A Wireless Self-Powered I-V Curve Tracer for On-Line Characterization of Individual PV Panels

The behavior of a photovoltaic (PV) generator under specific irradiance and temperature conditions is mainly described by its current–voltage (I-V) characteristic. Therefore, the I-V curve tracing has to be considered the most accurate and effective diagnostic tool for the proper identification of PV panels’ failures. In this article, an innovative I-V curve tracer for individual PV panels is presented. It ensures the following advantages. 1) An optimized I-V curve measurement due to an innovative tracing algorithm. 2) In-situ characterization during normal operation owing to a proper disconnection circuit. 3) Full portability thanks to a Li-ion battery power supply and a Bluetooth communication. The proposed tool has been designed and realized. The resulting prototype has been used to measure the I-V curve of a PV panel both in uniform and partial shading conditions.

A Wireless Self-Powered I-V Curve Tracer for On-Line Characterization of Individual PV Panels Monica De Riso , Ilaria Matacena , Pierluigi Guerriero , Member, IEEE, and Santolo Daliento Abstract-The behavior of a photovoltaic (PV) generator under specific irradiance and temperature conditions is mainly described by its current-voltage (I-V) characteristic.Therefore, the I-V curve tracing has to be considered the most accurate and effective diagnostic tool for the proper identification of PV panels' failures.In this article, an innovative I-V curve tracer for individual PV panels is presented.It ensures the following advantages.1) An optimized I-V curve measurement due to an innovative tracing algorithm.2) In-situ characterization during normal operation owing to a proper disconnection circuit.3) Full portability thanks to a Li-ion battery power supply and a Bluetooth communication.The proposed tool has been designed and realized.The resulting prototype has been used to measure the I-V curve of a PV panel both in uniform and partial shading conditions.

I. INTRODUCTION
W ITH the rapid escalation of the energy crisis, there is an increasing need to generate energy from renewable sources.In this scenario photovoltaic (PV) systems play a significant role.Monitoring and diagnostic tools are usually employed to ensure the optimum behavior of the PV field, thus leading to a fast return on investment.
An effective monitoring and diagnostic strategy should be able to guarantee the following features: fault detection, fault localization and classification, and energy loss quantification.Nowadays, the common diagnostic techniques are those based on imaging solutions, such as visual inspection and infrared thermography [1], [2], [3], and those based on the acquisition of the electrical quantities, such as current and voltage.The former has high performance in terms of fault detection and localization, but it does not provide an accurate fault classification and energy loss quantification.The latter technique, involving the use of a specific device called current-voltage (I-V) curve tracer, can be used to classify the fault and quantify the resulting energy loss on the faulty panel previously detected by the imaging solution.Unfortunately, most commercial I-V curve tracers are not able to perform on-line measurement.As a consequence, long and costly in-situ inspections are required, thus resulting in the shut-down of the PV plant and string disassembly.
In this work, a real-time I-V tracer for on-line measurements operating at panel level is proposed.It enables online acquisition of the I-V curve, i.e., during its normal operation, thanks to a disconnection system, allowing the string current to bypass the panel during the measurement time interval.Moreover, the tool is self-powered and wireless, hence the monitoring system does not require additional cabling.The measurement can be triggered on-demand by the user and the data are collected using a low power Bluetooth protocol.
The rest of this article is organized as follows.In Section II a background of the I-V tracing techniques available in literature is presented, in Section III a description of the proposed monitoring tool is given, with emphasis on the adopted circuit topology as well as the algorithm developed.In Section IV, the experimental results obtained from a prototyped I-V tracer are shown.Finally, Section V concludes this article.

II. BACKGROUND
The I-V curve measurement is performed connecting a variable load to the electrical terminals of the PV panel, thus forcing the operating point to span the whole I-V characteristic.The variable load can be obtained through different approaches that can be clustered in the following categories.
3) Active load based on dc-dc converters.4) Active load based on linear circuits (transistor-based).It is worth noting that an accurate I-V tracking is supposed to ensure the following features.
3) Even distribution of the points throughout the I-V curve regions.
In the following, the abovementioned approaches will be described in detail.

A. Variable Resistance
In the approaches based on variable resistive loads, the PV generator is loaded by a configurable matrix of resistances.
Each configuration corresponds to a specific global resistance value and forces the PV generator in a different operating point.The proper tuning of the load configuration allows to collect operating points along the whole I-V curve.As matter of example, the authors in [5] proposed a low cost I-V curve tracer based on the parallel combination of 48 different resistors and mechanical relays.Unfortunately, the number of matrix configurations are limited, and the values of the global resistance are fixed.Consequently, the curve tracing results in a small number of points which are not uniformly located along the I-V characteristic.Nevertheless, the approach based on variable resistance is widely used because it guarantees static and highspeed measurements.The solutions provided in [6] and [7] are indeed capable of sweeping the entire I-V curve taking each operating point in 1 ms.However, an important aspect to consider is that each resistor has to be rated for the maximum PV generator power, thus resulting in an increase of the system dimension and weight.For this reason, the I-V curve tracers based on variable resistance are employed only on small power rated PV panels.

B. Capacitive Load
As an alternative, [8], [9], [10], [11] proposed a simple and effective approach to perform the I-V curve tracing of a PV generator, which exploits a capacitive load instead of a pure resistive one.Initially a fully discharged capacitor is connected to the generator terminals, acting as a current source (i.e., it provides the short-circuit current).As the capacitor charges, the operating voltage increases while the charging current decreases.The measurement is concluded as the charging current falls to zero and the operating point reaches the open-circuit voltage.
This approach is typically implemented in commercial portable I-V tracers, thanks to its good features in terms of scalability and low specific volume and weight.This feature, in fact, permits commercial devices, such as DS-1000 [12], to measure the I-V curve of PV generators up to 100 kWp.However, the amount of energy that the capacitive load stores during the measurement time interval must be dissipated on a resistive load afterward, affecting the minimum time interval between two consecutive measurements.The capacitor charging time mainly depends on irradiation.Therefore, the capacitance value must be sized to meet the standard requirements in terms of measure time duration [28] even in case of low irradiation level.Moreover, the capacitive approach does not ensure a static I-V curve tracing since the load voltage continuously changes during the measurement, thus being affected by hysteretic effects [4].In addition, it is not possible to reach a uniform distribution of the points along the I-V curve because, as mentioned previously, the voltage slope depends on the charging current.

C. Active Load Based on Dc-Dc Converters
An active load can be used instead of the abovementioned passive ones.In particular, a dc-dc converter could be exploited to change the input resistance as function of the duty-cycle, as well explained in [13].More in detail, the boost converter is able to show an input resistance lower than the load resistance, while the buck converter is able to increase the input resistance with respect to the load.A combination of these two features can be easily accomplished using a dc-dc converter based on buck-boost topology.The results shown in [14], demonstrate that an I-V curve tracer based on boost converter is not capable of sweeping the portion of I-V curve near the short circuit condition.By contrast, the solutions provided in [15] and [16], based, respectively, on Cuk and SEPIC converters, solve this issue.
Since a dc-dc converter can reach and remain in a desired operating point in the PV curve; it guarantees a quasistatic measurement.In addition, its high dynamic behavior ensures a fast-sweeping time.On the other hand, the power switching leads to an inherent ripple in the operating voltage and current degrading the measurement accuracy.The reduction of the ripple typically requires the oversizing of reactive elements, thus affecting the size and weight of the overall system, and even the dynamics.
It should be considered that dc-dc converters devoted to maximum power point tracking (MPPT) could be exploited to periodically perform the I-V curve tracing (voltage sweep), without requiring any additional circuitry.Indeed, voltage sweep is typically implemented in commercial inverters to improve the MPPT in case of multiple maxima in the P-V curve due to mismatch.Nevertheless, the adoption of this technique for diagnostic purposes impacts on converter design.Those converters are designed to operate close to the MPP, thus they typically exhibit a stable operation for a limited input voltage range (MPPT range) and can experience stability issues when are forced to operate in the flat region of the I-V curve.As a consequence, without a specific design [17], they merely provide incomplete [18] and not static [19] I-V curve measurement.
In the last decades, several MPPT approaches have been proposed, including both low (i.e., centralized and string inverters) and high (i.e., distributed MPPT by means of power optimizers) granularity level techniques.As remarked in introduction, only the high granularity approach is able to match all the requirements of an effective monitoring and diagnostic system.Unfortunately, optimizers are not exempt from the abovementioned issues, likely affecting the performance in terms of classification and energy losses quantification.

D. Active Load Based on Linear Circuit
The I-V curves of a PV panel can be alternatively obtained using an electronic load based on transistors operating in linear region.The solutions presented in [20], [21], [22], [23], [24] are based on the exploitation of transistors as electronic load, such as MOSFET and bipolar junction transistor (BJT).The working point is given by the intersection between the I-V curve of the solar panel and the load curve that corresponds to the I-V characteristic of the transistor at a given gate to source voltage V GS [20] in case of MOSFET and base to emitter voltage V BE in case of BJT [22].This solution has the great advantage of a high-speed measurement that is done in static conditions because the device can quickly modify its resistance.
Unfortunately, the dependence of the operating resistance exhibited by the transistor on the control signal (i.e., V GS and V BE ) dynamically changes during the measurement time interval, thus not favoring a uniform distribution of the points along the I-V curve.To tackle this issue, a proper feedback network for the set-points in current and voltage can be employed, as presented in [24].Moreover, it should be noted that the entire power produced by the panel is dissipated by the transistor, thus requiring the need of additional heat-sinks.

E. Summary
A summary of the main differences among the different approaches listed in the previous subsection is reported in Table I.
The table highlights the following aspects.1) None of the mentioned solutions has the capability of performing an on-line measurement, meaning that the I-V curve tracing does require the shut-down of the PV plant and the eventual string/array disassembly.
2) The majority of the considered I-V curve tracers randomly distribute the measurement point along the I-V curve, thus reducing the detectability of potential faults.3) Only a few are able to ensure static measurements.On the contrary, the proposed I-V curve tracer embeds an effective disconnecting circuit, aiming to measure the I-V curve of the PV panel under test during its normal operation in the string.It embeds an innovative control algorithm assuring an even distribution of the points along the I-V characteristic.Moreover, it is based on linear topology.It does not require large and expensive reactive components, nor does it suffer from switching noise.The proposed I-V curve tracer also guarantees that each operating point of the I-V characteristic is acquired in static conditions, fulfilling all the requirements listed at the beginning of Section II.

III. SYSTEM DESCRIPTION
The proposed tool for I-V curve tracing embeds an active embeds an active load, a sensing section, a control unit, a communication module, a disconnection circuit and a power supply section.The block diagram of the tool is reported in Fig. 1.

A. Active Load
The active load implemented in the proposed monitoring tool allows imposing a variable current to the PV panel.
The amount of current that is forced to flow through the panel is linearly controlled by an external variable voltage source.To better understand its working principle, consider the circuit reported in Fig. 2. The circuit is made of two BJTs (Q 1 is PNP whereas Q 2 is NPN), a load resistor R and a pull-up resistor R BIAS .
The current generated by the panel, I PV , depends on the base-collector voltage applied to Q 1, which is equal to V DAC , according to (1) obtained from the Kirchhoff voltage law where Assuming that all the transistors are identical and working in forward active region, the voltage V BE2 can be considered equal to the voltage V EB1 .Therefore, the following equations can be inferred: Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
Supposing that the forward current gain α F is closer to 1, the emitter current can be considered equal to the collector current.Hence, the current that is forced to flow through the panel (i.e., I PV ) is linearly controlled by applying the external voltage V DAC .
It should be clarified that the bipolar transistors do not always work in forward active region during the I-V curve sweeping.To better explain this aspect, Fig. 3(a) and (b) reports the SPICE simulation results of the active load depicted in Fig. 2, where the BJTs are replaced with Darlington transistors due to their capability of providing higher current gain and higher input impedance.It is assumed that the PV panel under test is uniformly irradiated with open circuit voltage V OC = 9.6 V and short circuit current The load resistor R is set equal to 1 Ω.In Fig. 3(a), the blue curve depicts the external voltage V DAC applied to the basecollector of Q 1 , whereas the red curve depicts the PV current.V DAC is stepped from 0 to 7 V to sweep the I-V characteristic in eight operating points from V OC to I SC .In V OC , Q 1 works in the active forward region whereas Q 2 is in cutoff.As V DAC increases, the PV operating point moves towards the short circuit condition and both Darlington transistors are forced to work in forward active region.As can be seen in Fig. 3(a), V DAC and I PV match according to (3) as long as V DAC is less than 5 V.For larger voltage, Q 2 is pushed to work within the saturation region.This condition is well shown in Fig. 3(b), where the collector-emitter voltage of Q 2 saturates at 0.768 V.
As it can be deduced from (3), the sensitivity of the proposed circuit (i.e., dI PV /dV DAC ) depends on the resistance value R. In particular, supposing that the increment in the external control voltage (ΔV DAC ) is constant, a high resistance value results into a small current perturbation, whereas a low-resistance value corresponds to a larger current variation.
To better understand how the sensitivity of the circuit proposed in Fig. 2 impacts on the I-V curve tracing, consider the black solid curve reported in Fig. 4.This curve is a typical I-V characteristic arising from a uniform irradiated solar panel.It consists of a vertical branch, where small voltage increments correspond to large current increments, and a horizontal branch, where large increments in the voltage result into small increments in current.Consequently, the number of measured points (red dots in Fig. 4) could be either insufficient for scanning the horizontal branch or overabundant in the case of the vertical branch.This is the main drawback of using the circuit in Fig. 2. For such reasons, the circuit can be improved by incrementing the number of BJT branches, each containing a different resistance value such that I PV is finely adjusted by the combination of multiple current steps.The schematic of the active load proposed in this study is reported in Fig. 5(a).As can be seen, it is formed by two Darlington branches.The I PV current is composed by the sum of the current that flows through Branch A (the current term I 1 ) and through Branch B (current term I 2 ).With reference to Fig. 4, if R 1 is larger than R 2 , the current steps resulting from a constant voltage increment ΔV DAC1 (ΔI 1 ) can be used to trace the vertical portion of the I-V curve, whereas the current steps imposed by ΔV DAC2 (ΔI 2 ) can be used to trace the horizontal portion.The correct operation of this circuit requires the recognition of the knee, evaluated as the slope of the I-V curve in the MPP.In fact, when the slope of the I-V curve is larger than the knee, the point to be measured lies on the vertical branch, establishing an increase of V DAC1 .On the contrary, when the slope of the I-V curve is smaller than the knee, V DAC2 must be increased whereas V DAC1 must be kept constant.More details regarding the implemented algorithm for the I-V curve tracing are given in Section F. In addition to the Darlington branches, the circuit in Fig. 5(a) contains a third branch made of a single transistor (T 1 ) for the short circuit measurement.In fact, it is to be remarked that the Darlington branches reported in Fig. 5(b) are not able to impose the short circuit condition due to the presence of the saturation voltage across Q 24 .Therefore, T 1 is activated whenever the short circuit measurement must be performed.
The active load proposed in this article is a linear circuit.Therefore, the entire PV power is dissipated by the circuit during the measurement time interval.The amount of power to dissipate might become concerning as the PV panel rated power increases, inducing the Darlington transistors to work outside their safety operating area.A solution can be to exploit the modularity of the circuit.A single Darlington branch can be split into multiple branches whose global resistance value is equal to the desired one.As matter of example, assume a Darlington branch with R = 1 Ω.This latter can be obtained by parallelizing two identical Darlington branches with R = 2 Ω, each dissipating half of the power.This feature makes the proposed active load a scalable solution, allowing the I-V curve tracer to handle even higher power dissipation.Nevertheless, the main drawback is the increased weight and occupied area.

B. Disconnection Circuit
Since the solar panel must be bypassed during the measurement time interval, the I-V curve tracer is provided with a specific disconnecting circuit, comprising a MOSFET switch and a bypass diode (see Fig. 1).The control unit activates the disconnection circuit by turning off the MOSFET switch.The switch disrupts the current flow through the panel, thus forcing the activation of the bypass diode.During the I-V curve tracing, the on-board bypass diode provides an alternative path to the string current, thus preventing the interruption of the string energy generation.The amount of energy lost during the panel disconnection depends both on the duration and the frequency of the measurement.Since the proposed I-V tracer requires less than 1 s and the number of measurements over the day is limited, the energy loss can be considered negligible.In addition, even though the activation of the disconnection circuit causes the MPPT to run additional cycles, it does not affect the stability of the power system.
The disconnecting circuit is specifically designed to run obstruction to electromagnetic interference, thus improving the quality of the I-V curve tracing.It is well-known that PV systems are hugely impacted by the noise generated by the commutations of the power electronic switches of inverters and dc/dc converters [25].Therefore, electromagnetic compatibility issue is of big concerns.Since the proposed I-V curve tracer is based on linear circuit, it does not inject switching noise to the PV string.In addition, the disconnecting circuit guarantees full rejection of any electromagnetic interference coming from the PV string, thus not impacting the quality of the I-V curve measurement.

C. Power Supply Section
The power supply section is provided with a Li-ion battery, which is autonomously recharged by the panel.Such feature guarantees full portability of the proposed tool, thus not requiring additional wiring.

D. Control Unit
The control unit is based on microcontroller, and it is in charge to perform the following tasks.
1) Enable the disconnection circuit before starting the measurement.2) Provide the proper voltage signals (V DAC1 and V DAC2 ) to the active load using a multichannel digital to analogue converter (DAC).3) Simultaneously acquire the current sensor and voltage sensor outputs through two independent 12-bits analogue to digital converters (ADCs) with a sampling rate of 7 MS/s.4) Send the data to a remote controller.
Authorized licensed use limited to the terms of the applicable license agreement with IEEE.Restrictions apply.
The data are digitally filtered to remove the noise and partially processed on-chip to assure that they are acquired in static conditions.

E. Communication Module
The communication module is Class 1 Bluetooth device.It enables a wireless and low power transmission of data up to 1000 m at 18 dBm [26] between the on-board controller and the remote controller.The range capability makes the proposed I-V curve tracer employable even in large PV installations.

F. Control Algorithm
The control unit implements an innovative algorithm that guarantees an even distribution of the measured points along the I-V characteristic.The proposed algorithm performs the following the steps.
1) Prior measurement of I SC and V OC .
2) Calculation of the distance between the points along the vertical branch as and along the horizontal branch as where N is the total number of points to acquire, given as input parameter by the user.This step is performed assuming a square-shaped trace.3) Evaluation of the voltage steps ΔV DAC1 corresponding to ΔI DIST and imposing ΔV DAC2 equal to the minimum step increase.4) Acquisition of the second measured point (k = 2) corresponding to V DAC1 = ΔV DAC1 and V DAC2 = 0 with the assumption that it lies on the vertical branch of the I-V characteristic.In fact, V OC is stored as k = 1, whereas I SC is stored as k = N. 5) Identification of the I-V branch.When a new point is acquired, it is needed to identify which Darlington branch must be increased.This task is performed by calculating the slope at the actual working point as where {V(k−1), I(k−1)} is the data point acquired in the (k−1) th iteration and {V(k), I(k)} is the data point acquired in the kth iteration.The slope is then compared to the value of the slope obtained at the MPP.This latter is calculated as [27] knee If knee is greater than slope, the working point lies on the horizontal curve and V DAC2 must be incremented, otherwise, the working point lies on the vertical curve and V DAC1 must be incremented.

TABLE II COMPONENTS USED IN HARDWARE
T 1 [see Fig. 4(a)].On the other side, the open-circuit voltage is measured when V DAC1 = 0, V DAC2 = 0, and T 1 is OFF.The measurement procedure terminates when the number of points acquired reaches the number of points N set by the user.Moreover, in case of partial shading condition, the proposed algorithm is able to detect multiple knees, assuming that the slopes are equal to the one obtained in case of uniform light condition, as calculated in (7).

G. Prototype
The monitoring tool discussed in this section has been designed for monitoring PV panel with I SC of up to 10 A and V OC of up to 40 V.The resulting prototype is pictured in Fig. 7.The main components embedded in the prototyped board are listed in Table II.The proposed I-V curve tracer is also compared with some commercial portable devices in terms of I-V range, maximum sweeping time, capability of performing on-line measurement and cost.The information is collected from datasheets and website of the manufacturers.From Table III, it is clear that most of the commercial I-V tracers are conceived for PV strings and they suffer from the following limitations.

TABLE III COMPARISON WITH COMMERCIAL DEVICES
1) The PV system must be turned OFF.The shutdown implies loss of power generation.2) I-V curves measured at string level do not allow localizing the failure.3) They typically lack portability because based on heavy and bulky capacitive load.The need for capacitors makes them more expensive than the proposed device (the cost is €355) and since the sweeping time depends on the irradiance level, it can reach up to 5 s.By contrast, the proposed I-V tracer is conceived for the monitoring at panel level and it overcomes many limitations listed previously.The proposed device performs the measurement at panel level in less than 1 s.In addition, thanks to the disconnecting circuit, it prevents undesired stop of the PV string.The proposed device can be used in combination with other low-cost technique, such as unmanned aerial vehicles for infrared thermography.Once the faulty panels are identified, the I-V tracer has to be installed on one faulty panel at a time to obtain an accurate fault classification and energy loss quantification, thus resulting to be economically convenient even in large PV installations.

IV. EXPERIMENTS
The presented prototype has been employed to perform the I-V curve measurements shown in this section.The experiments are conducted on a c-Si 120Wp photovoltaic panel PV-MF120EC3 [28] installed on the rooftop of the Department of Electronics at the University of Naples Federico II.The experiments have been conducted both in the case of uniform light irradiance and partial shading condition.The performance of the proposed I-V tracer is tested in terms of sweeping time and the capability of evenly distributing the measured point along the I-V characteristic.For this latter, the relative standard deviation (RSD) of the distance in voltage and current between two adjacent points is given as performance index.

A. Validation of the Disconnection Circuit
As mentioned in Section III, the disconnection circuit aims to disconnect the PV panel under test during the curve tracing.In order to prove the effectiveness of the proposed circuit, an experimental setup has been arranged on a 1 kWp PV array.The proposed monitoring tool has been connected to the terminals of the PV panel under test and the terminals of the whole PV string, according to Fig. 1.The PV string has been indeed connected to a ZS electronic load.The string current I S and the enabling signal V D , which turns ON/OFF the disconnection circuit, are measured by means of digital oscilloscope Tektronix DPO3034.The experiment shows that I S , depicted in blue in Fig. 8, keeps flowing thanks to the activation of the internal bypass diode.
The red curve, instead, represents the disconnection signal V D , which is in charge of disconnecting the PV panel under test when set to the logic low level.Finally, the green curve is the current flowing through the PV panel under test, I PV .As it can be seen, it reproduces the transient behavior from the open circuit condition to the short circuit condition, whereas it is equal to IS outside the measurement window.This experiment clearly demonstrates that the proposed monitoring tool is capable of performing the I-V curve tracing in less than 1 s.This feature is required by the standard [29] in order to assume constant irradiance and temperature conditions.

B. Uniform Light Irradiance Condition
The first test is performed to evaluate the algorithm performance under uniform light irradiance.The experiments were repeated over the course of a day at different times in order to perform the measurement under different levels of irradiance.In addition, the number of points N is set equal to 128 for all the acquisitions.In Fig. 9, the results obtained under four different solar irradiance conditions are shown.
All curves consist of a number of points between 75 and 81.In principle the number of points should be equal to N, which is true only in case of ideal square-shaped I-V curve.In reality, the I-V curve results in a rounded shape near the knee, causing the number of acquired data to be always less than N.
Each data point shown in Fig. 9 is the result of a mean moving average filter performed on 16 consecutive samples acquired by the ADC, aiming to increase the signal-noise ratio.
As preannounced in the previous section, the implemented algorithm evenly distributes the measurement points on the vertical and horizontal branch.In order to understand such feature, the distances in current and voltage between adjacent samples of the I-V curve #2 are traced in Fig. 10.
The total number of samples is 75.The red diamond points represent the distance between two adjacent current points while the black square points represent the voltage difference between two adjacent points.The I-V characteristic is acquired starting from the open circuit condition; hence, the first 37 samples correspond to the vertical branch.If the red points are considered, the mean value and the standard deviation of the difference between two consecutive points is, respectively, equal to 0.11 and 0.008 A in the yellow region, resulting in a RSD of less than 10%.Similar considerations can be applied to the black points lying in the white region.The mean value and the standard deviation of the distance in voltage between two adjacent samples is, respectively, 0.38 and 0.029 V. Therefore, the RSD is roughly equal to 7%.The mean value (μ) and the standard deviation (σ) of the points depicted in Fig. 8 are collected in Tables IV and V.As it can be seen in Fig. 9, the I-V curves are characterized by a lack of points near the short circuit one.In fact, given the value of the resistance R 2 , the minimum measured voltage (i.e., the nearest  point toward the current I SC ) is given by the intersection between the load curve of equation I = (V−Vce sat )/R 2 and the I-V curve (corresponding to V MIN = R 2 I SC + V ce sat ).Nevertheless, such property does not lead to lack of information thanks to a linear trend of the I-V curve near I SC .In Fig. 9, the minimum voltage measured on the purple curve is closer to the short circuit point with respect to the others due to its lower I SC .
In addition, the value of R 2 affects the minimum measurable slope on the horizontal branch and it should be chosen equal to the value that causes a current variation at least equal to the ADC resolution.

C. Partial Shading Condition
The monitoring of the I-V curve is fundamental to track the state of health of the PV panel.As matter of example, a simple visual inspection of the I-V shape is useful to determine whether the PV panel under test experiences partial shading during its normal operation.When partial shading occurs, the bypass diode that protects the shaded submodule turns ON, resulting in a stepped I-V characteristics.Tracking this kind of events is pivotal for good maintenance.In fact, when the shadows are persistent (such as those caused by dust, leaves, and bird droppings), they induce localized hotspots [30] that accelerate the aging process and resulting in low power generation.
In Fig. 11, an example of data measured under partial shading is plotted.This condition is obtained partially shading some

TABLE VI CURRENT POINTS
cells of the PV panel under test to create small and localized shadows.The patterns are compatible with those projected by environmental obstacles usually present on the rooftop of the buildings, such as chimneys and antennas.
Small-area shadows are indeed considered to be the worst shading scenario, leading to marked yield reduction [31].As it can be seen in Fig. 11, due to the presence of the localized shadow, the current drastically drops from I SC to 1.6 A when the voltage is around 10 V, corresponding to the point where the bypass diode is deactivated.Moreover, it can be easily induced from the shown I-V curve that the PV panel is composed of two submodules, each protected by a bypass diode.Also under shaded conditions, the algorithm proposed in this article is able to evenly distribute the measured points.As in the case of uniform condition, the data points are the results of a mean moving average filter performed on 16 consecutive samples acquired by the ADC.In Fig. 12, the black square points represent the distance in voltage between two adjacent points and the red diamond points depict the distance in current between two consecutive measured points.The yellow zones highlight the regions corresponding to the vertical branches, while the white ones correspond to the horizontal branches.
As it shown in Fig. 12, the points representing the distance in voltage are distributed around 0.35 V in both the white regions with a standard deviation equal to 0.019 V, whereas the mean value of the distance in current is 0.055 A in the yellow shaded regions, with a standard deviation equal to 0.0053 A. In both cases, as already proven in Section B, the RSD is less than 10%.
The mean values and the standard deviations evaluated in the vertical and horizontal branches are reported in Tables VI and VII.It is worth to observe the lack of points near the local MPP in Fig. 11, corresponding to the range of voltage between 5 and 7 V approximately.Such a feature is caused by the assumption that the slope of the I-V curve in the MPP is the same both in uniform and partial shading condition.Therefore, the value obtained from (7) results in being underestimated in case of uniform light irradiance and overestimated in case of partial shading.
Hence, in the latter condition, where only a limiting number of solar cells are responsible of the photogeneration, the knees in the I-V plane results to be sharper with respect to the uniform light condition, thus causing an inaccurate recognition of the I-V branch.

V. CONCLUSION
In this article, an innovative I-V tracer based on Darlington transistors for on-field measurements was presented.This monitoring tool had some powerful features, such as the ability of tracing the I-V characteristics of the PV panel under test during its normal operation by means of a disconnection circuit.This tool was provided with a wireless and low power communication module, thus not requiring additional cabling.The implemented control algorithm was able to regularly distribute the working points along the I-V characteristic, thanks to the exploitation of two Darlington branches.The prototype of the I-V curve tracer was designed in order to test medium-power rated PV panels with I SC of up to 10 A and V OC of up to 40 V.
The effectiveness of the proposed monitoring tool was assessed by means of an experimental campaign carried out on a commercial 120 Wp PV panel.More in detail, the I-V characteristics were measured both under uniform light conditions and partial shading scenario.The data points were properly analyzed to prove the capability of evenly distributing the measured points along the I-V curve, guaranteeing an RSD less than 10%.In addition, the results clearly demonstrated that this I-V tracer can be considered a powerful real-time diagnostic instrument, thanks to the feature of performing the measurement in less than 1 s.

Fig. 1 .
Fig. 1.Functional block diagram of the proposed I-V tracer.

Fig. 3 .
Fig. 3. (a) External voltage applied to the active load V DAC is depicted in red, whereas the photovoltaic current (I PV ) is depicted in red.(b) Collector-emitter voltage of the NPN transistor.

Fig. 4 .
Fig. 4. Qualitative description of the distribution of the measured operating points (red dots) acquired by the proposed I-V curve tracer.The black solid line is the I-V curve of a generic uniformly irradiated PV panel.ΔI 1 is the current step imposed by Branch A [see Fig. 5(b)], used to trace the vertical portion of the curve, whereas ΔI 2 is the current step imposed by Branch B [see Fig. 5(b)] to trace the horizontal portion.The knee defines the separation between the vertical and horizontal side of the I-V curve.

Fig. 5 .
Fig. 5. (a) High-level schematic circuit of the proposed active load.(b) Breakdown of the blocks reported in (a).

Fig. 8 .
Fig. 8. Experimental curves.The blue curve depicts the string current, in green the PV panel under test current is represented, whereas the red waveform is the disconnection signal.

Fig. 12 .
Fig. 12.Current and voltage distances under partial shading.The points represent the difference between the kth sample and the k-1th one, with k = 1..N.

TABLE I COMPARISON
AMONG DIFFERENT VARIABLE LOAD APPROACHES