Wave energy harvesting systems mostly have low power production capability because of unoptimized design of the system components. A bidirectional flow impulse-turbine used in such a system has efficiency less than 40%, and it is required to design the turbine for a higher efficiency. Present work finds an optimal design and shows design-variable sensitivity to the turbine efficiency. The problem is solved using numerical analysis technique. The flow through the turbine was analyzed by solving the Reynolds averaged Navier-Stokes equations (RANSE). The design variables; namely number of rotor blades and number of guide vane, guide vane angle and guide vane profile were modified to maximize the turbine efficiency. Using the Latin hypercube sampling technique, sample points were selected from a design space defined by lower and upper limits of the variables. Then, several surrogates were constructed using the RANSE calculated results, and the turbine performance was optimized. The results show that guide vane angle is the most sensitive parameter, while the guide vane profile has negligible effect on efficiency. A hybrid genetic algorithm searched the optimal design point. The relative mean efficiency enhancement over a wide range of flow coefficient was approximately 24%, while it was 28% at maximum efficiency point. (C) 2018 Elsevier Ltd. All rights reserved.