【作者】 徐臣；

【导师】 余贻鑫；

【作者基本信息】 天津大学 ， 电力系统及其自动化， 2009， 博士

【摘要】 环境保护、技术进步、社会发展和国家政策使得智能电网成为世界电力工业发展的必然选择。由于迫切需要提高电网自愈能力、电力基础设施利用效率、电能质量和供电可靠性的原因,智能电网的建设,将首先在配电网和终端用户展开。智能配电网的研究和实践尚处于起步阶段,涉及众多技术领域,其中配电快速仿真与模拟(DFSM)是智能配电网的基础和核心功能之一。本文紧紧围绕着DFSM及其分布式智能系统的几个关键问题展开研究工作,目的是:应用分布式人工智能的相关理论和方法建立起DFSM的分布式智能系统;并在DFSM的核心功能及相关领域取得实质性的成果。配电网三相状态估计是DFSM的核心功能和研究难点之一。对于基于支路电流的配电网三相状态估计算法,采用Java语言编制了相应的应用程序。算例分析表明该算法可以实现三相解耦计算,能够处理多种类型的量测数据,为本文的后续研究奠定了基础。应用MP广义逆矩阵和加权最小二乘问题的唯一极小最小二乘解等数学方法,推导出配电网状态估计中量测误差和状态向量估计误差之间的数学关系表达式;提出了一种新颖的量测评估和配置优化方法,避免了组合爆炸的产生,能够简单准确地确定量测装置的类型和最佳安装地点。为配电网量测系统建设和持续性优化提供了理论上和方法上的支持。结合智能科学的理论和方法研究智能配电网,提出了一个改进的人类智能模型,阐明“信息→知识→智能”的转换过程即是智能形成的核心机制,用于指导DFSM的分布式智能系统的研究。将Agent和MAS的理论与方法应用于DFSM的分布式智能系统的研发,提出Agent的八元素实现结构模型,设计并初步实现了基于MAS的DFSM分布式智能系统(masDFSM)。应用Agent的八元素实现结构模型,将配电网三相状态估计算法程序嵌入Agent的动作库中,开发了基于MAS的配电网三相状态估计并行分布式计算软件(masDSE),为进一步开发DFSM中的其它仿真分析软件和提高计算性能提供了新的方法和可供借鉴的模板。对masDSE中的任务调度问题进行研究,建立了相对应的调度问题数学模型,提出了基于模糊C均值(FCM)聚类分析算法的任务调度优化方法,使得各计算节点的计算负载均衡程度最佳,进一步提高了masDSE的计算性能。更多还原

【Abstract】 Environmental protection, technological progress, social development and national policy, all the factors together makes the choice of Smart Grid be inevitable for the world electric power industry. Because of urgent needs to improve the self-healing ability, utilization efficiency of electric infrastructure, power quality and reliability, the construction and development of Smart Grid will start from distribution system and terminal consumers.The research and practice on Smart Grid is still at the initial stage, which involve a great number of technical fields. Distribution Fast Simulation and Modeling (DFSM) is the basis and one of the key functions of Smart Distribution. This dissertation focuses on the key issues of DFSM and its Distributed Intelligence System (DIS). The aim is to design and develop the DIS of DFSM, and achieve substantive results on kernel application and related fields in DFSM.The distribution state estimation is the core function of DFSM. A three-phase state estimation algorithm based on the branch current for distribution network is realized by practical programming using Java. Large numbers of tests prove that this algorithm can perform decoupled three-phase calculation, and deal with many types of meter data, which is the foundation of later studies.The theories of Moore-Penrose generalized inverse matrix and the only minimal least squares solution of the weighted least squares problem are applied to deduce the formula, which expresses the mathematical relationship between the meter data error and the state vector estimation error in distribution state estimation; and a novel method for meters evaluation and configuration optimization in distribution network is proposed. This method can avoid the combinatorial explosion problem, and determine the type and the optimal installation site of measurement devices simply and exactly, thus provide theoretical and methodological support for the construction and continuous optimize of measurement system.The theory and perspective of intelligent science are introduced into the study on Smart Grid, and an improved human intelligence model is proposed. It explores that the conversion process of“information→knowledge→intelligence”is the kernel mechanism of the formation of intelligence, which is used to direct the study on DIS of DFSM. The theory and methodology of Agent and Multi-Agent System (MAS) is applied in the study on DFSM and its DIS, an eight-elements structure model of Agent is proposed. Furthermore, DIS of DFSM based on MAS (masDFSM) is being designed and developed.By applying the eight-elements structure model of Agent, the distribution three-phase state estimation program is embedded in the action database of Agent, and a parallel distributed computing environment for distribution three-phase state estimation (masDSE) is established based on the MAS. This work provides new methods and templates for developing other simulation software and performance improvement of DFSM.A suitable mathematical model is built for the study of the task scheduling problem in masDSE. And a new optimal method for task scheduling based on fuzzy C-means clustering algorithms is proposed, which can achieve optimal balance of calculation load for all computers, thus further improve the whole performance of masDSE.更多还原