Transient Stability Parallel Computing For Large-Scale Power Systems

Nowadays the interconnection among areal power systems are being formed in China, which results in the exigent requirements of faster power system transient stability simulation. Thus, parallel computing for large scale power system transient stability analysis was studied in this dissertation. The main work focused on the improvement of serial algorithm, parallel algorithm, task scheduling and parallel software implemention on a high-performance platform and so on. The work of this dissertation can be listed as follows:1) An novel mixed serial algorithm is introduced to power system transient stability simulation, in which Shamanskii method and very dishonest newton algorithm(VDHN) are combined with variable step size technique. In the framework of simultaneous solution method of differential-algebra equations, this proposed algorithm utilizes the step-size control strategy on the basis of the local truncation error theory firstly. In addition, according to a class of Newton methods’convergence index, Shamanskii method is employed, which can control the update of Jacobin matrix adaptively. Furthermore, VDHN method applies to simplify the computation of voltage vector. Several test cases testify the validity and practicability of the proposed algorithm, which is especially efficient to improve the simulation when the fault of the network is serious. Some comments on its limitations are also provided.2) An OpenMP-based parallel Very Dishonest Newton (PVDHN) algorithm with variable step size is provided, running transient stability simulations on multi-core computers. Under the framework of simultaneous solution method of TSA, the step-size control strategy is used according to the local truncation error theory firstly. Then, computation of the generation units, which is the most time-consuming part of the simulation, is dynamically dispatched to several cores using an a dynamic scheduling scheme to obtain workload balancing based on OpenMP. Due to the convergence of Newton-type iterations, an adaptive Jacobian update control strategy is applied to reduce the sequential part of the simulation and the overhead generated by OpenMP. Several large scale test cases verify the validity and practicability of the proposed parallel algorithm, showing that the proposed approach achieves6.01comprehensived speed-up and a considerable reduction in parallel overheads.3) In order to satisfy the desire for faster than real-time transient stability simulation of large-scale power system, a parallel algorithm based on CPU-GPU (Graphics Processor Unit) platform is introduced. According to two-level block bordered diagonal form (BBDF) of the matrix, the algorithm decomposed the whole computation into three parts:1. the computation of dynamic units and injection currents;2. sub-network computation;3. boundary network computation. Part land2were computed on multi-core CPUs. Part3was solved by preconditioning Biconjugated gradient stabilized method on GPU. The precodnditoner was produced by sparse approximate inverse technique. The simulation results show that for a large network with12685nodes, the proposed algorithm runs6.63times faster than on a single CPU core, and the simulation time is less than the actual transient process time. It provides a new solution strategy for large-scale power system simulation.4) A multi-rated parallel algorithm based on CPU-GPU platform for the transient stability simulation of AC/DC power system is presented. A detailed model is applied to HVDC in which detailed HVDC controls and L/R Line dynamic is included. As a result, AC system uses large step size and HVDC uses small step size in the simulation. At the first parallel level, AC/DC power system is decomposed into two parts:the computation of AC power system is deployed on CPU, and HVDC is deployed on GPU. The second parallel level focuses on HVDC. Based on the principle of parallel-in-time method, the pipeline technique is implemented on GPU, which makes multi-integration steps of HVDC solved simultaneously. In order to increase the utilization level of the high-performce platform, a cloud computing proterotype system is developed based on Model-View-Controller (MVC) model. Case studies show that the proposed algorithm obtains a better efficiency, and the high-performance computing resources can be more eaily accessed by theThe realization methods in this dissertation provide a new approach to high performance computing technologies in the field of transient stability simultion in power industry.