Data Compression-Based Model-Free PI Algorithm for Sparse LQT Control in Interconnected Multimachine Power Systems

This study delves into the distributed linear quadratic tracking (LQT) problem within interconnected multimachine power systems (IMMPSs), and proposes a model-free policy iteration (PI) algorithm based on data compression technology for designing sparse controllers that align with the actual communication links in IMMPSs. Specifically, to address the practical limitation that communication links between subsystems of IMMPSs may be unavailable, we first formulate a sparse LQT problem in which the sparse patterns of controllers match the actual communication links. Meanwhile, in order to be applicable to real-time applications while overcoming model uncertainty caused by parameter variability common in IMMPSs models, we subsequently develop a data compression-based model-free PI algorithm for the abovementioned sparse LQT problem. The main advantages of this algorithm over existing algorithms for IMMPSs control are threefold: first, it has the ability to operate without a prior knowledge of system model, second, its embedded data compression significantly reduces the time consumption for controller design, making it suitable for real-time applications, and third, it designs controllers based on actual communication links, making it practical for applications where communication infrastructure may be constrained. Finally the efficacy of the proposed algorithm is verified through the IEEE 39-bus New England Power System.