Simulation-based long-term fault detection of solar thermal systems

Keizer, Corry de

kassel university press, ISBN: 978-3-86219-400-1, 2012, 143 Pages

URN: urn:nbn:de:0002-34012

Zugl.: Kassel, Univ., Diss. 2012

| Price and available forms -->

Content: In this PhD thesis, a simulation-based approach for detecting faults during the operation of low-temperature solar thermal systems is developed and experimentally verified. Effective long-term automated monitoring and fault detection can ensure a good long-term performance of solar thermal systems and quick reparation of occurring
malfunctions. A review of existing fault detection methods for solar thermal systems was carried out to analyse the different methods that are based on e.g. manual fault finding, algorithms, spectral analysis or simulation-based fault detection.

The developed fault detection approach is based on an hourly, daily or monthly comparison of measured and reference energy yields that are simulated with TRNSYS. For a correct detection of a fault it is essential to have detailed information on the uncertainties of measurement and simulations. An extensive analysis is carried out to determine the
local and global sensitivity of simulation parameters and measured input data. Furthermore, different methods to calculate uncertainty margins were applied and compared: a Monte-Carlo uncertainty analysis, a minimum-maximum analysis and an empirically derived linear method based on the Monte Carlo global sensitivity analysis.

The fault detection method was applied and evaluated for three field test systems with different hydraulics and a collector area between 15 and 1290 m2. Several faults were successfully detected. Faults in the solar loop can be detected as long as the fault causes a large enough energy loss in comparison to the uncertainties of the simulated and
measured energy yield. For the field test systems with the installed measurement equipment this is ca. 25 %. Furthermore, some faults may be easier detected with this approach. Fault diagnosis turns out to be more difficult and is expected to greatly improve by integrating the simulation-based fault detection approach in a larger framework
that includes data management and algorithm-based fault detection.

Publication is available in following forms:

Full text (pdf-file, printable, Open Access - 11.40 MB)
view PDF