Abnormal detection method of photovoltaic grid-connected circuit breaker based on improved AHA-ELM

Abstract

To address the limitations of traditional detection methods for PV grid-connected circuit breakers, which struggle to extract and classify complex nonlinear features, an anomaly detetion method based on an improred AHA and an optimized ELM is proposed. Firstly, the factors affecting the health status of PV grid-connected circuit breakers are studied, the common fault types and fault characteristics of circuit breakers are analyzed, and appropriate characteristic signals are selected as high-quality data sets that can reflect the working status of circuit breakers. Secondly, dynamic flight strategy and chaotic mapping mechanism are introduced to optimize the AHA, and the improved AHA is used to optimize the input weight and hidden layer bias of the ELM, so as to enhance the learning ability of the model for high-dimensional complex features. Then, the improved AHA-ELM algorithm is used to quickly obtain the data characteristics of circuit breakers under various faults, and the abnormal detection model of PV grid-connected circuit breakers is trained to detect the abnormal state of circuit breakers. Finally, based on a practical distribution network with photovoltaic grid connection, the simulation result show that the proposed method is feasible and effective.

Keywords

photovoltaic grid-connected circuit breaker / anomaly detection / extreme learning machine / artificial hummingbird algorithm / chaotic mapping