In the world of maritime and offshore operations, power transformers are the unsung heroes, ensuring that electricity flows smoothly to keep vessels and platforms running. But like any critical equipment, they need regular health checks. A recent study published in the journal ‘Technologies’ (formerly known as ‘Technologies’) offers a fresh approach to predicting the health of these transformers, which could have significant implications for the maritime industry.
Dr. Rania A. Ibrahim, from the Electrical and Control Engineering Department at the Arab Academy for Science, Technology and Maritime Transport in Egypt, led the research. She and her team have developed a new method for Dissolved Gas Analysis (DGA), a technique used to monitor the health of oil-immersed transformers. The key innovation here is the use of SHapley Additive exPlanations (SHAP), a tool that not only makes the predictions more interpretable but also helps in selecting the most relevant features for analysis.
So, what does this mean for maritime professionals? Well, transformers are the backbone of power distribution in ships and offshore platforms. Keeping them in good health is crucial to avoid costly downtime and maintenance. The current methods of DGA often focus on classification accuracy or single-model regression, which can be complex and lack transparency. Dr. Ibrahim’s approach, on the other hand, offers a more balanced method that combines interpretability and reduced complexity.
The team used a publicly available dataset to benchmark models from four different algorithmic families: linear, kernel/tree-based, boosting, and hybrid ensembles. They found that their proposed LightGBM–CatBoost hybrid ensemble, enhanced by SHAP-guided feature pruning, achieved superior predictive accuracy while reducing model complexity and improving transparency.
Dr. Ibrahim explained, “Unlike prior works carried out using the same dataset, our framework not only provides a balanced approach that combines interpretability and reduced complexity, but also surpasses previous regression-based approaches.” The results showed a reduction in Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) by 4.93% and 2.31%, respectively, and an enhancement in Health Index (HI) predictive accuracy by 1.45%.
For the maritime industry, this means more reliable predictions about the health of transformers, leading to better maintenance planning and reduced downtime. It also offers opportunities for commercial impact, as shipping companies and offshore operators could potentially save on maintenance costs and improve the efficiency of their power systems.
In the words of Dr. Ibrahim, “This approach could be a game-changer for the maritime industry, providing a more accurate and transparent way to monitor the health of power transformers.” As the industry continues to evolve, such advancements in predictive maintenance will be crucial in ensuring the smooth and efficient operation of maritime and offshore assets.