New Publication: Multi-objective Hyperparameter Optimization of Artificial Neural Networks for Optimal Feedforward Torque Control of Synchronous Machines

We are thrilled to announce our latest research, authored by Niklas Monzen, Florian Stroebl, Prof. Dr. Herbert Palm, and Prof. Dr.-Ing. Christoph M. Hackl, published in IEEE Open Journal of Industrial Electronics Society.

Our team successfully applied Multi-Objective Hyperparameter Optimization (MO-HPO) to identify the best Artificial Neural Network (ANN) architectures for the optimal feedforward torque control (OFTC) of synchronous machines.

This innovative method allows for systematically identifying Pareto-optimal ANNs, providing an optimal trade-off between accuracy and computational effort. We trained these identified Pareto-optimal ANNs, implemented them into a real-time system, and successfully tested them on a nonlinear reluctance synchronous machine (RSM) in our lab.

Drawing on the latest findings from ANN approximation theory, Niklas, Florian, Prof. Palm, and Prof. Hackl present valuable design guidelines for a Pareto-optimal, ANN-based OFTC design, significantly limiting the search space of ANN architectures.

📄 To the article: 10.1109/OJIES.2024.3356721