chenyewu AT

back to Research

Does centralized control always guarantee the best performance?

Most power systems are operated by central decision making entities - e.g., the independent system operators - based on the belief that the central control can always achieve the best performance. On further review, however, this argument makes an implicit and impractical assumption, in particular for large dynamic systems. It assumes that the central entity has all the information about the dynamic power system and also has unbounded computational power. In contrast, I seek to improve the performance of the current control framework without such an assumption, by understanding how local information exchange between different components in the system can help, specifically for frequency regulation (an ancillary service for power systems to balance supply and demand every few seconds). Unfortunately, arriving at an analytical understanding of the role of communication is a particularly delicate task for large dynamic systems, such as power systems, in that most distributed algorithms take some time to converge to the desired control strategies. To tackle this challenge, together with Professors Soummya Kar and Gabriela Hug, I proposed a consensus-based algorithm, which utilized the central control signals (the area-control-error signals) [1]. We demonstrated analytically that with only lightweight information exchange (one round local information exchange with communication neighbors after receiving the real time central signals), our approach stabilizes the system very cost effectively ($\epsilon$-close to the optimal dispatch). Moreover, simulation results showed that our approach stabilizes the system about 30% faster than the current framework.
I intend to further improve this approach by incorporating more practical issues in the model, e.g., the delay and synchronization of communication, and the nonlinear model of the components. My ultimate goals are to implement our approach in the real power systems, and design a realistic self-managing intelligent power system model without a central control entity.
Related Publications:
1. Chenye Wu, Soummya Kar, Gabriela Hug, Enhanced secondary frequency control via distributed peer-to-peer communication, under 2nd round review at IEEE Trans. on Power Systems.
Simulation source code can be downloaded here.