Dissertation > Excellent graduate degree dissertation topics show

Research of Power Flow Calculation and Reliability Evaluation for Large-Scale Power Systems Using Parallel Processing Techniques

Author: ZhangHuaiXun
Tutor: XieKaiGui
School: Chongqing University
Course: Electrical Engineering
Keywords: Large-scale power system parallel computation FGMRES method GESP technique state enumeration method
CLC: TM744
Type: Master's thesis
Year: 2009
Downloads: 390
Quote: 0
Read: Download Dissertation

Abstract

Parallel processing has been one of the key technologies in the area of computing science.With the rapid growth of electric energy demand, the inter-regional interconnection and long-distance transmission systems have emerged, the size of power system is becoming more and more huge. The complexity of the network makes power flow calculation and reliability assessment may need large amount of calculation, traditional serial processing technology is limited in time and space to process this problem. At the same time, researchers pay more attention to the parallel computing applications in power system. Improving the parallel speedup and parallel efficiency is the main purpose of the study. In view of this, according to the current problems in flow calculation and reliability assessment for large-scale power system, this thesis aim to study the the parallel computing model, algorithms and so on. Luckily, this thesis have made some preliminary results. Specifically, this article will cover the following:①Power Flow Computation Based on FGMRES(Flexible Generalized Minimal RESidual)-Newton MethodThe calculation of high-dimensional sparse power system equations is the most fundamental and core issue of power system analysis. In order to improve its convergence speed and reliability, this thesis proposed flow calculation algorithm based on FGMRES-Newton method. FGMRES iterative will be strictly dependent on the condition number of coefficient matrix of linear equations. In order to improve the condition number of the coefficient matrix, we need make pre-condition on the coefficient matrix. Preconditioner was generally transformed from the coefficient matrix, this article according to the change of coefficient matrix in calculation process, real-time generate new preconditioner in order to achieve the best pre-condition effect.Then, the purpose of flexible pre-condition can be achieved.As the coefficient matrix of flow equations is a high sparse matrix, in order to avoid filling large number of non-zero sub-element in the process of pre-condition,this article choose the incompletion LU arithmetic to calculation the pre-conditioner. And then analysis a number of power systems using FGMRES method, results of case studies show that FGMRES method can effectively reduce the number of iterations and improve the convergence reliability.②Distributed calculation algorithm for power flow of large-scale power systems using the GESP technique Parallel computing has become a main means for power flow calculation of large scale power systems. In order to obtain a good parallel speedup and efficiency, this paper presents a distributed algorithm for solving linear power flow iteration equations of Newton approach using the Gaussian-elimination-with-static-pivoting (GESP) technique.Based on the properties of coefficient matrix, such as the diagonal dominance and sparsity, the matrix can be split into several blocks with a smaller dimension and be stored in a distributed storage mode based on the border of supernodes. The pipeline technique has also been used to improve the efficiency of the proposed algorithm in the process of parallel LU decomposition. A distributed parallel algorithm for power flow was designed and applied to a number of power systems, such as the power systems with 3000 and 12000 buses. Results of case studies show that the major advantage of the proposed distributed GESP method is that it has better parallel speedup and efficiency when a power system has more than 2000 buses.③Parallel simulation of large-scale power system reliability evaluation based on state enumeration methodThe parallel simulation of bulk power system reliability evaluation is mainly studied in this thesis.Through using state enumeration approach, two parallel task allocation methodologies are presented. The two methods use the system state as the unit of task assignment in order to obtain good load balancing between different processors. According to the nodes’computing performance various in distributed cluster environment, proposed static task allocation scheme, in order to make each node’tasks match to their computing performance.On this basis, because of the assessment process of reliability for different system state difference in time, this thesis designed dynamic allocation scheme based on the processors working state in order to secondary distribution of tasks. The results obtained in tests with RBTS and IEEE-RTS79 have approximately the same reliability indices as those in series processing,but show high speed-up and parallel efficiency. With excellent scalability of parallel algorithms presented,this thesis evidently general reference into different parallel computational practice.

Related Dissertations

CLC: > Industrial Technology > Electrotechnical > Transmission and distribution engineering, power network and power system > Power system simulation and calculation > The calculation of the power system
© 2012 www.DissertationTopic.Net Mobile