Research On Construction And Parallel Computing Of Power System State Matrix Based On OpenMP

With the continuous expansion of the scale of the world's power grid,the rank of the corresponding state matrix of the power system and the difficulty of constructing the state matrix are increasing at the same time.This paper will introduce OpenMP technology,matrix partitioning and other methods to improve the speed of power system state matrix construction.Plug-in modeling technology can modularize all relevant components of power system.According to their state and non-state variables,it convert them into a set of transmission modules.This method can select the input and output signals at will,and can form state space equation intuitively with less modification to the model when adding or subtracting new power components.However,the state space equation processed by this method needs a rather tedious matrix operation to construct the state matrix,so the coefficient matrix in the construction process needs to be reorganized and expanded.After the interpolation modeling technology and matrix reorganization and expansion,the rank of matrix is still relatively big,and the inversion and algebraic operation of higher order matrix are time-consuming in the construction process.In order to speed up the operation of the correlation matrix,we can reduce the order of the high-order matrix by blocks,and transform the inversion of a large matrix into the inversion of several smaller-order matrices.Then we use the Gauss elimination method to operate directly on the initial matrix,which reduces the memory occupancy of the matrix operation to a certain extent.Faced with the fact that most coefficient matrices are sparse matrices in the process of constructing state matrices,Haig iteration method is used to reduce the shape and loan of state matrices,and then three-tuple sequential table method is used to compress storage to reduce the storage capacity and access time before operation.After writing the basic program in FORTRAN language,the appropriate code segments can be processed in parallel with OpenMP technology.Based on Visual Studio 2010 platform and Intel Fortran 13SP1 environment,this paper compiles and compiles the program.Through the power network examples of 69 generators and 260 generators,the validity of formula reorganization,matrix block reduction,sparse matrix shape reduction method and parallel computing is verified.