AI-Driven Proportionate Power Sharing in Virtual Synchronous Generators for Optimizing the Source Conditions and Efficiency

Virtual impedance-based power sharing between several interconnected grid-forming inverters (GFMI)s in a power electronics-dominated grid (PEDG) does not consider the available power reserves. This non-optimal power-sharing may result in the overloading of nearby inverter-based resources (IBR)s exhibiting lower effective line impedances in the network. This paper proposes an optimal and controllable power-sharing scheme based on a modular artificial neural network (ANN) that can consider various factors like power reserve, line losses, incentives from the grid, etc., and increases the system's scalability. An optimum power allocator (OPA) mechanism that monitors the output active powers and line losses of all the IBRs is proposed to address the issue of uncontrolled power sharing. The OPA dynamically adjusts power allocation weights using linear programming to optimize the overall loss minimization cost function. Moreover, an artificial neural network-based virtual impedance optimizer (ANN-VIO) is also proposed, which estimates the corresponding virtual impedance (VI) value for each GFMI unit in the PEDG. The primary controller of GFMI uses these estimated VI values to adjust its point of common coupling (PCC) voltage reference and attain optimum proportionate power sharing within the overall PEDG. Multiple case studies are performed to validate the effectiveness of the proposed GFMIs power-sharing scheme.