Modeling of Scroll Expander Based on Long Short-Term Memory Neural Network

Suitable modeling is the premise and basis for the optimal control of the scroll expander. This paper proposes a modeling method of scroll expander based on deep learning. The method selects the inlet pressure, the rotational speed and the inlet temperature of the scroll expander as the model inputs, and selects the output torque and the outlet volume flow rate as outputs; Processing gross errors in the data by the Butterworth filter; Finally, the long short-term memory (LSTM) neural network is used to train data to obtain the scroll expander neural network model. The LSTM neural network effectively overcomes the vanishing gradient problem in the recurrent neural network. Therefore, the error between the predicted value and the measured value of the model output is small, which satisfies the engineering control requirements.