INVESTIGATING PHOTOVOLTAIC SOLAR POWER PRODUCTION USING REMOTE SENSING TECHNOLOGY
| dc.creator | Daniel Czirjak | |
| dc.date.accessioned | 2022-01-25T19:12:51Z | |
| dc.date.available | 2022-01-25T19:12:51Z | |
| dc.date.issued | 2019 | |
| dc.description.abstract | Remote sensing platforms have consistently demonstrated the ability to detect, and in some cases identify, specific targets of interest and photovoltaic solar panels are shown to have a unique spectral signature that is consistent across multiple manufacturers and construction methods. Solar panels are proven to be detectable in hyperspectral imagery using common statistical target detection methods such as the Adaptive Cosine Estimator, and false alarms can be mitigated through the use of a spectral verification process that eliminates pixels that do not have the key spectral features of photovoltaic solar panel reflectance spectrum. The Normalized Solar Panel Index (NSPI) is described and is a key component in the false alarm mitigation process. After spectral verification, these solar panel arrays are confirmed on openly available literal imagery and can be measured using numerous open-source algorithms and tools. The measurements allow for the assessment of overall solar power generation capacity using an equation introduced that accounts for solar insolation, the area of solar panels, and the efficiency of the solar panels conversion of solar energy to power. Using a known location with readily available information, the methods outlined in this paper estimate the power generation capabilities within 6% of the rated power. Airborne hyperspectral imagery, however, does not have the capacity to collect data on a regional or global scale that will aid in the assessment of planetary photovoltaic solar power generation growth. The recently launched WorldView-3 16-band VNIR/SWIR multispectral satellite presents the opportunity to apply methods developed for hyperspectral imagery to detect photovoltaic solar arrays and assess solar power production over a much larger geographic extent, and can be used to measure progress on a global scale within UN Sustainable Development Goal #7: Affordable and Clean Energy. Using a target detection process based upon the adaptive cosine estimator (ACE), along with an adapted spectral verification routine using both the normalized solar panel index and the new normalized ethyl-vinyl acetate index, it is proven that solar arrays can be effectively detected using WorldView-3 imagery in vicinity of Palo Alto, CA. As part of the processing algorithm, false alarms are mitigated using both true detection and false alarm training libraries based upon in-scene spectral signatures compiled over multiple processing iterations. Using a location within the image scene that has a well characterized solar array, the processing algorithm is able to detect the array and estimate its solar power generation capacity to within 4% of the array’s rated power production. Based upon these results it is likely that this process can be adapted to other geographic regions across the globe to assess solar power generation installation and growth. lt is assessed that WorldView-3 VNIR/SWIR imagery and its associated processing is capable of detecting photovoltaic solar arrays using spectral detection techniques at a level comparable to that of AVIRIS-NG hyperspectral imagery if the target solar array is large in size relative to the pixel size of the MSI imagery. Under the right circumstances, smaller, residential-sized solar arrays are detectable with the MSI data, but HSI is better suited to this level of detection. This is primarily due to the higher spatial resolution of the HSI data; however, since WorldView-3 data has been released at its native spatial resolution of 3.7 meters, it likely that the MSI data performance against these smaller solar arrays will increase. This assessment, however, will need to be re-evaluated as the spatial resolution restriction has been lifted and a similar study must be conducted to evaluate the validity of the conclusions presented here. The research presented has definitively shown that remotely sensed data can be used to detect photovoltaic solar panels using both hyperspectral and 16-band multispectral data. The process involves the use of many different sources of data and provides a framework for developing large scale application of remotely sensed data to aid in the observation of progress towards attaining several different UN Sustainable Development Goals. As algorithms and processing techniques advance, this work will provide useful information to those who expand on these ideas and implement them in a practical manner in future research. | |
| dc.identifier.uri | https://hdl.handle.net/1920/12338 | |
| dc.title | INVESTIGATING PHOTOVOLTAIC SOLAR POWER PRODUCTION USING REMOTE SENSING TECHNOLOGY | |
| thesis.degree.discipline | Geoinformatics and Geospatial Intelligence | |
| thesis.degree.grantor | George Mason University | |
| thesis.degree.level | Ph.D. |
Files
Original bundle
1 - 1 of 1