基于BO-GRU-ELM的电网虚假数据注入攻击定位检测方法OA

With the deepening coupling of cyber and physical layers in power systems,the threat of cyberattacks has become increasingly severe.Among these threats,false data injection attack(FDIA)can stealthily tamper with measurement data and compromise state estimation in power systems.Consequently,FDIAs will cause severe impacts on the safety,stability,and economic operation of the power system.In this study,a hybrid FDIA model is developed to balance attack cost and benefit.Furthermore,a locational detection method for FDIA based on a Bayesian optimization-gated recurrent unit-extreme learning machine(BO-GRU-ELM)framework is proposed.The method integrates the temporal feature extraction capability of gated recurrent units(GRU)with the efficient multi-output classification capability of extreme learning machines(ELM).Based on these,a GRU-ELM-based detection algorithm is designed.In addition,with the F2-score serving as the optimization objective,BO is employed to perform global optimization of GRU-ELM hyperparameters,thereby further enhancing the detection performance.Finally,simulations are conducted on the improved 14-bus and 107-bus power systems based on actual grid data to validate the effectiveness of the proposed hybrid FDIA model.These results demonstrate that the proposed attack locational detection algorithm exhibits superior performance in terms of accuracy,robustness,and generalization capability.