Spectral-matching-ratio modelling based on ANNs and atmospheric parameters for the electrical characterization of multi-junction concentrator PV systems. (1st August 2018)

• Record Type: Journal Article
• Authors: Almonacid, Florencia; Fernández, Eduardo F.; Almonacid-Cruz, Bernardo; Rodrigo, Pedro M.
• Is Part Of: Energy. Volume 156(2018)
• Page Start: 409
• Page End: 417
• Publication Date: 2018-08-01
• Subjects: Photovoltaics -- Multi-junction solar cells -- Electrical characterization -- Spectral performance -- Mathematical modelling; Power resources -- Periodicals; Power (Mechanics) -- Periodicals; Energy consumption -- Periodicals; 333.7905
• DOI: 10.1016/j.energy.2018.05.105
• Languages: English
• ISSNs: 0360-5442

Abstract

Abstract: One of the most critical issues to evaluate the performance of multi-junction (MJ) concentrator photovoltaic (CPV) systems is related to its spectral dependence. The spectral matching ratio (SMR) index is nowadays widely used to evaluate the spectral impact on CPV systems. The limitation of the present models devoted to estimating the SMR is related to the difficulty of obtaining high-quality data of aerosols and water vapour. This paper aims to fill this gap by introducing a novel approach based on commonly available variables in atmospheric stations and/or databases. In particular, the impact of aerosols has been quantified trough the ratio DNI/GNI (i.e. direct and global normal irradiances), while the impact of water vapour has been quantified through the air temperature (Tair ) and relative humidity (Hr ). Due to the complexity for finding appropriate relationships between these variables and the SMR indexes, an artificial neural network (ANN)-based model has been used. The model shows a high quality in the evaluation of the spectral performance of MJ CPV systems through the estimation of the SMR indexes, with a correlation coefficient ranging from 0.79 to 0.98, a Root Mean Square Error ranging from 2.32% to 4.32% and a Mean Bias Error around 0%. Highlights: A method for characterizing the spectral matching ratios of CPV systems is developed. Artificial neural networks are used to capture the non-linear behaviour. Easy to get atmospheric variables are considered as inputs. Root mean square errors from 2.32 to 4.32% are obtained from the experimental data. The method opens the opportunity of evaluating the spectral impact in any location.