Application Of Information Fusion Technology Of Improved D-S Evidence Theory In Fault Diagnosis Of Rotating Equipments Of Generating Units

Supplying power for residents and enterprises uninterruptedly becomes increasingly important with the electricity consumption rising up rapidly in the 21 century, and it is also closely related to a sound development of national economy. Although new energies has been rising up in recent years, they are sparsely scattered in a short range and not mature in technique, therefore, traditional powers, especially thermal power remains in the dominant position and how to better identify the breakdown is put highly on our agenda. While the existing methods in identifying malfunction can not well meet the ever complicated mechanism of high-rating generator, it is necessary to make a research on the application of information fusion theory in settling problems on high-rating rotating equipments of generating units.In the following for the problem that result from D-S evidence theory is contrary to the real fault when there are conflicts among evidences in the process of information fusion, two methods were introduced which can improve traditional D-S evidence theory by modifying evidence sources. One of the methods is based on fuzzy membership function and average evidence distance to distribute weight index for each evidence, the availability of which was checked by experiments, and this method was used in the fault diagnose of turbine of rotating equipments of generating units. The other is based on evidence entropy of random set and similarity of evidence the availability of which was checked by experiments, and this method was used in the fault diagnose of generating motor of rotating equipments of generating units.In the last, as for the question that it is hard to achieve BPA (basic probability assignment) for D-S evidence theory, this paper tried to propose a new approach based on attribute reduction of rough set under random set theory and degree of rule and extended decision rule to achieve BPA for D-S evidence theory, and checked it with examples. D-S evidence theory rough set theory and evidence entropy adopted in this paper are all based on random set theory. By this way, a uniform mathematical framework is set up and it is also very convenient to research the whole fault diagnose.