Data Reconstruction of Wireless Sensor Network Based on Graph Signal

Abstract

The environmental and other factors can cause data missing in power systems; thus, data reconstruction is of great significance. In this paper, we model the observed signal as time-varying signal based on graph signal processing (GSP) and establish an optimization problem with the objective of minimizing the error between the true signal and the reconstructed signal at the sampling points and improving the smoothness of the reconstructed signal. To solve the optimization problem, Taylor series expansion is performed on the Hessian inverse matrix of the objective function, and truncated Taylor series is used as an approximation of the Hessian inverse matrix. In the simulation, the algorithm proposed in this paper is compared with the gradient descent algorithm, and the result shows that the proposed algorithm converges faster and the reconstructed signal is more accurate.