| Transformer is the key equipment of power system. Its reliable operation is an important guarantee of safety and stability of power system. Correctly judging transformer operating state is conducive to grasp the real-time working state of the transformer, and timely finding the transformer fault risk and reasonably arrange maintenance is significant. The development of transformer online monitoring technology makes transformer condition monitoring parameter more comprehensive, more accurately to judge the operating state of transformer. It is worthy to study how to make use of monitoring parameters to comprehensively judge the transformer state. The immune system has the unique properties such as self-organization learning and self-adaption judging, which is the significant reference to transformer state recognition. Based on immune theory, transformer state recognition can be more sensitive to distinguish its state and provide foundation for transformer control and repair strategy.First of all, this paper learn from the principle of biological immune recognition, to establish the framework of the transformer state recognition. On this basis, the paper analyzes and selects various kinds of characteristic indexes that reflect transformer running state. By selecting transformer winding insulation, electrical test and other kinds of indexes make up state vector, as the antigens of transformer state recognition. Transformer states are classified according to the index system standard. The shape-space are divided into four regions.Secondly, the self-adaptive learning algorithm of transformer immune recognition is devised. The transformer immune recognition system is divided into discriminant module and learning module. In the discriminant module, the detectors recognize transformer states by calculating the Euclidean distance between the antigen and antibody vector detector center.The learning module optimize the detector distribution by introducing the principle of mutation and clonal selection. The self-learning in learning module realizes the self-adaptive evolution in discriminant module. Through actual examples, the self-adaptive immune algorithm of transformer can realize running state recognition.Finally, this paper further improve the algorithm strategy of discriminant module in immune algorithm, add clearance mechanism and annexation mechanism to discriminant module of transformer. Optimize the spatial distribution of detection,thus further improve the efficiency and sensitivity of state recognition, devise an improved transformer immune recognition algorithm. By simulation, this paper proved that an improved self-adaptive immune algorithm has better applicability and expansibility. |
