FPGA Implementation of ADMM for Model Predictive Control in a DC/AC Converter

This work reports the design and evaluation of an implicit Model Predictive Control scheme tailored for regulating the output voltage of Distributed Energy Resources implemented on an Field Programmable Gate Array. Traditionally, the intrinsic complexity of the design and development stages for implementing optimization algorithms in hardware has precluded the use of implicit formulations of MPC schemes for the control of DERs. In this work, we evaluate the possibilities provided by modern High-Level Synthesis tools to facilitate the fast-prototyping of implicit MPC schemes in reconfigurable logic devices that can overcome typical software limitations in terms of execution time and real-time requirements. In specific, we report a case-study that simulates a DER system controlled through an MPC scheme that uses the iterative Alternating Directions Method of Multipliers (ADMM) to solve the related optimization problem required for each control sampling interval. Experimental results using a Xilinx ZCU104 platform show that implementing the core of the ADMM algorithm in hardware reduces up to 10x the execution time of an MPC iteration in comparison to a software implementation executing in the same platform. The results provide experimental evidence about the potential of FPGAs for the implementation of advanced control schemes for DERs operating in microgrids.