IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 48, NO. 3, JUNE 2001 585 Fuzzy-Controlled Li–Ion Battery Charge System with Active State-of-Charge Controller Guan-Chyun Hsieh, Senior Member, IEEE, Liang-Rui Chen, and Kuo-Shun Huang Abstract—A fuzzy-controlled active state-of-charge controller (FC-ASCC) for improving the charging behavior of a lithium–ion (Li–ion) battery is proposed. The proposed FC-ASCC is designed to replace the general constant-voltage charging mode by two kinds of modes: sense and charge. A fuzzy-controlled algorithm is built with the predicted charger performance to program the charging trajectory faster and to remain the charge operation in a proposed safe-charge area (SCA). A modeling work is conducted for analyzing and describing the Li–ion battery in charging process. A three-dimensional Y-mesh diagram for describing the charging trajectories of the proposed FC charger is simulated. A prototype of a Li–ion battery charger with FC-ASCC is simulated and realized to assess the predicted charging performance. Experiment shows that the charging speed of the proposed FC charger compared with the general one increases about 23% and the charger can safely work in the SCA. Index Terms—Charge mode, fuzzy-controlled state-of-charge controller, safe-charge area, sense mode. active I. INTRODUCTION T HE rapid progress in developing computer and communication systems promotes the design trend of portable electronic apparatuses which are light, thin, short, and small. The secondary batteries then become the significant power source for the portable electronic apparatuses. Nowadays, widely used secondary batteries such as NiCd and NiMH are not satisfying people’s requirements due to lack of high-energy capacity and bulky size. Furthermore, another consideration is possibly the environmental pollution, such as from the cadmium in a NiCd battery. However, many advantages, such as no memory effect, high operation voltage, and high energy density in weight (Wh/kg) and volume (Wh/L) forward the lithium–ion (Li–ion) battery in becoming the acceptable battery for portable electronic systems [1]–[8]. For charging processes, constant-current (CC) and constant-voltage (CV) techniques are still widely used for Li–ion batteries [2], [7], [9]. However, it always takes about 25% 40% of the complete charging period to fulfill about 75% 80% of the totally required charge Manuscript received March 23, 2000; revised February 14, 2001. Abstract published on the Internet February 15, 2001. G.-C. Hsieh is with the Department of Electronic Engineering, National Taiwan University of Science and Technology, Taipei 106, Taiwan R.O.C. (e-mail: gchsieh@et.ntust.edu.tw). L.-R. Chen is with the Department of Electronic Engineering, National Taiwan University of Science and Technology, Taipei 106, Taiwan R.O.C., and also with the Department of Electronic Engineering, Chien Kuo Institute of Technology, Changhua 500, Taiwan, R.O.C. K.-S. Huang was with the Department of Electronic Engineering, National Taiwan University of Science and Technology, Taipei 106, Taiwan R.O.C. He is now with Sunplus Technology Company, Ltd., Hsinchu 300, Taiwan, R.O.C. Publisher Item Identifier S 0278-0046(01)03813-8. capacity during the CC mode, and the time required for the remaining 20% 25% capacity to be charged in CV mode should be 3 the charging length in CC mode [2]. Due to complex electrochemical characteristics in a Li–ion battery, a precise model is difficult and complex to build [10]. In addition, human expertise in the charging process for a Li–ion battery is also not accurately realized by a usual control rule. Thus, a fuzzy control sense is adopted for solving the nature language of the Li–ion battery charger in man–machine interface [11], [12]. The estimation of the remnant charge in the secondary battery has been conducted by the open-circuit voltage monitoring and Coulomb counting methods, but it is not accurate [13]. James proposed a microprocessor-based estimator for detecting the state-of-charge in the battery by combining the two methods mentioned [14]. Stolitzka indicated that Jame’s method could provide 1% accuracy for the charge detection but kept the high cost for realization [13], [14]. In this paper, a fuzzy-controlled active state-of-charge controller (FC-ASCC) is presented to replace the usual CV mode. The proposed FC-ASCC can instantly monitor the charging state of the Li–ion battery for adaptively programming its charging performance. A fuzzy-controlled current source (FCCS) in FC-ASCC can adaptively provide a suitable charging current for the battery charge. The proposed FC-ASCC is conducted by a sense mode (SM) and a charge mode (CM) with a fuzzy-controlled strategy under a safe-charge area (SCA). In this paper, modeling and fuzzifying for describing the proposed system are conducted. A prototype of a Li–ion battery charger with two batteries in series package is designed and realized to assess the system performance. The charging time of the proposed FC-ASCC compared with the general CV mode has been reduced by about 23%. II. FC-ASCC Shown in the dotted block of Fig. 1(a) is the proposed FC-ASCC, which consists of a current/voltage converter (C/V-C), a fuzzy-logic controller (FLC), an FCCS, and a clock timer. All of them are driven by a constant voltage source (CVS). The proposed FC-ASCC is operated in an SCA, which is defined as a safe maximum in-charge voltage of the Li–ion battery. The periodical SM and CM conduct the FC-ASCC with and a charging period defined in Fig. 1(b) a sense period and (c). In SM, the open-circuit voltage detection (OVD) and the charging current detection (CCD) are the primary work for checking the state of charge in the Li–ion battery. In other words, the OVD is for detecting the open-circuit voltage of the package battery in , and the CCD is for sensing the 0278–0046/01$10.00 © 2001 IEEE 586 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 48, NO. 3, JUNE 2001 Fig. 1. (a) Configuration of the proposed FC charge system. (b) Control sequence for (a). (c) Working state diagram for (a). Fig. 2. Structure of FLC. required charging current in . After the SM, and are then converted into a coded number through the C/V-C. A suitable control algorithm from the FLC is then deduced to the FCCS to charge the Li–ion battery with in CM during the period . Remarkably, the minimum sense is determined by the computing speed of the period s is microprocessor and the number of charging periods determined according to the defined SCA of the Li–ion battery. HSIEH et al.: FUZZY-CONTROLLED Li–ION BATTERY CHARGE SYSTEM Fig. 3. 587 Equivalent charging models for one-cell Li–ion battery. (a) General charging strategy. (b) Proposed FC charging strategy. For ease of control, a digitally coded strategy is used in the , , and are proposed FC-ASCC. The coded , , and , respectively. The FLC in then given by Fig. 2 consists of a fuzzifier, a rule library, an inference engine, ( ), a defuzzifier, and three prescaling factors, ( ), and ( ). For ease of analysis and description in our proposed charging rule, we presume that all battery cells in the battery pack have the same characteristics. Thus, the and are normalization factors of the package battery defined as (1) and Fig. 4. Predicted charging trajectories of the general ( - - -) and proposed FC (—) charging processes for one Li–ion battery. Thus, the open-circuit voltage , sense current , and of the one-cell battery are, respectively, charging current given by (2) and are the number of batteries for series package where is deand shunt package, respectively. The scaling factor fined as (3) (4) (5) and (6) 588 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 48, NO. 3, JUNE 2001 Fig. 5. Membership functions of the charging strategy in FC-ASCC mode for (a) voltage sense, (b) current sense, and (c) the output of FLC. The coded numbers , , and are then represented by TABLE I CONTROL RULE TABLE FOR THE FC-ASCC (7) (8) and (9) , A set of fuzzy variables for the one-cell coded number ( ) is then given through the fuzzifier. According to the prescribed rule library and the fuzzy variables, a deduced fuzzy set from the inference engine is then obtained. Accordingly, after , a suitable charging current from the defuzzifying the FCCS is then provided for charging the Li–ion package battery in CM. A cyclic charging process continues until the sense curis smaller than or equal to 0.01 C, where the battery rent is fully charged [8]. III. CONTROL STRATEGY The proposed fuzzy-controlled and the general chargers for one Li–ion battery are, respectively, modeled in Fig. 3(a) and (b), in which the CC mode in each chargers is the same. The FC-ASCC mode is proposed to replace the general CV mode so as to proceed with the charging process in an SCA and reduce the charging time. The predicted charging trajectories of the two charge systems are, respectively, depicted in Fig. 4 for comparison. Remarkably, the charging performances are nearly the same in both two CC modes, but the charging speed in the FC-ASCC is obviously faster than that in the general CV mode. in the general CV mode and the CM The charging current of the FC-ASCC for one Li–ion battery cell can be estimated by (10) , , and are, respectively, the in-charging where voltage, the open-circuit voltage, and the inner resistance of the is detected Li–ion battery. In Fig. 3(b), the sense current with the nominal reference at the CCD of SM during V and can be described by (11) HSIEH et al.: FUZZY-CONTROLLED Li–ION BATTERY CHARGE SYSTEM 589 Fig. 6. Simulations of the proposed FC charging trajectories in 3-D Y-mesh diagram. (a) m = 0:2. (b) m = 0:5. (c) m = 0:8. Remarkably, (11) is valid when the detected at OVD of SM . Interestingly, in (11) is is not greater than 4.2 V during inherently temperature dependent [1], but this effect can be inand no temperature consideration is cluded in the measured and required in the proposed FC-ASCC. With the detected , a suitable charging current from FCCS is provided for charging the Li–ion battery in CM during . The required in the FCCS can then be described by charging current (12) is a function of the FLC. Since the maximum fault where over voltage threshold in the one-cell Li–ion battery is 4.25 V should be limited by [15], [16], the in-charging voltage (13) From (10) and (13), the required charging current can then be determined by in CM (14) Substituting (11) into (14), we have (15) Equation (15) is the design reference for the safe-charge consideration in determining the membership function in the FLC. The membership function of the charging strategy for the FC-ASCC is described in Fig. 5. Fig. 5(a) shows the degree of (coded as ) versus the voltage-sense fuzzy variable ( ) and is described by two the open-circuit voltage linguistic terms of Small and Big in trapezoid form. Fig. 5(b) ( ) is the degree of the current-sense fuzzy variable ( ) and is described by Small, Middle, and Big versus in triangular form. Fig. 5(c) is the degree of the output fuzzy ( ) versus ( ) and is described by sinvariable gleton types of Small, Middle, and Big. The singleton type used is to minimize the accounting capacity of the microprocessor used in the FC-ASCC. The singleton type Small is preset at (i.e., H) since the trickle current in the Li–ion battery is only 1% of the bulk current [15], where 590 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 48, NO. 3, JUNE 2001 is the maximum charge current (i.e., bulk current) and is defined as singleton type Big. The singleton type Middle is ( ). The value defined by is determined according to the charge trajectory to be used is and clearly depicted in Fig. 7, of which the selection of described in detail in Section IV. The fuzzy control rules for the FC-ASCC are briefly described as follows. is Small (i.e., V and 1) When ), it means that the energy to be charged to the battery is large. In this sense, the rule is is Small is Big (R.1) is Big (i.e., V and ) 2) When is Big (i.e., , and ), and the energy to be stored in the battery is still large and the required charging current in CM would be Big (i.e., and ). The rule is is Big is Big is Big (R.2) is Big (i.e., V and ) is Middle (i.e., and ), the energy to be stored in the battery is not still large and the required charging current in CM would and ). The be Middle (i.e., rule is 3) When and is Big is Middle is Middle (R.3) is Big (i.e., V and ) is Small (i.e., and ), the energy to be stored in the battery is small and the required charging current in CM would be and ) to avoid Small (i.e., overcharge. The rule is 4) When and is Big is Small is small (R.4) The complete fuzzy control rule for the FC-ASCC is tabulated in Table I. IV. DESIGN CONSIDERATIONS The complete charging period and the value of the membership function are the key factors in designing the proposed FC-ASCC. The complete period consists of the sense period and the charge period . Two samples conducted in SM include the detection of in and the during period in . The charge period is for charging the sense of is Li–ion battery in CM mode. The sense time primarily determined by the operation speed of the fuzzy inference and is much smaller than . For convenience in design, . The Li–ion battery voltage varying we presume that with respect to time is quite a linear portion in CC mode [3], is for describing a capacitor in [9]. Since Fig. 7. Required charging current i (t) in CM with respect to the sense current i (t) in SM for various ms. linear charge with a constant current , we can then approximately describe the Li–ion battery as a capacitor during the charging process for ease of analysis. In each charging period , the charge incremented in Li–ion battery can then be given by (16) is the charging current deduced from the FCCS at . Since is almost a constant current during the charging period , where Constant Therefore, the open-circuit voltage Li–ion battery can be given by (17) incremented in the (18) is the equivalent capacity of the Li–ion battery in where farads. From (16) and (18), we have (19) from the FC-ASCC is Since the deduced charging current distinct in each period , the totally accumulated charge can then be given by (20) where is the number of charging times in CM. The total increon the Li–ion battery during periods mental voltage is thus given by (21) For safe charging consideration, the in-charging voltage on the battery voltage during CM mode should not be over 4.25 V HSIEH et al.: FUZZY-CONTROLLED Li–ION BATTERY CHARGE SYSTEM 591 Fig. 8. (a) Realization block diagram of the proposed FC charge system. (b) Control time sequences for the charging strategy. [15], [16]. Thus, the final incremental voltage given by should be satisfied with (22) From (19) and (22), we have (23) The complete charging period by can then be estimated from (23) (24) For convenience in analysis, we take the results given from Section V for describing the design consideration. Fig. 6 is a three–dimensional (3-D) Y-mesh diagram for describing the charging trajectories of the proposed FC charger in simulation, in which three kinds of typical charging conditions such as , , and are shown. It is clearly shown that the charging trajectories during CC mode are all Fig. 9. Remanent charge capacity percent with respect to the sense current i (t) for Li–ion battery. the same, but are different for various s in FC-ASCC mode. in CM with respect to the sensed in The deduced SM for various s is simulated in Fig. 7. Interestingly, we 592 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 48, NO. 3, JUNE 2001 Fig. 10. Charging performance of the general (- - -) and proposed FC (—) charging systems. (a) Remanent capacity versus charging time. (b) Detected open-circuit voltage v (t) and the in-charging voltage v (t). (c) Charging trajectories of the deduced i (t). find that the charging trajectory in the FC-ASCC mode with is the same as that in the general CV mode. When , the deduced in FC-ASCC is less than the charging current in the general CV mode, conversely when . In order to increase the charging speed, is the preferred design reference for the proposed FC-ASCC. However, the maximum deduced current should be limited by value can be easily (15) for safe-charge consideration. The determined by plotting the equation (15) in Fig. 7, in which the shaded area is the design reference. V. REALIZATION AND EXPERIMENTATION A design example of the proposed FC-ASCC for the Li–ion battery is examined. Two Li–ion batteries (Panasonic CGR and 18 650) to be charged are packaged in series ( ). The final and the maximum permissible in-charge voltages are, respectively, specified as 8.4 V (4.2 V/cell) and 8.5 V (4.25 V/cell). For safe-charge consideration, the (bulk curmaximum and minimum charging currents (trickle current) are, respectively, specified rent) and A C) and 5 mA (i.e., as 0.5 A (i.e., C), where A is the rated charge current for the package battery. The physical block diagram of the proposed FC-ASCC and its control sequence are shown in Fig. 8. The realized FC-ASCC is composed of a timer, two switches, a constant voltage source, a constant current source, a fuzzy-controlled current source circuit, an A/D converter (ADC0804), a D/A converter (DAC0800), and an FLC with latches (8-bit microprocessor EM78447B). In this example, we can get , , and from (1)–(3). By experiments, the equivalent capacity of the battery F and its inner series is estimated as . Since the maximum charging current resistance A, from (24), the maximum charge period can be estimated as (25) s. The required sampling In this design, we choose , detecting , and fuzzy infertime, including sensing s; the charging time in CM is then ence, is about s. The permissible charging current in CM can then by be given from (15) with (26) The proper charging current in this design is then given from (26) as shown in the shaded area of Fig. 7. It is clearly shown HSIEH et al.: FUZZY-CONTROLLED Li–ION BATTERY CHARGE SYSTEM that the maximal value cannot be over 0.74 and we choose as the design reference. The remanent charge cafor this package pacity with respect to the sense current Li–ion battery is measured in Fig. 9, in which the sense curis inversely proportional to the remanent charge carent pacity. In words, it means that the remanent charge in the batis large. Thus, the remanent charge in the tery is small if . battery can then be easily estimated by the sense current The charging behaviors of the proposed FC-ASCC and the general charger are experimentally represented in Fig. 10. Fig. 10(a) shows the charged capacity with respect to the charging time. Fig. 10(b) shows the open-circuit voltage and the in-charging voltage in detection and charging states, respectively. Clearly, the open-circuit voltage is always less than in-charging voltage until both voltages are almost equal to each other when the battery is fully charged. Fig. 10(c) shows the charging trajectories . Obviously, the charging current for the FC-ASCC of the is greater than that in CV mode of the general charger before reaches 0.01 C. Thus, the charging the charging current times for the proposed FC charger and the general charger are 243 and 280 min, respectively. The charging performance in the FC-ASCC has been improved by about 23% compared with that in the usual CV mode. Remarkably, the charging times in CC mode for both systems are almost the same as 133 min. VI. CONCLUSION In this paper, an FC-ASCC has been proposed to replace the general CV charge mode in a charger system. The proposed FC charge system can actually increase the charge speed and continue the charge process in an SCA after the CC mode. Two processes, SM and CM, included in the FC-ASCC are for periodically sensing the charging state and keeping the charging trajectory in the SCA. A 3-D Y-mesh diagram was explored to describe the trajectory of the charge behavior and was referred to as the design reference. A prototype of the Li–ion charger was examined with simulation and realization. The experimental results are very close to the theoretical prediction. The proposed FC-ASCC, compared with the general CV mode, can improve the charging performance about 23%. REFERENCES [1] G. Nagasubramanian and R. G. Jungest, “Energy and power characteristics of lithium–ion cell,” J. Power Sources, vol. 72, pp. 189–193, 1998. [2] J. A. Carcone, “Performance of lithium–ion battery systems,” in Proc. WESCON Idea/Microelectron. Conf., Anaheim, CA, Sept. 1994, pp. 242–248. [3] B. Carter, J. Matsumoto, A. Prater, and D. Smith, “Lithium ion battery performance and charge control,” in Proc. 31st Intersociety Energy Conversion Engineering Conf., Washington, DC, Aug. 1996, pp. 363–368. [4] H. Oman, “Battery news from the 32nd Intersociety Energy Conversion Engineering Conference,” IEEE Aerosp. Electron. Syst. Mag., vol. 13, pp. 23–31, Mar. 1998. [5] M. J. Isaacson, M. E. Daman, and R. P. Hollandsworth, “Li–ion batteries for space applications,” in Proc. 32nd Intersociety Energy Conversion Engineering Conf., Honolulu, HI, July 1997, pp. 31–34. [6] S. Hossain, A. Tipton, S. Mayer, and M. Anderman, “Lithium ion cells for aerospace applications,” in Proc. 32nd Intersociety Energy Conversion Engineering Conf., Honolulu, HI, July 1997, pp. 35–48. [7] C. Lurie, “Evaluation of lithium ion cells for space applications,” in Proc. 32nd Intersociety Energy Conversion Engineering Conf., Honolulu, HI, July 1997, pp. 58–63. 593 [8] P. David and D. Stefano, “Survey of lithium–ion battery performance for potential use in NASA mission,” in Proc. 32nd Intersociety Energy Conversion Engineering Conf., Honolulu, HI, July 1997, pp. 39–41. [9] W. F. Bentley and D. K. Heacock, “Battery management consideration for multi-chemistry system,” IEEE Aerosp. Electron. Syst. Mag., vol. 11, pp. 23–26, May 1996. [10] S. Gold, “A PSPICE macromodel for lithium–ion batteries,” in Proc. IEEE 12th Annu. Battery Conf. Applications and Advances, Jan. 1997, pp. 215–222. [11] L. A. Zadeh, “Fuzzy logic computing with words,” IEEE Trans. Fuzzy Syst., vol. 4, pp. 103–111, May 1996. [12] R. C. Berkan and S. L. Trubatch, Fuzzy System Design Principles. Piscataway, NJ: IEEE Press, 1997. [13] D. Stolitzka and W. S. Dawson, “When is it intelligent to use a smart battery,” in Proc. 9th Annu. Battery Conf. Applications and Advances, CA, Jan. 1994, pp. 173–178. [14] J. H. Aylor, A. Thieme, and B. W. Johnson, “A battery state-of-charge indicator for electric wheelchairs,” IEEE Trans. Ind. Electron., vol. 39, pp. 398–409, Oct. 1992. [15] Product Data Handbook, Unitrode Corp., Merrimack, NH, 1997. [16] L. W. Hruska, “Smart battery and lithium–ion voltage profiles,” in Proc. IEEE 12th Annu. Battery Conf. Applications and Advances, Jan. 1997, pp. 205–210. [17] G. C. Hsieh, L. R. Chen, and K. S. Huang, “Fuzzy-controlled active state-of-charge controller for fasting the charging behavior of Li–ion battery,” in Proc. IEEE IECON’99, San Jose, CA, Nov. 1999, pp. 400–405. = Guan-Chyun Hsieh (S’81–M’87–SM’95) was born in Hua-Lien, Taiwan, R.O.C., in 1950. He received the B.S. degree from National Taiwan Institute of Technology, Taipei, Taiwan, R.O.C., the M.S. degree from National Chiao-Tung University, Shinchu, Taiwan, R.O.C., and the Ph.D. degree from National Taiwan University, Taipei, Taiwan, R.O.C., in 1976, 1981, and 1986, respectively, all in electronic engineering. He is currently a Professor in the Department of Electronic Engineering, National Taiwan University of Science and Technology, Taipei, Taiwan, R.O.C., where he joined the faculty in 1981. He has engaged in research and teaching in the areas of power electronics, electronic circuit design, control systems, and phase-locked servo systems. Dr. Hsieh received the 1993 Engineering Paper Award from the Chinese Institute of Engineers and the 1997 Best Joint Projects Award from the Ministry of Education. He is a member of the Chinese Institute of Engineers and the Chinese Institute of Electrical Engineering. Liang-Rui Chen was born in Changhua, Taiwan, R.O.C., in 1971. He received the B.S. degree in 1994 and the M.S. degree in 1996, both in electronic engineering, from National Taiwan University of Science and Technology, Taipei, Taiwan, R.O.C., where he is currently working toward the Ph.D. degree in the Department of Electronic Engineering. Since 1999, he has been a Lecturer in the Department of Electronic Engineering, Chien Kuo Institute of Technology, Changhua, Taiwan, R.O.C. His major research interests are fuzzy control, phase/frequencylocked servo systems, battery chargers, and electronic circuit design. Kuo-Shun Huang was born in Taipei, Taiwan, R.O.C., in 1970. He received the B.S. degree in electrical engineering and the M.S. degree in automation and control engineering from National Taiwan University of Science and Technology, Taipei, Taiwan, R.O.C., in 1997 and 1999, respectively. Since 1999, he has been an R&D Engineer with Sunplus Technology Company, Ltd., Hsinchu, Taiwan, R.O.C. His major research interests are intelligent control, battery chargers, and electronic circuit design.