Adaptive Critic-based Control of Voltage Source Converters in Microgrid Systems

Thumbnail Image
Seidi Khorramabadi, Sima
Power Electronics , Fuzzy Logic , Microgrids
Control of microgrids, as the main building blocks of the future smart power grid, is an important problem which has initiated many research activities in recent years. The microgrid should appear to the power grid as a single united entity, in which the majority of distributed energy resources are interfaced through voltage source converters (VSCs). In dynamic situations, specific structure, natural nonlinearity, and low physical inertia of VSCs may lead to higher sensitivity to network disturbances and power oscillations and in occasions result in violation of overall stability; hence the need for fast and flexible control techniques in the microgrid is evident. The design simplicity and easy implementation of PI controllers have resulted in their popularity in controlling VSCs; however their application is associated with a number of drawbacks such as poor harmonics attenuation and unsatisfactory operation in case of load changes and high penetration of distributed generators. In this Ph.D. thesis, three different control algorithms are proposed for VSCs in microgrid systems. The control systems are based on the adaptive critic-based control concept and employ an element called critic whose task is to evaluate the credibility of the performance and compare it with the desired goals. The critic’s evaluations are then used in an on-line procedure to update the controller parameters during dynamic transients. The critic-based control idea is used in conjunction with PI and neuro-fuzzy controllers. With the proposed approach, the need for precise design of the controller is removed, and because of the supervisory role of the critic, no complicated mathematical calculations are required for its design. This fact increases the degree of intelligence and adaptivity against changes such as high penetration of distributed generators and dynamically demanding situations like presence of motor loads and results in a self-tuning and non-model-based control system with high computational speed. The simulation results verify that the application of the proposed approach significantly improves the dynamic performance by reducing the convergence time, output oscillations, tracking error, and unwanted current harmonics and confirm the effective control in case of high penetration of distributed generators.
External DOI