Dynamic security assessment and control of modern power systems using intelligent system technologies

Research Doctorate - Doctor of Philosophy (PhD)Security is a basic yet essential requirement for operating a power system. Due to the continuous growth of load demand and unmatched infrastructure investment, modern power systems are being pushed to operate closer to their security boundary. This research focuses on the dynamic security analysis of power systems, which involves examining different stability criteria when the power system is subjected to a sudden disturbance. Conventional methods are mainly based on time-domain simulation technique, which usually suffers from insufficiently fast speed and inability to provide useful information about system dynamic characteristics and guidelines for control. Using intelligent system (IS) technologies, this research developed a series of new methods, which can effectively overcome the shortcomings of the conventional approaches. In the area of dynamic security assessment (DSA), an extreme learning machine (ELM)-based real-time DSA model is developed at first. The model is much faster in learning speed and hence can be on-line updated for performance enhancement. Subsequently, based on ensemble learning theory, an IS composed by multiple ELMs is developed for more accurate and reliable DSA. This IS is able to filter out potentially inaccurate results, when combined with a traditional DSA tool such as time-domain simulation it can enable reliable and real-time DSA. At last, an intelligent DSA framework is proposed for power systems with large-scale wind power. The proposed framework is able to accommodate fast changes of system operation condition due to the imbedded intermittent wind power generations. In the area of dynamic security control (DSC), methods for preventive transient stability control and emergency frequency stability control are developed. Specifically, a hybrid method for transient stability-constrained optimal power flow (TSC-OPF) is proposed, which is able to able to rigorously satisfy the transient stability constraints and efficiently search the global optimal solution. Subsequently, for more transparent and interpretable preventive control, two new methods based on decision tree (DT) and pattern discovery (PD) techniques are proposed, respectively. The DT-based method can infer stability control rules and the PD-based method is able to identify (statistically) the stable region, which are both interpretable for practical use. When combined with an OPF tool, efficient and transparent preventive control can be realized. Finally, emergency control against frequency instability is studied. A method for real-time predicting event-driven load shedding is proposed. The proposed algorithms and methodologies have all been successfully demonstrated on a number of benchmark test systems and compared with existing proposals in the literature (when applicable). Simulation results have verified their effectiveness and superiority over compared approaches