IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 1, JANUARY

Transcription

1 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 1, JANUARY Assessment of Perturb and Observe MPPT Algorithm Implementation Techniques for PV Pumping Applications Mohammed A. Elgendy, Bashar Zahawi, Senior Member, IEEE, and David J. Atkinson Abstract The energy utilization efficiency of commercial photovoltaic (PV) pumping systems can be significantly improved by employing simple perturb and observe (P&O) maximum power point tracking algorithms. Two such P&O implementation techniques, reference voltage perturbation and direct duty ratio perturbation, are commonly utilized in the literature but no clear criteria for the suitable choice of method or algorithm parameters have been presented. This paper presents a detailed theoretical and experimental comparison of the two P&O implementation techniques on the basis of system stability, performance characteristics, and energy utilization for standalone PV pumping systems. The influence of algorithm parameters on system behavior is investigated and the various advantages and drawbacks of each technique are identified for different weather conditions. Practical results obtained using a 1080-Wp PV array connected to a 1-kW permanent magnet dc motor-centrifugal pump set show very good agreement with the theoretical analysis and numerical simulations. Index Terms DC DC power conversion, maximum power point tracking (MPPT), photovoltaic (PV) power systems, photovoltaic (PV) pumping, stability. I. INTRODUCTION STANDALONE photovoltaic (PV) pumping systems have become a favorable solution for water supply, gaining more acceptance and market share, particularly in rural areas that have a substantial amount of insolation and have no access to an electric grid. The maximization of energy utilization of these systems via maximum power point tracking (MPPT) has not been sufficiently exploited in the literature. As a result, most commercial PV pumping systems either utilize inefficient MPPT control or do not utilize MPPT control at all. Directly connected systems for example operate at the intersection of current voltage curves of the PV array and the motor-pump set. This operating point may be far from the maximum power point (MPP) of the generator wasting a significant proportion of the available solar power. A simple dc dc converter controlled by an MPPT algorithm can be used as a pump controller to match the PV generator to the motor-pump set. A number of different MPPT algorithms have been proposed [1] [3]. The simplest is to operate Manuscript received September 30, 2010; revised June 24, 2011; accepted September 01, Date of current version December 16, The authors are with the School of Electrical, Electronic, and Computer Engineering, Newcastle University, Newcastle upon Tyne, NE1 7RU, England, U.K. ( mohammed.elgendy@ncl.ac.uk; bashar.zahawi@ncl.ac.uk; dave. atkinson@ncl.ac.uk). Color versions of one or more of the figures in this paper are available online at Digital Object Identifier /TSTE the PV array at constant voltage equal to the MPP voltage of the array at the standard test conditions (STCs) provided by the manufacturer (ignoring the effects of insolation and temperature variations on the MPP voltage). This value is used as a reference for a feedback control loop that usually employs a PI controller to adjust the duty ratio of the MPPT converter. The authors have previously investigated this approach and found that it offered significantly better energy utilization efficiencies (up to about 91%) compared to directly connected systems [4]. The utilization efficiency, however, can be further improved (at the cost of a small increase in implementation cost) by employing a hill-climbing MPPT technique such as the perturb and observe (P&O) algorithm. This is a simple algorithm that does not require previous knowledge of the PV generator characteristics or the measurement of solar intensity and cell temperature andiseasytoimplement with analogue and digital circuits. The algorithm perturbs the operating point of the PV generator by increasing or decreasing a control parameter by a small amount (step size) and measures the PV array output power before and after the perturbation. If the power increases, the algorithm continues to perturb the system in the same direction; otherwise the system is perturbed in the opposite direction. The number of perturbations made by the MPPT algorithm per second is known as the perturbation frequency or the MPPT frequency. Three techniques have been proposed in the literature for implementing the P&O algorithm: reference voltage perturbation [5] [11], reference current perturbation [12], [13], and direct duty ratio perturbation [6], [10], [14] [16]. Fig. 1 shows a block diagram for reference voltage perturbation in which the PV array output voltage reference is used as the control parameter in conjunction with a controller (usually a PI controller) to adjust the duty ratio of the MPPT power converter. The PI controller gains are tuned while operating the system at a constant voltage equal to the STC value of the MPP voltage. These gains are kept constant while the reference voltage is controlled by the MPPT algorithm. Similarly, the reference current perturbation approach uses the PV array output current reference as the control parameter. Due to its slow transient response to irradiance changes and high susceptibility to noise and PI controller oscillation, reference current control will not be considered in this paper. For direct duty ratio perturbation, the duty ratio of the MPPT converter is used directly as the control parameter (Fig. 2). To date, there has been no publication which includes a clear criteria for the choice of algorithm parameters and the impact of these parameters on system performance under practical /$ IEEE

2 22 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 1, JANUARY 2012 Fig. 1. Fig. 2. Fig. 3. Block diagram of MPPT with reference voltage control. Block diagram of MPPT with direct duty ratio control. Circuit diagram of experimental PV setup. operating conditions. The P&O algorithm implementation approaches have so far been utilized individually in the literature and the choice as to which method to employ continues to be an open question. This paper presents a detailed theoretical and experimental evaluation of the reference voltage perturbation and direct duty ratio perturbation P&O implementation techniques, operating under different weather conditions, in terms of overall energy utilization and the local stability of the system. The effects of different perturbation rates and step sizes are investigated and criteria for the choice of these parameters presented. II. EXPERIMENTAL SETUP AND METHODOLOGY An experimental PV pumping system prototype was constructed comprised of six 180 Wp SANYO HIP-J54BE2 solar modules (divided into two parallel branches of three series connected modules), a step-down dc dc converter, and a Sun- Pumps SCB motor-pump set consisting of a ten-stage centrifugal surface pump driven by a PM brushed dc motor, as detailed in a previous publication [4]. A simplified circuit diagram of this setup is illustrated in Fig. 3. The PV array was installed facing south at a fixed tilt angle of 54 with respect to the horizontal. This angle was chosen at the time of panel installation to obtain the maximum possible annual light incidence without the need for sun tracking equipment. The PV array current and voltage were measured with Hall Effect sensors: LTS15-NP and LV25-P, respectively. For experimental flexibility and ease of programming, a Texas Instruments TMS320F2812 DSP-based ezdsp kit was used for control and data acquisition. In a commercial product, a lower cost microcontroller would be more than adequate to implement the control algorithms under investigation. Meteorological parameters were recorded at a 1-s sampling rate utilizing a Vaisala MAWS201 weather station installed on the same roof on which the PV array is installed. Motor armature resistance and inductance were measured at 1.25 and 3.5 mh, receptively. A 470- F link capacitance was used together with a PWM frequency of 10 khz, ensuring converter operation in continuous current mode throughout the full range of duty ratio variations. To study the effects of the algorithm parameters on system performance, the experimental system was initially run at constant solar intensity and cell temperature for 30 s, ignoring variations in solar irradiance within 1%. System parameters were recorded with a sampling rate of 2 K samples/s. Transient behavior was examined by disconnecting one of the two branches of the PV array while the system is running, emulating a step decrease in solar irradiance to 50% of its value. A longer experimental test duration of 20 min was chosen to study the effects of solar irradiance and cell temperature variations on system behavior and to calculate the energy utilization for different weather conditions. In this test, parameters were recorded with a low acquisition rate of 10 samples/s to limit the host computer buffer size and the storage memory required for the acquisition files. This 20-min test was repeated many times at different weather conditions and results are shown to demonstrate the performance of the MPPT algorithm during periods of slow, smooth changes in irradiance conditions as well as faster irradiance changes that occur over a period of a few seconds. Such irradiance changes are common place in the U.K. where these tests were carried out. In this study, the energy utilization efficiency of the MPPT algorithm was calculated by dividing the integral of the measured PV generator output power by that of the maximum possible power output calculated at the same solar irradiance and cell temperature values. These calculated values are obtained from the PV generator numerical model based on the measured output characteristics of the PV array at different weather conditions. Further details of this approach are described in a recent publication by the authors [4]. A recent European standard EN50530 [17] has suggested a procedure for calculating the transient and steady-state energy utilization efficiency of an MPPT algorithm. However, the suggested approach requires measurements at specific, controlled irradiance values and rates of change. This makes it unsuitable for calculating the MPPT efficiency of a site PV installation such as the one used in this investigation, where the irradiance levels cannot be controlled as would be the case in a laboratory investigation. III. STABILITY ANALYSIS A system employing P&O MPP tracking is continually subjected to two excitation sources, one originating from variations in solar irradiance/cell temperature and the other from the perturbation of the tracking algorithm. When the system is operating under steady-state solar irradiance and temperature conditions, it is still subjected to continuous step changes in the control parameter at the selected perturbation rate. Although numerical simulations can predict the response to these step changes, they do not give a clear understanding of the transient response and the local stability of the system. An analytical solution is

3 ELGENDY et al.: ASSESSMENT OF P&O MPPT ALGORITHM IMPLEMENTATION TECHNIQUES FOR PV PUMPING APPLICATIONS 23 needed which is very difficult to develop for such a nonlinear time-varying system. Under constant and uniform irradiance and cell temperature levels, the system operates around the MPP and it is possible to develop a linearized analytical model to characterize system behavior around this equilibrium point. This can be accomplished by taking an average over a switching cycle assuming low output ripple. The averaged circuit is then linearized about the equilibrium point by expanding any nonlinear function into a Taylor series and retaining only the linear terms. The Taylor series of a nonlinear function in a neighborhood of an operating point is given by Fig. 4. Averaged circuit model of PV system with a motor-pump load. (1) This can be applied to the torque equation of the motor-pump set (2) which includes a nonlinear term representing the pump load torque [4] (2) Fig. 5. Array voltage to duty ratio small signal block diagram. where is the torque constant of the motor and and are constants that can be calculated from the Taylor series approximation of the summation of load and friction torques about the equilibrium point. At maximum power transfer, the PV generator can be replaced by a voltage or a current source whose internal resistance equals the resistance of its load. A current source with parallel resistance representation is chosen here since the PV generator current is proportional to the solar irradiance. The resultant averaged circuit model is shown in Fig. 4. This circuit can be modeled by the equations where lower-case symbols represent the quantities as a function of time. Replacing each quantity by a dc component (represented by an upper case symbol) plus a small perturbation component (represented by a lower case symbol with a circumflex) we get (3) (4) (5) Neglecting second-order perturbation terms, removing the steady-state quantities and transforming the equations to the -domain, we obtain (9) (10) (11) where upper-case symbols represent the quantities as functions of. From the above equations, the small signal block diagram (Fig. 5) can be constructed. A. Direct Duty Ratio Control The array voltage to duty ratio small signal transfer function corresponding to Fig. 5 is given by where (12) (6) and (7) (8) The coefficients of the transfer function can be calculated by substituting the system parameter values given in Table I. These parameters were obtained by a combination of measurement,

4 24 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 1, JANUARY 2012 TABLE I SYSTEM PARAMETERS AT NOMINAL OPERATING CONDITIONS; SOLAR IRRADIANCE OF 800 W/m AND 25 CCELL TEMPERATURE Fig. 8. Root locus for reference voltage control utilizing a controller. 1) Proportional Control: When using a controller, the small signal control equation is given by In this case, the open loop transfer function is given by while the closed loop transfer function is (13) Fig. 6. Array voltage response to a step increase in duty ratio. (14) Fig. 7. Block diagram of a PV system with reference voltage control. design choice, and simulation as detailed in [18]. This transfer function has a negative real pole at and two complex conjugate poles at. It has two zeros at and. The response for a step increase in duty ratio is illustrated in Fig. 6. The signal has a negative steady-state gain of about 198, the rise time is about 2.2 ms, the settling time is about 23 ms, and the peak overshoot is about 29%. This system can be approximated to an under damped second order system by eliminating the real pole with the zero at. For this system, the damping ratio and the natural frequency will be 0.34 and 635 rad/s, respectively. B. Reference Voltage Control For reference voltage control, an output voltage reference is used in conjunction with the controller to adjust the duty ratio (Fig. 7). Both proportional (P) and proportional-integral (PI) control strategies are examined below. It is accepted that other approaches such as PID control may enhance the transient response of the system to the MPPT perturbations, but will result in a more fluctuating steady-state response in the presence of noise [19]. Nonlinear controllers such as fuzzy logic [20], [21] can also be utilized to adjust the MPPT converter duty ratio. However, the implementation of such controllers requires the use of Microcontrollers/DSPs with high processing abilities and large memory sizes resulting in increased system cost. With unity feedback gain, the open loop transfer function has the same poles and zeros as those stated above for. From each of the three open loop poles, a root locus branch starts; two of these end at the two real zeros and the third at an infinite zero as shown in Fig. 8. The branches do not cross the imaginary axis and the system is always stable. The steady-state error in the step response when using a controller can be calculated by applying the final-value theorem to the error signal as follows: (15) This steady-state error can be reduced by using a higher value of. However, any noise superimposed on the PV generator voltage will be amplified by,changing the duty ratio and causing fluctuations in the array voltage. These fluctuations misguide the maximum power point tracker, when an MPPT algorithmisusedtoadjustthereference voltage. To reduce the effect of noise on the MPPT algorithm, a lowpass filter is required for the PV generator voltage feedback signal. Let us assume a 200-Hz cutoff frequency, first-order software filter is used. The small signal transfer function of the filter is given by (16) Using a first-order low-pass filter adds a pole to the open loop transfer function pulling the root locus to the right as shown in Fig. 8. This tends to lower system relative stability and slows down the settling time of the response. The system is stable for low values of and unstable for high values. The value of that makes the system marginally stable so that sustained oscillation occurs is about and the period of that oscillation is about s. Because is low for the stable

5 ELGENDY et al.: ASSESSMENT OF P&O MPPT ALGORITHM IMPLEMENTATION TECHNIQUES FOR PV PUMPING APPLICATIONS 25 Fig. 10. Array voltage response to a step increase in reference voltage. Fig. 9. Bode diagram of PV system with reference voltage control. range, the steady-state error is very high. According to (15), the minimum steady-state error in the stable range (corresponding to ) is about 25% of the steady-state output. Using ahigherorderfilter or one with a lower cutoff frequency will result in a lower value of and a higher steady-state error. 2) Proportional Integral Controller: The addition of an integral part to the controller eliminates the steady-state error and makes the system less susceptible to noise. However, it adds a pole at the origin to the open loop transfer function making the system less stable and requiring proper consideration to be given to the design of the controller gains. The small signal control equation of the PI controller is The open loop transfer function will be While the closed loop transfer function is given by (17) (18) (19) The PI controller gains are calculated from and obtained from the root locus, utilizing the second Ziegler Nichols method [19]. The obtained values using this method are and. At these values, the system has a gain margin of 3.25 db and a phase margin of 11.9 as given by the Bode plot shown in Fig. 9. Although the system is analytically stable, the stability of the practical system can not be guaranteed due to the low stability margins. There will always be high oscillation and noise present in a practical system even if further tuning trails are attempted. To solve this problem, the low-pass filter is removed from the feedback loop accepting higher noise levels as the cost of better system stability. In this case, at the design values of the PI controller gains ( and ), the system has an infinite gain margin and a 21.5 phase margin (Fig. 9). These margins are high enough to ensure stability against variations in component values of the system and variations in the reference voltage around the equilibrium point. The magnitude curve of the closed-loop Bode plot (Fig. 9) has a resonant peak of 8.6 db at 158 Hz. Corresponding to this resonant peak, the closed loop transfer function is expected to have a pair of complex conjugate poles with a damping ratio of about 0.2. The system bandwidth is about 220 Hz. The small signal closed loop transfer function is obtained by substituting with unity in (19) which is solved numerically showing that the system has two real poles at and and a pair of complex conjugate poles at. The real pole at is eliminated by a real zero reducing the system order to three. The damped natural frequency is close to the resonant frequency obtained from the frequency response analysis. However, the damping ratio of the complex conjugate poles (0.128) differs from that expected from the frequency response analysis due to the effect of the real pole at which cannot be neglected since its relative distance from the -axis is not so high. The step response of the closed loop system is shown in Fig. 10. The rise time is 1.5ms,andthesettlingtimeis25.8ms.Thepeakovershootis 29.7% of the steady-state output. IV. P&O ALGORITHM PARAMETERS The P&O algorithm continuously perturbs the operating point of the system causing the PV array terminal voltage to fluctuate around the MPP voltage even if the solar irradiance and the cell temperature are constants. Consequently, similar current and power fluctuations occur. These usually fluctuate between three levels provided that the perturbation frequency is low enough so that the system can reach a steady state before the next perturbation. The step size must also be high enough so that control is not affected by noise (and the oscillation resulting from the use of a PI controller in case of reference voltage perturbation) and to allow the perturbation to cause a measurable change in array output power. Noise has considerable effect on the MPPT algorithm performance, especially at low step sizes where the system response to noise is comparable to that of the MPPT perturbations. With reference voltage control, the noise superimposed on the array current and voltage feedback signals cannot be rejected using low-pass filters since using such filters would result in system instability as discussed in the above section. With direct duty ratio perturbation, the stability of the system is not affected by using such filters as they are placed outside the duty ratio-array voltage control loop. However, the signal delay resulting from using these filters may influence the decisions of

6 26 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 1, JANUARY 2012 Fig. 13. Armature current and motor speed waveforms with three-level operation; P&O MPPT algorithm with reference voltage perturbation. Fig. 11. Experimental verification of three-level operation of PV pumping system employing P&O MPPT algorithm with reference voltage perturbation. Fig. 12. Behavior of P&O MPPT algorithm with reference voltage perturbation in thee-level operation. the algorithm. A second-order low-pass filter with cutoff frequency of 200 Hz was chosen to ensure good noise rejection with an acceptable signal delay of about 1.5 ms. A. Three-Level Operation of the P&O Algorithm Fig. 11 shows the three-level operation of the experimental PV pumping system employing a P&O MPPT algorithm with reference voltage perturbation when the system was started and run at W/m solar irradiance and 27.9 C cell temperature with a low perturbation frequency of 1 Hz and a high step size of 10 V. System operation can be better explained with reference to the array power voltage curve at the same irradiance and cell temperature levels (Fig. 12). Array power was measured at about 900 W with the system operating at the initial reference voltage (162 V) represented by point A. The initial perturbation direction is to increase the reference voltage, so the reference voltage is increased by 10 V (the step size) to 172 V moving the operating point to point B. Array power is then measured after a perturbation period of 1 s has passed. Because power is decreased at point B (about 778 W), the P&O algorithm reverses the perturbation direction decreasing the reference voltage to 162 V and moving the operating point back to point A. Due to the power increase at point A, the algorithm continues to decrease the reference voltage to 152 V (point C) passing through the MPP (926 W at point O) located at V. Because the power at point C (915 W) is higher than that at A, the P&O algorithm continues to decrease the reference voltage to 142 V (point D) where output power falls to W. Consequently, the P&O algorithm reverses the perturbation direction increasing the reference voltage to 152 V (point C), then 162 V (point A) and the sequence is repeated until there is a change in solar irradiance/cell temperature. The dc motor draws a high starting current of about 35 A and reaches a steady-state value equal to the array current divided by the duty ratio of the power converter after about 50 ms, as shown in Fig. 13. The motor current oscillates for about 20 ms after each MPPT perturbation. The motor speed takes longer to settle due to the higher mechanical time constant of the motorpump set. Transients and high frequency motor current ripple are damped by the motor inertia, thus have less effect on motor speed(fig.13).theflow rate is proportional to motor speed while the load torque is a square function of motor speed. Similarly, when employing direct duty ratio perturbation the waveforms of the system fluctuate between three levels at a low perturbation frequency and a high step size (Fig. 14). For an irradiance level of W/m and a cell temperature of 43 C, the initial operating point at 50% duty ratio is to the right hand side of the MPP. The initial perturbation direction is to increase the duty ratio (thus moving the operating point to the left along the arrays power voltage curve) with a chosen step size of 5%. A higher array power was then measured after every perturbation period (1 s) and the algorithm increases the duty ratio until the operating point crosses the MPP and the array power starts to decrease (at a duty ratio of 90%). The algorithm reverses the perturbation direction decreasing the duty ratio and the operating point then fluctuates between three levels in the same manner described above for reference voltage perturbation. Although the tests presented in Figs. 11 and 14 were not carried out under the same solar irradiance and cell temperature conditions (they were in fact carried out on different days),

7 ELGENDY et al.: ASSESSMENT OF P&O MPPT ALGORITHM IMPLEMENTATION TECHNIQUES FOR PV PUMPING APPLICATIONS 27 Fig. 16. Experimental results showing the effect of algorithm parameters on the array voltage; P&O MPPT algorithm with direct duty ratio perturbation. Fig. 14. Experimental verification of three-level operation of PV pumping system utilizing P&O MPPT algorithm direct duty ratio perturbation. Fig. 17. Experimental results showing the effect of algorithm parameters on the array voltage; P&O MPPT algorithm with reference voltage perturbation. Fig. 15. Armature current and motor speed waveforms with three-level operation; P&O MPPT algorithm with direct duty ratio perturbation. they nevertheless give an indication of the different starting behaviors of the two implementation techniques. Unlike the MPP voltage which varies in a narrow range around its standard test conditions value, the optimum duty ratio can vary from 0% to 100% depending on the irradiance level. As a result, a system employing a P&O algorithm with direct duty ratio control has a slower transient response compared with reference voltage perturbation. For instance, if the optimum duty ratio is 90% and the system is started with 50% duty ratio, the system takes 40 perturbation cycles to reach the MPP with a step size of 1%. Armature current and motor speed waveforms are shown in Fig. 15. Compared with Fig. 13, the slower response of the system is evident together with the smaller oscillations in armature current due to the absence of a PI controller. B. Effect of Perturbation Rate and Step Size A lower step size results in lower steady-state oscillations but slows down the system response to radiation and temperature changes, and vice versa. For example, at W/m solar irradiance and 43 C cell temperature (Fig. 14), 5 s are enough for the P&O algorithm to shift array voltage to fluctuate around the MPP, with a 5% step size and 1-Hz perturbation frequency. However, at similar weather conditions (Fig. 16), 15 s were required with a 2% step size at the same perturbation frequency. The slow response at low step sizes can be resolved using a higher perturbation rate (Fig. 16). The transient time was less than2swhenusinga10-hzperturbation frequency with a 2% step size at an irradiance level of W/m and a cell temperature of 31.8 C. With direct duty ratio perturbation, higher perturbation rates up to the PWM rate (or the ADC rate if lower) can be used. However, if the perturbation period becomes lower than the settling time of the system response, the system is never allowed to reach a steady state and its response at a particular time is affected by previous perturbations resulting in chaos-like behavior. With reference voltage perturbation, a practical system with a low step size is more susceptible to any noise superimposed on the array current/voltage waveforms. This is because the response to noise in such a system is comparable to the effect of algorithm perturbations. The array output voltage fluctuates between many levels around the MPP as shown in Fig. 17. A higher perturbation frequency results in faster deviation from the MPP, faster recovery, and a faster response to irradiance and temperature changes. If the perturbation period becomes lower than the settling time of the system, the system is never allowed

8 28 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 1, JANUARY 2012 Fig. 18. Experimental results showing the system responses to a PV array branch disconnection ( %, Hz). to reach a steady state and the array output voltage fluctuates in a random-like pattern. Further increases in perturbation rate may cause the PI controller to lose its stability. An additional experimental test showing the effect of algorithm parameters on the transient response of the system was carried out by disconnecting one of the two branches of the PV array while the system is running. This mimics a step decrease in solar irradiance to 50% of its value but without the slight change in MPP voltage that occurs with real changes in solar irradiance. The MPP duty ratio changes significantly by a branch disconnection as well as by a step irradiance change and the algorithm parameter values have considerable effects on the system response, as shown in Figs However, because the MPP voltage remains unchanged, the P&O algorithm with reference voltage perturbation may be confused only for a single perturbation period (regardless of the algorithm parameter values) due to the power change resulting from the branch disconnection. For this reason such a test was not considered for the reference voltage perturbation. When one of the two PV array branches is disconnected at an irradiance level of W/m and a cell temperature of 41.3 C (Fig. 18), the algorithm takes eight perturbation cycles to adjust the duty ratio for the new maximum power level (2 s at a 4-Hz perturbation frequency) including four cycles resulting from two algorithm confusions. The algorithm attributes the power decrease due to branch disconnection to the preceding perturbation which was to decrease duty ratio (at s), accordingly, it reverses the direction of perturbation increasing the duty ratio at s and moving the operating point away from the MPP. At s, the measured power decreases and the algorithm corrects the perturbation direction. However, less power is measured at s confusing the algorithm once more and requiring another perturbation cycle to recover. When the array branch is disconnected at a similar irradiance level with a lower step size of 2% (Fig. 19), the P&O algorithm takes 4.25 s to adjust the duty ratio to the new maximum power level, including 1 s resulting from algorithm confusion due to system dynamics. The algorithm is not confused by branch disconnection because the perturbation preceding that disconnec- Fig. 19. Experimental results showing the system responses to a PV array branch disconnection ( %, Hz). Fig. 20. Experimental results showing the system responses to a PV array branch disconnection ( %, Hz). tion was to increase the duty ratio and the algorithm reverses the perturbation to the correct direction. Fig. 20 shows that the algorithm takes a shorter transient time of 1.1 s (including 0.6 s resulting from confusions due to branch disconnection and system dynamics) when the branch is disconnected at an W/m solar irradiance and a 40.5 C cell temperature with a 10-Hz perturbation frequency. Although the algorithm is confused three times, the transient response is still faster than that shown previously in Fig. 18 with a 4-Hz perturbation frequency. C. Choice of Perturbation Rate and Step Size Different values of energy utilization efficiency have been reported in the literature for systems employing a P&O MPPT algorithm ranging from 68% [22] to 81.5% [23] and 97.8% [1]. In this project, a higher energy utilization efficiency of up to 99% was calculated for the experimental system used in this investigation system. The main reason for this variation in energy

9 ELGENDY et al.: ASSESSMENT OF P&O MPPT ALGORITHM IMPLEMENTATION TECHNIQUES FOR PV PUMPING APPLICATIONS 29 utilization efficiency values is the choice of the P&O algorithm parameters, i.e., the step size and the perturbation frequency. There is no published general procedure for the optimum choice of P&O algorithm parameters and trial and error is often used when making this choice. Some attempts have been made to calculate the optimum values of P&O algorithm parameters. Femia et al. calculated the parameters of the P&O algorithm for a system with direct duty ratio perturbation [16] and for a system with reference voltage perturbation [6] in relation to the dynamic behavior of the specific systems used in those investigations. A perturbation period higher than the settling time of the system response was chosen considering a 10% steady-state error. With 10% steady-state error, three-level operation occurs only with high step sizes increasing the amplitude of the steady-state oscillation, reducing the energy utilization efficiency. Regarding step size calculations, Femia assumed that the optimum step size is the one at which the algorithm will not be confused by solar irradiance changes. This can be ensured if the magnitude of power change due to perturbation during the perturbation period is higher than that resulting from the irradiance change during the same period. Based on this assumption, Femia derived a formula to calculate the step size depending on some intrinsic parameters of the utilized solar cell such as the series resistance, the reverse saturation current, and the ideality factor as well as the rate of change of solar irradiance. These cell parameters need to be calculated initially as they are usually not given by the manufacturer. The rate of change of solar irradiance varies significantly from time to time for the same installation site and from one installation site to another. For example, Fig. 27 (cf. Section VI) shows that a rate of irradiance change of about 550 W/m per second was reached during a 20-min period of measurement, while the average irradiance change during the day on which these measurements were taken was less than 10 W/m per second. If the maximum rate is considered for step size calculations, it will result in very high steady-state oscillations and poor energy utilization. If the average rate is used in calculations instead, the algorithm is more likely to be confused during rapid irradiance changes. A simpler general procedure for the choice of the parameters of the P&O MPPT algorithm is described in this section. This approach does not require complex mathematical calculations or prior knowledge of site and/or array dependent parameters. Figs. 21 and 22 show flowcharts of the proposed procedure for direct duty ratio perturbation and reference voltage perturbation, respectively. For three-level operation, the perturbation period (for both perturbation techniques) must be chosen higher than the settling time of the system response to a single MPPT perturbation, as calculated in the above stability analysis, considering a 2% steady-state error. For commercial systems, this settling time can simply be measured using an oscilloscope by running the system at a high irradiance level with a low perturbation frequency (say 1 Hz) and a high step size (say 10%) to ensure three-level operation. If more than three levels are observed, the perturbation frequency is reduced to allow the system to settle to a steady state after each MPPT perturbation. A system with a low settling time, allows the use of a lower perturbation period and vice versa. The link capacitance should not be over-sized unless this is necessary to ensure certain current ripple and ef- Fig. 21. Choice of the parameters of the P&O algorithm with direct duty ratio perturbation. Fig. 22. Choice of the parameters of the P&O algorithm with reference voltage perturbation. fective series resistance of the capacitor as this will slow down the system response and increase the sampling time. The optimum step size should be calculated to give a low steady-state error at an acceptable speed of system response. For the system under consideration, the sizes of the PV array and the motor-pump set were chosen so that the motor voltage did not exceed 135 V at the STC irradiance. Taking into consideration that the MPP voltage at the STC is 162 V, the duty ratio should not normally exceed 90% %.Itisalsoassumed that the system would not pump water when the duty ratio is less than 10%, hence a value of %. With direct duty ratio perturbation, we recommend to set the initial duty ratio of the converter to a central value of 50% to shorten the time required by the P&O algorithm to change the duty ratio from the initial value to the steady-state value, unless it is essential to limit the motor starting current for motor-based loads. This

10 30 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 1, JANUARY 2012 time is referred to as the starting transient time of the P&O algorithm. With the above initial value and operating range of the duty ratio, the algorithm needs to increase/decrease the duty ratio by a maximum value of 40% during starting. Now, the question is how long should it take the P&O algorithm to change the duty ratio by this amount? A shorter algorithm transient time means better energy utilization during the transient stage but this is associated with a higher step size, higher steady-state oscillations and lower steady-state energy utilization. As a compromise, we suggest a maximum transient time of 2 s for the P&O algorithm to change the converter duty ratio from its initial value to the optimum value. This means that the required rate of change of the duty ratio can be obtained by dividing by the corresponding transient time of the P&O algorithm (2 s in this case). The percentage step size is the product of and the perturbation period. For example, with a 10-Hz perturbation rate, the duty ratio can be increased/decreased by 40% in 2 s by using a step size of 2%. A smaller change in duty ratio (smaller than 40%) is expected even for significant changes in irradiation producing an even faster response by the algorithm. For instance, the total change in duty ratio is about 18% when the irradiance increases from 500 to 1000 W/m producing an algorithm response time of less than 1s. With the reference voltage perturbation technique, the system has better transient response as the MPP voltage varies in a narrow range around its STC value equal to 81% of the STC open circuit voltage and varies from 65% to 85% of that voltage depending on irradiance and temperature levels, as discussed in [4]. This means that at worst, when the system is started at the STC value of the MPP voltage (also recommended to shorten the transient time), the P&O needs to decrease the reference voltage by 14% of the STC open circuit voltage (i.e., 28 V for the considered system) during starting. Assuming a 2-s transient time, this can be achieved with a step size of about 1.1 V. The reference voltage step size should be chosen high enough so that the algorithm is not confused by noise and oscillation resulting from the PI controller. V. ALGORITHM CONFUSION DUE TO IRRADIANCE CHANGES During solar irradiance changes, the P&O algorithm may be confused depending on the direction of the last perturbation. For example, if the irradiance increases just after the algorithm decreases the reference voltage, the power increase will be attributed to the perturbation and the algorithm will continue to decrease the reference voltage moving the operating point away from the maximum power point. For a step change in solar irradiance, the P&O algorithm recovers the correct direction after a single perturbation period (Fig. 23). For a ramped change in solar irradiance, the duration of the confusion period depends on the relative effect of the irradiance change and the perturbation on the array output power. For example, if the irradiance increases from 500 to 1000 W/m in 0.2 s with a step size of 2 V and a perturbation frequency of 10 Hz (Fig. 24), the confusion remains until the irradiance change stops (since the irradiance change is more significant). For a ramp decrease in solar irradiance, the algorithm confusion has less effect on the array Fig. 23. Simulation results showing the confusion of the P&O algorithm with reference voltage perturbation by a step increase in solar irradiance. Fig. 24. Simulation results showing the confusion of the P&O algorithm with reference voltage perturbation by a ramp increase in solar irradiance. output power as the algorithm reverses the perturbation direction after each sample. The recovery time depends on the distance between the operating point and the MPP, the step size, and the perturbation rate. With direct duty ratio control, the P&O algorithm may be confused during solar irradiance changes in the same manner discussed above for reference voltage control. For a step change in solar irradiance, the P&O algorithm recovers the correct direction after a single perturbation period. For a ramp change, the duration of the confusion period depends on the relative effect of the perturbation and the irradiance change on the PV array output power. For example, if the irradiance increases from 500 to 1000 W/m in 0.75 s with a step size of 1% and a perturbation frequency of 10 Hz (Fig. 25), the maximum array power increases at about 73 W per perturbation cycle due to irradiance increase while the maximum change in array power due to perturbation is about 24 W per cycle. As the irradiance change is more significant, the confusion is supposed to remain until the irradiance change stops. However, as the operating point moves away from the MPP, the system response becomes slower resulting in confusion as a result of system dynamics at s. The P&O algorithm then reverses the direction of perturbation to the correct direction, i.e., increasing the duty ratio. For a ramp decrease in solar irradiance, the array output power is

11 ELGENDY et al.: ASSESSMENT OF P&O MPPT ALGORITHM IMPLEMENTATION TECHNIQUES FOR PV PUMPING APPLICATIONS 31 Fig. 25. Simulation results showing the confusion of the P&O algorithm with direct duty ratio perturbation by a ramp increase in solar irradiance. less affected by the confusion of the P&O algorithm since the algorithm reverses the perturbation direction continually during a period when array power is decreasing. Fig. 26. Experimental system performance under slow changing irradiance; P&O algorithm with reference voltage perturbation. VI. ENERGY UTILIZATION The energy utilization efficiency of the experimental system was calculated for 20-min operation periods under slow changing irradiance and rapidly changing irradiance conditions. For the reference voltage perturbation, with slow changing solar irradiance (Fig. 26), the calculated energy utilization efficiency was 97.2% with a 2-V step size and a perturbation frequency of 5 Hz. The utilization efficiency is relatively low due to the effect of noise. In the presence of noise, the magnitude of the array voltage ripple may not be decreased by using lower step sizes. For example, no significant difference was observed in array voltage ripple when a 2-V step size was used (Fig. 26) compared to the steady irradiance periods in Fig. 27 where a 5-V step size is used. The calculated energy utilization efficiency was marginally lower during rapidly changing irradiance due to the energy loss during the confusion and recovery periods. For example, the energy utilization efficiency for the solar irradiance shown in Fig. 27 was about 97% at a step size of 5 V and a 5-Hz perturbation frequency. Slightly higher energy utilization efficiency was achieved with direct duty ratio control than that obtained with reference voltage control due to the lower impact of noise and the absence of the oscillation resulting from the PI controller. For instance, the calculated energy utilization efficiency of the experimental system for the 20-min operation periods under slow changing irradiance was about 99% when using a 2% step size and 10-Hz perturbation frequency (Fig. 28). It is worth noting that the peak peak array voltage ripple with the chosen algorithm parameters (Fig. 28) is about 6% of the MPP voltage compared to about 12% in the corresponding case in [16] (see [16, Fig. 14]). The energy utilization efficiency was lower in rapidly changing irradiance due to the energy loss during the confusion and the recovery periods, where the operating point is away from the MPP. For example, the energy utilization efficiency for the solar irradiance shown in Fig. 29 was calculated at Fig. 27. Experimental system performance under rapidly changing irradiance; P&O algorithm with reference voltage perturbation. 97.9%. Higher variations in the array voltage occur at lower irradiance levels since the power voltage curve of the array becomes flatter so that the perturbation causes a tiny change in the array power which becomes comparable to that caused by noise and/or system dynamics, confusing the algorithm. At higher irradiance levels, the energy utilization efficiency is increased. This is due to the change in control parameter producing a higher array power change magnitude and consequently lower algorithm confusion. In addition, higher energy utilization is expected for systems with lower time constants as the algorithm confusion resulting from system dynamics is reduced. The change in average array voltage with solar irradiance shown in Figs is unusually low. This is due to the superior temperature characteristics of the HIT PV array used in this investigation (open circuit voltage temperature coefficient of 0.26% per C) and the significant cooling effect of the high wind speeds at the installation site which frequently exceeds 20 m/s.

12 32 IEEE TRANSACTIONS ON SUSTAINABLE ENERGY, VOL. 3, NO. 1, JANUARY 2012 calculated at 97.2% for slow changing solar irradiance. The energy utilization efficiency is marginally lower at 97% for rapidly changing irradiance due to the energy loss during the confusion and recovery periods when irradiance changes. Direct duty ratio control offers better energy utilization and better stability characteristics at a slower transient response and worse performance at rapidly changing irradiance. System stability is not affected by using low-pass feedback filters. Higher energy utilization efficiency was achieved with direct duty ratio control due to the lower impact of noise and the absence of the oscillation resulting from the PI controller. The calculated energy utilization efficiency of the experimental system was about 99% and 97.9% for the slow and rapidly changing irradiance, respectively. In addition, direct duty ratio perturbation allows the use of high perturbation rates up to the PWM rate without losing the global stability of the system. Fig. 28. Experimental system performance under slow changing irradiance; direct duty ratio perturbation. Fig. 29. Experimental system performance under rapidly changing irradiance; direct duty ratio perturbation. VII. CONCLUSION The paper presents a comprehensive analysis and experimental evaluation of the reference voltage perturbation and direct duty ratio perturbation techniques for implementing the P&O MPPT algorithm. The effects of the perturbation rate and step size on system behavior were examined, the criteria for the choice of these parameters presented, and the energy utilization calculated at slow and rapidly changing weather conditions using a 1080-Wp experimental setup. With reference voltage perturbation, the system has a faster response to irradiance and temperature transients. However, it loses stability if operated at a high perturbation rate or if lowpass filters are used for noise rejection from the array current and voltage feedback signals. This noise has significant impact on the algorithm performance and consequently on the energy utilization, especially with low step sizes where the system response to noise is comparable to that of MPPT perturbations. The energy utilization efficiency of the experimental system was ACKNOWLEDGMENT The authors would like to thank the staff at Narec and in particular Dr S. McDonald for the use of their PV arrays and for their valuable support. REFERENCES [1] D. P. Hohm and M. E. Ropp, Comparative study of maximum power point tracking algorithms, Prog. Photovoltaics: Res. Appl., vol. 11, pp , [2] V.Salas,E.Olias,A.Barrado,and A. Lazaro, Review of the maximum power point tracking algorithms for stand-alone photovoltaic systems, Solar Energy Mater. Solar Cells, vol. 90, pp , [3] T. Esram and P. L. Chapman, Comparison of photovoltaic array maximum power point tracking techniques, IEEE Trans. Energy Convers., vol. 22, no. 2, pp , Jun [4] M.A.Elgendy,B.Zahawi,andD.J.Atkinson, Comparisonofdirectly connected and constant voltage controlled photovoltaic pumping systems, IEEE Trans. Sustain. Energy, vol. 1, no. 3, pp , Oct [5] O. Wasynczuk, Dynamic behavior of a class of photovoltaic power systems, IEEE Trans. Power App. Syst., vol. PAS-102, no. 9, pp , Sep [6] N. Femia, G. Petrone, G. Spagnuolo, and M. Vitelli, A technique for improving P&O MPPT performances of double-stage grid-connected photovoltaic systems, IEEE Trans. Ind. Electron., vol. 56, no. 11, pp , Nov [7] R.Alonso,P.Ibáñez,V.Martínez, E. Román, and A. Sanz, An innovative perturb, observe and check algorithm for partially shaded PV systems, presented at the 13th Eur. Conf. Power Electronics and Applications (EPE 09), Barcelona, Spain, [8] H. Patel and V. Agarwal, Maximum power point tracking scheme for PV systems operating under partially shaded conditions, IEEE Trans. Ind. Electron., vol. 55, no. 4, pp , Apr [9] D. Sera, R. Teodorescu, J. Hantschel, and M. Knoll, Optimized maximum power point tracker for fast-changing environmental conditions, IEEE Trans. Ind. Electron., vol. 55, no. 7, pp , Jul [10] A. Pandey, N. Dasgupta, and A. K. Mukerjee, High-performance algorithms for drift avoidance and fast tracking in solar MPPT system, IEEE Trans. Energy Convers., vol. 23, no. 2, pp , Jun [11] M. Fortunato, A. Giustiniani, G. Petrone, G. Spagnuolo, and M. Vitelli, Maximum power point tracking in a one-cycle-controlled single-stage photovoltaic inverter, IEEE Trans. Ind. Electron., vol. 55, no. 7, pp , Jul [12] W.T.Chee,T.C.Green,andA.H.-A. Carlos, Analysis of perturb and observe maximum power point tracking algorithm for photovoltaic applications, presented at the 2008 IEEE 2nd Int. Power and Energy Conf. (PECon 2008), Johor Bahru, Malaysia, [13] C. W. Tan, T. C. Green, and C. A. Hernandez-Aramburo, A currentmode controlled maximum power point tracking converter for building integrated photovoltaics, presented at the 2007 Eur. Conf. Power Electronics and Applications (EPE), Aalborg, Denmark, 2007.

13 ELGENDY et al.: ASSESSMENT OF P&O MPPT ALGORITHM IMPLEMENTATION TECHNIQUES FOR PV PUMPING APPLICATIONS 33 [14] E. Koutroulis, K. Kalaitzakis, and N. C. Voulgaris, Development of a microcontroller-based, photovoltaic maximum power point tracking control system, IEEE Trans. Power Electron., vol.16,no.1,pp , Jan [15] R. Gules, J. D. P. Pacheco, H. L. Hey, and J. Imhoff, A maximum power point tracking system with parallel connection for PV standalone applications, IEEE Trans. Ind. Electron., vol. 55, no. 7, pp , Jul [16] 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 , Jul [17] Overall Efficiency of Grid Connected Photovoltaic Inverters, BSEN 50530:2010, BSI, BSI Standards Publication, [18] M. A. Elgendy, Photovoltaic Pumping Systems With Microcontroller- Based MPPT Control, Ph.D. Thesis, School of Electrical, Electronics and Computer Engineering, Newcastle University, Newcastle, U.K., [19] K. Ogata, Modern Control Engineering, 4th ed. Englewood Cliffs, NJ: Prentice-Hall, [20] A. E. A. Nafeh, F. H. Fahmy, and E. M. Abou El-Zahab, Maximumpower operation of a stand-alone PV system using fuzzy logic control, Int. J. Numerical Modelling: Electronic Networks, Devices and Fields, vol. 15, pp , [21] A. M. A. Mahmoud, H. M. Mashaly, S. A. Kandil, H. El Khashab, and M. N. F. Nashed, Fuzzy logic implementation for photovoltaic maximum power tracking, in Proc. Industrial Electronics Conf. IECON, vol. 1, pp , 20. [22] T. Tafticht, K. Agbossou, M. L. Doumbia, and A. Chériti, An improved maximum power point tracking method for photovoltaic systems, Renew. Energy, vol. 33, pp , [23] K. H. Hussein, I. Muta, T. Hoshino, and M. Osakada, Maximum photovoltaic power tracking: An algorithm for rapidly changing atmospheric conditions, in Proc. Inst. Elect. Eng., Generation, Transmission and Distribution, 1995, vol. 142, pp Bashar Zahawi (M 96 SM 04) received the B.Sc. and Ph.D. degrees in electrical and electronic engineering from Newcastle University, England, in 1983 and From 1988 to 1993, he was a design engineer with a U.K. manufacturer of large variable speed drives and other power conversion equipment. In 1994, he was appointed as a Lecturer in Electrical Engineering at the University of Manchester and in 2003 he joined the School of Electrical, Electronic and Computer Engineering at the Newcastle University, England, as a Senior Lecturer. His research interests include small scale generation, power conversion, and the application of nonlinear dynamical methods to electrical circuits and systems. Dr. Zahawi is a chartered electrical engineer. David J. Atkinson received the B.Sc. degree in electrical and electronic engineering from Sunderland Polytechnic, England, in 1978, and the Ph.D. degree from Newcastle University, England, in He is currently a Senior Lecturer in the Power Electronics, Drives and Machines Research Group at the School of Electrical, Electronic and Computer Engineering, Newcastle University. He joined the university in 1987 after 17 years in industry with NEI Reyrolle Ltd. and British Gas Corporation. His research interests are mainly focussed on control of power electronics systems including electric drives and converters. Dr. Atkinson is a chartered electrical engineer. Mohammed A. Elgendy was born in Behera, Egypt, in He received the B.Sc. degree in 1997 from Menoufia University, Egypt, the M.Sc. degree in 2003 from Ain Shams University, Egypt, and the Ph.D. degree from Newcastle University, England, in 2010, all in electrical engineering. From June 1998 to May 2006, he was a Research Assistant at the New and Renewable Energy Department, Desert Research Centre, Cairo, Egypt. He is currently a Research Associate at the School of Electrical, Electronic & Computer Engineering, Newcastle University, England. His research focus is on control of power electronic converters for photovoltaic systems and other renewable generation schemes.