MEASUREMENT OF PV MAXIMUM POWER POINT TRACKING PERFORMANCE

M. Jantsch1, M. Real, H. Häberlin, C. Whitaker, K. Kurokawa, G. Blässer, P. Kremer, C.W.G. Verhoeve

Working Group 3: PV Systems, Technical Committee 82: Photovoltaics, International Electrotechnical Commission
1
contact: Netherlands Energy Research Foundation ECN, P.O. Box 1, NL-1755 ZG Petten, The Netherlands
Tel: +31.224.56.4178, Fax: +31.224.56.3214, Email: jantsch@ecn.nl

ABSTRACT: Methods to measure the accuracy, error, and efficiency of maximum power point trackers (MPPT) have
been identified and are presented in a schematic way, together with definitions of terms and calculations. These meth-
ods are the result of a review on how international institutes and private industries are determining the MPPT accuracy
and efficiency. The intention of this paper is to invite discussion, and to stimulate other experts to contribute and to
further refine the terms and procedures, as it is intended to generate an IEC-standard.
Keywords: MPPT - 1 : accuracy - 2 : efficiency - 3 : error - 4 : measure -5 : inverter - 6

1. INTRODUCTION
Many different ways exist to track the MPP which can be
Inverter efficiencies typically declared are calculated as the classified as either direct or indirect methods (see table 1).
fraction of AC output divided by DC input power. Inverter Direct methods include algorithms that use measured DC
manufacturers and system installers assume that the invert- input current and voltage or AC output power values, and,
ers are normally working at the maximum power point by varying the PV array operational points, determine the
(MPP) of the I-V curve of the PV array. actual MPP. Adjustment of MPP may occur continuously
or intermittent, and algorithms may well or not include
In practice, there are a number of factors which cause the
artificial MPP search movements.
actual operating point to vary from the true MPP. For
example, devices that use search algorithms to find the Indirect methods are those which use an outside signal to
MPP have to move constantly around this optimal point estimate the MPP. Such outside signals may be given by
thus operating the array off of MPP for some period of measuring the irradiance, the module temperature, the short
time. Search algorithms use finite time and voltage or circuit current, or the open circuit voltage of a reference
current steps that may cause some error. solar cell. A set of physical parameters has to be given,
and the MPP setpoint is derived from the monitored signal.
These MPPT inaccuracies conspire to reduce the conver-
sion efficiency of the PV array, and therefore, the entire Table 1: Overview: MPP tracking algorithms
system.
Maximum Power Point Tracking Algorithms
MPPT performance is important to system designers who
are guaranteeing a certain system performance and need to direct, controlled indirect, derived set-
know all of the system losses as well as to system operators maximum through: point on basis of:
who want to ensure that their system is operating per its maximise power design parameters
specifications. Thus an inverter or separate MPPT certifi- P = I⋅V → max operational parameters
cation should include MPPT performance.
make derivative zero system characteristics
When pressed, inverter manufacturers may claim an MPPT dP
/dV → 0 , dP/dI → 0
accuracy or efficiency, but this value is likely based on the make quotient sum zero
resolution of the MPPT search algorithm, not on a meas- V
/I + dV/dI → 0
ured performance. Appropriate methods for determining
MPPT performance - both for certification purposes and for
field verification - need to be defined, and, along with
1.2. MPPT Accuracy, Error, and Efficiency
them, the terms and calculations to be used.
Static and dynamic factors influencing MPPT behaviour
include:
1.1. MPPT Algorithms
• power (irradiance level),
Maximum power point tracking (MPPT) is performed by
• voltage (temperature; layout including well- or mis-
some battery charge controllers and by most grid connected
matched PV and MPPT voltage ranges),
PV inverters. The principle is to adjust the actual operating
voltage V (or current I) of the PV array so that the actual • fluctuations (clouds),
power P approaches the optimum value PMAX as closely as • PV technology (I-V curve shape)
possible (see figure 1).
• need (battery state of charge, in case of charge con-
I troller with MPPT).
I PMAX
IMAX,
V ⇒ VMAX Three terms can be used to describe how well an MPPT
P performs. They are functions of time (even under static
V conditions, due to MPPT search movements) and of addi-
Figure 1: Maximum Power Point Tracking Principle tional parameters.
Accuracy (static and dynamic) indicates how close to MPP average efficiency ηMPPT.A , i.e. the average of the MPPT
the MPPT operates the PV array and can be defined as a power efficiency ηMPPT.P , is equally weighted over time,
percentage of IMAX, VMAX, or PMAX: regardless of high or low irradiance and power level P(t).
aMPPT.X = X / XMAX
with X = I, V, or P 1.3. MPPT Assessment and Testing Methods Overview

Efficiency indicates the ratio of actual to available PV Measuring MPPT behaviour has its complication, since
array power (a particular case of accuracy) or energy [1]: two devices and their interaction are involved in the meas-
urement - the PV array and the MPPT. The actual operat-
ηMPPT.P = P / PMAX ing voltage and current of the PV array are readily meas-
ured but, it is not easy to determine VMAX and IMAX which
ηMPPT.E = E / EMAX (see chapter 1.2.1.) vary with irradiance, temperature, spectrum and other
Error (static and dynamic) indicates the absolute or relative conditions. Also, some inverters attempt to maximise the
difference between actual and MPP values of voltage, AC output rather than the DC input (PV array) power.
current or power: This approach may result in operating the PV array off of
MPP slightly but increasing the inverter efficiency such
εMPPT.X = X - XMAX (absolute) that the total sun light to AC efficiency is optimised.
or X / XMAX - 1 (relative) Table 2 gives an overview of the identified methods to
with X = I, V, or P measure MPPT performance, which are divided into labo-
ratory (indoor) and field (outdoor) measurements.
Accuracy and efficiency are essentially the same, however,
efficiency, ηMPPT.P, can be used to 'correct' the inverter's Table 2: Overview: MPPT Measurement Methods
conversion efficiency as shown in figure 2. Since the MPPT Measurement Methods
MPPT operates the PV array as a constant voltage or con-
stant current source, voltage or current error better de- Laboratory (Indoor) Field (Outdoor)
scribes what the MPPT is doing. Also, voltage or current assessment under switching between
error for a given MPPT varies only as a function of IMAX static conditions MPPT and I-V tracer
and VMAX whereas efficiency is additionally a function of
the PV array I-V curve shape (fill factor). assessment under using a calibrated
dynamic conditions reference module
PV Array Inverter ηMPPT P assessment of sampling MPPT
PMAX
=100% energetic efficiency input at high speed
<100% P
⇒ ⇒ ⇒ using manual mode
G P Out
V to obtain I-V curve
ηTOTAL = ηPV · ηMPPT.P · ηINV further tests analysing
monitoring data
PMAX[W] P[W] Out[W]
ηTOTAL = ----------------------- · ------------ · ----------
G[W/m2] · A[m2] PMAX[W] P[W]
2. LABORATORY MEASUREMENTS (INDOOR)
Figure 2: Efficiencies of MPPT and Power Conversion
To perform reproducible laboratory measurements, a PV
array simulator is necessary that generates DC power with
1.2.1. Instantaneous vs. Integral Assessment the I-V curve characteristic of a PV array. The exact
The MPPT's influence on the PV system performance simulation of such an I-V characteristic requires either a
depends both on its static performance - how closely it sophisticated control device or a network of diodes and
operates to a fixed MPP - and its dynamic performance - resistors capable of handling large amounts of power. The
how well it responds to changes in MPP. The impact of simulator must be able to simulate an array under a variety
static and dynamic effects can be accounted for using the of conditions (including different fill factors signifying
following terms, calculating integrals from t = 0 to TM : different cell technologies) with satisfactory static accuracy
as well as dynamic small-signal and large-variation re-
MPPT Energetic Efficiency: sponse to high frequencies. The simulator must not inter-
ηMPPT.E = E / EMAX (≤1) act with the MPPT in a way that is significantly different
= 0∫Tm P(t) dt / 0∫Tm PMAX(t) dt from a PV array.
MPPT Energetic Error (i.e. MPPT Energetic Loss):
εMPPT.E = E / EMAX - 1 ( ≤ 0 , indicates loss) 2.1. Assessment of Static MPPT Performance
= 0∫Tm P(t) dt / 0∫Tm PMAX(t) dt - 1
The purpose of this test is to measure the MPPT errors in
MPPT Average Efficiency: voltage ε MPPT.V , current εMPPT.I , and power εMPPT.P , under
ηMPPT.A = 1/Tm · 0∫Tm ηMPPT.P(t) · dt static conditions and as functions of important parameters.
= 1/Tm · 0∫Tm P(t) / PMAX(t) dt
Resulting plots display the MPPT errors as functions of
The energetic performance is correctly expressed by the VMAX and PMAX.
MPPT energetic efficiency ηMPPT.E , whereas the MPPT
The principle of MPPT performance measurements is quite such conditions predominate, this dynamic behaviour is
straight-forward (see figure 3). also an important issue.
PV Array Simulator The purpose of this test is to measure εMPPT.V , εMPPT.I , and
MPPT
I-V Tracer I εMPPT.P , under dynamic conditions and as functions of
⇒ V important parameters.
P
I-V Measurements are executed under varied parameters and
⇓ ⇓ defined conditions. The dynamic parameters could follow
I I triangle-shaped signal functions (sweeps), where dVmax/dt or
PMAX P dPmax
/dt can be varied, together with sweeping through
V V VMAX or PMAX at the same time. Responses to step func-
ηMPPT tions (e.g. stepping from 10% to 100% of nominal power)
Figure 3: Static laboratory measurements are however easier to achieve. Also are most MPPT algo-
rithms quite slow (response time of many seconds up to
The MPPT is connected to the PV array simulator and the minutes), while irradiance may change within 300 milli-
measured DC energy (or average current or voltage) over a seconds or less.
certain measuring time TM (e.g. a few seconds up to a few
minutes) 0∫Tm P(t) dt is compared to the expected DC Resulting plots display the MPPT errors as functions of
energy PMAX·TM (or IMAX or VMAX) which should have VMAX , PMAX , dVmax/dt , and dPmax/dt.
been extracted from the device if the MPPT had operated Measurement equipment comprises a dynamic program-
exactly in the MPP. mable PV array I-V curve simulator, a function generator,
The PV array simulator should have reproducible perform- and meters and instrumentation (possibly interconnected
ance, so that VMAX and IMAX can be measured easily. As for control and automated).
the MPPT continuously moves around the MPP, taking For dynamic assessment and measurements, a programma-
only one pair of I and V values, or the average of all I and ble PV array simulator with real-time signal output of
all V values, is not sufficient and reduces the measurement PMAX(t), VMAX(t) and IMAX(t) is necessary, so that actual
accuracy somewhat. Transient phenomena should have and MPP values can be directly compared [2],[3],[4], (see
decayed before starting to measure, and ηMPPT.E should be figure 4).
calculated on the basis of multiple samples within the
measurement period TM. Alternatively, εMPPT.V or εMPPT.I programmable PV Array Simulator
MPPT
can be calculated for each sample and averaged over the I
period TM. ⇒ V
IV(t) P
Laboratory measurements with PV array simulators have
the advantages of being convenient, fast, and reproducible. ⇓ ⇓
However, PV array simulators can be quite expensive and I I
some practical problems have been observed during labo- PMAX(t) P(t)
ratory tests: V V
ηMPPT
• Simulators based on controlled switching mode devices
Figure 4: Dynamic laboratory measurements
can inject DC ripple on the I-V curve. This ripple can
adversely affect MPPT behaviour. Moreover, under
such conditions, determination of the exact value of 2.3. Assessment of Energetic Efficiency
PMAX may be more difficult
The purpose of these tests is to determine the MPPT ener-
• If the PV array simulator consists of a network of many getic efficiency ηMPPT.E under various sets of conditions,
diodes (and resistors), care must be taken that the di- covering ambiental factors and system design parameters.
odes are at the same temperature during measurement
of MPP and during the actual MPPT operation. Measurement equipment comprises a dynamic program-
mable PV array I-V curve simulator, a function generator
• The combination of a PV array simulator and an MPPT (profile re-player) to reproduce sequences for simulator
may oscillate, or (depending on the design) the MPPT control, and meters and instrumentation (possibly intercon-
may influence and alter the I-V characteristic of the nected for control and automated).
simulator. Controlled switching mode simulators are
especially prone to this problem. Measurements are executed under varied parameters and
defined conditions (sequences). It is proposed to let the
This static test can be repeated at a variety of PV operating dynamic sequences represent various climatic conditions,
conditions to provide curves of ηMPPT.E , εMPPT.V , and PV technologies, and system design parameters (such as
εMPPT.I . voltage and power (mis)match between PV array and
MPPT).
2.2. Assessment of Dynamic MPPT Performance Measuring under static conditions indicates energy loss due
to continuous MPPT search.
The dynamic behaviour of the MPPT algorithm - e.g. on
cloudy days with frequent and rapid changes of irradiance - Measuring under dynamic conditions indicates energy loss
is not reflected in the static figures. In locations where due to quick changes in irradiance.
Resulting figures quantify, for varied sets of conditions, the In order to give accurate results, it is essential that the
effect of MPPT behaviour on the energetic performance of ambient conditions do not change significantly between the
the PV system it is a part of. The argument functions to I-V curve trace and the normal MPPT operation. Varia-
integration may be displayed over time as well, to verify tions in irradiance can be corrected by simultaneously
correctness of the achieved result. measuring the irradiance level and module temperature.
Practical problems encountered with actual measurements
2.4. Further Tests include the following:
Additional tests may be considered to complete the as- • If the time needed to measure the I-V characteristic of
sessment of MPPT device behaviour. the PV array is relatively short, fast semiconductor
switches are necessary. These switches create some
2.4.1. Night-Day and Day-Night Transitions voltage drop that may not be equal for operation with
the MPPT and the I-V tracer. If the tracer operates
This test provides an assessment of the behaviour around very fast (e.g. 1 ms or even faster), MPPT operation is
startup and shutdown times. not affected significantly in most cases. However, fast
Resulting figures describe how efficient available daylight operation of the I-V curve tracer makes it more sensi-
is used, and how effectively futile operation at night is tive to influences of noise picked up during measure-
avoided. ment. Due to inductance of cables to the PV generator,
switching the PV current can result in peak voltages
and in resonant oscillations with involved capacitors.
2.4.2. Ability to Cope With Irregularities Also, some cell technologies, especially those with
This test provides an assessment of how effectively the high minority carrier lifetimes, will not give accurate
MPPT device can cope with possibly occurring irregulari- results if I-V curves are swept too quickly.
ties, such as a partially shaded PV array with conducting
• If the time needed to measure the I-V characteristic of
by-pass diodes and a 'double-kneed' I-V curve, and cloud
the PV array is relatively high (e.g. several 10 ms or
enhancement phenomena, where irradiance can easily jump
longer), MPPT operation may have to be halted during
up to 1400 W/m2 (over 1900 W/m2 have been measured in
the I-V curve tracing, making a restart of the MPPT
the mountains).
necessary, which can be a rather annoying and time
These tests cover robustness of the employed MPPT algo- consuming process. On the other hand, noise problems
rithm, and of the device itself. are much easier to handle in this case, and mechanical
switches or manual connections can be used which cre-
ate less voltage drop.

3. FIELD MEASUREMENTS (OUTDOOR)

3.2. Using a Calibrated Reference Module
Outdoor measurements have the advantage that actual
MPPT behaviour will be observed with the real PV array In this method, I-V curves are taken on a pre-calibrated
thus avoiding potentially unrealistic interactions between reference module to determine the actual MPP, while the
the MPPT and PV array simulator. Obtaining the neces- PV array itself is operating normally with its MPPT (see
sary range of parameters outdoors requires co-operative figure 6).
weather as well as access to a variety of PV technologies. PV Array Reference Module
The following methods have been identified to determine I-V Tracer I
MPPT
the MPPT energetic efficiency ηMPPT.E .
⇒ V
P

3.1. Switching Between MPPT and I-V Tracer ⇓ ⇓

This method has been used by several private companies I I
⇒ G PMAX P
and institutes [5],[6]. The principle is straight-forward.
The operating point of the PV array under normal operation V V
with the MPPT is measured and compared to quasi- ηMPPT
simultaneous measurements from an I-V curve tracer (see Figure 6: Using a calibrated reference module
figure 5).
In order to estimate the MPP of a PV array from the MPP
PV Array of one reference module to the MPP, extensive calibration
I-V Tracer I
MPPT measures have to be performed on the PV array and the
reference module.
⇒ V
P
The reference module must operate under average array
conditions (inclination, orientation, mounting method, field
⇓ ⇓
of view, wind speed, ambient temperature, etc.). Between
I I
⇒ G PMAX P tracing I-V curves the reference module must also be oper-
ated at a voltage equivalent to the PV array so that it
V V
ηMPPT
achieves a similar temperature. Differences in soiling
Figure 5: Switching between inverter and IV tracer between the reference module and the PV array must be
minimised as well. In principle it is possible to measure noting the actual power, the MPP can effectively be deter-
not only static but also dynamic ηMPPT with this method. mined.
Possible reasons for errors in determining PMAX with this In a second step, the MPPT is switched to the automatic
method are: MPPT mode, and operating values are then compared to
the MPP. Irradiance should be recorded to compensate for
• Temperature differences of a few degrees between the actual variations during that measurement (see figure 8).
reference module and other parts of the PV array are
possible. PV Array
MPPT
I
• Partial shading of the PV array can occur during the ⇒ V
measurements. P

• Soiling of parts of the PV array may not be equal to Manual MPPT

that of the reference module. ⇓ ⇓
I I
⇒ G PMAX P
V V
3.3. Sampling MPPT Input at High Speed ηMPPT
This measurement principle is simple, has low cost and is Figure 8: Using manual mode to obtain I-V curve data
well suited for field measurements (see figure 7).
PV Array
This method is by far the simplest to implement and can be
MPPT done as a quick check of MPPT health. It is, however,
I much more effective on inverters that provide manually
⇒ V adjustable voltage operation than those that provide manu-
P
ally adjustable current especially under changing irradiance
⇓ conditions. Additionally, manual operating control capa-
I I bility is not widely available so the use of this method is
⇒ G ? P limited.
V V
ηMPPT ?
3.5. Analysing Monitoring Data
Figure 7: Sampling MPPT input at high speed
There are several approaches for using data collected from
The voltage V and current I at the MPPT input are continu- operating PV systems to determine how accurate the MPPT
ously measured at a high sampling speed, typically every is operating. To a first order approximation, the MPP
5 ms (to avoid 100 or 120 Hz aliasing errors). current IMAX is dependent solely on the in-plane irradiance
G. If the MPPT works correctly, the input current I should
For reference purposes the irradiance should be measured be close to IMAX. Since the proportionality between G and
simultaneously with a fast response device (i.e. a silicon- IMAX is a linear relation, it is valid not only for instantane-
based pyranometer, not a thermopile-based unit), to com- ous values, but also for integrals such as sums or average
pensate for variations in irradiance. data (see figure 9).
In most MPPT designs, the MPPT causes low frequency PV Array
variations of the DC voltage in the order of 0.1 to 1 Hz. A MPPT
I
small 100 or 120 Hz ripple, coming from the AC power
⇒ V
pulsation in case of single phase inverters, can be superim- P
posed over this signal [7],[8].
P ⇓
P
By studying plots of power and voltage over time it can be MP
I
determined whether the MPP is correctly tracked. Oscilla- ⇒ ⇐ P
tions (be it MPPT search or ripple) show correlated varia- G V
G
tion of P and V below the MPP (i.e. power rises with volt-
ηMPPT ?
age), while P and V are anti-correlated above the MPP (i.e.
power falls with rising voltage). Figure 9: Analysing monitoring data
PMAX can be obtained from a regression of successive The proportionality between G and I is a necessary condi-
readings where V and I do not change direction (i.e. con- tion for the correct operation of the MPPT, but in order to
tinue rising or falling) while P does change direction. It is be completely sure that the I-values coincide with the IMPP-
necessary that the sampled values oscillate around the values it is necessary to check that at least one of the points
MPP. on the plot of I versus G is the IMPP-value for the corre-
sponding irradiance G. This can be done by one measure-
ment of the PV array IV characteristics.
3.4. Using Manual Mode to Obtain I-V Curve Data
Also, a scatter diagram (based on 10 minutes averaged
Some MPPTs provide the capability to manually adjust measuring values) with the temperature corrected PV array
their operating point, i.e. they allow the operator to pre-set efficiency and the PV array voltage versus the irradiance
either the voltage or current flowing into the MPPT and can be displayed. A diagram of irradiance, PV array volt-
therefore determine where on its I-V curve the PV array is age, and PV array power versus time makes the working of
actually operating. By manually sweeping the values and MPPT evident as well. This way, static as well as dynamic
MPPT inaccuracies can easily be identified, and many 9th Symposium Photovoltaic Solar Energy, Staffel-
problems with MPPT have been solved in the past (see stein, Germany, 1994, pages 492...496
poster P5B.11, [9]). Assessing the PV array operating
voltage along with PV array efficiency is essential to avoid [7] M. Jantsch, C.W.G. Verhoeve: "Inverters with Three
erroneously concluding MPPT inaccuracy. Phase Output and without Electrolyte Capacitor for
Improved Lifetime, Efficiency and Costs of Grid Con-
Monitoring data can also be used with more sophisticated nected Systems", P7B.6, 14th EC PVSEC, Barcelona,
modelling or I-V curve translation techniques and a well Spain, 1997
characterised array. For example the methods used by ESTI
[10] and by Sandia [11] both can be used to estimate IMAX [8] M. Jantsch, C.W.G. Verhoeve: "AC PV Module
and VMAX for a given set of ambient conditions, for com- Inverters with Full Sine Wave Burst Operation Mode
parison to the array operating data recorded in the data file. for Improved Efficiency of Grid Connected Systems at
Low Irradiance", PL5.3, 14th EC PVSEC, Barcelona,
Operational data from projects monitored according to the Spain, 1997
Guidelines for the Assessment of Photovoltaic Plants
(Document A - Photovoltaic Systems Monitoring) [12] [9] C.W.A. Baltus: "Analytical Monitoring of Losses in
contain the hourly averages of both G and I. Other projects, PV Systems", P5B.11, 14th EC PVSEC, Barcelona,
such as PVUSA and TEAM-UP in the USA, also collect Spain, 1997
the appropriate weather and system performance data. [10] G. Blaesser: "PV Array Data Translation Procedure",
Like in method 3.2 (using a calibrated reference module), IEC TC82 WG2+3, Ispra April 1996 (also: Proc. 13th
partial shading and differences of irradiation and soiling EC PVSEC, Nice, France, 1995)
within the PV array may influence the results obtained. [11] D. King: "Photovoltaic Module and Array Perform-
ance Characterisation Methods for All System Operat-
ing Conditions", Proceeding of NREL/SNL Photovol-
taics Program Review Meeting, November 18-22,
4. SUMMARY AND CONCLUSIONS 1996, Lakewood, CO, USA
Draft definitions of terms and calculations are presented for [12] Guidelines for the Assessment of Photovoltaic Plants,
discussion. A number of measurement methods are avail- Document A: Analysis and Presentation of Monitoring
able and useful for determining the performance of an Data, JRC Ispra, Report EUR 16338 EN, 1995
inverter's MPPT. The given situation, available equipment,
user's needs, and other factors will determine which proce-
dure will be most appropriate. The authors encourage other
researchers to provide feedback in the form of variations,
corrections, additional steps and precautions, or entirely
different procedures.
Our intent is to compile this input and develop a set of
consensus procedures for inclusion in an upcoming IEC
standard.

5. REFERENCES
[1] M. Jantsch: "Nutzungsgrad photovoltaischer Anlagen",
VDI-Verlag, Reihe 6, Nr. 336, 1996
[2] A. Veltman, R. Klöckner, K. Geers, J. v. Twisk:
"Equipment for Testing of Maximum Power Point
Trackers under Variable, Programmable Conditions",
10th EC PVSEC, Lissabon, Portugal, 1991
[3] K. Takigawa, H. Kobayashi: "A Development of High
Accuracy and Wide Range Modular PV Array Simu-
lator", Proceedings 9th International PVSEC, Miya-
zaki, Japan, 1996
[4] G. Keller, G. Klein, V. Löwenstein: "Darstellung unter-
schiedlicher Hardware-Simulatoren von Photovoltaik-
generatoren", 8th Symposium Photovoltaic Solar En-
ergy, Staffelstein, Germany, 1993, pages 478...482
[5] C. Cornelius, D. Tegtmeyer: "Portabler PV-Kenn-
linienanalysator", 8th Symposium Photovoltaic Solar
Energy, Staffelstein, Germany, 1993, pages 509...511
[6] B. Decker, D. Tegtmeyer: "Leistungsvermessung von
PV-Modulen unter realen Tageslichtbedingungen",