Tanima Bal is a Ph.D. student of Electrical and Computer Engineering department, where she has been pursuing her doctoral studies since January 2025 under the guidance of Dr. Philip Pong. Her research interests include power-system monitoring, machine-learning applications in power systems, and soft-computing techniques for developing advanced control algorithms. Her work focuses particularly on fault detection, wide-area monitoring and control, and energy-management strategies for net-zero-energy buildings.
Her current research focuses on developing intelligent, multi-sensor-based arc-fault detection methods for U.S. naval shipboard power systems, with the goal of improving the accuracy, reliability, and speed of fault detection while reducing false alarms and unnecessary tripping. Through this work, she aims to contribute to the development of more reliable, intelligent, and resilient power systems.
Outside of academics, she enjoys exploring different cuisines and learning about diverse food cultures. She is also passionate about baking and decorating cakes, which allows her to express her creativity and explore new ideas beyond her academic work.
What would you say that could be the next big thing in your area of research?
Based on my literature review, many arc-fault detection systems developed for naval shipboard power systems still rely on fixed-threshold algorithms. Although these methods are relatively simple to implement, they may not adapt effectively to changing operating conditions and can result in false alarms or unnecessary tripping, potentially causing serious operational and financial consequences.
I believe the next major advancement will be the integration of artificial intelligence and machine learning into real-time arc-fault detection and protection systems. Data-driven methods can analyze complex patterns from multiple sensors, adapt to different operating conditions, and distinguish actual arc faults from normal electrical transients more accurately.
Although significant progress has already been made in AI-based detection, implementing these methods in safety-critical, real-world naval applications remains challenging. As issues related to data availability, model reliability, computational requirements, and explainability are addressed, I expect intelligent and adaptive protection systems to become much more practical and widely adopted in the coming years.
Some say that AI should not be trusted in critical areas such as power systems, where a single glitch could cause catastrophic damage. What is your opinion?
I believe the concern is valid, but AI should not be excluded from critical power-system applications. Modern power systems generate enormous amounts of data from renewable energy resources, microgrids, intelligent electronic devices, and energy markets. AI can use these data to improve forecasting, predictive maintenance, fault detection, system monitoring, and cybersecurity. However, AI also introduces risks. Its performance depends heavily on the quality and availability of training data, which can be limited for rare power-system faults. An AI model may also produce unreliable decisions under unfamiliar operating conditions or cyber manipulation, while some complex models lack transparency and explainability.
Therefore, AI should not be trusted blindly or immediately given complete control over critical operations. It should initially serve as a decision-support or supervisory tool, supported by conventional protection methods, physics-based constraints, human oversight, continuous monitoring, and fail-safe mechanisms. With extensive validation and strong cybersecurity, AI can complement the protection systems and human expertise.
As one of ECE students who have taken DS 675 Machine Learning, how did you like the course? How do you feel it can help with your PhD research?
DS 675 provided me with a strong foundation in artificial intelligence and machine learning. The course helped me understand the complete machine learning workflow, including data cleaning, preprocessing, exploration analysis, feature selection, model development, and performance evaluation. I also learned how to select appropriate algorithms based on the characteristics and complexity of a particular problem. Through the course project, I gained practical experience with linear and nonlinear support vector machines, perceptrons, and other classification methods. This helped me understand how different algorithms identify complex relationships and nonlinear decision boundaries within data, which is particularly relevant because power systems are highly dynamic and nonlinear.
These skills directly support my Ph.D. research on arc fault detection in naval shipboard power systems. I plan to analyze electrical and multi-sensor data to develop machine learning models that can distinguish arc faults from normal operating events. The knowledge gained from DS 675 will help me build, evaluate, and eventually implement a more accurate and reliable real-time detection system.
