An artificial intelligence integrated real-time and dynamic voltage control framework

Abstract

The ability to regulate voltage levels along the whole power system within accepted limits in a reliable and economically efficient way is a goal that utilities have worked toward for many years. Previous approaches to voltage control, such as the Optimal Reactive Power Flow (ORPF) methods, are deficient in both quick adaptation to unexpected contingencies and keeping the number of control adjustments reasonable.

This thesis describes a new framework, named the Artificial Intelligence (AI) Integrated Real-Time and Dynamic Voltage Control Framework (AIIRTDVCF), for regulating voltage in a timely and cost effective way. This system consists of four components: • A planner to generate a predictive schedule of voltage control activities based on load forecast; • A monitor to observe the performance of the predictive schedule; • A evaluator to rank the priority order of alternative control actions to alleviate an emergency situation; • A decision maker to adapt the existing schedule to cope with unexpected events in real time.

In the planner, a Multi-Stage Optimal Reactive Power Flow (MSORPF) model has been developed to generate the predictive schedule. This model considers both the time-separated constraints, such as power flow equations, and the time-linked constraints, such as maximum allowable daily operating times of devices. The latter kind of constraints cannot be accommodated in traditional snap-shot ORPF methods. A computationally effective solution method for this MSORPF model, combining heuristic rules with the Sequential Quadratic Programming (SQP) algorithm, is proposed.

In the monitor, the Newton-based power flow algorithm is used to detect existing or potential voltage violations due to unexpected events happened in a dynamic and uncertain power system.

In the evaluator, the Simple Multi-Attribute Rating Technique (SMART) was adapted to rank the over-all performances of alternative control actions considering both the cost coefficient and the schedule consistency. The consideration of the schedule consistency overcomes the drawback of other existing methods: impractically large number of adjustments.

In the decision maker, a fuzzy sensitivity-based voltage correction algorithm was devised to adapt existing schedule in real-time. This method can accommodate the vagueness in the real-time voltage control, such as the inexactness associated with sensitivity coefficients and the soft operational limits.

The proposed framework is validated on both a 9-bus test system and the IEEE 30-bus system in normal and abnormal scenarios. Simulation results show that the performance of this framework reaches the expectation: regulating voltage in a timely and cost effective way. Specifically, it quickly detects abnormal conditions and provides real-time voltage control solutions that are optimum in both short-term and long-term efficiency. Furthermore, its architecture and problem-solving methods can be translated to design other scheduling systems, especially in a dynamic and uncertain environment.