Virtual Inertia Estimation for Grid-forming Converter based on Deep Reinforcement Learning Method

A large amount of renewable energy is integrated into the power system through converters, which leads to the reduced system inertia and would deteriorate the frequency stability. Grid forming converters is proposed to deal with the inertia deficiency of converter by mimicking synchronous generator. In this context, estimating and evaluating the converter-based inertia is crucial for power system operators to understand the frequency-regulation potential of converter-integration power grids. However, due to the differences in frequency response characteristics, existing inertia estimation methods applied to synchronous generators are not suitable for virtual inertia based on digital control. To solve this problem, a non-intrusive online inertia estimation method considering frequency response based on deep reinforcement learning method is proposed. Firstly, the theory and delay characteristics of grid forming converters are analyzed. Then, a serial-parallel identification structure is designed within the framework of deep reinforcement learning. The proximal policy optimization method is utilized to establish network considering frequency measurement to estimate inertia parameters. Finally, results in numerical simulation and hardware-in-the-Ioop experiments verify the effectiveness and accuracy of the proposed method.