Validation and comparison of three power prediction models for CPV modules

Abstract

Accurate predictive energy modelling of a solar farm requires a thorough understanding of solar spectral variations, along with the spectral response and optical properties of the photovoltaic system. This paper investigates the minimum data required to accurately predict power output from CPV modules, comparing modelled output to both measured data and the existing method used by Sandia PV Array Performance Model (SAPM). Three models were derived based on various weather inputs. A Detailed Spectral Model (DS) uses SMARTS, inputting measured air mass, aerosols, ozone and water content, and incorporating measured DNI to account for cloudy days. The Sub-System Algebraic Model (SSA) removes the need for instantaneous spectrum calculations by creating equations for each sub-cell and DNI, based on the same inputs as the DS. These two models rely heavily on aerosol, which is not readily available. Alternatively, an Empirical model (EMP) may be used to determine the relationship between measured output power and easily measureable weather data (ambient temperature, air mass, direct normal irradiance and water content). These non-linear DS, SSA and EMP models have a bias error of 3.09 %, 4.24 % and −0.67 %, respectively. It was also found that the SSA model can be used in lieu of the SAPM.