Abstract: Abstract: As a key component of a reversible pump-turbine unit, a pump-turbine runner rotates at two opposite directions under the pump and turbine modes. The flow states in the runner are quite different under the pump and turbine modes, which creates a great challenge for the design of a pump-turbine runner. Nowadays, a pump-turbine runner is always designed in one mode and verified with the other mode. Since there are several parameters affecting the runner's performances and one parameter might have contrary effects on runner performances under the pump and turbine modes, it is hard to obtain a runner with high performance at both the pump and turbine modes by a trial and error method. In the present paper, a 3D inverse design method, CFD calculations, and multi-objective optimization strategy were coupled, and an optimal design system was built to improve the performances of a reversible pump-turbine runner. In the system, by using a 3D inverse design method, the runner shape was obtained by inputting corresponding parameters. A validated CFD method was used to estimate runner performances under different operating conditions. A multi-objective optimization strategy was then adopted to search runners with good performances at both operating modes in the design space based on a response surface model between input parameters and runner performances. The optimal design system was capable of solving some difficulties in pump-turbine runner design, such as a long design cycle and difficulties in obtaining good performances at both pump and turbine modes. The system was used to optimally design a middle-high head reversible pump-turbine runner, in which blade loading and blade lean were chosen as the input parameters, runner efficiency at the pump design point and the turbine rated point were set as optimization objectives, and the head at pump design point was set as a constraint. Since the optimization calculations were based on a response surface model, performances of nearly 10 000 runners under different operating modes were estimated in several minutes. Three runners from the Pareto frontier were chosen for detailed validation. CFD results showed that the optimal runner had a high hydraulic efficiency of over 95% at both the pump and turbine modes, which indicated a great improvement in runner efficiency in the turbine mode and the maintenance of high efficiency at the pump mode as compared with the initial design. The effects of the input parameters on runner interior characteristics under different operating modes were also analyzed in this paper. The results showed that fore-loaded blade loading distribution on the shroud was helpful in improving the flow state at the inlet under the pump mode by reducing the low-pressure area on the suction side. The blade lean at the high pressure side not only changed the work distribution along the span wise direction, but also improved the flow state by bringing down the pressure difference at the inlet in the turbine mode, and played an important role in improving runner efficiency, especially in the turbine mode.