Meta-modeling Power Inverters for Fast Evaluation of Energy Flow Controllers on Resource Limited Devices

In recent publications, energy flow controllers have been trained using simulation based optimization with a complex inverter simulation model, which leads to rather long training times. In this paper, the training and results of a computationally efficient symbolic regression meta-model for the power inverter is presented, which can be applied for faster training and adaptation of energy flow controllers. The results show that the performance of energy flow controllers can be assessed correctly using the meta-model in almost all simulated scenarios. Moreover, the speedup of the controller evaluation using the meta-model results in approximately a factor 24 compared to the evaluation using the complex inverter simulation model.