energies

Article
Modeling of German Low Voltage Cables with Ground
Return Path
Johanna Geis-Schroer 1, *,† , Sebastian Hubschneider 1, *,† , Lukas Held 1,† , Frederik Gielnik 1 ,
Michael Armbruster 2 , Michael Suriyah 1 and Thomas Leibfried 1

1 Institute of Electrical Energy Systems and High-Voltage Technology (IEH), Karlsruhe Institute of
Technology (KIT), 76137 Karlsruhe, Germany; [email protected] (L.H.); [email protected] (F.G.);
[email protected] (M.S.); [email protected] (T.L.)
2 Stadtwerke Buehl GmbH, 77815 Buehl, Germany; [email protected]
* Correspondence: [email protected] (J.G.-S.); [email protected] (S.H.)
† These authors contributed equally to this work.

Abstract: In this contribution, measurement data of phase, neutral, and ground currents from real
low voltage (LV) feeders in Germany is presented and analyzed. The data obtained is used to review
and evaluate common modeling approaches for LV systems. An alternative modeling approach
for detailed cable and ground modeling, which allows for the consideration of typical German LV
earthing conditions and asymmetrical cable design, is proposed. Further, analytical calculation
methods for model parameters are described and compared to laboratory measurement results of real
LV cables. The models are then evaluated in terms of parameter sensitivity and parameter relevance,
focusing on the influence of conventionally performed simplifications, such as neglecting house
junction cables, shunt admittances, or temperature dependencies. By comparing measurement data





 from a real LV feeder to simulation results, the proposed modeling approach is validated.

Citation: Geis-Schroer, J.;

Keywords: distribution systems; multiphase and unbalanced analysis; low voltage grids; low voltage
Hubschneider, S.; Held, L.; Gielnik, F.;
Armbruster, M.; Suriyah, M.;
cable; ground return path; asymmetric modeling
Leibfried, T. Modeling of German
Low Voltage Cables with Ground
Return Path. Energies 2021, 14, 1265.
https://doi.org/10.3390/en14051265 1. Introduction
Traditionally, distribution system operators (DSOs) assume symmetrical load con-
Academic Editor: Ilhan Kocar ditions for low voltage grid planning [1]. Despite the single-phase connection of most
domestic appliances, equal load distribution is achieved by considering wider grid seg-
Received: 15 December 2020
ments and statistically averaging a large number of costumers. When also assuming
Accepted: 15 February 2021
symmetrically designed electrical equipment, this allows for modeling low voltage (LV)
Published: 25 February 2021
grids as simple, single-phase equivalent systems.
In the course of the ongoing German energy transition, the number of distributed
Publisher’s Note: MDPI stays neutral
photovoltaic (PV) systems installed in low voltage grids are rising. About 70% of those
with regard to jurisdictional claims in
are small single- or two-phase rooftop systems [2]. Looking at this in more detail, the
published maps and institutional affil-
previously mentioned DSO’s assumption of symmetrical load conditions is not feasible
iations.
anymore. Measurement results from Reference [2] show that 93% of single-phase PV
inverters are connected to the same phase, typically phase a. Increasing penetration rates of
electric vehicles and battery storage systems—of which, once again, a large share has single-
phase or two-phase grid connections—further amplify asymmetrical load situations [3].
Copyright: © 2021 by the authors.
The discrepancy between the assumptions during grid planning and present asymmet-
Licensee MDPI, Basel, Switzerland.
rical grid reality causes new challenges in grid operation, for example concerning voltage
This article is an open access article
quality according to European standard DIN EN 50160 or neutral conductor ampacity [4].
distributed under the terms and
In order to evaluate these issues and develop potential solutions, there is a growing interest
conditions of the Creative Commons
in detailed modeling of distribution grids, used for integration studies, stability studies and
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
real-time simulation environments. Not only load modeling, but also detailed modeling of
4.0/).
lines, is mandatory for unbalanced system studies. With a share of 80% and increasing, un-

Energies 2021, 14, 1265. https://doi.org/10.3390/en14051265 https://www.mdpi.com/journal/energies

Energies 2021, 14, 1265 2 of 34

derground cables are the dominating line type in German LV grids [5]. The most common
cable types, four-core NAYY and NAYCWY, are depicted in Figure 1.
Authors, like Oeding et al. [6], present tables for positive and zero sequence impedance
parameters for different cable types, intended for rough calculations of minimum and
maximum fault currents. Based on the assumption of equal mutual inductances between
phases—which seems unrealistic for four-core cables (see Figure 1)—these parameters
do not seem to be suitable for modeling of LV systems. Furthermore, the underlying
assumptions regarding system grounding are derived from transmission grids [6] and
seem questionable for German LV grids.
In this contribution, we present measurement data of phase, neutral and ground
currents from real LV substations and feeders in Section 2. In Section 3, we use this
data to review and evaluate common modeling approaches for LV systems. Lastly, we
present an alternative modeling approach, which allows for the consideration of typical
German LV earthing conditions and asymmetrical cable design. In Section 4, we present
calculation methods for our cable model parameters, and validate these through laboratory
measurements of real LV cables. Section 5 focuses on modeling grounding impedances at
substations and costumer nodes. In Section 6, we evaluate our model in terms of parameter
sensitivity and parameter relevance, focusing on the influence of commonly performed
simplifications, such as neglecting house junction cables, temperature dependencies, or
shunt admittance. In order to validate our modeling approach, we compare simulation
results to measurement data from a real LV feeder in Section 7. Our conclusions are drawn
in the final section.

outer sheath (PVC)

concentric conductor
core insulation (copper)
(PVC)
solid
sector conductor
(aluminum)

core
covering
(PVC)

(a) (b)

Figure 1. Standard cable types in German low voltage (LV) grids [7]. (a) NAYY-J cable type [8].
(b) NAYCWY cable type [9].

2. Monitoring of Real LV Feeders

This section presents measurement data recorded in real low voltage distribution
grids. The grids of two different distribution system operators in South Western Germany,
Stadtwerke Buehl GmbH and Netze BW GmbH, are located in areas with different geo-
graphical characteristics. While substations A to D are located in the plain Rhine valley,
substations E to H are located in the Black Forest highlands. Both grids are part of larger
grid areas and include several medium voltage/low voltage (MV/LV) substations. The
corresponding low voltage grids differ in line equipment (cable versus overhead line), as
well as number of feeders and type of customers (residential versus industrial areas).
In the following, we show and compare results from two different types of measure-
ments. Section 2.1 focuses on line currents of single feeders (see Figure 2a), while Section 2.2
analyzes data of current measurements at substation busbars (see Figure 2b). For acquiring
measurements, we used a Tektronix 8 channel 12 bit mixed signal oscilloscope of type
MSO58. Hall-effect current clamps PAC16 by Chauvin Arnoux were used for measuring
the currents.
Energies 2021, 14, 1265 3 of 34

Both measurements focus on return currents through neutral conductor and ground,
where ground is considered to be a return path through building foundation grounding,
soil and transformer foundation grounding, marked with the subscript G in variable names.
This additional return path through ground results from the TN-C-S design of the German
low voltage grid (see Section 3). In this TN-C-S design, the neutral conductor is grounded
at both the transformer neutral point and every customer node. The neutral conductor is
usually combined with the PE (protective earth) conductor and, therefore, labeled as PEN
(protective earth neutral) conductor.

a b cY a b cY

xfmr

xfmr
iY,xfmr
a a
b b
c c
PEN PEN
ia,f1
Feeder 3

Feeder 2

Feeder 1

Feeder 3

Feeder 2

Feeder 1
ib,f1
ic,f1
iPEN,f1 iPEN,f3 iPEN,f2 iPEN,f1

(a) Measurement setup at LV feeders. (b) Measurement setup at PEN busbars.

Figure 2. Measurement schemes (xfmr: transformer).

2.1. Detailed Measurement of Line Currents

Figure 2a depicts the measurement scheme exemplarily for feeder 1. The three phase
currents ia,f1 , ib,f1 , and ic,f1 and the PEN current iPEN,f1 of one single feeder are recorded at
the same time. The direction of current flow is defined in direction of the substation.

2.1.1. Breakdown of Feeder Currents

Figure 3 exemplarily shows measurements of currents of all six feeders of substation
D over 20 ms. For all feeders, the three phase currents ia (brown), ib (black), and ic (grey)
are significantly imbalanced; hence, the PEN conductor is carrying the return current iPEN
(blue). The residual ground current iG (yellow) was calculated according to Equation (1),
assuming the considered feeder to be the only one at the substation (also see Section 2.3).

iG = −(ia + ib + ic + iPEN ). (1)

Table 1 shows the root mean square (RMS) values IPEN and IG corresponding to
measured PEN currents iPEN and calculated ground currents iG , respectively, for several
feeders in different MV/LV substations. Due to safety reasons and difficulties to access all
four conductors with current clamps, measurements could not be performed at all feeders
of all substations.
Besides the ratio IG /IPEN of true RMS values, Table 1 also provides the ratio IG,1 /IPEN,1 ,
which only considers RMS values of the fundamental oscillation. The ratio of ground cur-
rents to PEN conductor currents, IG /IPEN , varies between 0.06 and 0.66 and is equal to or
lower than IG,1 /IPEN,1 —which reaches up to 0.93—for all considered feeders. This is con-
sistent with Figure 3, where iG shows less harmonic distortion than iPEN for all six feeders
of substation D. The difference between both ratios allows us to draw the conclusion of
different loads being responsible for the ground currents in different shares.
Energies 2021, 14, 1265 4 of 34

Figure 3. Measurement results from all six feeders at substation D.

Due to the small number of measurement samples, it is not possible to draw general
conclusions regarding the influence of line type, number of cable cabinets and customers
on the share of return current flowing through ground. However, the average of IG /IPEN
for feeders with more than ten costumer nodes (0.44) is noticeably higher compared to
the one of feeders with less than ten costumer nodes (0.23). A possible explanation is that
grounding impedances of costumer nodes provide parallel grounding resistances, which
leads to a decreased overall ground return path impedance.

Table 1. Current measurements of different feeders at exemplary LV substations.

Sub- Feeder Line Cabinets Customers Grid IG IG,1

IPEN IG
Station (#) Type (∑) (∑) Type IPEN IPEN,1
1 Cable/OL 2 22 Res. 23.3 A 5.43 A 0.23 0.24
A
2 Cable 2 10 Res. 6.2 A 1.19 A 0.19 0.19
B 1 Cable 1 7 Res. 5.6 A 1.11 A 0.20 0.30
1 Cable 1 1 Ind. 6.0 A 3.24 A 0.54 0.63
C
2 Cable 1 4 Ind. 6.9 A 1.83 A 0.26 0.29
1 Cable/OL 0 8 Res. 8.7 A 0.54 A 0.06 0.07
2 Cable 1 14 Res. 16.3 A 6.53 A 0.40 0.42
3 Cable 0 14 Res. 6.8 A 2.49 A 0.37 0.52
D
4 Cable 1 21 Res. 13.4 A 8.83 A 0.66 0.93
5 Cable 1 19 Res. 15.5 A 6.18 A 0.40 0.46
6 OL 0 8 Res. 6.0 A 0.94 A 0.16 0.15
1 Cable/OL 0 13 Res. 9.5 A 2.93 A 0.31 0.35
2 Cable 0 9 Res. 3.6 A 0.95 A 0.26 0.35
E
3 Cable/OL 0 8 Res. 15.8 A 1.90 A 0.12 0.12
4 Cable 0 3 Res. 2.4 A 0.67 A 0.27 0.26
F 1 Cable 0 1 Hotel 9.5 A 1.29 A 0.14 0.19
1 OL 0 7 Res. 13.2 A 0.77 A 0.06 0.06
G
2 OL 0 3 Res. 5.7 A 0.54 A 0.10 0.10

2.1.2. Time Dependency

Table 2 shows measurements results of substation A, feeder 2. The RMS currents
IPEN and IG , as well as the ratio between these currents, as described earlier, are given
for different times. Typical changes in phase loads—and in load imbalance—during the
considered time period of 82 min led to significant variations in measured PEN current
Energies 2021, 14, 1265 5 of 34

IPEN and calculated ground current IG . In contrast, the ratios of ground to PEN current
IG /IPEN and IG,1 /IPEN,1 both remain similar over time with a relative standard deviation
of 8% and 11%, respectively.
Hence, the measurement results show that a changing load situation does not lead
to major changes in the ratio, which indicates a nearly rigid grounding situation and,
thus, constant shares of ground currents for the considered LV feeder. We assume the
remaining deviations in ratio IG /IPEN over time to result from neutral current composition
by different customers and, thus, different grounding paths and impedances.

Table 2. Time dependency of current measurements at substation A, feeder 2.

Time IPEN IG IG /IPEN IG,1 /IPEN,1

11:23 a.m. 6.91 A 1.13 A 0.16 0.16
11:25 a.m. 2.24 A 0.42 A 0.19 0.19
11:39 a.m. 2.37 A 0.47 A 0.20 0.17
12:40 p.m. 6.24 A 1.19 A 0.19 0.19
12:43 p.m. 3.57 A 0.72 A 0.20 0.21
12:45 p.m. 5.59 A 1.05 A 0.19 0.21
Mean 4.49 A 0.83 A 0.19 0.19
RSD 0.45 0.41 0.08 0.11

2.2. Measurements at Substation PEN Busbars

The analysis in Section 2.1 focuses on individual feeders. This section, in contrast,
focuses on measurements at the substations’ PEN busbar at the LV side of the respective
MV/LV transformer. The PEN conductors of all feeders and the transformer’s neutral
conductor are connected to this busbar. The current clamps were installed as depicted in
Figure 2b.
Figure 4 shows exemplary measurement results from substation D. The currents of
PEN conductors from Feeder 1 to Feeder 6 are described by iPEN,f1 to iPEN,f6 (grey). The
sum of these currents adds to iPEN,∑ (blue). Curve iY,xfmr (green) shows the neutral point
current of the transformer.

Figure 4. Measurement of ground/neutral currents at LV PEN busbar of substation D.

As seen earlier for feeder currents in Section 2.1, there is a relevant deviation between
the neutral point current of the transformer and the sum of the feeders’ PEN currents. The
overall sum of the currents in the PEN busbar has to add to zero. Hence, the residual
current iG,∑ , yellow in Figure 4, results from Equation (2).

iG,∑ = iY,xfmr − iPEN,∑ . (2)

Energies 2021, 14, 1265 6 of 34

Table 3 shows the RMS values of the already introduced currents iY,xfmr , iPEN,∑ and
iG,∑ for seven different substations B to H with 2–10 feeders and 9–111 customers. In
addition, it lists the ratio IG,∑ to IPEN,∑ , as well as, again, the ratio of their fundamentals
IG,∑ ,1 and IPEN,∑ ,1 . Harmonic distortion in all currents is clearly visible in Figure 4. For the
considered substations, the ratios between true RMS values of ground and PEN currents
range between 0.01 and 0.21. Ratios of fundamental RMS values are very similar, showing
absolute differences of only up to 0.01 compared to true RMS ratios. Both ratios are
significantly lower compared to the ratios calculated for single feeders in Table 1.

Table 3. Measurements at PEN busbars of different substations.

Sub- Feeders Customers IG,∑ IG,∑ ,1

IY,xfmr IPEN,∑ IG,∑
Station (∑) (∑) IPEN,∑ IPEN,∑ ,1
B 10 111 34.62 A 34.41 A 1.62 A 0.05 0.04
C 6 12 27.41 A 26.17 A 1.39 A 0.05 0.06
D 6 84 57.00 A 54.01 A 4.28 A 0.08 0.07
E 6 48 52.52 A 51.50 A 1.05 A 0.02 0.02
F 8 19 37.32 A 37.20 A 0.47 A 0.01 0.00
G 2 10 8.39 A 8.10 A 0.46 A 0.06 0.05
H 2 9 5.80 A 5.10 A 1.09 A 0.21 0.20

2.3. Conclusions
From a general point of view, the measurement data shows that power flow in German
low voltage grids is typically significantly unbalanced, which leads to transformer neutral
point currents of up to the same magnitude as phase currents.
Furthermore, the analysis indicates that some part of the balancing currents takes
different paths than the PEN conductors, i.e., through grounding devices and soil. Consid-
ering single feeders, the theoretical ground current as residual current of phase and PEN
conductor currents ranges between 6% and 93% of the PEN current. As it reduces voltage
drops across the PEN conductor and influences the cable’s magnetic field, it is, therefore,
relevant for detailed modeling of LV grids. The ratio between both currents is not constant,
but—compared to absolute current values—only slightly differs in time. Additionally, the
ratio varies between LV feeders.
The ratio ground to PEN current is typically smaller for substations than for feeders.
We assume this to result from compensating ground currents between different feeders
belonging to the same substation, as ground currents are not bound to physical conductors.
Considering these observations, accounting for the ground return path is significant
when aiming for accurate low voltage feeder modeling.

3. Modeling Approaches of Low Voltage Cables and Grounding

When trying to accurately model line impedances in LV grids, a common problem is
that DSOs do not provide detailed information about type, condition (age), and length of
underground cables and overhead lines. However, even if this information is available,
one main challenge remains due to the design of the German distribution network as
TN-C-S system.

3.1. German LV Grid as TN-C-S System

As shown in Figure 5, there is a combined PEN conductor between the substation and
the building connection point. Inside of the building, it is split up into separate PE and n
conductors. The PEN conductor is grounded at the transformer neutral point, and the PE
conductor—hence, indirectly the n conductor—is connected to the main earth bar in the
customer buildings. Due to this, a ground return path is provided in parallel to the PEN
Energies 2021, 14, 1265 7 of 34

conductor. The measurement data presented in Section 2 clearly shows the relevance of
accounting for this ground return path when modeling unbalanced current flow through
low voltage cables.
MV/LV substation
public distribution grid: TN-C system
a
b
c
PEN

private costumer’s grid: TN-S system

transformer grounding a
b
c
n
distribution panel PE

main earth bar

lightning protection, metallic pipes, ...

grounding conductor
grounding electrode

Figure 5. TN-C-S system, based on Reference [1,10].

Modeling this ground return path is a nontrivial task, mainly because soil resistivity
ρE varies with soil type, humidity, and temperature—Table 4 gives exemplary values in
comparison with non-soil materials. Furthermore, in Europe’s mainly moderate climate
zone, soil resistivity values follow a sinusoidal dependency reaching their mean value in
May and November, while the maximum and minimum values in April and August are
approximately 30% higher and lower [11]. In addition, the realization and the effectiveness
of the grounding electrode varies depending the age of the building (see Section 5).

Table 4. Resistivity values for different soil types and materials, based on Reference [12–14].

Soil Type ρE in Ωm (Mean Value)

Marshy soil 30
Clay or humus soil, arable land 100
Sandy soil (humid) 200
Sandy soil (dry) 1100
Stony soil 1000
Concrete (1 part of cement, 3 parts of sand) 150
Aluminum (20 °C) 2.8 × 10−5
Stainless steel strip for foundation grounding ≤1.5 × 10−4

3.2. Conventional Modeling Approaches

This subsection gives a brief overview of modeling approaches of low voltage cables
in related work, presenting and evaluating underlying assumptions.

3.2.1. Handling of Shunt Admittances

Shunt admittances between cable conductors and to ground consist of conductances
and capacitances. For dielectric material not being excessively aged by improper cable
operating conditions, conductance G is negligibly small compared to ωC [15]. Hence,
in literature there is general agreement on neglecting conductances when modeling low
voltage cables [16–19].
Energies 2021, 14, 1265 8 of 34

According to authors, like Oeding and Oswald [6], Schwab [1], and Kersting [20],
capacitances cannot be generally classified as negligible in low voltage cables due to small
distances between conductors and the high permittivity of polyvinyl chloride (PVC).
The authors of Reference [16–19], however, neglect both conductances and capaci-
tances, pointing out that the influence of shunt admittances on voltage deltas along low
voltage cables is rather small compared to the influence of line resistances and induc-
tances. In CIGRÉ benchmark grids, neglecting capacitances is justified by referring to
comparatively small line lengths in low voltage grids [21]. Since no author quantifies the
exact influence of neglecting capacitances on simulated node voltages, it remains unclear
whether modeling capacitances is relevant for detailed modeling of low voltage cables.

3.2.2. Temperature-Independent Parameters

As low voltage cables are usually embedded about 1 m below ground surface, daily
temperature variations do not influence the temperature of the surrounding soil [22]. Nev-
ertheless, heat losses of cables in operation lead to increased conductor and insulation
temperatures. Depending on the load situation, temperatures inside the cable may reach
the maximum rated temperature of 70 °C for PVC-insulated cables [23]. The electrical
conductivity of the conductor shows a significant temperature dependency [24]. However,
common modeling approaches assume conductor resistance to be constant during opera-
tion, relying on reference values valid for a temperature of 20 °C. In literature, there has
been no discussion on the effect of this simplifying assumption on simulation accuracy.

3.2.3. Modeling in Symmetrical Components

As can be seen from the most typical German four-core low voltage cable types in
Figure 1, the cable design shows asymmetry. We expect that magnetic and electrical cou-
pling, i.e., mutual inductances and capacitances, between horizontally/vertically conductor
pairs significantly differ from those of diagonal conductor pairs.
However, when assuming mutual inductances to be symmetrical and applying sym-
metrical components, line models can be described by three-dimensional diagonal phase
impedance matrices, which greatly simplify load flow calculations or impedance estimation
methods [6,16,25]. Using these reduced models only requires knowledge of four param-
eters: line resistance and reactance in both positive and zero components. Authors, like
Oeding and Oswald [6], provide tables with corresponding reference values for different
German low voltage cable types, yet pointing out that these parameters are originally
intended for rough calculations of minimum and maximum fault currents and, thus, not
necessarily suitable for accurate modeling of low voltage grids.
Urquhart [25] demonstrates, for a British low voltage cable with similar asymmetrical
design to the typical German four-core NAYCWY in Figure 1b, that, approximating the
impedance matrix only in terms of positive and zero sequences, can lead to considerable
errors of up to 17% in calculation of voltage drops.
Another possible source of inaccuracies with modeling in symmetrical components
results from underlying assumptions with respect to grounding. Oeding and Oswald [6]
present two different zero sequence impedance values for each cable type. The first one is
based on neglecting the ground return path. Our analysis of measurement data from real
German LV grids in Section 2 clearly shows that assuming the neutral conductor as the only
return path for balancing currents is not feasible when aiming at accurate modeling of real
load scenarios. The second zero impedance value provided in Reference [6] is derived from
modeling the ground return path according to Carson’s approach and applying Kron’s
reduction. In the following, we will briefly outline the latter approach, and evaluate its
suitability for modeling of German LV grids.

3.2.4. Carson’s Equations for Ground Return Path Modeling

The classical approach for ground return path modeling starts from a five-conductor
system, where a fictitious fifth conductor in parallel to phase and neutral conductors
Energies 2021, 14, 1265 9 of 34

represents the ground return path (see Figure 6). There is magnetic coupling between
all five conductors. The sum of currents in the cable and ground is assumed to be zero,
such that:
I a + I b + I c + I n + I G = 0. (3)

Va Ia Zaa V a0

+ +
Vb Ib Zbb Zab V b0
+ +
Vc Ic Zcc Zbc V c0
+ +
Vn In Znn V n0
+ +
VG ZGG ZcG 0
VG
IG
+ +
− −
reference ground

Figure 6. Five-conductor line model, including the ground return path, based on Reference [20,25].

Voltages in Figure 6 are defined relative to the transformer’s grounded neutral point,
i.e., reference ground. As Figure 6 depicts a cable segment not directly connected to the
0
transformer, V G and V G are local ground voltages of magnitudes unequal to zero. Voltage
drop across the cable segment is described by Equation (4).

V a − V a0 Zaa Zab Zac Zan ZaG Ia

     
 V b − V 0   Zba Zbb Zbc Zbn ZbG 
b   Ib 
 
 V − V 0  =  Zcn
 
 c c  Zcb Zcc Zcn  ·  I c .
ZcG  (4)
 
 V − V0   Z

n n na Znb Znc Znn Z  I 
nG n
VG − VG 0 ZGa ZGb ZGc ZGn ZGG IG

Primitive Circuit Impedances

Re-arranging Equation (4) for voltage differences relative to local ground voltages
and using Equation (3) leads to Equation (5) [20,26], where Ẑii and Ẑij refer to so-called
“primitive” self and mutual impedances, into which the impedance of the ground return
path is merged (see Equations (6) and (7)).
 0 
V aG − V aG

Ẑaa Ẑab Ẑac Ẑan
  
Ia
 V bG − V 0   Ẑba Ẑbb Ẑbc Ẑbn   Ib 
bG   ·  ,
 V − V 0  =  Ẑ (5)
 
cG cG ca Ẑcb Ẑcc Ẑcn   I c 
0
V nG − V nG Ẑna Ẑnb Ẑnc Ẑnn In

Ẑii = Zii − ZGi − ZiG + ZGG , (6)

and Ẑij = Zij − ZGj − ZiG + ZGG = Ẑ ji (7)
with i, j = 1, 2, 3, N, ZGi = ZiG and ZGj = Z jG .

Originally presented for telegraph overhead lines in 1926 [27], Carson’s equations
enable to directly compute these primitive impedances. Carson assumes the soil to be of
constant specific resistance and infinite in radius and depth, and models ground currents
as images of conductor currents in soil. Furthermore, he assumes that conductor diameters
are small compared to distances between conductors. He solves for the impedances by
considering electromagnetic fields between conductors and images [25].
By only considering those terms from the original equations that are relevant at
technical frequencies, Kersting [20] derives simplified equations. Transforming these into
SI units leads to Equations (8) and (9) [25], which only contain the variables conductor
resistance Rii in Ω
m , specific soil resistance ρE in Ω m, grid frequency f in Hz and geometric
dimensions Dii and Dij in m. Geometric mean radii Dii and geometric mean distances
Energies 2021, 14, 1265 10 of 34

Dij can easily be calculated for round conductor arrangements in low voltage cables (see
Section 4.2.2).
!
0 µ 0 · ω µ 0 · ω 658.9
Ẑii = Rii0 + +j ln , (8)
2·π
p
8 Dii · f /ρE
!
0 µ0 · ω µ0 · ω 658.9
Ẑij = +j ln . (9)
2·π
p
8 Dij · f /ρE

Hence, although originally derived for round overhead lines, Equations (8) and (9)
are often used for modeling of low voltage cables with ground return path—for instance,
by the authors of Reference [17,19,28], as well as in CIGRÉ benchmark grids [21].

Kron’s Reduction
Using common power flow software and modeling in symmetrical components do not
allow for an explicit representation of the neutral conductor [20,21,25]. The 4 × 4 system in
Equation (5), thus, often needs to be reduced to a three-dimensional system. This cannot be
done solely through simple transformations of equations, but requires further assumptions.
Usually, Kron’s reduction is applied, which implicates assuming a perfectly grounded
neutral point at every node [20,25].
In general, assuming a perfectly grounded neutral at every node clearly seems un-
realistic for low voltage grids. Although there may be grounding electrodes at sub-
stations and costumer nodes, such as in the German TN-C-S system, there are most
likely also nodes without grounding, such as junctions between two cable types. The
authors of Reference [25,28], therefore, do not apply Kron’s reduction to calculated primi-
tive impedances.

Applicability To German Low Voltage Grids

Even without using Kron’s reduction, considerable issues with applying Carson’s
equations to models of German low voltage cables remain. While assumptions required
for the derivation of Carson’s equations may be considered realistic for long overhead
transmission lines—which have no neutral conductor, and ground is the only return
path [15]—they appear questionable for comparatively short low voltage cables with
closely spaced conductors.
As Carson models ground currents as images of line currents, a key assumption
for applying his equations to cables is that cable and ground currents sum up to zero
at any point along the cable, such as described by Equation (3). This does not allow for
the possibility of ground currents following different routes than in parallel to the cable.
In reality, for instance, metallic underground pipes might provide a path with a smaller
impedance [17]. Moreover, Carson’s approach excludes circular ground currents between
neighboring grounded nodes [25]. In reality, such circular currents can occur between
buildings, substations or street lighting installations, even if these do not belong to the
same feeder.
Lastly, our analysis of measurement data from low voltage substations and feeders
in Section 2 shows that Carson’s hypothesis of ground currents being bound to paths in
parallel to the feeder may not hold in reality. Technically, Equation (3) is used to determine
the theoretical ground currents for a single feeder in Section 2.1. Following Carson’s
approach, and thus assuming all theoretical return currents to reach the substation busbar,
the ratios of ground to PEN currents of single feeders would correspond to the ratio of
ground to PEN currents calculated for the entire substation. However, the substations
ground current is considerably lower than the feeders theoretical ground currents. This
indicates the presence of the previously described circular ground currents between feeders
and busbars.
Energies 2021, 14, 1265 11 of 34

3.3. Alternative Modeling Approach: Ground as Single Electrical Point

Questioning the applicability of Carson’s equations to low voltage grid modeling,
Olivier et al. [17] suggest fellow researchers to try out an alternative approach “for the sake
of simplicity”: To model low voltage cable segments as four-line π-systems, considering the
ground as single electrical point, to which all grounded neutral points are connected—with
or without grounding impedances.
As all grounding impedances refer to the same electrical point, independently of
the grounded node’s distance to the transformer, we believe this alternative approach
suggested in Reference [17] to represent earthing conditions of the German TN-C-S system
more realistically than Carson’s equations. Notably, the alternative approach also allows
for circular currents between grounded nodes.
The modeling approach used in our work is, therefore, based on Olivier et al. [17], and
depicted in Figure 7. At the end of the line segment of length l, the neutral conductor is
grounded with a grounding impedance ZG,build , referring to grounding of an exemplary
connected costumer building. As sketched in grey, reference ground is not defined to be the
transformer’s neutral point, but we model the transformer’s neutral point to be grounded
with an impedance ZG,xfmr .

0 l
Raa 0 l
Laa

1 0 1 0
2 Cac l 0 l
Lab 2 Cac l
0 l
Rbb 0 l
Lbb

0 l 0 l 0 l
Lbc
Rcc Lcc

1 0 1 0
2 Ccn l 0 l 0 l 2 Ccn l
Rnn Lnn

1 0 1 0
ZG,xfmr 2 CnG l 2 CnG l ZG,build

Figure 7. Conceptual sketch of derived line model (line section of length l, mutual inductances, and
capacitances incomplete).

To be able to assess the influence of model simplifications later in Section 6, we initially

create our model to be as realistic as possible. As all four cable conductors are represented
explicitly, we enable modeling unequal mutual inductances between conductors. In con-
trast to conventional modeling approaches, we also account for capacitances between cable
conductors and stray capacitances to ground (blue).
With respect to parameters to be determined for our low voltage grid simulation
models, we distinguish between the following two groups:
Cable parameters The depicted cable parameters in Figure 7, Rii0 , Lii0 , Lij0 , and Cij0 , only
depend on cable design and materials. For these parameters, we present calculation
methods in Section 4.2 and evaluate these with laboratory measurements of real low
voltage cables in Section 4.3.
Ground/Ground return path Capacitance to ground CiG 0 , as well as impedances

ZG,xfmr and ZG,build , vary depending on the cable’s real installation conditions and
can, therefore, not be determined through laboratory measurements. We present
exemplary reference values in Sections 4.2.3 and 5.

4. Low Voltage Cable Parameter Determination

For the alternative modeling approach presented in Section 3.3, no general reference
values for parametrization of cable models are suggested in literature. Theoretically, an
accurate determination of required parameters is possible through numerical analysis
Energies 2021, 14, 1265 12 of 34

of the field differential equations (e.g., by using finite element method [26] or method
of moments [29]). This requires not only detailed information on cable dimensions but
also on the exact installation and environmental conditions—ideally, a three-dimensional
soil model [25]. However, huge computational efforts seem disproportionate considering
significant manufacturing tolerances, as seen in measurements. In addition, practical field
conditions impede an exact ground return path modeling.
In this work, we use practical approaches for calculation of cable parameters, deriving
required information from standards and data sheets by manufacturers, and present them
in the following Section 4.2. Then, the resulting cable parameters are validated with data
obtained by measurements of real cables, before analyzing the findings with respect to
dependencies between different cable models.
Our test objects are of different types of underground cables typically used in German
LV grids and summarized in Table 5. Three of these show an asymmetric four-core design
(A, B, C), one has a symmetric three-core design (D).

Table 5. Test cables for measurements.

Cable Type Manufacturer Length

A NAYY-J 4 × 150 Bayka [30] ≈100 m
B NAYY-J 4 × 50 Bayka [30] ≈25 m
C NAYCWY 4 × 50 SE/25 Bayka [31] ≈25 m
D NAYCWY 3 × 50 SE/50 Faber [32] ≈25 m

4.1. Assumptions and Simplifications

Though aiming at a detailed model of cables, there are certain simplifications concern-
ing cable lay length and shield we make and discuss shortly in the following sections.

4.1.1. Influence of Cable Lay Length

Due to the cable lay, the length of the individual conductors is higher than the cable
length, to which impedance and admittance per-unit parameters are multiplied to. Unfortu-
nately, neither manufacturer data sheets nor German cable standards provide information
on lay lengths and corresponding correction factors. In accordance with Reference [26], we
assume the influence of the cable lay to be negligible compared to other influences on the
calculated parameters.

4.1.2. Idealization of the Concentric Conductor

The concentric copper conductor in cables of type NAYCWY is composed by several
physically distanced copper wires (see Figure 8). As a copper band around the wires
ensures conductivity between them, we consider the concentric copper conductor as ideal
cylindrical electrical shield (see Figure 9b,c). We refer to the concentric conductor as shield
(S), regardless of its actual functionality as PE (four-core NAYCWY) or PEN (three-core
NAYCWY) conductor. The PEN sector-shaped conductor of NAYY cables is referred to as n
(see Figure 1a).

Figure 8. NAYCWY 4 × 50 SE/25 test cable (sheath cut open).

Energies 2021, 14, 1265 13 of 34

S
S
a a
a
b n b n b

c c
c

(a) NAYY-J (A, B). (b) Four-core NAYCWY (C). (c) Three-core NAYCWY (D).

Figure 9. Ideal cable designs (corresponding test cables).

4.2. Calculation of Cable Parameters

In the following, we provide approaches to calculate cable parameters based on
geometric dimensions of conductors and insulation.

4.2.1. Ohmic Resistances

For economic reasons, we expect cable manufacturers to design conductors close to
the maximum permissible values given in standard DIN VDE 0295 [33], which are listed as
direct current (DC) resistances per unit length in Ω
m for a temperature of 20 °C. For other
temperatures ϑ, the standard defines the equation as given in Equation (10):

0 0 1 1
RDC,ϑ = RDC,20 [1 + α(ϑ − 20 °C)] where α = αAl = 0.00403 or α = αCu = 0.00393 . (10)
K K
The specified cross-sectional areas in datasheets of cables refer to these resistance
values, but do not correspond to the real geometric cross-sectional area. For shielded
NAYCWY cables, the indicated cross-sectional area of the concentric conductor refers to
the resistance value of an aluminum conductor with equivalent cross-sectional area. For
example, the name extension “25” in type NAYCWY 4 × 50 SE/25 defines the maximum
permissible DC resistance of the copper shield equal to the DC resistance of an aluminum
conductor with a cross-sectional area of 25 mm2 [33].
For three-phase alternating current (AC) operation, the resistance is defined according
to (11), where factors FS > 1 and FP > 1 represent the dependency on skin and proximity
effect, respectively [24].
0 0
RAC,ϑ = FS · FP · RDC,ϑ . (11)
Both effects depend on the skin depth δ, which varies with typical cable operating
temperatures between 12 mm and 13 mm (see Table 6). For the calculation of FS and FP ,
Oeding and Oswald [6] provide approximation formulas for symmetrical arrangements
of three round conductors. The authors derive FS and FP from the parameter x, which
puts conductor radius r into relation with skin depth δ. In order to apply the formulas
from Reference [6] to sector-shaped conductors, we assume the radius r of a sector-shaped
conductor to equal the radius of a round conductor with an equivalent cross-sectional
area (see Equation (12)). Taking into account the European standard DIN EN 60228 [33]
and once again assuming economic interests of the manufacturers, the real geometric
cross-sectional area differs from qnom —which is the nominal value as defined by the cable
type. Shafieipour et al. [29] estimate qreal to be about 88% of qnom .
q
qreal
r π
x= = . (12)
2δ 2δ
For FP , the relation between conductor diameter and distance to other conductors
needs to be considered. We adapted the approximation formula provided in Reference [6]
to account for asymmetrical arrangements of sector-shaped conductors. Table 6 gives
Energies 2021, 14, 1265 14 of 34

an overview of calculated factors FS and FP for the cross-sectional areas in focus of our
investigations.
As mentioned in Section 3.2.3, Oeding and Oswald [6] provide reference impedance
values in symmetrical components. The indicated positive sequence resistances for the cable
types in focus correspond to applying Equation (11) for a temperature of 20 °C and factors
FS and FP from Table 6. Neglecting the comparatively small temperature dependency of FS
and FP , we, thus, obtain Equation (13) for the AC resistance of sector-shaped conductors.
For concentric copper conductors, literature does not suggest AC reference values.
However, as they consist of a multitude of small copper wires, we assume skin and
proximity effect to be negligible. Hence, we assume the AC resistance to equal the DC
resistance according to Equation (14).

0 0
RAC,ϑ = RAC,20, [6] ·[1 + αAl (ϑ − 20 °C)] sector-shaped aluminum conductor (13)
| {z }
pos. sEquation reference value
0 0
RAC,ϑ = RDC,20, [33] ·[1 + αCu (ϑ − 20 °C)] concentric copper conductor (14)
| {z }
for aluminum conductor of equal qnom

Table 6. Evaluation of skin and proximity effect factors for f = 50 Hz and µ = µ0 , approximated for
aluminum sector conductors.

ϑ δ qnom x FS FP FS · FP
50 mm2 0.16 1.0002 1.0004 1.0006
0 °C 11.54 mm
150 mm2 0.28 1.0021 1.0054 1.0075
50 mm2 0.16 1.0002 1.0003 1.0005
20 °C 12.04 mm
150 mm2 0.27 1.0018 1.0036 1.0053
50 mm2 0.14 1.0001 1.0002 1.0003
70 °C 13.19 mm
150 mm2 0.25 1.0012 1.0025 1.0037

4.2.2. Self and Mutual Inductances

As mentioned earlier, we cannot use reference values in symmetrical components nor
Carson’s equations for consideration of self and mutual inductances in our models. In
the following, we briefly outline how to use standard methods from Reference [34,35]—
developed for overhead lines—in order to compute the required parameters.
In Reference [34,35], Grigsby and Glover calculate the magnetic flux linkage of an
arrangement of several round conductors of infinite length. They assume uniform current
distribution, which is a valid approach for the underground cables we focus in our work
(see x in Table 6). According to Reference [6], a decrease in the magnetic flux due to skin
effect only occurs for x ≥ 0.3, while Reference [15] assumes relevance only for x ≥ 1.
From the method presented in Reference [34,35], the inductance per unit length in
H
m results from Equations (15) and (16), which only depend on the geometric parameters
geometric mean radius (GMR) Dii and geometric mean distance (GMD) Dij . It is to mention
that we do not expect this to lead to significant model inaccuracies, as permeability of
aluminum, copper, PVC, and common soil materials only slightly differs from vacuum.

 
µ0 1
Self inductance of a conductor i: Lii0 = ln , (15)
2π Dii
!
µ0 1
Mutual inductance between conductors i and j: Lij0 = ln . (16)
2π Dij
Energies 2021, 14, 1265 15 of 34

To use Equations (15) and (16) on the conductor arrangements depicted in Figure 9, we
need to determine the five parameters listed below. Since Carson’s equations also require
the calculation of GMR and GMD, approaches for LV cables are available in literature.

GMR sector-shaped conductor DCC := Daa = Dbb = Dcc (= Dnn )

GMR shield DSS
horizontally/vertically (
Dab = Dbc = Dac three-core cable
GMD neighboring sector- DHV :=
Dab = Dbc = Dan = Dcn four-core cable
shaped conductors
GMD diagonal conductors DDiag := Dac = Dbn four-core cable
sector-shaped
GMD DCS := DaS = DbS = DcS (= DnS )
conductor–shield

Geometric Mean Radius of Sector-Shaped Conductors

Urquhart et al. [26] approximate the GMR of a sector-shaped conductor DCC as equal
to the one of a round conductor of equal cross-sectional area. Using the real geometric
cross-sectional area qreal as introduced before, this leads to the calculation of DCC according
to Equation (17).
r r
− 41 0.88 · qnom 1 qreal
DCC = e = e− 4 . (17)
π π

Geometric Mean Distance Between Sector-Shaped Conductors

The GMD of two sector-shaped conductors can be approximated as the distance of
their rotational axes [26]. Equations (18) and (19) calculate the GMD in dependence of
m, which is the conductors rotational axis’ distance to the rotational axis of the cable (see
Figure 10). Urquhart et al. [26] refer to British cables according to the British standard
BS 3988:197 [36], using the parameters back radius, sector depth and corner radius. Adapting
the approach from Reference [26] to German LV cables, we interpret these parameters
as r1 , d and r2 , respectively, as defined in the German standard DIN VDE 0276-603 [37].
Furthermore, θ summarizes the offset of the rotational axis due to the insulation thickness,
where dins describes the thickness of the insulation layer according to Reference [37] and α
equals the conductors’ angle in the center of the cable—which is 90° (four-core cable) or
120° (three-core cable).
 
d dins 1
m = r1 − + max{ − r1 + d + r2 − 1 , 0} (18)
2 sin(α/2) sin(α/2)
| {z }
θ

(√
3·m three-core cable
DHV = √ (19)
2·m four-core cable
DDiag = 2 · m (20)

Geometric Mean Radius and Distance of the Concentric Shield

According to Reference [6], the GMD between any conductor and a surrounding
concentric shield equals to the GMR of the shield, which corresponds to its radius. In
NAYCWY cables, the shield is located on top of the core covering. Hence, the shield’s GMR
can be computed according to Equation (21), where dcc is the thickness of the core covering
(see Figure 10).

DCS = DSS = r1 + θ + dins + dcc . (21)

Energies 2021, 14, 1265 16 of 34

dcc

d
r1 dins

r2 m

α
θ

Figure 10. Sectional view and parameter definition of a four-core underground cable based on
Reference [26,37].

4.2.3. Capacitances
In terms of shunt admittances, we only focus on capacitances between conductors, ne-
glecting shunt conductances. While DC conductances can be easily approximated through
DC conductivity and thickness of the insulation material, due to complicated frequency-
and temperature-dependent polarization losses, this is not true for AC conductances [38].
However, conductances play a subordinate role (see Section 3.2.1) and, thus, are not fur-
ther considered.
With respect to capacitances, the significant temperature dependency of permittivity
er of the insulation material PVC needs to be taken into account [24]. For 20 °C, er varies
between 3 and 4, depending on the exact composition of PVC. For maximum operation
temperature of 70 °C, er is approximately 8, which results in twice the capacitance value
compared to the one for 20 °C. Due to the lack of further information, we assume a linear
interpolation between those values.
Shafieipour et al. [29] show for a shielded four-core cable, comparing simple closed-
form equations to finite element method (FEM) analysis, that the maximum error is around
5%. As the analysis in Reference [29] refers to conductors shaped as ideal quarter circles,
we geometrically approximate sector-shaped conductors as a quarter or one third of a circle
for our simplified calculations (see Figure 11).
We consider horizontally/vertically sector-shaped conductor pairs (e.g., a–b and b–n
in Figure 11) as parallel plate capacitors (see Equation (22)). As there is no closed-form
expression for capacitances in diagonal sector-shaped conductor pairs, derived from the
FEM analysis in Reference [29], we approximate the values as shown in Equation (23).

0 r1
CHV =e , (22)
2 · dins
0 0
CDiag ≈ 0.05 · CHV . (23)

Furthermore, in accordance with Reference [29], we consider conductor–shield pairs

in four- and three-core NAYCWY cables as a quarter and one-third of a cylindrical capacitor,
respectively (Equation (24)).

2·π· e
1
3 · r1 + dins + dcc
three-sector cable
 !


 ln
r1

0
CCS = 1 2·π· e
(24)
4 · four-sector cable.

r1 + dins + dcc
 !

ln


r1

All conductors in operation additionally have an electrical field to ground. While, for
0 between all four conductors and ground,
NAYY-J cables, there are stray capacitances CCG
Energies 2021, 14, 1265 17 of 34

i.e., the surrounding soil, for NAYCWY cables, the field only exists between the concentric
conductor—considered as ideal shield—and ground (CSG 0 ). The electrical field limitation

outside the cable is undetermined and strongly depends on environmental conditions [24].
An upward estimation is possible by assuming a limitation of the electrical field to the
outer side of the cable’s sheath. The resulting values match with the general reference
µF µF
values of 0.48 km and 0.66 km for qnom of 50 mm2 and 150 mm2 , respectively, that can be
derived from symmetrical component modeling in Reference [6] and are, thus, utilized for
our models.

S
a n
CDiag
r1
dins
dcc
b c
CHV

CCS
Figure 11. Sector-shaped conductors as ideal quarter circles in four-core NAYCWY cable.

4.3. Experimental Determination of Cable Parameters

As mentioned earlier, our investigations include measurements of the parameters
from four real cables (see Table 5) for validation of the presented calculation approaches.
Test cables A and B are of the same type with different cross-sectional areas, allowing
for conclusions regarding scalability. On the contrary, cables B, C, and D are of different
types with equal cross-sectional areas, thus allowing for comparison of the influence of
cable designs.
Although measurements are ideally carried out on infinitely long cables laid out in
a straight line in non-magnetic surroundings, this is obviously not applicable in reality.
Installation and operation of measurement devices require both ends of the cable to be close
to each other, in order to reduce parasitic effects and noise on long measurement lines. Thus,
for practical reasons, the cables were installed in a U-bend with a spacing of approximately
1.5 m between the two halves of the cable. Cable length is limited by economic reasons,
available space and cable handling, while an adequate minimal length is required for
reducing parasitic effects at the cable ends and in order to obtain technically measurable
results. During the measurements, cables B, C, and D were installed approximately 0.35 m
above laboratory floor. Cable A was installed directly on a concrete surface.
Due to low measurement currents, compared to nominal currents of the cable, we
assume the conductors’ temperature to be constant during measurement. Thus, for cables B,
C, and D with a length of 25 m, the temperature equals the laboratory’s ambient temperature
of 20 °C. Cable A, which is installed outside the laboratory, was measured at an ambient
temperature of 24 °C. For higher conductor temperatures, we injected currents of up to
400 A. Temperature is measured at the conductor ends, under the sheath, with a Pt100
resistance thermometer.

4.3.1. Resistances and Inductances

Figure 12 depicts the measurement setup for acquiring resistances, as well as self and
mutual inductances. Configured as a current source, the four-quadrant amplifier (4QA)
APS 10,000 by Spitzenberger&Spies drives current I i ≈ 30 A through conductor i, while
there is no current injected in conductor j. Current I i and the induced voltages V i1i2 and
Energies 2021, 14, 1265 18 of 34

V j1j2 across the conductors i and j are measured by the power measuring device LMG500
by ZES Zimmer.
0
We determine the measured series resistance Rii,m 0
and self inductance Lii,m of conduc-
0
tor i and the measured mutual inductance Lij,m between conductor i and j, respectively,
per unit length, according to Equations (25)–(27). In the equations, Lii and Lij are the self
and mutual conductor inductances, respectively, while Lii,p and Lij,p summarize parasitic
inductances due to non-optimal cable installation. The latter cannot be identified sepa-
rately, thus having to be considered analyzing the results (see Section 4.4.2). Cable length
in meters is described by l.

0 V i1i2 1
Rii,m = Re( )· , (25)
Ii l
0 V 1 1 1 1
Lii,m = Im( i1i2 ) · · = ( Lii + Lii,p ) · · , (26)
Ii ω l ω l
0
V j1j2 1 1 1 1
Lij,m = · · = ( Lij + Lij,p ) · · . (27)
Ii ω l ω l

External current source Measuring device

4QA LMG500
Ij

V i1i2 V
Rii Lii Lii,p
i2
Conductor i
i1

Measured cable Lij + Lij,p

NAYCWY 4×50/SE 25

j1 j2
Conductor j
R jj L jj L jj,p
V j1j2 V

Figure 12. Test setup for measuring resistances and inductances.

4.3.2. Capacitances
For capacitance measurements, we opted for the Omicron DIRANA dielectric response
analyzer [39]. During measurement of capacitance between two conductors, the other
conductors were grounded. Furthermore, the cable was installed adequately far from metal
surfaces. Hence, stray capacitances are close to zero and are neglected.

4.4. Comparison and Analysis of Calculated and Measured Cable Parameters

Although aiming for a model valid for grid frequency of 50 Hz, in general, we per-
formed measurements in the range of 40 Hz to 60 Hz in order to identify potential interfer-
ence at grid frequency.
The four test cables from Table 5 differ in conductor designs and cross-sectional areas.
Nevertheless, for all cable types, comparing the results leads to similar relations between
calculated values according to Section 4.2 and measured values according to Section 4.3.
Thus, drawn conclusions apply for all cables under investigation. In the following, we
present and discuss exemplary results.
Energies 2021, 14, 1265 19 of 34

4.4.1. Conductor Resistances

Figure 13 compares measured resistances to those calculated, for frequencies between
DC and 57 Hz. The calculated values are, in accordance with Section 4.2.1, derived from
the German standard DIN VDE 0295 [33] and corrected for temperature deviation from
20 °C, skin and proximity effect (see Equations (10) and (11)). During measurements, only
one conductor was loaded. The error bars consider measurement tolerance of the LMG500
and uncertainties in temperature measurement.

0.23 Conductor a
Conductor b
0.225 Conductor c
Conductor n
R0 in Ω/km

0.22 calculated

0.215

0.21

0.205
DC 47 50 53 57
Frequency in Hz

Figure 13. Calculated and measured resistances at ϑ = 24 ± 2 °C (NAYY-J 4 × 150 SE, cable A).

The figure shows that measured DC resistances of all four conductors are in good
agreement with the calculated ones. For calculated parameters of AC resistances, skin
and proximity effect have a rather low influence at the considered frequencies. However,
there is a significant discrepancy between calculated and measured values of up to 8% at a
frequency of 50 Hz. Hence, measurement results show a stronger frequency dependency
than calculated AC resistances. We assume the differences to result from eddy currents
induced in unloaded neighboring cable conductors and, thus, as part of proximity effect.
In Section 4.2.1, the proximity effect factor FP is calculated by an approach for symmetrical
arrangements of three round conductors. In contrast, the measured low voltage cable
in Figure 13 is composed by four asymmetrically arranged sector-shaped conductors.
In addition, as only one conductor is loaded, the cable load is significantly unbalanced
during measurement. We consider this discrepancy to be a potential reason for inaccurate
representation of proximity effect.
For neighboring conductors, considering the large cross-section and comparatively
thin insulation, the magnetic field facing the loaded conductor is approximately 4.6 times
higher than the one on the far side (dins = 1.8 mm, qreal = 132 mm2 ), which leads to
strongly inhomogeneous currents in the sector-shaped conductors. Applying a contrary
current on a neighboring conductor or symmetric three-phase currents to the cables led
to deviations in measurement results, which supports the assumption of induced eddy
currents being responsible for effective losses.
Eddy currents possibly induced in the floor and other metallic components in labo-
ratory surroundings might also lead to increased measured resistances. Unfortunately, a
more detailed breakdown is not possible, as material conductivity and dimensions cannot
be evaluated reasonably.
Furthermore, Figure 13 shows—for every measured frequency—an offset of approx-
imately 2% between conductors c and n. As measurement methods were the same for
all conductors, and repeated measurements led to the same results, we assume these
deviations to result from manufacturing tolerances.
For calculation of resistances, a linear temperature dependency of aluminum with
a temperature coefficient of αAlu = 0.004 03 K1 is assumed. Regarding temperature de-
pendency of our test objects, measurement results for conductor n of test cable A show
a linear dependency for three different temperatures (Figure 14). While we expected an
Energies 2021, 14, 1265 20 of 34

increase in resistance of 10% compared to 24 °C for a conductor temperature of 48 °C, the

measured increase is approximately 6%. This discrepancy can be explained by difficulties
in accurately determining conductor temperature and maintaining a constant conductor
temperature (with respect to time, as well as temperature gradients, across the cable length)
due to effects of wind and solar radiation during measuring.
0.25
n (ϑ = 24 ± 2°C)
n (ϑ = 39 ± 3°C)
n (ϑ = 48 ± 3°C)
0.245

0.24
R0 in Ω/km

0.235

0.23

0.225

0.22
45 50 55 60
Frequency in Hz

Figure 14. Measured resistances at different temperatures (NAYY-J 4 × 150 SE, cable A).

Summarizing, we can state that the calculated DC resistance values correspond well
to those measured in real exemplary cables. However, considering asymmetric load situa-
tions, four-core cable designs, as well as non-negligible losses and resulting temperature
deviations, measurement results diverge from the results of usual calculations. Hence, for
detailed simulations of German LV systems, an in-detail consideration of resistance param-
eters is required. Particularly the influence of eddy currents induced in other conductors
requires further investigations.

4.4.2. Self and Mutual Inductances

Figure 15 compares measured (dots) and calculated mutual inductances (hatched areas,
Appendix A, Table A1) for test cables A, C, and D. Again, error bars refer to measurement
tolerances according to the LMG500 data sheet.
Cable B only differs from C in its nonexistent shield. Subsequently, as the results of
test cable B are very similar to those of cable C, measurements confirm the assumption that
the nonmagnetic shield does not affect mutual coupling between sector-shaped conductors.
Nevertheless, for all test cables, the results depicted in Figure 15 show a nearly
constant, slightly frequency dependent offset between measured and calculated values—
the highest for test cable D, where measured mutual inductances are approximately 10%
above computed ones. For both test cables B and C, deviation is approximately 3–4%. For
the three-core cable D, deviation is actually negative. For self inductances (see Figure 16),
we observed the same phenomena. The reasons for these discrepancies are still not clearly
identifiable. Yet, there are several possible explanations:
Stray magnetic flux Due to the finite length of test cables, stray magnetic flux at the
cable ends can distort measured inductances. Nevertheless, this influence should
actually become less relevant with an increasing length of the test object.
Test cable setup As described in Section 4.3, test cables were installed in a U-bend,
forming a kind of one-turn loop. This most likely leads to an increase in both self
inductance and mutual coupling between the conductors.
Laboratory environment Influences on the magnetic field due to the floor and other
possibly magnetizable surfaces and components in the test cable surroundings cannot
be excluded.
Energies 2021, 14, 1265 21 of 34

Approximate formulas Equations (15) and (16) for calculation of self and mutual
inductances were originally developed for round and infinitely long conductors,
relatively far distanced from each other compared to their radius. Thus, for sector-
shaped conductors with large cross-sectional areas and comparatively small geometric
mean distances between them, the accuracy of the approximation is questionable.

0.96 a–b
0.94 b-c
a–n
0.92
c–n
0.9 a–c
L0 in mH/km

b–n
0.88
0.86
0.84 Horizontal/vertical sector-shaped conductor pairs (calculated)

0.82
0.8
0.78
Diagonal sector-shaped conductor pairs (calculated)
0.76
45 50 55 60 65 70
Frequency in Hz

(a) NAYY 4 × 150 SE (cable A)

0.99 a–b
0.98 b–c
a–n
0.97 c–n
0.96 a–c
L0 in mH/km

Horizontal/vertical sector-shaped conductor pairs (calculated) b–n

0.95 a–S
0.94 b–S
c–S
0.93 n–S
0.92
0.91
0.9
Sector-shaped conductor–shield pairs (calculated)
0.89
0.88
0.87
Diagonal sector-shaped conductor pairs (calculated)
0.86
45 50 55 60 65 70
Frequency in Hz

(b) NAYCWY 4 × 50 SE/25 (cable C)

0.96 a–b
0.95 b–c
a–c
0.94 a–S
b–S
0.93 c–S
L0 in mH/km

0.92 n–S

0.91
0.9
0.89
0.88
0.87
0.86
45 50 55 60 65 70
Frequency in Hz

(c) NAYCWY 3 × 50 SE/50 (cable D)

Figure 15. Comparison of calculated and measured mutual inductances.
Energies 2021, 14, 1265 22 of 34

1.2 Conductor a
Conductor b
Conductor c
Conductor n
1.15

L0 in mH/km
1.1

Sector-shaped conductor (calculated)

1.05
45 50 55 60 65 70
Frequency in Hz

Figure 16. Measured self inductances (NAYY-J 4 × 150 SE, cable A).

Lastly, considering thermal influences, and thus comparing measurement results of

test cable D at approximately 24 °C to results obtained at higher conductor temperatures,
only insignificantly small differences in measurement results—all within the measure-
ment tolerance range—could be detected. This validates our assumption of temperature-
independent permeability of cable and soil materials.

4.4.3. Capacitances
Table 7 summarizes mean values of measured capacitances for 50 Hz and different
types of conductor pairs. Between pairs of the same type (e.g., c–n versus a–b as hori-
zontally/vertically neighboring pairs), we could measure differences of up to 3% in the
same cable. This most likely reflects manufacturing tolerance in insulation thickness. It
can be stated that all measured values in Table 7 coincide with calculated ranges values in
Appendix A, Table A1. This is valid regardless of the idealization of the sector-shaped
conductors as ideal third or quarter circles.

µF
Table 7. Measured capacitances in km at 50 Hz and ϑ ≈ 20–24 °C.

Cable Type 0
CHV 0
CDiag 0
CCS
A NAYY-J 4 × 150 0.17 0.01 –
B NAYY-J 4 × 50 0.13 0.01 –
C NAYCWY 4 × 50 SE/25 0.11 0.01 0.22
D NAYCWY 3 × 50 SE/50 0.09–0.10 – 0.35

As only one conductor was heated by the heat transformer, there was a temperature
gradient across the PVC layers to neighboring conductors, which further complicated an
accurate analysis of temperature dependency of er . Nonetheless, our results confirm that
temperature dependency generally is noticeable. For a temperature of the conductor of
approximately 38 °C, we could measure an increased capacitance of approximately 6%
compared to a conductor temperature of 24 °C.

5. Grounding Impedance of Grid Nodes

The grounding electrode at customer nodes is realized in various ways. For older
buildings, grounding is often realized through metallic water and gas pipes entering
the ground. For new buildings, in Germany it is mandatory to install a ring of stainless
steel strip or round steel in the external walls of the building foundation as a foundation
grounding electrode [40]. Hence, estimating a grounding system’s impedance depends
on a multitude of parameters and differs between buildings. However, some general
assumptions can be made, and are presented in the following sections.

5.1. System Identification and Parameter Definition

Generally, the impedance triangle depicted in Figure 17 is valid for any grounding
system. Resistance RC and reactance XC refer to the grounding conductor, which connects a
building’s main earth bar to the metallic grounding electrode (see Figure 5). The impedance
Energies 2021, 14, 1265 23 of 34

of the metallic grounding electrode is represented by RM and XM . The complex grounding

impedance is, thus, calculated as

ZG = RG + jXG (28)

The reactive component XG in Equation (28) is the sum of the grounding conductor’s
reactance XC and the reactance of the metallic grounding electrode XM (see Figure 5).
This reactance XG only needs to be considered when analyzing high frequency transients,
such as those caused by lightning and switching surges, and is—compared to the resistive
component RG —negligible at 50 Hz [13].
The resistive component RG consists of the grounding conductor resistance RC , the
metallic grounding electrode resistance RM and the so-called propagation resistance RP . The
latter describes how efficient the grounding system can transmit the current and, thus,
models the resistance between the earthing electrode and the reference ground. Hence, RP
depends on the soil resistivity ρE , as well as the type and the geometric dimensions of the
grounding electrode [13].

ZG XM
XG
XC

RP RM RC

RG

Figure 17. A grounding system’s impedance triangle, based on Reference [41].

As RC and RM are negligible compared to RP [13], and further neglecting the reactive
component XG , we model the grounding impedance ZG as approximately equal to RP :

Z G ≈ RG ≈ RP . (29)

5.2. Grounding Impedance of Typical Households

As stated in Section 1, due to material, temperature, and humidity variations, among
others, there is a significant uncertainty in determining the value of ρE . Therefore, it does
not seem reasonable to calculate grounding resistances RP for different types of grounding
systems and environmental conditions in detail. In order to get a reference value for
simulation models, RP is determined for the foundation electrode of a typical German
single family house with foundation dimensions of 10 m × 15 m × 0.25 m [42]. Approxi-
mate formulas for grounding resistance calculations are specified in Reference [41,43]. In
both cases, RP is proportional to ρE (see Table 8). Thus, we assume a standard value of
ρE = 100 Ω m [11].
Equation (30) from Reference [43] uses the approximation of a hemispherical ground-
ing electrode. Variable d corresponds to the diameter of a sphere of equal volume. Applying
Equation (30) to the previously described example leads to a value of RP ≈ 6.1 Ω (Table 8).
The authors of Reference [41] base their formula, Equation (31), on calculations for
soil-embedded circular ring electrodes. The correction factor 1.05 takes into account that
the foundation electrode is embedded in concrete. Concrete is less conductive than the
surrounding soil. The geometrical constant g accounts for the influence of grounding
electrode design and distance from the terrain surface. In order to determine g, a few more
parameter values (e.g., thickness of the stainless steel strip and depth below the frost line)
need to be known. Using exemplary values from Reference [41] leads to a resistance of
RP ≈ 5.4 Ω (Table 8). For the sake of simplicity, and as an exact value cannot be found, for
our models, we assume an average grounding impedance at costumer connection points to
equal ZG,build = 6 Ω.
Energies 2021, 14, 1265 24 of 34

Table 8. Comparison of two different methods to approximate RP of a foundation grounding device.

Reference Formula Approximation Based on Result

ρE
[43] RP = (30) Hemispherical grounding electrode RP ≈ 6.1 Ω
π·d
ρE
[41] RP = 1.05 (31) Soil-embedded circular ring electrode RP ≈ 5.4 Ω
g

5.3. Grounding Impedance of Substations

As depicted in Figure 7, in the approach presented in the present article, the neutral
point of the substation transformer is not modeled as “perfectly grounded”, but with an
impedance ZG,xfmr . For this impedance, we assume a value of 2 Ω, as DSOs usually consider
a grounding impedance of 2 Ω as sufficiently low when designing MV/LV-substation
grounding systems.

6. Parameter Relevance and Sensitivity

In order to evaluate the approach’s parameter relevance and sensitivity to parame-
ter changes, we integrated our modeling approach in the simulation of a test feeder in
MATLAB/Simulink.

6.1. Setup of Test Feeder Model for Parameter Analysis

Figure 18 depicts the layout of the simulated LV feeder, which is based on feeder A-2
(see Section 6, Table 1). Having a total length of more than 1 km, feeder A-2 classifies as
“extremely weak” according to the analysis of German LV grids in Reference [7]. Therefore,
we expect significant voltage asymmetry under unbalanced load conditions.

13 15 19
14
50 m 16
50 m
30 m

40 m
LV busbar
LV/MV- 20 m
transformer
3
721 m 2 275 m 30 m 4 10 m 5 40 m 20 m 10 m 8 30 m
9
10 m

10 m
20 m

10 m
7
6
30 m

1 12 17 18

11
NAYCWY 3×150/95 Cable distribution cabinet
NAYY-J 4×150 10
Cable joint
NAYY-J 4×50
Costumer connection

Figure 18. Sketch of test feeder

6.1.1. Modeling of Transformer, Cable Segments and Grounding

Since our analysis focuses on modeling of cable segments and grounding impedance,
we model the transformer as an ideal three-phase 230 V, 50 Hz voltage source. Its neutral
point, and thus the LV PEN busbar, is grounded with an impedance of ZG,xfmr = 2 Ω
(see Section 5.3). As suggested in Section 5, foundation grounding of costumer nodes is
modeled through an impedance of ZG,build = 6 Ω.
Usually, voltage drop across house junction cables is neglected when modeling and
simulating LV feeders [7,25]. As outlined in Figure 18, we explicitly model house junction
cables in the reference scenario, which allows for comparing the effect of other model
simplifications and parameter variations to the effect of neglecting house junction cables
(see Section 6.2.2). The cable segments are modeled according to Section 3, Figure 7 as four-
line π-model, and model parameters calculated according to Section 4. The line between
nodes 1 and 2 is modeled as a cable of type NAYCWY 3 × 150 SE/95.
Energies 2021, 14, 1265 25 of 34

6.1.2. Modeling of Load Distribution

By an unbalanced assignment of six single-phase PV feed-ins (four to phase a, two to
phase b) and one single-phase electric vehicle (EV) charging process at phase c in addition to
the base load of costumer nodes, we induce a significantly asymmetrical load scenario (see
Table 9). Since the active power values of ±4.6 kW correspond to the maximum permitted
single-phase active power load or feed-in according to German DSO grid codes [44], the
scenario may be considered “worst-case”, but not necessarily unrealistic.

Table 9. Allocation of loads to costumer nodes and phases (Base scenario K0).

Base Load PV Plant EV

PQ Load PQ Load PQ Load
Node
P = + 100 W P = − 4600 W P = + 4600 W
cos ϕ = 0.9 (ind.) cos ϕ = 1 cos ϕ = 1
10 a, b, c – –
11 a, b, c b –
12 a, b, c a –
13 a, b, c b –
14 a, b, c – –
15 a, b, c a –
16 a, b, c – –
17 a, b, c a –
18 a, b, c – –
19 a, b, c a c

In the reference scenario—modeled with the new detailed modeling approach as de-
scribed in the previous paragraphs, without any parameter variations or model
simplifications—line-to-neutral voltages at the farthest costumer node, node 19, are cal-
culated to V19,an = 1.15 pu, V19,bn = 0.86 pu and V19,cn = 1.07 pu. Hence, according to
European standard DIN EN 50160 [45], the voltage of phase a is outside the permitted
voltage range of 0.85 pu to 1.10 pu.
Considering current magnitudes, due to the load situation, neutral current IPEN = 81 A
in the cable segments between node 1 and 3 is higher than the phase currents Ia = 67 A,
Ib = 29 A and Ic = 33 A, which is a typical current distribution for highly unbalanced load
conditions. Ground current IG through the transformer’s grounding impedance is 10 A,
thus equaling approximately 11% of neutral current IPEN , which is in accordance with our
field measurements (see Section 2).
In the following, we analyze the changes of simulation results relatively to the results
of the reference scenario, varying cable and ground return path modeling.

6.2. Modifications in Modeling the Cable Components

Figure 19 plots line-to-neutral phase voltages at node 19 in the reference scenario as
solid horizontal lines. Simulated voltages of scenarios S1 to S4 are plotted as dots, each
scenario referring to a simulation with a modification in cable models. In the following,
only magnitudes of currents and voltages are considered.
Energies 2021, 14, 1265 26 of 34

Van S0
1.15 Vbn reference
Vcn scenario
1.10

1.05

Voltage in pu
1.00
Permissible
0.95 voltage range
acc.
0.90 DIN EN 50160 [45]

0.85

ap. . ind
.
0 °C °C ndin
g Ω
w /o c mut o house cables a ϑ = 3 = 70 grou =2 =1
Ω
S1a b w / o
w / t i on S3 S 3b ϑ
a w / o
b R G,bu
ild
c R G,xfm
r
S 1 S 2 jun c S 4 S 4 S4

Figure 19. Voltages at node 19 for different model modifications.

6.2.1. Neglecting Electrical or Magnetic Coupling between Conductors (S1a, S1b)

In the reference scenario, the currents flowing through capacitances between conduc-
tors are in the range of a few mA. Removing those capacitances, and thus the electrical
coupling, from the model (scenario S1a), the resulting missing reactive power compensa-
tion leads to a slight increase of inductive reactive power feed-in of approximately 25 var
at node 1. At node 19, compared to the reference scenario, a negligible difference in phase
voltages occurs.
However, a change in phase voltages of up to 0.1 pu is visible for scenario S1b, where
mutual inductances, and thus the magnetic coupling between conductors, have been
removed. Remarkably, the order of phase voltages also changes such that V19,cn exceeds
V19,an in this scenario only.

6.2.2. Neglecting House Junction Cables (S2)

For scenario S2, all house junction cables are removed from the model. Due to resulting
omitted line losses, active power feed-in at node 1 is decreased by approximately 40 W per
phase. Furthermore, there is a slight decrease in voltage asymmetry at node 19. With an
increased phase voltage of approximately 0.01 pu, the biggest change is visible for phase b.

6.2.3. Variation in Conductor Temperature (S3a, S3b)

Scenarios S3a and S3b are used to assess the relevance of considering higher conduc-
tor temperatures. During summer and for load currents equal to around 50 % of rated
conductor currents, conductor temperatures of 30 °C, assumed for scenario S3a, are quite
common [23]. As shown in Figure 19, changes in phase voltages at node 19 are of the same
magnitude as those neglecting house junction cables in scenario S2, only now increasing
voltage asymmetry due to higher conductor impedances. Since the impedance of the
neutral conductor also increases, current through ground return path, IG , increases by
approximately 4%.
As a worst-case assessment, scenario S3b assumes that all conductors in all cable
segments have a temperature of 70 °C, typically the maximum rated temperature. This
leads to a 3% increase in phase voltages at node 19, and a significant rise in voltage
asymmetry. In scenario S3b, voltage V19,bn falls below power quality limits defined in the
European standard DIN EN 50160 [45].

6.2.4. Modifications in Modeling the Ground Return Path (S4a–S4c)

In the following, we analyze the influence of customer node and transformer ground-
ing impedances. For scenario S4a, all grounding impedances have been deleted from
the model. Thus, the neutral conductor is the only remaining return path and IG = 0 A.
This leads to increased asymmetries in node voltages, as the voltage drop across the PEN
Energies 2021, 14, 1265 27 of 34

conductor rises—V19,bn drops by 0.015 pu, while V19,an and V19,cn rise by 0.009 pu and
0.007 pu, respectively.
Decreasing RG,build from 6 Ω to 2 Ω in Scenario S4b—which is a typical value for
urban areas with many foundation grounding electrodes close to each other [11]—does
not influence simulated voltage levels much. However, decreasing RG,xfmr from 2 Ω to 1 Ω
(S4c) leads to slightly smaller voltage deltas along the lines.

6.3. Modifications in Connection Phase

Lastly, we assess how interchanging loads between phases a, b and c effects overall
simulation results. Therefore, we define four different load configuration scenarios (see
Table 10). K0 corresponds to the load distribution introduced in Table 9, Section 6.1.2, and
is considered as base scenario. In scenarios K1, K2 and K3, loads are connected to the
same costumer nodes, but to different phases. For scenario K2, the base scenario’s load
allocation to phases is shifted in a circular manner; thus, the phase sequence remains equal.
In contrast, K1 and K3 are two permutations with reversed phase sequence.

Table 10. Configurations for distribution of load groups to connection phases.

K0 K1 K2 K3
a 4PV 4PV 2PV 1EV
b 1EV 2PV 4PV 4PV
c 2PV 1EV 1EV 2PV

All scenarios K are simulated for both, a symmetrically designed feeder—all line seg-
ments modeled as symmetrically designed three-core NAYCWY cables (see Section 6.3.1)—
and a feeder composed by asymmetrically designed NAYY-J four-core cables only (see
Section 6.3.2).

6.3.1. Line Model: Symmetrically Designed NAYCWY Three-Core Cable

Figure 20a shows the simulated voltage magnitudes for using symmetrical NAYCWY
three-core line models. Here, again, colored dots indicate the voltage levels Van , Vbn and
Vcn at node 19 of the previously introduced feeder (see Figure 18).
For the scenarios with similar phase sequences, K0 / K2 and K1 / K3, voltage magni-
tudes remain at the same level. However, comparing the results from these two groups,
differences in voltage magnitudes can be observed. We assume this to be due to phase
currents adding up differently in the neutral conductor for different phase sequences.
This leads to different voltage drops across the neutral conductor and, thus, to different
magnitudes in phase–neutral voltages.

6.3.2. Line Model: Asymmetrically Designed NAYY-J Four-Core Cable

Figure 20b shows the simulated voltage magnitudes for using asymmetrical NAYY-J
four-core line models, again for node 19 of the same test feeder.
Remarkably, in contrast to simulations of the symmetrically designed grid in
Section 6.3.1, changes in voltage also can be observed when phase sequence remains
the same (such as between K0 and K2). In addition, when phase sequence changes (such as
between K0 and K2), changes in voltage magnitudes are considerably bigger than in the
symmetrically designed grid. We assume this to result from unequal mutual inductances
between conductors.
Energies 2021, 14, 1265 28 of 34

1.15 Van
Vbn
1.10 Vcn
1.05

Voltage in pu
1.00
0.95
0.90
0.85
K0 K1 K2 K3

(a) Symmetrical lines (only NAYCWY three-sector cables).

1.15 Van
Vbn
1.10 Vcn
1.05
Voltage in pu

1.00
0.95
0.90
0.85
K0 K1 K2 K3

(b) Asymmetrical lines (only NAYY-J cables).

Figure 20. Voltages at node 19 for configurations K0–K4.

6.4. Conclusions of Simulation Model Variations

In this section, simulation design and model parameters were compared for a multi-
tude of variations. Concluding the results shown, some assertions can be summarized:
Capacitances: For our investigations, capacitances are neglectable at 50 Hz, at least
compared to the impact of other modeling inaccuracies.
Inductances Mutual inductances between conductors can have a substantial impact
on simulation results, particularly for asymmetrically designed lines.
House junction cables: Typical LV grid simulations neglect house junction cables.
However, as results show, for detailed modeling of LV grids the impact is not negligible
due to the rather high impedance of typically utilized cables.
Conductor temperature: Assuming a conductor temperature of 70 °C might be unreal-
istic, but is still relevant for worst-case scenarios, since changes in conductor resistance
due to temperature increase have a major influence on phase voltages. Especially for
highly loaded grids, modeling of cable temperature should be considered.
Ground return path: Our investigations have shown that modeling of the ground
return path is relevant when simulating grids with multiple groundings, since delet-
ing all grounding impedances in our model significantly increases voltage asymme-
try. Small changes in the range of a few Ω in the costumer buildings grounding
impedances ZG,build , however, only lead to small changes in simulated voltages. This
validates the idea of using exemplarily calculated reference values for parametrization
of ZG,build (see Section 5.2).
Load distribution & phase sequence: As shown, distribution of loads to phases and
corresponding phase sequence have a major impact on simulation results. This impact
is considerably bigger for asymmetrically designed lines. Thus, for worst-case studies
aiming at estimating most critical voltage values, careful considerations of potential
load distribution have to be carried out.

7. Model Validation with Data from Real LV Feeder

In this section, we validate our cable and grounding models with the help of mea-
surements of the real LV feeder A-2, which is also used for the studies described in
Section 6. As reference values, we rely on measurement data acquired at the MV/LV
transformer busbar (xfmr) and in between nodes 2 and 3 (load), where the cable can be
accessed with measurement equipment. Measurements contain magnitudes and phase
Energies 2021, 14, 1265 29 of 34

angles of voltages, as well as active and reactive powers for each phase, recorded with time
synchronous power quality analyzers [46].
For the simulation, over the period of 420 seconds, voltages at the LV busbar are set to
the measured values. Measured powers are fed in at the measuring point, representing
the remainder of the feeder (nodes 3 to 19) as an equivalent load. This setup results in a
simulated feeder consisting of a voltage source with given voltages, two different lines,
i.e., NAYCWY 3 × 150/95 (721 m) and NAYY-J 4 × 150 (130 m), and an equivalent PQ load.
The loads fed in during the simulated period are depicted in Figure 21.
Cable segments are parametrized according to Section 4, assuming a temperature of
20 °C. As measurements of the feeder A-2 show a ratio of 0.19 between ground and PEN
currents (see Table 1), we adjusted grounding impedances to reproduce this ratio.

1000

500
•
0

−500 •••••••••••••••••••••••••••••••••••••••...................................... � ... ... .············· ······ ········· -..

..
••••••••••• �

- ............
············································� :
························:
•••
. ....
• t•····-41111
• .

••••••••••••••••••••••••••••••••••••••• ········································································································�1······················································., •••••••

..

−1000 ·············································································· .......................••...•.••.......•............................................. .--····································································. ..... ••••••••

- Pa,load - Pb,load - Pc,load

−1500 1111111

Q a,load 1111111

Qb,load 1111111

Q c,load
−2000
0 50 100 150 200 250 300 350 400
Tim lll S

Figure 21. Active and reactive phase loads.

Simulation results are shown in Figure 22, which depicts the measured voltage magni-
tudes at the transformer (brown/black/grey, dotted) and measuring point
(brown/black/grey, solid), as well as the results of the simulation (blue), separated by
phase. We can state that measured voltages Vin,load,meas and simulated voltages Vin,load,sim
correlate quite well. This is visible and valid particularly for events of heavy load shifts, as
the sudden changes of voltages are reproduced by our simulation model—even in cases,
where the load change is applied on another phase. However, in all curves, an offset
between measurement results and simulated voltages is visible, which depends on the
momentary load level and load distribution.
Possible reasons for the deviations between measured and simulated phase voltages
are listed as follows:
Inaccurate model parameters As discussed in Sections 4.4.1 and 4.4.2, calculated resis-
tance and inductance parameters differ from those values measured at test cables. This
indicates that discrepancies between assumptions made for calculation of parameters
and reality do exist. In addition, conductor temperature might have been different to
the assumed 20 °C.
Grounding model & impedances Only one feeder is simulated; thus, ground currents
are forced to return to the transformer. In reality, ground currents possibly take
different paths, thus influencing real voltages (see Section 3.3). In addition, grounding
impedances are parametrized based on measurements of another day, when ratio of
PEN to ground currents most likely was different.
Feeder setup simplifications For the simulations carried out in this Section, the origi-
nal feeder A-2 (see Figure 18) was reduced to three nodes. Hence, additional inaccura-
cies concerning individual grounding impedances and phase loads—and resulting
ground currents—can be seen as possible reasons for discrepancies between simulated
and measured phase voltages.
Measurement results & feeder data Even though measurements were carried out
with reasonable care, inaccuracies in measurement data and noise, as well as in
available grid data, cannot completely be excluded.
Energies 2021, 14, 1265 30 of 34

Nevertheless, our modeling approach succeeds in replicating measured voltage levels

quite well. Hence, we consider our approach as promising for detailed grid simulations.
We will carry out further investigations and additional validations comparing simulation
results and real measurement data in the future.

(a) Phase a

(b) Phase b

(c) Phase c

Figure 22. Comparison of measured and simulated phase voltages.

8. Conclusions
In this paper, we have taken a closer look on accurate modeling of low voltage
cables and grounding conditions in German low voltage grids. Taking all results into
consideration, we sum up and conclude our work in the following.

8.1. Derivation and Parametrization of an Alternative Modeling Approach

Our analysis of measurement data from real low voltage substations in
Section 2 has underlined the importance of modeling the ground return path in Ger-
man low voltage grids. Conventionally, Carson’s approach (see Section 3.2.4) is used to
account for this ground return path. However, the measurement data in Section 2 suggests
that Carson’s underlying hypothesis of ground currents bound to paths in parallel to cables
does not hold in German low voltage grids. Thus, we have chosen an alternative modeling
approach, based on modeling ground as single electrical point. Further investigations
and a harmonic analysis of neutral and ground currents are required in order to get more
detailed knowledge of grounding of German TN-C-S systems.
This alternative modeling approach, in contrast to common strongly simplified ap-
proaches, also accounts for capacitances, asymmetric magnetic coupling in four-core cables
and temperature dependent resistances. Since literature does not provide any suitable
Energies 2021, 14, 1265 31 of 34

parameters for this alternative modeling approach, we have presented calculation methods
based on geometric dimensions of conductors and insulation.
For evaluating our calculated parameters, we have conducted measurements on four
real low voltage cables of different conductor designs and cross-sectional areas. Measured
values of calculated capacitances and DC resistances coincide for all cable types. For
inductances and resistances, in contrast, we have found discrepancies in the range of
3–10% (see Section 4.3). In order to clearly identify the reasons, future work will focus on
analyzing magnetic field conditions and proximity effect in asymmetrical arrangements of
sector-shaped conductors and unbalanced load scenarios.

8.2. Analysis of Parameter Relevance and Sensitivity

By integrating the presented modeling approach in a simulation environment based
on a real LV feeder in Section 6, we have analyzed the approach’s parameter relevance and
sensitivity to parameter changes. We have found that detailed modeling of house junction
cables—which is commonly neglected—can be relevant depending on the required level of
detail of simulations. The influence of neglecting capacitances, in contrast, is negligible,
at least at mains nominal frequency of 50 Hz. Further work needs to be carried out to
analyze the relevance of modeling capacitances at higher frequencies, i.e., when accounting
for harmonics.
Furthermore, investigations show that temperature-dependent line resistances signifi-
cantly affect voltage drops across lines (see Section 4.2). In addition, the load distribution
and phase sequence play an important role (see Section 6.3). Both should be considered for
worst-case studies on voltage quality.

Author Contributions: Conceptualization, S.H., L.H., and J.G.-S.; methodology, S.H., L.H., and
J.G.-S.; software, J.G.-S. and F.G.; validation, S.H., J.G.-S., F.G., and L.H.; formal analysis, J.G.-S., S.H.,
L.H., and F.G.; investigation, J.G.-S., S.H., L.H., and F.G.; resources, T.L., M.S., M.A.; data curation,
J.G.-S.; writing—original draft preparation, J.G.-S., S.H., L.H., and F.G.; writing—review and editing,
J.G.-S., S.H., L.H., F.G., M.S., and T.L.; visualization, J.G.-S., S.H., L.H., and F.G.; supervision, M.S.
and T.L.; project administration, M.S. and T.L.; funding acquisition, S.H., M.S., and T.L. All authors
have read and agreed to the published version of the manuscript.
Funding: This research was funded by the German Federal Ministry for Economic Affairs and Energy
within the research project flexQgrid, grant number 03EI4002F.
Acknowledgments: Many thanks to our partners, the distribution system operators Netze BW GmbH
and Stadtwerke Buehl GmbH, as well as their employees, who supported us during measurements
in their distribution grids. We greatly appreciate the provision of grid data and good cooperation.
Additionally, we acknowledge support by the KIT-Publication Fund of the Karlsruhe Institute
of Technology.
Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design
of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or
in the decision to publish the results.
Energies 2021, 14, 1265 32 of 34

Appendix A. Model Parameters Calculated acc. Section 4

Table A1. Model parameters acc. Section 4, calculated for ϑ = 20 °C. Datasheets: Refs. [30–32,47].

NAYY-J 4 × 50 SE NAYCWY 3 × 50 SE/50 NAYCWY 4 × 50 SE/25

Ω
Resistances in km (from Reference [6])
0
RCC 0.641 0.641 0.641
0
RSS – 0.641 1.20
mH
Self inductances in km
0
LCC 1.168 1.168 1.168
0
LSS – 0.909–0.919 0.882–0.892
mH
Mutual inductances in km
0
LHV 0.940–0.951 0.940–0.951 0.940–0.951
0
LDiag 0.871–0.882 – 0.871–0.882
0
LCS – 0.909–0.919 0.882–0.892
µF
Capacitances between conductors km
0
CHv 0.09–0.15 0.08–0.12. 0.09–0,15
0
CDiag 0.00–0.01 – 0.00–0.01
0
CCS – 0.25–0.38 0.22–0.33
µF
Capacitances to ground km (from Reference [6])
0
CCG 0.48 – –
0
CSG – 0.48 0.48
NAYY-J 4 × 150 SE NAYCWY 3 × 150 SE/95
Ω
Resistances in km (from Reference [6])
0
RCC 0.207 0.207
0
RSS – 0.320
mH
Self inductances in km
0
LCC 1.058 1.058
0
LSS – 0.819–0.827
mH
Mutual inductances in km
0
LHV 0.848–0.856 0.847–0.855
0
LDiag 0.779–0.787 –
0
LCS – 0.819–0.827
µF
Mutual capacitances between conductors in km
0
CHV 0.12–0.19 0.1..0.16
0
CDiag 0.01 –
0
CCS – 0.33–0.49
µF
Capacitances to ground in km (from Reference [6])
0
CCG 0.66 –
0
CSG – 0.66

Appendix A.1. Meaning of Subscripts

CC / SS Sector-shaped conductor / shield to itself (self impedance: resistance and self inductance)
LS Sector-shaped conductor–shield pairs
HV Horizontal/vertical sector-shaped conductor pairs
Diag Diagonal sector-shaped conductor pairs
Energies 2021, 14, 1265 33 of 34

Indicated ranges in Table A1 reflect tolerances given for d and dins in standard [37], as
well as uncertainties regarding the value of er,PVC (between 3 and 4 according to Ref. [24]).
For the simulations and the sensitivity analysis in Section 6, we used the maximum values.

References
1. Schwab, A.J. Elektroenergiesysteme, 4th ed.; Springer Vieweg: Wiesbaden, Germany, 2015.
2. Pardatscher, R. Planungskriterien und Spannungsqualität in Mittel- und Niederspannungsnetzen mit hoher Photovoltaik-
Einspeisung. Ph.D. Thesis, Technical University of Munich, Munich, Germany, 2015.
3. de Castro, D.B.; Rezania, R.; Litzlbauer, M. V2G-Strategies: Auswirkung verschiedener Elektromobilitätsszenarien auf die
Spannungsqualität von Niederspannungsnetzen unter Betrachtung der Phasenunsymmetrie. In Proceedings of the 12th
Symposium Energieinnovation, Graz, Austria, 15–17 February 2012.
4. Wagler, M.; Witzmann, R. Effects of asymmetrically connected PV and battery systems on the node voltages and pen-conductor
currents in low-voltage grids. In Proceedings of the 24th International Conference & Exhibition on Electricity Distribution
(CIRED), Glasgow, UK, 12–15 June 2017.
5. Bundesverband der Energie- und Wasserwirtschaft e. V. (BDEW). Länge des deutschen Stromnetzes. Available online:
https://www.bdew.de/media/documents/PI_20181204_Zeitreihe-Stromnetze.pdf (accessed on 5 March 2020).
6. Oeding, D.; Oswald, B.R. Elektrische Kraftwerke und Netze, 8th ed.; Springer Vieweg: Wiesbaden, Germany, 2016.
7. Kerber, G. Aufnahmefähigkeit von Niederspannungsverteilnetzen für die Einspeisung aus Photovoltaikkleinanlagen. Ph.D.
Thesis, Technical University of Munich, Munich, Germany, 2011.
8. elektro4000.de. Available online: https://www.elektro4000.de/Kabel-Leitungen/Kabel/Starkstromkabel/Verschiedene-Diverse-
K-L-NAYY-J-4x150-SE-Eca-S-NAYY-J-4x150-SE-Eca::4302.html (accessed on 6 March 2020).
9. elektro4000.de. Available online: https://www.elektro4000.de/Kabel-Leitungen/Kabel/Starkstromkabel/Verschiedene-Diverse-
K-L-NAYCWY-4x-25-RE-16-S-NAYCWY-4x-25-RE-16::1024075.html (accessed on 6 March 2020).
10. Kiefer, G.; Schmolke, H. VDE 0100 und die Praxis, 16th ed.; VDE-Verlag: Berlin, Germany, 2017.
11. Niemand, T.; Schröder, A. Erdungsanlagen, 2nd ed.; VDE Verlag: Berlin, Germany, 2016.
12. Kasicki, I. Elektrotechnik für Architekten, Bauingenieure und Gebäudetechniker—Grundlagen und Anwendung in der Gebäudeplanung,
2nd ed.; Springer Vieweg: Wiesbaden, Germany, 2018.
13. Markiewicz, H.; Klajn, A. Erdungsysteme—Grundlagen der Berechnung und Auslegung. In Schriftenreihe “Erdung und
elektromagnetische Verträglichkeit”; Deutsches Kupferinstitut: Düsseldorf, Germany, 2003.
14. DEHN + SÖHNE GmbH + Co.KG. Produktdatenblatt Bänder. Available online: https://www.dehn.de/store/f/1845384/
Artikelnummer_PDF/810225.pdf (accessed on 1 April 2020).
15. Crastan, V. Elektrische Energieversorgung 1 Netzelemente, Modellierung, stationäres Verhalten, Bemessung, Schalt- und Schutztechnik, 4th
ed.; Springer Vieweg: Wiesbaden, Germany, 2015.
16. Benoit, C. Models for Investigation of Flexibility Benefits in Unbalanced Low Voltage Smart Grids. Ph.D. Thesis, Université
Grenoble Alpes, Grenoble, France, 2015.
17. Olivier, F.; Fontenau, R.; Ernst, D. Modelling of three-phase four-wire low-voltage cables taking into account the neutral
connection to the earth. In Proceedings of the CIRED Workshop, Ljubljana, Slovenia, 7–8 June 2018.
18. Neusel-Lange, N. Dezentrale Zustandsüberwachung für intelligente Niederspannungsnetze. Ph.D. Thesis, University of
Wuppertal, Wuppertal, Germany, 2013.
19. Beharrysingh, S. Phase Unbalance on Low-Voltage Electricity Networks and Its Mitigation Using Static Balancers. Ph.D. Thesis,
Loughborough University, Leicestershire, Großbritannien, UK, 2014.
20. Kersting, W.H. Distribution System Modeling and Analysis, 4th ed.; CRC Press: Boca Raton, FL, USA, 2017.
21. Strunz, K.; Abbasi, E.; Abbey, C.; Andrieu, C.; Annakkage, U.; Barsali, S. Task Force C6.04.02: Benchmark Systems for Network
Integration of Renewable and Distributed Energy Resources; Elektra, CIGRÉ: Paris, France, 2014.
22. Stegner, J. Bestimmung thermischer Materialkennwerte von Erdkabelbettungen. Ph.D. Thesis, Technical University of Darmstadt,
Darmstadt, Germany, 2016.
23. Schuster, A. Ladeinfrastruktur und Verteilnetzbelastungen durch Elektromobilität. Ph.D. Thesis, Vienna University of Technology,
Vienna, Austria, 2013.
24. Heinhold, L.; Stubbe, R. Kabel und Leitungen für Starkstrom, 5th ed.; Publicis MCD Verlag: Erlangen, Germany, 1999.
25. Urquhart, A.J. Accuracy of Low Voltage Electricity Distribution Network Modelling. Ph.D. Thesis, Loughborough University,
Leicestershire, UK, 2016.
26. Urquhart, A.J.; Thomson, M. Series impedance of distribution bcables with sector-shaped conductors. IET Gener. Transm. Distrib.
2015, 9. [CrossRef]
27. Carson, J.R. Wave Propagation in Overhead Wires with Ground Return. Bell Syst. Tech. J. 1926, 5, 539–554. [CrossRef]
28. Ciric, R.M.; Feltrin, A.P.; Ochoa, L.F. Power Flow in Four-Wire DistributionNetworks—General Approach. IEEE Trans. Power
Syst. 2003, 18, 1283–1290. [CrossRef]
29. Shafieipour, M.; Chenb, Z.; Menshovc, A.; Silva, J.D.; Okhmatovski, V. Efficiently computing the electrical parameters of cables
with arbitrary cross-sections using the method-of-moments. Electr. Power Syst. Res. 2018, 162, 37–49. [CrossRef]
Energies 2021, 14, 1265 34 of 34

30. Bayerische Kabelwerke AG (Bayka). BayEnergy® CPR Eca Starkstromkabel, PVC-Isoliert, PVC-Mantel NAYY-J—4-5-adrig nach
DIN VDE 0276 Teil 603. Available online: http://www.bayerische-kabelwerke.de/Produkte/Starkstromkabel/Aluminiumkabel_
1_kV/?docId=cablespecies_98 (accessed on 5 May 2020).
31. Bayerische Kabelwerke AG (Bayka). BayEnergy® Starkstromkabel, PVC-/VPE-isoliert, konzentrischer Leiter, PVC-Mantel
NAYCWY - 4 1/2-adrig, (EMV) nach DIN VDE 0276 Teil 603. Available online: http://www.bayerische-kabelwerke.de/
Produkte/Starkstromkabel/Aluminiumkabel_1_kV/?docId=cablespecies_841 (accessed on 5 May 2020).
32. Klaus Faber AG. NAYCWY 03X50SE/50 SW. Available online: https://shop.faberkabel.de/Starkstromkabel-1-30-kV/
Niederspannungskabel/Starkstromkabel-NAYCWY/090206.html (accessed on 5 May 2020).
33. DIN EN 60228 VDE 0295:2005-09, Leiter für Kabel und isolierte Leitungen (IEC 60228:2004); VDE: Berlin, Germany, 2005.
34. Grigsby, L.L. Electric Power Generation, Transmission, and Distribution, 3rd ed.; CRC Press: Boca Raton, FL, USA, 2012.
35. Glover, J.D.; Overbye, T.J.; Sarma, M.S. Power System Analysis & Design, 6th ed.; Cengage Learning: Boston, MA, USA 2017.
36. BS 3988:1970 Wrought Aluminium for Electrical Purposes—Solid Conductors For Insulated Cables; British Standards Institution:
London, UK, 1970.
37. DIN VDE 0276-603 VDE 0276-603:2010-03, Starkstromkabel Teil 603: Energieverteilungskabel Mit Nennspannung 0.6/1 kV; VDE: Berlin,
Germany, 2010.
38. Küchler, A. High Voltage Engineering, 5th ed.; Springer Vieweg: Wiesbaden, Germany, 2018.
39. OMICRON Electronics GmbH. Dirana Brochure. Available online: https://www.omicronenergy.com/download/document/D7
35659E-887A-407F-BE93-8E452256A3DC/ (accessed on 20 November 2020).
40. IEC 60364-5-54:2011 Low-Voltage Electrical Installations—Part 5-54: Selection and Erection of Electrical Equipment—Earthing Arrange-
ments and Protective Conductors; IEC: Geneva, Switzerland, 2011.
41. Hering, E. Berechnung der Erdungswiderstände ringförmiger Erder. Elektropraktiker 2000, 54, 12.
42. Fertighaus.de. Wie viel Grundstück braucht man für ein Haus? Available online: https://www.fertighaus.de/ratgeber/
grundstueck/wie-viel-grundstueck-braucht-man-fuer-ein-haus/ (accessed on 27 June 2020).
43. DEHN + SÖHNE GmbH + Co.KG. Blitzplaner. Available online: https://www.heinze.de/pdfdownload/?pdf=/m2/25/62225/
pdf/97/15230897px510x692.pdf (accessed on 1 April 2020).
44. Bundesverband der Energie- und Wasserwirtschaft e. V. (BDEW). TAB 2019—Technische Anschlussbedingungen für den Anschluss an
das Niederspannungsnetz; BDEW: Berlin, Germany, 2019.
45. DIN EN 50160:2011-02, Merkmale der Spannung in öffentlichen Elektrizitätsversorgungsnetzen (EN 50160:2010); EN: Brussels, Bel-
gium, 2011.
46. A-Eberle GmbH & Co. KG. Bedienungsanleitung, Datenblatt Netzanalysator, Transientenrekorder PQ-Box 200 & 300. Avail-
able online: https://www.a-eberle.de/de/downloads/power-quality/datenblaetter/mobile-analysatoren (accessed on 13
December 2020).
47. Bayerische Kabelwerke AG (Bayka). BayEnergy® Starkstromkabel, PVC-/VPE-isoliert, konzentrischer Leiter, PVC-Mantel
NAYCWY—3-adrig nach DIN VDE 0276 Teil 603. Available online: http://www.bayerische-kabelwerke.de/Produkte/
Starkstromkabel/Aluminiumkabel_1_kV/?docId=cablespecies_101 (accessed on 5 May 2020).