Energy Efficient State Estimation For Smart Grid

State estimation is a crucial part of power grid Supervisory Control And Data Acquisition System and undertakes the task of providing data for other analysis of power grid.With the rapid development of smart grid,smart measuring devices are interconnected more closely with each other,and the amount of data generated by the measure system increases exponentially.The state estimation for power grid is facing new challenges.To process large amount of measurements,serial state estimation iterative algorithm is time consuming,and can not meet the demand of real-time state estimation of power system;as intelligent measure devices increase the possibility of measurements being attacked,the tradit ional serial bad data detection method is slow and can not detect false data injection attacks.Although paralized serial algorithm is faster,it needs large scale computing cluster and can not avoid time-consuming iterative computation.The use of clusters also produces high energy consumption and maintenance costs,which reduces the efficiency of parallel algorithms.In order to estimate smart grid status for fast and low energy,this paper analyzes and captures the characteristics of false data detection and state estimation,which can be realized through neural network.The false data detection algorithm based on low-rank matrix recovery can be converted into convex optimization problem,and the decomposition matrix can be obtained by alternating iter ation.This article uses neural network respectively completed the convex optimization solut ion and least squares fitting.The neural network is built and trained offline,and for actual calculation,the original energy-consuming iterative process will be replaced by the neural network forward calculation,which greatly accelerates the computation speed.Due to the short time required for the forward calculation of neural network,the method proposed in this paper can be operated on a single machine platfor m even if the large-scale power grid is processed,thus avoiding the energy consumption required by large-scale computing cluster.The experimental results show that the neural network based false data defense algorithm is more accurate,and measurement recovers faster,compared with the literature algorithm.The state estimation algorithm based on the neural network proposed in this paper is approximately 205 times faster than the serial state estimation algorithm,and the calculation speed is about 17.4 times higher than that of the cluster of 20 nodes,and the energy consumption is about 1/193.