Rapid diagnosis and active disturbance rejection control of compressor surge based on hybrid deep learning

Abstract

ObjectiveIn order to improve the safe and stable operation level of compressor equipment, this paper puts forward a rapid diagnosis method of surge states based on hybrid deep learning parameter identification, and proposes an active disturbance rejection control (ADRC) strategy to realize compressor anti-surge. MethodFirst, a long-short-term memory neural network (LSTM) is used to process the time series relationship of the input and output data for compressor parameter identification; the interval probability estimation ability of Gaussian process regression (GPR) is integrated; a combination of LSTM and GPR (LSTM-GPR) is proposed; and a hybrid deep learning parameter identification algorithm is used to realize the rapid diagnosis of the compressor surge state. Then, based on the ADRC method, the parameters of the compressor's throttle valve are controlled, and the accurate control of the surge state of the compressor is realized through the compensation of the throttle valve parameters by the control amount. ResultsThe results show that the hybrid deep learning parameter identification algorithm can accurately identify the critical Greitzer parameters of the compressor and quickly and accurately judge whether it is in a surge state, and the ADRC-based control strategy can effectively allow the compressor to exit the surge state, which is faster and more effective than traditional PID control and nonlinear feedback control without losing the working range of the compressor. ConclusionThe proposed parameter identification and ADRC method can be applied to the surge diagnosis and active control of compressors to improve their safety and stability.

Keywords

• compressor
• surge diagnosis
• hybrid deep learning model
• active disturbance rejection control (adrc)