Ever-growing energy demands and unacceptable emissions from fossil fuel combustion are major driving forces for expanding alternative, green energy sources. Hydro-power is one promising ecological alternative to meet these energy requirements. Such systems do not require any weir or dam and, thus, can be employed with minimum ecological impact. However, available designs are not yet suitable as efficient water energy converters, in particular for conditions corresponding to low water speeds (as mostly found in practice), due to low power output. Savonius turbines are particularly robust and cost-efficient, but show a poor efficiency. This article aims at maximizing the output power of a hydraulic Savonius turbine by modifying the blade profile. The main difference in this work compared to previous studies is that the blade shape of the concave and convex sides evolve independently from each other (no constant blade thickness). Twelve geometrical parameters are involved during shape optimization. To obtain optimal conditions, many transient computational fluid dynamics (CFD) simulations are performed using the industrial flow simulation code Star-CCM+, driven by the in-house optimization library OPAL++ relying on evolutionary algorithms. The optimization process takes into account the output power coefficient (C-p) as a target function. Compared to the traditional Savonius turbine, a relative increase of Cp by almost 12% is obtained at a tip speed ratio of 1.1. The performance of the optimal geometry was compared to the standard design over the whole range of operation. This comparison revealed additionally that the performance was improved by almost 15% at a tip speed ratio of 1.2. The performance of the optimal design is superior for the whole range of operation, particularly at a tip speed ratio exceeding 0.8. Finally, it was checked that the optimal design is still self-starting. (C) 2018 Elsevier Ltd. All rights reserved.