Data Learning-based Frequency Risk Assessment in a High-penetrated Renewable Power System

Renewable energy is gradually replacing traditional power plants to provide electricity for users, but it also brings potential risks to the safe operation of the power grid. Thus, in the planning stage, it is necessary to comprehensively evaluate the probability of violation risk of maximum frequency deviation. Planning method based on Monte Carlo simulation (MCS) is inefficient, while artificial neural network (ANN) can make fast and effective prediction by learning data. Therefore, this paper proposed an MCS-ANN algorithm to realize the rapid assessment of violation risk of regional maximum frequency deviation. Firstly, a large number of stochastic disturbances were generated, and only a small part of disturbances were used for MCS. Then, these data were sent to the neural network for training, and most of the remaining disturbances were sent to the trained neural network for output prediction. The above training and prediction processes were repeated. The average of multiple prediction results was used as the final prediction output, and the probability distribution of each risk interval was obtained. Finally, the effectiveness of the proposed MCS-ANN algorithm was verified on IEEE 10-machine 39-node system.