Study On The Problem Of Transient Power Quality

The problem of power quality has aroused great concern of power companies and consumers. Recognizing, locating and classifying the problem are of great significance. The paper analyzes automatic recognition, location and classification of transient power quality problem, and some improvements have been achieved.Different types of load will change the resonant frequency of power systems to different ways. According to this characteristic, a recognition algorithm using the entropy feature vectors of the wavelet transform special layer is developed for capacitor switching disturbance at low voltage, furthermore, the location of the disturbance can be determined accurately.The disturbance power and energy is a commonly used method to locate the power quality disturbances. But it is susceptive to other disturbances occurring at the same time and it makes mistake when the disturbance injects energy to the system. This paper considers only the initial disturbance energy and voltage as criterion, which lessens greatly the impact of other disturbance and obviates the need for the classification before the location. All types of disturbances are summarized and a right location method is proposed.When classifying the disturbances, the linear decision method and K-nearest neighbor method have the same shortcomings: enormous amounts of computation and storage as well as impaired classification accuracy. This paper improves the Bayes optimal method, which changing it from the parameter classification method to a non-parameter one and making it applicable to the classification of limited intercrossed samples. Based on the conclusion, this paper improves the methods of linear decision and K-nearest neighbor, overcoming their shortcomings and improving the computation accuracy. When applied to the Probabilistic Neural Networks, the approach improves its two inherent shortcomings.