A Data-Driven Approach for Estimating Inertia and Damping of a Single-Area Power System

The inertia of the electrical grid is crucial for ensuring system stability. The increasing integration of renewable energy resources gradually decreases the inertia of the system, leading to greater frequency deviation under disturbances. Sim-ilarly, due to fluctuating demand and intermittent generation, system inertia varies considerably. In this context, the accurate estimation of inertia is crucial which is challenging through conventional mathematical methods. This paper proposes a con-volution neural network approach for estimating the inertia and damping of an electrical grid. The neural network-based method utilizes a non-disruptive test signal to change the dynamics of the power network and estimates the system inertia and damping coefficient from the local frequency measurements. The introduced method determines the inertia and damping with high accuracy even under the impact of noise and is compared with the performance of the multilayer perceptron, support vector machine regressor, and gradient-boosting machine regressor in terms of accuracy, root mean squared error, and mean absolute error to validate the results. The proposed technique can assist system operators in providing fast-frequency support and system protection schemes.