Research On Power Grid Fault Diagnosis And Its Framework

At present our country’s economic and construction are entering a rapid development stage, the dependence on power system of all kinds of industry has been increasing. The stable operation of the power system has become the main task of our country which related to the national economy and people’s livelihood. But large-scale blackouts always occurred because of the operation mistakes, man-made destruction and natural force. So the self-healing of power system which including the rapid diagnosis after faults, fault isolation and self-healing become a major research direction of smart grid. As the precondition of self-healing, the power system fault diagnosis has been the focus of the research topic at home and abroad.On the basis of summing up the research experience of the power system fault diagnosis and learn from the good developed ideas of artificial intelligence, this paper did the further research include the aspects of power grid topology, knowledge representation of fault diagnosis, alarm message missing and system structure. The detailed research work and contribution are shown as following:(1) Two dimension adjacent lists for the power gird topology was introduced in order to overcome the weak point that topology matrix of the power grid lead to low flexibility and bad extension. Because power grid topology is the necessary means to find outage area, the efficiency of topology is very important to all the fault diagnosis system. IEC61970adopts CIM/SVG as the standards of graphic modeling for the power system, but its application is only limited in graphic integration and static topology expression at present. This paper proposed a new topology generation and searching scheme based on CIM which can reflect the actual circuit diagram. The method avoids directly searching on the line diagram-XML file which is the bottleneck restrictions of this system. The system makes SVG operation program no longer as an independent system, but as the data source of the power network topology by cooperating with SCADA. This paper also put forward a new two dimension adjacent lists data structure based on this topology method, which makes searching algorithm more real-time and reliable than the traditional method.(2) The characteristic of faults was transformed to knowledge which the computer can accept. Reasonable fault diagnosis algorithm was extracted by mapping from fault knowledge. Knowledge is the basis of artificial intelligence. In order to make the computer has intelligence which can simulate the behavior of human, it must possess knowledge. So the knowledge representation of fault diagnosis is decisive to the fault diagnosis algorithm. In order to make the fault diagnosis algorithm faster and exacter, this paper provides simplified and completed fault knowledge to intelligent diagnosis method. Firstly, provide the thought that use completed fault knowledge to diagnose power gird fault and realize it. Secondly, defines the topology knowledge and protection knowledge, put forward informationization method for both of them. Finally, improve the efficiency of fault diagnosis by optimizing the knowledge and use petri net to prove it.(3) Creates a new power system fault diagnosis considering absence of alarm messages of circuit breakers on the border of outage area. Because the alarm messages of circuit beakers are used not only to diagnose fault sections but also to calculate outage area, it becomes the key point of fault diagnosis algorithm. There is little report about fault diagnosis algorithm under the circumstance that alarm messages of circuit beakers are missing. This paper puts forward a new method which can search the outage area under the circumstance that alarm messages of circuit beakers are missing. Based on graph theory, one or more outrage area can be partitioned by cut-set of nodes. Multi-objective optimization method skyline query was applied to sort the outrage area decreasingly based on the precision rate of fault section and the descending order was provided to the diagnosis algorithm. The developed diagnosis method could make the diagnosis program much faster, and can insure the fault section will not be omitted. Petri net was applied as the diagnosis tool and several examples showed the accuracy of skyline.(4) Because the power grid fault diagnosis system’s real time demand is rigorous, it depends too much on the information transmission capacity of communication digital link. When the fault occurs, the automatic devices of the power system would produce a mass of alarm messages and datagram congestion will be happen. This can lead frame loss or data distortion. In order to reduce the reliance on communication and promote the efficiency of fault diagnosis system, this paper makes the following improvement to the traditional fault diagnosis framework. On the data acquisition layer, this paper use data grid to collect and pretreatment the information from transformer substation. It can provide uniform data view to the fault diagnosis program. This method not only solves the communication problem of fault diagnosis system, but also avoids excessive data in the control center. On the arithmetic layer, this paper designs a distributed fault diagnosis framework according to the characteristics of data grid. The distributed system can provides high performance computing strategy to fault diagnosis arithmetic, so the fault diagnosis framework can meet the practical engineering requirements.From the above, this paper put forward a fault diagnosis project which includes fault knowledge representation, fault diagnosis arithmetic and fault diagnosis framework based on comprehensive investigation and engineering practice. This paper will enrich the research status of fault diagnosis domain and will enhance the fault diagnosis performance of the power system.