| dc.description.abstract | This thesis examines hourly aggregated data from 501 photovoltaic (PV) installations, builds a better knowledge foundation about the geographical performance of PV systems in Norway, and provides a groundwork for how PV datasets with limited metadata can be analyzed. Metadata is supplemented with inferred tilt and azimuth by analyzing the power and irradiance relationship at different orientations, with 1-degree intervals. When tested with a known PV installation, the result shows a median accuracy of 12.2 and 14.1 degrees for tilt and azimuth, respectively. To analyze the performance of PV installations, the power output data is filtered with a linear filter (RANSAC) and a non-linear polynomial filter. The latter shows promising results, as long as specific requirements regarding the number of available timestamps are available. Unknown capacity units are inferred by selecting highly probable units (Wp, kW_p, and MW_p) and finding highly probable specific yields. Installations, where highly probable specific yields are not found using these units have been removed from further analysis. | |
