MODEL-BASED CONDITION MONITORING TECHNIQUES FOR BALANCE OF PLANT ANALYSIS USING TEMPO

摘要

The OECD Halden Reactor Project (HRP) has taken an active role in advancing condition monitoring techniques for maintenance support. Model-Based Condition Monitoring (MBCM) is one such technique that includes physical modelling, empirical modelling, and knowledge-based modelling. The potential of MBCM to provide advanced monitoring and diagnostic functionality has been demonstrated in many studies, pilot projects, and industrial small-scale applications. The Thermal Performance Monitoring and Optimisation (TEMPO) system has been developed at the HRP and utilizes MBCM for balance of plant simulation. The data-reconciliation method used in TEMPO relies on fitting a simulation of the turbine cycle to the actual plant data. The difference between measurements and calculated values (residuals) are monitored to detect deviations. Each measurement point is assigned an uncertainty. How well the simulation fits to the measurements is compared to the given uncertainty. Traditionally this comparison is directly used to determine if there is a fault in the measurement. By using a time series analysis of plant data, changes below single point statistical significance can be found. Variations in both individual residuals and the global object function, i.e. the sum of all residuals, are small and their values mostly static. Thus, trending the global object function value is important in order to identify possible faults. Comparing residuals with past behaviour enhances fault detection compared with a statistical analysis of each data point.