Aiming at the problem that the current non-intrusive load monitoring (NILM) method has insufficient recognition accuracy for low-power electrical equipment, an improved method is proposed in this paper. This method is based on improving the initial clustering center of FCM. In addition to using active power feature, the fundamental wave power factor and voltage-current third harmonic content difference are selected as new features. The gray wolf algorithm (GWO) and the simplex method (SF) are introduced to optimize the clustering process, and the number of load types is determined by fuzzy clustering to realize the load recognition. The experimental results show that the method has good robustness and high accuracy in different scenarios with the increasing of load types.