In countries with cold winters and substantial snowfall, such as Sweden, snow losses can limit the energy generated from PV systems by up to 100% monthly and up to 20% annually. Therefore, snow loss modeling is required to effectively predict the generated PV electricity in these locations. This thesis investigates the implementation of snow loss models to improve theaccuracy of hourly simulated Photovoltaic (PV) power generation. Four different snow loss models were identified as potential options, which was selected based on the purpose and constraints of the PV power simulation model used. The evaluated models include the Marion Model, The Modified Marion Model, The SunPower Model, and the Combined model, which is a combination of the three other models. The assessment of the snow loss models was conducted using 29 PV reference systems, predominantly located in Sweden (26 systems), with two additional systems in Norway and one in Estonia. The reference systems include known system characteristics, such as PV module parameters, tilt, azimuth, and measured PV power generation data, depending on the systems, between the years 2017-2023. Initially, all PV reference systems were simulated without applying the snow loss model, followed by simulations incorporating the snow loss models. The performance of the snow loss models were evaluated by the difference in the coefficient of determination, R2, before and after the respective snow loss model was implemented. Furthermore, all tested models were optimised based on various parameters. For all models except the SunPower model, the optimisation involved adjusting the snow clearing coefficient, and the snow depth threshold, THsnowfall. Following the model optimisation and comparison, all snow loss models demonstrated improved accuracy in compared to the baseline simulations. Among these models, the Modified Marion model was recommended due to its low complexity and its notably improved accuracy. Specifically, the Modified Marion model yielded an average monthly improvement in R2 values ranging from 0.1 to 0.14 for all winter months except for March (0.004), with an overall average improvement of 0.0094. The estimated annual snow losses using the Modified Marion model ranged from 0.02% to 12% over the period from 2017 to 2023, with most systems experiencing values between 2% and 6%. Finally, the monthly losses were estimated to reach up to 100% for the northernmost systems. The main challenges of the recommended snow loss model include lower performance in March compared to other winter months for most systems, as well as an overall decreased accuracy for the northernmost systems, where substantial snowfall is present. However, for systems with moderate snowfall, the model generally demonstrated increased performance, which can be of value for Distribution System Owners conducting PV power simulations for grid planning and for solar power forecasting.
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:uu-531435 |
| Date | January 2024 |
| Creators | Öhgren, Gustav |
| Publisher | Uppsala universitet, Institutionen för samhällsbyggnad och industriell teknik |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
| Format | application/pdf |
| Rights | info:eu-repo/semantics/openAccess |
| Relation | UPTEC ES, 1650-8300 ; 24011 |
Page generated in 0.0018 seconds