Convolutional Neural Network-based Inertia Estimation using Local Frequency Measurements

Increasing installation of renewable energy resources makes the power system inertia a time-varying quantity. Furthermore, converter-dominated grids have different dynamics compared to conventional grids and therefore estimates of the inertia constant using existing dynamic power system models are unsuitable. In this paper, a novel inertia estimation technique based on convolutional neural networks that use local frequency measurements is proposed. The model uses a non-intrusive excitation signal to perturb the system and measure frequency using a phase-locked loop. The estimated inertia constants, within 10% of actual values, have an accuracy of 97.35% and root mean square error of 0.2309. Furthermore, the model evaluated on unknown frequency measurements during the testing phase estimated the inertia constant with a root mean square error of 0.1763. The proposed model-free approach can estimate the inertia constant with just local frequency measurements and can be applied over traditional inertia estimation methods that do not incorporate the dynamic impact of renewable energy sources.