Photovoltaic system modeling: A validation study at high latitudes with implementation of a novel DNI quality control method. (1st July 2020)

• Record Type: Journal Article
• Authors: Böök, Herman; Poikonen, Antti; Aarva, Antti; Mielonen, Tero; Pitkänen, Mikko R.A.; Lindfors, Anders V.
• Is Part Of: Solar energy. Volume 204(2020)
• Page Start: 316
• Page End: 329
• Publication Date: 2020-07-01
• Subjects: PV modeling -- PV monitoring -- PV performance -- PVGIS -- Direct normal irradiance -- Quality control; Solar energy -- Periodicals; Solar engines -- Periodicals; 621.47
• DOI: 10.1016/j.solener.2020.04.068
• Languages: English
• ISSNs: 0038-092X

Abstract

Highlights: A widely used PV model (Huld et al.) is analyzed and adjusted in a Nordic context. Modeled PV output, minute by minute, is in good agreement with in-situ measurements. Results useful for monitoring PV health: e.g., shadowing effects are recognized. Finnish sites suffer considerable losses in PV production due to snow cover. A Quality Control method is presented for GHI and DHI -based calculated DNI. Abstract: As the share of photovoltaics (PV) in electricity production increases, accurate modeling and forecasting of its output becomes critical. PV output is however affected by multiple factors, making accurate modeling of these systems a non-trivial task. This study provides new insight on the performance of a popular parametric PV output model in a Nordic context. The model and its subcomponents – including estimations on plane-of-array radiation, PV module temperature, and PV output – are here evaluated against dedicated measurements at two Finnish sites. In addition, a novel Quality Control (QC) approach is introduced for handling calculated direct normal irradiance (DNI) values. All reviewed methods show good agreement with the references. The proposed QC efficiently filters unrealistic calculated DNI data, providing a potential approach for DNI QC implementation. In snow-free conditions, the selected PV module temperature scheme has a mean absolute error (MAE) around 2 °C, while the bias is below 1 °C. The plane-of-array radiation model has a MAE of 10–15 W/m 2 with a bias of smaller than ±10 W/m 2, depending on the site location and available input data. Altogether, the PV model is shown to provide relevant PV system performance information while demonstrating precise performance in snow-free conditions. By implementing site-specific parameter optimization, MAE was reduced from 19 to 14 W/kWp and bias lowered from 15 to 7 W/kWp. For the locations studied, the estimated PV output losses due to snow cover for the winter period 2017–2018 are estimated to be up to 1.5 months of summer production.