Acceleration of a Compute-Intensive Algorithm for Power Electronic Converter Control Using Versal AI Engines

Modern power conversion applications require increased computation power. Recently, an optimal fault-tolerant control technique for cascaded H-bridge power converters, utilizing a compute-intensive algorithm, was presented. However, limitations of conventional computation platforms impose serious constraints on the real time operation of such algorithms. Novel AI Engines emerge as excellent candidates for these computeintensive applications. This article evaluates the novel AI Engines of the Versal Adaptive Compute Acceleration Platform (ACAP) from AMD (formerly Xilinx) for accelerating the aforementioned compute-intensive real-time control algorithm of the power converter. The algorithm is implemented on devices of the Versal ACAP and Zynq-7000 families. For the first case, AI Engines and the Programmable Logic (PL) are used and, for the second, the Processing System (PS) and PL. Several heterogeneous test configurations are designed and the implementation results are compared in terms of throughput, latency, power consumption and hardware cost. The obtained results highlight the superiority of the Versal architectural for this application and demonstrate the potential of AI Engines in power converter control.