| With the rapid development of power system’s dispatching automation, a higher requirements of operation security, economy, power quality and rehabilitation measures after accidents for power system has been proposed. While the grid size expended, grid measuring equipment increased, as well as a variety of EMS advanced application software (PAS) implementation, resulting in the a great amount of collected data by the SCADA system. As a basis for PAS state’s other applications, State estimated calculation after topology analysis requires stable, efficient and rapid, therefore, rapid calculation of the measured data stability, presents a big challenge to state estimation, while the development of parallel computing make it possible to calculate quickly of the power system state estimation in real time.In this paper, a large-scale power system state estimation is calculated as the background, using basic weighted least squares of state estimation. Through the algorithm simulation and analysis, LU decomposition method for solving high-dimensional sparse linear correction equations occupy the main time algorithm of the state estimates. Because the forward and backward substitution of LU decomposition method is not suitable for parallel generation, and the characteristic of symmetric and positive definite for matrix of information in state estimate, This paper proposes to replace LU decomposition method with the use of conjugate gradient for iterative method. At first, proposed a weighted least squares state estimates of the improved algorithm based on preconditioned conjugate gradient method, i.e. PCG-WLS.At the same time, pretreatment techniques are been studied in order to accelerate the convergence rate conjugate gradient method. Due to the high information matrix sparsity, combined with the sparse matrix techniques, information matrix block parallel computing, in a continuous state estimation calculation, whereby only need to transmission the part of need change of sub-matrix portion, so as to improve computational efficiency. Finally, with the GPU’s massively parallel computing performance, to achieve the estimated state GPU-based parallel computing, simulation experiments, get good results, the validite the effectiveness of the parallel.Through simulate preliminary experiments, proposed pre-conjugate gradient method and WLS state estimation algorithm has a better degree of integration, and its parallel computing method should have a good estimate of the project to accelerate the effect in the actual state of the system. |
