Learning-Based Supervisory Control for Peak Power Management of Distributed Power Cycling Test Systems

With the shift towards electrification in transport, significant research has focused on efficient and reliable drive systems. Accelerated power cycling tests (APCT) are critical but often require complex setups and large sample sizes. This paper introduces an optimization-based supervisory controller, integrating a neural network based model, to manage multiple test setups in medium-to-large facilities. The proposed approach improves test efficiency by reducing the peak power consumption and resource demands. Simulations on a three-setup system and experiments on a two-setup system are presented, along with implementation details.