Probabilistic modeling of a parabolic trough collector power plant – An uncertainty and sensitivity analysis

Abstract

In this work, an uncertainty and sensitivity analysis for the annual performance of a parabolic trough collector plant based on a probabilistic modeling approach of the solar-to-thermal energy conversion process has been accomplished. Realistic probability functions have been assigned to the most relevant solar field performance parameters. The Latin Hypercube sampling method has been used to create equal probable parameter combinations. The so obtained sample matrix has been used to run multiple annual electricity yield simulations in SimulCET, a validated parabolic trough collector plant simulation software, developed by the National Renewable Energy Center (CENER) in Spain García-Barberena et al., 2012. This procedure has led to a representative distribution for the annual plant performance, given the uncertainty in the input data. For this study the parabolic trough power plant model has been run in solar driven operation mode, without the use of thermal storage or fossil fuel back up. While being aware of the great influence of the solar irradiation on the power plant performance, only one single reference meteorological year has been used as solar input data. This has been done in order to emphasize the influence of technical design- as well as solar field maintenance parameters, factors that can be controlled or affected by mankind. In order to assess and rank the impact of each varied model parameter a multiple linear regression has been performed. The standardized regression coefficients, the Pearson correlation coefficients as well as the coefficient of multiple determination R2 are discussed. Varied parameters are the collector mirror reflectance, the collector mirror cleanliness factor, the collector glass tube transmittance, the collector receiver tube absorptance, and the collector receiver tube heat loss characteristic. Based on existing and published bibliography, a set of parameter distributions and ranges have been chosen for this work and the simulation results show that the cleanliness factor has the strongest influence on the model output. The cleanliness is followed (in this sequence) by the mirror reflectance, the glass tube transmittance, the receiver tube absorptance and, finally, by the receiver tube heat loss characteristic.