Data-driven inference of unknown tilt and azimuth of distributed PV systems. (15th November 2020)

• Record Type: Journal Article
• Authors: Meng, B.; Loonen, R.C.G.M.; Hensen, J.L.M.
• Is Part Of: Solar energy. Volume 211(2020)
• Page Start: 418
• Page End: 432
• Publication Date: 2020-11-15
• Subjects: PV orientation estimation -- Data-driven method -- Distributed photovoltaics -- Solar irradiance; Solar energy -- Periodicals; Solar engines -- Periodicals; 621.47
• DOI: 10.1016/j.solener.2020.09.077
• Languages: English
• ISSNs: 0038-092X

Abstract

Highlights: A novel method to estimate tilt and azimuth of photovoltaic systems. Only requires hourly power generation data and off-site irradiance data. Validated with photovoltaic systems up to 200 km away from the weather station. Tilt and azimuth estimated with mean absolute errors of 4.3° and 4.5° respectively. Source code written in Python is available online to facilitate implementation. Abstract: Information about the orientation (i.e. tilt and azimuth angles) of PV modules is a fundamental input for PV performance studies. However, this type of metadata is difficult to obtain for distributed in-use PV systems, which considerably impedes monitoring and diagnostics of PV systems and power grid management. Recently proposed parameterization methods to derive PV tilt and azimuth have limited practical applicability because they rely on data that is often not accessible. Hence, the aim of this research is to develop a novel method to infer tilt and azimuth angles of distributed PV systems, utilizing widely available data. The proposed method, which is based on a curve-matching procedure, is designed to be scalable because it only requires PV generation data and off-site irradiance data at a 1-hour time interval. The accuracy of this method has been tested using notional PV systems with a wide variety of orientations, as well as with data from real PV roofs distributed across the Netherlands. These tests show that the proposed method can parameterize azimuth and tilt angles of PV panels as far as 195 km away from the irradiance measurement site with mean absolute errors of 4.5° and 4.3° respectively. A demonstration case of the proposed approach written in Python is uploaded online for other researchers to use.