Fault Diagnosis Of Cascaded H-Bridge Power Conversion System Method Study

Power conversion system,as an important component of energy storage systems,are important equipment for achieving complementary output from wind,light,and storage,and solving the problem of new energy security and consumption.The cascaded H-bridge energy storage converter has many advantages such as small battery commutation,high cycle efficiency,and easy expansion,making it the object of research and engineering practice by many scholars.However,the cascaded H-bridge power conversion system has a large number of transistors and is susceptible to external environmental influences during long-term operation,making the converter prone to malfunctions.Once a transistor fails,if it is not disposed of in a timely manner for a long time,it will cause distortion of the output current on the AC side,reduce power quality,and affect the normal operation of non faulty power transistors,further causing instability of the cascaded H-bridge transformer energy storage converter,and threatening the safety of the power system.Therefore,it is of great significance to take effective fault diagnosis and isolation measures for cascaded H-bridge power conversion system.Due to the high similarity of different fault behaviors in cascaded H-bridge power conversion system,the diagnostic effect of the same method applied to different voltage levels of energy storage converters is also different.Therefore,this thesis focuses on cascaded energy storage converters and studies fault diagnosis methods applicable to different voltage levels,providing theoretical support for improving the reliability and stability of energy storage systems.For low-voltage power conversion system,due to their low grid connection voltage level and limited number of cascaded H-bridge units,feature extraction and shallow learning methods can be used for fault localization.This thesis takes the seven level power conversion system as the research object,proposes a method of ensemble empirical mode decomposition with multi-scale arrangement entropy to extract similar fault features,and optimizes the support vector machine classification model using the sparrow algorithm.The main content includes:(1)Adopting the method of ensemble empirical mode decomposition to select the optimal mode to reduce the interference of on-site working condition noise on fault diagnosis;(2)Adopting multi-scale arrangement entropy for feature extraction of similar waveforms on the AC side,and then inputting the extracted fault features into the classification model to obtain the accuracy of fault diagnosis;(3)Create a semi physical experimental platform to verify the applicability of the method proposed in this article under off grid conditions.For high-voltage power conversion system,due to their high grid connection voltage level and large number of cascaded H-bridge units,the degree of fault similarity is further increased,and the accuracy of traditional feature extraction+shallow learning methods is low.Therefore,this thesis takes the highvoltage direct mounted power conversion system as the research object and proposes a multi signal source adaptive fusion fault diagnosis method for high-voltage Power conversion system.The main content includes:(1)Data processing,which converts one-dimensional time series voltage and current waveforms into two-dimensional timefrequency domain through wavelet transform;(2)Using convolutional neural networks(CNN)to extract fault features from multiple signal sources;(3)The fault features of multiple signal sources are weighted separately,and the weight proportion is adaptively adjusted based on the transformer fitting effect until the optimal fitting result is achieved.This thesis studies the fault diagnosis methods of two typical multi-level cascaded H-bridge Power conversion system under battery charging and discharging modes.The research results indicate that the proposed method for low-voltage fault diagnosis has high diagnostic accuracy for low-voltage cascaded H-bridge power conversion system in parallel off grid mode,but has poor diagnostic performance for high-voltage Power conversion system;The method of adaptive fusion of multiple signals from CNN transformers can adapt to fault diagnosis of various level power conversion system,and the diagnostic result is effective.