Research On Distributed Power Flow Calculation Based On Spark

The modern power system is an extremely large and complex network.In terms of current computer performance,the development of power systems has faced many bottlenecks and problems,as well as power system power flow calculations.The traditional serial power flow calculation method can meet the calculation of small-scale networks,but it is difficult to cope with the system analysis requirements of large-scale and complex power flow calculation.Distributed parallel computing can divide a computing task into several subtasks,and then mobilize multiple machines and small tasks at the same time,which can make up for the shortcomings of serial computing.In order to solve the above problems,the parallel computing method can be used to solve the problem that the serial computing takes a long time due to the increase of the number of power nodes in the power flow calculation process,and at the same time,the convergence of the power flow calculation can be ensured.This paper focuses on the distributed power flow calculation method and technology using the Spark platform.The main results of the research work and the following are as follows:(1)The classical algorithm of traditional serial power flow calculation such as Newton-Raphson method and PQ decomposition method(fast decoupling method)is analyzed,and the main problems and calculation difficulties of power flow calculation are analyzed;The method of taking the inverse matrix of the Jacobian matrix and then solving the correction value by the inverse matrix is designed,The distributed Jacobian matrix inversion method is designed,and the inverse matrix size is reduced from the original n-1 order and m order matrix to the n-1 order matrix.(2)In order to solve the fast solution of large-scale matrix multiplication,the blocksolving method of large matrix is designed.The large-scale matrix is divided into smallscale sub-matrices,and then distributed according to the matrix multiplication rule.Finally,the calculation results are summarized.The advantages of distributed parallel computing methods can be fully utilized to improve computing power.(3)Using a virtual machine to build a four-node Spark platform on a desktop computer,one node is used as the master node and the other three nodes are used as computing nodes.The power flow calculation algorithm for parallel computing PQ decomposition is written in Scala language.(4)The algorithm is tested with power system data of different scales from tens of thousands to tens of thousands of nodes.The experimental results show that the distributed power flow calculation based on Spark platform is obviously better than the serial power flow calculation,and has good scalability,which proves this The algorithm has certain practical value in large-scale power flow calculation.