Non-intrusive load monitoring method is a critical technology for realizing power system intelligence

which helps to optimize energy management and promote efficient energy utilization. In order to cope with the existing problems of feature redundancy

limited recognition accuracy and computational inefficiency

a novel non-intrusive load identification method based on optimal signatures and improved random forest is proposed. Firstly

the optimal feature combination is autonomously determined by recursive feature elimination method to reduce the information redundancy. Then load identification is realized by constructing a weighted random forest model. The weights are established by utilizing out-of-bag data. Construction parameters of random forest are optimized using the improved whale algorithm. Ultimately

the experimental results prove the accuracy and superiority of the proposed load identification method.