| PhD Viva

Name of the Speaker: Ms. Lavanya (EE18D414)
Guide: Dr. Swarup K S
Online meeting link: https://meet.google.com/gww-kwyo-qvv
Date/Time: 17th June 2025(Tuesday), 3 PM
Title: Event-Driven and Continuous Inertia Estimation in Modern Power Systems

Abstract :

Grid inertia plays a crucial role in damping frequency fluctuations and ensuring resilience during disturbances. A significant amount of grid inertia comes from conventional generators, including coal, gas, nuclear, and hydro plants. However, these plants are being retired and replaced by inverter-based renewable plants that have little to no inertia. As a result, grid inertia reduces considerably. With lower inertia, the power system's stability is impacted, increasing the likelihood of outages. Estimating and monitoring grid inertia is therefore essential to ensure grid reliability. Grid inertia can be determined using the measurements reported by Phasor Measurement Units (PMUs). In this talk, various methods for grid inertia estimation using PMU data are discussed.

A novel method that uses energy variations to estimate grid inertia from power imbalance events is presented. Traditional event-based approaches determine inertia from the calculation of the Rate of Change of Frequency (RoCoF). However, computing RoCoF accurately is challenging due to inter-area oscillations. The method uses energy variations to estimate inertia, which eliminates the need to calculate RoCoF. Also, it models the governor's response, improving the accuracy of the results.

Additionally, a methodology for calculating the inertia of an area from tie-line flows is presented. The estimated inertia includes the load response in addition to the inertia, termed effective inertia. The load response is spontaneous and acts like additional inertia by reducing frequency deviation during disturbances. However, this response varies with system operating conditions, disturbance size, and location. An Artificial Neural Network (ANN) model is developed to predict effective inertia from actual inertia, which can be used to forecast frequency dynamics following outages accurately.

Given that grid inertia can vary depending on operating conditions, an algorithm for continuous real-time estimation of inertia using the measurements from normal operating conditions is also presented. Unlike System Identification techniques, which are often computationally expensive and assume constant mechanical power, the presented method offers improved efficiency and accuracy. A Q-learning-based approach is also introduced that identifies variations in mechanical power, ensuring accurate inertia estimation.