Battery modeling Presentation Battery modeling Presentation MengJie - PowerPoint PPT Presentation

Battery modeling Presentation Battery modeling Presentation MengJie Huang Cheng ‐ Ru Chang Cheng Ru Chang

A new BMS system based on cell redundancy Antonio Manenti, Andrea Abba, Alessandro Merati, Sergio M. Savaresi IEEE Transactions on Industrial Electronics

Outline Outline • Introduction Introduction • Switch network • Signal acquisition Si l i i i • Balancing Algorithm • SOC estimation • Prototyping Prototyping • Conclusion

Introduction Introduction • Each cell in battery pack has different characteristics Each cell in battery pack has different characteristics • Disconnected the cell when a single cell reaches its limit limit • Balancing in both charge and discharge • BMS should identify and bypass damaged cell permanently

Architecture Architecture • Previous work Previous work – DC ‐ DC converter, PWM • Standard Li ‐ ion cell Standard Li ion cell – 6 connected at the same time, only 1 disconnected – 4.2V of full charge voltage – 4400mAh of capacity – 10A of maximum continuous current load – 3A of maximum charge current

Switch network Switch network • Switch resistance directly impacts on the y p performance of the system • Switch have to interrupt current flow in both charge and discharge phase charge and discharge phase • Connect switch – NMOS switches NMOS switches (low on ‐ state resistance) • Bypass switch – PMOS switches • Only one cell is bypassed

Protection system Protection system • Prevent floating situation Prevent floating situation • BJT in open ‐ collector with pull ‐ up resister p p

Border cell Border cell • Bottom cell 0 Bottom cell 0 – Using both NMOS ‐ based switches – More efficient due to great conductivity More efficient due to great conductivity • Top cell N ‐ 1 – both PMOS ‐ based switches

Terminal voltage jumping Terminal voltage jumping • Due to pack reconfiguration Due to pack reconfiguration • But not a issue since – Reconfiguration needs 100us R fi ti d 100 – Standard load (electric motor) has slower dynamics (10ms) dynamics (10ms) – Load control system between BMS and load can handle and level voltage jumps handle and level voltage jumps.

Acquisition Acquisition • Worst case Worst case – 25V if all connected cell are fully charged (6 cells) – 6mV resolution for 12 ‐ bit ADC • Hardware solution – 0~5V  2.5~4.2V 0 5V  2.5 4.2V – 6mV  2.4mV • Software solution Software solution – Oversampling to reduce noise • Finally 6mV  600uV Finally, 6mV  600uV

Microcontroller Microcontroller • Microchip (dsPIC30F3014) Microchip (dsPIC30F3014) • Large pinout • 12 ‐ bit ADC 2 bi C

Balancing algorithm Balancing algorithm • ACQ ACQ – Cell voltage, pack voltage, current • Voltage mode – No current acquisition • SOC mode – OCV, impedance, neural network, fuzzy logic in t k f l i i previous work • Ԑ vm Ԑ sm : deviation Ԑ vm, Ԑ sm : deviation • m:cell index

Balancing algorithm Balancing algorithm • Charge and discharge Charge and discharge – Find min and max deviation • Selected cell is bypassed ,and previously bypassed l b d one is reconnected

SOC estimation algorithm SOC estimation algorithm • Coulomb ‐ counting – Initial value of SOC Initial value of SOC – Only on the current measurement • Model ‐ based – Need a good cell model – Need voltage and current input

Voltage mode vs SOC mode Voltage mode vs SOC mode

Refresh time calculation Refresh time calculation • Ts: SOC estimation time interval Ts: SOC estimation time interval • Tref: pack configuration refresh time interval • Too large Tref l f – Loss accuracy • Too small Tref – Increase the stress of the system and cells due to spikes (Voltage jumping)

Refresh time calculation Refresh time calculation • Q is the integrated absolute error in SOC Q is the integrated absolute error in SOC • Q is low when balancing effect is high

Refresh time calculation Refresh time calculation • ά is a coefficient related to the discharge rate ά is a coefficient related to the discharge rate T1 T2 T3

Refresh time calculation Refresh time calculation • The SOC mean value The SOC mean value • The deviation of the SOC of the m ‐ th cell with respect to average SOC results p g =

Refresh time calculation Refresh time calculation • Q is proportional to Tref Q is proportional to Tref • Increase Tref worsen the balancing effect • Increase N worsen the balancing quality h b l i li

Theoretical trend vs Measured result Theoretical trend vs Measured result • Quality factor versus number of cells(N) and Quality factor versus number of cells(N) and refresh time (Tref) • Discharged at 1C • Discharged at 1C

Efficiency Efficiency • Switches that are connected in series to the Switches that are connected in series to the current flow could overheating of devices and determine a efficiency loss determine a efficiency loss • Best case – Fully charged cell with a low current F ll h d ll ith l t

Conclusion Conclusion • Optimal balancing of the battery pack during Optimal balancing of the battery pack during operation

A supervisory control strategy for series hybrid electric vehicles with series hybrid electric vehicles with two energy storage systems Pierluigi Pisu and Giorgio Rizzoni V hi l P Vehicle Power and Propulsion, 2005 d P l i 2005 IEEE Conference

Series Hybrid Electric Vehicle Series Hybrid Electric Vehicle Fig. 1 Schematic representation of a series hybrid configuration. Fig. 2 Schematic representation of a connection of two electrical a connection of two electrical power sources configuration.

Energy Management Control Problem Energy Management Control Problem • The overall fuel consumption over a given trip: The overall fuel consumption over a given trip: • The local criteria becomes at all times:

Equivalent Fuel Consumption Minimization Strategy – Physical Viewpoint h i l i i • The main idea of the strategy is: The main idea of the strategy is: A present discharge of the RESS corresponds to a future consumption that will be necessary to future consumption that will be necessary to recharge the RESS; A present RESS charge corresponds to a future fuel A present RESS charge corresponds to a future fuel savings because this energy will be available in the future to be used at a lower cost. • The instantaneous fuel consumption:

Equivalent Fuel Consumption Minimization Strategy – Physical Viewpoint h i l i i Fig. 3 Energy path for equivalent fuel: (a) consumption during RESS discharge; (b) consumption during RESS recharge.

Mathematical Formulation: Discharging Mode for a Single Component RESS • The future cost of discharging The future cost of discharging can be represented as: •

Mathematical Formulation: Discharging Mode for a Single Component RESS d f i l • The total energy recharged in the future is:

Mathematical Formulation: Discharging Mode for a Single Component RESS d f i l • The cost of the total energy recharged in the The cost of the total energy recharged in the future is

Mathematical Formulation: Discharging Mode for a Single Component RESS d f i l • After manipulating and approximating we get After manipulating and approximating, we get the future cost of :

Mathematical Formulation: Discharging Mode for a Single Component RESS d f i l • The instantaneous fuel flow rate caused by The instantaneous fuel flow rate caused by RESS:

Mathematical Formulation: Charging Mode f for a Single Component RESS i l • The instantaneous fuel flow rate caused by The instantaneous fuel flow rate caused by RESS:

Equivalent Fuel Consumption of a Single Component RESS l

Simulation Result Simulation Result Fig.7(a) Batteries Fig.7(b) Battery pack SOE for HDUD cycle current for HDUD cycle Fig. 6 HDUD driving cycle cycle Fig.7(c) Ultracapacitors Fig. 8(d) Ultracapacitors SOE for HDUD cycle current for HDUD cycle

Conclusions Conclusions • it requires the only knowledge of the efficiency maps for the various systems in the powertrain architecture for the various systems in the powertrain architecture, and their torque and power limits; • it requires a limited number of inputs that include the SOEi of the RESSi ( i=1,2) and the torque requested at the wheels by the driver (this can be calculated from y ( f the accelerator and brake pedal position); • it is easy to implement in real ‐ time because the it i t i l t i l ti b th optimal power split can be determined by an easy and fast minimization of the function

Conclusions Conclusions • in many cases, the optimal power split can be a y cases, t e opt a po e sp t ca be pre ‐ calculated and saved in a multi ‐ dimensional map as a function of the input variables, avoiding on ‐ line minimization procedures and therefore, l d d h f reducing the computational time; • it is quite robust to estimation errors in the it i it b t t ti ti i th recharging and charging efficiencies and in the power split. power split. • It can be easily extended to any number of RESS in parallel. p