Performance change prediction of stoner pipeline simulator simulation centrifugal compressor based on Whale Optimization Algorithm-enhanced back-propagation neural network

Centrifugal compressors are the primary energy-consuming equipment in long-distance gas pipelines, accounting for approximately 40%–50% of the total operational costs. Thus, an accurate understanding of the performance parameters, specifically the shaft power, of centrifugal compressors is crucial for reducing operational expenses. Current research primarily focuses on predicting performance parameters such as the pressure ratio, with limited studies on shaft power prediction. This research aims to enhance the accuracy of centrifugal compressor performance prediction by establishing four prediction models: support vector machine (SVM), back-propagation neural network (BPNN), Bayesian optimization back-propagation neural network (BO-BPNN), and Whale Optimization Algorithm back-propagation neural network (WOA-BPNN). Eight sets of training data and two sets of testing data are utilized for interpolation and extrapolation, while the stoner pipeline simulator (SPS) simulation software is employed for validation purposes. The results demonstrate that the WOA-BPNN model exhibits higher precision and accuracy in predicting performance. By improving the weights and thresholds of the BPNN based on the WOA, this model avoids the risk of falling into local optima and demonstrates good generalization performance and applicability. Compared to the SVM and BPNN models, the WOA-BPNN model showcases lower mean absolute errors, root mean square errors, and mean absolute percentage errors in predictions, be it interpolation, extrapolation, or simulation of actual operating conditions using SPS software. The predicted results show minimal differences from the actual operating conditions, while the absolute average percentage error can be controlled around 1.24%, which is acceptable. The established WOA-BPNN model outperforms the other models across all evaluation metrics. This method enhances the accuracy and applicability of performance prediction for centrifugal compressors.