Stefan van Sterkenburg Stefan.vansterkenburg@han.nl Stefan.van.sterken burgr@han.nl
Contents Introduction of Lithium batteries Development of measurement equipment Electric / thermal battery model Aging of batteries Conclusions Questions
Introduction Why Lithium batteries? Price ( /kwh) Energy density (Wh/kg) Power density (kw/kg) Cycle life (# deep cycles) Lead-acid 50-300 25-30 100-250 300-1000 NiMH 200-400 50-80 250-500 500-1500 Lithium 300-1000 100-200 200-1000 1000-5000 Why not Lithium batteries? - More dangerous (inflambale, explosive) - More expensive - Need for additional electronics to guard batteries and for balancing (BMS) Extra complexity and extra costs
How does a Lithium-ion cell work? A Lithium cell exists of two half-cells separated by a liquid solvent with lithium salts or by a polymer One half cell contains the Lithium (mostly embedded in Graphite or Lithium Titanate) The other half cell contains often Lithium metal oxide salts that may capture Lithium ions such as Li x COO 2, Li x FePO 4 or Li x MnO 2
Different types of Lithium cells Chemistry Cell voltage [V] Energy density [Wh/kg] Carbon / LiCoO 2 3.7 195 (LiPo) 1000 Carbon / LiFePO 4 3.3 90-130 >3000 Cycle life [# deep cycles] Carbon / NiCoAlO 2 3.6 205 2000-3000 Carbon /Mn 2 O 4 3.8 150 >5000 Lithium Sulphur Next generation cells (still in development) Energy density = 300 500 [Wh/kg] Focus of our research: LiFePO4
Measurement setup Specifications: - Measurement system for battery pack that conisists of at most 12 cells - The cell voltage and temperature of each cell can be measured. - 14-bit resolution (0.38mV), sample rate up to1 ksample/s - Control of power supply for charging (up to 400 [A]) - Control of electronic load for discharging (up to 500 [A]) - Temperature chamber from -20 C up to 85 C - Implements a balancing system + temperature and current calibration - Software implements a SIL-test environment
Block scheme of setup
Block scheme of control unit
Userinterface software
Example of measurement script file // CC-charge tsample = 1; I0 = -100; endconditionucellmax = 3650; StartConstantCurrentMode; // CV charge Uspcell = 3650; ki = 300; endconditionucellmax = 0; endconditioncurrentmax = -1; StartConstantUcellmaxMode; // Sample time = 1 [s] // Chargecurrent = 100 [A] // Stop when Ucell = 3.65 [V] // Start current of 100 [A] // Setpoint voltage = 3.65 [mv] // Int. constant of PID-control // Disable check on cell voltage // Stop at charge current of 1 [A] // Start constant voltage mode
Electric cell modeling Battery models relate the output voltage to the current and parameters such as temperature and state of charge (soc) and state of health (soh). General equivalent network model of a battery cell consists in general of an open voltage source (open circuit voltage) and an impedance. Purpose of many models is to determine the soc and soh from measurements of I, U t and T.
Open circuit voltage Open circuit voltage is measured from a charge / discharge measurement at relatively low C-rate (voltage loss across internal impedance is compensated).
Impedance (1) The impedance takes into account the voltage losses. The following losses are present: 1. Ohmic losses (resistance of current collectors and electrolyte) 2. Activation losses (some voltage is needed to stimulate the oxidation or reduction reactions, see Butler-Volmer-equation)
Impedance (2) 3. Mass diffusion losses. The longer the current runs, the more depletion of active material occurs at the separator and voltage is lost in order to stimulate the mass diffusion
Impedance determined from EIS-plot and step-response
Complete electric battery model Ohmic and activation losses are taken into account by Rb. Diffusion losses are taken into account by (multiple) R//C networks. Parameters depend on soc, T and sign of current!
Thermal cell model We used 1-D thermal models instead of 3-D models because of: - the complexity of 3-D models - too many parameters of 3-D model are not known (material constants, exact geometry of inner part of the cell) - literature shows that temperature rise within a cell is relatively constant (the variance is less than 10%).
Thermal cell model Heat is generated by the overvoltage and reversable heat exchange: P i = I { h i + T i du ocv / dt }. where: h i = overvoltage = (U ocv U i ) Heat is transferred by heat conduction and/or convection to other cells and surroundings: Temperatue can be calculated as:
Data flow diagram of battery model
Battery pack simulator - Device that can be used to test battery management sytsems - Emulates up to 248 cell voltages and temperature sensor signals - Cell parameters such as internal resistance, capacity, initial soc and ambient can be inputet for each cell individually - Sample rate of 2 ksamples/s could be achieved, so all major electric and thermal behaviour can be measured.
Results Simulated and measured voltage and temperature of a battery-pack that consists of 12 cells. The pack is loaded by a current measured from a vehicle that drives 2 subsequent the NEDC cycles.
Temperature at pole and core of a cell - The measured rise of temperature at the end of the NEDC cannot be explained from the heat dissipation and capacity - Because of this, we started a new research to explain this phenomenom - First, we opened a cell to look inside...
Assembly of a prismatic cell
We expanded our model with: Two extra heat capacities representing the poles Extra heat sources that model the heat dissipation caused by contact resistances. Extra heat resistances between core and poles A cable model
Parameter estimation and cable model Test setup to measure the heat resistance between Poles and core Test setup to measure the electric contact resistances
Results:
Aging
Aging research method - NREL aging model is used. This is a partly emperical, partly physical aging model. It calculates the loss of capacity and resistance growth as a function of time (calender life) and number of cycles (cycle life). - Find literature with aging experiments. We found about 10 articles with usable experiments about aging, mostly related to LiFePO4 cells. - Determine the NREL model parameters from literature. - Translate the NREL to a predective aging model from which the aging can be determined as a function of the application.
NREL aging model The following equations apply for resistance growth and capacity loss: R = a 1 t 0.5 + a 2 N Q = min (Q li, Q sites ) with: Q Li = b 0 + b 1 t 0.5 + b 2 N and Q sites = c 0 + c 1 N All parameters depend on T, Voc and some on DOD. R and Q are rate variables, therefore we assume the Arrhenius dependence for the temperature dependency, the Tafel dependence for the voltage dependency and the Wöhler dependency for the DOD-dependency. It applies:
Found in literature (1) Evaluation of Battery Capacity Loss Characteristics, Kohei Nunotani et al., VPPC2011
Found in literature (2) Battery Pack Design, Validation and Assembly Guide using A123 Systems AMP20m1HD-A Nanophosphate cells
Found in literature (3) Cycling degradation of an automotive LiFePO4 lithium-ion battery, Y. Zhanga et al., Journal of Power sources, 2011.
Results: article b1 [V/K] b1 [-] b2 [K -1 ] b2 [-] b2 [-] Evaluation of Battery Capacity Loss Characteristics, Kohei Nunotani et al., VPPC2011 1.9 10-3 1.7 10-3 4.9 10-4 1.6 10-3 4.7 10-4 2.4 10-3 0.5 0.016 0.18 0.012 0.18 0.40 (E b2 =0.24[eV]) b2 =-2.8 10-6 0.072 Battery Pack Design, Validation and Assembly Guide A123 Systems Cycle-life model for graphite-lifepo4 cells, J Wang et al. 1.4 10-3 b2 =8.9 10-4 1.5 10-3 b2 =9.1 10-4 b2 =1.2 10-3 (E b2 =0.33[eV]) b2 =8.0 10-4 (if T=25 C and =0)
Conclusions: This research have been very instructive for students and co-workers that have worked on this subject. We have a good understanding of thermal and electrical behaviour of (LiFePO4) cells and battery packs. The knowledge gained during Raak-PRO EPT can be usefull for companies in the designing of battery packs. Aging of batteries is very complex. Literature research did not result in unambiguous results. More research needs to be done.