Experimental validation of a model-based diagnosis algorithm dedicated to a SOFC μ-CHP system

The increasing demand of low cost, high performance and durability for SOFCs, especially for stationary applications, drives the current research efforts to the development of robust and generalizable diagnosis techniques. The purpose of these techniques is to detect, isolate and identify faults within the entire SOFC system, distinguishing among the stack and the BoP components. Coupled with appropriate recovery strategies, diagnosis prevents the undesired system shutdowns during faulty conditions, increasing lifetime and reducing maintenance costs. The present study focuses on the experimental validation of a model-based diagnostic tool developed for an SOFC μ-CHP system. The tool is based on a mathematical model, which simulates the system in different nominal operating conditions and it is crucial both for the monitoring and the detection phase. The first part of this work describes the development of the mathematical model through the exploitation of experimental data gathered during dedicated tests. The second part focuses on the definition of the inference algorithm, which can distinguish among several type of faults and it is critical to determine which system variable should be monitored. Furthermore, the complete tool is tested on-line inducing controlled faulty operation on the system.