Achieving high stable Power Transfer Efficiency (PTE) in Wireless Power Transfer (WPT) systems remains challenging due to the nonlinear, multimodal nature of the optimization space. Conventional algorithms such as Genetic Algorithms (GA), Differential Evolution (DE), and Simulated Annealing (SA) often face premature convergence, sensitivity to parameter settings, and inconsistent performance across runs. To overcome these issues, this study introduces the Evolutionary-Swarm Hybrid Algorithm (ESHA), which integrates DE for directional exploration, GA crossover for population diversity, SA for adaptive convergence, and Lévy Flights for stochastic global search. ESHA was implemented on a WPT system with a fixed 20 cm transmission distance and compared with GA, DE, and SA using three performance indicators: PTE, convergence speed, and computational efficiency. Results show that ESHA achieved a maximum PTE of 97.18%, surpassing GA (96.81%), DE (96.65%), and SA (96.19%), while maintaining zero variance across independent runs. It converged in an average of 31.2 iterations, slightly faster than GA (33.15) and SA (32.1), and comparable to DE (31.3). Execution time was 0.4738 s, close to GA (0.4654 s) and only marginally higher than DE (0.4262 s) and SA (0.4329 s). Statistical validation confirmed significant improvements in PTE (p < 0.05).