Home / Current Issue / Paper 1712654
A Novel Hybrid AI Model for Proactive Solar Grid Maintenance: Enhanced Fault Diagnosis and Predictive Remaining Useful Life Estimation
Subject area: Science,Engineering and Technology · Area of research: Artificial intelligence
DOI: https://doi.org/10.64388/IREV9I6-1712654
Abstract
This study creates a new mixed artificial intelligence (AI) system to solve the important problem of making sure that modern solar photovoltaic (PV) grids are reliable and work efficiently. Large-scale solar systems are naturally variable and prone to faults, making traditional care methods, which are often reactive and ineffective, hard to handle. To get around these problems, the best parts of an Artificial Neural Network (ANN), a Support Vector Machine (SVM), and a CatBoost regressor were put together to make a multi-tiered AI model. The ANN is a complex feature extraction device that looks at raw data from a simulated 100 kVA PV system to find patterns. The set of features that were created is then sent to the SVM for accurate problem detection, a job that it does very well. Lastly, a CatBoost regressor is used to guess how much useful life (RUL) a part still has, which makes it possible for maintenance to be truly proactive. The model was tested on a dataset that shows how things really are in the real world, with different types of faults and operating factors. Based on real-world data, the mixed model was 97.5% accurate at classifying, which is much better than independent models like SVM (82.3%) and ANN (88.1%). Also, the framework was able to pinpoint faults to specific words in the array and make accurate RUL predictions. For example, it predicted that a shading fault would last 22.5 days, which was very close to the actual 23 days. This study shows a scalable and effective way to do real-time tracking and preventative maintenance. This has huge implications for making solar energy systems and the smart grids they support last longer, be more reliable, and work more efficiently.
Keywords
Hybrid AI Model, Solar Photovoltaic (PV), Smart Grid, Fault Diagnosis, Predictive Maintenance, Remaining Useful Life. (RUL)
How to cite this paper
@article{1712654,
author = {Michael Obi Ogar-Abang, Julie C. Ogbulezie, Gertrude Austin Fischer, Samuel Okon Essang},
title = {A Novel Hybrid AI Model for Proactive Solar Grid Maintenance: Enhanced Fault Diagnosis and Predictive Remaining Useful Life Estimation },
journal = {Iconic Research And Engineering Journals},
year = {2025},
volume = {9},
number = {6},
pages = {1079-1089},
issn = {2456-8880},
url = {https://www.irejournals.com/formatedpaper/1712654.pdf},
abstract = {This study creates a new mixed artificial intelligence (AI) system to solve the important problem of making sure that modern solar photovoltaic (PV) grids are reliable and work efficiently. Large-scale solar systems are naturally variable and prone to faults, making traditional care methods, which are often reactive and ineffective, hard to handle. To get around these problems, the best parts of an Artificial Neural Network (ANN), a Support Vector Machine (SVM), and a CatBoost regressor were put together to make a multi-tiered AI model. The ANN is a complex feature extraction device that looks at raw data from a simulated 100 kVA PV system to find patterns. The set of features that were created is then sent to the SVM for accurate problem detection, a job that it does very well. Lastly, a CatBoost regressor is used to guess how much useful life (RUL) a part still has, which makes it possible for maintenance to be truly proactive. The model was tested on a dataset that shows how things really are in the real world, with different types of faults and operating factors. Based on real-world data, the mixed model was 97.5% accurate at classifying, which is much better than independent models like SVM (82.3%) and ANN (88.1%). Also, the framework was able to pinpoint faults to specific words in the array and make accurate RUL predictions. For example, it predicted that a shading fault would last 22.5 days, which was very close to the actual 23 days. This study shows a scalable and effective way to do real-time tracking and preventative maintenance. This has huge implications for making solar energy systems and the smart grids they support last longer, be more reliable, and work more efficiently.},
keywords = {Hybrid AI Model, Solar Photovoltaic (PV), Smart Grid, Fault Diagnosis, Predictive Maintenance, Remaining Useful Life. (RUL)},
month = {December},
doi = {https://doi.org/10.64388/IREV9I6-1712654}
}