基于伴随灵敏度的暂态稳定预防控制算法及其并行化研究

发布时间：2018-06-08 06:03

本文选题：预防控制 + 暂态稳定　；参考：《浙江大学》2017年硕士论文

【摘要】：预防控制作为保障电力系统安全稳定运行的第一道防线,是电力系统分析与控制中的重要研究课题。暂态稳定预防控制的目的是寻求一个经济性最优的稳定运行点,在该状态下系统不但要求满足静态安全的各种约束,还要求在发生预想故障后能够保持其暂态过程的稳定性。因其包含有描述系统暂态过程的微分-代数方程组,所以该问题在数学上是一个高维、强非线性、非凸的动态优化问题,当考虑的系统规模较大、故障数量较多时,该问题的求解往往会面临"维数灾",从而大大增加计算的复杂度。然而,随着风电、光伏等新能源的大量并网,现代电力系统的运行方式和状态越加多变,这就对电力系统稳定控制的实时性提出更高的要求,如何更快地得到一个可靠的暂态稳定预防控制决策就是本文研究的主要目的。本文的主要研究内容及其成果如下:1)将暂态稳定预防控制建模为考虑暂态稳定约束的最优潮流问题,利用约束转换技术将含微分代数方程约束的优化问题转换成常规的非线性优化问题,能够极大缓解"维数灾"现象,并利用优化层-仿真层的双层结构实现算法:仿真层利用伴随灵敏度分析方法实现动态灵敏度的快速计算,并在暂态仿真中采用非诚实牛顿法、在反向积分中采用雅克比矩阵重用来加速计算;优化层基于预测-校正内点法来求解转换后的非线性优化问题,利用BFGS方法来近似暂态稳定约束的海森矩阵,并根据其稀疏特性提出一种分块生成的方法来提高海森矩阵的准确性和稀疏性。在多个不同规模的算例中对算法进行测试,结果表明所提出的算法具有收敛性好、计算高效的优点,相较于其他算法表现出巨大的速度优势。2)针对上述串行算法中耗时最大的两个部分,即伴随灵敏度计算和内点法中修正方程求解,采用并行计算技术来提高其计算速度;针对不同故障的伴随灵敏度计算耗时存在较大差异的现象,提出一种基于任务队列的计算任务动态分配策略来提高其并行计算中的负载平衡效率;针对修正方程的系数矩阵非零元增多的情况,分析并利用其系数矩阵的结构特点,提出一种分块求解的方法将其主要计算任务转移到多右边项的回代上,从而可采取并行计算来提高求解速度。提出了包含两部分并行的暂态稳定预防控制并行算法。测试结果表明,该并行算法具有很好的负载平衡效率和加速比,能够有效降低考虑多预想故障的大规模电力系统暂态稳定预防控制问题的求解时间,使其达到在线控制的要求。
[Abstract]:As the first line of defense to ensure the safe and stable operation of the power system, the prevention and control is an important research topic in the analysis and control of the power system. The purpose of the transient stability prevention and control is to seek a stable operating point with the best economy. In this state, the system requires not only to meet the constraints of static security, but also to be preformed. The stability of the transient process can be maintained after a failure. Because it contains a group of differential algebraic equations describing the transient process of the system, the problem is mathematically a high dimension, strong nonlinear and non convex dynamic optimization problem. When the system is large and the number of faults is large, the problem will often face the "dimension disaster", As a result, the complexity of the calculation is greatly increased. However, with a large number of new energy sources such as wind power and photovoltaic, the operation mode and state of the modern power system are more and more changeable, which puts forward higher requirements for the real-time performance of the power system stability control, and how to get a reliable transient stability prevention and control decision faster is the study of this paper. The main contents and achievements of this paper are as follows: 1) the transient stability prevention and control is modeled as the optimal power flow problem considering the transient stability constraints, and the constrained transformation technology is used to transform the optimization problem with the constraint of differential algebraic equations into the conventional nonlinear optimization problem, which can greatly alleviate the "dimensionality disaster" phenomenon and benefit the problem. Using the double layer structure of the optimization layer simulation layer, the simulation layer uses the adjoint sensitivity analysis method to realize the fast calculation of the dynamic sensitivity. In the transient simulation, the dishonest Newton method is used, and the Jacobian matrix is used to accelerate the calculation in the reverse integration. The BFGS method is used to approximate the haessen matrix of transient stability constraints, and a block generation method is proposed to improve the accuracy and sparsity of the haessen matrix according to its sparsity. The algorithm is tested in a number of different scale examples. The results show that the proposed algorithm has good convergence and high efficiency. In view of the two most time-consuming parts in the serial algorithm, which are the two most time-consuming parts in the serial algorithm, that is, the adjoint sensitivity calculation and the correction equation in the interior point method, the parallel computing technology is used to improve the calculation speed. A dynamic assignment strategy based on task queues is proposed to improve the load balance efficiency in the parallel computing. In the case of the non zero increasing of the coefficient matrix of the modified equation, the structure characteristics of the coefficient matrix are analyzed and used, and a method of dividing the main computing task into the multiple right item is proposed. In the back generation, parallel computing can be adopted to improve the solution speed. A parallel algorithm for transient stability prevention and control with two parts parallel is proposed. The test results show that the parallel algorithm has good load balance efficiency and acceleration ratio, and can effectively reduce the transient stability prevention and control of large power system considering the multi preconceived obstacle. The time to solve the problem is to meet the requirement of on-line control.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM712

【相似文献】