Critical inertia assessment for renewable power system with deep learning method

Inertia is a critical physical indicator for the power system, essential for maintaining frequency stability. To accurately assess critical inertia for renewable power system, this study proposes a critical inertia assessment method based on an improved conservative convolutional neural network (IC-CNN). Firstly, frequency stability indicators are extracted from the constructed system frequency dynamics considering synchronous generators (SGs) and virtual synchronous generators (VSGs). Secondly, an improved conservative CNN model for frequency features prediction is developed. Finally, critical inertia assessment framework is proposed. The lower-level model uses the IC-CNN to predict system frequency features, while the upper-level model iteratively adjusts based on the feature predictions to achieve an accurate critical inertia assessment. Results on a modified IEEE 39-bus system demonstrate that the proposed method effectively assesses critical inertia and demonstrates better performance compared to traditional method.