The Design Of The Power Fault Recorder In Intelligent Substation

When the power system failure or oscillation, fault recorder is one of the most important electrical devices to record the entire process automatically. With the construction of intelligent substation, power system fault diagnosis and recovery has shown some new features, and the research for intelligent substation fault recorder has great application value.The graduation design is aimed to design fault recorder in smart substation based on IEC 61850. Firstly, this graduation design analyzes and compares principle of fault recorder in smart substation and conventional substation, and proposes the main technical indicators of fault recorder; then analyzes three kinds of smart substation network solution, and on this basis, proposes the program which is applicable to this graduation design, meanwhile designs access solutions for packets; nextly, analyzes information model and its modeling based on the IEC 61850 standard, and three kinds of information transmission models are introduced, in view of this, establishs fault recorder model and designs its System Configuration Document; lastly, hardware and software has been designed, in hardware design, this graduation design has designed select the chip model of IPC, in software design, establishs the function of each module and their specific workflow.The design has been focused on software design, in which, network data processing module, the packet processing module, data synchronization module and criterion of fault recorder startup module is the key. Network data processing module has completed the decoding of network packets, and the packet processing module has completed the parsing of SV packets and GOOSE packets. In criterion of fault recorder startup module, the traditional algorithms is analyzed and introduced, at the same time, two kinds of innovative algorithms are put forward, one is criterion startup algorithm based on PNN neural network, the other is criterion startup algorithm based on BP and Elman neural network. After verification, both run well. T...