Classification of Power Quality Disturbances Based on Multi-domain Parallel Feature Extraction and Optimization Strategy

With the extensive application of renewable energy generation and lots of new nonlinear loads such as electric vehicles, the power quality disturbances (PQDs) in the power grid become more complex and diverse. So how to extract complex features from big data of various PQDs and improve the classification accuracy by artificial intelligence technology has become a key issue in the field of power quality analysis. This paper presents a classification method based on multi-domain parallel feature extraction and optimization strategy. Multi-domain feature extraction is proposed to overcome the disadvantage of traditional feature extraction methods and enhance the diversity and completeness of original feature set. Feature selection strategy based on mutual information and fisher sore is used to reduce the cross-coupling of the original features, which obtains an efficient feature subset. Finally, the classifier based on extreme learning machine (ELM) is trained by optimized feature set and outputs the classification results. Numerical simulations illustrate efficiency and feasibility of the method.