AI-Based Condition Monitoring and Lifetime Estimation of Various Electrical Assets

The determination of whether low, medium, and high voltage electrical assets suffer from high levels of partial discharge (PD) has received great attention to ascertain the safety of neighboring utilities and achieve economic satisfaction. Such assets include power transformers, induction motors, and power cables. Despite the high robustness of these capital assets, which seldom leads to damage, these devices can still undergo significant degradation, particularly due to PD events. As a result, monitoring such systems for PD has become critically important. To this end, this research focuses on the condition monitoring and lifetime estimation of critical electrical assets, including power cables, induction motors, power transformers, and Li-ion batteries through the integration of Artificial Intelligence (AI) and Finite Element Analysis (FEA). Initially, FEA is employed to accurately model faults within these electrical components, providing a detailed and nuanced understanding of potential failure mechanisms. Subsequently, AI methodologies are applied to effectively detect and classify these faults, through accelerated aging tests conducted in the laboratory and historical data provided by Shell-Qatar (industry partner). Moreover, a key innovation of this research is the application of physics-informed neural networks to estimate the lifetime of power transformers, cables, and induction motors, a method that integrates domain knowledge of physical laws with the predictive capabilities of neural networks, thus offering a more accurate and reliable estimation of asset longevity. An AI-based lifetime predictive model was also developed as means to track the aging behaviour of industrial induction motors situated in Shell-Qatar's facilities. Such a model was able to consider the complex intricacies in the data observed from the stochastic nature of the seasonality effects and sudden load changes. To this end, this comprehensive study not only aims to advance the field of smart predictive maintenance but also seeks to significantly improve the reliability and efficiency of managing electrical assets. By bridging the gap between traditional condition monitoring techniques and cutting-edge AI methodologies, this research has the potential to set a new standard in the industry, ensuring safer, more efficient, and cost-effective operation of critical electrical infrastructure.