Comparison of data-based methods for monitoring of air leakages into oxyfuel power plants

Air leakages compromise the CO2 capture rate and auxiliary power consumption of oxyfuel power plants. Constructive measures can significantly improve the leakage rate in newly built plants. However, the mitigation of increasing leakage rates during the plant lifetime is crucial for high plant efficiency. In this paper, we apply three statistical methods on experimental process data gathered in an air leakage test in Vattenfall's Oxyfuel Pilot Plant in Schwarze Pumpe, Germany. The performance of the methods in identifying increasing leakage rates and localizing the leakage source is investigated. It was found that all three methods can identify and localize even small increases of the leakage rate. A combination of all three methods allows taking advantage of the individual features of each method. Additional installation of CO2, O2, H2O, and SO2 measurements in the oxidizer can considerably enhance localization performance. Finally, it is shown that the results can be transferred to commercial-scale oxyfuel pilot plants by generating training data with thermodynamic plant models.