Flexible and provacy conserving optimal parameter tuning for large scale power systems

Abstract

Reliable and secure operation of power systems becomes increasingly challenging as the share of volatile generation rises, leading to largely changing dynamics of the system. Typically, the architecture and structure of controllers in power systems, such as voltage controllers of power generators, are fixed during the design and buildup of the network. Replacing existing controllers is often undesired, challenging, or not possible at all. Setpoint adjustments, as well as tuning of the controller parameters, are possibilities to counteract large disturbances and changing dynamics which do not require changing the existing controllers. We consider approaches for fast, computationally efficient, and privacy conserving adaptation of parameters of structured controllers for large scale power systems based on H1 optimization, also referred to as structured H1 controller synthesis. The approach allows the dependency of the system model on the controller parameters to be nonlinear. Two methods for structured H1 controller synthesis are proposed, and conditions are established that guarantee that the approaches leads to stability of the closed loop system. The results are verified experimentally in a field test microgrid consisting of six inverters and a load bank, as well as multiple simulation studies. The proposed methods improve the system robustness, as well as the time-response to step disturbances and allow structured controller tuning even for large power networks. The results are compared to other methods for structured H1 synthesis, focusing on computation time and the obtained H1 norm, underlining the efficiency of the introduced methods. In addition, comparisons with approaches which introduce wide-area controllers to the system are made. It is shown that the proposed approach achieves an improved time-domain response, compared to existing wide-area control approaches. Finally, we introduce a hierarchical approach for the controller tuning, which exploits model reduction. The approach increases data privacy and scalability of the tuning, compared to centralized methods. Thereby, we derive conditions for the success of the approach and introduce a tailored approach for distributed model reduction of power systems. We apply the proposed hierarchical tuning method on the IEEE 68 bus system to show its effectiveness. We show that a similar system performance can be obtained as with a centralized method. The scalability of the approach is underlined considering a large scale power system with more than 2500 states and 1500 controller parameters. The approaches set the base for a series of future developments and can be expanded to other classes of systems, such as transport systems, water systems etc.