【作者】 杨霖；

【导师】 郭志忠；

【作者基本信息】 哈尔滨工业大学 ， 电力系统及其自动化， 2008， 博士

【摘要】 基于面向时间过程思想研究配电网分析与控制问题无论在基础理论研究还是实际工程应用方面都具有十分重要的意义。在动态系统的观测过程中,采集往往是以数据时间切片的形式,所以长期以来网络能量损耗的计算均采用时间切片的积分形式计算,并无电能量损耗直接的评估方法,能量损耗计算所需的数据信息和计算量都较大,计算效率不高。而且,在工程实际中电力系统是广域快速动态系统,常规的基于时间断面信息的反馈控制方法无法解决控制变量受经济性、安全性指标的制约不可能对每个时间断面做出反应的问题,控制中心决策过程缓慢与电网需要动态快速响应的矛盾也难以协调。因此,本文提出了时间过程特征状态的概念,应用面向时间过程的思想进行理论分析并面向时间区间制定优化控制策略。考虑到单纯地分析时间断面数据会抛弃断面间的联系,无法准确描述系统的动态运行过程,而逐点计算的方法在断面数据较多的情况下会遭遇计算量过大的困境,时间过程特征状态的概念将面向时间过程的优化控制问题对应等价为面向特征断面的优化控制问题,化繁为简,动态问题采用静态方法求解,有助于依靠以点带面的思想制定适用于某个时间区间的静态控制策略,然后通过时序融合算法得到整个时间区间时间分段方案,从而生成整个时间区间有效的动态控制策略。首先,基于时间过程特征状态的概念,本文在配电网分析与优化问题中引入了时序融合思想,为面向时间区间制定优化方案问题提供了理想的过程分段规则。在此基础上,通过对负荷波动对网络能量损耗的灵敏度分析,建立了给定时段内配网有功能量损耗的评估模型,根据过程特征状态对给定时间区间内网络的有功能量损耗进行直接评估,使有功能量损耗计算问题的求解在考虑负荷变化的情况下,在快速性与准确性之间达到了平衡。此外,该评估模型将网络有功能量损耗分解成3个部分,包括1个基本项和两个修正项,各部分物理意义清晰,可根据系统实际的量测配置分别简单求取,极大地改善了算法适应性。其次,在保证安全性的前提下,以网络能量损耗作为经济性指标制定了面向时间过程的配电网电容器规划与控制策略。电容器规划与控制问题的数学模型相似,本文将这两种问题结合起来进行讨论。受到安全和经济等因素的影响,电容器的安装数量、位置、类型和容量都不能根据负荷变化无限制地更改,即应尽量避免控制变量过于频繁的调节。因此,离散控制设备动作次数约束造成动态无功优化问题的时空强耦合,使问题的复杂度大大增加,本文以包括配置电容器的节能降耗收益与控制变量调节代价的综合运行费用最小为目标,面向时间过程特征状态建立了求解电容器优化配置问题的数学模型,从而得到了一段时间区间内的电容器优化问题静态配置方案。然后,结合时序融合算法所提供的时间分段方案,可以得到整个时间区间的动态优化策略。在电容器动态投切问题中,为便于时间切片数据的融合,本文将多节点电容器动态投切问题分解为一系列单节点电容器动态投切子问题,然后通过迭代依次求解子问题的方式得到整个时间区间内各节点电容器的最优动作时间和投入容量。再次,基于时间过程特征状态的概念,本文提出了面向时间区间的配电网静态重构和动态重构算法。在考虑负荷变化的情况下,依靠有功能量损耗评估模型,提出一个以能量损耗最小为目标的静态重构策略,在寻优过程中将开关状态的变化问题转化为回路电流源的叠加问题,仅需要一次弱环网潮流计算,有效降低了求解面向时间过程的重构问题所需的计算量。针对动态重构问题,考虑到开关操作次数的约束,本文构造了开关动作成本函数并以罚函数的形式扩展到动态重构的模型当中,面向时间过程特征状态建立了一个最大收益原则的时序融合算法,将传统的数学规划算法与现代智能算法相结合,根据每次开关操作在整个时间区间内的降损收益,确定了一段时间内网络动态重构所需的开关操作时间序列,取得较为理想的优化结果。最后,基于时间过程特征状态的概念,提出了面向时间过程的配电网多目标综合优化方法,结合电容器投切与网络重构两种优化措施,将降低系统能量损耗与均衡负荷共同作为优化目标,通过对网络重构与电容器投切两个子问题的交替迭代求解,得到了比单一优化方式更加理想的优化方案。更多还原

【Abstract】 Time-process oriented theory has great significance to promote technological development of distribution network analysis and control method and introduce a new breakthrough point for the related basic theoretical fields. Usually, the traditional energy loss computation in electrical networks demands for numerous metered data and great amount of computational time when the accurate energy loss evaluation is required. On the other hand, electrical network is a wide-area fast dynamic system in engineering practice, so conventional feedback control methods based on the time-point information are impossible to solve such a typical problem, that is control variables usually can not respond to all system operation states at each time point following the changed load, considering some potential economic or security constraints. Furthermore, the time-process oriented control approach can also be used to solve such contradiction: most human-involved decision-makings for system operating manners cannot be made quickly, but the system state is continually changing rapidly and demands instant responses from the control center. Therefore, solving the time-process oriented control problem is particularly significant because making a real-time control scheme is extremely unlikely in most cases even if all the information we need is probably available today. It is necessary to introduce the time-process oriented theory into the distribution network analysis and control fields so that the core objective of power system operation can be further achieved in a more secure, more reliable, higher quality and more economical way. This paper presents a concept of system process state characterization to deal with the huge dynamic information heaps. As we know, the common time-domain analysis method based on information at a certain point in time will neglect the relationship among contiguous time series, and corresponding computational efforts upon each point in time will become more unacceptable when more time series information is offered. So, the concept of system state characterization can help people avoid performing power flow calculations for every interval of the load curves and finish the endless tolerance of the iterative procedures, because its idea successfully converts the complex dynamic optimization problem into a relatively simple static optimization problem. The whole process of system behavior could be optimized simultaneously while the characteristic state is optimized by some approaches such as capacitor placement and network reconfiguration.Firstly, this paper introduces a novel concept about time-series fusion, and develops the sensitivity relationship model between the network energy loss and load curve based on the conception of system process state characterization, and establishes a new energy loss formula for electrical networks. In this formula, electric components’energy losses in a given duration are divided into three parts. The primary energy losses can be quantified by the power flow calculation with average loads at nodes, and its linear and quadratic correction value can be produced by using an approximate algorithm whose accuracy is determined by the current system’s actual measurement configuration. The categorization of load nodes according to the type of installed measurements further improves the adaptability of this algorithm and makes it suitable for networks with incomplete measurements as well as those advanced networks with modern measurements.Secondly, this paper takes the network energy loss as a major economic indicator to construct a dynamic capacitor placement and switching control principle on the premise of security restriction. Considering the mathematical model for dynamic capacitor placement and switching control is similar in nature, these two issues are combined for discussion in this paper. In most cases, the control scheme of capacitor placement will be made out of the following considerations: Subject to some security and economic factors, the installation number, location and type of capacitors can not be changed without restraint, and that means control variables should not be adjusted too often. However, since loads change on an hourly basis or even shorter, the optimal number, location and size of capacitors or even the network configuration may change accordingly in order to achieve the greatest energy conservation. Consequently, the action number of discrete control system makes the original reactive power optimal problem strong-coupling in time and space, which significantly increases complexity for direct solution. So the proposed method in this paper take the maximum energy saving as major objective in which the regulating costs of control variables are also converted into the forms of energy loss, and a optimal static control scheme of capacitor placement can be made in a given period. On this basis, time series fusion algorithm will provide the suitable sub-time dividing scheme so that the whole time interval of dynamic optimal control solution series will be obtained. Specifically, multi-node capacitor dynamic switching problem is decomposed into a series of single-node capacitor dynamic switching subproblem in this paper so as to simplify the process of time series fusion, and then the problem will be solved in an iteration manner.Thirdly, this paper presents static and dynamic distribution network reconfiguration algorithms based on time interval according to the conception of system process state characterization. In the optimization process, based on sensitivity analysis, loop-analysis and superimpose theorem, opening a tie switch is equivalent to superimposing an electric current source on the loop. Consequently, the computational effort required in solving such large-scale dynamic optimization problems is decreased greatly by converting the original problems into some simpler static optimization problems with appropriate linearization and stepwise correction idea. For dynamic reconfiguration, in considering the load changes in engineering practice, the mathematic model takes the minimum energy loss as the objective in which the regulating costs of control variables is also converted into the forms of energy loss. Therefore, depondeing on the traditional mathematical programming method combined with the modern intelligent algorithm, time series fusion algorithm can determine the optimal network struction in the each subsection and finally obtain the optimal switches control operation series required for dynamic network reconfiguration plan.Finally, this paper proposes a distribution network comprehensive optimization to solve capacitor switching and network reconfiguration problem at same time based on system process state characterization. In the method, minimum energy loss combined with predefined index for load balance is taken as objective, and an alternating iteration algorithm is proposed in the paper to improve the effect of comprehensive optimization. The experiment has worked out satisfactorily.更多还原