Dataset for Physics informed Ensemble Learning in Power System Security Assessment

This dataset includes two files: the "operation status 1" records the power system operates on most secure occasions. And the "operation status 2" records the power system handles insecure events. The operation mechanism in 1 is different from 2.In this section, comparative tests are carried out on the modified IEEE 39 system with variable renewable generations and power demand. The N-1 contingency of the power line which causes topology change is taken into consideration. Specifically, wind generation substitutes the machines on buses 35-39 with the same capacity to include the influences from the uncertainty from renewables. The system control facing uncertain renewable generations is decided based on alternating current optimal power flow (ACOPF) with minimum economic dispatch objective. The convergent events of both simulations are utilized to train IPF-GNN. The successful scenarios from ACOPF are applied to train C-GNN. Both successful and unsuccessful ones in ACPF and ACOPF are pooled to train the SA-GNN. Notably, the power flow (PF) dataset comprised 20,000 samples for training and 2,000 for validation, while the optimal power flow (OPF) dataset included 2,000 validation samples and 8000 initial training datasets for the validation of the framework without AL. Approximately 30\% of the instances within these datasets represented insecure scenarios