Validation and benchmark of a novel low-cost measurement system of direct and diffuse irradiance at six sites worldwide

Kurzfassung

Direct normal irradiance (DNI) and diffuse horizontal irradiance (DHI) are important measurands to characterize atmospheric conditions. Several commercial measurement systems of DNI and DHI are available. Sun trackers with pyranometers and pyrheliometers are expensive and require constant maintenance due to their susceptibility to failures and soiling. Simpler, i.e. less expensive and more robust systems, that are commercially available, typically bring considerable deviations under certain atmospheric conditions. To meet such challenges, we developed PyranoCam, a novel measurement system for DNI, DHI and global horizontal irradiance (GHI). It consists of a pyranometer and an all-sky camera and uses a combined physical and machine-learning model. In our contribution we demonstrate PyranoCam's performance by a three-fold experimental campaign as follows. First, we validate PyranoCam at six sites around the world including various climates using sun trackers with ISO 9060:2018 Class A pyrheliometers and shaded pyranometers as references. PyranoCam exhibits competitive relative root mean square errors of 3.4% to 7.4% (DHI) and 2.1% to 4.5% (DNI) at all sites including four sites that were not used for training the model. Second, at two of the sites, we benchmark PyranoCam against three commercial systems (Rotating Shadowband Irradiometer, (RSI); Delta-T SPN1; EKO MS-90). PyranoCam is the most accurate system in the test. It is followed closely by the RSI. SPN1 and MS-90 show significant deviations. Third, by an extended measurement setup at one site, we investigate influences such as the spectral composition of GHI, DNI and DHI on the accuracy of each test system. The deviations of SPN1's DHI and DNI increase sharply in situations with high circumsolar irradiance. The MS-90's DNI errors strongly depend on the reference DNI. For RSP and PyranoCam, comparatively small deviations are found under all atmospheric conditions. In particular for PyranoCam, a high transferability between atmospheric conditions is indicated.