Model-Based Sensor Placement for Component Condition Monitoring and Fault Diagnosis In Fossil Energy Systems

With the current energy related concerns, it is imperative that the efficiency and effectiveness of fossil-fuel-based power plants are improved tremendously.  One of the reasons for the loss of efficiency and effectiveness of such plants is due to low availability caused by the shutdown of a section of the plant or the whole plant after short span of operation. This shutdown of equipment/plant may occur immediately or over a period of time. The root cause for productivity losses, the “faults”, can be anticipated or unanticipated. Even though many of the faults are anticipated, the true evolution of faults depends upon the current dynamical state of the process and hence, cannot be properly predicted based on past operational experiences or even using high-fidelity steady-state models. If the faults can be identified in a timely manner then it may be possible to avoid serious damages to the equipment/plant. It is also possible to plan for the shutdown in advance. This improves the plant efficiency by improving the plant availability and reducing economic losses that result due to damage to valuable equipment. In this work, this problem is addressed by optimal placement of sensors. In this paper, we propose an innovative two-tier framework for the identification of locations, types, and optimum number of sensors for detection of system-level and component level faults in a coal-based IGCC (Integrated Gasification Combined Cycle) plant.  This is a challenging problem because of the significant interaction between the various components and nonlinearity in the process. Several illustrative examples will be discussed to demonstrate the power of the proposed approach in identifying maximally informative sensor locations.