基于松弛牛顿法的互联电网并行分布式潮流计算方法

分布式潮流计算是解决多区域互联电网一体化潮流问题的有效方法.以松弛牛顿法潮流求解公式为出发点,首先利用矩阵分裂法,将互联电网分解为相互独立的子系统,然后利用矩阵求逆运算的Sherman-Morrison-Woodbery公式对各子系统进行协调求解,从而导出了一种新的互联电网并行分布式潮流计算方法.与基于网络分解及协调计算的分布式潮流计算方法相比,该方法具有更好的收敛性.利用基于PVM的分布式计算环境,以IEEE 118节点系统为例,对所提算法进行了装配、测试以及分析,初步验证了所提算法的有效性.     

基于辅助问题和序列二次规划法的电网分区并行最优潮流算法

针对传统集中式潮流计算存在的问题,建立基于电网分区的分解协调模型.采用构造辅助问题(AP)的方法进行分布式并行计算,将大型互联电网最优潮流问题分解为多个子网的并行潮流优化问题,在每个子网中,采用序列二次规划法求解子系统的的优化问题.基于地域的系统分解与协调符合电网市场的发展方向.仿真结果表明,该算法具有收敛性和快速性     

基于异步迭代的多区域互联系统动态潮流分解协调计算

在全局电网分解协调计算中，为增加各子系统计算的独立性，并减小协调计算对通信系统的要求，提出了一种新的基于异步迭代的互联系统分解协调动态潮流计算方法，这种方法允许每个参与分布式计算的子系统自由选择本系统内的Vθ节点，采用内、外两层迭代的方式求解整个互联系统的动态潮流。以各子系统单独潮流计算作为内层迭代，再利用边界条件构造关于边界节点状态量的不动点迭代格式，通过不断修改各子系统相应外边界节点的等值注入作为外层迭代来统一全网潮流解。所提出的算法经IEEE9节点试验系统验证，可获得全网动态潮流解。     

具有指数收敛速度的电力系统全分布式潮流算法

为了满足互联电力系统潮流计算对数据隐私的需求,提出了一种全分布式潮流算法。该方法由内环迭代与外环迭代组成,外环迭代基于牛顿-拉夫逊法计算雅可比矩阵,内环迭代采用全分布式算法在互联电网各分区分别求解各自的潮流修正方程。该方法不需要协调层对分布式计算进行分解协调,各分区仅需要与邻居分区交换潮流方程修正量的信息,外环迭代收敛特性与全局潮流相同,内环迭代保证收敛且具有指数收敛速度。通过IEEE39节点和118节点算例的测试表明,该方法具有较高的收敛性,适合于没有协调层的分布式潮流计算。     

互联电网的直流最优潮流分解算法研究

研究了大系统互联电网的最优潮流优化策略,基于部分对偶理论分析了电网分区的分解协调模型,提出了一种基于直流最优潮流模型的互联电网多区域分解最优潮流并行求解算法,将一个大的电网互联系统分解成多个区域子问题,每个区域子问题是个典型的二次规划问题,使用直流最优潮流模型来求解互联电网的最优潮流分布,讨论了分区优化收敛条件。通过交换输出电价和边界节点相位角,完成区域间的信息交换。使用上述分解算法对IEEERTS-96算例的多个互联区域进行了分析,结果表明本文算法是一种有效的求解算法,适合大区电网互联后在线分布式动态OPF计算。在电力系统有极大的应用前景。     

基于拉格朗日对偶松弛的多区域柔性直流互联电网无功优化

现有方法对互联后的电网进行无功优化时难以满足大规模电网实时计算、快速反应的需要,并且面临全网数据收集的难题。为了解决上述问题,采用分解协调算法的思想,提出一种基于拉格朗日对偶松弛的多区域柔性直流互联电网无功优化方法。首先根据支路潮流模型,利用二阶锥松弛和二次旋转锥松弛方法建立了多区域柔性直流互联电网的集中式无功优化模型。然后利用拉格朗日对偶松弛理论在集中式优化模型的基础上,提出了可以并行计算的多区域柔性直流互联电网无功优化方法。通过算例计算,验证了所提算法的有效性和正确性。此外,还对比分析了所述的分解协调算法和基于交替方向乘子法(ADMM)的分布式优化算法的计算结果,证明所用算法在计算时间上更具优势。     

计及频率变化的异步迭代分布式动态潮流

为便于分布式调度员培训仿真系统（DTS）中分布式潮流算法的应用,提出了一种计及频率变化的基于异步迭代的分布式动态潮流算法。该算法将互联系统采用协调层一子系统、发电机单元一网络单元2个不同层次的模型来表示。在上级调度中心建立协调层,由发电机单元实现频率变化时电网不平衡功率在各子系统发电机节点和负荷节点之间的分配,并进行频率仿真;在网络单元采用基于异步迭代模式的分布式算法进行全网的分布式潮流计算,通过网络单元与发电机单元之间的交替求解,求得全网频率变化仿真与潮流分布。以IEEE118节点扩展系统为例,对所提出算法进行验证,结果表明该算法能够准确地仿真分布式系统的频率变化并求解全网潮流分布。     

考虑限幅约束的多区域互联电网全分布式潮流算法

多区域互联电网中有大量限幅非光滑约束,增加了潮流模型的非凸性。针对传统基于启发式规则处理限幅的分布式潮流算法在计算过程中容易出现收敛性问题,提出了一种能鲁棒性处理非光滑约束的互联电网全分布式潮流计算方法。首先根据分层分区调度模式对电网进行分区,并将模型中非光滑约束进行光滑处理,然后基于具有二阶收敛速度的双层交替方向非精确牛顿法,将潮流问题转化为求解最优步长增量的问题。基于零空间法(Null-Space)对状态变量的系数矩阵进行降维处理,采用共轭梯度(Conjugate Gradient, CG)算法更新每个分区的对偶乘子,乘子更新过程中利用二阶信息提高了算法的收敛性。多区域间进行分布式计算时无需协调层参与,仅需通信少量边界信息,因此通信负担轻。最后,以30节点和182节点系统为测试算例,验证了所提方法在设置恶劣初值和处理非光滑约束时具有较高的精确性和较好的收敛性。     

基于改进Jacobian-Free Newton-GMRES(m)的电力系统分布式潮流计算

为了提高互联电网分布式潮流算法的实用性,应尽可能减少协调计算中的交换信息.文中采用隐函数方式表示分布式潮流的边界协调方程,并采用具有自适应预处理能力的Newton-GMRES(m)方法构建分解协调算法.该方法具有Jacobian-Free特性,协调计算中只需要交换边界节点状态信息.以IEEE标准系统和大规模实际系统为算例的测试结果表明,新的分布式潮流算法具有较高的收敛性,数据交换接口简单,实用性强,适合于求解连续变化的分布式潮流问题.     

基于分解协调算法的互联电力系统状态估计

提出了基于分解协调算法求解多区域互联系统的状态估计算法。互联系统计算中心根据虚拟节点两端子系统状态估计所得功率残差加权平均值计算各子系统虚拟节点注入功率的变化量,各子系统节点状态量通过注入功率和功率估计值对节点状态量的灵敏度矩阵修正子系统状态估计结果,得到互联系统的状态估计解。仿真计算结果说明,分布式状态估计算法计算准确、速度快,且在互联系统计算中心无法获得部分子系统数据的情况下,能够最大限度地准确计算出互联系统的部分可观测结果。     

考虑可控负荷的多区域电力系统分布式模型预测负荷频率控制

由于同步发电机的惯性较大,导致传统的集中式负荷频率控制模式反应不够迅速,而用户侧具有快速响应能力的可控负荷资源为系统的频率调节提供了新机遇。研究了考虑用户侧可控负荷资源主动参与系统频率调节的多区域互联电力系统分布式模型预测负荷频率控制问题。通过建立的含可控负荷的多区域互联电力系统负荷频率响应模型及自动发电控制模型,基于连续时域交替方向乘子法和分布式模型预测控制方法,提出了一种用户侧可控负荷资源主动参与的多区域互联电力系统分布式模型预测最优负荷频率控制模型。基于修改的IEEE39节点三区域互联电力系统进行仿真验证,结果表明所提考虑可控负荷的分布式模型预测控制策略可显著减少系统恢复至稳态所需的时间。分布式控制策略的控制自由度更高,增强了系统的可控性。     

基于网络分割的电力系统潮流分解协调计算

因现有计算模式的速度已无法满足现代大规模电力系统实时计算的要求,故引入对等计算(P2P)模式以提供低廉而充足的计算力。为此研究了网络计算环境下的电力系统潮流计算模型,结合基于支路切割的网络分割方法和基于注入电流的潮流模型,提出了基于网络分割的电力系统潮流分解协调算法,将大规模互联电力系统分解成若干子网络,子网络间只需交换边界母线的电压状态就可完成分布式潮流计算,保证各个子网络潮流计算模型的独立性。对IEEE标准系统进行潮流计算的结果表明该法具有较高的收敛速度和计算精度,适合网络计算环境。     

A parallel distributed computing framework for Newton–Raphson load flow analysis of large interconnected power systems

This paper proposes a simple parallel and distributed computing framework for the conventional Newton–Raphson load flow (NRLF) solution of large interconnected power systems. The proposed approach is based on message-passing distributed-memory architecture with separate workstations, and involves the piecewise analysis of power systems utilizing the network tearing procedure. The NRLF solution method, applied to each torn system at the selected buses, employs the matrix inversion lemma consisting of the factorization, forward elimination and back substitution procedures. The computational requirements of the state-of-the art parallel algorithm to obtain the correction vector involved in the back substitution procedure is reduced with the proposed approach in which the back substitution is carried out in parallel taking into account the split buses, rather than the order in which the forward elimination is performed. The investigations are carried out on the IEEE 118 bus standard test system in a Redhat Linux based 100 Mbps Ethernet LAN environment. The investigations reveal that the proposed method is significantly faster than the conventional NRLF and also the NRLF based on the state-of-the-art parallel algorithm, and thus finds potential applications for the real-time load flow solution of both regulated and deregulated power systems distributed over large geographical areas.     

基于P2P的保护定值在线校核混合并行算法

考虑到大型互联电网规模的逐渐扩大,尤其是"三华"特高压同步电网的顺利投运,传统的集中式计算会遇到硬件计算能力的瓶颈,提出了一种基于P2P的保护定值在线校核混合并行算法。充分利用P2P网络技术的对等通信,实现区域间信息的对等交互,重点介绍了MPI+OpenMP的混合并行编程模型以及算法的设计,对在线校核进行了并行性分析,实现了在线校核进程级与线程级的两级并行。最后,在基于P2P技术的分布式并行计算平台上,对混合并行算法进行了测试比较,结果表明所提出的算法正确且有效。     

基于Jacobian-Free Newton-GMRES(m)方法的电力系统分布式暂态仿真算法

分布式暂态仿真是实现市场环境下互联电力系统在线一体化仿真分析的有效途径.文中研究了电力系统分布式暂态仿真计算模型,提出了基于Jacobian-Free Newton-GMRES(m)方法的协调求解算法.该算法只需要交换边界母线状态信息,接口简单,实用性强.为了提高算法收敛速度,减少协调求解所需通信次数,提出了自适应预处理和连续修正预处理矩阵、预估边界条件初值及多时步同时协调等改进方法并将其应用于新算法中.测试结果表明,新的暂态仿真分解协调算法收敛快,通信次数少,非常适合在基于广域网络的分布式环境中实现.     

Two-level optimal load–frequency control for multi-area power systems

In large-scale power systems, classical centralized control approaches may fail due to geographically distribution of information and decentralized controllers result in sub-optimal solution for load–frequency control (LFC) problems. In this paper, a two-level structure is presented to obtain optimal solution for LFC problems and also reduce the computational complexity of centralized controllers. In this approach, an interconnected multi-area power system is decomposed into several sub-systems (areas) at the first-level. Then an optimization problem in each area is solved separately, with respect to its local information and interaction signals coming from other areas. At the second-level, by updating the interaction signals and using an iterative procedure, the local controllers will converge to the overall optimal solution. By parallel solving of areas, the computational time of the algorithm is reduced in contrast to centralized controllers. This approach is applicable to any interconnected large-scale power system. However, for simulation purposes, a three-are power system is presented to show advantages and optimality of the proposed algorithm.     

Unified Fast Decoupled Load Flow in a Parallel Distributed Computation Framework

Abstract

This research proposes a parallel and distributed computation approach for the fast decoupled load flow (FDLF) analysis applicable to both transmission and distribution networks of large electrical power systems. In this approach, the overall repetitive FDLF computations are performed by the respective control centers identified for each sub-system, the sub-systems being formed by partitioning the system through a network tearing procedure splitting the identified buses. The data of each subsystem are stored at their own control centers and hence the subsystems are solved using message passing distributed memory architecture. The computations performed by the control centers in various sub-systems are coordinated from a regional control center. The investigations on IEEE 118 bus standard test system indicate that the proposed approach provides faster solution than the conventional approach, without affecting the solution accuracy. The effectiveness of the proposed technique over the conventional approach become more visible when it is applied for the load flow analysis of large scale systems spread over a wide geographical area.     

Unbundling of transmission and ancillary services. I. Technicalissues

This paper presents derivations of basic formulae for computing the contributions of each economic transaction to the network power flows throughout the system in steady-state operation of interconnected electric power systems. It is shown that the net system power imbalance caused by each transaction can be obtained as a function of all transactions present on the system. In addition, formulae are proposed for calculating the contributions of every ancillary generation unit to each transaction. This generation is needed to balance the system in response to economic transactions. Formulae supporting this are based on reformulating the load flow problem in terms of distributed slack bus     

Distributed Transient Stability Simulation of Power Systems Based on a Jacobian-Free Newton-GMRES Method

As power systems becoming more closely interconnected and are being deregulated in energy markets, distributed simulations among different dispatch centers are highly required for online full system analysis and control applications. In this paper a new algorithm for distributed transient stability simulation of interconnected power systems is presented. Based on a Jacobian-free Newton-GMRES(m) method, this algorithm requires only exchanges of states of boundary buses among different regions. Therefore, it has strong scalability in distributed computing environments built on heterogeneous computing resources. Moreover, several accelerating methods are developed to enhance its efficiency, including continuous preconditioning with adaptive preconditioners, predicting boundary conditions and multistep coordination. The standard IEEE 39-bus system and a real power system with 1165 buses were used as test systems. The test results show that these accelerating methods greatly enhance the convergence rate of the proposed algorithm and reduce communication costs remarkably, which proves the novel algorithm is feasible and can be adopted in wide area networks with high-latency.