Study on Stability Margin Trend of Power System Using GHNet

Under the guidance of peak carbon dioxide emissions and carbon neutrality, the proportion of the new energy generation will continue to rise significantly. The randomness, volatility and intermittency of new energy make it difficult to arrange the operation mode of the power system, which also brings great challenges to the early warning and pre-control of the dispatching operation. For the power grid stability trend assessment facing the uncertainty of new energy, the core problem lies in finding the typical and dangerous changing processes from numerous possibilities with high efficiency. This paper proposes the grid hierarchy net (GHNet) model and the chain rule of neural networks to search for the most stable and unstable processes to define the security range trend. Firstly, the GHNet model is constructed and well trained. Then, the restraint conditions of new energy are determined by the statistical results of the actual power and predictive power. Finally, a sensitivity analysis is performed by using chain rule and the extreme changing process will be found out by adjusting the predictive operation point according to the sensitivity results. A 114-bus system is used to verify the feasibility and effectiveness of this method.