Journal of Physics: Conference Series

PAPER • OPEN ACCESS You may also like

- Solar Cycle Variation of Coronal
PV generator nominal power estimation using a Temperature, Emission Measure, and Soft
X-Ray Irradiance Observed with Yohkoh
ground sensor and the PVLIB online irradiance Soft X-Ray Telescope
Aki Takeda, Loren Acton and Nicole
Albanese
database
- A thermal radiation exchange model of
whole-body UV phototherapy
To cite this article: J R Angulo et al 2022 J. Phys.: Conf. Ser. 2180 012005 A J Coleman, G A Aneju, P Freeman et al.

- Effects of synchronous irradiance

monitoring and correction of
current–voltage curves on the outdoor
performance measurements of
View the article online for updates and enhancements. photovoltaic modules
Yoshihiro Hishikawa, Takuya Doi, Michiya
Higa et al.

This content was downloaded from IP address 190.232.106.60 on 01/03/2022 at 07:17

Peruvian Workshop on Solar Energy (JOPES 2021) IOP Publishing
Journal of Physics: Conference Series 2180 (2022) 012005 doi:10.1088/1742-6596/2180/1/012005

PV generator nominal power estimation using a ground sensor

and the PVLIB online irradiance database

J R Angulo1, L A Conde1, E. Muñoz1, J de la Casa2 and J A Töfflinger1*

1
Materials science and Renewable Energies Group, Science Department, Pontificia
Universidad Católica del Perú, Av. Universitaria 1801, Lima 32, Peru
2
IDEA Research Group, Electronics and Automation Engineering Department,
Universidad de Jaén, Campus Las Lagunillas s/n. 23071–Jaén, Spain
*
Correspondence: [email protected]

Abstract. The nominal power of a photovoltaic generator is a crucial parameter. For instance, it
is a required input parameter for modelling the energy produced by a PV system. Determining
the nominal power in-situ requires measuring the irradiance and module temperature. Usually,
the irradiance is measured with a sensor placed on the ground and at the plane of the array. The
present work first compares the irradiance measurements of a calibrated PV module with the
openly available irradiance database PVLIB. Second, it contrasts the daily estimated nominal
power values considering both irradiance sources. The results indicate that the irradiance
database is suitable to calculate the nominal power.

1. Introduction
In 2020, more than 127 GW of photovoltaic (PV) capacity or nominal power had been installed, and
global solar installations could hit a record of 180 GW by the end of 2021 [1]. Grid-connected, utility-
∗
scale systems are the primary type of installed configuration. The nominal power (𝑃𝑀 ) represents the
maximum power point of a PV generator under standard test conditions (STC): irradiance of 1000 W/m2,
∗
module temperature 25°C and AM 1.5 spectrum [2]. Generally, the theoretical 𝑃𝑀 of a PV generator is
the sum of the maximum power of the installed modules under STC according to their datasheet.
However, this value is an overestimating approximation since the PV generators present losses due to
module mismatch and cable resistivity, among others. Therefore, it is essential to consider such intrinsic
losses in the nominal power estimation since knowing its real value in operating conditions is of great
∗
interest for quality assurance of the photovoltaic plant [3]. Verifying the effectively installed 𝑃𝑀 of any
PV system allows identifying problems in the installation or during operation.
∗
Following the procedure suggested by Martinez et al. [4], estimating 𝑃𝑀 of a PV generator requires
measuring the plane of array irradiance, module temperature, and DC output power. An important factor
∗
affecting the 𝑃𝑀 uncertainty can be the irradiance measurement in the tilted plane of the PV generator.
Using a calibrated PV module of the same technology as the PV generator as an irradiance sensor has
the advantage of considering the spectral response [5], decreasing the spectral mismatch error in
modelling the output power. However, installing a PV module as a sensor requires calibration in a
certified laboratory with a solar simulator. Such laboratories are yet scarcely accessible in many parts
of the world, particularly in developing countries.
The present work investigates the possibility of using the irradiance values from satellite-based, open-
access databases, such as PVLIB [6], as an alternative to in-situ ground-based measurements. To do this,
Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
Peruvian Workshop on Solar Energy (JOPES 2021) IOP Publishing
Journal of Physics: Conference Series 2180 (2022) 012005 doi:10.1088/1742-6596/2180/1/012005

section 2 presents the PV plant under study and the experimental setup. Section 3 shows the procedure
followed to calculate the nominal power using the ground sensor. Section 4 discusses the nominal power
values resulting from the irradiance database in comparison to the ground measurements.

2. Experimental setup
Figure 1 shows the studied PV generator located in Granada, Spain, with a nominal power of 109.4 kW
according to the manufacturer data sheet [7]. Its PV modules have a fixed tilted angle of 30° and are
oriented south. The PV generator's main electrical characteristics are summarized in Table 1.
The experimental campaign took place from March to septembre 2018. Three main parameters were
recorded every 30 seconds: (1) Direct Current power 𝑃𝐷𝐶 with a YOKOHAMA WT1600 wattmeter, (2)
module temperature (𝑇𝑚 ) with a PT100 sensor, and (3) ground irradiance (𝐺 𝑀𝑜𝑑 ) with a calibrated PV
module.
Adittionally, PVLIB is an online database for Python developed by Sandia National Laboratories. It
contains a repository with weather data to model PV systems [8] and it is comparable in quality with
commercial packages such as PVsyst [9]. In this sense, the information provided by the PVlib in the
irradiance values has been considered as input in the following section to calculate the nominal power.

Figure 1. View of the 109.4 kW photovoltaic generator.

Table 1. Electrical parameters of the PV generator at STC

according to the manufacturer.
Units Value
Power at maximum power point kW 109.4
Current at the maximum power point A 257.6
Voltage at the maximum power point V 574.2
Power temperature coefficient %/°C -0.43
Number of modules per string
18
connected in series
Number of strings connected in parallel 32

3. Nominal power estimation

The total number of days during this experiment was 135. Most of the days had partially cloudy sky
conditions such as the example shown in figure 2. However, to calculate the nominal power following
[4], it is crucial to consider only the days with clear skies, such as shown in figure 3. This requirement
is also indicated in the ASTM E2848-13 standard [10]. For our experimental campaign, 37 of the 135
days met these clear-sky conditions.

2
Peruvian Workshop on Solar Energy (JOPES 2021) IOP Publishing
Journal of Physics: Conference Series 2180 (2022) 012005 doi:10.1088/1742-6596/2180/1/012005

1200 120
1200 Irradiance 120 Irradiance
DC Power 1000 DC Power 100
1000 100
800 80
800 80
G [W/m2]

PDC[kW]

G [W/m2]

PDC[kW]
600 60 600 60

400 40 400 40

200 20 200 20

0 0 0 0
07:00 09:00 11:00 13:00 15:00 17:00 19:00 21:00 07:00 09:00 11:00 13:00 15:00 17:00 19:00 21:00
Hours Hours

Figure 2. Exemplary, partially cloudy day Figure 3. Exemplary, clear sky day
(18/05/2018). (16/05/2018).
120
106
100
Data (2018-09-24)
Nominal Power [kW]
P(G, T → 25°C)[kW]

Linear Fit 105

80

60 104

40
103
Ground Sensor
20 P*
M = (104.22 ± 0.03) kW ---- Mean= 104.13 kW
R-Square = 0.99 102
---- ±1.5% (1.56 kW)
0
0 200 400 600 800 1000 0 5 10 15 20 25 30 35
2
G [W/m ] Day

Figure 4. Linear fit of DC power corrected to Figure 5. Calculation of the nominal power
25 ºC for irradiance above 800 W/m2. for each day in clear sky conditions.
∗
To estimate 𝑃𝑀 , the corrected DC power (𝑃𝑇→ 25 ℃ ) is calculated using equation 1:

𝑃𝐷𝐶
𝑃𝑇→ 25 ℃ = (1)
1 + 𝛾(𝑇𝑚 − 25℃)

Here, γ is the power temperature coefficient provided in the manufacturer's module datasheet, 𝑃𝐷𝐶 is the
measured DC power and 𝑇𝑚 the module temperature. Finally, we perform a linear fit of the data using
∗
equation 2 to obtain the nominal power 𝑃𝑀 .

𝐺
∗
𝑃𝑇→ 25 ℃ = 𝑃𝑀 (2)
𝐺∗

Here, 𝐺 ∗ is the irradiance under STC (1000 W/m2), 𝐺 is the plane of array irradiance. The linear fit only
considers irradiance values between 800 W/m2 and 1000 W/m2.
Figure 4 demonstrates the resulting linear fit of the temperature-corrected DC power versus irradiance
in a clear-sky day. The slope of the fit represents the nominal power. On this particular day, we obtain
∗ ∗
𝑃𝑀 = 104.22 kW. This is repeated for all 37 clear-sky days, the resulting daily 𝑃𝑀 values are depicted
∗
in figure 4. The average of all values is 104.13 kW. This average 𝑃𝑀 is about 5% lower than the
manufacturer's total module maximum power at STC of 109.3 kW (see table 1). This difference is within

3
Peruvian Workshop on Solar Energy (JOPES 2021) IOP Publishing
Journal of Physics: Conference Series 2180 (2022) 012005 doi:10.1088/1742-6596/2180/1/012005

a reasonable range considering intrinsic losses and possibly degradation of the PV modules after eight
years of operation [7].

4. Results and discussions

Figure 6 shows a scatter diagram of the irradiance measured with the ground sensor 𝐺 𝑀𝑜𝑑 and compared
with the PVLIB 𝐺 𝐿𝑖𝑏 online database. Only clear-sky day data and irradiance values for 𝐺 𝑀𝑜𝑑 >
100 W/m2 are considered here. A linear fit of the 𝐺 𝑀𝑜𝑑 and 𝐺 𝐿𝑖𝑏 data within the entire irradiance range
results in a slope M =1.0289 indicating that 𝐺 𝐿𝑖𝑏 overestimates the irradiance by almost 3% on average.
The relatively low spread of the data is represented with an R-squared of 0.984. We can observe several
outliers [11], where 𝐺 𝐿𝑖𝑏 considerably overestimates the irradiance, possibly due to some cloud effects
or shadowing of the ground sensor 𝐺 𝑀𝑜𝑑 .
1200

1000

800
GLib [W/m2]

600

400
Data of clear-sky days
Linear Fit ( GLib=M* GMod )
200
M = 1.0289 ± 0.0009
R-square = 0.984
0
0 200 400 600 800 1000 1200
GMod [W/m2]

Figure 6. Scatter plot of 𝐺 𝑀𝑜𝑑 from PV module ground sensor and 𝐺 𝐿𝑖𝑏 from PVLIB database for 37
clear-sky days.
∗
Figure 7 depicts the 𝑃𝑀 estimated for both irradiance data sources on a daily basis. The estimated values
resulting from the ground sensor show little variation over time in comparison to the values obtained
from the PVLIB database. In the latter case, a seasonability is apparent which could be attributed to days
∗
where the irradiance is underestimated and overestimated, resulting in larger and lower 𝑃𝑀 , respectively.
112 112
Ground Sensor
Ground Sensor
PVLIB Database
Nominal Power P*M [kW]

PVLIB Database
Nominal Power P*M [kW]

108 108

104
104

100
100

96
96
0 5 10 15 20 25 30 35
Ground Sensor PVLIB Database
Clear-sky Day

Figure 7. Nominal power for each day Figure 8. Boxplot analysis considering the data
calculated with irradiance values using obtained in figure 7.
ground sensor and online database.

Figure 8 shows the box plot analysis for both cases. The ground irradiance sensor 𝐺 𝑀𝒐𝒅 leads to an
∗ ∗
average 𝑃𝑀 = 104.04 ± 0.79 kW, with error < 1%. For 𝐺 𝐿𝑖𝑏 , the estimation results in 𝑃𝑀 = 103.46 ±

4
Peruvian Workshop on Solar Energy (JOPES 2021) IOP Publishing
Journal of Physics: Conference Series 2180 (2022) 012005 doi:10.1088/1742-6596/2180/1/012005

3.5 kW, thus, with a higher relative error of about 3%. This higher error most likely originates from the
over- and under estimated irradiance of the PVLIB database.
In [4], the uncertainty of calculating the nominal power in a 1.3 kW PV string is analyzed by tracing the
I-V curve and following the IEC-60981 standard, resulting in 5% uncertainties. Therefore, our result in
the average nominal power and the corresponding uncertainties indicate that using the PVLIB irradiance
database can be a suitable alternative to using a ground sensor.

5. Conclusions
The nominal power of a photovoltaic generator in operation has been estimated. The experimental
campaign was 135 days; only 37 days had clear sky conditions allowing to calculate the daily nominal
power. We consider two cases: (1) ground PV module sensor 𝐺 𝑀𝒐𝒅 and (2) PVLIB database 𝐺 𝐿𝑖𝑏 . The
average nominal power values were 104.04 kW and 103.46 kW for the cases (1) and (2), respectively.
The main difference is evidenced in the uncertainty shown in standard deviation, which is 0.79 kW and
3.5 kW for cases (1) and (2), respectively.
Although the online irradiance database does not have the same degree of accuracy as a ground sensor,
the estimated nominal power showed consistency. We can conclude that for the location of the studied
PV generator using the irradiance from the PVLIB database can be a suitable alternative to in-situ ground
sensor measurements. As future work, it would be important to confirm whether the methodology
applying online irradiance database to estimate the nominal power can be applicable also to other
locations with different climate regions.

Acknowledgments
This work received financial support by CONCYTEC-FONDECYT within the framework of the call
E063-2019-01.BM, contract N°013-2020-FONDECYT-BM. José Angulo acknowledges the financial
support given by CONCYTEC under the Ph.D. scholarship program with contract N°236-2015-
FONDECYT. Additionally, part of this work has been financed by the “Agencia Andaluza de
Cooperación Internacional para el Desarrollo” of the Junta de Andalucía (Andalusian Autonomous
Government), through the project “Emergiendo con el Sol” under expedient code 2012DEC026.
Finally, the authors are grateful to the company GERION INGENIERIA for giving access to the
facilities and permission for monitoring the PV plant.

References
[1] Welter P 2021 The Solar Power Magazine Photon 5 1-56
[2] Muñoz-García M A, Marin O, Alonso-García M C, and Chenlo F 2012 Characterization of thin
film PV modules under standard test conditions: Results of indoor and outdoor measurements and
the effects of sunlight exposure Sol. Energy 86 no. 10 3049–56
[3] Kumar S R, Gafaro F, Daka A, and Raturi A 2017 Modelling and analysis of grid integration for
high shares of solar PV in small isolated systems – A case of Kiribati Renew. Energy 108 589–
597
[4] Martínez-Moreno F, Lorenzo E, Muñoz J, and Moretón R 2012 On the testing of large PV arrays
Prog. Photovoltaics Res. Appl. 20 no. 1 100–105
[5] Carrillo J M, Martínez-Moreno F, Lorenzo C, and Lorenzo E 2017 Uncertainties on the outdoor
characterization of PV modules and the calibration of reference modules Sol. Energy 155 880–
892
[6] Gurupira T and Rix A J 2016 Photovoltaic System Modelling using PVLib-Python Fourth South
African Sol. Energy Conf. SASEC 2016, Univ. Stellenbosch, South Africa 1 no. 1 1–10
[7] Muñoz-Cerón E, Lomas J C, Aguilera J, and de la Casa J 2018 Influence of Operation and
Maintenance expenditures in the feasibility of photovoltaic projects: The case of a tracking pv
plant in Spain Energy Policy 121 506–518
[8] Stein J S, Holmgren W F, Forbess J, and Hansen C W 2017 PVLIB: Open source photovoltaic
performance modeling functions for Matlab and Python 2017 IEEE 44th Photovolt. Spec. Conf.

5
Peruvian Workshop on Solar Energy (JOPES 2021) IOP Publishing
Journal of Physics: Conference Series 2180 (2022) 012005 doi:10.1088/1742-6596/2180/1/012005

PVSC 2017 1–6

[9] Gurupira T, Rix A J, Gurupira T, and Rix A J 2017 Pv Simulation Software Comparisons: Pvsyst,
Nrel Sam and Pvlib
[10] American Society for Testing Materials 2018 Standard Test Method for Reporting Photovoltaic
Non-Concentrator System E2848 −13 1–11
[11] Zhao Y, Lehman B, Ball R, Mosesian J, and De Palma J F 2015 Outlier detection rules for fault
detection in solar photovoltaic arrays Conference Proceedings - IEEE Applied Power Electronics
Conference and Exposition - APEC, 2013 2913–2920