Reducing the carbon footprint has motivated the adoption of clean energy generation, with wave energy emerging as a promising alternative. However, the design of wave energy farms poses significant challenges. In this work, we focus on the optimal spatial placement of buoys (WECs) to maximize the power output of a farm. A major limitation in this optimization process is the high computational cost of evaluating objective functions through hydrodynamic simulators. To address this issue, we investigate the use of machine learning models to predict the power output of wave energy farms. We perform a comparative study of four models and use the best-performing one as a surrogate objective function to optimize buoy placement with two well-known metaheuristics: Differential Evolution (DE) and Particle Swarm Optimization (PSO). Experimental results demonstrate that employing support vector machines as surrogate models, combined with DE for optimization, provides an effective alternative for designing wave energy farms.