Research On Power System Fault Calculation Based On Multi-core Parallel Computing

With the increasing scale of power system and the number of nodes, the computation burden of fault calculation is increasing gradually. The process of solving massive fault calculation usually needs consume a large amount of time, which cannot meet the rapid solution requirements in practical engineering. At present, enhancing the performance of CPU frequency is becoming more and more difficult. So expecting to rely solely on the improvement of the performance of CPU to accelerate the computing speed is becoming more and more difficult. With the advent and development of multi-core CPU, parallel computing provides a new idea and method for the analysis of fault calculation of power system.The traditional single thread programs of fault calculation cannot obtain direct parallel acceleration effect in multi-core computer, because only one core is operating and the rest of the cores are in idle state. So it is necessary to use the parallel programming method to develop the multi thread program, in order to make the program execute in parallel in multi-core computer.The principle and method of fault calculation of power system is introduced in this paper, and then the methods and steps of the design of fault calculation program are analyzed. Then the theory of parallel computing is introduced. Parallel calculation is divided into shared memory model and distributed memory model. The programming method and the applicable conditions of them are analyzed. The technology of shared memory model based on OpenMP is selected, and the performance of the technology and methods is introduced and analyzed.The parallel processing method of fault calculation program is studied and realized. Several parts of the fault calculation program are parallel developed. The data of each power system district is parallel read and processed separately. According to the characteristics of switching node admittance matrix to the node impedance matrix, the column vectors of impedance matrix are generated independently by different cores in parallel. After the formation of impedance matrix, the calculation process of a large number of fault is executed in parallel.The paper uses an actual power system model with about 3000 nodes to test the parallel fault calculation program. The results show that: multi-core parallel calculation method based on the OpenMP technology can effectively improve the efficiency of calculation. And with the increasing of the cores, parallel speedup will be improved accordingly.