Modular Multilevel Converter (MMC) energy balancing is a nontrivial open-loop control problem, since the choice of a technically meaningful output leads to an internal dynamics. The method used in this paper relies on an MMC arm energy model, which allows algebraic parametrization of almost all system variables and the rest can be obtained by integration of a small subsystem of low order. The solution introduced in this paper is to plan appropriate trajectories for all variables such that the balancing goal is met. The planned trajectories are given to the control system, which leads to a feedforward balancing effort supporting the standard feedback balancing control. In contrast to the previous approach, an analytical solution was obtained providing for efficient real-time trajectory planning. Test bench measurements confirm improved energy control performance compared to standard feedback balancing.