Abstract

Prediction of the performance of centrifugal compressors, the traditional methods using BP neural network. This single neural network for forecasting problem is not high enough precision, slow convergence and easy to fall into local optimal solution. In order to more accurately predict the performance of centrifugal compressors, the implicit commit identify problems early. Are the immune algorithm, genetic algorithm, wavelet theory, the combination of neural networks, established immune genetic algorithm optimization of wavelet neural network model (IGA-WNN). Realized to predict the performance of centrifugal compressor, and the predicted results with the BP neural network model prediction results and the wavelet neural network model prediction results were compared. Simulation results show that: the prediction model, can achieve the centrifugal compressor performance prediction and monitoring. Which, IGA-WNN optimal prediction results: with a simple algorithm, structural stability, the convergence speed and generalization ability of the advantages of prediction accuracy of 99% over traditional methods of prediction accuracy of 15%, with a certain Theoretical study and practical value.