Biological hydrogen production in stirred-tank bioreactors depends strongly on mixing conditions and impeller energy consumption. In this study, a data-driven framework was developed to optimize the impeller-speed schedule using hourly hydrogen-production data and the process energy balance. The decision variables included the impeller speed in the first stage, the impeller speed in the second stage, and the switching time. The problem was solved in a single-objective mode, aiming to maximize net energy, and in a multi-objective mode, aiming to increase hydrogen production while reducing mixing-energy consumption, using a genetic algorithm and NSGA-II, respectively. The single-objective results showed that operating the impeller at 100 rpm during the first 10 h and reducing the speed to 10 rpm during the remaining 14 h produced the maximum net energy of 26.0737 kJ/L_POME. This solution was reproduced in 30 independent runs and matched the result obtained from the exhaustive evaluation of all 368 feasible schedules. In the multi-objective case, a Pareto front containing 22 non-dominated solutions was obtained. The knee-point solution consisted of operating at 100 rpm during the first 7 h and at 10 rpm during the remainder of the process, providing a suitable balance between hydrogen production and energy consumption. The results demonstrated that stagewise control of impeller speed can maintain favorable hydrogen production while reducing mixing-energy consumption.