The accuracy of a thermal performance test is typically estimated by performing an uncertainty analysis calculation in accordance with ASME PTC 19.1 or equivalent. The resultant test uncertainty estimate is often used as a key factor in the commercial contract, in that many contracts allow a test tolerance and define the test tolerance to be equal to the test uncertainty. As such, the calculated test uncertainty needs to accurately reflect all of the technical factors that contribute to the uncertainty. The test uncertainty is a measure of the test quality, and, in many circumstances, the test setup must be designed such that the uncertainty remains lower than test code limits and/or commercial tolerances.
Traditional uncertainty calculations have included an estimate of the measurement uncertainties and the propagation of those uncertainties to the test result. In addition to addressing measurement uncertainties, ASME PTC 19.1 makes reference to other potential errors of method, such as “the assumptions or constants contained in the calculation routines” and “using an empirically derived correlation”. Experience suggests that these errors of method can in some circumstances dominate the overall test uncertainty. Previous studies (POWER2011-55123 and POWER2012-54609) introduced and quantified a number of operational factors and correction curve factors of this type.
To facilitate testing over a range of boundary conditions, the industry norm is for the equipment supplier to provide correction curves, typically created using thermodynamic models of the power plant to predict the response of the system to changes in boundary conditions. As noted in various PTC codes (PTC-22, PTC-46, and PTC-6) it is advisable to run the test at conditions as close to the rated conditions as possible to minimize the influence of the correction curves. Experience suggests that large deviations from rated conditions, and the associated influence of the correction curves, can result in decreased accuracy in the final corrected result. A discussion of these types of situations via case studies is discussed, as well as a means by which to reduce the uncertainty contributions from correction curves considerably.