Battery health diagnostics: Bridging the gap between academia and industry

Abstract

Diagnostics of battery health, which encompass evaluation metrics such as state of health, remaining useful lifetime, and end of life, are critical across various applications, from electric vehicles to emergency backup systems and grid-scale energy storage. Diagnostic evaluations not only inform about the state of the battery system but also help minimize downtime, leading to reduced maintenance costs and fewer safety hazards. Researchers have made significant advancements using lab data and sophisticated algorithms. Nonetheless, bridging the gap between academic findings and their industrial application remains a significant hurdle. Herein, we initially highlight the importance of diverse data sources for achieving the prediction task. We then discuss academic breakthroughs, separating them into categories like mechanistic models, data-driven machine learning, and multi-model fusion techniques. Inspired by these progressions, several studies focus on the real-world battery diagnostics using field data, which are subsequently analyzed and discussed. We emphasize the challenges associated with translating these lab-focused models into dependable, field-applicable predictions. Finally, we investigate the frontier of battery health diagnostics, shining a light on innovative methodologies designed for the ever-changing energy sector. It's crucial to harmonize tangible, real-world data with emerging technology, such as cloud-based big data, physics-integrated deep learning, immediate model verification, and continuous lifelong machine learning. Bridging the gap between laboratory research and field application is essential for genuine technological progress, ensuring that battery systems are effortlessly integrated into all-encompassing energy solutions.