Loading...
Soiling Detection on Solar Panels Using Artificial Intelligence
Kaya, Hussein Mohammad Ali
Kaya, Hussein Mohammad Ali
Date
2024-12
Authors
Advisor
Type
Thesis
Degree
Citations
Altmetric:
Description
A Master of Science thesis in Engineering Systems Management by Hussein Mohammad Ali Kaya entitled, “Soiling Detection on Solar Panels Using Artificial Intelligence”, submitted in December 2024. Thesis advisor is Dr. Zied Bahroun and thesis co-advisor is Dr. Noha Hussein. Soft copy is available (Thesis, Completion Certificate, Approval Signatures, and AUS Archives Consent Form).
Abstract
The demand of energy has shown a significant increase worldwide over the past few years. Solar energy is one source that can be the solution of our future. One of the most significant issues that has a substantial impact on the solar panel is soiling. Soiling accumulation generates losses in energy efficiency and decreases the electricity output. Several research papers have worked on proposing systems to investigate this issue. However, there is a lack of analysis regarding the inspection tools that lead to selecting the optimal ones. In this research, a new system of inspection tools and a model is proposed to detect the soiling on solar panel. The objective of this study is to provide a low-cost system that integrates between low-cost inspection tools and an accurate model to assist in precise detection of soiling on solar panel. Different inspection tools were examined experimentally to assess their performance and ability to detect soiling. Two setups were conducted, a low-cost system setup and a high-cost system setup, Additionally, a machine learning was utilized to train different models, to come up with a model with high accuracy for processing the collected data to detect soiling. Finally, Multi-Attribute Utility Theory (MAUT) was applied to obtain the most feasible and optimal combination of tools for the proposed system. As a result, a configuration comprising a voltage sensor (0-25V), high-cost current sensor 30A, and low-cost dust sensor GP2Y1010AU0F, was selected using MAUT and trained using machine learning. The Gaussian process regression model demonstrated high accuracy value for the proposed system, achieving an R-squared value of 99% and an RMSE value of 0.0022784.