Improving Battery Safety by Advanced BMS Diagnostics and Model-based Hardware-in-the-Loop Testing

Transcription

1 Battery Ageing Battery Models Battery Diagnostics Battery Pack Design Electromobility Stationary Energy Storage Energy System Analysis Improving Battery Safety by Advanced BMS Diagnostics and Model-based Hardware-in-the-Loop Testing Hendrik Zappen Chair for Electrochemical Energy Conversion and 1 Storage Systems

2 Overall Battery System Safety Depends on Different Aspects Battery Cell Safety Electrical Safety Functional Safety ISO Source: newscience.ul.com Source: Wikimedia Commons 2

3 Overall Battery System Safety Depends on Different Aspects Battery Cell Safety Electrical Safety Functional Safety Today s Talk ISO Source: newscience.ul.com Source: Wikimedia Commons 3

4 Battery Packs A Short Introduction Battery packs are sophisticated systems Consist of various mandatory components: Cells & cell modules Cell- & module interconnectors BMS electronic modules (V, I, T, etc.) Contactors (electromechanical or solid state) Precharge resistor Main & other system protecting fuses Wire harness Popular opinion: If something can go wrong, it sooner or later will. Identification of possible fault spots needed! 4

5 Battery Packs Components Cells & cell modules Cell- & module interconnectors BMS electronic modules (V, I, T, etc.) Contactors (electromechanical or solid state) Precharge resistor Main & other system protecting fuses Wire harness 5

6 Battery Packs Thermal Fault Spots Battery packs components have a dynamic thermal behaviour Potential fault spots are numerous Cells themselves (inner short etc.) Interconnectors (faulty tightening, welding, etc.) Cell voltage cables (torn, lose joints, etc.) Contactors (worn blades, sticky contacts, etc.) Many faults have thermal symptoms Very few temperature sensors available (ca. 2 or less per 6 cells) Task: Find a way to gain temperature info without additional dedicated Temperature sensors! 6

7 Battery Packs Maybe Utilize the Wiring Harness? Modular BMS Layout Various layouts for battery packs Modular BMS -> Short cabling. Direct Modular BMS -> Very short / no cabling. Central BMS -> Long cabling. Direct Modular BMS Layout BMS Electric Vehicle Voltage Sensing Cable Length Min. / cm Max. / cm BMS Topology Vehicle A Modular Vehicle B 3 21 Direct Modular Vehicle C Central Vehicle D 3 5 Direct Modular Vehicle E Central Central BMS Layout 7

8 Wire Temperature Acquisition The Effect Test setup includes PVC insulated 2-wire cable assembly Heating defined cable s portion Local deviation of permittivity towards larger value ( becomes more capacitive, because C ~ e r ) Wire s impedance locally decreases (Z L LL CC ) Local reflection and transmission factors develop at temperature event s entry and exit point 8

9 Wire Temperature Acquisition Signal Propagation Temperature event along cable leads to reflections of travelling signals Voltage step is driven into utilized cable, then travels down the line At impedance disruption Z L -> Z F a signal portion is reflected back Due to negative reflection factor, voltage sum decreases Transmitted signal travels further, gets reflected at temperature event s exit (Z F -> Z L ) and travels to generator Due to positive reflection factor, voltage sum increases At cable s end, remaining signal is reflected back Additive signal due to cable s open end Cable s dimension properties easy to determine t Event,Start t Event,End tcable,end 9

10 Wire Temperature Acquisition Calculation If temperature rises, local capacity increases, yielding in a lowered voltage amplitude 3.5 Capacity vs. Temperature Knowing group delay, distance of event can be derived Temperature can be calculated from area between two curves using known line impedance: C = 2 Z L 0 V NNNN t dd Plot capacity vs. temperature shows reasonable correlation Capacity / pf Temperature / C 10

11 Introducing Temperature and Position Acquisition Device GUI shows interesting part of graph, including: Reference curve (red), acquired curve (blue) Determined info on fault length and coordinates Derived temperature according to integral calculations <150 prototype device could be potentially integrated into BMS 11

12 Temperature Acquisition Along PVC Cable Test Results Heating induced via special heatwire, wound equally distanced along 20cm, starting from 30cm Test procedure included 9 temperatures from 50 C in 10K incr. Each 5 successive acquisitions Good temperature accuracy Best localisation accuracy >80 C Position deviations result from: Low T: small signal variation High T: fuzzy reflection at true eventstart 12

13 Battery Packs Components Cells & cell modules Cell- & module interconnectors BMS electronic modules (V, I, T, etc.) Contactors (electromechanical or solid state) Precharge resistor Main & other system protecting fuses Wire harness 13

14 BMS: Reaction to Faults and Plausability Checks - Example Alert-Flag: Voltage Cell 7 < 3.0 V (Undervoltage limit 1) Possibility 1: Voltage of Cell 7 is really too low Low SOC Cell defect (Internal short circuit, Damaged Cell) Polarisation due to Load Curent Cell balancing not working properly Possibility 2: False Alarm Malfunction of BMS Voltage Sense HW Defect of Sense Wire Software Error Measurement Error due to Cell balancing enabled Cell Voltage / mv Time / s 14

15 BMS: Reaction to Faults and Plausability Checks - Example Alert-Flag: Voltage Cell 7 < 3.0 V (Undervoltage limit 1) Possibility 1: Voltage of Cell 7 is really too low Low SOC Cell defect (Internal short circuit, Damaged Cell) Polarisation due to Load Curent Cell balancing not working properly Possibility 2: False Alarm Malfunction of BMS Voltage Sense HW Defect of Sense Wire Software Error Measurement Error due to Cell balancing enabled Cell Voltage / mv Time / s 15

16 What might an Internal Short Circuit look like in the Measurement Data? Event 1 Event 2 (different cell of same type, different test profile) Current Temperature Source: PhD thesis S. Käbitz, ISEA RWTH Aachen,

17 How to Detect Cell Faults Proposed Fault Detection methods in literature Active Signal Excitation or Additional Sensors Passive ( Software-only ) Non-electric Electric Time Domain Frequency Domain Active Ultrasound Acoustic Gas-detection Pressure (Spatially Resolved) Temperature Meas. EIS ( Electrochemical Impedance Spectroscopy ) Pulse Excitation (Time Domain) Correlation Based Methods Model Based Methods Passive EIS 17

18 How to Detect Cell Faults Proposed Fault Detection methods in literature Active Signal Excitation or Additional Sensors Passive ( Software-only ) Non-electric Electric Time Domain Frequency Domain Active Ultrasound Acoustic Gas-detection Pressure (Spatially Resolved) Temperature Meas. EIS ( Electrochemical Impedance Spectroscopy ) Pulse Excitation (Time Domain) Correlation Based Methods Model Based Methods Passive EIS 18

19 What can the Battery Cell s Open Circuit Voltage (OCV) Tell us? OCV: State-of-Charge dependent battery voltage under no-load conditions and after decay of all overvoltage effects Changes in OCV during no-load conditions in a battery pack: Quiescent Current Flow Current Draw of BMS components Parasitic Currents (Moisture, Isolation Defects) Cells: (Apparent) Self-Discharge Cells: High Resistance Internal Short Circuits? Cell Voltage Measurement Accuracy of Common Cell Monitoring IC s: About 1 mv (TI bq76pl536a, Linear Technology LTC6804) OCV change of 1 mv: Equals roughly 100 mah charge turnover for a 50 Ah cell For 1mA current 50 h detection time 3600 mv 3599 mv 19

20 How to Detect Divergent Cell Behaviour Faster? Step 1: Enhancing BMS measurement resolution and noise suppression by oversampling and averaging. Voltage / mv Time / h Voltage / mv Time / h 20

21 How to Detect Divergent Cell Behaviour Faster? Step 1: Enhancing BMS measurement resolution and noise suppression by oversampling and averaging. Step 2: Trend detection algorithm rejects influence of common-mode discharge (mainly due to quiescent current of connected BMS) to individual cell voltages by subtraction of a moving-average reference signal. Voltage / mv Time / h 21

22 How to Detect Divergent Cell Behaviour Faster? Step 1: Enhancing BMS measurement resolution and noise suppression by oversampling and averaging. Step 2: Trend detection algorithm rejects influence of common-mode discharge (mainly due to quiescent current of connected BMS) to individual cell voltages by subtraction of a moving-average reference signal. Step 3: Use of windowing and linear regression for slope estimation of trend signals rejects remaining residual noise. d/dt Voltage / mv Voltage / mv Time / h Time / h Voltage Change / mv/h Time / h 22

23 How to Detect Divergent Cell Behaviour Faster? Step 1: Enhancing BMS measurement resolution and noise suppression by oversampling and averaging. Step 2: Trend detection algorithm rejects influence of common-mode discharge (mainly due to quiescent current of connected BMS) to individual cell voltages by subtraction of a moving-average reference signal. Step 3: Use of windowing and linear regression for slope estimation of trend signals rejects remaining residual noise. Voltage Change / mv/h d/dt Voltage / mv Voltage / mv Time / h Time / h Time / h Cell 2 voltage decreases faster than average of all cells by about 50 µv/h 23

24 Validation Implementation on typical BMS Hardware Emulation with (High-value) Resistor Discharge on one cell Sensitity of method tested with different cases After Charging After Discharging After 48h Rest 12 hour test window (Scenario: Overnight parking stop ) In case of previous Charge/Discharge: No OCV conditions, but divergent behavior also detectable 8 Cell Module, 50 Ah Mitsubishi imiev Detection Treshhold: Equivalent DCH Rate After Charging C/10000 After Discharging C/12000 After 48h rest C/

25 Battery Packs The BMS Cells & cell modules Cell- & module interconnectors BMS electronic modules (V, I, T, etc.) Contactors (electromechanical or solid state) Precharge resistor Main & other system protecting fuses Wire harness 25

26 Hardware-in-the-Loop Testing of BMS Emulation during HIL testing 26

27 Requirements on Battery Cell Models for HIL Testing Electrical and thermal behavior emulated as closely as possible Simulation on individual cell level Real-time capability Impedance based battery models particularly suitable Taking into account statistical differences in cell behaviour especially useful regarding the modelling of aged battery packs Modelling of inhomogenous heat generation Estimation of cell balancing usage Source: Baumhoefer et. al, Journal of Power Sources,

28 Battery Model Development (Electrical & Thermal) No. of cells Capacity / Ah Scale Factor for R i 28

29 Battery Model Development (Electrical & Thermal) No. of cells Real-time capable thermal-electrical battery simulation framework, developed at ISEA, is available for free under open source license: Capacity / Ah Scale Factor for R i 29

30 Conclusion Many different types of faults in different components can possibly happen inside a battery pack. Advanced diagnostic methods can lower the risk of catastrophic failure by early detection of error patterns. High precision battery models are an important tool for BMS validation. Inclusion of statistical distributions helps to understand the effect of battery aging to BMS functions. 30

31 Thank you for your attention Contact We thank Hendrik Zappen Phone: Chair for Electrochemical Energy Conversion and Storage Systems Univ.-Prof. Dr. rer. nat. Dirk Uwe Sauer RWTH Aachen University Jaegerstrasse 17/ Aachen GERMANY 31