Detection of False Data Injection Attacks in Smart Grid Based on Machine Learning

The false injection attack on the internal state estimation module of the smart grid energy management system enables the attackers to forge false power, voltage and topology information, which make the false data successfully bypass the residual check module. And they can be upload it to the control center to affect the control decision-making and endanger the normal operation of the power system. Firstly, this paper introduces the principle, attack conditions and impact of false injection attack, and introduces in detail an AC model which is more widely used than DC model. The AC model can generate false data injection attacks against different power system topologies. On this basis, in order to detect the false data injection attack with high robustness and effectively, we propose an attack detection method based on machine learning. The random forest classifier in the integrated learning algorithm is used to detect and classify the false data injection attack, and the state information of the attacked power system is identified by integrating multiple decision trees. A large number of experiments show that the method has good recognition performance in different IEEE node systems.