Condition Monitoring for Parabolic Trough Fields

Abstract

Condition monitoring of solar thermal parabolic trough power plants using intelligent data analysis methods is a promising option to increase the reliability and yield of the plant while reducing costs. However, important measurement variables, such as the mirror soiling, as well as the loop fluid flow are not recorded by the built-in measurement equipment and are thus not regularly available to the operator. Determining these two variables using existing measurement technology and intelligent data analysis methods is a key aspect of this work. The determination of the two measured quantities mirror soiling and loop fluid flow represents an inverse problem, where the causal quantities must be inferred from the measured effects in the solar field. The existing measurement technology in the solar field is the basis for both approaches. For the determination of soiling, models from the field of machine learning are developed, which determine the gloss values of the mirrors. Gloss is used as representative for mirror soiling. Operational and meteorological data of the Andasol-3 power plant are used for that approach. The determination of the loop fluid flow uses the time difference of temperature gradients that arrive with the flow through the loop and are detected at different collectors. Transient simulations are used to create correction functions that allow the determination of fluid transit time between collectors. The presented method can determine the soiling values well and achieves a coefficient of determination of R²=0.77. The results of the models are compared with the current cleaning schedule of the Andasol-3 power plant. It is shown that using the method, 12.2% additional cleanings of soiled mirrors can be detected while reducing unnecessary cleanings of clean mirrors by 14.3%. The fluid flow determination method is able to reproduce well the relative mass flow distributions in the solar field and has only small deviations from the simulation results. Absolute values of fluid transit time are determined with a relative uncertainty of 2.3% for thermal oil and 3.8% for molten salt. The volumetric flow rate can be determined with a relative uncertainty of 3.0% for thermal oil and 2.0% for molten salt. With the two methods, one for the determination of the mirror soiling and one for the determination of fluid flow, it can be shown that these key measurement quantities can be calculated with the existing measurement technology. These two quantities provide valuable additional information for the power plant operator without causing additional costs for measurement instrumentation.