Artificial Intelligence-based Control Methods for Optimal and Stable Operation of Converter-dominated Microgrids

Abstract

The microgrid as a major player in future smart grids includes power-electronic-based distributed generation (DG) units, loads, energy storage system (ESS), and lines. The microgrid can operate either island or connected to the main grid. The voltage and frequency references in island microgrid are adjusted by individual DGs while in grid-connected mode these references are dictated to the DGs by the upstream grid. The droop control and virtual synchronous generator (VSG) control are well-known methodologies to control several converters in an island microgrid. The small-signal stability of a microgrid is defined as its ability to move from one permissible operating point to another permissible operating point after being subjected to a small-signal disturbance. The droop control coefficients, virtual impedances, and VSG parameters should be tuned in a feasible range to maintain the stability of microgrid. Despite the remarkable achievements, the state-of-the-art microgrid control methods face three major challenges: (1) These methods have not optimized the virtual impedances by considering the microgrid small-signal stability and power sharing in all operating points, inappropriate application of virtual impedances can jeopardize the microgrid stability; (2) VSG provides virtual inertia and damping in the microgrid including static and dynamic loads, however, inappropriate tuning of these parameters can threaten the microgrid stability, microgrid frequency, voltage, and reactive power sharing; (3) The application of artificial neural networks in online control of converters and VSGs is necessary to fulfil the stability and dynamic performance requirements in future microgrids. First and foremost, this thesis introduces a new perspective on microgrid control methods, which suggests to analyse the stability of all operating points and define an optimization problem according to the dynamics and stability preferences of microgrid. This optimization method concludes the stable operation of microgrid in all operating points and a desirable dynamic performance, simultaneously.Secondly, the thesis reports a novel method to optimize the virtual inertia, virtual damping, current state-feedback factor, and virtual impedances to enhance the microgrid small-signal stability. Moreover, the reactive power sharing, frequency Nadir, and voltage of buses are enhanced. Finally, the thesis introduces an online optimal control method based on adaptive network-based fuzzy inference system (ANFIS). In this method, the controller learns the optimal control policy for each value of active and reactive power and generates the optimal value of virtual inductance accordingly. The reactive power circulation among converters is minimized and the voltage drops on virtual inductances are negligible. Moreover, the small signal stability of microgrid is enhanced by the proposed control method.