Advances in HEV Battery Management Systems

Size: px

Start display at page:

Download "Advances in HEV Battery Management Systems"

Amelia Sims
5 years ago
Views:

1 Advances in HEV Battery Management Systems Martin Klein Compact Power, Inc. (a subsidiary of LGChem) Gregory L. Plett University of Colorado at Colorado Springs

2 Outline Importance of Battery Management Systems in HEVs HEV Battery Pack Overview BMS Functions State-of-Charge Estimation BMS and State-of-Charge Estimation Battery Pack Dynamics and Cell State Estimation Sigma-Point Kalman Filter Application Results of Testing Conclusion 2

3 HEV Battery Pack Overview Environment Vehicle (Mounting) The Battery Pack is the heaviest and costliest component of an HEV Propulsion System Cooling System Cell Cell Cell Cell Cell BMS Junction Module HCU (HEV Control Unit) High Voltage DC/DC Converter Cell Physical Material... Data* 3 Cell Pack Housing Energy* ata and Energy includes Physical Input and output of cable and connector

4 Battery Management System: Functions Manages delivery/acceptance of energy from/to cells Manages the cooling system Hi-voltage relay control; pre-charge circuit control Data conditioning Diagnostics Battery-to-host vehicle communications Cell-state monitoring (V, I, T) State-of-Charge and State-of-Health estimating 4

5 accura te Importance of SOC, SOH Estimation Optimal & smooth blending of battery power with IC Engine Maximum battery life depends on SOC, Temperature control, and SOH Optimized battery size (# of cells) Cell quantity directly translates to $, weight Also affects volume, reliability HEVs typically operate in SOC range of 20% - 80%: what happens if SOC is not accurate? 5

6 If SOC is too optimistic SOC reports available charge > reality: Propulsion system may demand power > available power/energy of the cells 8 Pack discharges deeper than expected 8 Leaves battery with insufficient energy to start the vehicle BMS may detect over-current condition and abruptly reduce power (perception of poor drivability) Potential cell damage if no over-discharge protection available Designers may simply add cells for headroom to compensate 6

7 If SOC is too pessimistic SOC reports available charge < reality: Propulsion system may unnecessarily limit power demand 8 Excessive reliance on ICE -> lower fuel economy 8 Customer dissatisfaction During deceleration/regen events, BMS may allow cells to accept energy in excess of true available capacity 8 Potential cell damage BMS may detect over-current condition and abruptly reduce power (perception of poor drivability) Potential cell damage if no over-discharge protection available Designers may simply add cells for headroom to compensate 7

8 Importance of SOC, SOH Estimation One final thought: accura te Good SOC estimation is important for Pure EV, PHEV, and HEV, BUT Most critical for HEVs, which do not have a plug-in feature that can reset SOC frequently 8

9 Battery Pack Dynamics and Cell State Estimation The BMS must estimate quantities that Describe the present battery pack condition but May not be directly measured States are quantities that change quickly (e.g., SOC, cell & pack voltages, currents, temp) Parameters are quantities that change slowly (e.g., cell capacities, aging effects) HEV batteries are subject to very dynamic power cycling hence rarely in electro-chemical equilibrium Noise Factors: Hysteresis, Polarization, Time Constants 9

areas of curve difficult to estimate Tino method: SOC [V I x R]/OCV Better, but discounts effects of high R at low T or at very low

10 SOC Estimation Methods: Survey Voltage vs. Voltage/Power Curve Measure V, then report SOC per point on curve Issues: 8 misses effects of IxR losses, hysteresis; 8 Wide flat areas of curve difficult to estimate Tino method: SOC [V I x R]/OCV Better, but discounts effects of high R at low T or at very low SOC Cuolomb Counting Keeps track of energy in, out of cells OK for short periods of operation when initial conditions are known or can be frequently reset Subject to drift due to current sensors fluctuations, other losses 10

11 Kalman Filtering Kalman Filter: Excellent estimator for linear systems Extended Kalman Filter (EKF) common approach for state estimation of nonlinear systems Linearizes equations at sample points using Taylor Series expansions Must compute derivatives Sigma-Point Kalman Filter (SPKF): improves on EKF for superior state estimation of non-linear systems. Better approximation of covariances Use current for short term SOC dynamics Use voltage measurement for longer-term SOC dynamics Kalman Filter optimally combines these to give best SOC estimates 11

12 Where: SPKF: State-Space Model State-Space model of the cell s dynamics: x k+1 = f(x k,u k,w k ) (1) y k = g(x k,u k,v k ) (2) Equation (1) is the State Equation Equation (2) is the Output Equation f( ) and g( ) are functions specified by the cell model And: x k is the state vector at time index k u k is the system input vector, typically containing i k : instantaneous cell current T k : cell Temperature C k : nominal Cell Capacity and/or R k: internal cell resistance estimate w k and v k are Gaussian random process (model sensor noise) 12

13 SPKF: Cell Model x k+1 = f(x k,u k,w k ) (1) y k = g(x k,u k,v k ) (2) Per the previous slide we learned f( ) and g( ) are functions specified by the cell model x k is the state vector and u k is the system input vector Therefore a cell model must be defined! We have developed an Enhanced Self-Correcting (ESC) Cell Model that includes in the state vector: Voltage, Current, Temperature Polarization, Hysteresis SOC, Ohmic Losses 13

14 SPKF-Based SOC Estimation Test Results Proof of Concept testing was performed using actual pack 40 high-power Li-ion battery cells (3.7V/cell; 4.7Ah) Cell were characterized prior to test (parameters) Cells were exercised using Aerovironment ABC 150 high-power cycler 1C Constant-current charging Dynamic drive cycle Rest periods Raw input and output data was collected throughout the test, then post-processed per several estimation methods Two Coulomb counting methods SPKF-Based (BMS) C-code PC-based Tino 14

15 SPKF-Based SOC Estimation Test Results Results: Tino performed poorly during transients, OK at steady-state Coulomb-counting showed divergence SPKF method did very well during transients; consistently converged At room temp, SPKF s RMS estimation error for SOC approx. 2%. 15

16 Additional Work Initial Proof-of-Concept testing very favorable, but limited in range of testing Further algorithm testing pushing the bounds of operation (V, I, T, charge, discharge rates, deliberately poor SOC initialization) also showed excellent convergence 1 CPI is establishing hardware-in-the-loop validation models and systems to evaluate actual BMS, cells and the SOC algorithm in real time 1 Paper to be published this year. 16

17 Conclusion Accurate real-time battery management is critical in Optimizing battery size, weight, cost, and reliability Providing acceptable, transparent vehicle performance State-of-Charge estimation is the most critical BMS function Many SOC algorithms have been considered, but few can handle the wide dynamic range of a non-plug-in HEV An SOC algorithm based on Sigma-Point Kalman Filtering has been developed which demonstrates superior estimation under dynamic conditions Further evaluation is in progress to further evaluate this algorithm using HIL techniques 17

KALMAN FILTER APPLICATIONS

ECE555: Applied Kalman Filtering 1 1 KALMAN FILTER APPLICATIONS 1.1: Examples of Kalman filters To wrap up the course, we look at several of the applications introduced in notes chapter 1, but in more