IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 28, NO.

7, JULY 2013

3195

Adaptive Step Size With Adaptive-PerturbationFrequency Digital MPPT Controller for a Single-Sensor Photovoltaic Solar System
Yuncong Jiang, Student Member, IEEE, Jaber A. Abu Qahouq, Senior Member, IEEE, and Tim A. Haskew, Senior Member, IEEE

AbstractThis paper presents a load-current-based maximum power point tracking (MPPT) digital controller with an adaptivestep-size and adaptive-perturbation-frequency algorithm. Only one sensor is needed in the controller circuitry since the MPPT controller is only utilizing the load current information. By utilizing a variable step-size algorithm, the speed, accuracy, and efciency of the PV system MPPT are improved when compared to the xed step-size load-current-based algorithm. Furthermore, the proposed adaptive algorithm utilizes a novel variable perturbation frequency scheme which further improves the controller speed. The concept and operation of the load-current adaptive-step-size and adaptive-perturbation-frequency MPPT controller are presented, analyzed, and veried by results obtained from a proof-of-concept experimental prototype. Index TermsAdaptive-perturbation-frequency perturb and observe (P&O) algorithm, adaptive step size, dcdc converters, maximum power point tracking (MPPT), photovoltaic (PV), solar energy.

I. INTRODUCTION

HOTOVOLTAIC (PV) panels are used to convert solar energy into electric power. The output characteristics of a solar PV panel are dependent on operating conditions such as irradiance level and surrounding temperature [1][4]. Maximum power points (MPPs) exist on the PV panel characteristic curves where the output power from the solar panel is maximum [5][8]. Maximum power point tracking (MPPT) algorithms and techniques such as perturb and observe (P&O) algorithm [9][13], incremental conductance (InCond) algorithm [14][16], ripple correlation control (RCC) algorithm [17], fractional voltage/current MPPT method [18] and neural-network (NN)-based MPPT control [19][21] have been developed to extract the maximum power from the PV panel. The P&O method, which locates the MPP using the slope of the P V character-

Manuscript received April 30, 2012; revised August 7, 2012; accepted September 14, 2012. Date of current version December 24, 2012. This work was supported in part by the U.S. Department of Agriculture (USDA) under USDA Project #58-3148-0-204. Any opinions, ndings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reect the views of the funding agency. Recommended for publication by Associate Editor T. Shimizu. The authors are with the Department of Electrical and Computer Engineering, University of Alabama, Tuscaloosa, AL 35487 USA (e-mail: yjiang7@crimson.ua.edu; jaberq@eng.ua.edu; thaskew@eng.ua.edu). Color versions of one or more of the gures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identier 10.1109/TPEL.2012.2220158

istics curve, is widely used due to its simplicity and ease of implementation. A disadvantage of the P&O algorithm is that the PV panel operation points oscillate around the MPP which causes energy loss. InCond algorithms overcome the drawbacks of P&O algorithms by eliminating the oscillations around the MPPs. However, since the InCond MPPT algorithm needs realtime calculation of the slope of the PV panel power curve, it is more complicated to be implemented in controller compared to the P&O algorithm. The RCC MPPT algorithm utilizes the derivatives of the power converters voltage and current ripples to determine the position of the PV panel operating point. One drawback of this method is that if the power converters switching frequency varies, it is required to redesign the highpass lter circuit which is used to obtain time derivatives of PV panel voltage and current. The fractional voltage/current method sets the optimal voltage/current reference as a fraction of the PV solar panels open-circuit voltage or short-circuit current, and therefore, it does not track the real MPP. Although this method has an acceptable tracking performance under steadystate conditions, it might fail to converge to new MPP under transient conditions. The NN-based MPPT controller improves the tracking efciency of the system by utilizing a multilayer control structure; however, this method involves computational iterations and increases the calculation load of the controller. All the MPPT methods discussed earlier require the sensing of PV panel voltage and current and need multiplication function to obtain the PV panel power values which increase the size and the power consumption of the controller. The load-current-based MPPT method with the xed stepsize (FXS) perturbation P&O algorithm has been proposed [22][24] to realize MPPT functionality by sensing only the load current (this method is discussed further in the next section) which eliminates the need for a multiplier that is required to obtain the power value in the conventional power-based MPPT methods. Adaptive-perturbation-step-size algorithms [25][27] are studied to provide fast dynamic convergence speed and high steady-state tracking efciency. Different from [27], in which a xed scaling factor and a xed MPPT frequency algorithm is used, Mei et al. [25] utilize an adaptive scaling factor and the xed frequency MPPT algorithm to optimize the controller speed during transient. However, the algorithm in [27] becomes more complicated: the algorithm requires the knowledge of the location of the PV panel operation point and the controller is switching between adaptive duty cycle control and xed duty cycle control.

0885-8993/$31.00 2012 IEEE

3196

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 28, NO. 7, JULY 2013

The tradeoff between fast dynamic response and highefciency steady-state operation with lower oscillations around the MPP for the load-current-based MPPT method is addressed in this paper by developing and using an adaptive-step-size and perturbation-frequency MPPT algorithm. Unlike commonly used variable step-size MPPT algorithms [25][27] which perturb the power stage duty cycle or voltage reference at xed MPPT frequency, the proposed load-current adaptive step-size and perturbation frequency (LCASF) MPPT algorithm adaptively generates large perturbations with longer perturbation periods during transients to achieve fast and stable dynamic response and generates small perturbations with shorter perturbation periods that results in high efciency and lower oscillations during steady-state operation. Generally, in a PV system with MPPT control, large perturbation requires longer settling time after perturbation is triggered and small perturbations require shorter settling time. In previous works, which either implement the FXS MPPT algorithm or an adaptive-step-size MPPT algorithm, the period of perturbation is xed. This xed perturbation time period is chosen to ensure that the system has enough time to settle down when the largest perturbation is triggered. However, this results in longer MPPT controller response time when the operating point is close to MPP. This issue is addressed in the proposed LCASF MPPT algorithm by utilizing an adaptive-perturbationfrequency scheme with higher perturbation frequency when the perturbation is smaller, and vice versa. Digital controllers are increasingly being used in a renewable energy system control because of their ability to perform advanced control algorithms among other advantages such as easy to be recongured and upgraded [28][34]. Therefore, the LCASF MPPT controller is realized by a digital controller. This digital controller is implemented in this paper by using a microcontroller (MCU), but can be also implemented by eldprogrammable gate array for examples. Section II discusses the proposed LCASF MPPT controller algorithm and its theoretical background. The proof-of-concept experimental prototype results are presented in Section III to evaluate the MPPT controller performance under steady-state and transient conditions. Section IV compares the proposed MPPT control algorithm with other MPPT algorithms. Conclusion is given in Section V. II. LCASF MPPT ALGORITHM Fig. 1 shows a PV solar system block diagram with the proposed LCASF MPPT controller. The power conversion process from the source (PV panel) to the load (battery load or resistive load) interfaced through a dcdc converter with efciency equal to . The dcdc converter regulates the voltage and current of the solar panel and thus it regulates the output power. The MPPT controller keeps adjusting the duty cycle of the power converter to reach the MPP of the solar panel. A. Review and Analysis of Load Current MPPT Concept In the conventional power-based P&O MPPT algorithm, the derivative of power to voltage dP/dV of a PV panel is used as a

Fig. 1. Block diagram of a PV solar power system with load current MPPT control.

tracking parameter. The MPP is reached when dPin =0 dVin (1)

where Pin is the power from the PV panel. For the system with a battery load, the load voltage VL is constant. The input power Pin can be expressed as a function of load current Io assuming that a buck converter topology is used for the power stage Pin = Io VL / Vin = VL /D. Substituting (2) and (3) into (1) yields dPin VL dIo VL dIo dD VL dIo = = = dVin dVin dD dVin dD =
2 dIo VL 2 dD . Vin

(2) (3)

VL 2 Vin (4)

Thus, at MPP (V = VM PP ), when dPin /dVin = 0, the following is true: dIo /dVin = 0 at D = DM PP = VL /VM PP . (5)

Based on (5), the MPP of the PV system can be tracked by tracking the sign of dIo /dD. For the system with a resistive load R, Pin can be expressed by
2 Pin = Io R/

(6) (7)

dPin R dIo 2Io R dIo = 2Io = 2 dD . dVin dVin Vin

At MPP, (7) is equal to zero, which yields also dIout /dVin = 0 at D = DM PP . The tracking of zero slope at MPP is valid in the system with a resistive load. The relationship between input power and output current with power stage duty cycle is illustrated in Fig. 2. B. Proposed LCASF MPPT Algorithm In the conventional load-current-based P&O algorithm, the duty cycle perturbation step size D is a xed value. During steady-state MPPT operation, small D reduces the power

JIANG et al.: ADAPTIVE STEP SIZE WITH ADAPTIVE-PERTURBATION-FREQUENCY DIGITAL MPPT CONTROLLER

3197

1) Adaptive Perturbation Step Size: In the digital controller algorithm of Fig. 3, during the current algorithm perturbation cycle/iteration, the power converter duty cycle and load current are denoted by D(k ) and I (k ), respectively. The duty cycle and load current from the previous perturbation cycle/iteration are denoted by D(k 1) and I (k 1), respectively. The changes in the load current and duty cycle from one cycle to the next are dened as Idi = I (k ) I (k 1)
Fig. 2. Relationship between input power and output current with power converter duty cycle.

(8) (9)

Ddi = D(k ) D(k 1).

If the signs of Idi and Ddi are the same, the duty cycle is incremented by D and the variable X is set to 1 in order for the algorithm to remember the last perturbation direction of the duty cycle. If the signs of Idi and Ddi are opposite, the duty cycle is decremented by D and the variable X is set to 0 in order for the algorithm to remember the last perturbation direction of the duty cycles. As shown in Fig. 3, the value of X is used in the case when Idi = 0, which could happen because the analog-to-digital converter (ADC)s resolution is not high enough to see the change in the load current change as a result of the last duty cycle perturbation. In this case, the duty cycles are perturbed in the same direction as in the past iteration and the current values are not swapped. This step is equivalent to increasing the duty cycles perturbation step size. The duty cycle value is always compared and limited to a minimum value Dm in and a maximum value Dm ax . The duty cycle for the next algorithm cycle is given by D(k + 1) = D(k ) C Idi = D(k ) D Ddi (10)

Fig. 3.

LCASF MPPT algorithm owchart.

losses caused by the oscillations around the MPP. During transient MPPT operation, larger D is preferred for faster convergence to the new MPP. Variable step-size algorithms are usually developed in order to achieve tradeoff between the speed and the accuracy of the tracking [22][25]. The proposed LCASF MPPT digital controller algorithm owchart is shown in Fig. 3. The control strategy of this algorithm is to continuously adjust the duty cycle perturbation values and adjust the perturbation frequency while observing the load current Io . The presented algorithm mainly consists of two schemes: the adaptive determination of the perturbation values D and the adaptive determination of perturbation periods T . After D and T values are obtained, the duty cycle of the power stage is perturbed by D and after waiting T period of time, the MPPT controller starts the next perturbation.

where C is a scaling factor constant that can be used to tune the adaptive perturbation step sizes in the design. The MPPT variable step-size operation using load current is realized by (10). The duty cycle perturbation step size decreases as the load current versus duty cycle curve (see Fig. 2) slope decreases. This is desired because the decrease in the curve slope indicates approaching the MPP and the steady-state operation. The closer the PV operating point from the new MPP, the smaller the step size and the farther the PV operating point from the MPP, the larger the step size. Equation (10) results in smaller perturbation step size in steady state and therefore less oscillations and higher efciency and larger perturbation step sizes during dynamic operation and therefore faster convergence speed to the MPP. Equation (11) sets the rules for calculating the scaling factor C dI C Dm in dD D m i n (11) dI C Dm ax . dD D m a x The minimum limit on the duty cycle step size Dm in is set in (12) for the following reasons. The rst is to prevent D from approaching zero which will cause Idi /D , and hence will cause the duty cycle D to saturate. The other reason

3198

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 28, NO. 7, JULY 2013

5%, a change is applied to the duty cycle from 20% to 25% and the changes in the current is recorded in order to calculate the value of C in (13). The change in current dI based on Dm ax perturbation recorded from Fig. 4(b) is 0.3 A dI dD

C = Dm ax

= 0.05/(0.3/0.05) = 0.00833.
Dm ax

(13) The value of Dm in is selected based on 1) the digital pulsewidth modulator hardware resolution and 2) the current sensing resolution. The value of Dm ax is selected such that the output voltage deviation is an acceptable value for the specic hardware/components which can be obtained by theoretical analysis or by experimental measurements. 2) Adaptive Perturbation Frequency: For variable step-size MPPT algorithms, as well as FXS MPPT algorithms, the perturbation time period T is a xed value. For examples, T = 0.1 s in [25] and 0.25 s in [27]. By far, almost no literature has addressed the MPPT frequency for the adaptive MPPT controller. One common rule is that large perturbation time periods are used to ensure steady-state operation before the next perturbation before taking the new current and/or voltage measurements in order to avoid oscillations and probably algorithm divergence. The perturbation time period is usually selected for the worst case duty cycle perturbation (Dm ax ) in adaptive-step-size algorithms. This worst case selection results in slower dynamic response of the MPPT controller. The settling time for this power stage design can be obtained theoretically from the output current-to-duty cycle transfer function Gid (s), as in (14), of the power stage and can be veried experimentally. Different power converters have different Gid (s). For a dcdc buck converter used in this paper (1 + Resr Co s) Vpv Io 1 + (L Vo /Io + Resr Co ) s + L Co s2 Vo (14)

Fig. 4. Sweep of power converter duty cycle versus (a) PV panel power and (b) load current.

is the digital controller hardware resolution limitation (ADCs resolution limitation). The guideline for Dm in value selection is that the minimum current change caused by Dm in should be large enough for the ADC to detect. Assuming an N -bit ADC and the maximum voltage of the ADC channel is VADC [I (D + Dm in ) I (D)] G VADC /(2N ) (12)

Gid =

where G is the gain of current sensor which is equal to 1 in this paper. Dm in will vary with different solar panel characteristics and different ADCs. In this paper, Dm in is calculated as 1% for a 12-bit ADC with VADC of 3 V and 2.5-A maximum load current. In Section III, this Dm in value usually results in a minimum change of 10 mV from the current sensor. The change in current dI as a function of the duty cycle can be measured theoretically, by simulation, and/or experimentally. Fig. 4 shows plots of PV solar panel power and load current as a function of the duty cycle value of the power stage. Fig. 4(b) plot is used to calculate the value of C . Because it is desired to have larger duty cycle step size values when the system operates far from the MPP and smaller duty cycle step size values when the system operates closer to the MPP, C value is obtained using the portion of Fig. 4(b) curve which is far from the MPP. Note that load current change becomes smaller (because of the bell-shaped curve) as the system operation approaches the MPP point, and therefore, D becomes smaller (for the xed C value). Therefore, since for this paper system design Dm ax =

where Vpv is the PV panel voltage which is the power stage input voltage, Co is the dcdc buck converter output capacitor, Lo is the dcdc buck converter output inductor, Resr is the output capacitor equivalent series resistance, Vo is the dcdc buck converter output voltage, and Io is the dcdc buck converter load current. The input to the aforementioned transfer function is D (duty cycle step size) and the output is the load current. In this paper, an Agilent solar array simulator is used to simulate the PV solar panel which has the following specications: Vo c = 15 V, Vm p = 10 V, Isc = 1.3 A, Im p = 1 A. The power stage design parameters are as follows: Co is 220 H, Resr is 0.01 , and Lo is 100 H. Different step sizes D result in different output voltage settling times that determine the adaptive perturbation time period. Therefore, the transfer function in (15) can be used to determine the settling time theoretically. One practical method to nd the settling time from (14), as an alternative to nding inverse Laplace transform to obtain the time-domain function, is to use a software like Simulink/MATLAB software

JIANG et al.: ADAPTIVE STEP SIZE WITH ADAPTIVE-PERTURBATION-FREQUENCY DIGITAL MPPT CONTROLLER

3199

TABLE I SYSTEM SETTLING TIME WITH VARIOUS D

Fig. 6.

Illustration of the experimental prototype. TABLE II MPPT ALGORITHMS PARAMETERS

Fig. 5.

MPPT perturbation time periods versus perturbation step sizes.

package. In such software package, the input to the transfer function block is a D step function and the output is Io . Table I shows the system settling time obtained from simulation and experimental obtained for different step sizes for this paper power stage design with 10-V input from PV panel, 5-V load voltage, and 4-A load current. For safety concern to avoid instability, an additional 20% safety margin is added to the experiment results. The combination of the experiment test and simulation test leads to the nal selection of MPPT frequency in Fig. 5, a plot of the PV system settling time versus different perturbation step sizes obtained from the experimental prototype. In the proposed LCASF algorithm, the perturbation period is varied as a function of the perturbation step size (which is also variable) as shown in Tm ppt = f (|D|). (15)

The relationship between MPPT perturbation time periods and perturbation step sizes can be approximated by a linear or a nonlinear function. Equation (16) gives the linear approximation based on Fig. 5 data points, where the line slope is 1.36444 and T -intercept is 4.9753 Tm ppt = f (|D|) = 1.3644 |D| + 4.9753. III. PROOF-OF-CONCEPT EXPERIMENTAL RESULTS A proof of experimental prototype is designed and built in the laboratory for testing and verication of the LCASF MPPT controller as illustrated in Fig. 6. The power converter topology used is a synchronous dcdc buck converter, operating in continuous conduction mode with 100-kHz switching frequency and PWM control. PV panel ratings and power converter parameters are identical with the parameters in the system response time analysis in the last section. (16)

The LCASF MPPT control algorithm is implemented using TMSF28335 MCU from Texas Instruments, which has a 12-bit ADC. The ADC clock frequency is set at 25 MHz. The load current value is sensed by a current sensor and fed to the MCU ADC port. A rolling-average algorithm is used to average 100 samples of the load current before the next perturbation and this average value is used in the algorithm as I(k). Electronic load (Chroma 63030) is used for two types of loads: a battery load (5 V) and a resistive load (2 ). As is the case in all MPPT controllers, it is assumed that the PV system load(s) is(are) able to absorb all the power available from the PV panel(s) in order to track the MPP. If the load can only absorbed part of the PV panel power that is available at the MPP, the PV panel will be able to only supply a power that the load is able to absorb in this case. In addition to the proposed LCASF algorithm, the FXS algorithm and the variable step-size algorithm without adapting the perturbation period (LCA) are tested. Table I shows the perturbation step size(s) and perturbation period(s) used in each of the algorithms. In the second row of Table II, the boundary of cycle perturbation step size is from 1% to 5% for an adaptivestep-size algorithm. The FXS algorithm step size is 1% and 5%, respectively. In the third row of Table I, 5 (ms) corresponds to the MPPT period when lower bound of duty cycle perturbation (1%) is triggered. The total of 15 (ms) corresponds to the MPPT period when upper bound of duty cycle perturbation (5%) is triggered. The LCASF MPPT period will be adaptively varied within the range of 515 ms, determined by (16). The last row in the table shows the corresponding gure part (see for Figs. 7 and 8) for the experimental results presented next. Fig. 7 shows waveforms of PV panel voltage, PV panel current, and load current under PV solar panel transient conditions with a battery load. For the PV solar input transient test, panel

3200

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 28, NO. 7, JULY 2013

Fig. 7. PV panel voltage, current, and load current waveforms under input transient response of controller with battery load for (a) LCASF algorithm, (b) LCA algorithm, (c) 1% FXS algorithm, and (d) 5% FXS algorithm.

Fig. 8. PV panel voltage, PV panel current, and load current waveforms under input transient response of controller with resistive load for (a) LCASF algorithm, (b) LCA algorithm, (c) 1% FXS algorithm, and (d) 5% FXS algorithm.

JIANG et al.: ADAPTIVE STEP SIZE WITH ADAPTIVE-PERTURBATION-FREQUENCY DIGITAL MPPT CONTROLLER

3201

TABLE III PERFORMANCE COMPARISON

parameters are switched from the initial values (Vo c = 15 V, Vm p = 10 V, Isc = 1.3 A, Im p = 1 A) to new values (Vo c = 12 V, Vm p = 9 V, Isc = 1.1 A, Im p = 0.9 A). From Fig. 7(a) (d), the solar panel was operating around the MPP (10 V, 0.9 A) before the simulated irradiance transient was triggered. As observed from the gure, the PV panel current dropped sharply at the beginning of the transient indicating that the operating point of the PV panel was far away from the MPP. In all the gures, the MPPT controller successfully adjusts the duty cycle of the power converter and brings the operating point of the PV panel closer to its MPP. Fig. 8 shows the MPPT controller response under the same transient conditions, but with a resistive load connected. Obviously, larger irradiance change results in larger change in the current dI which based on (10) will result in larger D (limited by the controller to Dm ax ). This will cause an initial large perturbation step size which will eventually get smaller (limited by the controller to Dm in ) as the new MPP is approached since dI approaches zero at/near the MPP. Table III summarizes the performance parameters of the MPPT algorithms tested. These parameters are tracking speed and tracking efciency, where the tracking efciency is dened as in = PPV /PM PP . (17)

Fig. 9. Comparisons of numbers of perturbation cycles of LCASF algorithm with 1% xed-step algorithm and 5% xed-step algorithm.

Fig. 10. PV panel voltage, current, load current, and voltage waveforms under load voltage transient with LCASF controller.

PPV is the actual power generated by the PV panel and PM PP is the ideal maximum power of the PV panel. The FXS with 5% perturbation step size has the fastest tracking time under transient with poor steady-state tracking efciency. The 1% FXS controller has slower response to transient but yields good tracking efciency under steady-state conditions. Both LCA and LCASF algorithms yield more than 95% tracking efciency during steady-state conditions. As observed from these results, for instance under a battery load, the LCA algorithm provides a tradeoff such that it has faster convergence speed (110 ms) to the new MPP than the xed 1% perturbation algorithm (200 ms) and higher tracking efciency (96%) than the FXS algorithm with large step size (88%). When the adaptive perturbation period scheme is activated together with the adaptive duty cycle perturbation scheme (LCASF algorithm), the convergence speed is further and signicantly improved (3050 ms faster). The LCASF MPPT controller algorithm results in convergence speed that is much closer

to the speed with 5% xed duty cycle perturbation step size while achieving the smaller oscillations and better tracking efciency with 1% xed duty cycle perturbation. The duty cycle values progression for the results of Fig. 7 is plotted in Fig. 9. Fig. 9 shows the comparison between the LCASF algorithm with the FXS (1% and 5%) algorithm. LCASF controller tracks around the same speed as the 5% FXS controller while keeps small oscillations as in 1% FXS after reaching steady-state conditions. In Fig. 10, a battery load voltage change is triggered in the system to test the performance of the MPPT controller. The battery load voltage is stepped up from 4.5 to 5 V. The change in battery voltage leads to a step down in load current before the LCASF controller adjusted the duty cycle in order to converge to the new MPP. The PV solar cell or panel power can range from few watts to much higher than this (hundreds and thousands of watts). This papers proof-of-concept experimental prototype uses a PV solar panel with relatively low power to validate the proposed MPPT controller. The performance of the proposed controller is independent of the PV panel rating, and therefore, it can be used in higher power PV systems.

3202

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 28, NO. 7, JULY 2013

IV. ADDITIONAL COMPARISONS WITH OTHER MPPT ALGORITHMS The following is a summary between the proposed LCASF MPPT controller algorithm and the other MPPT algorithms discussed in Section I: 1) General comparison between the proposed LCASF MPPT control and the other MPPT algorithms [9][27] in terms of MPPT tracking speed: Unlike the LCASF MPPT controller, all the other existing MPPT controllers that are discussed in the literature use a xed MPPT perturbation period, even when they utilize variable duty cycle (or references voltage) perturbation step size, which results in signicantly longer MPP tracking time because this xed perturbation period is set long enough for the case when the duty cycle perturbation step size is largest. However, the LCASF MPPT controller algorithm adapts the MPPT perturbation period to the duty cycle perturbation step size such that the MPPT perturbation period is smaller when the duty cycle perturbation step size is smaller and vice versa. For example, if a xed 15-ms MPPT perturbation period is used and 13 duty cycle steps are needed (regardless if they are xed or variable step sizes) in order to reach the new MPP, approximately 200-ms tracking time would be needed. However, when the MPPT perturbation period is varied with the duty cycle perturbation step size, the MPPT tracking time will be reduced to, for example, about half (80 ms as can be observed from Table III). This numerical example comparison assumes that the smallest allowed duty cycle perturbation step size is the same for all algorithms (1%). It should be noted again here that while the MPPT perturbation period affects the MPP tracking time/speed for the same duty cycle perturbation step size, the duty cycle perturbation step size affects the MPP tracking efciency. 2) Additional comparison between the proposed LCASF MPPT control algorithm and the conventional P&O control algorithm [9][13]. a) The conventional P&O MPPT control algorithm requires sensing both the PV panel voltage and current and requires a multiplier in order to obtain the PV panel power value. The proposed LCASF MPPT controller requires sensing only the load current (one sensor compared to two) and does not require a multiplier which results in hardware size, cost, and controller power consumption. b) In terms of tracking efciency, the conventional P&O MPPT algorithm and the LCASF MPPT algorithm have the same MPP tracking efciency (see experimental results also) when the smallest allowed the duty cycle perturbation step size is the same in both algorithms (e.g., 1%). However, if the step size in the case of the conventional P&O MPPT algorithm is made larger (e.g., 45%) in order to match the LCASF MPPT algorithm tracking speed, the LCASF MPPT controller will have signicantly higher MPP tracking efciency

(e.g., 97% compared to 88% or 82% as shown in Table III) for the same controller resolution. Therefore, the proposed LCASF MPPT control algorithm provides signicantly improved tradeoff between MPP tracking speed and tracking efciency compared to the conventional P&O MPPT control algorithm while using hardware with less cost and size. 3) Additional comparison between the proposed LCASF MPPT control algorithm and the conventional InCond MPPT algorithm [14][16]. a) The conventional InCond MPPT algorithm requires sensing both the PV panel voltage and current and requires a multiplier in order to obtain the PV panel power value. The proposed LCASF MPPT controller requires sensing only the load current (one sensor compared to two) and does not require a multiplier which results in hardware size, cost, and controller power consumption. Moreover, the InCond MPPT algorithm involves more complicated calculations (especially with variable step size) compared to both the proposed LCASF MPPT algorithm and the conventional P&O MPPT algorithm, which means that the proposed LCASF MPPT controller has even lower hardware processing power compared to the InCond MPPT algorithm. b) The tracking speed of the LCASF MPPT controller is faster than it is for the InCond MPPT algorithm (duty cycle perturbation step-size step) because it utilizes the proposed variable MPPT perturbation period as discussed earlier in point 1. The MPP tracking efciency for both algorithms is expected to be comparable when the minimum duty cycle perturbation step-size of both controllers is the same. Moreover, the same can be said on the MPP tracking speed comparison between the LCASF MPPT algorithm and the InCond MPPT algorithm as discussed earlier in point 2(b) for the conventional P&O MPPT algorithm. 4) Additional comparison between the proposed LCASF MPPT control algorithm and the conventional RCC MPPT algorithm [17]. a) The RCC MPPT algorithm is similar to the P&O MPPT algorithm except that it is executed with the same frequency as the switching frequency of its MPPT dcdc power converter (the MPPT perturbation frequency is equal to the power converter switching frequency). Therefore, the proposed RCC MPPT implementations in the literature are only analog controller implementations with xed duty cycle perturbation step sizes. Moreover, higher duty cycle perturbation frequency requires a dcdc power converter with higher switching frequency. The unsuitability of using digital control in the case of the RCC MPPT algorithm is a disadvantage because the ability to easily update the MPPT controller (e.g., when PV panel changes and power

JIANG et al.: ADAPTIVE STEP SIZE WITH ADAPTIVE-PERTURBATION-FREQUENCY DIGITAL MPPT CONTROLLER

3203

when the converter changes) is lost and because the advantage of the ability to easily implement system communication protocols is also lost. b) Based on the above, unlike the LCASF MPPT algorithm, the RCC MPPT algorithm requires a highspeed or bandwidth voltage sensor, a high-speed current sensor, and a high-speed multiplier for power calculations. Especially, if the RCC MPPT algorithm needs to be implemented in a digital controller, its size, cost, and controller power consumption would be signicantly higher than the LCASF MPPT controller and the conventional P&O MPPT controller. c) Even though the MPPT perturbation frequency is usually higher in the RCC MPPT algorithm compared to the LCASF MPPT algorithm, during load transients and PV solar irradiance transients, the RCC MPPT controller will not be able to converge to the new MPP until the dcdc power converter reaches a new steady state (i.e., depends on the power converter settling time or the overall PV system settling time), which limits the actual MPP tracking speed. Since the proposed LCASF MPPT algorithm uses variable MPP perturbation period, which is equal to the settling time, the RCC MPPT algorithm tracking speed is not necessarily faster than the LCASF MPPT algorithm tracking speed (depends on the design) but it is faster than the MPP tracking speed of the conventional P&O MPPT algorithm. In summary, the RCC MPPT algorithm requires signicantly much more hardware resources in order to have a chance to achieve faster MPP tracking speed than the LCASF MPPT algorithm. d) The MPP tracking efciency of the LCASF MPPT controller and the LCASF MPPT controller is expected to be comparable with appropriate design. Note that the LCASF MPPT controller MPP tracking efciency can be increased from the 97% (e.g., to 99%) obtained from this papers proof-ofconcept experimental prototype results by increasing the ADCs resolution and the minimum duty cycle perturbation step size. 5) Additional comparison between the proposed LCASF MPPT control algorithm and the conventional fractional voltage and fractional current method [18]. While these two methods are simple and need one voltage or current sensor, they do not track the actual MPP in the PV system (they provide an approximation to the MPP, which implies low MPP tracking efciency under variable operating conditions). The LCASF MPPT controller tracks the real MPP with high MPP tracking efciency by also using one current sensor. 6) Additional comparison between the proposed LCASF MPPT control algorithm and the NN-based MPPT algorithm [19][21]. a) The NN MPPT algorithm is far more complicated than any other MPPT control algorithm discussed

in this paper. The NN MPPT algorithm method builds a nonlinear solar panel model and calculates the MPP of the solar panel under each operating condition. The NN MPPT method requires sensing irradiance level, temperature, panel current, and voltage. Thus, the NN MPPT method requires more hardware resources, especially sensors and processing power compared to the most MPPT algorithms including the LCASF MPPT control algorithm and therefore results in much higher hardware cost, size, and power consumption. The algorithm itself requires tens to hundreds of nodes to perform iterations. While the research continues to be open for better and simpler NN MPPT algorithms, as of this paper date such NN MPPT methods are not practically used. b) The LCASF MPPT control algorithms MPP tracking speed is faster than that of the NN-based MPPT algorithms mainly because the NN-based MPPT algorithms use a xed algorithm update period, 100 Hz, for example, as in [21]. c) While the NN MPPT algorithm yields no oscillations around the MPP because the MPP is calculated, the reported MPP tracking efciency is in the range of 94%97% [20], which is not higher than the proposed LCASF MPPT control algorithm (97% in this papers proof-of-concept experimental prototype results). Under varying irradiance conditions when the NN MPPT algorithm is used, the calculated training model, which is based on historical and real-time sensed data points, might have errors that result in degrading the MPP tracking efciency. From the aforementioned comparison, it can be concluded that the proposed LCASF MPPT controller when compared with the other MPPT algorithms results in the best tradeoff between MPP tracking speed, MPP tracking efciency, cost, size, and controller power consumption. This is because a new variable MPPT perturbation period method is utilized with a variable duty cycle perturbation step-size function, with only one sensor, no multipliers, and relatively a simple algorithm.

V. CONCLUSION This paper presented a load-current-based variable-step-size and variable-perturbation-frequency MPPT digital controller. In addition to utilizing a function to adapt the duty cycle perturbation, the proposed MPPT controller adapts its perturbation frequency as a function of the variable duty cycle perturbation value. As the results presented in this paper showed, the duty cycle adaptive-step-size scheme used in the proposed MPPT controller yields a good tradeoff between the convergence speed and tracking efciency compared to the FXS algorithm. Furthermore, the novel adaptive perturbation frequency scheme used in the proposed controller results in faster convergence speed compared to existing adaptive-step-size algorithms. The

3204

IEEE TRANSACTIONS ON POWER ELECTRONICS, VOL. 28, NO. 7, JULY 2013

proposed adaptive perturbation frequency scheme could also be used with other MPPT algorithms.

REFERENCES
[1] H. S.-H. Chung, K. K. Tse, S. Y. R. Hui, C. M. Mok, and M. T. Ho, A novel maximum power point tracking technique for solar panels using a SEPIC or Cuk converter, IEEE Trans. Power Electron., vol. 18, no. 3, pp. 717724, May 2003. [2] A. I. Bratcu, I. Munteanu, S. Bacha, D. Picault, and B. Raison, Cascaded DCDC converter photovoltaic systems: Power optimization issues, IEEE Trans. Ind. Electron., vol. 58, no. 2, pp. 403411, Feb. 2011. [3] Y. H. Ji, D. Y. Jung, J. G. Kim, J. H. Kim, T. W. Lee, and C. Y. Won, A real maximum power point tracking method for mismatching compensation in PV array under partially shaded conditions, IEEE Trans. Power Electron., vol. 26, no. 4, pp. 10011009, Apr. 2011. [4] S. Jain and V. Agarwal, A single-stage grid connected inverter topology for solar PV systems with maximum power point tracking, IEEE Trans. Power Electron., vol. 22, no. 5, pp. 19281940, Sep. 2007. [5] L. Zhang, K. Sun, Y. Xing, L. Feng, and H. J. Ge, A modular gridconnected photovoltaic generation system based on DC bus, IEEE Trans. Power Electron., vol. 26, no. 2, pp. 523531, Feb. 2011. [6] Z. Liang, R. Guo, J. Li, and A. Q. Huang, A high-efciency PV moduleintegrated DC/DC converter for PV energy harvest in FREEDM systems, IEEE Trans. Power Electron., vol. 26, no. 3, pp. 897909, Mar. 2011. [7] L. Zhang, W. G. Hurley, and W. H. W ole, A new approach to achieve maximum power point tracking for PV system with a variable inductor, IEEE Trans. Power Electron., vol. 26, no. 4, pp. 10311037, Apr. 2011. [8] L. Zhou, Y. Chen, K. Guo, and F. Jia, New approach for MPPT control of photovoltaic system with mutative-scale dual-carrier chaotic search, IEEE Trans. Power Electron., vol. 26, no. 4, pp. 10381048, Apr. 2011. [9] G. Petrone, G. Spagnuolo, and M. Vitelli, A multivariable perturb-andobserve maximum power point tracking technique applied to a singlestage photovoltaic inverter, IEEE Trans. Ind. Electron., vol. 58, no. 1, pp. 7684, Jan. 2011. [10] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, Optimization of perturb and observe maximum power point tracking method, IEEE Trans. Power Electron., vol. 20, no. 4, pp. 963973, Jul. 2005. [11] B. Yang, Y. Zhao, and X. He, Design and analysis of a grid-connected photovoltaic power system, IEEE Trans. Power Electron., vol. 25, no. 4, pp. 9921000, Apr. 2010. [12] T. Esram and P. L. Chapman, Comparison of photovoltaic array maximum power point tracking techniques, IEEE Trans. Energy Convers., vol. 22, no. 2, pp. 439449, Jun. 2007. [13] A. K. Abdelsalam, A. M. Massoud, S. Ahmed, and P. N. Enjeti, High-performance adaptive Perturb and Observe MPPT technique for photovoltaic-based microgrids, IEEE Trans. Power Electron., vol. 26, no. 4, pp. 10101021, Apr. 2011. [14] W. Wu, N. Pongrantananukul, W. Qiu, K. Ruston, T. Kasparis, and I. Batarseh, DSP-based multiple peak power tracking for expandable power system, in Proc. Appl. Power Electron. Conf. Expo., 2003, pp. 525 539. [15] A. Safari and S. Mekhilef, Simulation and hardware implementation of incremental conductance MPPT with direct control method using cuk converter, IEEE Trans. Ind. Electron., vol. 58, no. 4, pp. 11541161, Apr. 2011. [16] X. Zhou, D. Song, Y. Ma, and D. Cheng, The simulation and design for MPPT of PV system based on incremental conductance method, in Proc. Int. Conf. Inf. Eng., Aug. 2010, vol. 2, pp. 314317. [17] S. L. Brunton, C. W. Rowley, S. R. Kulkarni, and C. Clarkson, Maximum power point tracking for photovoltaic optimization using ripple-based extremum seeking control, IEEE Trans. Power Electron., vol. 25, no. 10, pp. 25312540, Oct. 2010. [18] M. A. S. Masoum, H. Dehbonei, and E. F. Fuchs, Theoretical and experimental analyses of photovoltaic systems with voltage- and current-based maximum power-point tracking, IEEE Trans. Energy Convers., vol. 17, no. 4, pp. 514522, Dec. 2002. [19] W. M. Lin, C. M. Hong, and C. H. Chen, Neural-network-based MPPT control of a stand-alone hybrid power generation system, IEEE Trans. Power Electron., vol. 26, no. 12, pp. 35713581, Dec. 2011. [20] K. Samangkool and S. Premrudeepreechacharn, Maximum power point tracking using neural networks for grid-connected photovoltaic system, in Proc. Int. Conf. Future Power Syst., Nov. 2005, p. 4.

[21] M. Tsai, C. Tseng, G. Hong, and S. Lin, A novel MPPT control design for PV modules using neural network compensator, in Proc. Int. Symp. Ind. Electron., May 2012, pp. 17421747. [22] D. Shmilovitz, On the control of photovoltaic maximum power point tracker via output parameters, IEE Proc. Elect. Power Appl., vol. 152, no. 2, pp. 239248, Mar. 2005. [23] D. Shmilovitz, Photovoltaic maximum power point tracking employing load parameters, in Proc. IEEE Int. Symp. Ind. Electron., Jun. 2005, vol. 3, pp. 10371042. [24] A. S. Kislovski and R. Redl, Maximum power tracking using positive feedback, in Proc. 25th Annu. IEEE Power Electron. Spec. Conf., Jun. 1994, pp. 10651068. [25] Q. Mei, M. Shan, L. Liu, and J. M. Guerrero, A novel improved variable step-size incremental-resistance MPPT method for PV systems, IEEE Trans. Ind. Electron., vol. 58, no. 6, pp. 24272434, Jun. 2011. [26] L. Piegari and R. Rizzo, Adaptive perturb and observe algorithm for photovoltaic maximum power point tracking, IET Renewable Power Generation, vol. 4, no. 4, pp. 317328, Jul. 2010. [27] F. Liu, S. Duan, B. Liu, and Y. Kang, A variable step size INC MPPT method for PV systems, IEEE Trans. Ind. Electron., vol. 55, no. 7, pp. 26222628, Jul. 2008. [28] R. Kadri, J. P. Gaubert, and G. Champenois, An improved maximum power point tracking for photovoltaic grid-connected inverter based on voltage-oriented control, IEEE Trans. Ind. Electron., vol. 55, no. 7, pp. 6675, Jan. 2011. [29] K. Kirubakaran, S. Jain, and R. K. Nema, DSP-controlled power electronic interface for fuel-cell-based distributed generation, IEEE Trans. Power Electron., vol. 26, no. 12, pp. 38533864, Dec. 2011. [30] K. Ishaque, Z. Salam, M. Amjad, and S. Mekhilef, An improved particle swarm optimization (PSO)-based MPPT for PV with reduced steady-state oscillation, IEEE Trans. Power Electron., vol. 27, no. 8, pp. 36273638, Aug. 2012. [31] Z. Qian, O. A. Rahman, H. A. Atrash, and I. Batarseh, Modeling and control of three-port DC/DC converter interface for satellite applications, IEEE Trans. Power Electron., vol. 25, no. 3, pp. 637649, Mar. 2010. [32] M. P. Kazmierkowski, M. Jasinski, and G. Wrona, DSP-based control of grid-connected power converters operating under grid distortions, IEEE Trans. Ind. Informat., vol. 7, no. 2, pp. 204211, May 2011. [33] T. A. Ocran, J. Cao, B. Cao, and X. Sun, Articial neural network maximum power point tracker for solar electric vehicle, Tsinghua Sci. Technol., vol. 10, no. 2, pp. 204208, Apr. 2005. [34] B. N. Alajmi, K. H. Ahmed, S. J. Finney, and B. W. Williams, Fuzzylogic-control approach of a modied hill-climbing method for maximum power point in microgrid standalone photovoltaic system, IEEE Trans. Power Electron., vol. 26, no. 4, pp. 10221030, Apr. 2011.

Yuncong Jiang (S10) received the B.S. degree (Hons.) in electrical engineering from the Shanghai University of Electric Power, Shanghai, China, in 2009. He is currently working toward the Ph.D. degree in the Department of Electrical and Computer Engineering, University of Alabama, Tuscaloosa. He worked on projects funded by National Science Foundation, NASA, and the U.S. Department of Agriculture. His research interests include analog and digital control in power electronics, renewable energy research, power delivery system design, maximum power point tracking techniques in solar power systems, and dcac inverters design. He served as a peer reviewer for Applied Power Electronics Conference from 2010 to 2012. He received the Graduate Council Fellowships and the Research and Creative Activity Fellowship from the University of Alabama in 2009 and 2011, respectively.

JIANG et al.: ADAPTIVE STEP SIZE WITH ADAPTIVE-PERTURBATION-FREQUENCY DIGITAL MPPT CONTROLLER

3205

Jaber A. Abu Qahouq (SM07) received the B.Sc. degree (rst class Hons.) from Princess Sumaya University/Royal Scientic Society (RSS), Amman, Jordan, in 1998, and the M.S. and Ph.D. degrees from the University of Central Florida (UCF), Orlando, in 2000 and 2003, respectively, all in electrical engineering/electronics. He is currently an Assistant Professor at the Department of Electrical and Computer Engineering, University of Alabama, Tuscaloosa. He was with Intel Corporation from August 2005 to January 2008. From January 2004 to August 2005, he was a Visiting Assistant Professor and Instructor with the School of Electrical Engineering and Computer Science, UCF, and from January 2002 to December 2003, he was an Adjunct Faculty at UCF. He was a Research Assistant/Associate with the Florida Power Electronics Center, UCF, from 1999 to 2003. From 1998 to 1999, he was with the RSS, Electronic Services and Training Center, Amman. He has published many refereed journal and conference papers (more than 100 as of August 2012) and holds several patents (13 U.S. patents granted and many others pending as of August 2012). His research interests include high-frequency high-density low-voltage high-current fast-transient dcdc converters, soft-switching power conversion, digital control in power electronics, adaptive power control, platform power delivery distribution architectures and related power management schemes, power factor correction acdc converters, and dcac inverters, renewable energy systems, power devices, nanotechnology-based electronic devices, and unmanned aerial vehicles, among others. Dr. Abu Qahouq is a Member of Eta Kappa Nu and Phi Kappa Phi. He is actively serving as a Reviewer to several journal and conference publications including IEEE TRANSACTIONS and conferences; he chaired several conference sessions over the years. He severed as a technical program committee member of the IEEE PESC 2007 conference, as a review committee member of the IEEE ISCAS 2008 conference, as a technical program committee member of the IEEE ICECS 2009, 2010, and 2011 conferences, as a technical program committee member of the IEEE ECCE 2010, 2011, and 2012 conferences, a technical program committee member of the IEEE APEC 2013 conference, and as a member of Intels mobile platforms patent committee (when he was with Intel Corporation) among others. He led and worked on several projects funded by the National Science Foundation, U.S. Department of Agriculture, Army Research Lab, NASA, Astec Power/Emerson, Intel, the UCF, and the University of Alabama. He received the Division Recognition Award from the Systems Technology Lab, Corporate Technology Group, Intel Corporation in 2006, the IET premium award for the best paper in 2009, the IEEE Outstanding Graduate Student Award in 2002, and the King of Jordan Royal Watch in 1998, among others.

Tim A. Haskew (S86M87SM02) received the B.E.E., M.S., and Ph.D. degrees from Auburn University, Auburn, AL, in 1987, 1988, and 1991, respectively. He is the Head of the Department of Electrical and Computer Engineering, University of Alabama, Tuscaloosa, and holds the rank of Professor. He also serves as the Director of the Electromechanical Systems Laboratory. He has authored or coauthored more than 50 refereed publications and two book chapters. His research interests include the electrical and mechanical disciplines, and he is specically focused on power electronics, motion control, electric machinery, energy conversion, power systems, and system control. Dr. Haskew is a Member of a number of IEEE professional societies including the Power Electronics Society, Industry Applications Society, and the Power and Energy Society. He was active on the Applied Power Electronics Conference (APEC) committee from 2004 to 2010, and he served as the APEC Seminar Chair from 2005 to 2010.