Detection And Location Of False Data Injection Attacks Based On OPTICS-LOF In Secure PMU Configuration

Aiming at the malicious behavior of false data injection attacks(FDIA)in supervisory control and data acquisition(SCADA)can avoid bad data detection mechanism and tamper with state estimation result,this paper developed a parallel dynamic state estimation scheme to achieve the goal of FDIA detection and location,which starts from secure phasor measurement unit(SPMU)and uses state estimation consistency detection and clustering as means.Firstly,in order to solve the effect of measurement noise on state estimation,a robust algorithm is introduced to improve it,a Huber cost function is established,and a design correction factor is used to reconstructs the measurement noise covariance to inhibit the impact of bad data.The applicability of the robust algorithm is verified by simulation,which provides a basis for consistency detection of parallel state estimation.Secondly,in view of the iteration problem of simulate anneal arithmetic in SPMU configuration optimization,the zero injection node merge algorithm is used to improve the initial solution of the simulated annealing method,and random disturbance is added to the optimization process to satisfy the redundancy requirement of state estimation and minimize the number of SPMU under the premise of system observability.Through simulation and comparison experiments,SPMU is configured for different systems to prove the effectiveness of the method.Furthermore,combined with the characteristics of independent operation of measurement systems,the robust and unscented Kalman filter is used for parallel dynamic state estimation to judge the operation state of SCADA system at each moment according to the consistency of estimated results.Aiming at the principle of FDIA avoiding the traditional measurement residual detection,a measurement estimation model is established and the two-dimensional measurement deviation affected by the attack is defined.Considering that the traditional detection method threshold is empirical and cannot be located,the ordering points to identify the clustering structure and local outlier factor(OPTICS-LOF)is adopted to mine the characteristics of measurement deviation in this paper,and false data in the SCADA system is detected and located according to the result queue and outlier index.Finally,the IEEE30 node test system validates the effectiveness of the research scheme under different attack scenarios.Through simulation analysis,it is proved that the method adopted in this paper can deal with systems with large sample data,which provides theoretical basis and technical support for ensuring the safety and stable operation of power system information.