Development Of Power Plant Fault Diagnosis Expert System And On-Line Life Monitoring Software

On the background of"Supervisory Information System of ERDOS electric power metallurgy Co. Ltd project (2×300MW)", On-line life monitoring models of power plant superheater, reheater, drum, turbine rotor and fault diagnosis expert system are studied in the thesis. The main contens are as follows:1. Life loss reason of power plant superheater, reheater, drum and turbine rotor is studied. Based on the analysis, considering the request of real time calculation, on-line life models are given.2. General creep life loss and low cycle fatigue life loss calculation modules are designed and implemented. Creep life loss calculation is based on Robinson principle and ASME standard is adopted in low cycle fatigue life loss calculation. For the aim of generalization, creep life loss and low cycle fatigue life loss calculation modules are designed based on object-oriented principle in the thesis. These modules have been used in the life calculation of superheater, reheater, drum and turbine rotor.3. The structure of expert system, knowledge acquisition and knowledge presentation are researched. The implementation of expert system knowledge base, reasoning algorithm and confidence level calculation is also studied. Based on the theory, fault diagnosis expert system based on rule and fuzzy reasoning is developed using c++ language, which has been used in SIS of ERDOS power plant. Fuzzy c-means and decision tree method are appied in determining the threshold of symptom. At the same time, knowledge base of condenser vacuum fault is built on the basis of mechanism knowledge, expert experience and relative reference. The knowledge base has been used in the testing of expert system and diagnosis result is very good. Knowledge base is also developed in the thesis with which user can easily maintain expert system knowledge base.4. Using of design patterns in SIS development is explored. Designed patterns are the summary of object oriented software development, which can help improve software architecture and enhance reuse of modules. In the development of creep life loss and low cycle fatigue life loss calculation modules, template pattern, strategy pattern and singleton pattern are used and get good results.