Status : Verified
| Personal Name | Dalagan, Anne Glydel C. |
|---|---|
| Resource Title | Spatial Inventory of Solar Photovoltaic (PV) Installations Using Remote Sensing and Machine Learning |
| Date Issued | December 2024 |
| Abstract | The declining costs of photovoltaic technologies have accelerated the expansion of solar ianstallations in the Philippines, necessitating an effective detection method to delineate both utility-scale PV (UPV) and distributed PV (DPV) systems, which is essential for status monitoring and implementation of relevant programs by stakeholders and decision-makers. This study aims to detect and delineate PV installations in Central Luzon using satellite imagery sources with different spatial resolutions (Sentinel-2, Planetscope, and Kompsat-3) and machine learning, with the aid of open-source GIS and remote sensing software. A semi-automated approach combining pixel-based classification (PBC) and object-based classification (OBC) was introduced to enhance accuracy. Training and validation data were extracted from the satellite images. The study also implemented a post-processing procedure using a set of spectral rules to refine the classification results. The performance for each classification approach (PBC and OBC) was evaluated using the three classifiers: Support Vector Machine (SVM), Random Forest, and Naïve Bayes. The findings revealed that all image sources showed similar superior performance in classifying UPVs, with Sentinel-2 achieving the highest F1-score of 96.83%. However, for DPVs, Kompsat-3 had the highest number of detected installations (75) and achieved delineation accuracies of 0.42 (Pampanga) and 0.5 (Tarlac), with the latter meeting the threshold of 0.5 despite the limited coverage of the study area. In contrast, Sentinel-2 struggled to detect PVs smaller than 7 pixels.vi Estimated power capacities showed errors of 7% and 9.4% for UPVs and 10% and 15.6% for DPVs when using STC irradiance and average irradiance, respectively. This research demonstrates the effectiveness of integrating remote sensing, and machine learning for PV mapping, providing insights for better monitoring and management of solar energy infrastructure. Future works may explore appl |
| Degree Course | MS Geomatics Engineering |
| Language | English |
| Keyword | Solar Photovoltaics; Remote Sensing; Pixel-based classification; Object-based classification |
| Material Type | Thesis/Dissertation |
Preliminary Pages
757.88 Kb
Category : P - Author wishes to publish the work personally.
Access Permission : Limited Access