Presentation Details
| Impact of Multi-instrument Quality Control on Reliability of Solar Measurements Oliver Osvald, Peter Mihalik, Rastislav Hrncirik, Marketa Hulik Jansova, Tomas Cebecauer, Katarina Blstak Catlosova, Assa Camara, Martin Jakubik. Solargis s.r.o., Bratislava, Slovakia |
Abstract
Contemporary solar projects increasingly include redundant measurements of the same type, e.g., two pyranometers measuring global horizontal irradiance (GHI). This work introduces and quantitatively evaluates an automated multi-instrument quality control (MIQC) method that exploits this redundancy to detect issues that single-instrument and physical-limit checks systematically miss. The algorithm compares collocated pairs against a time-dependent threshold accounting for temporal resolution, solar elevation, atmospheric variability, and typical dataset discrepancy, then attributes each flag to the most likely affected instrument. Evaluated on Class A redundancies from publicly accessible World Bank ESMAP and NREL SRRL sites, MIQC reduces inter-instrument RMSD by 29% and MAD by 10% on average over a baseline that already includes Solargis physical and clear-sky QC. Segmentation by issue type confirms that MIQC's newly captured anomalies are dominated by soiling (34%), calibration drift (19%), and dew (12%) — the issues that propagate directly into biased yield estimates and site-adapted satellite time series. We further show that performance is governed by sensor collocation, for which we propose a coordinate-free diagnostic (the collocation ratio), and by instrument class, which requires threshold relaxation when lower-class sensors (RSI, SPN) are included. Building on the collocation ratio, we extend the method to setups with N > 2 sensors via a pairwise grouping rule for measurement fleets.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.
No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the author.