Sparrow optimization gated recurrent convolutional network for temperature modeling of wave rotor refrigeration process

Temperature modeling plays an important role in the wave rotor refrigeration process control and optimization. However, considering data-driven nonlinear and time-delay modeling, how to determine the structure of the model is a challenging problem. To solve this problem, a novel sparrow optimization gated recurrent convolutional network (SGRC) deep learning method is proposed. Firstly, to exploit the advantages of convolutional neural network (CNN), the sample data is converted into grids along the time axis similar to the image input, which contains model structure and dynamic time-delay information. The multivariate and dynamic time-delay information is input into the CNN to extract the multivariate model structure features of the data. Then, after flattening the data into one-dimensional time series, input it into gated recurrent unit (GRU) layers to learn the temporal dependencies of the wave rotor refrigeration. The hyperparameters of the SGRC network are optimized using the sparrow search algorithm (SSA). Finally, simulation results based on wave rotor refrigeration industry data show that the proposed SGRC method achieves superior performance compared to traditional machine learning and other deep learning approaches, exhibiting lower RMSE and MAE values while attaining a higher R² score. This enhancement significantly improves the generalization capability of the temperature model in the wave rotor refrigeration process.