A feasibility study on applying meta-heuristic optimization and Gaussian process regression for predicting the performance of pantograph-catenary system

As the pantograph-catenary system provides electric energy for high-speed trains, it is vital to evaluate the contact force (CF) between pantograph and catenary for stable energy supply. The magnitude and variation range of CF determines the quality of current receiving and safe operation of the train. Therefore, a rapid and accurate prediction of CF is of great significance. However, collecting CF data through experiments is challenging, and obtaining timely results using numerical simulations is not always feasible. In this study, we propose an efficient simulation-based surrogate approach based on Gaussian process regression (GPR), combined with meta-heuristic optimization, to predict key parameters of pantograph-catenary system, which are responsible for the energy transfer quality. Firstly, a pantograph-catenary model is established and validated using finite element method (FEM), which serves to generate training and test data. Secondly, Gaussian process regression is utilized for estimation. A new developed meta-heuristic optimization, i.e., binary hunger game search (HGS), is applied on feature selection. To enhance the performance of HGS, chaos mechanism is embedded, resulting in Chaos-HGS GPR (CHGS-GPR). Finally, the predictive results of CHGS-GPR are evaluated. It is found that the proposed CHGS-GPR provides rather accurate prediction for the mean value of CF, and can be extended to the preliminary design of railway lines, real-time evaluation, and control of train operations.