【作者】 王达；

【导师】 薛禹胜；

【作者基本信息】 山东大学 ， 电力系统及其自动化， 2009， 博士

【摘要】 跨大区电网互联和电力市场化的推进在为我国经济建设与发展提供强有力支持的同时,也对电力系统的安全、稳定与可靠性运行带来了新的挑战。而电力系统的安全、稳定与国家建设与稳定休戚相关,因此,如何防御大停电灾难,保证电网安全、稳定、优质的运行就成为一个重大和迫切的课题。闭环离散控制是电力系统在大扰动后保持运行可靠性的主要手段,包括切机、切负荷、解列等等,可以分为以故障驱动的紧急控制(EC)及由受扰轨迹驱动的校正控制(CC)两类。当前对二者的研究主要集中在如何以最小的控制量或控制代价来满足安全稳定约束,方案的制定依赖于系统仿真和工程经验,决策过程都是在确定的运行工况、模型参数及故障场景下,孤立进行的。闭环离散控制的决策要素包括系统模型与参数、工况、预想故障集、目标函数、约束条件、控制变量与控制时机,其优化决策离不开可靠、高效的安全稳定量化分析算法。扩展等面积准则(Extended Equal Area Criterion)提出了高维受扰轨迹摆次稳定性的概念以及评估每摆稳定裕度的量化方法,深刻的揭示了非自治非线性多机电力系统暂态稳定性的机理和规律,并能提供满足工程实用要求的稳定裕度指标。本文利用基于EEAC理论的电力系统安全稳定量化分析与优化决策软件包FASTEST,从兼顾电力系统安全性与经济性的角度出发,对EC优化、CC优化及二者的协调展开了较为系统的研究。EC是故障驱动的前馈控制,时间滞后很小,控制性价比高;但由于其决策表是按典型场景预先准备的,实际场景与之失配可能引起严重过控或欠控。CC是轨迹驱动的反馈控制,其执行时刻在不安全现象出现以后,其性价比要差;但反馈控制律可以消除过控及欠控问题。本文以切机、切负荷和解列措施为例,深入比较了EC与CC在启动判据、动作时机、控制代价、控制律及控制精度等方面的不同,讨论了它们在确定工况及故障下的优化决策。电力系统本身是随机的,各种随机因素按受扰场景可分为工况类和故障类两种。不同地点和类型的故障对稳定性的破坏程度不同,即使相同的故障,在不同的电网结构、机组组合、母线注入和系统备用下的控制代价也不同。按典型的工况和故障的确定性仿真及保守原则制定控制策略,其决策不但不经济,还有可能带来危险的负效应。以旋转备用水平及频率特性为例,分析了随机因素对某实际电网频率安全性的影响。若用EC来应对小概率高风险的不确定因素,则在大多情况下将发生过控;若用CC来应对概率大的不确定因素,则机会成本很高。应基于风险概念来协调这两种在物理和经济特性上都互补的控制方式。配置EC以减轻CC的强度,追加CC来减轻EC的保守性,并避免系统在预先未考虑到的严重故障下崩溃。二者的协调非但关系到防御体系的强壮性,也影响到防御行为的经济性。以某实际电网为例,考虑切除时间和负荷模型的不确定性,对其切机方案进行协调优化,仿真结果表明优化后的控制方案风险代价显著减小。切负荷(LS)控制既可由故障驱动(即紧急LS),也可由轨迹驱动(即校正LS)。大量不确定因素突出了它们之间协调的重要性。这是一个时变系统的动态规划问题,需要处理各类稳定性约束、高维混合决策空间中的寻优及控制措施之间的非线性交互等难题,只能采用解耦优化-聚合协调的方法求解。首先用稳定性量化分析软件FASTEST分析不确定因素下的稳定性,然后分别优化故障驱动的LS及轨迹驱动的LS,在优化其中之一时,将另一个的最新迭代结果作为仿真的外部场景,迭代到预想故障集的风险总代价不再明显减小为止。考虑某实际系统的预想故障、负荷模型和负荷水平等不确定因素,仿真证实该协调优化方法在保证系统稳定性的前提下,显著减少了控制的风险。解列控制既可由故障驱动(即第二道防线中的故障解列),也可由轨迹驱动(即第三道防线中的失步解列)。对于完全确定性的理想情况,只需要考虑故障解列,以减少对系统的冲击;但客观存在的不确定因素模糊了连锁解列的充要条件,突出了失步解列及两者协调的重要性。将解列后,使各孤立子系统稳定的最小控制代价与故障概率的乘积之和作为解列风险,并按该指标协调两种解列方案。采用“在线预算,实时匹配”的方法提高故障解列的自适应性;由控制中心站、就地主站和解列装置组成自适应失步解列系统。其中,控制中心站负责应对振荡中心的转移;就地主站按断面功率或风险代价在线判断最优的解列割集。通过仿真验证了其有效性。更多还原

【Abstract】 The advancement of interconnected power grids and electricity markets provides strong supports for economic construction and development of our country. Meanwhile, it also brings new challenges to security, stability and reliability of power systems. Because security and stability of power systems are closely related to construction and stability of our country, it is imperative to develop new strategies to defend power systems blackouts, ensure security, stability and economical operation of power grids.Closed-loop discrete control is the main means of maintaining operational reliability of power systems after large disturbance, including generator tripping, load shedding, splitting, and so on. These measures can be divided into fault-driven emergency control and trajectory-driven correction control. Current researches focus on how to minimize control quantity or control cost of meeting stability constraints. Control schemes rely on system simulations and engineering experiences, and the decision-making is carried on with deterministic operation conditions, models, parameters, and fault scenarios.The decision-making factors of closed-loop discrete control include models and parameters, operation conditions, assumed faults, objective functions, constraints, control variables and execution time. Control strategies can not be optimized without reliable and efficient security and stability quantitative analysis algorithms. Extended Equal Area Criterion proposes the swing stability concept of high dimensional disturbed trajectories and the methods of assessing stability margin of every swing. It profoundly reveals transient stability mechanism and law of non-autonomous and nonlinear multi-machine power systems. Considering security and economy of power systems, the dissertation optimizes EC and CC, and coordinates them by using FASTEST software package. EC is a feed-forward control driven by faults. Its lag time is very small and cost-performance ratio is large. But its decision tables are made according to typical scenarios. The scenario errors maybe lead to serious over-control or deficient-control. CC is a feed-back control driven by trajectories. It is executed after insecure phenomena occur, so its cost-performance ratio is small. But its feed-back law can avoid over-control and deficient-control. The dissertation compares the criterion, execution time, cost and precision of EC and CC, as well as their generator tripping, load shedding and network splitting actions. Their optimized strategies with deterministic operation conditions and faults are also discussed.Power systems are random. Diverse stochastic factors are divided according to operation conditions and faults. Faults with different locations and types impact on stability of power systems differently. The control costs of same type faults are different when they occur at different grid topologies, unit commitment, injection power and system reserve. It is not economical and maybe brings about negative control effects to make control strategies by deterministic simulation with typical conditions and faults according to conservative principles. With consideration of non-determinism of tie-line power, clearing time, load level and load models, the influence of stochastic factors on power angle stability, frequency security and voltage security is analyzed by simulations on a Chinese power systems.If an emergency control is used to cope with uncertain factors occurring in small probability and high risk, it will most likely lead to over-controlled results in most of operational scenarios. If correction control is used in this case to cope with the factors that have a large probability to occur, the cost would be high. It is emphasized that the coordination of these two control modes with complementary characteristics physically as well as economically should be based on risk analysis. EC can reduce control strength of CC; CC can alleviate conservatism of EC and avoid power systems blackouts under unpredicted serious faults. Their coordination can improve adaptability and economy of defend systems. Considering the non-determinism of clearing time and load model, simulation studies undertaken on a power system show that the risk cost of optimized generator tripping scheme is reduced obviously.Load shedding can be driven by either faults (emergency load shedding) or trajectories (corrective load shedding). To deal with many non-deterministic factors, it is important to coordinate them. According to a decoupling-optimization and global-coordination framework, power system stability and security are analyzed with FASTEST software package considering the non-deterministic factors, and then fault-driven load shedding and trajectory-driven load shedding are optimized respectively and interactively. During the course of optimizing one of them, the newest results of another are taken as constrained scenarios. Iteration is conducted until the total risk of all assumed faults is not reduced obviously. With consideration of the non-deterministic factors of the assumed faults, load model and load level, simulation studies undertaken on a power system show that the optimal coordinative load shedding scheme reduces the risk cost obviously with a premise of maintaining stability and security of the power system.Splitting can be driven by either faults (fault-driven splitting performed at the second defense line), or trajectories (out-of-step splitting performed at the third defense line). Under absolutely deterministic conditions, only fault-driven splitting is needed to reduce impact on power systems. However, there are many of non-deterministic factors which weaken the necessity of fault-driven splitting but emphasize out-of-step splitting and the importance of coordinating them. Multiplying the control cost of all isolated sub-systems with the splitting probability is newly defined as the risk cost of splitting control proposed as an index in this paper, and the coordination is also based on this index. Then "on-line pre-decision and real-time match" is used to improve the adaptability of faults-driven splitting. The paper also proposes an adaptive out-of-step system that consists of a control center, local splitting stations and splitting devices. The control center would detect the location of an oscillation center automatically, and coordinate fault-driven splitting and out-of-step splitting according to the splitting risk. Local splitting stations would be responsible for on-line selecting the cutting sets of the power grid. Simulation studies undertaken on power systems show that this optimal and coordinative splitting scheme is effective and practical.This project is jointly supported by National Science Foundation of China (No. 50595413) and State Grid Corporation of China (No. SGKJ[2007]98&187&2009). Except for theoretical value, it is also expect to provide practical decision-making tools for interconnection and marketization of china power grids by improving in engineering earlier.更多还原