SOLAR PV POWER PLANT UNDERGROUND

CABLE SIZING CASE STUDY

Carson Bates PK Sen
Colorado School of Mines Colorado School of Mines
Golden, CO, USA Golden, CO, USA
[email protected] [email protected]

Abstract—A cable ampacity study for a PV power plant is temperatures, and these measurements were compared with the
evaluated based on temperature measurements taken during temperatures anticipated from the design calculations.
normal operation for one year. Differences between the design and
the installation are discussed regarding their impact on cable
temperature. Soil thermal resistivity measurements are reviewed, II. DATA COLLECTION
highlighting the need for the design engineer to properly specify The case study focuses on four sections of installed cable:
the soil conditions for the measurement. The measured Section 1-A, 1-B, 2-A, and 2-B. These cable sections and their
temperature was significantly less than the temperature calculated installation details are shown in Figure 1, Figure 2, and Figure
using the Neher-McGrath method.
3.
Keywords—cable ampacity, Neher-McGrath method, PV studies,
solar power generation, power cable sizing

I. SOLAR PV POWER STATION CASE STUDY

A solar photovoltaic (PV) power station located in Arizona
was studied with the goal of comparing the pre-installation
engineering analysis of cable temperature and ampacity with
the values measured after the system was installed. The rated
power output of the PV plant is approximately 10 MW using
single axis tracking. In order to provide an optimal design for
reliability and cost, the cables were specified using the Neher-
McGrath calculation method and designed to operate near their
maximum temperature rating. Several measurements of the soil
thermal resistivity were made on soil samples from the
installation site. Many of these measurements yielded soil
resistivity values higher than 90 °C*cm/W used to calculate the Figure 1: Cable installation plan view
tables in the National Electrical Code (NEC)[1]. These tables
are commonly used by engineers to determine cable ampacity,
sometimes without knowledge or consideration of the actual
soil resistivity. Many of the measured resistivities also
exceeded the highest value of RHO, 120°C*cm/W, used in the
tables of IEEE Std. 835 [2].
The highest thermal resistivity resulting from the testing on-
site was 329°C*cm/W, which was given at a soil moisture of
0%. The engineers designing the cable system for the
photovoltaic (PV) system used a soil resistivity value of
270°C*cm/W for their design. This was the average of all the
tests taken at 0% soil moisture content. Furthermore, the design
used a load factor of 53% that was based on measured data from
similar projects and calculations from a commercially available
software titled PVsyst. These parameters were used by the
design engineer in the Neher-McGrath method to specify
conductor size and such installation details as cable spacing and
depth. The cable was selected based on a maximum operating
temperature 75˚C, which is the limit of the cable terminations.
Figure 2: Logger 1 cable installation detail and sensor placement
Sensors were installed at the site to measure actual cable

978-1-5386-7551-9/19/$31.00 ©2019 IEEE

 Figure 4: Cable current and temperature on the hottest day

Figure 3: Logger 2 cable installation detail and sensor placement

III. SOIL THERMAL RESISTIVITY
Two data loggers were used, built on the Arduino
For a given conductor size, the soil thermal resistivity is the
MEGA2560 platform. The data loggers store the measurements
most important factor for determining cable ampacity, and it
on a local Secure Digital (SD) card. Since the PV plant is
typically varies more than any other design parameter. It must
unmanned, retrieval of this data is difficult. So, a cellular
be measured at every project site since soil varies by location—
modem was added to the logger to transmit the data to an offsite
it cannot be predicted accurately. However, it is often not
server for more frequent access. The electric current
measured prior to the design of cable systems in industrial and
magnitudes for the circuits of each measured section were
commercial projects. If it is measured, it is often difficult to
recorded by the local inverter data logger. The inverter stores
determine the appropriate soil resistivity for ampacity
the DC current every fifteen minutes for each input module.
calculations.
There are eight parallel input modules for each inverter and the
current is measured from the common DC bus to each of the Prior to the design for this case study, the resistivity was
eight input modules. The DC combiner box circuits are measured in a laboratory for seven soil samples taken from the
connected in parallel to the common DC bus. So, the current project site. The resistivity was measured seven additional
logged by the inverter loses some accuracy from the current for times at the project site (in-situ) after the cables were installed
each combiner box, but this is not considered significant since as part of this case study. The first four measurements were
the combiner circuits have similar power output. performed on native soil in four undisturbed tube samples. The
fifth and sixth measurements were performed on native soil that
The currents measured by the inverter were time-aligned
was compacted to 92% modified proctor. The seventh
with the temperatures recorded by the data logger since each
measurement was performed on backfill samples that were
was measured at a different sample rate. For cable ampacity
lightly tamped, i.e. uncompacted. These measurements were all
calculations, the primary factor is maximum current, so the
performed before and during the cable system design. Seven
exact time is not critical; however, it is helpful to examine the
additional measurement were performed during the initial setup
time constant of the cable/soil system. Figure 4 shows the
of this experiment using a stainless steel probe that is a scaled
recorded data on hottest day of the year, August 14, for two
down version of the field needle described in IEEE 442 [3]. The
temperature sensors and the measured current for Inverter A-1.
results are shown in Figure 5.
 National Resource Conservation Service (NRCS)
(https://www.wcc.nrcs.usda.gov/scan/) [5]. If limited data is
available, a moisture content of 2% provides a sufficiently
conservative design.

IV. DESIGN ANALYSIS

The design engineer selected the following values for the
software program used to perform the Neher-McGrath
calculations:
• Thermal resistivity = 270°C-cm/W (based on the average
of the four thermal resistivities measurements, all at 0%
moisture content, from two geotechnical reports)
Figure 5: Soil thermal resistivity measurements
• Ambient soil temperature = 29°C
The values vary significantly, ranging from 46 to
329°C*cm/W. Generally, using the in-situ measurements is the • Load factor = 53% (based on calculations from PVsyst [6]
best practice because they represent the most likely operating and measured data from similar projects
condition after the installation is complete. The lab values are
• Conductor temperature limit = 75°C (based on the DC
helpful to provide some adjustment to the in-situ value.
Combiner Junction Boxes terminal temperature, NEC
Typically, this involves adjustment for moisture content and
110.14(C)(1) [1])
compaction. These two parameters have a large impact on the
resistivity. In this case, the in-situ tests did not include A single cable ampacity was used in the design and was
measurement of the moisture content or compaction. However, determined based on the cable section with the highest
the geotechnical report indicates the moisture content ranged temperature. Often, a cable is routed through various
from 2% to 3% on measurements performed in December 2011. environments that consist of different burial depths, number of
In-situ testing was done in February 2017, and it was assumed adjacent cables, ambient temperatures, and soil thermal
that the moisture content of 3% in December was the same in resistivities. The engineer may choose to calculate the ampacity
February. Even though the measured resistivity values were for every condition along the cable route for every cable, but
much lower, the design engineer used an average of the highest this is extremely laborious. Alternatively, the engineer limits
soil resistivity measurements for sizing the cable: a fully dried the number of cable sections to model and considers only the
and minimally compacted soil. “worst case scenarios” when determining the ampacity.
Compaction and moisture content must be determined For this PV plant, one section involved 13 combiner box
before selecting a thermal resistivity for use in ampacity “home runs” in parallel with another 11 combiner box “home
calculations. A standard compaction for utility or industrial runs” as the cable routes approached the inverter. The physical
installations is 85% to 95% of standard proctor. This range is spacing between parallel circuits is determined by numerous
primarily determined by the cost of installation with lower factors such as: installation costs, available area, cable size, and
compaction requiring less effort and lower cost. Higher calculated ampacity. The cables were modeled with the spacing
compaction reduces the soil thermal resistivity. The soil is shown in Figure 6.
compacted not only to reduce thermal resistivity but to provide
a harder surface for vehicular traffic or additional construction
above the cables. If vehicle traffic or construction above the
cable is not expected, a compaction of 85% of Standard Proctor
may allow for better revegetation but this is generally
undesirable in PV applications [4]. Regardless of the amount of
compaction, it is common to backfill the cable in one-foot lifts
(increments), compacting between lifts using a vibratory or
tamping foot compactor. Specifying a higher level of soil Figure 6: Cable installation detail
compaction for the cable backfill—such as 90% of Standard
Proctor—reduces the soil resistivity, allowing the designer to The current in each cable was determined based on the
use smaller conductor sizes. Moisture content cannot be maximum power point current multiplied by 125%. The
specified since it is controlled by site conditions. The soil analysis was performed using commercially available software
moisture content should be measured at the project site. If it (AmpCalc). This software utilizes the Neher-McGrath method
cannot be measured, soil moisture data is available from various to determine the conductor temperature for each cable. The
sources on the Internet. This data can assist the designer in results are given in Table 1.
determining the minimum moisture content. One source of this
data is Soil Climate Analysis Network (SCAN) from USDA’s
 Table 1: Software calculated temperatures

Trench Number of Maximum Current

(separated Circuits Cable
by 48”) Temperature

1 13 75°C 220A

2 11 73°C 220A

During the construction of the PV plant, a backfill material

was proposed with a dry thermal resistivity that exceeded the
Figure 7
modeled thermal resistivity. Backfill material is often selected
based on its proximity to the project location. The electrical
engineer was called upon to perform the ampacity study based
on this proposed backfill. The software model was updated with
the backfill thermal resistivity at 0% moisture content
(303°C·cm/W). The cable currents were also refined based on
the exact number of PV module strings rather than an average
for the plant. Some cables in the model increased in current
while others decreased, depending on the number of PV module
strings. As the ampacity analysis was performed, it was clear
the increased soil resistivity posed a problem, so the design
engineer proposed moving the routing of two circuits to reduce
the cable temperatures. The results given in Table 2.
Table 2: Software calculated temperatures with increased soil resistivity

Figure 8
Trench Number of Maximum Current
(separated Circuits Cable Since the temperature sensors were placed on the cable jacket,
by 48”) Temperature Fourier’s Law was used to determine the conductor temperature
across the jacket thermal resistance. The conductor temperature
1 11 87°C 258A is equal to: Tc=Tj+I²R·ρj·ln(rj /rc)/(2π) where R=8.60·10-7 Ω/cm
(53°C Al), ρj=350°C·cm/W, rc=0.908” (750 kcmil compact),
2 11 86°C 258A
rj=1.178” (2kV XLPE). The measurements are listed in Table 3
Table 3: Maximum temperature measurements
With the proposed backfill, the software results predicted
that the cable temperatures would exceed the design limit of Logger Number Maximum Maximum Current
75°C. Nevertheless, construction moved forward with the intent of Jacket Conductor
of measuring the final cable temperatures. Circuits Temp Temp

V. TEMPERATURE MEASUREMENT 1 (similar 11- 51.4°C 54.9°C 210A

to 48”gap-
Temperature sensors were installed on the cables and the
software) 11
backfill was placed in four-inch lifts with manual compaction.
After installation and compaction, the solar plant was operated
2 (deeper 12- 57.3°C 60.7°C 208A
normally, and the temperatures were measured. Installation was
& closer) 9”gap-
completed in February 2017 and measurements from the hottest
12
part of the year, June 24 through August 25, are shown in Figure
7 and Figure 8. Figure 7 is from Logger 1, which measures
temperatures at locations similar to the software model. Figure
Table 4 compares the results of the software model, using
8 is from Logger 2, which measures temperatures of the deepest
both values of soil resistivity, with the measurements made by
cables (about 5 feet deep, sloping up to 3 feet over a distance of
Logger 1.
10 feet).
 Table 4: Software calculated versus measured temperatures The measurements suggest a maximum ambient
temperature of 43.46°C. One common method for estimating
Maximum Cable Current the maximum temperature of the site is to use nearby SCAN
Temperature data from the NRCS [5]. The nearest recording site to the
installation site is recording station 2185, named Essex, and
Software Rho 270 75°C 220A located in San Bernardino County, CA. The site has been
recording data since 2012. Data from 2012 to 2017 was
Software Rho 303 87°C 258A compiled and analyzed for the maximum temperature and
minimum moisture content. Figure 10 and Figure 11 summarize
Logger 1 54.9°C 210A the data from this recording site.

Logger 1 60.7°C 208A

Clearly, the measurements are lower than either of the

software model results. The current is slightly lower for the
measurements than for the software model, but this doesn’t
explain the large difference in maximum temperature. Several
other parameters could contribute to this difference: soil Figure 10: Historical maximum soil temperature at five depths
thermal resistivity, ambient soil temperature, cable spacing, or
load factor. Each of these parameters is discussed in the
following paragraphs.
The soil thermal resistivity was measured onsite during the
installation of the measurement equipment. The resistivity
measurements were significantly lower than the values used for
the software model. The mean of the measured values was
equal to 157°C*cm/W with a 99% confidence interval of 118
to 197°C*cm/W. Since the in-situ measurements were taken at
Figure 11: Historical maximum soil moisture content at five depths
a single location, the measurements do not account for site
variability. The moisture content was not measured during the
“ST” stands for Soil Temperature, and “SM” stands for Soil
in-situ testing. Since the measurements were taken in February,
Moisture. Both prefixes are followed by the soil depth in inches
it is expected this represents a low moisture content due to lack
where the measurement was made as listed in the legends of the
of rainfall during the winter months. Considering all of this, the
graphs.
upper limit of the 99% confidence interval, 197°C*cm/W, is
much lower than either of the resistivities used in the design The primary value of interest is the maximum temperature
calculations (270°C*cm/W and 303°C*cm/W). This is the most near the depth of the installed cables, which is 40” since the
likely explanation for the difference between the maximum middle of the trench is approximately 36” below grade. When
measured temperature and the calculated maximum modeling cable ampacity, the design must account for the
temperature. highest temperature at the lowest moisture content, because it
is associated with the highest thermal resistivity/ambient
The ambient soil temperature was not measured at site. The
temperature combination of the soil. From the SCAN site data,
closest measurement that could be considered ambient was the
the highest yearly 99th percentile soil temperature is 34.6°C and
sensor located 18” above the top of the cables for data logger 2.
the lowest yearly 1st percentile moisture content is 1%. To
It seemed to vary with cable current rather than accurately
summarize: the maximum measured temperature 18” above the
measuring ambient temperature but was suggestive of the
cables is 43.5°C, the maximum temperature from a nearby
maximum possible ambient temperature at the site. Figure 9
SCAN site is 34.6°C, and the software modeled temperature is
shows the measurements from this sensor on the hottest day.
29°C. Since the modeled ambient temperature was less than the
actual ambient temperature, the resulting cable temperature
should have been greater than the calculated temperature. The
cable temperature was not greater, so the ambient temperature
cannot be the reason for the discrepancy between the measured
and calculated cable temperatures.
Cable spacing could also explain the difference between the
measured temperature and the calculated temperature. The
spacing between cables and between circuits is specified by the
Figure 9: Suggestive soil ambient temperature on the hottest day engineer in the form of drawing details. These details are used
during construction by the tradesperson. It is difficult to at 50% load factor. The design load factor is 53% based on
maintain the specified cable spacing in direct-buried cable PVsyst analysis and measured data from similar projects. The
installations since the tradesperson must visually place the actual load factor can be calculated for the experimental data by
cables and avoid disturbing their spacing while backfilling. dividing the average current by the maximum current for a
Installations in conduit or duct banks reduce the variability in given period. The daily and weekly load factors are shown in
cable spacing. The project utilized direct-buried cable, and the Figure 13. The number of daylight hours divided by 24 hours is
spacing deviated from the design by more than the expected also shown. The daylight hours are calculated based on the
tolerance of two inches. This is shown in Figure 12 latitude and day of the year.

Figure 13: Measured daily and weekly load factor

The measured data indicates a maximum daily load factor of

48% and weekly load factor of 46%. This is 10% less than the
Figure 12: Specified cable spacing versus installed location design load factor. The expected operating temperature should
be correspondingly lower.
The installed depth was shallower than the designed
depth. This will result in lower temperatures than expected. The
VI. CONCLUSION
spacing was not consistent--some cables were closer than five
inches and others were farther. Many cables of the same DC In summary, the installation differed from the design
circuit have a spacing of only two-inches. Additionally, the parameters. Table 5 compares the differences and predicts the
center cables are closer than the specified spacing. It is expected impact on cable temperature.
that the net result of the spacing discrepancies would be higher Table 5: Installation differences and their impact on temperature
operating temperatures than calculated. Since these installation
deviations counteract each other in temperature effects, it is
Parameter Design Measured Temperature
difficult to determine whether the resulting temperature will be
Impact from
higher or lower than the calculated temperature. The assumed
Design to
result is neutral. Neither a higher nor lower temperature should
Measured
result from these deviations that occurred during installation.
Regardless of the net temperature impact resulting Soil thermal 270 and 197 Lower
from these deviations, the installation practices can result in a resistivity 303°C*cm/W °C*cm/W
final installation that varies from the engineer’s design. This
possibility should be evaluated by the design engineer. It may Ambient soil 29°C 34.6°C Higher
be prudent to evaluate the installation at a more conservative temperature
condition, such as closer spacing and deeper burial depth, than
the specified installation. Any critical spacing should be Cable 5” horizontal, Varies from Neutral
emphasized to the installer. spacing 6” vertical 2” to 12”
Load factor is another parameter that could result in Load factor 53% 48% Lower
differences between the measured and the expected cable
temperature. Its impact is nonlinear, so an exact ratio between
load factor and ampacity is not possible. However, ampacity is The measured temperature is lower than the calculated
more sensitive to soil resistivity when using lower load factors. temperature, so it is clear that the differences between the
That is, an increase in soil resistivity at 100% load factor affects design parameters and the actual site conditions resulted in a
the ampacity less than the equivalent increase in soil resistivity net impact of lower cable temperatures. The design engineer did
not include any additional safety factors or design margins. Had
he done so, there would have been an even greater over-
estimation of temperature and over-sizing of the cables,
resulting in unnecessary expense. It appears that the Neher-
McGrath method provides an already conservative design and
does not require additional safety factors or design margins.
The original design analysis indicated the cables would
operate near their temperature rating. Following that, backfill
soil with a soil resistivity higher than the original design was
installed. Ampacity calculations using this higher soil
resistivity predicted operation above the temperature rating of
the cables. The measurements revealed that the actual operating
temperature was much lower than the temperature calculated
using the Neher-McGrath method.

VII. REFERENCES
[1] National Fire Protection Association, NFPA 70 National Electrical Code
2014. 2013.
[2] “IEEE Standard Power Cable Ampacity Tables,” IEEE Std 835-1994, pp.
1–3151, Dec. 1994.
[3] “IEEE Guide for Soil Thermal Resistivity Measurements,” IEEE Std 442-
1981, pp. 1–16, Jun. 1981.
[4] F. Media, “Optimizing Soil Compaction and Other Strategies - Forester
Network,” Forester Network, 01-Sep-2004. [Online]. Available:
https://foresternetwork.com/daily/construction/optimizing-soil-
compaction-and-other-strategies/
[5] “NRCS National Water and Climate Center - Soil Climate Analysis
Network.” [Online]. Available: https://www.wcc.nrcs.usda.gov/scan/
[6] “PVsyst Photovoltaic Software.” [Online]. Available:
http://www.pvsyst.com/en/

VIII. BIOGRAPHIES
Carson Bates (M’09) received the B.S. degree in
engineering with electrical specialty in 2010, the M.S. in
electrical engineering in 2013, and the Ph.D. in electrical
engineering in 2018, all from the Colorado School of Mines,
Golden, CO. He currently works as a full-time engineer at NEI
Electric Power Engineering in Wheat Ridge, CO.