Resilient Model based Predictive Control Scheme Inspired by Artificial Intelligence Methods for Grid-Interactive Inverters

Abstract

This paper presents an intelligent predictive control schemes that integrates model and data-driven schemes for enhancing the resiliency of grid-interactive inverters to mitigate the impact of dynamic grid condition on model-based control performance. Conventional model predictive control (MPC) techniques feature several advantages such as fast dynamic response, single loop optimization instead of cascaded control schemes, and several others that are enabled by enhancements in micro-controllers for control of power electronics converters. These inherent features of MPC enable design of control schemes with advance functionalities for grid-interactive inverters. MPC efficacy is highly dependent on prediction accuracy of control variables. The prediction accuracy for a predictive controlled grid-interactive inverter depends on many factors including the controller knowledge on filter model parameters and variation of grid impedance. The variation of grid impedance can impact the current prediction accuracy due to the effect of the equivalent impedance on the effective impedance the inverter experiences at its point of common coupling (PCC). The grid impedance variation is expected in future power electronics dominated grid (PEDG) with multiple point of common coupling (MPCC). The proposed resilient artificial intelligence (AI) inspired MPC scheme addresses these challenges towards improving the performance of grid-interactive inverters in PEDG. This is done through the introduction of a Learned Impedance Factor to the MPC cost formulation equation. In this paper an overview of the proposed integrated data-driven and model-based control scheme is provided, and results demonstrate the proposed controller improves the THD and tracking error compared to conventional MPC that is purely model-based.