A Simple Fault Diagnosis Method for Power Switch System Based on Mean Conditional Entropy

This paper presents a simple open-circuit fault diagnosis method based on conditional entropy for two-level three-phase voltage source inverter(VSI), which is a typical power switching system. Referring to the concept of entropy in information theory, this paper firstly divides the amplitude of current signal into several intervals, and counts the frequency of phase current data falling into each interval in a current period. According to the frequency, the corresponding probability of each interval under normal and fault modes is estimated to calculate the conditional entropy between two phases in the switching system. Besides, in order to make up for the deficiency of traditional conditional entropy, the concept of “mean conditional entropy” is proposed. According to the difference of mean conditional entropy of phase currents in normal and fault modes, the threshold value is selected. Then, whether a fault occurs in the inverter can be judged. The proposed method can locate to the fault leg effectively with small amount of calculation. Simulation experiments show its advantages of strong resistance to both load disturbance and speed fluctuations.