A Study on the dynamic SOC compensation of an Ultracapacitor module for the Hybrid Energy Storage System

Transcription

1 A Study on the dynamic SO compensation of an Ultracapacitor module for the Hybrid Energy Storage System Hyun-Sik Song 1, Jin-Beom Jung 2, Baek-Haeng-Lee 2, Dong-Hyun Shin 2, Byoung-Hoon Kim 2, Yu-heol Park 2, Hoon-Heo 1, Hee-Jun Kim 3 1 Dept. of ontrol and Instrumentation Eng., Korea University, Korea 2 Electronic System R&D enter, Korea Automotive Technology Institute, Korea 3 Dept. of Electrical Eng., Hanyang University, Korea, hjkim@hanyang.ac.kr Abstract There have been a number of methods proposed for accurate state of charge (SO) estimation of a battery as an important role of the battery management System (BMS). However, adequate test data of batteries and BMS development costs must be taken into consideration before application in actual vehicles. In addition, different algorithms need to be adopted according to the type of energy storage device and system configuration. For a mild-he hybrid energy storage device that uses the ultracapacitor module and the RLA battery, the small capacitance of ultracapacitor module makes accurate SO difficult. Accordingly, this paper introduces a dynamic SO estimation algorithm capable of SO compensation of the ultracapacitor module even when there is current input and output. A cycle profile that simulates the operating conditions of a mild-he was applied to a vehicle simulator to verify the effectiveness of the proposed algorithm based on testing. I. INTRODUTION Energy storage devices for Hybrid Electric ehicles (HEs) are required to have excellent power density and energy efficiency characteristics and guarantee long life cycle and usage even under frequent high-current charging and discharging conditions. According to these needs, energy storage devices with high power and longevity, such as Ni- MH or Li-Ion, have been proposed and used. In the case of mild-hes that only use limited functions of HEs including idle-stop/start or regenerative braking, there have been a number of studies on using alue-regulated Lead-Acid (RLA) batteries that yield relatively high price-performance ratios. However, using a RLA battery as a stand-alone energy storage device for a mild-he has disadvantages in terms of lifetime, power characteristics and weight, calling for improvements in the performance of the RLA battery, which has its limitations. In turn, recent studies have suggested combining the RLA battery with the ultracapacitor module capable of high efficiency and high-current charging and discharging to implement an energy storage device with improved efficiency and prolonged lifetime. [1][2][3] In order to effectively use the RLA battery and the ultracapacitor module as a hybrid energy storage device, accurate State of harge (SO) estimation must be performed by the Battery Management System (BMS) that manages and monitors the energy storage device. SO estimation is one of the most important issues for application products that use energy storage devices. Known methods for SO estimation include using chemical reaction, voltage, current integration and internal impedance. [4] The chemical method involves estimating SO by numerically representing the linear reduction of electrolyte s specific gravity (SG) during battery discharging. SG is the ratio of the weight of a solution to the weight of an equal volume of water at a specified temperature. However, the chemical method can only be used when electrolyte can be directly accessed, such as in the case of lead-acid batteries. Another drawback of the chemical method is that it can yield different results according to the ambient temperature and the amount of electrolyte. The voltage method uses the correlation between SO and the battery voltage during charging and discharging. Although this method provides accurate SO based on Open ircuit oltage (O) after a sufficient time for stabilization, different results are produced depending on the ambient temperature, discharging rate (-rate), usage time, and variation in the characteristics of battery s internal resistance. Also referred to as oulomb counting or Ampere-hour counting, the current method calculates SO by integrating battery s input and output current. However, this method requires that an accurate account of initial capacity and current leakage and the sensing are accumulated over time, gradually increasing the SO. The internal impedance method estimates SO by measuring the variation in internal impedance according to battery s charging and discharging. Drawbacks of the internal impedance method are that it is difficult to measure battery s impedance while the cell is reacting and that impedance is sensitive to temperature. Since precise SO estimation is difficult with aforementioned methods, new approaches such as Fuzzy logic, Kalman filtering and Neural network, are being proposed, but they call for accuracy in battery modeling and there are drawbacks such as significant amounts of test data and increased costs associated with using 32-bit MUs for fast numerical processing. Therefore, accurate SO estimation requires research efforts that utilize advantages of each method and do not use relatively large amounts of test data while taking into consideration the cost perspective. In a hybrid energy storage device, the ultracapacitor module is used as the main power source due to its /09/$ IEEE

2 outstanding power characteristics. Moreover, storage device s efficiency can be improved and usage prolonged by controlling the switch according to ultracapacitor module s SO. [1] Therefore, in addition to battery s SO accuracy, ultracapacitor module s SO estimation that involves relatively small capacitance so that SO is easily changed even with small amounts of input and output current is more important than anything for applying the switch control logic. The conventional approaches of calculating ultracapacitor module s SO include the Ampere-hour counting method that integrates input and output current, and the O detection method that detects ultracapacitor module s voltage with no current flowing and under a sufficiently stable state to estimates the SO value that corresponds to the detected voltage. However, since an ultracapacitor module has a much smaller capacitance than a battery, even a minute current results in accumulation of SO for the Ampere-hour counting method. As for the O detection method, accurate SO estimation is feasible only if measurement is made after the ultracapacitor module has stabilized. SO estimation based on voltage measurement while input or output current is present accompanies. In order to perform accurate SO estimation of the ultracapacitor module in hybrid energy storage system that use an ultracapacitor module, a RLA battery and a switch, this paper proposes an accurate SO estimation algorithm that can be compensated while in vehicle operation. It primarily uses Ampere-hour counting and when ignition key (IG) is turned on, initial SO compensation is carried out according to the O as well as voltage-based SO compensation in the operation region determined by testing. The accuracy and effectiveness of the proposed algorithm was tested and verified with a vehicle simulator using a cycle profile that simulates the driving conditions of mild-hes. II. DYNAMI SO OMPENSATION OF AN ULTRAAPAITOR MODULE A. System onfiguration of Hybrid Energy Storage System The hybrid energy storage system used for this study consists of an ultracapacitor module, a 36 RLA battery, a switch and the BMS, as shown in Fig. 1. The 42 ultracapacitor module was built by serially connecting 18 cells developed by NESSAP, 2.7, 1700F capacity per cell. [5] Made by GS-Yuasa, the RLA battery has a capacity of 20Ah and a nominal voltage of 36 and is the same model used in Toyota rown. The switch between the ultracapacitor module and the battery uses the NAIS EB 100 relay developed by Matsushita for 42 systems. [1] For the system implemented for this study, the ultracapacitor module is always connected to the 42 power net and used independently. When high energy is required for cranking or starting after an idle-stop, the BMS controls the switch for parallel operation with the battery according to the SO of the RAL battery and the ultracapacitor module as well as the mode of the ehicle ontrol Unit (U). During Fig. 1. Hybrid energy storage device configuration [1] regenerative braking and vehicle acceleration, ultracapacitor module s energy is transferred to the battery in advance to receive more energy. In the case of power assist, the switch or the D/D converter is controlled to transfer the energy to the battery. The main function of the BMS controller is to monitor the SO of the ultracapacitor module and the RLA battery to estimate the SO of the hybrid energy storage device, supply the ISG (integrated starter generator) with the energy required for vehicle s idle stop/start, and retrieve the energy from regenerative braking. For this, the BMS controller measures the current, voltage and temperature needed to calculate the SO, transfers hybrid energy storage device s available discharge and charge energy to the U, and manages energy by controlling the switch. The BMS consists of the power supply circuit for supplying power to controllers and various sensor devices, the sensing circuit for detecting vehicle status information input and the voltage, current and temperature of the energy storage device, the digital output circuit for providing information related to the status of switch control, energy storage devices and BMS controller, and the AN communication circuit for monitoring the U and BMS. Taking into consideration the cost and interface aspects, current was measured using Nikkohm s 0.1mΩ shunt resistor by amplifying the voltage across the two shunt terminals. The voltage was dropped to a 5 level through resistance distribution. The temperature was measured using thermistor of NT type, which is 10kΩ at 25. Since current and voltage is closely related with accurate SO estimation, a 16- bit AD was used for increased precision. Information exchange between BMS monitoring and the U was implemented based on AN2.0. Infineon s 16-bit controller X164 was used for vehicle application and fast algorithm computation. The BMS configuration is shown in Fig. 2. B. SO Estimation of an Ultracapacitor module As mentioned earlier, a hybrid energy storage system that uses the ultracapacitor module as the primary power source controls the switch and the D/D converter based on the

3 Fig. 2. BMS onfiguration SO of the ultracapacitor module. In order to apply the logic, ultracapacitor module s SO estimation that involves relatively small capacitance so that SO is easily changed even with small amounts of input and output current is extremely important, as well as battery s SO accuracy. Whereas the Ampere-hour counting method and the O methods are widely used for the SO estimation of the ultracapacitor module, their disadvantages were explained earlier. Therefore, this paper proposes an algorithm that uses the Ampere-hour counting method but compensating the initial SO with the O and further compensating for the current integration by completely blocking current flow or detecting voltage in a low-current region. In this paper, the SO compensation described above shall be referred to as dynamic SO compensation. 1) SO ompensation of an Ultracapacitor module Accurate SO estimation of a vehicle based on O is feasible when sufficient time has passed for stabilization after battery usage and all of the loads are open. For this, experiments regarding the SO-O relationship must be conducted in advance for charging and discharging at different temperatures and the result data must be compiled at nonvolatile memory. Fig. 3 displays the SO-O curves for charging and discharging at various temperatures. As the graph indicates, there are discrepancies in the SO-O relationship for charging and discharging depending on the temperature. For initial SO capacity compensation, determination must be first made concerning the stabilization time after the IG is turned off and before it is turned back on. When the IG is turned off, the BMS stores the SO( SO ) based on Ampere-hour counting and the time count based on Real-Time lock (RT) in flash memory and then powers down the system. When the IG is turned on, the BMS measures ultracapacitor module s voltage and the ambient temperature, and retrieves from the look-up table the SO ( SO O ) that corresponds to the voltage and the temperature. The BMS then compares the RT stored in flash memory and the present RT to determine the time required to stabilize the ultracapacitor module. If sufficient time for stabilization has not passed, the BMS determines that the ultracapacitor module is not stable and uses the SO ( SO Flash ) stored in flash memory as the initial SO ( SO ). In other words, the BMS uses SO that was last calculated by Ampere-hour counting prior to the IG off. If sufficient time has passed, the BMS determines that the ultracapacitor module is stabilized and uses the SO ( SO O ) based on O as the initial SO ( SO ). The SO determined by the BMS is the initial SO used for Ampere-hour counting. As indicated by equations (4) and (5), since the capacitance of the ultracapacitor module is continuously integrated, SO (%) is converted into capacity (Ah) according to equation (1). SO 100 r = (1) Where r is the rated capacitance of the ultracapacitor module that was tested while discharging from fully charged condition (44) to the cutoff voltage at 1/5. Fig. 4 illustrates the flow chart for obtaining SO. Fig. 3. Ultracapacitor module s SO-O curves for charging/discharging at various temperatures Fig. 4. Flow chart for calculating the initial capacity

4 2) Dynamic SO ompensation of an Ultracapacitor module If the input/output current has been maintained at 0 for a certain period of time for stabilization, or if the input/output current is not 0 and the current is below a specific value during forward charging/discharging, the dynamic SO compensation of an ultracapacitor module involves calculating the SO ( SO ) value based on the voltage and comparing it to the SO( SO ) value calculated according to Amperehour counting to provide compensation. Therefore, it is necessary to conduct charging and discharging tests for current levels in order to determine how much input/output current should flow through the ultracapacitor module for a sufficiently small voltage, which is the difference in voltage before and after rest time, and the discrepancy in the SO value that corresponds to the voltage. It is also necessary to check at what point compensation should commence when conversion is made to charging/ discharging under the current threshold. For the first experiment, charging/discharging was performed at a predetermined current level for a specific duration (5 seconds) and voltage was measured after stabilization (30 minutes) to select an appropriate current level that produced a small. The experiment was conducted with SO 60% at ambient temperature of 25 for 2.5A, 5A, 7.5A and 10A of current. Detailed experimental results are shown in Table I and II. The SO ( SO ) that corresponds to the voltage before and after stabilization can be obtained from equations (2) and (3). SO ratio for the voltage difference before and after stabilization for 5A charging/discharging was less than ±1%. TABLE I SO before and after stabilization for 5A charging (SO 60%) Test urrent [A] 5sec [] 30min [] oltage [] SO [%] TABLE II SO before and after stabilization for 5A discharging (SO 60%) Test urrent [A] 5sec [] 30min [] oltage [] SO [%] (rest test ) SO = 100 (2) E E = (max min ) (3) 2 is the capacitance of 18 serially connected 1700F capacitors, rest is the voltage after stabilization, test is the voltage at the termination of the experiment with the corresponding current, max is the fully charged voltage, min is the cutoff voltage, and E is the difference in energy between the maximum energy at fully charged voltage and the minimum energy at the cutoff voltage. E is from equation (3). In the case of the 5A, 5 sec experiment of Table I, the SO ( SO ) before and after stabilization becomes 0.51% from equation (2). Test results indicate that the SO is proportional to the current, producing s of 0.5% and 1% at ±5A and ±10A, respectively. In order to examine whether the SO displayed similar tendencies outside of the SO 60% region, additional tests were conducted for SO 10% and SO 90%. As shown in Fig. 5, the SO 10% was smaller than the others for charging, but the SO 10% was bigger than the others for 5A below discharging region. It was also found that the Fig. 5. SO corresponding to the voltage difference before and after stabilization according to test current levels Applying the results of the first test is problematic because there is no time for stabilization during actual vehicle operation. Therefore, in order to find a region that allows compensation during operation, a second experiment was conducted to examine whether SO comes within a desired range if a test current (0, ±2.5A, ±5A) is maintained for a few seconds after maximum charging/discharging. Taking into consideration the operating conditions of the mild-he (cranking 5kW, regenerative braking 3.3kW), the maximum range of charging/discharging current that can be applied to the ultracapacitor module was set at 150A for charging and 250A for discharging. Table III lists the SO s for voltage differences after charging for 5 seconds at 150A from the cutoff voltage of 27, discharging for 4 seconds at 250A from the fully charged voltage of 44, maintaining for 3, 5, 7 and 10 seconds at stabilization current and after undergoing stabilization. The experimental results indicate that if the stabilization current flows in the same direction as the initial current, the voltage becomes greater than in the case of the opposite direction.

5 urrent[a] TABLE III SO due to stabilization current Rest urrent[a] SO[%] 3 sec 5 sec 7 sec 10 sec Fig. 6. harging at 150A followed by charging at 5A It can be deduced that the maximum would occur with 5A of current flowing in the same direction as the initial current and the average SO would be around 3%. The SO caused by the maximum voltage that can occur with the current maintaining below ±5A decreased by up to 0.52% as the duration of the current held below 5A increased from 5 to 10 seconds, but the difference was not significant. Since designing the algorithm so that SO compensation only occurs when the duration of the current held below 5A exceeds 10 seconds rather reduces the number of SO compensations, it is advisable to set the duration of current maintenance at about 5 seconds. Fig.s 6 and 7 illustrate the graphs of voltage and current when stabilization current is flowing in the same direction as the initial charging/discharging current that is generated when SO becomes the maximum. According to Fig. 6, resting after charging at 150A for 5 seconds and then at 5A for another 5 seconds, the voltage at the end of 5A charging is 36.42, which drops down to after 30 minutes of stabilization. Using equations (2) and (3), the SO becomes 3.35%. Likewise, according to Fig. 7, resting after discharging at -250A for 4 seconds and then at -5A for another 5 seconds, the voltage at the end of -5A discharging is 31.48, which increases to after 30 minutes of stabilization, yielding an SO of -3.8%. In this study, if the current is maintained at 5A or below for 5 seconds or longer, the current voltage of the ultracapacitor module is measure to obtain the SO value based on the voltage, according to which SO compensation is performed. ompensation only takes place when there is a difference of 5% or greater between the current value ( SO ) calculated by Ampere-hour counting and the compensation target value ( SO ) calculated with the voltage. This is to take into consideration the innate (up to 3.8%) of the compensation mechanism that uses instantaneous voltage. However, there is a transient SO variation when setting SO based on voltage as the compensation SO value ( SO ). Since the energy storage system and the vehicle can affect other control algorithms and instantaneous SO variation can cause unstability, it would be appropriate to perform small levels of (+) compensation during charging and Fig. 7. Discharging at 250A followed by discharging at 5A ( ) compensation during discharging for a specified duration t in a consistent manner. The dynamic SO compensation algorithm for the ultracapacitor module described earlier is shown in Fig. 8 and the associated equations can be expressed as follows. Step 1. After detection input/output current of the ultracapacitor module, calculate the instantaneous capacity of remaining by Ampere-hour counting. AS + i t f 3600 t 1 t s e t = (4) t is the Ampere-hour(Ah) accumulated up to the present moment, ASt 1is the Ampere-sec(As) accumulated up to the previous point, i t is the amount of current at the present moment, S f e is the charging/discharging efficiency. SO + t = (5) r is the initial capacity when the ignition key is turned on, and r is the rated capacitance.

6 Step 2. If the magnitude of input/output current ( i cap ) is less than the predefined value ( i ), check if the current is maintained for a specified duration ( t ). Step 3. If the magnitude of the current is maintained for the specified duration t, calculate the instantaneous voltage SO ( SO ) based on the ultracapacitor module voltage from the SO_O table. Step 4. alculate the of the remaining capacity with the instantaneous SO and the SO based on instantaneous voltage. SO = SO SO (6) Step 5. If the of the remaining capacity is greater than the predefined range (±5%), set SO as the remaining capacity ( SO ) and perform continuous compensation for the accumulated of SO over a specified duration t in small amounts. SO SO is compensated SO. SO t = SO ± (7) Step 6. When compensation for SO is complete, once again monitor whether input/output current is maintained below the specified current level for a specified duration ( t ). If it is, repeat the calculation of SO and perform SO compensation. Fig. 8. Flow chart for dynamic SO compensation III. EXPERIMENTAL RESULTS Accurate SO calculation is very important because inability to precisely calculate energy storage system s SO lead to inefficient energy management, which has adverse effects on vehicle operation. Accordingly, the levels of accuracy of the current, voltage and temperature used for the BMS to calculate SO are also important. The in current measurement was about 1% in the -250A ~ 150A range and about 2% in the low current range (-10A ~ 10A). The voltage was within 0.5% in the 27 ~ 44 region. As for temperature, the was within 3% between -40 and 90. In order to minimize the s associated with the external environment, all of the tests were conducted in a chamber at 25. To use as references for determining the accuracy of the algorithm explained below, we used Yokogawa s current transducer (751574), which is capable of measuring current with less than 0.05%, and NetDAQ 2640 to measure voltage within an range of 0.01%. In order to verify the effectiveness of the proposed algorithm, we created a random test waveform and conducted tests. As shown in Fig. 9, when charging/discharging took place between 100A and -200A and the current was changed to ±5A for 5 seconds, the SO according to the instantaneous voltage was smaller than SO, and (+) compensation was performed for SO. In Fig. 10, when charging/discharging took place as similar pattern and the current was changed to ±5A for 5 seconds, the SO according to the instantaneous voltage was greater than SO, and ( ) compensation was performed for SO. In order to examine how much the proposed algorithm affected the SO accuracy of the hybrid energy storage system, a power profile during idle stop/start and regenerative braking, phenomena that frequently occur during mild-he driving, was created and modified as shown in Fig. 11 and put through a testing of 70 cycles, after which the remaining capacity of the ultracapacitor module was checked. The cycle profile used for test was modified by adding a regenerative function to the ZPA (zero-power assist) efficiency and life test profile of the Freedom AR 42 Battery Test Manual published by INEEL and DOE. [7] The 70 cycle test was started after fully charging the ultracapacitor module and then discharging it for two hours down to 1/5 and SO 60%. In order to examine the remaining capacity after 70 cycles, the ultracapacitor module was discharged at 1/5 to the cutoff voltage. As shown in Table Ⅳ, the difference between the SO variation of the capacity before and after the cycle test calculated with the reference and by the BMS was 0.87%, which is within the general design specification of ±5% in HE applications. TABLE Ⅳ SO estimation result SO variation before and after testing Error Reference BMS SO % 24.61% 0.87 %

7 Fig. 12 is ultracapacitor module s SO graph calculated by the BMS during the profile cycle test. As shown in the Fig., the SO by Ampere-hour counting is divergent owing to current accumulation, the SO by voltage method has big in case of high-current, and the SO by dynamic SO estimation algorithm well follows Ampere-hour counting in high-current and skillfully adjusts as voltage method in lowcurrent. Fig. 9. Plus compensation I. ONLUSION In order to improve the accuracy of SO estimation of the ultracapacitor module used in hybrid energy storage systems, this paper introduced a compensation algorithm that uses compensation by O when the ignition key is turned on as well as ultracapacitor module voltage when the input/output current is below a certain level during operation. The study also verified the accuracy of the proposed dynamic SO algorithm (within 0.87% of the reference) based on a modified cycle profile test that simulated mild-he driving conditions. Even if the battery and the ultracapacitor module are modified in a hybrid energy storage system that uses the ultracapacitor module as the main power source, dynamic SO estimation can be accurately performed with a few sets of test data described in this paper. Fig. 10. Minus compensation Fig. 11. ycle profile for SO accuracy REFERENES [1] Baek-Haeng Lee, Dong-Hyun Shin, Hyun-Sik Song et al, Development of an Advanced Hybrid Energy Storage System for Hybrid Electric ehicles, Journal of Power Electronics, ol. 9, No. 1, 51-60, January 2009 [2] Paul Bentley, David A. Stone, Nigel Schofield, The parallel combination of a RLA cell and supercapacitor for use as a hybrid vehicle peak power buffer, Journal of Power Sources, olume 147, Issues 1-2, 9 September 2005, Pages [3] Adam W. Stienecker, Thomas Stuart, yrus Ashtiani, An ultracapacitor circuit for reducing sulfation in lead acid batteries for Mild Hybrid Electric ehicles J. Power Sources 156 (2006) [4] [5] Do Yang Jung, Young Ho Kim, Sun Wook Kim, Suck-Hyun Lee, Development of ultracapacitor modules for 42- automotive electrical systems, J. Power Sources 114 (2003) [6] Baek-Hang Lee, Hyun-Sik Song, Dong-Hyun Shin et al., Development of 10KW energy storage system and control device for 42 power supply system, Korean Ministry of ommerce, Industry and Energy, January 2008 [7] J. Barnes, FreedomAR 42 Battery Test Manual, INEEL and DOE/ID-11070, April Fig. 12. The result of SO cycle test