Integrating Machine Learning with Binary Integer Programming for Optimal PMU Placement in Dynamic Power Networks

PMUs or synchronized PMUs are rapidly becoming essential in today’s power systems acting as monitors in the advancement of mainstream grid supervisory, protective, and controlling systems. This paper presents an improvement of the existing methodology for PMU placement by incorporating Machine Learning algorithm coupled with binary integer programming in order to feed real-time grid conditions and demand forecasts into the algorithm. Addressing four primary constraints: 1) observability in fully and partially observable cases; 2) conventional measurements across various cases; 3) network robustness against single PMU or a communication line; and finally, 4) flexibility with regards to increasing or decreasing failure rates, this augmented approach includes, but does not limit to, machine learning-based predicative analysis for a dynamic power grid management. Employed on IEEE 14-bus, a 21-bus $400-\mathrm{kV}$ real power system, as well as the IEEE 30-bus system, strategy helps in creating a cost-effective, secure and smart monitoring system.