Research On The Price Modification Attack Detection Method In Smart Grid

With the development of smart grid technology,more and more researchers turn their attention to the security problems in power grid.When the two-way communication of information and power help the power system make more efficient scheduling and management,it also bring new security risks.Smart meter is an intelligent terminal of the power system,it make an important role in demand response,real-time pricing and some other technologies.But it is also an attractive object to attackers because of its low power consumption and computing ability.Price modification attack is a novel attack on power system load scheduling by sending malicious modified electricity price data to smart meters.Considering the work principle of PMA and the demand response mechanism in power system,the online detection method of PMA is studied in this paper.This paper mainly focus on the following aspect: designing a fast and effective online attack detection method.In order to provide an accurate data base for online attack detection methods,we first propose a load forecasting method considering the similarity of user power consumption.By analyzing the user’s historical electricity data,the method cluster analysis on the intelligent meter data and construct the user grouping according to the similarity between different users.In each group,the linear regression forecasting method is used to predict the load of the subsystem,and then the prediction results of each subsystem are aggregated to realize the accurate prediction of the overall load of the power grid.In our online PMA detection method,we accumulated the generalized likelihood ratio of demand data and load data in each time slot to ensure the scale of the computing task is consistent.Considering that the attack parameters are unknown,we design an iterative parameter approximation method to solve it and the correctness is proved.In order to improve the robustness of the algorithm,we design sliding window strategy and weighted coefficients associated with user demand to reform the decision function.To verify the effectiveness and practicability of the method,we carry out simulation experiments on a real user load data set.The results show that the grouping strategy considering the customers’ behavior similarity can effectively reduce the prediction error generated in the load prediction.The results also show that our method can guarantee high accuracy in attack detection rate and low false alarm rate,with low detection delay at the same time.