配电系统对用户供电可靠性水平有着显著影响。随着用户对供电质量要求的提高，配电系统的可靠性评估问题也越来越受到人们的重视。目前配电系统可靠性评估方法主要采用解析法和蒙特卡洛模拟法，但是只能进行灵敏度分析和可靠性指标计算，给出配电系统的整体可靠性水平，难以有效地识别配电系统具体可靠性的薄弱环节。贝叶斯网是概率论和图论相结合的产物，它一方面用图论的语言直接揭示问题的结构，另一方面又按照概率论的原则对问题的结构加以利用，降低推理的计算复杂度。近年来，贝叶斯网在众多不同领域得到广泛应用。本文建立了适用于中压配电系统可靠性评估的贝叶斯网络模型。根据中压配电系统的结构特点，首先引入数据结构中的邻接表来保存配电网拓扑结构，然后在此拓扑结构上利用深度优先搜索技术找出中压配电系统的各种故障模式，据此将中压配电系统划分为不同的区域，并将其作为贝叶斯网络的子模块，从而建立了中压配电系统的贝叶斯网络模型。基于该贝叶斯网络，采用变量消元算法对其进行因果推理，得到了各负荷点及系统的可靠性指标，进行诊断推理可得到各种因果假设下的条件概率，为工程技术人员采取相应的增强措施提供参考。由于该贝叶斯网络以区域为单位建模，从而可以减少对元件逐一分析的计算量、简化贝叶斯网络结构、提高贝叶斯网推理的效率。基于贝叶斯网络相关理论和可靠性评估模型，本文还开发了中压配电系统可靠性分析软件，该软件使用方便、易于扩展，可广泛应用于配电系统的可靠性评估。最后通过对IEEE提出的配电系统测试系统（RBTS）算例的研究，验证了所建贝叶斯网络模型和推理算法的正确性及有效性。

The contribution of distribution systems to overall customer reliability is quite significant. There is a growing concern in distribution-system-reliability assessment due to the improvement of customer requirement. At present, Monte Carlo method and Analytical method are widely used in power system reliability assessment. However, they can only calculate the reliability indices, make sensitivity analysis, give out the overall reliability level of the system but can’t effectively identify the weak elements.Bayesian network is a product of the combination of probability theory and graph theory. On one hand, it reveals the problem’s structure by using the language of graph theory directly. On the other hand, it makes use of the problem’s structure in accordance with the principles of probability theory to reduce the computational complexity of reasoning. In recent years, Bayesian network is widely used in many different fields.This paper firstly builds a Bayesian network model which is suitable for reliability assessment of mid-voltage distribution system. According to the characters of mid-voltage distribution system, the adjacency list is introduced to save its topology structure. Then based on this topology structure, failure modes of mid-voltage distribution system are found by using depth first search. The mid-voltage distribution system is divided into different regions as a Bayesian network module. Therefore, the Bayesian network model of mid-voltage distribution system is built. Based on this Bayesian network, the reliability indexes of system and various load points can be obtained by the casual inference and the weakest link can be obtained by the diagnosis inference. Since the Bayesian network is built with region units, the computational complexity is reduced and the structure of Bayesian network is simplified, so the efficiency of Bayesian reasoning is improved.Based on Bayesian network theory and reliability assessment model, this paper also develops the reliability analysis software. This software is easy to use and extend, so it can be widely used in distribution system reliability assessment.At the end of this paper, the model is applied to IEEE reliability test system (RBTS). The results show that the Bayesian network model and the reasoning algorithm are correct and effective.