Incremental Learning-based Online Risk Assessment of Power System Operation with High Renewable Energy Penetration

Recently, the goal of implementing the carbon neutrality by the mid-century has become an international development consensus, as a result, renewable energy sources have been widely deployed in the power system. The power system can effectively utilize renewable energy sources to alleviate the imbalance between supply and demand. While renewable energy sources contribute greatly to energy saving and emission reduction, its stochastic and intermittent nature also bring about new challenges to the safe and stable operation of the power system. Therefore, this paper takes the power system with high renewable energy penetration as the research object, generates real-time operation data by using the method of intra-day rolling optimization scheduling. Firstly, risk assessment indexes are calculated based on the scheduled results of the power system at each moment. Secondly, the power system risk assessment model is trained based on long and short-term memory neural networks to realize the prediction of risk assessment indexes. Then, the neural network based on incremental learning paradigm updates the model online to improve its prediction performance. Finally, the accuracy of the risk assessment model is verified based on a case study of the IEEE-33 node system. The results show that the online risk assessment method for power system operation based on the incremental learning paradigm performs well in predicting the operational risk indicators. It can realize the online updating of the model, which can provide guidance for the actual operation of the power system.