Research On Online Decision Making System For Smart Substation Operation

At present,the problem of operation and maintenance of smart substation is the urgent problem to be solved.The traditional method which is relied on manual inspection is not suitable for smart substation anymore.Taking artificial intelligence method,especially the expert system,to solve the problem of operation and maintenance of smart substation has become the consensus of many expert.But there are also problems of expert system.For example there is abundance of work to complete the expert system facts,knowledge acquisition depends on the knowledge of human experts,the formation process didn't take use of the description ability of SCD files,the expert system can't learn from cases.This paper aims at analyzing and summarizing the status of smart substation operation and maintenance problem and solutions.And discussing the intelligent methods to generate an expert system through facts automatically generation and rules selflearning method.First,through reading materials on operation and maintenance of smart substation,by comparing the existing methods,analyzing the advantages and disadvantages of the existing methods.An idea to generate data base and rule base automatically is proposed.Aiming at the complex connection between intelligent substation and the huge number of equipments,a scalable model of intelligent substation based on semantic network is proposed.This thesis presents a method to generate the fact database of the expert system by using SCD file,without the over burden during the design and realization process of the fact base.Through designing the template structure of fact base,SCD file parsing and the fact content conversion to generate the fact cases of smart substation.Fact cases include substation equipment objects,primary device topology connections,secondary devices and secondary device dependencies,and redundant relationships between devices and functions.Instead of getting rules from human experts,we presents a decisiontree-based self-learning method for rules of expert system.Based on the existing rules extraction method based on decision tree,combined with the actual operation and maintenance of smart substation problems,we optimized the self-learning method on the particular area.Through this method,we could deduce rules from large number of cases quickly and accurately.Finally,based on real smart substation,we test the fact automatic generation method and self-learning method of rules.For facts generation,the test includes testing the reasonability of fact base structure and the integrity of fact instance which are generated from SCD file.For rules' selflearning,the test includes testing of the efficiency of rule learning process and testing the rules' accuracy,comprehensiveness and interpretability and so on.The test result show that the automatic generation of facts is comprehensive and accurate,the process of rules' self-learning method is efficient,those learned rules are highly accurate and interpretable.