余国清,罗可

摘要： 在变电站智能告警专家系统中，针对业务知识的不确定性和复杂性，文中设计了一种不确定性知识的表示方法，可以使采用该方法构建的知识库更有利于推理机的推理和知识库的自学习。该方法根据领域知识将变电站信号进行分层分类，以子知识库的形式进行知识的组织，克服了知识的复杂性。而且它采用一种模糊产生式表示业务知识，将模糊推理规则映射为模糊产生式，解决了知识的不确定性问题。实验分析表明，该方法适合进行变电站关联信号的告警推理，结合系统对知识的自学习机制，可以有效降低告警的差错率和虚警率。

关键词：专家系统；知识库；模糊产生式；模糊推理

In intelligent alarm expert system for substation, the business knowledge is uncertain and complex. Focusing on these characteristics, this paper designs a method of representation for uncertain knowledge. The knowledge library built using this method is more conductive to carry on fuzzy reasonging and self-learning. Based on the domain knowledge, this method classifies signals of the substation, and organizes the knowledge in the form of sub-knowledge library，to overcome the complexity of knowledge. In order to solve the problem of knowledge's uncertainty, it represents business knowledge by a kind of fuzzy production, and maps the fuzzy reasoning rule to fuzzy production. Experiment results show that this method is suitable for alarm reasoning on related signals of the substation. Combined with the self-learning mechanism for knowledge of the system, the method can reduce the rate of missing alarm and false alarm effectively.

Expert system; Knowledge library; Fuzzy production; Fuzzy reasoning

TM76文献标识码：A

1 引言

为进一步实现变电站无人值班管理，迫切需要在监控系统中建立告警和异常处理的专家系统，对变电站信息进行智能化管理，从中提取告警信息，用以辅助工作人员进行故障判断及处理。

一个完整的专家系统由6部分组成：知识库、数据库、推理机、知识获取机制、解释机制和人机接口，其中知识库和推理机是核心部分[1]，好的知识表示和知识库组织方法可以有效提高系统进行匹配和推理的效率。

为解决一些不确定性问题，模糊逻辑已被引入到故障诊断方法的研究中[2]。本文设计了一种有效的不确定性知识的表示与组织方式。变电站业务知识虽然数量庞大且复杂，可以将它们按设备来源和信号性质进行分类细化；……