A hybrid model based on chaos particle swarm optimization for significant wave height prediction

Abstract

Short-term prediction of significant wave height (SWH) has crucial impacts on operation safety of offshore structures and marine navigations. However, conventional intelligent models have limitations in predicting non-linear situations. This paper introduces a hybrid algorithm combining chaos particle swarm optimization (CPSO) with a support vector regression (SVR) model to enhance the generalization and nonlinear handling capabilities for SWH prediction. Additionally, Principal Component Analysis (PCA) is incorporated to reduce information redundancy. To validate the proposed model's predictive performance, several alternatives are tested, including the single SVR model, PCA-SVR, and PCA-GA (Genetic Algorithm)-SVR models. Additionally, the PCA-GWO (Grey Wolf Optimizer)-SVR and PCA-CPSO-SVR models are compared to assess the effects of GWO and CPSO techniques. Significant improvements were observed when comparing CPSO-SVR with other algorithms. Prediction efficiency was evaluated using mean absolute error (MAE), root mean square error (RMSE), and the correlation coefficient (R). Across different test set lengths, the PCA-CPSO-SVR model reduced RMSE by 54.12 % to 74.88 % compared to the benchmark. These results demonstrate the hybrid PCA-CPSO-SVR model's strong generalization ability and superior predictive capacity for non-stationary waves.