Вы находитесь на странице: 1из 9

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 2, NO.

1, JANUARY 2012

47

A Maximum Power Point Tracker for Long-Term Logging of PV Module Performance


Uwe Zimmermann, Member, IEEE, and Marika Edoff
AbstractWe present a monitoring system for the eld test of photovoltaic modules. The system is designed for the monitoring of individual modules under maximum power point (MPP) conditions that allow the extraction of the energy yield of different modules under optimum conditions. It can also be used to monitor the performance and long-term stability of modules under realistic eld conditions. The monitoring system consists of an individual MPP tracker attached to each module under test. The measurement data are transmitted to a central multichannel data logger by means of analog voltages proportional to the current at MPP, i.e., Im p , and voltage at MPP, i.e., Vm p , of the modules. Index TermsEmbedded software, maximum-power-point tracking, performance evaluation, photovoltaic cells, power MOSFET.
Fig. 1. Overview over the PV module monitoring system. The shaded area represents the MPPT unit, which is attached to the PV module under test. The data are captured with a central multichannel logger.

I. INTRODUCTION

ONITORING the performance of photovoltaic (PV) modules under eld conditions is important for both the testing of individual modules and the comparison of energy yield for different types of solar cell modules at certain geographic locations [1]. Some past studies perform such testing using small-scale PV installations connected to commercial balance of system components, like grid-connected inverters or electric loads, while others use dedicated test equipment [2][4]. Under real-world conditions, the electrical parameters of the maximum power point (MPP) change constantly with changing irradiance, caused by the changing angle of incidence of the solar radiation, by atmospheric perturbations and clouds and by local shading. Dedicated test equipment should, therefore, consist of a maximum power point tracker (MPPT) and an electronic load that dissipates the electric power generated by the PV modules [5], [6]. In recent years, research has been done on improving the capabilities of MPPT algorithms, partly driven by the availability of more powerful control circuitry (see, e.g., [7][13]). However, even a quite simple hill-climbing or perturb-and-observe algorithm can give sufcient accuracy and response speed to assess the power output of PV modules to within a few percent of error margin over a wide range of irradiances [14].

The goal of the work presented in this paper is to develop a modular MPPT unit to be used in a eld-test installation on the roof top of our institute. With this installation, we are currently comparing the annual energy yield of four different types of commercial PV modules in the Swedish climate. The MPPT units were planned to be exible enough to easily adapt to PV modules of different sizes and, thus, power levels. Because of the inherent differences between, e.g., wafer-based and thin-lm PV modules, there was also a need to handle different levels of module voltage and current, from Vo c 20 V and Isc 5 A for a typical 80-W crystalline silicon module to Vo c 80 V and Isc 1.5 A for a typical 80-W thin-lm module. The electronic load was realized using power MOSFETs in a constant-voltage feedback circuit. Power MOSFETs are regularly used as the active element in electronic loads because of the convenient control of the load resistance by means of a varied gate potential. These devices are also characterized by their ruggedness and scalability [15][17]. II. MEASUREMENT SYSTEM An overview over the measurement system for a single module can be seen in Fig. 1. The modules are installed on the roof top of the Angstr m Laboratory in Uppsala, Sweden. Each PV o module is connected to an individual MPPT unit, which keeps the module under optimum working conditions independent of the actual irradiance. The connection of the module at the input of the MPPT unit is shown in more detail in Fig. 2, indicating the variable load and the shunt resistor R41 . The unit acquires the instantaneous values of the PV module voltage Vm o d = Vin,+ Vin, and the voltage drop over the shunt resistor caused by the module current Vin, = Im o d R41 . The load itself is regulated by a control signal from the MPPT, closing a feedback loop. Within the MPPT unit, the module voltage and current are translated to analog voltage

Manuscript received September 6, 2011; revised October 18, 2011; accepted October 23, 2011. Date of publication December 8, 2011; date of current version January 30, 2012. This work was supported by the Swedish Electrical Utilities R&D CompanyELFORSK ABwith contributions from the Swedish Energy Agency within the SolEl-programme (http://www.solelprogrammet.se/) under Project 40015. The authors are with the Division of Solid State Electronics, Department of Engineering Sciences, Uppsala University, 75121 Uppsala, Sweden (e-mail: uwe.zimmermann@angstrom.uu.se; marika.edoff@angstrom.uu.se). 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/JPHOTOV.2011.2174031

2156-3381/$26.00 2011 IEEE

48

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 2, NO. 1, JANUARY 2012

Fig. 2. Module current Im o d is measured as the voltage drop over the shunt resistor R 4 1 , while the module voltage is measured as V m o d = V in , + V in , . The electronic load is a constant-voltage current sink. The set point is derived by means of a low-pass lter (LPF) from the PWM output of a microcontroller and compared with the scaled measured voltage V V P V of the PV module.

Fig. 4. Inverting amplier to measure the load current of the PV module connected to the measurement system.

amplier in the prototypes was set to give VIPV = 1V R32 R32 R41 Vin, = Im o d = Im o d R31 R31 A (2)

Fig. 3. Differential amplier to measure the terminal voltage of the PV module connected to the measurement system.

where VIPV stands for a voltage signal representing the measured current. Scaling the full-range input current 010 A to a 010 V signal, the load current is transmitted to the central data logger by means of a shielded twisted-pair cable. Changing the ratio of the resistors allows us to easily adopt the unit to different input current ranges. The signal VIPV is scaled to match the input range of the ADC. C. Maximum Power Point Tracker Controller The central part of the MPPT unit is a C, which samples the measured current IPV and voltage VPV of the PV module and controls the electronic load according to an MPPT algorithm. For the prototype units described here, the Atmel ATtiny45 running on an internal clock frequency of 8 MHz at 5-V supply voltage was chosen as the C [18]. Two of the four ADC channels of this controller are used to sample the derived voltage signals VIPV and VVPV with 10-bit resolution. In place of digital-to-analog converter (DAC) outputs, the controller features four independent pulsewidth modulation (PWM) channels with 8-bit resolution. One of these channels is low-pass ltered and is used to control the electronic load, which is described in Section II-D. The low-pass lter is a thirdorder voltage-controlled voltage source or general SallenKey lter [19] with a cutoff frequency of 600 Hz and a dc voltage gain of 2, while the PWM signal has a repetition frequency of 15 kHz. This gives a 010 V signal that is used as a set point for the electronic load. The currently implemented algorithm of the MPPT controller takes up only 14% of the available program memory of the ATtiny45. The principle of the algorithm is shown as the example of the powervoltage characteristics of a PV module in Fig. 5 and as a ow chart in Fig. 6. This algorithm is described as perturb-and-observe or hill-climbing in the literature [20]. It is based on the measurement of the power output P1,2 of the PV module for two different set-point voltages V1 = Vset and V2 = Vset + 2V , respectively. The relation between the power levels P1 and P2 determines the action of the MPPT.

signals which are remotely logged by an Agilent 34970A data logger equipped with 34901A 20-channel armature multiplexers in our measurement lab. The signals from all modules, readouts from temperature sensors at the backside of the modules, and from four independent irradiance sensors are scanned in 10-s intervals. A. Voltage Measurement The PV module voltage is measured at the connection terminal by means of a classic differential amplier shown in Fig. 3. With the resistor values R21 = R22 = 100 k and R23 = R25 = 10 k, the gain of the amplier in the prototypes was set to result in an output signal VVPV VVPV = R23,25 Vm o d . (Vin,+ Vin, ) = R21,22 10 (1)

Scaling the full-range input voltage of 0100 V to a 010 V signal, the module voltage is transmitted to the central data logger by means of a shielded twisted-pair cable. Changing the ratio of the resistors allows us to easily adopt the unit to different input voltage ranges. The voltage signal VVPV formed by the resistive voltage divider R26 and R27 matches the input range of the analog-to-digital converter (ADC) of the microcontroller (C). B. Current Measurement The load current is measured in the form of the voltage drop over the shunt resistor R41 = 0.1 by an inverting amplier, as shown in Fig. 4. With this value for the shunt resistor and the resistors R31 = 10 k and R32 = 100 k, the gain of the

ZIMMERMANN AND EDOFF: MAXIMUM POWER POINT TRACKER FOR LONG-TERM LOGGING OF PV MODULE PERFORMANCE

49

Fig. 5. MPPT process visualized on the power-voltage characteristics of a PV module. In region (I), the voltage set point is increased in steps of V from V 1 to V 1 per iteration, climbing up the power curve toward the MPP in region (III). In region (II), the algorithm lowers the voltage in steps of V toward the MPP. The step size V is exaggerated in this sketchin the prototype devices, the step size is around 0.42% of V o c .

1) P1 < P2 : The set-point voltage is increased V1 = V1 + V (see region (I) in Fig. 5 and label (A) in Fig. 6). The test at label (C) denes the upper limit of the set-point voltage. 2) P1 > P2 : The set-point voltage is decreased V1 = V1 V (see region (II) in Fig. 5 and label (B) in Fig. 6). The test at label (D) denes the lower limit of the set-point voltage. This results in overlapping voltage intervals moving toward the MPP. At the MPP, the algorithm will either settle in an equilibrium position, where P1 = P2 as depicted in region (III) in Fig. 5, or continue to alternately step around the location of the MPP. Making the voltage step V small enough will ensure that the algorithm will come sufciently close to the MPP, even though it will never rest at the MPP itself. In the prototype systems, the width of the voltage step V was in the range of 0.42% of the Vo c , depending on the actual Vo c rating of the module. As a consequence of sudden changes in the irradiance, Vo c of the module can quickly drop below the set-point voltage. In this case, the measured voltage over the PV module will drop below the set-point voltage, and the algorithm would not be able to track the change in the MPP because P1 = P2 = 0 W. Additionally, leakage currents and noise present in the measured signals can guide the algorithm into the wrong direction also at low irradiance levels. Therefore, the decision at label (E) in Fig. 6 resets the algorithm to the starting point V1 = 0 V if such a condition is detected. D. Electronic Load The electronic load shown in Fig. 2 is constructed as a constant-voltage current sink, using an operational amplier as error amplier [15]. The error amplier compares the voltage signal VVPV , which is derived from the module voltage with the set-point value generated by the C, and controls the load resistance accordingly in order to minimize the voltage error. The 8-bit resolution of the PWM signal that is used to derive the setpoint value limits the resolution of the set point to 0.4% of the full-range value. The RC network C14 = 1 F and R17 = 50 k improves the stability of the closed feedback loop.

Fig. 6. Flow chart of the utilized MPPT algorithm. The labels (A)(E) are described in the text.

The output voltage of the error amplier is applied as gateto-source voltage VGS of the power MOSFET Q41 . The drain source channel resistance of the MOSFET in series with the current-sensing shunt resistor R41 acts as a variable resistive load for the PV module. The choice of the MOSFET depends on the voltage and power rating of the PV module that is to be monitored. In commercial power MOSFETs, the drain contact is generally connected to the metal surfaces of the package; therefore, the MOSFET needs to be mounted electrically isolated onto an appropriate heat sink. This results in a higher thermal resistance from the junction to the heat sink which has to be

50

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 2, NO. 1, JANUARY 2012

TABLE I POWER MOSFET PARAMETERS

FETs that are originally optimized for switch-mode operation. We could not detect any indication of instabilities that could have been caused by inhomogeneous current spreading within the individual MOSFETs. The power-handling capabilities of the electronic load can be enhanced by using several power MOSFETs in parallel, as shown in Fig. 7. The advantage of the parallel connection of several MOSFETs is not only the increased current-handling capacity but mostly the enhanced heat conduction to the heat sink. The individual gate resistors labeled R18,[a,b...] prevent detrimental high-frequency oscillations caused by the gate capacitances [25]. Thermally controlled current sharing between the devices cannot be relied on in the case of linear operation of MOSFETs [26]. Instead, additional source resistors labeled R40,[a,b...] need to be added to compensate for the wide variation in gate threshold voltage Vth(GS) normally found in commercial devices. Otherwise, the transistor with the lowest threshold voltage will carry the major part of the load current and, hence, still dissipates most of the power of the electronic load. In two out of four prototype units, the single IRFP3710 MOSFET has later been replaced with three parallel IRF520I MOSFETs in order to counteract problems caused by overheating of the MOSFET. The source resistors were set to R40 a , b , c = 1 and the gate resistors to R18 a , b , c = 1 k. III. RESULTS AND DISCUSSION A. Circuit Simulations

Fig. 7. Power-handling capacity of the MPPT unit can be increased by parallel connection of several power MOSFETs.

considered. In our setup, PV modules with output power from 50 to 80 W at Vo c values from 20 to 80 V were rst connected to MPPT units equipped with a single International Rectier IRFP3710 power MOSFET each. This MOSFET was chosen because of its power- and voltage-handling capabilities even at higher junction temperatures of up to 175 C (see Table I) [21]. The lowest obtainable module voltage Vm o d is limited by the voltage drop over the MOSFET and the shunt resistor R41 = 0.1 Vm o d,m in R41 ISC + RDS,m in (3)

Computer simulations of the electronic load were performed using LTspice [27]. A spice model of a 36-cell 80-W crystalline silicon PV module (Vo c = 21.5 V, Isc = 5.1 A) was connected to a model of the electronic load with a MOSFET of the type IRFP3710 dissipating the electric power of the module, according to the circuit diagram in Fig. 2. The simulation results are visualized in Fig. 8, where the set-point voltage of the electronic load was swept from 0 to 25 V, with the Vo c of the simulated PV module at 21.5 V. The plot in Fig. 8(a) shows the voltage of the PV module Vm o d which follows the set-point voltage (red, dotted line) in the range 0.65 V = Vm o d,m in Vm o d Vo c = 21.5 V. The corresponding drain current IDS through the electronic load is plotted in Fig. 8(b) as a function of the set-point voltage. Above Vo c , the leakage currents through the MOSFET and the residual input currents into the operational amplier (opamp) circuits are negligible. Fig. 8(c) displays the power delivered by the PV module (black, upper curve) and the power dissipated in the MOSFET (red, lower curve). The difference between these two curves corresponds to the power dissipated in the shunt resistor Pshunt R41 Isc = 0.5 W. Finally, Fig. 8(d) and (e) shows the gatesource voltage, i.e., VGS , applied by the feedback circuit and the resulting resistance of the electronic load, respectively. For close-to-short-circuit conditions, the maximum VGS is limited by the supply voltage and the drive capability of the opamp, giving a total load

where RDS,m in is determined by the characteristics of the particular MOSFET and the maximum applied gate voltage VGS . While this limits the ability to measure the full short-circuit curVo c , for our rent Isc of the PV module, with Vm o d,m in 1 V prototype units, it does not affect the performance in determining the MPP. Since the MOSFET in the electronic load is constantly operating in a regime where drainsource resistance RDS is dominated by the channel resistance of the device, the use of a special linear power MOSFET would be preferred [23]. In these devices, the equal current spreading over the individual cells of the device is improved by increasing the resistance of the individual source contacts, while maintaining the general properties of power MOSFETs [16], [24]. However, in the prototype units described here, we used easily available standard power MOS-

ZIMMERMANN AND EDOFF: MAXIMUM POWER POINT TRACKER FOR LONG-TERM LOGGING OF PV MODULE PERFORMANCE

51

Fig. 9. Spice simulation results of the constant-voltage load with two power MOSFETs with slightly different V (th )G S connected in parallel. The load is connected to an 80-W PV module, and the set point for the module voltage is swept from 0 to 25 V. (a) Source resistors were set to R 4 0 a = R 4 0 b = 1 . (b) R 4 0 a , b = 1 were practically omitted. TABLE II SIMULATION RESULTS FOR PARALLEL-CONNECTED MOSFETS IN THE ELECTRONIC LOAD

Fig. 8. Spice simulation result of the constant-voltage load using a single MOSFET. The load is connected to an 80-W PV module, and the set point for the module voltage is swept from 0 to 25 V.

resistance of about Rload = 0.12 . With an increasing set-point voltage, the applied VGS decreases toward the V(th),GS = 3 V of the MOSFET used in the simulation. The achievable dynamic range of load resistances in this setup covers more than ve orders of magnitude, 0.12 Rload 50 k for a gate voltage range of 4 V VGS 3 V. In order to study the possibility of increasing the powerhandling capability of the electronic load by using several discrete MOSFETs connected in parallel, simulations with two MOSFETs of the type IRFP3710 dissipating the electric power of the module were performed in the connection shown in Fig. 7. In order to simulate the inuence of the variation in commercial MOSFET parameters, the Vth(GS) of the two MOSFETs were set to different values, i.e., Q41a : V(th)GS = 2.7 V and Q41b : V(th)GS = 3.3 V, which is well within the expected spread of threshold voltages from the datasheet and measurements on actual devices [21]. The two sets of graphs in Fig. 9 and the data in Table II show the simulation results for two different values

of individual source resistors for the parallel-connected MOSFETs. Setting the value of the source resistors to R40a,b = 1 causes a stabilizing negative feedback of the actual drain current onto the VGS of the individual MOSFETs. The results of the corresponding Spice simulation are displayed in Fig. 9(b). The upper graph shows the individual VGS of the two MOSFETs Q41a,b . For close-to-short-circuit conditions, at low set-point voltages, the attainable maximum value of VGS is now additionally limited by the voltage drop over the source resistors; for two MOSFETs in parallel, we get VGS,m ax VDD Vop R40 ISC = 8 V. 2 (4)

52

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 2, NO. 1, JANUARY 2012

Together with the increased voltage drop over the electronic load that makes it impossible to obtain measurements at low module voltages Vm o d , for the values in this simulation, we get Vm o d,m in R41 ISC + (RDS,m in + R40 ) ISC = 3.2 V 2 (5)

where R41 = 0.1 is the shunt resistor for the measurement of the PV module current. However, under no practical conditions, the MPP can be located at such a low module voltage for the types of PV modules that are considered here. As can be seen in the upper graph of Fig. 9(b), the feedback provided by the individual source resistors causes the VGS for the two MOSFETs to separate when the main feedback of the electronic load circuit reduces the common gate voltage at setpoint voltages above VPV ,m in . The middle graph again shows the individual drain current ID of the two MOSFETs, as well as the sum of both currents, corresponding to the current IPV delivered by the module. Although not exactly equally shared, the difference between the ID of the two MOSFETs is in the order of 10%. MOSFET Q41,a with the lower V(th)GS carries the higher ID . The lower graph of Fig. 9(b) depicts the power delivered by the PV module and the power dissipation in the MOSFETs. Following the unequal current sharing, Q41,a also dissipates more power than Q41,b , but this imbalance also stays within 10%. The sum of the power dissipated in both MOSFET is shown by the dashed line in this graph, as well as the power delivered by the PV module. The signicant difference of up to 17 W between these two curves represents the power dissipated in the two additional source resistors R40a,b . Therefore, these resistors have to be designed accordingly. In Fig. 9(a), the source resistors were set to a value of R40a,b = 1 equivalent to omitting any source resistor as recommended in [15] and [28]. This results in the exact same VGS being applied to both MOSFETs shown in the upper graph of Fig. 9(a). For close-to-short-circuit conditions, the electronic load drives the gate voltage to VGS = 10.5 V as limited by the supply voltage of VDD = 12 V and the voltage drop in the output stage of the opamp TL074. This voltage is sufciently high above the threshold voltage to drive both MOSFET into the conductive state. The lowest obtainable module voltage Vm o d now becomes [also see (3)] Vm o d,m in R41 ISC + RDS,m in ISC = 0.6 V. 2 (6)

power dissipation in the two MOSFETs shown in the lower graph. The full-load power is dissipated in the single transistor Q41a . The resulting self-heating of Q41a will, in practice, lead to a shift of its characteristics, but for power MOSFETs like the IRFP3710 considered here, the transfer characteristics ID (VGS ) shows a positive temperature coefcient for VGS < 6 V [21]. This will, therefore, not counteract the imbalance between the two MOSFETs, unlike the case for switched applications with signicantly higher VGS and negative temperature coefcients of ID (VGS ) [24]. In the lower graph, the electrical power delivered by the PV module is shown as well. The difference between the module power and the sum of the power dissipated in each of the two MOSFETs is the power which is dissipated in the current-sensing resistor R41 . This comparison shows that it is important to use sufciently high values for the source resistors in the parallel connection of power MOSFETs used as resistive loads. The drawback of the use of source resistors is the decreased voltage range limited by the minimum achievable value of the load resistance and the power dissipation in the source resistors, both of which have to be considered in the design process of the electronic load.

B. Field Test At the time of writing, the rst MPPT units have been in operation on the roof of our laboratory for a full year. The units are contained in individual IP65-rated aluminum boxes that are equipped with aluminum heat sinks for the power MOSFETs. Each unit is mounted in the shade behind the PV module to which it is connected. Measurement data have been collected in 10-s intervals for a comparative study on energy yield for different types of PV modules. The co-occurrence of particular high irradiances and power output from the attached PV modules and of less efcient cooling of the power MOSFETs because of high air temperatures and low wind speeds caused an overheating of the IRFP3710 MOSFETs in two of the prototype units after half a year of operation. In these units, the single MOSFETs were then replaced by three parallel-connected IRF520I MOSFETs, as described in Section II-D. Apart from a higher reliability of these units, the change of the MOSFETs did not affect the performance of the measurement systems. Fig. 10 shows typical measurement data of an 80-W module on May 10, 2011, which was a particularly sunny day in Uppsala. The irradiance in the top part of this gure is measured with a Kipp&Zonen CMP11 pyranometer in the tilted plane of the module. The module itself is tilted at a slope of 42 , oriented south. Around solar noon, the glass roof of another building reects sunlight into the pyranometer, causing the spikes in the irradiation data, marked in the graph with the letter A (see the photo in Fig. 11 for the appearance of the reections on the wall beneath the test site). However, the spots of increased irradiance are not covering the entire size of the PV modules, which is responsible for the signicantly smaller response to these spikes seen in the power output of the modules, shown in the middle part of Fig. 10.

Increasing the set-point voltage, the feedback loop tends to reduce the VGS of both MOSFET simultaneously in order to increase the RDS of the channel and thus increase the voltage over the load. As can be seen from the graph in the middle of Fig. 9(a), the current is evenly shared by the two MOSFET for the short-circuit condition, where the sharing ratio only depends on possible variations in the R(on)DS which were neglected in this simulation. However, with increasing set-point voltage, the current sharing becomes unbalanced because of the variation in V(th)GS and MOSFET Q41a with the lower V(th)GS carries practically the full-load current. This is, consequently, also reected in the

ZIMMERMANN AND EDOFF: MAXIMUM POWER POINT TRACKER FOR LONG-TERM LOGGING OF PV MODULE PERFORMANCE

53

Fig. 12. Closeup of the dynamic reaction of the MPPT unit to the shading of a module. The shading is caused by a nearby chimney, and the measurement data are collected in 10-s intervals.

Fig. 10. Measurement data of an 80-W PV module on a clear day. The global irradiance at the top is measured in the plane of the module. The graph in the middle shows the electrical power output of the module, while the lower graph shows the corresponding load resistance constituted by the MPPT unit; markers represent individual measurement points every 120 s. The spikes A in the irradiance are caused by reections from another building, while the dip B in the power output is caused by shading from a chimney.

Fig. 11. Reections of a nearby glass roof and shading of a chimney inuence the irradiance measurements of our test site.

Later in the afternoon, a chimney on the roof casts a narrow shadow over the modules which does not extend to the position of the pyranometer. The effect of this shadow can be seen at the dip, marked B in the middle part of Fig. 10, showing the electrical power output of the module as measured by the MPPT unit. The lower part of Fig. 10 illustrates the effective load resistance Rload = RDS + Rshunt calculated from the voltage VPV and current IPV logged by the measurement system. For this particular module, the load resistance is dynamically adjusted in the range from 4 to 4 k between sunrise and sunset. Fig. 12 shows the effect of this shadowing event on another module in more detail. It takes about 2 min for the shadow of the chimney to reach the maximum coverage on the module, while the module remains in the shadow for about 40 min.

During the time of the maximum shadowing, the module still delivers about 5 W of electrical power generated by diffuse illumination. The MPPT unit adjusts the load resistance from 8 before the shadow hits the module to up to 60 during the time of shadowing. In order to observe the reaction of the MPPT units to sudden changes in irradiance of the attached PV module, a shadowing test was conducted. The result of this test with a crystalline silicon PV module is displayed in Fig. 13. Out of the 36 silicon solar cells of the module, two were quickly and simultaneously covered with a sheet of translucent paper (80 g/m2 ). During the test, the sampling rate of the measurement signals was set to a faster rate of one sample every 5 s. The irradiance during the experiment was stable at 1015 W/m2 . Within one sample interval of 5 s, a new stable operating point is found, and the load resistance has changed from 3.2 under full illumination to 19 under partial shadowed conditions. At the same time, Vm p has increased from 16 V under full illumination to 20 V under partial shadowed conditions. This corresponds to the location of a local maximum in the power output, which is marked as 2b in the simulated characteristics of a shadowed silicon PV module of similar construction in Fig. 14. According to the simulated characteristics, the global maximum, which is labeled 2a in Fig. 14, is found at a signicantly lower voltage of Vm p,2a 10 V with an output power of Pm p,2a 40 W, twice as high as the power in point 2b. The reason for the tendency of the MPPT algorithm during this test to track the local maximum at the higher voltage lies in the positive slope dP/dV of the power curve at the voltage of the original MPP before the shadowing, which is marked as point 1 in Fig. 14. However, this is only the case for very abrupt changes in the irradiance of a few of a modules cells. For practical purposes, the algorithm has proven to be sufciently fast in adapting to natural changes in irradiance, as shown in the case of the shading by the chimney in Fig. 12. Furthermore, a different algorithm could be implemented in the

54

IEEE JOURNAL OF PHOTOVOLTAICS, VOL. 2, NO. 1, JANUARY 2012

Fig. 13. Dynamic reaction of the MPPT unit to the shading of a module. In a crystalline silicon PV module, two out of 36 cells were shaded with a piece of paper, and the measurement data are collected in 5-s intervals.

tions. It can also give valuable feedback to the manufacturer of PV modules regarding long-term reliability under certain test conditions. Using a small C, a simple MPPT algorithm, and a power MOSFET as variable electronic load, we have constructed a tool to monitor the outdoor performance of individual PV modules. In spite of its simplicity, the measurement unit has proven its long-term stability and reliability during one year of operation in the form of prototype units at ambient temperatures between 25 C and +30 C. The use of a constant-voltage electronic load linearizes the control problem and circumvents the variability of the threshold voltage of power MOSFETs. This allows the utilization of the full range of the output signal of the DAC and gives the best possible resolution along the voltage axis of the characteristics of the PV module. The presented circuit is fast enough to track quick changes in irradiance and power output caused by partial shading of the attached PV modules. However, because of the implemented MPPT algorithm, only a local maximum in the power output of the shaded PV module is found, but this could be taken care of by implementing a different control strategy in the software. The dynamic range of the measurement signal is sufcient to reasonably cover daily and seasonal variations in power output. The adaptation to the current and voltage characteristics of different modules is easily accomplished by the dimensioning of two resistive voltage dividers. For a reliable construction of the electronic load, it is essential to take the worst case of ambient temperature and power output of the connected PV module into consideration. In order to accomplish the effective removal of the generated heat from the junction of the power MOSFET to the heat sink, the parallel connection of several smaller MOSFETs has proven to be advantageous over a single MOSFET of higher power rating. In order to compensate for variances in the threshold voltage of these devices, the use of separate source resistors is necessary.

ACKNOWLEDGMENT The authors would like to thank A. Girard for his valuable help in the assembly of the MPPT units.
Fig. 14. Simulation result of the partial shading of a PV module. In a crystalline silicon PV module, the irradiance was lowered from 1000 to 200 W/m 2 for two out of 36 cells, simulating the shading with a piece of translucent paper. The dashed curve shows the unaffected IV curve, while the solid curve shows the partially shaded case with two distinct local maxima in the power output (marked as 2a and 2b).

REFERENCES
[1] A. Virtuani and H. M llejans, E. Dunlop, Comparison of indoor and outu door performance measurements of recent commercially available solar modules, Prog. Photovoltaics: Res. Appl., vol. 19, pp. 1120, 2011. [2] E. van Dyk, T. Strand, and R. Hansen, Technical evaluation of two 6-kW mono-Si photovoltaic systems at the National Renewable Energy Laboratory, in Proc. 25th IEEE Conf. Rec. Photovoltaic Spec. Conf., May 1996, pp. 15331536. [3] S. Pietruszko, B. Fetlinski, and M. Bialecki, Analysis of the performance of grid connected photovoltaic system, in Proc. 34th IEEE Photovoltaic Spec. Conf., Jun. 2009, pp. 4851. [4] E. Koutroulis and K. Kalaitzakis, Development of an integrated dataacquisition system for renewable energy sources systems monitoring, Renewable Energy, vol. 28, no. 1, pp. 139152, Jan. 2003. [5] K. Emery, The rating of photovoltaic performance, IEEE Trans. Electron Devices, vol. 46, no. 10, pp. 19281931, Oct. 1999.

software to avoid the locking to local maxima, as described in the relevant literature (see, e.g., [20]). IV. CONCLUSION Outdoor testing and monitoring of PV modules is an important part in the evaluation of the economic and technological viability of different types of modules at different loca-

ZIMMERMANN AND EDOFF: MAXIMUM POWER POINT TRACKER FOR LONG-TERM LOGGING OF PV MODULE PERFORMANCE

55

[6] O.-M. Midtgard, T. Saetre, G. Yordanov, A. Imenes, and C. L. Nge, A qualitative examination of performance and energy yield of photovoltaic modules in Southern Norway, Renewable Energy, vol. 35, no. 6, pp. 12661274, Jun. 2010. [7] V. Agarwal and S. Jain, New current control based MPPT technique for single stage grid connected PV systems, Energy Convers. Manage., vol. 48, no. 2, pp. 625644, Feb. 2007. [8] M. F. Ansari, S. Chatterji, and A. Iqbal, A fuzzy logic control scheme for a solar photovoltaic system for a maximum power point tracker, Int. J. Sustainable Energy, vol. 29, no. 4, pp. 245255, 2010. [9] S. Brunton, C. Rowley, S. 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. [10] L.-R. Chen, C.-H. Tsai, Y.-L. Lin, and Y.-S. Lai, A biological swarm chasing algorithm for tracking the PV maximum power point, IEEE Trans. Energy Convers., vol. 25, no. 2, pp. 484493, Jun. 2010. [11] N. Dasgupta, A. Pandey, and A. Mukerjee, Voltage-sensing-based photovoltaic MPPT with improved tracking and drift avoidance capabilities, Solar Energy Mater. Solar Cells, vol. 92, no. 12, pp. 15521558, Dec. 2008. [12] O. Lopez Lapena, M. Penella, and M. Gasulla, A new MPPT method for low-power solar energy harvesting, IEEE Trans. Ind. Electron., vol. 57, no. 9, pp. 31293138, Sep. 2010. [13] G. Yu, Y. Jung, J. Choi, and G. Kim, A novel two-mode MPPT control algorithm based on comparative study of existing algorithms, Solar Energy, vol. 76, no. 4, pp. 455463, 2004. [14] T. Esram and P. Chapman, Comparison of photovoltaic array maximum power point tracking techniques, IEEE Trans. Energy Convers., vol. 22, no. 2, pp. 439449, Jun. 2007. [15] A. Garrig s and J. M. Blanes, Power MOSFET is core of regulated-dc o electronic load, EDN Mag., pp. 9293, 2005. [16] A. Sattar and V. T. V. Tsukanov, MOSFETs withstand stress of linearmode operation, Power Electron. Technol., vol. 33, pp. 3439, Apr. 2007. [17] R. Severns, in MOSPOWER Applications Handbook, J. Armijos, Ed. Santa Clara, CA: Siliconix, 1985. [18] Datasheet ATtiny24/45/85, Atmel Corp, San Jose, CA, 2011, rev. 2586NAVR-04/11. [Online]. Available: http://www.atmel.com [19] D. Lancaster, Active Filter Cookbook, 2nd ed. ed. Amsterdam, The Netherlands: Elsevier, 1995. [20] H. Desai and H. Patel, Maximum power point algorithm in PV generation: An overview, in Proc. 7th Int. Conf. Power Electron. Drive Syst., Nov. 2007, pp. 624630. [21] Datasheet and Spice Model of IRFP3710 (rev. 2011). Int. Rectier, El Segundo, CA [Online]. Available: http://www.irf.com [22] Datasheet of IRF520, SGS-Thomson Microelectronics (1993). [Online]. Available: http://www.st.com [23] Y. Kuai and S. Yuvarajan, An electronic load for testing photovoltaic panels, J. Power Sources, vol. 154, no. 1, pp. 308313, 2006.

[24] B. J. Baliga, Power Semiconductor Devices. Boston, MA: PWS, 1996. [25] N. Mohan, T. M. Undeland, and W. P. Robbins, Power Electronics, 3rd ed. New York: Wiley, 2002. [26] J. M. Jacob, Power Electronics: Principles and Applications, 1st ed. Albany, NY: Delmar, 2001. [27] LTspice IV. Linear Technol. Inc. (2011). [Online]. Available: http://www. linear.com/designtools/software/ [28] V. Leite and F. Chenlo, An Improved Electronic Circuit for Tracing the I-V Characteristics of Photovoltaic Modules and Strings, presented at the Proc. Int. Conf. Renewable Energies Power Quality, Granada, Spain, 2010.

Uwe Zimmermann (M06) received the M.Sc. degree in solid-state physics from the University of Kiel, Kiel, Germany, in 1997 and the Ph.D. degree in solidstate electronics from the Royal Institute of Technology, Stockholm, Sweden, in 2003. During Ph.D. studies, he was involved in research on high-voltage silicon carbide diodes. Since 2003, he has been a Researcher of Cu(in,Ga)Se2 based thinlm solar cells with Uppsala University, Uppsala, Sweden, where he is involved in teaching and R&D activities in photovoltaics and electronics. Between 2005 and 2008, he was part-time employed by the thin-lm solar cell manufacturer Solibro.

Marika Edoff received the M.Sc. degree in electrical engineering and the Ph.D. degree in solid-state electronics from the Royal Institute of Technology, Stockholm, Sweden, in 1990 and 1997, respectively. Since 2003, she has been leading research activities on solid-state thin-lm solar cells with Uppsala University, Uppsala, Sweden. She was one of the four founders of Solibro, where she has also been part-time employed since 2005. Her research interests include Cu(In,Ga)Se2 based thin-lm solar cells with a focus on material synthesis and device characterization.

Вам также может понравиться