Artificial Neural Network-Based Inertia Estimation

The increase of non-synchronous renewable energy penetration is causing a gradual drop in the system inertia which has been challenging for system operators and can pose a threat to system stability. This paper presents a perturbation-based method to estimate the system inertia. The perturbations are small in magnitude and short in duration. The method relies on an HVDC converter to inject a perturbation which can be triggered periodically to track the changes in system inertia. The system's response to the perturbation is analyzed using supervised learning to produce an estimate of the system inertia. The method has been evaluated using the simulated IEEE-39 bus system in DIgSILENT™ PowerFactory® and is verified using simulated signals distorted with noise typically found in measurements.