32 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 7, NO. 1, JANUARY 2016

Transcription

1 32 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 7, NO. 1, JANUARY 2016 An Online Frequency Tracking Algorithm Using Terminal Voltage Spectroscopy for Battery Optimal Charging Ala A. Hussein, Member, IEEE, Abbas A. Fardoun, Senior Member, IEEE, and Samantha S. Stephen Abstract This paper proposes an online tracking algorithm to allocate and track the optimal charging frequency for common batteries in real time under any condition. The optimal frequency refers to the frequency of the sinusoidal component of the charging current at which the internal ac impedance of the battery is minimal. The ac impedance depends on the physical properties of the battery s electrodes and electrolyte. It also varies nonlinearly with temperature, state-of-charge (SOC), and state-of-health (SOH). Although the ac impedance can be determined offline using sophisticated ac battery models or measurement equipment, there is no direct way to measure it in real time. The proposed technique has the capability of allocating and tracking the optimal frequency in real time without using an ac battery model or measurement equipment. Moreover, the proposed technique is simple, inexpensive to implement, and applicable to any battery cell or pack. Derivation of the proposed technique is presented and followed by simulation and experimental verification. Index Terms Charging efficiency, constant current constant voltage (CCCV), pulse charging, sinusoidal ripple charging (SRC), voltage spectroscopy. I. INTRODUCTION T HE SERVICE life of a battery is a major concern in many applications that use batteries as a primary or secondary source of energy. The battery often reaches its end-of-service prematurely leading to increased battery replacement costs [1] [3]. One way to improve the lifetime of the battery is by optimizing the charging process. According to [4], the charging process must address the chemical reaction occurring inside the battery. That is, the charging process must work with the electrochemical process to optimize the chemical reaction by minimizing the internal heat generated. Many charging techniques have been used to charge batteries in general. Among these techniques are the constant current constant voltage (CCCV), pulse and sinusoidal-with-ripple charging techniques, which are briefly discussed herein. Manuscript received May 06, 2015; revised July 16, 2015; accepted August 16, Date of publication September 11, 2015; date of current version December 11, This work was supported in part by UAE University under research Grant G and in part by Japan Cooperation Center Petroleum (JCCP) under Contract 21N125. Paper no. TSTE (Corresponding author: Ala A. Hussein.) The authors are with the Department of Electrical Engineering, United Arab Emirates University, Al Ain, UAE ( ala.hussein@uaeu.ac.ae). Color versions of one or more of the figures in this paper are available online at Digital Object Identifier /TSTE A. CCCV Charging Techniques ([4] [10]) In CCCV charging techniques, a constant current is used to charge the battery to a predefined voltage. Once the predefined voltage is reached, the charger applies a constant voltage across the battery s terminals until the current drops below a specified value at which moment the charging process is terminated. Although these techniques prevent overcharge to occur, they are time consuming. Furthermore, they generate internal heat in the cell, which results in inefficient charging process and eventually reduced lifetime. B. Pulse Charging Techniques ([11] [13]) These techniques use a pulse current to charge the battery allowing the battery to rest periodically. These rest periods allow the electrolyte s ions to distribute more evenly resulting in a more efficient charging process compared with the CCCV method. Moreover, some other pulse techniques use a negative pulse for a very short time during the rest period to allow further uniform ion distribution. Although introducing a negative pulse is claimed to improve the charging process, the implementation cost of such method is high due to the need of a bidirectional charger to allow bidirectional current flow. Although CCCV and pulse techniques have been widely used in many commercial applications, they do not totally optimize the charging process due to the high internal losses they generate. C. Sinusoidal-With-Ripple ([14]) According to [14], using a sinusoidal current with ripple has several advantages such as improving the charging efficiency, charging speed, and lifetime of the battery. This technique uses a sinusoidal current profile with ripple or dc shift, where the dc shift is equal to the amplitude of the sinusoidal waveform as illustrated in the following equation: i(t) = I M + I M cos(2πft). (1) In order to allocate the optimal frequency, an impedance analyzer was used to measure the impedance magnitude at a certain range of frequencies, and the optimal frequency was set to the value at which the impedance magnitude has a minimum value [14]. In [15], a technique was proposed to allocate and track the optimal frequency using a phase-locked loop IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See for more information.

2 HUSSEIN et al.: ONLINE FREQUENCY TRACKING ALGORITHM USING TERMINAL VOLTAGE SPECTROSCOPY 33 TABLE I OCV-SOC LOOKUP TABLE Fig. 1. Simple ac cell model. (PLL). Although the method in [15] allocates and tracks the optimal frequency without using measurement equipment (i.e., impedance analyzer), its convergence time is linearly proportional to the distance between the initial guess and the true value of the optimal frequency because the frequency is incremented or decremented using a fixed step-size in the iteration phase. In this paper, an algorithm with a variable step-size for allocating and tracking the optimal charging frequency is proposed. The main advantage of the proposed algorithm is that it allows convergence in a short time with very accurate results by varying the input current s frequency and monitoring the terminal voltage of the battery. In addition, the proposed algorithm can be used to charge any battery type with the highest possible efficiency. This paper is organized as follows. In Section II, a discussion on ac battery cell models is presented. In Section III, the proposed algorithm is derived and compared with fixed step-size algorithms. Simulations results are presented in Section IV followed by experimental verification in Section V and finally conclusion in Section VI. II. AC CELL MODEL A battery cell consists of two electrodes (negative anode and positive cathode), electrolyte, and a porous separator to isolate the two electrodes. The transport, kinetic, and mass-transfer properties of these components such as the thickness, density, specific heat, thermal conductivity, particle radius, and diffusion coefficient vary significantly with the cell s state-of-charge (SOC), temperature, and age [16]. In order to develop a tool that allows understanding and analyzing the electrical behavior of the battery cell, an equivalent-circuit model is used. Fig. 1 shows an ac cell model for a Li-ion battery cell [15], [17] [24]. In Fig. 1, the resistors account for the Ohmic losses (I 2 R losses) in the form of internal heat. The resistor-capacitor (RC) network models the transient behavior of the cell. The inductor L accounts for the inductive part of the cell s internal impedance. The open-circuit voltage (OCV) represents the terminal voltage of the ideal cell (if the internal impedance is zero) and is also known as the electromotive force. All the mentioned parameters change considerably with the SOC, state-of-health (SOH), and environmental conditions. To optimize the charging process, the internal losses inside the cell must be minimized to make the charging more efficient. That is, instead of charging the cell with a constant current as in traditional chargers, the cell will be charged with an ac current at an optimal frequency at which the ac impedance of the cell is minimized. The optimal charging current profile is given in the following equation: i(t) = I M + I M cos(2πf opt t). (2) In order to minimize Z ac, the frequency of the current is varied and the instantaneous output and input power are calculated as follows: p out (t) =OCV (t) i (t) (3) p in (t) =v (t) i (t). (4) Assuming that the OCV is constant in one charging current cycle, i.e., 1/f s, the average output and input power over one charging cycle are given in the following equations: The charging efficiency is P out =(OCV)(I) avg (5) P in =(V ) avg (I) avg. (6) η = P out 100%. (7) P in Substituting (5) and (6) in (7), the charging efficiency can be calculated at any frequency by dividing the OCV (the terminal voltage of the ideal cell) by the average terminal voltage of the cell as given in the following equation: η = OCV 100%. (8) (V ) avg Obviously, the charging efficiency in (8) is inversely proportional to (V ) avg. A simplified OCV-SOC lookup table is given in Table I [5]. In order to use the model in Fig. 1 in determining the optimal charging frequency, the values of the parameters that were used are the same as those used in [14]. However, since these parameters vary with SOC, load, and environmental conditions, a random noise was added as shown in the following equations: OCV =2.2+ΔOCV (V) (9) R o = ΔR o (mω) (10) R f = ΔR f (mω) (11) C f = ΔC f (mf) (12) L = ΔL(µH) (13) where 0 ΔOCV 1.4; 0 ΔR o 50; 0 ΔR f 20; 0 ΔC f 30; 0 ΔL 1. In (9), the minimum and maximum values for OCV are 2.2 and 3.6 V, respectively, which are equivalent to SOC of 0% and 100%, respectively, according to Table I. In (10) (13), the variation in the parameters is due to possible variation in SOC, SOH,

3 34 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 7, NO. 1, JANUARY 2016 Fig. 2. Z ac as a function of frequency. temperature, etc., which have a direct impact on the cell s parameters. Fig. 2 shows typical battery impedance Z ac as function of frequency for battery cell. Fig. 2 assumes battery parameter values shown in (9) (13). Z ac has a minimal value at optimal frequency f opt. An algorithm that tracks the optimal charging frequency without calculating the ac impedance is detailed in the next section. III. PROPOSED ALGORITHM The concept of the proposed algorithm is based on the fact that at the optimal frequency, the slope of the voltage frequency curve is equal to zero (the terminal voltage has a minimum value at the optimal frequency). The process of allocating the optimal frequency starts by setting an initial estimate for the optimal frequency (e.g., 100 Hz). The slope of the voltage frequency curve is calculated at f = 100 Hz. If the slope is negative, that means the optimal frequency is greater than 100 Hz, and thus the frequency is incremented. The slope is calculated again at the new frequency; if the slope is negative, the frequency is incremented again until the slope becomes positive. Once the sign of the slope changes, the process is terminated. On the other hand, if the slope is positive, that means that the initial guess of the optimal frequency is greater than the true value, and as such, the frequency is decremented. The process is terminated once a change in the slope sign is detected. In order to examine the effectiveness of using a variable step-size algorithm, a fixed step-size algorithm is derived and compared with the proposed algorithm. A. Fixed Step-Size Algorithm Using a small step requires long convergence time but results in a small or no steady-state error, whereas using a large step increases the convergence speed but results in large steady-state error. A fixed step size of 50 Hz (small) and 250 Hz (large) were used in the algorithm shown in Fig. 3. The results are presented in Section IV. B. Variable Step-Size Algorithm In order to improve the convergence time and accuracy, a variable step-size algorithm is used. The value of the step size Fig. 3. High-level flowchart of the proposed algorithm. can range between a minimum value (step) min, and a maximum value (step) max. The step value is chosen based on the voltage slope. Around the optimal frequency, the slope is very small; therefore, a small step is more suitable within this zone. On the other hand, the slope is high when the frequency is far from the optimal frequency; therefore, using a large step is more suitable to reduce the number of iterations needed to ensure a faster convergence time. The upper and lower limits of the step size for the proposed algorithm are given in the following equation: (step) min step (step) max. (14) The minimum and upper thresholds are set to 50 and 400 Hz, respectively. These lower and upper limits were chosen for a Liion battery cell by trial and error. These values can be adjusted for other batteries or to get different convergence speed or steady-state error. For the variable step-size algorithm, the step size is calculated according to (14) and (15) and using Table II M = 1 ΔV N Δf (15) where V is the terminal voltage of the battery (in Volts), f is the input current frequency (in Hz), ΔV/Δf is the slope of the voltage frequency curve, N is a constant, and M is a variable

4 HUSSEIN et al.: ONLINE FREQUENCY TRACKING ALGORITHM USING TERMINAL VOLTAGE SPECTROSCOPY 35 TABLE II VARIABLE STEP SIZE VALUES that is used to determine the step size. The value of N in (15) that gives the best results was found by trial and error. For a negative slope, the best value is N =7 10 9, whereas for a positive slope, the best value is N = Table II lists the step sizes used in the simulation in Section IV. A flowchart of the proposed algorithm is shown in Fig. 3. In Fig. 3, the process starts by setting an initial value of the optimal frequency and measuring the voltage at this frequency. Then, the frequency is incremented by 50 Hz and the voltage is measured again at the new frequency. The slope of the voltage frequency curve is calculated then, and based on the slope value, the step size is selected according to the criterion explained in (14) and (15) and Table II. Then, the frequency is updated and the new voltage is recorded again. This process is repeated until a change in the voltage slope sign is detected. A change in the slope sign indicates that the optimal frequency has converged to the true value. The process is terminated once convergence is achieved. C. Voltage Spectroscopy The use of impedance spectroscopy is an already established technique for determining the optimal frequency for ac charging [22], [23]. However, this method requires complex impedance calculations; therefore, in this paper, an alternative technique is proposed which is voltage spectroscopy. In this technique, the battery is charged using ac sinusoidal ripple current for a range of frequencies over short-time duration (e.g., 1000 s). The average of the battery terminal voltage is computed. The averaged voltage will have the lowest value due to the minimum internal voltage drop at the resonant frequency of the battery. This frequency will correspond to the optimal frequency using impedance spectroscopy. The voltage drop is the difference between the OCV and the terminal voltage of the cell. As the impedance increases, the internal voltage drop across it increases due to the increase in the voltage difference (ΔV = V OCV) as shown in Fig. 4. Here, we assume that the change in battery OCV is negligible or remains constant for a short time period. At the optimal frequency, where the cell battery impedance is the least, the voltage drop is also going to be the lowest. If the voltage drop is minimal at the optimal frequency, then the voltage measurement technique could be used for tracking the optimal point, in a similar manner how electrochemical impedance spectroscopy (EIS) is used for tracking. Therefore, the voltage spectrum could be an alternative to impedance spectroscopy. So, the aim of our research is to test and verify the voltage measuring technique for optimal frequency detection, which is expected to be an easier, less expensive, and less time-consuming technique than EIS and PLL control. Fig. 4. Filtered terminal voltage of a generic battery at optimal and non-optimal charging frequencies. Fig. 5. Top: 0.5-A peak-to-peak SRC charging current at 0.1 Hz used to charge a Li-ion battery cell. Bottom: measured terminal voltage of the Li-ion cell. The nonsinusoidal curve (bottom) indicates the smoothed voltage using a moving average filter with a time window of 10 s (the filtering function was applied twice over the same window). As shown in Fig. 4, the battery s terminal voltage at the optimal frequency (f = f opt ) is less than that at a different frequency (f f opt ). Hence, the charging efficiency (8) at f = f opt is higher than that at f f opt. The lower terminal voltage at the optimal frequency is due to the lower voltage drop across the battery impedance at f opt. Fig. 5 illustrates the concept of the moving average filter. The moving average filter is similar to a rolling window. It creates a series of averages of different subsets of the full data set. It is used to find the averaged terminal voltage of the battery when the input (charging) current is sinusoidal with dc shift. The width of the window must be adjusted according to the charging frequency. That is, the number of samples averaged in each cycle must follow the Nyquist theorem to ensure accurate sampling. IV. SIMULATION RESULTS In order to verify its effectiveness, the proposed algorithm was compared with two other algorithms with fixed step size of 50 and 250 Hz. For all the algorithms, the initial guess was set at 100 Hz. Fig. 6 shows the performance of the algorithms. A summary of the results obtained by running the algorithms on the six cells is given in Table III, where k and Δ are the iterations number and steady-state error in Hertz, respectively. Obviously, the proposed algorithm showed a more accurate performance compared to the fixed 250 Hz step-size algorithm, and a less convergence time compared to the fixed 50 Hz

5 36 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 7, NO. 1, JANUARY 2016 TABLE III SUMMARY OF RESULTS Fig kW TrueData-EIS impedance analyzer (Fuel Con). step-size algorithm. In terms of the charging efficiency and using (8), the proposed algorithm and the fixed 50 Hz step-size algorithm were able to achieve a higher efficiency compared to the fixed 250 Hz step-size algorithm. However, the proposed algorithm reached the peak-efficiency point in less time compared to the fixed 50 Hz step-size algorithm. One remark on the proposed algorithm is on the zero steadystate error that was achieved. The reason behind that is that the cell s terminal voltage was calculated at frequencies of 50 Hz increments, which is equal to the minimum step size. For the same reason, the fixed 50-Hz step-size algorithm showed zero steady-state error. However, if the terminal voltage is measured at more frequency points, the step size must be decreased for a zero steady-state error. V. EXPERIMENTAL VERIFICATION Fig. 6. Results of allocating the optimal charging frequency for cells 1 to 6. A 3.8-V 2600-mAh off-the-shelf Samsung Li-ion battery cell was used to verify the concept of voltage spectroscopy used in the proposed algorithm. First, the ac impedance spectrum of the cell was obtained using an ac impedance analyzer (TrueData- EIS FuelCon impedance analyzer shown in Fig. 7). The ac impedance spectrum of the Li-ion cell is shown in Fig. 8. The optimal frequency inferred from the spectrum is roughly around 1000 Hz, as at this frequency, the cell ac impedance has the minimum value of 118 mω. The cell is charged using a sinusoidal ripple charging (SRC) profile for a specified duration, and the battery terminal voltage over time is recorded. The battery charging equipment used for test is NuVant EZSTAT-pro battery tester as shown in Fig. 9.

6 HUSSEIN et al.: ONLINE FREQUENCY TRACKING ALGORITHM USING TERMINAL VOLTAGE SPECTROSCOPY 37 Fig. 8. AC impedance spectrum for the tested Li-ion battery. Fig. 9. (a) 3.8-V 2600-mAh Samsung Li-ion battery used in the test. (b) Battery charging test bench. Fig. 10. Li-ion charging test at 0.5-A peak-to-peak sinusoidal charging current. (a) Entire test. (b) Zoomed-in portion. The cell is charged using SRC at different frequencies. The cell s charging procedure includes the following steps. Step 1) The battery is discharged to the lower cutoff voltage using constant current. Step 2) The battery is discharged using constant voltage until the current becomes negligible (to ensure that the cell reached 0% SOC). Step 3) The battery is allowed to rest for 30 min. Step 4) The battery is charged using SRC for a certain time at a prespecified frequency. Step 5) The cell is allowed to rest until its voltage becomes completely stable. Step 6) Steps 1) to 5) are repeated at every test frequency. The charging procedure was performed at the following frequencies: 0.1, 10, 1000, 5000, and 8000 Hz. The test was performed twice at each test frequency with peak-to-peak charging currents of 0.5 A ( cos(2πft)) and 1 A ( cos(2πft)). The sinusoidal voltage waveforms at 0.5 and 1 A are filtered using a moving average filter over a 10-s period window as shown in Figs. 10 and 11, respectively. Table IV lists the measured terminal voltages of the Li-ion cell at each test frequency. The charging process was extended to 3000 s for the two different charging rates to check the performance of the proposed algorithm and the same voltage patterns were obtained. As shown in Figs. 10, 11 and Table IV, the optimum performance (minimum terminal voltage) occurred at the optimal frequency of 1000 Hz. In other words, the voltage drop inside the battery is minimized when the SRC charging is performed at the optimal frequency. Fig. 12 shows the voltage spectrum Fig. 11. Li-ion charging test at 1-A peak-to-peak sinusoidal charging current. (a) Entire test. (b) Zoomed-in portion. obtained for the Li-ion cell. This plot shows the battery terminal voltage as a function of sinusoidal charging frequency. To verify the validity of the proposed algorithm across different battery technologies, the proposed experiment was applied

7 38 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 7, NO. 1, JANUARY 2016 TABLE IV LI-ION BATTERY CHARGING RESULTS Fig. 14. Lead-acid charging test at 1-A peak-to-peak sinusoidal charging current. TABLE V LEAD-ACID BATTERY CHARGING RESULTS Fig. 12. Sinusoidal charging frequency versus battery terminal voltage using 0.5-A peak-to-peak and 1-A peak-to-peak currents for the Li-ion battery. Fig. 13. AC impedance spectrum for the tested lead-acid battery. to a lead-acid battery using the same charging procedure used for the Li-ion battery. A 2-V 8000-mAh off-the-shelf lead-acid battery was tested. Fig. 13 shows the ac impedance spectrum for this battery. The ac impedance spectrum shown in Fig. 13 indicates the minimum impedance magnitude of mω at 100 Hz. The voltage curves of a 1-A peak-to-peak charging tests are shown in Fig. 14. The following frequencies were used in the test: 1, 10, 100, 500, and 1000 Hz. The results in Fig. 14 show that the voltage curve at the optimal frequncy was the least compared to curves at other frequncies, which gives the conclusion that the proposed method can be also used with batteries other than Li ion such as lead acid. Table V shows that the terminal voltage measured at the end of 500 and 1500 s is consistently the lowest for optimal SRC charging at 100 Hz. The voltage spectrum for lead-acid battery is shown in Fig. 15. Additionally, the optimal SRC charging method was compared with the traditional CCCV method for both Li-ion and lead-acid batteries at the optimal frequencies found using the Fig. 15. Sinusoidal charging frequency versus battery terminal voltage using 1-A peak-to-peak current for the lead-acid battery. TABLE VI COMPARATIVE SUMMARY BETWEEN SRC AND CCCV CHARGING METHODS proposed algorithm. Each battery was charged to a certain voltage then discharged using constant current. The charge and discharge Ampere-hours were recorded. The results are summarized in Table VI. In Table VI, the effective energy stored in the battery is higher when the battery was charged using SRC method compared to CCCV method. In terms of charging efficiency, which was found by dividing the Ampere-hour Out by the Amperehour In, the charging efficiency for the Li-ion battery was

8 HUSSEIN et al.: ONLINE FREQUENCY TRACKING ALGORITHM USING TERMINAL VOLTAGE SPECTROSCOPY % using CCCV and 98.78% using SRC. For the lead-acid battery, the efficiency was 73.13% using CCCV and 82.59% using SRC. It shall be noted that the accurate results presented are subject to constraints such as the battery charging device accuracy and scan rate. The device used for charging has a slew rate of 0.2V/µs and a bandwidth of 8 khz. To ensure battery quality, brand new batteries were used and all tests were carried out at room temperature. VI. SUMMARY AND CONCLUSION Since the battery is a highly nonlinear system, and due to the time-variant nature of its internal parameters, tracking the change in these parameters using a battery model is very difficult. In addition, estimating these parameters using measurement equipment is expensive and not practical in real-time applications. In this paper, an online tracking algorithm has been proposed to allocate and track the optimal charging frequency without the use of predetermined battery model or measurement equipment. The proposed algorithm uses a variable step size in the iteration phase to improve the convergence time and accuracy. When evaluated on six Li-ion battery cells with unknown conditions, the proposed algorithm showed an improvement of 250%, 111%, 350%, 77%, 143%, and 83% in the convergence time compared to a 50-Hz fixed step-size algorithm, and an error reduction of 13%, 10%, 16%, 13%, 22%, and 17% compared to a 250-Hz fixed step-size algorithm. Additionally, the voltage spectroscopy technique has been experimentally verified on Li-ion and lead-acid batteries. The tested batteries showed consistently least terminal voltage when charged using SRC under optimal frequency. Also, it was shown that complex impedance calculations for tracking the optimal charging frequency can be avoided. However, it has been observed during testing that when the batteries are charged for long periods, the charging curves will not stay in the same order. To overcome this issue, the algorithm must be able to converge in a short time by having a good initial estimate for the frequency and well-tuned algorithm parameters. Also, the frequency must be tracked and updated continuously to guarantee accurate results. Finally, the proposed algorithm is expected to be capable to allocate and track the optimal frequency accurately and efficiently for any battery provided that the algorithm parameters that determine the step size (M and N) are selected carefully, and that a good initial guess for the optimal charging frequency is used. REFERENCES [1] A. A. Hussein, A. A. Fardoun, and S. S. Stephen, An on-line tracking algorithm for Li-ion batteries optimal charging frequency, presented at the IEEE Power Energy Soc. (PES) Gen. Meet., Denver, CO, USA, Jul , [2] L. Lu, X. Han, J. Li, J. Hua, and M Ouyang, A review on the key issues for lithium-ion battery management in electric vehicles, J. Power Sources, vol. 226, pp , [3] A. Väyrynen and J. Salminen, Lithium ion battery production, J. Chem. Thermodyn., vol. 46, pp , [4] R. C. Cope and Y. Podrazhansky, The art of battery charging, in Proc. 14th Battery Conf. Appl. Adv., Long Beach, CA, USA, Jan. 1999, pp [5] A. A. Hussein and I. Batarseh, A review of charging algorithms for nickel and lithium battery chargers, IEEE Trans. Veh. Technol., vol. 60, no. 3, pp , Mar [6] Y. L. Ke and Y. C. Chuang, A novel high-efficiency battery charger with a buck zero-voltage-switching resonant converter, IEEE Trans. Energy Convers., vol. 22, no. 4, pp , Dec [7] Y. S. Hwang, C. C. Wang, F. C. Yang, and J. J. Chen, New compact CMOS Li Ion battery charger using charge-pump technique for portable applications, IEEE Trans. Circuits Syst. I, Reg. Papers, vol. 54, no. 4, pp , Apr [8] J. J. Chen, F. C. Yang, C. C. Lai, Y. S. Hwang, and R. G. Lee, A high-efficiency multimode Li Ion battery charger with variable current source and controlling previous-stage supply voltage, IEEE Trans. Ind. Electron., vol. 56, no. 7, pp , Jul [9] B. D. Valle, C. T. Wentz, and R. Sarpeshkar, An area and power-efficient analog Li-ion battery charger circuit, IEEE Trans. Biomed. Circuits. Syst., vol. 5, no. 2, pp , Apr [10] B.-Y. Chen and Y. S. Lai, New digital-controlled technique for battery charger with constant current and voltage control without current feedback, IEEE Trans. Ind. Electron., vol. 59, no. 3, pp , Mar [11] Z. Jiang and R. A. Dougal, Synergetic control of power converters for pulse current charging of advanced batteries from a fuel cell power source, IEEE Trans. Power Electron., vol. 19, no. 4, pp , Jul [12] L. R. Chen, J. J. Chen, N. Y. Chiu, and J. Y. Han, Current-pumped battery charger, IEEE Trans. Ind. Electron., vol. 55, no. 6, pp , Jun [13] L. R. Chen, C. M. Young, N. Y. Chu, and C. S. Liu, Phase-locked bidirectional converter with pulse charge function for 42V/14V dual-voltage PowerNet, IEEE Trans. Ind. Electron., vol. 58, no. 5, pp , May [14] L. R. Chen, S. L. Wu, D. T. Shieh, and T. R. Chen, Sinusoidal-ripplecurrent charging strategy and optimal charging frequency study for Li- Ion batteries, IEEE Trans. Ind. Electron., vol. 60, no. 1, pp , Jan [15] L. R. Chen and C. H. Wu, Sinusoidal ripple current charging system with PLL function, in Proc. IEEE 23rd Int. Symp. Ind. Electron. (ISIE), 2014, pp [16] V. Srinivasan and C. Y. Wang, Analysis of electrochemical and thermal behavior of Li-Ion cells, J. Electrochem. Soc., vol. 150, no. 1, pp. A98 A106, [17] F. Huet, A review of impedance measurements for determination of the state-of-charge or state-of-health of secondary batteries, J. Power Sources, vol. 70, no. 1, pp , Jan [18] M. Coleman, C. K. Lee, C. Zhu, and W. G. Hurley, State-of-charge determination from EMF voltage estimation: Using impedance, terminal voltage, and current for lead-acid and lithium-ion batteries, IEEE Trans. Ind. Electron., vol. 54, no. 5, pp , Oct [19] W. Waag, S. Kabitz, and D. U. Sauer, Experimental investigation of the lithium-ion battery impedance characteristic at various conditions and aging states and its influence on the application, Appl. Energy, vol. 102, pp , [20] T. Momma, M. Matsunaga, D. Mukoyama, and T. Osaka, Ac impedance analysis of lithium ion battery under temperature control, J. Power Sources, vol. 216, pp , [21] F. Croce et al., An electrochemical impedance spectroscopic study of the transport properties of LiNi0.75Co0.25O2, Electrochem. Commun., vol. 1, no. 12, pp , Dec [22] D. Andrea, M. Meiler, K. Steiner, C. Wimmer, T. Soczka-Guth, and D. U. Sauer, Characterization of high-power lithium-ion batteries by electrochemical impedance spectroscopy. I. Experimental investigation, J. Power Sources, vol. 196, no. 12, pp , Jun. 15, [23] D. Andre, M. Meiler, K. Steiner, H. Walz, T. Soczka-Guth, and D. U. Sauer, Characterization of high-power lithium-ion batteries by electrochemical impedance spectroscopy. II: Modelling, J. Power Sources, vol. 196, no. 12, pp , Jun. 15, [24] J. Gomez, R. Nelson, E. E. Kalua, M. H. Weatherspoon, and J. P. Zheng, Equivalent circuit model parameters of a high-power Li-ion battery: Thermal and state of charge effects, J. Power Sources, vol. 196, no. 10, pp , May 15, 2011.

9 40 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 7, NO. 1, JANUARY 2016 Ala A. Hussein (S 07 M 12) received the B.S. degree in electrical engineering from Jordan University of Science and Technology, Irbid, Jordan, in 2005, and the M.S. and Ph.D. degrees in electrical engineering from the University of Central Florida, Orlando, FL, USA, in 2008 and 2011, respectively. From 2011 to 2012, he was with Coda Automotive (previously EnergyCS), Inc., Los Angeles, CA, USA, where he led the company s efforts in the developing advanced control techniques to improve the reliability of electric vehicles batteries. In 2012, he joined as an Assistant Professor with the Department of Electrical Engineering, United Arab Emirates University, Al Ain, UAE. He has over nine years of combined experience in research, teaching, and industry in the USA and UAE. He has extensive experience in battery charging, modeling, and control. He is the author and coauthor of over 20 technical papers in refereed journals and conference proceedings. He is a Periodic Reviewer of a number of international journals and conferences. Samantha S. Stephen received the B.E. (Hons.) degree from Birla Institute of Science and Technology, Pilani, India, in 2011, and the M.S. degree from the University of Sheffield, Sheffield, U.K., in 2014, both in electronics and electrical engineering. She is currently working as a Research Assistant with the Department of Electrical Engineering, United Arab Emirates University, Al Ain, UAE. Her research interests include battery modeling and renewable energy related technology. Abbas A. Fardoun (M 90 SM 05) was born in Lebanon. He received the B.S. degree in electrical engineering from the University of Houston, Houston, TX, USA, in 1988, and the M.S. and Ph.D. degrees in electrical engineering from the University of Colorado, Boulder, CO, USA, in 1990 and 1994, respectively. He was with the Advanced Energy Inc., Fort Collins, CO, USA, from 1994 to 1996, where he was involved with high-frequency power supply design. From 1996 to 1998, he was with Delphi where he worked on the development of sinusoidal drive for electrical power steering applications. From 1998 to 2006, he was with TRW Automotive, Livonia, MI, USA, working on various aspects of electrical power steering development. Since 2006, he has been with the Department of Electrical Engineering, United Arab Emirates University (UAEU), Al Ain, UAE, where he is currently a Full Professor and Director of the Renewable Energy Laboratory. Dr. Fardoun was the recipient of the Hariri Foundation Distinguished Graduate Award in He holds seven awarded patents.