A Benchmark of Simple Diffuse and Direct Irradiance Measurement Systems

Abstract

Accurate measurements of direct normal irradiance (DNI) and diffuse horizontal irradiance (DHI) are required in solar energy projects at various stages. Several rather simple, i.e. economic and robust, measurement systems for DNI and DHI are available on the market. Such sensors may yield large deviations under various atmospheric conditions, or be susceptible to soiling or user mistakes. Previous validations and benchmarking of simple radiometers for solar energy applications did not include all systems which are relevant in 2023. Also, most previous benchmarking did not measure and analyze the influence of atmospheric parameters like circumsolar irradiance which may directly influence the measurements of these sensors. Thus, it can be difficult to identify an adequate sensor for a project’s site and application. We aim to close this gap by benchmarking relevant measurement systems, i.e. Rotating Shadowband Irradiometer (RSI), RSP 4G; Delta-T SPN1, EKO MS-90, PyranoCam, Sunto CaptPro, at six sites located in various climates. In the following, these sensors are referred to as RSI, RSP, SPN1, MS-90, PyranoCam and CaptPro, respectively. We include PyranoCam which is a novel measurement system for DNI, DHI and global horizontal irradiance (GHI) presented in a previous publication (Blum et al. 2022). It consists of a pyranometer and a fisheye camera that takes images of the whole sky and employs a combined physical and a machine-learning model. At two sites, all sensors are compared while different subgroups of the sensors are included at the other sites. At one site we also measure effective wavelengths of the global, direct and diffuse irradiance, circumsolar irradiance, cloud cover, Linke turbidity index, and other meteorological parameters. Based on that, we identify how atmospheric and other relevant parameters influence the measurements of each sensor. We detected notable differences in the performances of the tested systems. The rotating shadowband systems (RSI, RSP) and PyranoCam yield similar deviations, as indicated by root mean square deviation (RMSD), mean absolute deviation (MAD) and bias, and are clearly the most accurate systems at both sites evaluated so far. SPN1 and MS-90 yield considerable deviations. Finally, CaptPro clearly yields the largest deviations. For SPN1 and MS-90, the metrics vary considerably between the test sites. This indicates a dependence of these radiometers’ performances on the local atmospheric conditions. Using the mentioned additional meteorological measurements, we connect the deviations of each sensor to the actual physical causes. We show for example that MS-90’s and SPN1’s DNI measurements have the strongest shortcomings in situations with large circumsolar irradiance. This in turn is more pronounced in turbid conditions or for GHI enhancement events caused by clouds. Our study allows to estimate the sensors’ performance for a specific application and climate more accurately. It also provides insight on ways to improve the sensor technologies.