An Intelligent Operational Reliability Assessment Approach Considering Sample Imbalance

Operational reliability assessment is an important approach to evaluate the potential risk in the new power systems. However, the accumulated computation burden from calculating thousands of scenes in operational reliability assessment still challenges the advanced solving algorithm. To address this issue, data-driven approaches using neural networks are proposed. However, existing approaches suffer from imbalanced samples because the possibility of load shedding is extremely low. To this end, this paper proposes an intelligent operational reliability assessment approach considering sample imbalance. Specifically, the imbalanced load-shedding sample extension method is proposed based on the generative adversarial network to deal with the imbalance of the load-shedding samples in the training dataset. Besides, a novel loss function is constructed with a specific focus on the load-shedding buses. This learning strategy helps the neural networks capture the features of the imbalanced shedding loads in different buses. Finally, simulation results in the IEEE 39-bus and 118-bus systems demonstrate that the proposed method can achieve an obvious improvement in evaluation accuracy and facilitate online evaluation.