An ever growing share of renewable energy resources in the distribution grid imposes fluctuating and hardly predictable feed-in and thus demands new control strategies. On the other hand, combined with controllable, shiftable loads and battery capacity, these energy units set up new flexibility potentials for ICT-based control. So far, many approaches for harnessing this potential neglect the indispensable grid compliance of scheduling results due to the high computational complexity. We present a hybrid approach that enables distributed, agent-based algorithms for predictive energy planning to incorporate grid friendly behavior into agents' decision routines. We propose a scheme using a covariance matrix adaption evolution strategy (CMA-ES) for deciding on grid compliant solutions in a many-objective approach. The integration with an agent-based greedy algorithm for decentralized predictive scheduling is demonstrated and the effectiveness of the approach is shown by several simulation experiments.