【作者】 王明敏；

【导师】 王建全；

【作者基本信息】 浙江大学 ， 电力系统及其自动化， 2008， 硕士

【摘要】 大规模交直流联合运行的电力系统的形成,对电力系统暂态稳定分析速度提出更高的要求。暂态稳定性的快速正确评估,对于电力系统的安全稳定运行具有重要的意义。目前广泛采用数值积分法对预想事故进行暂态稳定性分析,这种方法的主要缺点是计算速度慢。在电力系统暂态稳定计算中,大量的计算量存在于网络方程中的矩阵运算,深入研究电力系统暂态稳定计算中节点编号的优化以提高计算速度是本论文的主要研究内容。本文首先分析了稀疏技术在电力系统网络计算中的应用以及各种已有稀疏矩阵节点编号和稀疏矢量节点编号方法的特点、效率及其存在的问题。根据节点优化编号本质上属于组合优化的特点,提出了基于粒子群优化算法的节点优化编号方案。在不同规模的系统上证明了算法的有效性,并以该算法的结果作为近似最优解,与传统的各节点编号算法进行比较,说明了对于大规模电网各传统节点编号算法还有待改进。同时,指出了当系统网络结构或独立矢量、解矢量发生局部变化时,对基于粒子群优化的最优节点编号作局部修正仍然是比较好的近似最优解。文中还指出了暂态稳定计算中,网络方程独立矢量中仅发电机节点的注入电流为非零元,解矢量也只需要得到发电机节点电压的特点,根据这个特点,提出了在独立矢量和解矢量确定的情况下的MSNOL、MD-MSL1、MD-ML1、MD-MNPS等节点编号优化算法,并与其他算法得到的结果进行了比较,证实了这些算法的有效性。最后,提出可将改进的节点编号优化算法应用于网络的分区中,说明了基于因子树的网络分割方案,将节点编号优化方案用于优化平衡分区的计算负荷及在分区中进一步减少前代回代计算量。更多还原

【Abstract】 As the scale of the interconnection of power systems is enlarging increasingly, the requirement for faster electromechanical transient simulation is more urgent. The main analysis method of transient stability is numerical integration. Though lots of methods were presented so far, none of them is fast and reliable enough for practical use. The application of sparsity technology evidently makes the matrix calculation of power system more efficient. The main works of this paper focus on the electrical network node ordering in transient stability solution.The paper firstly analyzes the characteristic and efficiency of the node ordering algorithms that already presented recently. A new method of node ordering that based on swarm particle optimization was proposed and its efficiency had been demonstrated by simulation tests of different systems. And the results of the node ordering algorithm based on PSO and the traditional method are being compared, which demonstrate that the efficiency of the traditional node ordering method has to be improved. According to the characteristic of network equation, some improvement of the traditional node ordering is proposed in the event that independent vector and solution vector are determined. Test result on many power systems indicates that the methods proposed like MSNOL、MD-MSL1、MD-ML1、MD-MNPS have effective performance. These methods are then used in network partition that based on factorization path tree partitioning, to assess the partitioning scheme in the original algorithm of factorization path tree partitioning to make the distribution of computing load between each processor more reasonable.更多还原