SAE Technical paper 2010 Conference

SOH Recognition of Li-ion Aviation Batteries
Via Passive Diagnostic Device
John E. James and Dr. Boris Tsenter
GEM Power LLC
ABSTRACT
Aviation battery maintenance is trending toward on-condition maintenance. Nickel-Cadmium (NiCd), Valve Regulated Lead-Acid (VRLA), or Lithium-Ion (Li-ion) batteries are used to start engines, provide emergency back-up power, and assure ground power capability for maintenance and pre-flight checkout. As these functions are mission essential, recognition of battery state of health (SOH) is critical. SOH includes information regarding battery energy, power and residual cycle life along with monitoring overall battery safety. This paper describes an SOH recognition technique for on-board Li-Ion aviation batteries and discusses a passive diagnostic device (PDD), that analyzes input data derived from normal system parameters such as battery current, voltage and ambient temperature. These parameters are monitored in a totally passive mode eliminating the need for active signals to the battery. Active signals are restricted or even prohibited in order to avoid any interference with the vehicle electrical system. A procedure for sampling and analyzing transient and stationary battery voltage and current and establishing a matrix of battery parameters (MoP) is discussed. The basis for SOH recognition is based on a matrix of parameters containing values for ohm and chemical resistance, instantaneous and dynamic open circuit voltage, and Tafel coefficients of electrode reactions. Because Li-ion battery charging is provided under precise controlled conditions, data are sampled and processed both in charging and discharging modes. Advantages discussed are the capability of the PDD to provide early signs of impending battery failure or simply the inability of the battery to carry out a necessary function, or the need for off-line battery maintenance.
INTRODUCTION
The PDD for a typical Li-ion battery is part of a developing battery management system, which the block diagram in Figure 1 illustrates.
Charging subsystem Prognostic subsystem Discharging subsystem Cell equalization subsystem Individual cell control Diagnostic subsystem Safety control Temperature control
Figure 1 Block diagram battery management system
The PDD can be considered a component of diagnostic and safety subsystems, without the generation of any active signal. The creation of a PDD for a typical Li-ion battery provides a challenge. A lack of access to individual cells requires a very accurate recording and sophisticated interpretation of the variation of terminal voltage and battery current. However, the requirement for precision charging voltage and current for Li-ion chemistries presents the opportunity to sample and process information related to battery condition and to compare it with information from a MoP derived during discharging. Seven 18650 Li metal-oxide cells connected in series has been chosen as the basic initial design for study. Changes in Li cell design and chemistry will probably require some adjustments in the PDD algorithm but the basic elements of the PDD like
power and state of charge (SOC) recognition along with cycle life prediction and safety features should remain constant between chemistries.
EQUIVALENT SCHEMATIC AND MEASUREMENT PROCEDURE
A Li-ion battery equivalent schematic (Figure 2) is the core element of a battery management system in the current development stage. A A BBBRchACdlARchCCcdlCOCV1+_RcohmIhcCICCIchAICAOCV1-OCV battery cellRcohm-cell ohm resistanceRcL-cell inductance resistanceRchAanode chemical resistanceRchCcathode chemical resistanceCdlA-double layercapacitor of anodeCdlC-double layercapacitor of catodeIin-input charging currentIout-output charging currentIchA,IchC-transient currents through catode and anode chem. resistanceICA,ICC-transient currentsthrough catode and anode double layer capacitorn-cell numberRohm=nRcohm -battery ohm resistanceRch=n(RchA+RchC)-battery chemical resistanceCdl=n(CdlA+CdlC) -battery EDLCIdl-curent through battery double layer capacitorIch-curent through battery chemical resistanceIoutRcLIinRchCdlnOCV1+RohmIdlIchIoutRLIin
Figure 2 Equivalent schematic of individual cell (A) and battery (B)
As part of the creation of a battery first principles model, the equivalent schematic is the initial element of the first principles model development [1]. Analysis of the equivalent schematic leads to a conclusion regarding a measurement procedure, time intervals and the desired reliability of measured parameters. Each electrode is modeled as a parallel circuit of electrical double layer capacitor (EDLC) and chemical resistance connected in series to ohmic resistance. Current across the EDLC is transient and associated with a change in electrical charge on the solid layer of the electrode and on the ionic layer of the electrode. Chemical resistance relates to the energy dissipation during the process of charge transfer in electrochemical reactions. This kind of resistance may have an activation or diffusion nature associated with ion transfer or phase transition or phase transformation kinetics [2]. This paper recognizes both linear and Tafel electrode kinetics, which can be extrapolated to another mechanism of charge transfer. Linear electrode kinetics result in an exponential current drop across EDLC per expression:
ICdl=Iexp(-t/tc) (1)
The transient time tc, which is product of chemical resistance and EDLC (RchemCdl) is defined as a time constant. This is a time period, when current across an EDLC drops 2.7 (e) times. For a Li-metaloxide 18650 battery tc is 40-50 ms. Transient current across EDLC causes unstationary current across the chemical resistance (Ichem) as a difference between input current I and the current across ICdl :
Ichem= I(1- exp(-t/( tc)) (2)
Tafel’s electrochemical kinetic results in a hyperbolic current drop across of the EDLC instead of an exponential one.
ICdl=IbCdl/(t-t0) (3)
Here b is a Tafel pre-logarithmic coefficient and t0 is the transition time from linear to exponential kinetics. Reading parameters from individual cell to full battery level is calculated with a proper time interval. If, for example the objective is to capture 95% current across the EDLC, the registration time period should be no more than t≤0.05 tc. It is around 2.5ms in case of 8650 battery. Under a requirement to capture no less than 95% current across chemical resistance, time interval has to be no less than t≥3tc, which is 120-150ms for 18650 battery. As soon as most of the transient current passes across EDLC, battery EDLC is calculated as a summary of individual cells. The same is true for determining battery chemical resistance. Capturing it at time interval t>> tc allows the determination of battery chemical resistance as the summary of one cell. Constant chemical resistance is assumed for linear electrode kinetics. For non linear kinetics, dynamic chemical resistance is introduced as the ratio of delta chemical polarization and delta current. Determination of battery equivalent schematic parameters such as ohmic and chemical resistance along with EDLC results from sampling battery transient current and voltage at proper time intervals.
For example, under a discharging load change, the relationship between transient voltage and current is:
V(2) - V(1)= (I2-I1)Rohm +I2Rchem((1-exp(-t/( tc)) (4)
Here I1, I2 and V2, V1 are discharging currents and voltages under low (I1,V1) and high (I2,V2) loads. Along with above mentioned transient measurement, some parameters are sampled in the stationary mode. This mode can be defined as a time interval much longer than the time constant. Internal resistance for example is obtained from charging data based on an expression equally true for both constant current and constant voltage modes for linear electrode kinetics.
Vch=OCV +IRohm +IRchem (5)
The summary of Rohm and Rchem is defined as DC battery impedance. If the OCV is obtained through direct integration of current with respect to time and Rohm is known through transient measurement of charge current, equation (5) results in a measurement of chemical resistance. Under high current, where chemical resistance is a function of current, the equation for charging voltage balance is:
Vch=OCV +IRohm +Vchem (6)
Here Vchem is chemical polarization. In the case of Tafel electrode kinetics, (6) can be written as:
Vchem =a +blg(I) (7)
Tafel coefficients a and b can be considered effective for the porous three dimensional structure of battery electrodes. Tafel coefficients can be obtained under given OCV and ohmic resistance according to expression 7.
STATE OF CHARGE RECOGNITION
There is a strong relationship between reversible OCV and SOC for typical Li-ion batteries. While obtaining reversible OCV under unstationary conditions of battery operation is impossible, dynamic OCV is used. Dynamic OCV can be detected under any sequence for battery SOC and is sampled under discharge as a summary of battery discharging voltage and irreversible voltage losses attributed to ohmic and chemical polarization. Conceptually, dynamic OCV also works in charging mode, where dynamic OCV is found as difference between charging voltage and lost irreversible voltages. Figure 3 illustrates the relationship between dynamic and reversible OCV obtaining during the discharge of 18650 Li –Co0.2Ni0.8O2 cells. Using dynamic OCV in place of reversible OCV, results in an error of less than 7% in the determination of battery SOC.
Figure 3. Relationship between reversible and dynamic OCV and DoD for Li-metal-oxide battery cell 3.43.453.53.553.63.653.73.753.83.853.93.9544.054.14.154.200.050.10.150.20.250.30.350.40.450.50.550.60.650.70.750.80.850.90.951OV,VDoDReversible OCVDynamic OCV
SUFFICIENT BATTERY POWER RECOGNITION
This feature is useful when maximum required power (Pmax) and maximum delivered capacity are known. Information concerning Pmax allows the calculation of maximum discharging current Imax. The expression for delivered maximum power is:
Pmax = Imax (OCVmin-IRohm-ΔVch) (8)
Here, OCVmin is battery OCV corresponding with maximum delivered capacity. By using ΔVch from equation (7) under current Imax it is possible to find Rohm, which provides required maximum power. Under a current 3A and SOC 50% required maximum power is 75W. For example a 7x18650 battery is consistent with Rohm≤0.4ohm.
BATTERY FAILURE PREDICTION FEATURES
For Li-ion batteries, especially high power batteries like A123 or VL4V, a trend in the value of EDLC coincides with battery cycle life. A drop in EDLC is likely caused by an aging and shrinking electrode surface, which is initially large for high power Li-ion batteries. EDLC is sampled as the ratio of the product of current and time to voltage difference for the same time interval [3]. While the time constant for a typical Li-ion chemistry is about 10ms, sampling EDLC within 2ms time period creates an error of no more than 20%. The chart in Fig.4 is an example of the MoP’s predictive capability, showing the determination of the residual life cycles for the battery. Based on EDLC information, a value of 12F indicates the A123 battery will deliver around 2000 cycles and the VL4V battery 3000 cycles prior to failure. As soon as the EDLC reaches 2F for the A123 and 5F for the VL4V battery, both batteries will be on the brink of failure. EDLC for a low power, metal-oxide 18650 Li cell does not change over cycling as significantly as high power cells (Fig.5). It can be explained with the assumption of less surface energy and therefore a more stable electrode surface in the low power cell. A drop in delivered capacity occurs due to swelling of the positive electrode active material [1,4], which doesn’t directly
affect the MoP. A combination of monitoring the trend of EDLC and ohmic resistance can predict the end of cycle life for 18650 metal-oxide low power cell.
0100020003000400027121722Residual cyclesEDLC,F
Fig.4 Residual cycles vs. EDLC over cycling VL4V(Δ) Li-metaloxide and A123 (0) Li-nanophosphate batteries
00.010.020.030.040.050.060.070.000.501.001.502.002.500100200300400500600700800Rohm/cell,Rchem/cell, ohmAh‐out ,
EDLC/cell,FCyclesAh OutEDLC/cellRohm/cellRchem/cell
Fig.5 Ah-out, Rohm/cell, Rchem/cell, EDLC/cell vs. cycles for 18650 Li-metaloxide battery.
SAFETY CONTROL
Safety control at the battery level with the PDD is provided by monitoring potential shunt and conditions resulting in Li electroplating. A voltage drop (dV/dt<0) under constant current charging mode and current drop (dI/dt>0) under constant voltage charging mode can be used as indicators of potential shorting in a battery cell [5]. An abnormal rise in battery chemical resistance corrected to temperature is used as an indicator of potential lithium metal electroplating.
CONCLUSION
Analysis of transient voltage and current, from the battery during charging and discharging, allows for the recognition of Li-ion battery power, state of charge, potential internal shunt and electroplating along with prediction of battery failure without any interaction with airplane electrical system.
ACKNOWLEDGMENTS
Sean Field – NAVAIR Mark Hurley – NAVAIR Nathan Kumbar - NAVAIR
REFERENCES:
1. J.E. James and B.Y. Tsenter. “State of Health Recognition for Aircraft Batteries. Dynamic Equivalent Schematic and First Principles Model Considerations”. SAE Technical Paper 2008-01-2933.
2. J. Ma, C. Wang, S. Wroblewski. “Kinetic Characteristics of Mixed Conductive Electrodes for Lithium -ion Batteries”. Journal of Power Sources, v. 164, 2, 2007, pp. 849-856
3. State of Health Recognition of Secondary Batteries. US patent 7,605,591.
4. J. Vetter, P. Novák, M.R. Wagner, C. Veit, K.-C. Möller, J.O. Besenhard, M. Winter, M.Wohlfahrt- Mehrens, C. Vogler, A. Hammouche. “Ageing Mechanisms in Lithium-Ion Batteries”. Journal of Power Sources, v.147, 1-2, 9, 2005, pp. 269-281
5. Shunt Recognition in Lithium Batteries. US Patent 5,729,116.