Since the hard clustering method is susceptible to interference factors such as voltage and current in the non-intrusive load identification, this paper proposes a load identification method based on discrete fuzzy numbers. This method takes the characteristic of active load P and reactive power Q as the starting point. And it uses the finite links in discrete fuzzy numbers as the basis of evaluation level to construct an evaluation method based on probability and statistics. Then it converts the electrical equipment with similar characteristics into a discrete fuzzy matrix, and combines the matrix centroid and the ratio of the evaluation criteria to form the final evaluation value, thereby realizing the load identification. Compared with the hard clustering method, this method does not rely solely on the P-Q dimensional load characteristics. Instead, it uses the evaluation method of the load object to obtain the identification result. Finally, it is proved by real experiments that the results obtained by this method are consistent with the actual load switching results and have certain anti-interference ability.