of ensemble learning, this paper studies the multi-model combination load disaggregation method that combines the improved PSO mentioned above, integer programming (IP) algorithm and delta feature (DF) model in parallel, and the experimental group is designed through the control variate method. Experimental results show that the multi-model combination method has a good performance with a high average accuracy of 87% for load disaggregation.