A Neural Network-based Online Estimation of Stochastic Inertia in a Power System

Inertia provides a first line of defence against frequency contingencies in a power system. It limits the rate of change of frequency (RoCoF) and provides significant time for the operation of primary frequency control, in particular governor control. This ensures adherence to frequency excursion limits as specified in the grid codes and, hence, power system frequency stability is preserved. Conventionally, inertia has been provided by synchronous generators (SGs). In modern power systems, SGs are being replaced with inertia-less stochastic renewable energy resources (RES). Penetration of RES into power systems results in a low and time-varying inertia, which potentially exposes the power system to the risks of blackouts caused by frequency instability. This calls for the online estimation of inertia to aid the design of adaptive protection schemes for the prevention of frequency instabilities. Therefore, this paper proposes an artificial neural network (ANN) based approach to estimate power system inertia. The approach can estimate inertia under both transient and ambient conditions. Moreover, the method is computationally inexpensive and explicitly extracts inertia which is an added advantage over most modal-based estimation approaches.