配电系统是由多种配电设备(或元件)和配电设施所组成的、变换电压和直接向终端用户分配电能的一个电力网络系统，是连接发、输电系统与用户的桥梁。据不完全统计，80%以上的用户停电事件是由配电系统的故障引起的，因此，对配电系统的可靠性进行深入研究，无论是在理论上还是在实际应用中都具有重要的意义。配电系统的网络结构可用节点-支路表示，其节点数成千上万甚至达到百万级别，而且配电系统的网络分布范围广、网络参数组合数巨大，因此可靠性评估计算具有复杂性、分布性和大规模性等特点，随着分布式发电(Distributed Generation，DG)大规模的接入，又增加了动态性和不确定性等特点。本文围绕配电系统可靠性评估的相关问题展开了一系列研究，充分考虑了老化故障对元件故障率的影响、DG接入后系统中存在的各种不确定性因素，进而研究了含DG与不含DG系统的可靠性评估方法(一组非线性函数的解析求解问题或模拟计算问题)，其中涉及交流潮流计算(一组非线性方程组的求解问题)，含DG时还涉及稳定性分析计算(非线性方程组和微分方程组的联立求解问题)。采用本文提出的可靠性与潮流交互计算的可靠性评估流程，能够同时给出配电系统在电气性能与可靠性两方面存在的薄弱环节，为规划人员制定规划改造方案提供重要的参考，从而为实现系统的安全、可靠和经济运行提供保障。本文的主要内容如下：1. 提出一种基于竞争故障分布(CF分布)的元件故障率估算方法。针对元件老化现象日益加剧，对系统可靠性影响越来越大的现状，基于竞争风险理论，构造一种新的CF分布替代常规的Weibull分布用于模拟元件的老化故障。首先，分析CF分布的主要特征；然后采用EM算法对该分布中的未知参数进行估计，并证明其似然方程唯一一致根的存在性；最后给出用于检测指数分布、Weibull分布和CF分布三者合理性的似然比检验。CF分布用于元件故障数据分析，除了考虑元件的老化故障外，同时也将发生频率较高的偶然故障纳入研究之中，更加切合元件故障的实际情况。其结果不仅能为系统可靠性评估提供准确的可靠性参数输入，而且能进一步指导不同区域设备负载率的调整，并能为元件的检修和更新计划提供决策依据。2. 提出一种计及容量约束的系统可靠性评估方法。首次将系统可靠性计算和潮流计算二者进行关联，不仅在网络拓扑结构描述上采用一致的节点-支路模型，而且实现了可靠性和潮流的交互迭代求解，从而使得网络既满足电气性能方面的约束又满足供电可靠性方面的需求；所提出的基于图论和事故模拟技术的可靠性评估方法，可实现对与事故相关的各种动态变化的模拟，所得出的系统可靠性指标是元件可靠性参数和各类开关位置及类型的非线性函数，在增加经济性约束条件的基础上可转化为一个可靠性优化计算问题，从而确定出各类开关的最优位置和数量，为可靠性的成本效益分析奠定坚实的基础。3. 提出一种计及DG不确定性的系统可靠性评估的通用解析方法。为了消除随机性很强的太阳辐照度和负荷之间的相关性，将一年的研究周期分为多个时间帧，且对每个时间帧独立进行研究；为了处理太阳辐照度和风速的不确定性，分别采用Beta分布和Weibull分布来模拟太阳辐照度和风速；为了能采用解析方法，将连续形式的概率密度分布函数离散化，建立了负荷和DG的多状态模型。最后通过引入DG的供电概率指标对不含DG的系统可靠性指标进行相应的修正，从而推导出了含DG的系统可靠性指标。所提解析方法具有物理概念清晰、建模简单和计算效率高的特点。4. 提出一种计及DG不确定性的系统可靠性评估的快速模拟方法。为了提高模拟法的计算效率，提出网络分块简化方法并构造两种故障影响矩阵，可快速地确定出故障发生后各网络分块的供电恢复情况；应用伪序贯蒙特卡洛模拟方法，对系统状态进行非序贯蒙特卡洛抽样，而仅对描述整个停电过程的故障相邻状态子序列进行序贯蒙特卡洛模拟，这样既保留了序贯蒙特卡洛模拟方法的准确性，同时也具有非序贯蒙特卡洛模拟方法的计算效率。

Power distribution system is an electric network system which composed by a variety of power distribution equipment (or component) and distribution facilities, transforms voltage and directly distributes electric energy to end-users. It is a bridge linking the transmission system and end-users. According to incomplete statistics, more than 80% of the users’ power outage was caused by the failure of distribution system. Therefore, the in-depth study of distribution system reliability has important significance both in theory and in practical applications.The network structure of distribution system can be represented by nodes- branch. The number of nodes reaches thousands or even million levels. In addition, the network is distributed widely, and the combination of network parameters is huge. Therefore, the reliability evaluation of distribution system has the feather of complex, distributed and large-scale. As the large-scale integration of Distributed Generation (DG) into the system, the system also has the characteristics of dynamic and uncertainty.This paper carries out a series of studies surrounding the related issues about the reliability evaluation of distribution system. The equipment’ aging failures and various uncertainties of system after the integration of DG into system are fully considered. The reliability evaluation method of system with DG or not are studied (a set of non-linear function), which involves the AC power flow calculation (a set of nonlinear equations); stability analysis is also involved when with DG (non-linear equations and differential equations). The interactive calculation of reliability and stability analysis is also achieved. In this way, it is able to justify the weak links of sytem both in electrical performance and reliability, and provides an important reference for planners to develop planning and rehabilitation programs. Therefore, it provides an indemnification to achieve the security, reliability and economic operation of system. The main contents are as follows.1. An estimation method of component failure rate based on competing failure distribution (CF distribution) is proposed. For the status quo of increasing aging of components, a new CF distribution is constructed as a possible alternative to the conventional Weibull distribution. First, the main characteristics of CF distribution are analysized; then the unknown parameters is estimated through EM algorithm, and the existence of a consistent root of the likelihood equations is verfied; at last, the likelihood ratio test to choose between an exponential distribution, Weibull distribution and CF distribution are presented. Thus, in addition to the aging failures, the occasional failures are also included in the lifetime data analysis. The results not only can provide the accurate parameters inputs for the system reliability evaluation, but also can further guide the adjustment of equipment loading rate of different regions. In addition, it can also provide decision making basis for the maintenance and updating of components.2. A reliability evaluation method considering the capacity constraints is proposed. The reliability calculation and power flow calculation is firstly connected, not only the consistent node-branch model is used for the network topology description, but also the interative soluation is achived. In this way, not only the electrical performance constraints are fulfilled but also the reliability requirments are fulfilled. The proposed reliability evaluation method based on the graph theory and accident simulation technology can realized the simulation of contingency-related dynamic behavior. And the resulted reliability indexes are the nonlinear function of switches locations and types. When the economic constraints are added, it can be transformed into a reliability optimization problem, and the location and number of various switches can be determined. Therefore, it lays a solid foundation for the reliability cost-benefit analysis.3. A general analytical method of system reliability evaluation considering DG uncertainty is proposed. To nullify the correlation between load and solar radiation, the study period is divided into several time segments; to handle the uncertainty of the solar irradiance and wind speed, the Beta distribution and Weibull distribution is uesed respectively; in order to used the analytical method, the multi-state model of load and DG is established by discrete the continuous function of probability distribution function; at last, the reliability indexes are derived by induction of DG supply probability index to modify the reliability index of system without DG. The proposed analytical method has the feather of physical concept clear, modeling simple and computation efficiency.4. A fast simulation method of system reliability evaluation considering DG uncertainty is proposed. Network section simplified method and two fault impact metrix are constructed in order to improve the computation efficiency. The Pseudo-Sequential Monte Carlo simulation method is first applied to reliability evaluation, which is based on the non-sequential Monte Carlo sampling of system states and on the chronological simulation of only the sub-sequences associated with failed states. In this way, the accuracy of the sequential Monte Carlo simulation method is retained and the computational efficiency of non-sequential Monte Carlo simulation method.