SENSOR: embedded fiber-optic sensing systems for improved battery management

Abstract

Hybrid and electric vehicle battery systems today use conservative design approaches and multiple redundant layers of safety to compensate for the lack of real-time, directly sensed cell internal conditions during operation. The harsh electrochemical environment and other demanding constraints prevent the use of conventional sensors, so the true conditions and state inside the batteries through external readings like current and voltage. The uncertainty of these inferences means that batteries must be used conservatively to guarantee reliability and safety, which prevents utilizing batteries to their true capabilities and is a key contributor to their high cost. Under the ARPA-E AMPED program for advanced battery management systems, PARC and LG Chem Power (LGCPI) are developing SENSOR (Smart Embedded Network of Sensors with an Optical Readout), an optically based smart monitoring system prototype targeting batteries for XEVs. The system will use fiber optic sensors embedded inside Li-ion battery cells to measure parameters indicative of cell state in conjunction with PARC’s low-cost, compact wavelength-shift detection technology and intelligent algorithms to enable effective real-time performance management and optimized battery design. To achieve this, our team is leveraging PARCs demonstrated compact wavelength shift detector (WSD) for optical demodulation and particle-filtering based battery prognostic algorithms. The WSD has shown promising results in laboratory-level demonstrations for dynamic strain sensing using fiber Bragg gratings [Bellmann 10]. It combines a laterally varying optical transmission filter with a position-sensitive photo detector that can resolve wavelength shifts as small as 50fm at 100Hz. The concept can be extended to measure a variety of relevant internal cell-state parameters using multiplexed, embedded optical fiber sensors with sufficient accuracy for BMS. These sensors will be read out by an advanced version of our compact WSD for multiplexed fiber optic sensors that can separately quantify these parameters. Initial versions of our teams particle-filtering algorithms [Saha 12] have used external performance parameters (voltage, current, and temperature) to make battery end-of-discharge predictions with 5% accuracy. Advanced embedded versions of those algorithms will use the internal cell-state measurements for even higher accuracy state estimates that can enable significant reductions attributable to conservative battery oversizing and over-engineering. This project will culminate in demonstration of a prototype commercial EV-grade module with embedded optical sensors, advanced optical readout unit, and accurate state-estimation algorithms. The prototype will be subjected to validation tests consistent with industry standards for certifying/adopting new cell and module designs by LGCPI. This talk will give an overview of the project, the underlying enabling technologies, and then cover some promising initial experimental results from the project, including internal cell signal data and state estimation using fiber optic sensors embedded in Li-ion pouch cells over charge-discharge cycles. References: [Bellmann 10] Bellmann K., Kiesel P., and Johnson, N., Compact and fast read-out for wavelength-encoded biosensors, Proceedings of the SPIE, v. 7593-32, 2010. [Saha 12] Saha B., Cuong C., and Goebel K., Optimizing battery life for Electric UAVs using a Bayesian framework, IEEE Aerospace Conference 2012.