MF-LSTM for fault diagnosis of new energy electrical equipment

Abstract To enhance the accuracy and reliability of fault diagnosis in new energy electrical equipment, we present a fault diagnosis method based on MF-LSTM (multidimensional fuzzy long short-term memory). It incorporates thermal imaging feature distance optimization for the separation of thermal map feature patterns, then introduces a logic sequence optimization structure to enhance the adaptability of the LSTM model. Additionally, a multifuzzy algorithm is introduced to improve the model’s global understanding of fault features. Results demonstrate that it enhances fault diagnosis accuracy, achieving a precision of 92.31%, surpassing other algorithms and exhibiting superior recognition performance across various fault categories.