Article

Lifetime Assessment of PILC Cables with Regard to Thermal

Aging Based on a Medium Voltage Distribution Network
Benchmark and Representative Load Scenarios in the Course of
the Expansion of Distributed Energy Resources
Martin Zapf 1,*, Tobias Blenk 1, Ann-Catrin Müller 1, Hermann Pengg 2, Ivana Mladenovic 3 and Christian Weindl 1

1 Institute of High Voltage Technology, Energy System & Asset Diagnostics (IHEA), Coburg University of
Applied Sciences and Arts, 96450 Coburg, Germany; [email protected] (T.B.);
[email protected] (A.-C.M.); [email protected] (C.W.)
2 AUDI AG, 85045 Ingolstadt, Germany; [email protected]

3 Siemens AG, 91058 Erlangen, Germany; [email protected]

* Correspondence: [email protected]

Abstract: The decentralized feed-ins from distributed energy resources (DER) represent a significant
change in the manner in which the power grid is used. If this leads to high loads on electrical equip-
ment, its aging can be accelerated. This applies in particular with regard to the thermal aging of
older generations of power cables, namely paper insulated lead covered (PILC) cables. This type of
power cable can still be found frequently in medium voltage (MV) networks. If aging of these cables
is significantly accelerated in the presence of DER, distribution system operators (DSO) could face
Citation: Zapf, M.; Blenk, T.; Müller,
unplanned premature cable failures and a high replacement demand and costs. Therefore, this pa-
A.-C.; Pengg, H.; Mladenovic, I.;
Weindl, C. Lifetime Assessment of
per investigates the thermal aging of PILC cables in a MV distribution network benchmark for dif-
PILC Cables with Regard to Thermal ferent load scenarios, using standardized load profiles and representative expansion scenarios for
Aging Based on a Medium Voltage wind power and photovoltaics plants in particularly affected network areas in Germany. A main
Distribution Network Benchmark objective of this paper is to present a methodology for estimating the thermal degradation of PILC
and Representative Load Scenarios cables. An approach is used to draw simplified conclusions from the loading of cables to their con-
in the Course of the Expansion of ductor or insulation temperature. For this purpose, mainly Joule losses are considered. In addition,
Distributed Energy Resources. thermal time constants are used for the heating and cooling processes. Based on the insulation tem-
Energies 2021, 14, 494. perature, thermal aging is determined using the Arrhenius law or the Montsinger rule. However, it
https://doi.org/10.3390/en14020494
is important to note that there is an urgent need for research on reference data in this area. For this
reason, the results of the lifetime estimation presented in this paper should only be considered as
Received: 27 November 2020
an approximation if the selected reference data from the literature for the aging model are actually
Accepted: 11 January 2021
applicable. The lifetime assessment is performed for a highly utilized line segment of the network
Published: 18 January 2021
benchmark. Accordingly, extreme values are examined. Different operational control strategies of
Publisher’s Note: MDPI stays neu- DSO to limit cable utilization are investigated. The results show that the expansion of DER can lead
tral with regard to jurisdictional to a short but high cable utilization, although the average utilization does not increase or increases
claims in published maps and insti- only slightly. This can lead to significantly lower cable lifetimes. The possible influence of these
tutional affiliations. temporarily high loads is shown by comparing the resulting cable lifetime with previous situations
without DER. It is also shown that DSO could already reduce excessive aging of PILC cables by
preventing overloads in a few hours of a year. In addition to these specific results, general findings
on the network load due to the influence of DER are obtained, which are of interest for congestion
Copyright: © 2021 by the authors. management.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
Keywords: distribution network benchmark; load scenarios; thermal network method; thermal time
distributed under the terms and con-
constant; Arrhenius law; Montsinger rule; line and network load; thermal aging
ditions of the Creative Commons At-
tribution (CC BY) license (http://cre-
ativecommons.org/licenses/by/4.0/).

Energies 2021, 14, 494. https://doi.org/10.3390/en14020494 www.mdpi.com/journal/energies

Energies 2021, 14, 494 2 of 26

1. Introduction
In order to limit climate change, a transition of the energy system to generation tech-
nologies without direct emissions of greenhouse gases is a basic prerequisite. Key tech-
nologies in this respect are wind power and photovoltaic (PV) plants. These distributed
energy resources (DER) are characterized by fluctuating power generation due to their
dependence on the natural availability of solar radiation and wind. Furthermore, these
plants are primarily connected to the distribution grid due to their power size. The decen-
tralized feed-ins of DER represent a significant change in the manner in which the power
grid is used. In the past, the usage of the grid was structured top-down, with feed-ins at
high voltage levels and loads at lower voltage levels. In sum, the expansion of DER leads
to a change in the utilization of the power grid—especially of the distribution network.
Distribution system operators (DSO) have to adapt the correlating new challenges. One
challenge is that a changing load behavior of distribution networks in turn influences the
aging of electrical equipment. For example, the aging of cables is dependent on their load.
Medium voltage (MV) networks are particularly affected, as a relatively high proportion
of rather outdated cables are installed. Further aging has a significant impact on the num-
ber of failures in the near future. As grid renewal is limited for economic reasons and due
to capacity constraints of DSO, operational control measures to limit the utilization of
electrical equipment should be considered, taking into account the remaining service life.
In the past, estimates of the remaining lifetime of electrical equipment were based on rel-
atively even, low, and predictable utilization rates. PV and wind power plants have highly
volatile feed-in profiles. The expansion of these technologies in distribution grids can lead
to short but high peak loads for electrical equipment as well as a higher overall utilization,
which influences the aging behavior. The lifetime assessment of grid assets is complex, as
DSO on the one hand often lack knowledge about the utilization and the remaining life-
time of individual electrical equipment. On the other hand, suitable models for estimating
the aging with regard to variable stresses are necessary [1].
This paper describes the thermal aging of paper insulated lead covered (PILC) cables.
This type of power cable can be found frequently in MV networks, even if this type was
mainly replaced by cross-linked polyethylene (XLPE) cables for new installations since
the 1970s. In 1997, PILC cables represented around 50% of the cable line length of the MV
grid in Germany. In 2019, the majority of these cables were older than 40 years [2]. In 2005,
screened and belted PILC (36% and 22%) and XLPE (33%) cables were the most commonly
used MV cable types worldwide [3].
This paper starts with a general description of stresses on cables (see Sections 2). In
this context, thermal stress is the main focus of interest. The Arrhenius law or Montsinger
rule are used to describe the degeneration based on thermal stress (see Section 3). To in-
vestigate the influence of DER to the grid, a simulation of a network benchmark is exe-
cuted. Various load scenarios are considered for the network benchmark, using standard-
ized load profiles and representative expansion scenarios for wind power and PV plants
in particularly affected MV network areas in Germany (see Sections 4 and 5). The aging
behavior for the different load scenarios is not presented for the entire network bench-
mark. Instead, a lifetime assessment for a highly utilized line segment is shown. Accord-
ingly, extreme values are examined. The main objective of this paper is to provide a meth-
odology for estimating the thermal degradation of PILC cables and to show the potential
influence of temporarily high loads. To this end, thermal aging of cables in the course of
the integration of DER into the grid is compared with previous situations without DER
(see Section 6).
Energies 2021, 14, 494 3 of 26

2. Thermal Stress
Electrical equipment is exposed to electrical, thermal and mechanical stresses. This
can be caused by external influences, such as solar radiation, external heat sources, light-
ning discharges, neighbouring conductors, wind, and ice loads. Stresses are also induced
by internal influences such as electric field intensity, Joule losses, dielectric and magnetic
losses, magnetic forces, thermal expansion forces and vibrations. In addition, atmospheric
stresses can also occur due to chemical influences and pollution. Depending on the dura-
tion of exposure, a distinction can be made between long-term and short-term stresses.
Continuous as well as short-term loads place thermal and mechanical stress on electrical
equipment, which has an influence on their insulation properties and aging [4]. For the
oil-impregnated insulation of PILC cables thermal degradation is the main aging mecha-
nism, therefore, only this is considered in the following [1].

2.1. Heat Sources

Internal heat sources for cables are Joule losses as well as hysteresis, eddy-current
and dielectric losses. Additional thermal stress is caused by external heat sources such as
other cables or supply lines in the vicinity [4,5]. Joule losses 𝑃𝑃𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 are load-dependent,
caused by resistances 𝑅𝑅 and the apparent currents 𝐼𝐼𝑎𝑎 :
2
𝑃𝑃𝐿𝐿𝐿𝐿𝐿𝐿𝐿𝐿 = �𝐼𝐼𝑎𝑎 � ∙ 𝑅𝑅𝐷𝐷𝐷𝐷 ∙ 𝑘𝑘𝑐𝑐
with: (1)

𝑅𝑅 = 𝑅𝑅𝜗𝜗𝑅𝑅 ∙ [1 + 𝛼𝛼 ∙ (𝜗𝜗𝑛𝑛 − 𝜗𝜗𝑅𝑅 )]

where 𝑅𝑅𝜗𝜗𝑅𝑅 is the resistance for a reference temperature—in most cases 20 °C (𝑅𝑅20 ), since
this corresponds to the average ambient temperature in regions under consideration [4]—
and 𝛼𝛼 is the material specific temperature coefficient, which indicates a linear resistance
change as a function of temperature (𝛼𝛼𝐶𝐶𝐶𝐶 = 3.85 × 10−3 ). There may be a significant in-
crease in resistance due to the skin and proximity effect. Because of these current displace-
ment effects, the AC resistance of a conductor is greater than its DC resistance 𝑅𝑅𝐷𝐷𝐷𝐷 by a
current displacement factor 𝑘𝑘𝑐𝑐 . This increase should to be taken into account for conduc-
tor cross-sections larger than 185 mm² [6]. Eddy-current and hysteresis losses of cables can
be considered as a function of the Joule losses via loss factors [6,7]. Dielectric losses are
negligible for low voltages, e.g., this is recommended for voltages below 44 kV in the Na-
tional Electrical Code (USA) [8]. Following the previous explanations, the internal heat
sources of MV cables can be limited to Joule losses and losses converted to them.

2.2. Permissible Amapcity

The maximum permissible current of cables is specified in standards [9,10]. Manu-
facturers also provide information and recommendations for current-carrying capacities
under different conditions [11]. The basis for the first international standard IEC 60287
was the formula system developed by J.H. Neher and M.H. McGrath in 1957 [8]. This
standard only covers the steady-state at continuous load. Transient and cyclic loads are to
be calculated according to IEC 60853 [12]. In Germany, also DIN VDE 0276 for power
distribution cables has to be considered. The standards allow a comparable calculation of
the load capacity of cables systems, verified by measurements. The methodology in the
standards is built on analytical approximation formulas, which are often conservative.
These are based on a thermal equivalent circuit diagram with heat sources and thermal
resistors (cf. [13,14]). Today, numerical methods can be used for more accurate calcula-
tions, e.g., the finite element method (FEM, see IEC TR 62095). The normative specifica-
tions in DIN VDE 0276 are the result of calculations, tests and many years of experience.
The rated values are based on boundary conditions that are usually found in Central Eu-
rope, such as ambient conditions of 30 °C air temperature and 20 °C ground temperature.
In addition, a daily load factor of 0.7 is generally assumed for underground cables. This
Energies 2021, 14, 494 4 of 26

relies on a load profile that is common for DSO. For overhead lines and cables installed in
the air, a continuous load, i.e., a load factor of 1.0, is taken into account, since their thermal
time constants are relatively short compared to underground cables [15].
To consider the operating and ambient conditions as well as the type of installation,
a permissible current 𝐼𝐼𝑝𝑝 , which deviates from the rated current 𝐼𝐼𝑟𝑟 , must be determined
using conversion factors ∏ 𝑓𝑓 [10,15,16]:

𝐼𝐼𝑝𝑝 = 𝐼𝐼𝑟𝑟 ∙ � 𝑓𝑓 (2)

If the maximum allowable temperatures are maintained, cables can also be operated
in cyclical operation, e.g., at a daily load factor of 0.7, above their rated current. Depending
on the preload and the overload duration, a potential overload can be determined without
exceeding certain temperatures [11,15,16]. In practice, DSO often assume that short-term
and infrequent cable overloads of up to 20% are acceptable [11,17].
In Section 6, a lifetime assessment is presented for the cable type NEKEBA 3 × 120
mm² 12/20 KV (Umax = 24 kV). Corresponding data are given in Table 1.

Table 1. Data for cable type NEKEBA 3 × 120 mm² 12/20 kV (Umax = 24 kV) [18].

Rated current per conductor 𝐼𝐼𝑟𝑟 302 A

Direct current resistance at 65 °C per conductor 𝑅𝑅𝐷𝐷𝐷𝐷 0.18 Ω/km
Active resistance at 65 °C per conductor 𝑅𝑅𝐴𝐴𝐴𝐴 0.184 Ω/km
Joule losses per cable based on active resistance 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 50.3 kW/km

Factors for the conversion of the rated current into the permissible current are shown
in Table 2. The rated values are based on a load factor of 0.7, a ground temperature of 20
°C and a specific ground resistance of 1 km/W. The conversion factors from the table show
that for soils with a resistivity of 2.5 km/W instead of 1 km/W the permissible current is
lower by a factor between 0.82 and 0.85 to maintain the maximum allowable temperature.
Furthermore, for every 5 °C lower soil temperature, the acceptable current is higher by a
factor between 1.03 and 1.054.

Table 2. Conversion factors from [18] and typical seasonal ground temperatures in Central Europe
from [12,15,19].

Maximum Allowable Temperature of 65 °C, Installation in Ground and Load Factor of 1.0
𝒇𝒇𝟏𝟏 ∙ 𝒇𝒇𝟐𝟐 = ∏𝒇𝒇
Ground Temperature Specific Ground Resistance
20 °C 1 km/W 0.90 × 0.91 = 0.82
(Typical for July, August, September) 2.5 km/W 0.74 × 0.91 = 0.67
15 °C 1 km/W 0.94 × 0.91 = 0.86
(Typical for May, June, October) 2.5 km/W 0.78 × 0.91 = 0.71
10 °C 1 km/W 0.97 × 0.91 = 0.88
(Typical for March, April, November) 2.5 km/W 0.82 × 0.91 = 0.75
5 °C 1 km/W 1.00 × 0.91 = 0.91
(Typical for December, January, February) 2.5 km/W 0.85 × 0.91 = 0.77

2.3. Thermal Analysis

The heating of a medium is determined by the amount of internal and external heat
sources as well as by heat storage and release. The energy balance of supplied, stored and
transferred heat according to first law of thermodynamics is the basis for thermal analysis.
Heat transfer is caused due to temperature differences. The direction of heat transfer is
always from a location with higher temperature to a location with lower temperature. The
heat transfer can take place by heat conduction, convection or radiation [5]. A possibility
of thermal analysis is the so-called thermal network method. This method is based on the
analogy between variables and constants of the thermal and electrical field, summarized
Energies 2021, 14, 494 5 of 26

in Table 3. Accordingly, the thermal processes can be described with a thermal network
analogous to an electric network [20].

Table 3. Analogy between variables and constants of the thermal and electrical field [4,5].
Electrical Field Thermal Field
Parameter Formula Symbol Unit Parameter Formula Symbol Unit
Potential difference U (V) Temperature difference ∆𝜗𝜗 (K)
Electric current I (A) Heat flow Pth (W)
Electrical resistance R (Ω) Thermal resistance Rth (K/W)
Electrical capacity C (As/V) Thermal capacity Cth (Ws/K)
Electrical energy W (Ws) Heat Qth (Ws)

Thermal networks consist of sources and sinks as well as thermal resistances and
capacities. If 𝐶𝐶𝑡𝑡ℎ is related to the mass of medium 𝑚𝑚, the corresponding specific thermal
capacity 𝑐𝑐 is obtained:
Δ𝑄𝑄
𝐶𝐶𝑡𝑡ℎ = = 𝑐𝑐 ∙ 𝑚𝑚 (3)
Δ𝜗𝜗
The thermal capacity indicates how much heat has to be supplied to or removed from
a medium to cause a certain temperature change. The thermal resistance of the surround-
ing environment represents the greatest uncertainty for the thermal analysis of under-
ground power cables. This is due to the fact that conductivity of the ground is highly
dependent on parameters such as humidity and temperature, which are subject to local
and temporal fluctuations. The thermal resistance of soil can vary between 0.4 and 4
km/W, with typical values between 0.8 and 1.3 km/W [21].
In addition, drying-out of the soil at higher temperatures sets in (cf. [13,14,22,23]).
According to IEC 60287 (2005), in stationary observations, drying-out takes place at a crit-
ical excess temperature of 15 °C. Above this temperature, a specific soil resistance of 2.5
km/W and below 1.0 km/W is assumed [7].
The energy balance of a cable consists of heat sources, such as Joule losses 𝑄𝑄𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 , the
stored heat 𝑄𝑄𝑠𝑠𝑠𝑠 and the transferred heat 𝑄𝑄𝑡𝑡 during a differential time period 𝑑𝑑𝑑𝑑:
∆𝜗𝜗 ∙ 𝑑𝑑𝑑𝑑
𝑄𝑄𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 = 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 ∙ 𝑑𝑑𝑑𝑑 = 𝑄𝑄𝑠𝑠𝑠𝑠 + 𝑄𝑄𝑡𝑡 = 𝐶𝐶𝑡𝑡ℎ ∙ 𝑑𝑑𝑑𝑑 +
𝑅𝑅𝑡𝑡ℎ
(4)
with:
∆𝜗𝜗 = 𝜗𝜗𝑚𝑚 − 𝜗𝜗𝑠𝑠
where ∆𝜗𝜗 is the difference between the medium temperature 𝜗𝜗𝑚𝑚 and the surrounding
temperature 𝜗𝜗𝑠𝑠 . The first-order differential equation with constant coefficients is as fol-
lows [24,25]:
𝑑𝑑𝑑𝑑 ∆𝜗𝜗
𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 = 𝐶𝐶𝑡𝑡ℎ ∙ +
𝑑𝑑𝑑𝑑 𝑅𝑅𝑡𝑡ℎ
multiplied by 𝑅𝑅𝑡𝑡ℎ (5)
𝑑𝑑𝑑𝑑
𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 ∙ 𝑅𝑅𝑡𝑡ℎ = 𝐶𝐶𝑡𝑡ℎ ∙ 𝑅𝑅𝑡𝑡ℎ ∙ + ∆𝜗𝜗
𝑑𝑑𝑑𝑑
Simplified with constant surrounding temperature 𝜗𝜗𝑆𝑆 = 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐.:
Energies 2021, 14, 494 6 of 26

𝑑𝑑[∆𝜗𝜗(𝑡𝑡)]
𝑅𝑅𝑡𝑡ℎ ∙ 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 (𝑡𝑡) = 𝜏𝜏 ∙ + ∆𝜗𝜗(𝑡𝑡)
𝑑𝑑𝑑𝑑
with:
𝜏𝜏 = 𝑅𝑅𝑡𝑡ℎ ∙ 𝐶𝐶𝑡𝑡ℎ (6)
converted:
𝑑𝑑∆𝜗𝜗(𝑡𝑡)
∆𝜗𝜗(𝑡𝑡) = 𝑅𝑅𝑡𝑡ℎ ∙ 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 (𝑡𝑡) − 𝜏𝜏 ∙
𝑑𝑑𝑑𝑑
where 𝜏𝜏 is the thermal time constant. For the time-dependent temperature 𝜗𝜗(𝑡𝑡) of the
medium results (index 𝑚𝑚 is omitted for clarity):
−𝑡𝑡
𝜗𝜗(𝑡𝑡) = 𝜗𝜗0 + ∆𝜗𝜗(𝑡𝑡) = 𝜗𝜗0 + (𝜗𝜗∞ − 𝜗𝜗0 ) ∙ �1 − 𝑒𝑒 � 𝜏𝜏 � � (7)

where 𝜗𝜗∞ is the steady-state temperature and 𝜗𝜗0 is the initial temperature. The temper-
ature rises until there is a balance between the energy supplied and the energy removed
by heat transfer. At this point the stationary or steady-state temperature is reached [26].
Conversely, in the case of cooling, the temperature drops until this balance is attained. For
a given constant heat loss, the steady-state temperature difference ∆𝜗𝜗∞ is calculated as
follows:
∆𝜗𝜗∞ = (𝜗𝜗∞ − 𝜗𝜗0 ) = 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 ∙ 𝑅𝑅𝑡𝑡ℎ (8)
Exemplary temperature curves during heating and cooling are shown in Figure 1.

Figure 1. Temperature curve during heating (left) and cooling (right).

As previously stated, the thermal time constant 𝜏𝜏 results from the product of the
thermal capacity and resistance:
𝜏𝜏 = 𝑅𝑅𝑡𝑡ℎ ∙ 𝐶𝐶𝑡𝑡ℎ (9)
The thermal time constant determines the course of the medium temperature over
time and thus reflects how quickly steady-state is reached after a change in load. The final
temperature is attained by approx. 63% after a period of ∆𝑡𝑡 = 𝜏𝜏 and by approx. 95% after
∆𝑡𝑡 = 3 ∙ 𝜏𝜏. If these thermal dynamics are not taken into account, the temperature of the
medium will be systematically overestimated when it heats up and underestimated when
it cools down. These systematic errors are particularly noticeable in the case of fluctuating
loads and high thermal time constants [27]. In [28] it is stated: “[…] omitting thermal tran-
sients for the examined cables and the relevant daily load cycle leads to a large underes-
timation of service life […]”.
An exact analysis with the thermal network method requires the subdivision into
layers (cf. [6]) with different material properties (𝐶𝐶𝑡𝑡ℎ . and 𝑅𝑅𝑡𝑡ℎ ) and heat sources, which
represent the nodes of the thermal network [5]. Without any other heat sources except a
constant Joule loss 𝑃𝑃𝑙𝑙𝑜𝑜𝑜𝑜𝑜𝑜 , the time dependent temperature difference ∆𝜗𝜗(𝑡𝑡) between a ca-
ble conductor and a surrounding reference point can be calculated as follows:
Energies 2021, 14, 494 7 of 26

∆𝜗𝜗(𝑡𝑡) = 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 ∙ 𝑍𝑍𝑡𝑡ℎ (𝑡𝑡)

with:
𝑗𝑗
𝜗𝜗(𝑡𝑡) −𝑡𝑡
� � (10)
𝑍𝑍𝑡𝑡ℎ (𝑡𝑡) = = � 𝑅𝑅𝑡𝑡ℎ,𝑖𝑖 ∙ �1 − 𝑒𝑒 𝜏𝜏𝑖𝑖 �
𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙
𝑖𝑖=1

with:
𝜏𝜏𝑖𝑖 = 𝑅𝑅𝑡𝑡ℎ,𝑖𝑖 ∙ 𝐶𝐶𝑡𝑡ℎ,𝑖𝑖
where 𝜏𝜏𝑖𝑖 is the thermal time constant of the individual layers 𝑖𝑖 and 𝑍𝑍𝑡𝑡ℎ,𝑐𝑐𝑐𝑐 (𝑡𝑡) is the time
dependent thermal impedance between the conductor 𝑐𝑐 and the surrounding reference
𝑗𝑗
point 𝑠𝑠. 𝑍𝑍𝑡𝑡ℎ,𝑐𝑐𝑐𝑐 (𝑡𝑡) takes the value of the steady-state thermal resistance ∑𝑖𝑖 𝑅𝑅𝑡𝑡ℎ,𝑖𝑖 after the
transient processes are subsided.
The thermal network can be displayed with the Foster model that consists of a series
connection of RC elements. Another possibility is the physical equivalent circuit diagram,
called Cauer model (see Figure 2). A conversion of the models can be done by partial
fraction decomposition or polynomial division [26,29]. In the Foster model, R and C values
cannot be determined from the real components, unlike in the Cauer model. The determi-
nation must to be done by measured or simulated temperature courses. Thus, the physical
reference is lost. The Foster model represents a behavioral model only [26]. Both models
will not be described in detail at this point, they only serve to illustrate the formulas, es-
pecially the derivation of 𝑍𝑍𝑡𝑡ℎ (𝑡𝑡).

Foster Thermal Network Model Cauer Thermal Network Model

𝐶𝐶𝑡𝑡ℎ,2 𝐶𝐶𝑡𝑡ℎ,2 𝐶𝐶𝑡𝑡ℎ,3 𝑍𝑍𝑡𝑡ℎ,𝑐𝑐𝐿𝐿 (𝑑𝑑)

𝜗𝜗𝑐𝑐 𝑑𝑑 𝜗𝜗𝑐𝑐 𝑑𝑑

𝑅𝑅𝑡𝑡ℎ,1 𝑅𝑅𝑡𝑡ℎ,2 𝑅𝑅𝑡𝑡ℎ,3

𝑃𝑃𝑙𝑙𝐿𝐿𝐿𝐿𝐿𝐿 (𝑑𝑑) 𝑅𝑅𝑡𝑡ℎ,1 𝑅𝑅𝑡𝑡ℎ,2 𝑅𝑅𝑡𝑡ℎ,3 𝑃𝑃𝑙𝑙𝐿𝐿𝐿𝐿𝐿𝐿 (𝑑𝑑)
𝑍𝑍𝑡𝑡ℎ,𝑐𝑐𝐿𝐿 (𝑑𝑑) 𝜗𝜗𝐿𝐿 𝐶𝐶𝑡𝑡ℎ,1 𝐶𝐶𝑡𝑡ℎ,2 𝐶𝐶𝑡𝑡ℎ,3 𝜗𝜗𝐿𝐿
∆𝜗𝜗(𝑑𝑑) ∆𝜗𝜗(𝑑𝑑)

Figure 2. Thermal network models, following [26].

For the evaluation of the thermal aging of PILC cables the temperature of the insula-
tion is the basis. The conductor temperature represents the maximal occurring tempera-
ture of the surrounding insulation. The steady-state temperature of the conductor results
from the total thermal resistance of a cable conductor to a surrounding reference point
𝑅𝑅𝑡𝑡ℎ , e.g., the soil for underground power cables. This value can be computed with the
rated steady state temperature difference ∆𝜗𝜗∞,𝑟𝑟 and the adapted Joule losses 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,𝑎𝑎 :
𝑗𝑗
∆𝜗𝜗𝑐𝑐,∞,𝑟𝑟
𝑅𝑅𝑡𝑡ℎ = � 𝑅𝑅𝑡𝑡ℎ,𝑖𝑖 =
𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,𝑎𝑎
𝑖𝑖=1

with:
(11)
𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 ∙ (∏𝑓𝑓)2
𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,𝑎𝑎 =
𝑛𝑛𝑐𝑐
𝜗𝜗𝑚𝑚𝑚𝑚𝑚𝑚 = 𝜗𝜗𝑐𝑐,∞,𝑟𝑟 = ∆𝜗𝜗𝑐𝑐,∞,𝑟𝑟 + 𝜗𝜗𝑠𝑠
converted:
Energies 2021, 14, 494 8 of 26

∆𝜗𝜗𝑐𝑐,∞,𝑟𝑟 = 𝜗𝜗𝑐𝑐,∞,𝑟𝑟 − 𝜗𝜗𝑠𝑠 = 𝑅𝑅𝑡𝑡ℎ ∙ 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,𝑎𝑎

where 𝐼𝐼𝑟𝑟 is the rated current, 𝑅𝑅𝑡𝑡ℎ is the thermal resistance, 𝜗𝜗𝑚𝑚𝑚𝑚𝑚𝑚 is the maximum allow-
able temperature (65 °C for PILC cables), 𝑛𝑛𝑐𝑐 is the number of conductors per cable, 𝜗𝜗𝑐𝑐,∞,𝑟𝑟
is the rated steady state temperature of the conductor and 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 are Joule losses per
cable. Taking into account the losses from Table 1, at a surrounding temperature of 20 °C
and a correction factor of 0.82 to consider a continuous load according to Table 2, the ther-
mal resistance of the corresponding PILC cable type is 3.99 K/W. This value is based on a
thermal resistance of the surrounding soil of 1 km/W.
The minimal existing thermal time constant according to Formula (9) can be assessed
in a simplified way with the thermal capacity 𝐶𝐶𝑡𝑡ℎ of the conductor. In this case, the heat-
ing of the individual cable layers is not taken into account [30]. This results in an under-
estimation of the thermal time constant. The heat capacity of the PILC cable can be calcu-
lated as follows:
1.2 cm2 ∙ 100 cm ∙ 8.96 g/cm3 ∙ kg
𝐶𝐶𝑡𝑡ℎ = 𝑚𝑚𝑐𝑐 ∙ c = 𝐴𝐴 ∙ 𝑙𝑙 ∙ 𝜚𝜚 ∙ 𝑐𝑐 = ∙ 385 Ws/kg ∙ K (12)
1000 g
where 𝑚𝑚𝑐𝑐 is the mass of the conductor resulting from the cross-sectional area 𝐴𝐴, the cable
length 𝑙𝑙 and the density of the conductor material 𝜚𝜚. For the cable type NEKEBA 3 × 120
mm² 12/20 kV (Umax = 24 kV) the mass of the conductor is 1.075 kg, based on 8.96 g/cm³ for
the density of copper and the cross section of 1.2 cm² per conductor. With the specific heat
capacity for copper of 385 Ws/kg∙K, the conductor has a thermal capacity of 413.95 Ws/km.
This results in a thermal time constant of 1651.66 s or approx. 30 min.
Typical thermal time constants of essential electrical equipment are [27]:
 <15 min for overhead lines,
 <120 min for cables and
 <150 min for transformers.
This dimension of time constant for cables are approximately confirmed by exem-
plary simulations found in literature, e.g., in [21] for a 400 kV cable with 630 mm² copper
conductor and in [12] for a 30 kV XLPE cable with 240 mm² aluminum conductor. Further
research is necessary to assess typical thermal time constants for cables under certain am-
bient conditions.
The steady-state temperature of the cable conductor 𝜗𝜗𝑐𝑐,∞,𝑛𝑛 at a time step 𝑛𝑛 with a
load deviating from the permissible current 𝐼𝐼𝑝𝑝 can be calculated as follows:
𝜗𝜗𝑐𝑐,∞,𝑛𝑛 = ∆𝜗𝜗𝑐𝑐,∞,𝑛𝑛 + 𝜗𝜗𝑠𝑠 = 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,𝑎𝑎,𝑛𝑛 ∙ 𝑅𝑅𝑡𝑡ℎ + 𝜗𝜗𝑠𝑠
with:
𝐼𝐼𝑛𝑛 2 ∙ 𝑅𝑅(𝜗𝜗𝑛𝑛−1 )
𝜗𝜗𝑐𝑐,∞,𝑛𝑛 = ∙ ∆𝜗𝜗∞,𝑟𝑟 + 𝜗𝜗𝑠𝑠
(𝐼𝐼𝑟𝑟 ∙ ∏𝑓𝑓)2 ∙ 𝑅𝑅(𝜗𝜗𝑚𝑚𝑚𝑚𝑚𝑚 )
(13)
with 𝐼𝐼𝑝𝑝 = 𝐼𝐼𝑟𝑟 ∙ ∏𝑓𝑓 and 𝑅𝑅(𝜗𝜗) according to Formula (1)
with
𝐼𝐼𝑛𝑛 2 ∙ [1 + 𝛼𝛼 ∙ (𝑇𝑇𝑛𝑛−1 − 293 K)]
𝜗𝜗𝑐𝑐,∞,𝑛𝑛 = ∙ ∆𝜗𝜗∞,𝑟𝑟 + 𝜗𝜗𝑠𝑠
𝐼𝐼𝑝𝑝 2 ∙ [1 + 𝛼𝛼 ∙ (𝑇𝑇𝑚𝑚𝑚𝑚𝑚𝑚 − 293 K)]
where 𝐼𝐼𝑛𝑛 is the continuous conductor current during the time step 𝑛𝑛, 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,𝑎𝑎,𝑛𝑛 are the
adapted Joule losses and 𝑅𝑅 is the electrical resistance. For simplification, in the calcula-
tion the temperature-dependent resistance is determined with the temperature of the pre-
vious time step 𝑛𝑛 − 1.
The conductor temperature considering the thermal time constant explained in Sec-
tion 2.2 can be calculated as follows for each time step that is assumed to be 300 s (5 min)
[4]:
Energies 2021, 14, 494 9 of 26

−300s
𝜗𝜗𝑐𝑐,𝑛𝑛 = 𝜗𝜗𝑐𝑐,𝑛𝑛−1 + (𝜗𝜗𝑐𝑐,∞,𝑛𝑛 − 𝜗𝜗𝑐𝑐,𝑛𝑛−1 ) ∙ �1 − exp � �� (14)
𝜏𝜏
An increase in the steady-state temperature of the cable conductor 𝜗𝜗𝑐𝑐,∞ due to
changes of the specific ground resistance caused by soil dehydration can be determined
using correction factors ∏𝑓𝑓𝑔𝑔𝑔𝑔 from Table 2 and the following equation:
(∏𝑓𝑓)2𝑔𝑔𝑔𝑔1 (∏𝑓𝑓)2 𝑔𝑔𝑔𝑔1
𝜗𝜗𝑐𝑐,∞ = 𝑃𝑃𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙 ∙ 𝑅𝑅𝑡𝑡ℎ ∙ + 𝜗𝜗𝑠𝑠 = ∆𝜗𝜗𝑐𝑐,∞ ∙ + 𝜗𝜗𝑠𝑠
(∏𝑓𝑓)2𝑔𝑔𝑔𝑔2 (∏𝑓𝑓)2 𝑔𝑔𝑔𝑔2

At specific ground resistance (gr) of 1.0 km/W instead of 2.5 km/W: (15)

0.822
𝜗𝜗𝑐𝑐,∞ = 45 K ∙ + 20 °C = 87 °C
0.672
Furthermore, the drying-out of the soil influences the thermal time constant. On the
one hand, the thermal resistance increases. On the other hand, the heat capacity decreases
(cf. [31] regarding the specific heat capacity and density of moist and dry soil). An increase
or decrease of the time constant depends on the magnitude of the change of these two
parameters. Basically, at constant power losses, the steady-state temperature rises with
increasing thermal resistance due to drying-out of the soil.
In addition, depending on seasonal deviations of the soil temperature 𝜗𝜗𝑠𝑠 different
permissible currents 𝐼𝐼𝑝𝑝,𝜗𝜗𝑠𝑠 can be calculated, with which the rated steady state tempera-
ture or maximum allowable temperature is not exceeded. The permissible current can be
varied depending on the season with appropriate correction factors (∏𝑓𝑓)𝜗𝜗 , e.g., according
to Table 2, as follows:
(∏𝑓𝑓)𝜗𝜗𝑟𝑟
𝐼𝐼𝑝𝑝,𝜗𝜗𝑠𝑠 = 𝐼𝐼𝑟𝑟 ∙ (16)
(∏𝑓𝑓)𝜗𝜗𝑠𝑠
where 𝜗𝜗𝑟𝑟 is the temperature of the soil according to the rated current and 𝜗𝜗𝑠𝑠 is the tem-
perature of the soil under consideration.

3. Lifetime Assessment of PILC Cables with Regard to Thermal Aging—Methodology

For paper insulated cables, typical age for replacement is about 50 years but varies
between 15 and 80 years amongst different DSO. The replacement is mostly based on a
technical, economical or strategic criteria (cf. [32–34] regarding different maintenance
strategies). The often quoted generally accepted lifetime of cables of 40 years is derived
from present experience of operated cables. Power cable manufacturers do not specify any
expected period of cable lifetime [3].
Aging can be described as a gradual deterioration of the physical and electrical
strength of the equipment until failure occurs and the component loses functionality,
which marks the end of the equipment’s life [35]. This represents the last part of the so-
called bathtub phenomenon. The bathtub curve reflects the development of failure rates
with few failures at the beginning of the life cycle, often due to manufacturing defects or
bad assembly, followed by a long period during which failures are rare and an increasing
number of failures at the end of the lifetime [3].
For cables, a large proportion of defaults are attributable to cable joints. Relating to
statistical data of a large DSO in The Netherlands, 45% of failures observed for MV cable
systems can be traced back to them [36]. Failures in this regard are not investigated here.
As described in Section 2, thermal degradation is the main aging mechanism for PILC
cables. Further degradation, e.g., due to electrical aging, can be determined by multi-stress
models (cf. [37–40]). However, this is not examined further at this point. Instead, a meth-
odology for a lifetime assessment is described exclusively with regard to thermal aging.
When determining cable temperatures for thermal aging also increased dielectric losses
due to thermal degradation of the insulation (cf. [36,38,41]) are neglected.
Energies 2021, 14, 494 10 of 26

The correlation between lifetime and temperature of the considered medium is de-
scribed by the thermal aging reaction rate 𝑘𝑘, given by the Arrhenius law [42]:
𝐵𝐵
𝑘𝑘 = 𝐴𝐴 ∙ 𝑒𝑒 �𝑇𝑇 �
with: (17)
−𝐸𝐸𝑎𝑎
𝐵𝐵 =
𝑅𝑅
where 𝑅𝑅 is the universal gas constant and 𝑇𝑇 is the absolute temperature. 𝐴𝐴 and 𝐵𝐵 are
component specific constants that are particularly derived from empirical studies [43].
Experimental evidence indicates for transformers with cellulose insulation that 𝐴𝐴 =
9.8 × 10−18 and 𝐵𝐵 = 15,000. With Arrhenius reaction rate theory, per unit life and per-
centage loss of life can be calculated as follows [44]:
𝐵𝐵
Per Unit Life = 𝐴𝐴 ∙ 𝑒𝑒 �𝑇𝑇 � = 𝐿𝐿(𝑇𝑇)
𝐹𝐹𝐸𝐸𝐸𝐸𝐸𝐸 ∙ ∆𝑡𝑡 ∙ 100 ∆𝑡𝑡 ∙ 100
Percentage Loss of Life = 𝐷𝐷𝑡𝑡 (𝑇𝑇) = =
𝐿𝐿(𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 ) 𝐿𝐿(𝑇𝑇)
with:
𝐿𝐿(𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 ) (18)
𝐿𝐿(𝑇𝑇) =
𝐹𝐹𝐸𝐸𝐸𝐸𝐸𝐸
∑𝑚𝑚
𝑛𝑛=1 𝐹𝐹𝐴𝐴𝐴𝐴,𝑛𝑛 ∙ ∆𝑡𝑡𝑛𝑛
𝐹𝐹𝐸𝐸𝐸𝐸𝐸𝐸 =
∑𝑚𝑚
𝑛𝑛=1 ∆𝑡𝑡𝑛𝑛

𝐵𝐵 𝐵𝐵
𝐿𝐿(𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 ) � − �
𝐹𝐹𝐴𝐴𝐴𝐴 = = 𝑒𝑒 𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 𝑇𝑇
𝐿𝐿(𝑇𝑇)
where 𝐿𝐿(𝑇𝑇) is the lifetime at the occurred absolute temperature 𝑇𝑇 and 𝐷𝐷𝑡𝑡 �𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 � is the
percentage loss of life or degradation rate throughout the total time period ∆𝑡𝑡 with re-
gard to the reference lifetime 𝐿𝐿(𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 ) at the reference absolute temperature 𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 . 𝐹𝐹𝐴𝐴𝐴𝐴 is
the aging acceleration factor—the rate at which insulation aging is accelerated compared
with the aging rate at reference temperature. 𝐹𝐹𝐸𝐸𝐸𝐸𝐸𝐸 is the equivalent aging factor for the
total number of time intervals 𝑚𝑚 and 𝐹𝐹𝐴𝐴𝐴𝐴,𝑛𝑛 is the aging acceleration factor for the tem-
peratures that exists during the time interval ∆𝑡𝑡𝑛𝑛 .
The aforementioned formulas according to the Arrhenius law are also applicable for
the aging behavior of power cables. The constants 𝐴𝐴 and 𝐵𝐵 could be estimated from his-
toric load data, but these data are usually not available [43]. A novel way to determine
these Arrhenius parameters is presented in [19,43], using a simulation of historic cable
utilization. The study is carried out for cables with a reference temperature of 90 °C and a
corresponding lifetime of 40 years and is therefore not suitable for PILC cables.
The Montsinger rule as an adaption of the Arrhenius law for transformers can also
be applied for PILC cables to determine the aging behavior [38,45–47]:
𝜗𝜗𝑐𝑐 −𝜗𝜗𝑟𝑟𝑟𝑟𝑟𝑟
𝐷𝐷𝑦𝑦 (𝜗𝜗𝑐𝑐 ) �
𝜗𝜗𝑑𝑑
�
𝐹𝐹𝐴𝐴𝐴𝐴 = =2
𝐷𝐷𝑦𝑦 (𝜗𝜗𝑟𝑟𝑟𝑟𝑟𝑟 )
𝜗𝜗𝑐𝑐 −𝜗𝜗𝑟𝑟𝑟𝑟𝑟𝑟
� �
𝜗𝜗𝑑𝑑
𝐷𝐷𝑦𝑦 (𝜗𝜗𝑐𝑐 ) = 𝐷𝐷𝑦𝑦 (𝜗𝜗𝑟𝑟𝑟𝑟𝑟𝑟 ) ∙ 2 (19)
𝜗𝜗𝑐𝑐,𝑛𝑛 −𝜗𝜗𝑟𝑟𝑟𝑟𝑟𝑟
� �
𝜗𝜗𝑑𝑑
∑𝑚𝑚
𝑛𝑛 2
∙ 𝑡𝑡 ∙ 100
𝑥𝑥𝑛𝑛
𝐷𝐷𝑡𝑡 �𝜗𝜗𝑐𝑐,𝑛𝑛 � =
𝐿𝐿(𝜗𝜗𝑟𝑟𝑟𝑟𝑟𝑟 )
where 𝑥𝑥𝑛𝑛 is the number of time steps, 𝐷𝐷𝑦𝑦 (𝜗𝜗𝑐𝑐 ) is the yearly degradation rate at conductor
temperature 𝜗𝜗𝑐𝑐 , 𝐷𝐷𝑦𝑦 (𝜗𝜗𝑟𝑟𝑟𝑟𝑟𝑟 ) is the yearly degradation rate at reference temperature and 𝜗𝜗𝑑𝑑
Energies 2021, 14, 494 11 of 26

is the temperature increase that doubles the degradation rate. Montsinger has demon-
strated a correlation for the aging of paper-oil insulation systems. “He noted that the rate
of deterioration of mechanical properties doubled for each 5–10 °C increase in tempera-
ture. The doubling factor was not a constant, being about 6 °C in the temperature range
from 100–110 °C and 8 °C for temperatures above 120 °C. However, people tend to re-
member the doubling factor as a constant and the present IEC Loading Guide uses 6 °C”
[44].
A comparison between Montsinger rule and Arrhenius law reveals a significant dif-
ference, shown in Figure 3. There is only agreement for a small temperature range. Oth-
erwise, the Montsinger rule always yields higher aging in the given case (cf. [46]).

Aging Acceleration Factor 𝐹𝐹AA

20
15
10
5
ln(𝐹𝐹A 𝐴𝐴 )

0
Montsinger
–5
Arrhenius
–10
–15
–20
14
28
42
56
70
84
98
0

112
126
140
154
168
182
196
210
224
238
Temperature (°C)

Figure 3. Aging acceleration factor according to Arrhenius reaction rate theory (B = 15,000) and
Montsinger rule (𝑇𝑇𝑑𝑑 = 6 °C) at a reference hottest point temperature of 110 °C.

For the calculation of the degeneration rate with regard to thermal aging, the load-
dependent conductor temperature 𝜗𝜗𝑐𝑐,𝑛𝑛 is required analogous to the hottest point temper-
atures for transformers. The corresponding values are determined with Formula (14),
which reflect the conductor temperature under consideration of the thermal time constant
at the end of a time step, which is assumed to comprise 300 s (5 min). The resolution of 5
min was chosen to ensure high accuracy for dynamic temperature calculation. These time
steps differ from the generation fluctuations considered for DER, for which hourly varia-
tions are assumed.
For the lifetime assessment, there is a need for reference data for the service life of
the cable at a given temperature. Together with the Arrhenius parameters or Montsinger
rule, these variables provide the necessary database for thermal aging. Currently, there is
little or no data on this. Further research is needed in this respect, especially since the
reference values have a major influence on the final outcome (see Figure 4). At this point,
the Montsinger rule and related reference data from the literature are compared with the
Arrhenius law and related estimated reference data. For the latter, a service life of 15 years
at maximum load is assumed, based on the minimum lifetime for PILC cables under re-
placement [36]. A continuous load of 100% as well as 70% is regarded as maximum load.
In addition, for the Arrhenius parameter 𝐵𝐵 the value according to the transformer insu-
lation from [44] is used. Based on these assumptions, the following calculation for the
degeneration according to the Arrhenius Law results:
Energies 2021, 14, 494 12 of 26

𝐵𝐵 𝐵𝐵 15,000 15,000
� − � � − �
𝜗𝜗𝑟𝑟𝑟𝑟𝑟𝑟 +273 𝜗𝜗𝑐𝑐,𝑛𝑛 +273 𝜗𝜗𝑟𝑟𝑟𝑟𝑟𝑟 +273 𝜗𝜗𝑐𝑐,𝑛𝑛 +273
∑𝑚𝑚
𝑛𝑛 𝑒𝑒 ∑𝑚𝑚
𝑛𝑛 𝑒𝑒
∙ ∆𝑡𝑡 ∙ 100 ∙ ∆𝑡𝑡 ∙ 100 (20)
𝑥𝑥𝑛𝑛 𝑥𝑥𝑛𝑛
𝐷𝐷𝑡𝑡 �𝜗𝜗𝑐𝑐,𝑛𝑛 � = =
𝐿𝐿(𝜗𝜗𝑟𝑟𝑟𝑟𝑟𝑟 ) 15 𝑎𝑎
In [45] reference data for the Montsinger rule is given. A yearly degradation rate of
0.5% for PILC cables at a reference temperature of 15 °C and 𝜗𝜗𝑑𝑑 = 6.5 °C is assumed. This
corresponds to an expected lifetime at this temperature of 200 years:
𝐹𝐹𝐸𝐸𝐸𝐸𝐸𝐸 ∙ ∆𝑡𝑡 ∙ 100 1 ∙ 1𝑎𝑎 ∙ 100
𝐿𝐿(𝑇𝑇) = = (21)
𝐷𝐷𝑦𝑦 �𝑇𝑇𝑟𝑟𝑟𝑟𝑟𝑟 � 0.5

A conductor temperature of approx. 15 °C is reached if the surrounding temperature

is 10 °C and a load of 30% is present. In relation to the typical age for replacement of PILC
cables of 50 years and average utilization rates in the medium-voltage grids of 20 to 35%,
a reference lifetime of 200 years (cf. [45]) seems very high. At 15 °C, however, almost no
thermal aging takes place. Instead, other aging mechanisms dominate the service life. This
explains the high lifetime values when only thermal aging is considered at such low tem-
peratures.
Figure 4 shows the lifetime curves related to the different aging models with the ref-
erence data and assumptions presented above.

Montsinger 15 °C/200a Arrhenius 37 °C/15a/B=15000

Arrhenius 65 °C/15a/B=15000 Arrhenius 65 °C/15a/B=7500

200
150
Lifetime [a]
100
50
0

10 15 20 25 30 35 40 45 50 55 60 65 70

Conductor Temperature [°C]

Montsinger 15 °C/200a Arrhenius 37 °C/15a/B=15000

Arrhenius 65 °C/15a/B=15000 Arrhenius 65 °C/15a/B=7500

200
150
Lifetime [a]

100
50
0

30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
Utilization Regarding Permissable Current at 10°C Ground Temperature [%]

Figure 4. Lifetime regarding different reference data and aging models.

Energies 2021, 14, 494 13 of 26

The curves according to Arrhenius law reflect a cable service life of 15 years with an
assumed constant load of 70% (37 °C conductor temperature) on the one hand and of 100%
(65 °C conductor temperature) on the other hand as a reference point. Arrhenius constant
B has a considerable influence on the slope of the lifetime curve. For comparison, in Figure
4 also a curve with B = 7500 instead of B = 15,000 is included. The reference data for the
Arrhenius model are based only on simplified assumptions at this point. For this reason,
the Montsinger rule and the reference data given in [45] are used for the lifetime assess-
ment of PILC cables presented in this paper. Lifetime assessment for transformers is also
typically performed according to this model. According to the explanations in this chap-
ter, the selected database probably provides extreme values for the thermal aging of PILC
cables.
The large deviation of the curves illustrates the significance of the reference data.
There is a need for further research on this topic. As more accurate reference data become
available, this should be used to review the results of the lifetime assessment in Section 6.

4. MV Distribution Network Benchmark

The CIGRE Task Force C6.04.02 has developed benchmark systems to facilitate the
analysis and validation of methods and techniques that enable the economic, robust, and
environmentally responsible integration of DER. In [48] network benchmarks for different
voltage levels and versions for North American style with 60 Hz and European style with
50 Hz as a common basis for testing are presented. Out of this, the medium voltage (MV)
distribution network benchmark with 50 Hz is used to evaluate the aging of PILC cables
in case of different load scenarios, shown in Figure 5.

HV-MV Subtransmission Ne twork 110 kV

110/20 kV 110/20 kV

1 12
2.8 km
Fe e der 1 2 Fe e der 2
4.9 km
Conside red
4.4 km Line Se gment 2
3 13
0.6 km
4
1.3 km 3.0 km
0.6 km 0.5 km S 3
5 11 14
0.3 km
8 S1
10 2.0 km
9 1.7 km
0.8 km 0.3 km 7
Load
S2
0.2 km Transformer
6
1.5 km Bus
S witc h/CB

Figure 5. Topology of European MV distribution network benchmark, following [48].

Energies 2021, 14, 494 14 of 26

The selected network benchmark represents a typical European medium-voltage

grid that is derived from a physical network in southern Germany, which supplies a small
town and the surrounding rural area. Depending on the switch position, the network can
be operated radially or meshed. At this point the radially version with open switches is
considered. The MV distribution network has a nominal voltage of 20 kV and is fed from
the 110 kV subtransmission network via separate transformers. XLPE cables (type
NA2XS2A) in the area of feeder 1 and aluminum overhead lines in the area of feeder 2 are
used as conductors. As an investigation of PILC cables is carried out here, the cable type
originally used in the network benchmark is adapted. PILC cables (type NEKEBA 3 × 120
mm² 12/20 kV) with the parameters from Table 4 are chosen for all line segments.

Table 4. PILC cable tested for thermal aging.

𝑹𝑹𝑫𝑫𝑫𝑫 at 20 °C
Type X (Ohm/km) C (nF/km) Un (kV) R0/R1 (p.u.) X0/X1 (p.u.) C0 (nF/km)
(Ohm/km)
NEKEBA 3 × 120 mm² 0.157 0.123 338.0 20.0 9.48 3.29 338.0

In the MV distribution network benchmark different residential and commercial/in-

dustrial loads are included. Table 5 gives the values of the coincident peak loads for each
node of the network. Load values given for nodes 1 and 12 are much higher than those for
the other ones, because these loads represent additional feeders that are not modeled in
detail. The loads are assumed to be symmetric [48]. Line segments and their correspond-
ing lengths that connect the different network nodes are shown in Table 6.

Table 5. Load parameters of European MV distribution network benchmark [48].

Apparent Power, S (kVA) Power Factor

Node
Residential Commercial/Industrial Residential Commercial/Industrial
1 15,300 5100 0.98 0.95
2 --- --- --- ---
3 285 265 0.97 0.85
4 445 --- 0.97 ---
5 750 --- 0.97 ---
6 565 --- 0.97 ---
7 --- 90 --- 0.85
8 605 --- 0.97 ---
9 --- 675 --- 0.85
10 490 80 0.97 0.85
11 340 --- 0.97 ---
12 15,300 5280 0.98 0.95
13 --- 40 --- 0.85
14 215 390 0.97 0.85

In order to limit the scope of the investigations, only a line segment with high utili-
zation is analyzed according to the load scenarios from Section 5. Line segment 2 is se-
lected for the lifetime assessment, which is marked in Figure 5.
Several DER are additionally integrated into the network benchmark. According to
the wind power-dominated and photovoltaic dominated scenario, explained in Section 5,
the total installed apparent power of all wind power and photovoltaic (PV) plants is cho-
sen. Other DER are neglected because of their limited influence. Single power plants are
distributed randomly to nodes of feeder 1 (including node 1) as shown in Table 7. In the
wind power-dominated scenario, it is necessary to reinforce line segments 1 and 2 to avoid
high grid bottlenecks. For this purpose, a parallel PILC cable with an identical configura-
Energies 2021, 14, 494 15 of 26

tion according to Table 4 is set up. It is assumed that all plants are connected to the me-
dium-voltage grid. For power generators in this voltage level, a reactive power control
according to VDE 4110 is necessary. A voltage-dependent control is selected as shown in
Figure 6.

Table 6. Connections and line lengths of MV distribution network benchmark [48].

Line Segment Node from Node to Length (km)

1 1 2 2.82
2 2 3 4.42
3 3 4 0.61
4 4 5 0.56
5 5 6 1.54
6 6 7 0.24
7 7 8 1.67
8 8 9 0.32
9 9 10 0.77
10 10 11 0.33
11 11 4 0.49
12 3 8 1.3
13 12 13 4.89
14 13 14 2.99
15 14 8 2

Table 7. Distribution of wind power and photovoltaic plants in the MV distribution network.

Installed Apparent Power, S (kVA)

Node Name/Type
Wind Power-Dominated PV-Dominated
Node 1 PV1 2.11 2.83
Node 11 PV11 2.11 2.83
Node 4 PV4 2.11 2.83
Node 5 PV5 2.11 2.83
Node 10 PV10 2.11 2.83
Node 9 PV9 2.11 2.83
Node 8 PV8 2.11 2.83
Node 5 W5 6.5 2.115
Node 3 W3 6.5 2.115
Node 8 W8 6.5 2.115
Node 10 W10 1.5 2.115
Wind power plants (W) 21.0 8.5
Total
Photovoltaic plants (PV) 14.8 19.8
Energies 2021, 14, 494 16 of 26

Voltage-Dependent Reactive Power Control

0.4
0.33
0.3

0.2

0.1

Q (p.u.)
0.0
0.9 0.95 0.98 1.03 1.08 1.1
–0.1

–0.2

–0.3
–0.33
–0.4
U (p.u.)

Figure 6. Voltage-dependent reactive power control for power generators in the medium-voltage
grid.

5. Representative Load Scenarios

For the lifetime assessment of line segment 2 from the network benchmark different
load scenarios are investigated. On the one hand, a scenario is examined without DER,
i.e., with a load only by the connected power consumers. On the other hand, the influence
of DER is considered by a future prognosis about their penetration.
In [49] three scenarios of the future expansion of DER in Germany until 2032 are pre-
sented. These scenarios differ in the total installed capacity and the local distribution of
the plants. The scenario “EEG 2014”, for example, is taken into account as a minimal sce-
nario with 128 GW of DER in 2032 (60 GW wind power plants, 59 GW PV plants, 9 GW
others). This scenario is used for the lifetime assessment of line segment 2, as the required
data is the only publicly available. More recent political objectives in Germany are forcing
a comparatively higher level of expansion.
To distribute the total installed apparent power to federal states, grid areas and volt-
age levels, in [49], an analysis of existing networks and previously installed power plants
in Germany is used. Based on this, the expansion of DER is assigned to representative
medium-voltage (MV) and low-voltage (LV) model network classes
(“Modellnetzklassen”) using stochastic methods. Each of these classes (10 LV and 8 MV
classes) represent distribution grids with a similar typical distribution of loads and feed-
ins. The ratio of the average installed DER capacity to the annual maximum load for each
withdrawal point in the network is used as an adequate indicator of how strongly the load
flow in a distribution grid is influenced by DER. For example, this ratio is above 2 in MV
model network classes 7 and 8 (see Table 8) related to the installed capacity of photovoltaic
plants [49]. The higher the ratio, the greater the need for grid expansion or the number of
grid bottlenecks [50]. In [50], grid area classes (“Netzgebietsklassen”) are formed that rep-
resent combinations of different model network classes from [49]. Three classes with a
strong generation-related need for future grid expansion are detected, listed in Table 8.
With reference to the aforementioned studies, the following scenarios are used for
the lifetime assessment presented in Section 6:
 PV-dominated in Bavaria (MV model network class 8 from [49]),
 Wind power-dominated in Schleswig-Holstein (MV model network class 6 from
[49]).
In [49] the network structure parameters of the MV model network classes are as-
sumed to be 8 branches with 15 withdrawal points each, i.e., a total of 120 withdrawal
points. Multiplied with the average installed DER capacity per withdrawal point, the total
capacity of wind power and PV plants for the corresponding scenarios or grid area classes
emerge, shown in Table 9. These scenarios reflect the expected expansion of DER for the
year 2032 in particularly affected MV network areas in Germany.
Energies 2021, 14, 494 17 of 26

Table 8. Grid area classes [50].

Grid Area Classes Characteristics Model Network Classes from [49]

High feed-ins by photovoltaic and wind power plants;
Wind power-
Low load per withdrawal point; 4 (LV) + 6 (MV)
dominated
Typical for Schleswig-Holstein
Moderate feed-ins by photovoltaic systems and high
feed-ins by wind power plants;
Weak Load Low load per withdrawal point; 3 (LV) + 6 (MV)
Typical for Eastern Germany (Mecklenburg-
Vorpommern and Brandenburg)
High feed-ins by photovoltaic systems and low feed-ins
by wind energy plants;
Bavaria: 7 (LV) + 8 (MV)
PV-dominated High load per withdrawal point;
Baden-Württemberg: 10 (LV) + 7 (MV)
Typical for Southern Germany (Baden-Württemberg
and Bavaria)

Table 9. Expected value of installed capacity of DER in in particularly affected MV distribution networks in 2032.

Average Installed DER Capacity per Withdrawal

Scenarios/Grid Total DER Capacity
Point According to the Scenario “EEG 2014” from [49]
Area Classes
Photovoltaic (kW) Wind Power (kW) Photovoltaic (MW) Wind Power (MW)
Wind power-
123.2 178.9 Approx. 15 Approx. 21
dominated
PV-dominated 165 70.5 Approx. 20 Approx. 8

In order to create a profile that provides a holistic and realistic view of the loads on
a cable, a load flow analysis (Newton–Raphson method) is carried out over one year. Load
profiles are required for this purpose. For consumers, the standard load profiles H0 for
households and G0 for commercial/industry are selected from [51,52]. The profile H0 is
adjusted over the year according to the dynamization function in [51]. The corresponding
per unit load curves during the period from Friday to Sunday related to the second week
of 2012 are shown in Figures 7 and 8, as well as the annual load curves 2012.

H0 Standard Load Profile - Friday till Sunday H0 Standard Load Profile - Annual
1.2 1.2
1 1
0.8 0.8
S (p.u.)

0.6 0.6
S (p.u.)

0.4 0.4
0.2 0.2
0 0
0

1098
1647
2196
2745
3294
3843
4392
4941
5490
6039
6588
7137
7686
8235
549
00:00 Sun
04:00 Sun
08:00 Sun
12:00 Sun
16:00 Sun
20:00 Sun
00:00 Sat
04:00 Sat
08:00 Sat
12:00 Sat
16:00 Sat
20:00 Sat
00:00 Fri
04:00 Fri
08:00 Fri
12:00 Fri
16:00 Fri
20:00 Fri

Time (h)

Figure 7. Load profile H0 for households, following [51].

Energies 2021, 14, 494 18 of 26

G0 Standard Load Profile - Friday till Sunday G0 Standard Load Profile - Annual
1.2 1.2
1 1
0.8 0.8
0.6 0.6

S (p.u.)
S (p.u.)

0.4 0.4
0.2 0.2
0 0
00:00 Fri
04:30 Fri
09:00 Fri
13:30 Fri
18:00 Fri
22:30 Fri
03:00 Sat
07:30 Sat
12:00 Sat
16:30 Sat
21:00 Sat
01:30 Sun
06:00 Sun
10:30 Sun
15:00 Sun
19:30 Sun

1172
1758
2343
2929
3515
4101
4686
5272
5858
6444
7029
7615
8201
586
Time (h)

Figure 8. Load profile G0 for commercial/industry, following [51].

Since the values in Table 5 reflect coincident peak loads, the maximum values of the
individual loads are converted with a coincidence factor of 0.85. Using the two standard
load profiles H0 and G0, this results in the corresponding simultaneous peak values.
For DER, the generic feed-in profiles for 2030 from [53] are used. This is a federal
state specific standard data set for wind power and PV plants for different years of con-
sideration (2020–2050), which was created on the basis of a consistent and transparent
modeling. These feed-in profiles reflect hourly fluctuations in power generation of wind
power and PV plants. The selected feed-in profiles for Bavaria and Schleswig-Holstein are
shown as annual load duration curves in Figure 9.

Annual Load Duration Curve of Feed-in Profiles

for Wind Power and PV Plants
90%
Feed-in Power in Relation to Installed Apparent Power

80%

70%

60%

50%

40%

30%

20%

10%

0%
1

1133
1416
1699
1982
2265
2548
2831
3114
3397
3680
3963
4246
4529
4812
5095
5378
5661
5944
6227
6510
6793
7076
7359
7642
7925
8208
8491
284
567
850

Time (h)

PV Plants Schleswig-Holstein PV Plants Bavaria

Wind Power Plants Schleswig-Holstein Wind Power Plants Bavaria

Figure 9. Annual load duration curve of feed-in profiles from [53] for wind power and PV plants.

6. Results
For an exemplary PILC cable (see Tables 1 and 4) of line segment 2 from the MV
distribution network benchmark (see Figure 5 in Section 4), a lifetime assessment with
regard to thermal aging is performed. The results are presented in the following. The basis
for determining the thermal aging of cables is the conductor or insulation temperature
(see Section 2) that results from their utilization.
Energies 2021, 14, 494 19 of 26

High utilization or overloads in radially operated distribution networks occur espe-

cially in substations and the adjacent line segments, such as line segment 2 in the network
benchmark. The utilization of this line segment is determined by means of a load flow
analysis for three different load scenarios, namely without DER, wind power- and PV-
dominated (see Section 5). The corresponding annual load duration curves of line segment
2 in relation to the permissible current (see Formula (2): 𝐼𝐼𝑝𝑝 = 302 A ∙ 0.88) are shown in
Figure 10. The load limits of the transformer of feeder 1 are not violated, and the maximum
utilization is approx. 100%. In the scenarios with DER the highest voltage deviations of
approx. −3% and +2% relate to the nominal voltage level occurred at node 11 of the net-
work benchmark. Also at this node, the maximum voltage deviation in the scenario with-
out DER is about −4%.
Without DER the annual load duration curve is relatively flat, with a maximum uti-
lization of approx. 50%. In contrast, the corresponding curves of the scenarios with DER
show significant peak values, especially the PV dominated scenario. Peak Loads occur
temporarily due to the type of power generation and reach up to 120% of the maximum
permissible current in the PV-dominated scenario. Such short-term peaks are taken into
account at this point for the lifetime assessment to illustrate the corresponding effect on
service life, as some DSO also allow such overloads in practice (see Section 2.2). Alterna-
tively, however, the service life is also evaluated here if the feed-ins are reduced by the
DSO so that line segment 2 is utilized to a maximum of 100%.

Utilization of Line Segment 2

120%
110%
100%
90%
Load related to Ip

80%
70%
60%
50%
40%
30%
20%
10%
0%
1
314
627
940
1253
1566
1879
2192
2505
2818
3131
3444
3757
4070
4383
4696
5009
5322
5635
5948
6261
6574
6887
7200
7513
7826
8139
8452

Time (h)
Without DER Wind Power-Dominated PV-Dominated

Figure 10. Annual load duration curves of line segment 2 for different load scenarios.

Figure 11 shows the correlation of the utilization of line segment 2 with the network
load of feeder 1 (see Figure 5) for the scenarios wind power-dominated and without DER.
Alternatively, the correlation with the network load of the entire MV distribution network
benchmark is shown in Figure 12. The network load 𝑃𝑃𝑁𝑁 is calculated as follows:

𝑃𝑃𝑁𝑁 = 𝑚𝑚𝑚𝑚𝑚𝑚 �� 𝑃𝑃𝐿𝐿 , � 𝑃𝑃𝐹𝐹 � (22)

where ∑ 𝑃𝑃𝐿𝐿 is the sum of all withdrawals from loads and network losses and ∑ 𝑃𝑃𝐹𝐹 is the
sum of all feed-ins in a network area, e.g., feeder 1. The illustrations indicate that without
decentralized feed-ins, the network load of an entire distribution network correlates well
with the load of a selected individual line. With the penetration of DER, this is usually no
longer the case, especially if there are different degrees of expansion of decentralized
power plants in subnetwork areas. In the selected example, DER are installed exclusively
in feeder 1 to represent an extreme scenario. If DER are in place, the correlation between
Energies 2021, 14, 494 20 of 26

the network load and the utilization of single line segments exists only if the network load
is formed for interconnected network areas (INA, cf. [54]). However, this does not neces-
sarily apply to all line segments of an INA. In areas with few or no feed-ins within an INA,
there can be no correlation. This is the case, for example, for line segment 5 and the net-
work load of feeder 1, as shown in Figure 13. For line segment 5, which has a very low
utilization rate, the line load correlates with the sum of all withdrawals from loads and
network losses ∑ 𝑃𝑃𝐿𝐿 (see Formula (22)) from Feeder 1. The correlations presented here can
be useful for congestion management in order to assess in which network situations bot-
tlenecks occur.

Correlation Between Line Load (L2) and Network Correlation Between Line Load (L2) and Network
Load - With DER Load - Without DER
100% 25 60% 5

Network Load (MW)

Network Load (MW)

80% 20 50% 4
40%
Line Load

Line Load
60% 15 3
30%
40% 10 2
20%
20% 5 10% 1
0% 0 0% 0
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85
Time (h) Time (h)

Line Load (L2) Network Load (Feeder 1) Line Load (L2) Network Load (Feeder 1)

Figure 11. Correlation of the load of line segment 2 with the network load of feeder 1 for the scenarios wind power-
dominated (left) and without DER (right).

Correlation Between Line Load (L2) and Network Correlation Between Line Load (L2) and Network
Load - With DER Load - Wihtout DER
100% 50 60% 50
Network Load (MW)

Network Load (MW)

80% 40 40
40%
Line Load

Line Load

60% 30 30
40% 20 20
20%
20% 10 10
0% 0 0% 0
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85
Time (h) Time (h)
Line Load (L2) Load Line 2
Network Load (MV Distribution Network Benchmark) Network Load (MV Distribution Network Benchmark)

Figure 12. Correlation of the load of line segment 2 with the network load of the entire MV distribution network bench-
mark for the scenarios wind power-dominated (left) and without DER (right).

Correlation Between Line Load (L5) and Network Load - Correlation Between Line Load (L5) and Sum of all
With DER Withdrawals from Loads and Network Losses - With DER
2.0% 25 2.0% 5
Network Load (MW)
Network Load (MW)

20 4
1.5% 1.5%
Line Load
Line Load

15 3
1.0% 1.0%
10 2
0.5% 0.5%
5 1

0.0% 0 0.0% 0
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85 1 5 9 13 17 21 25 29 33 37 41 45 49 53 57 61 65 69 73 77 81 85

Time (h) Time (h)

Line Load (L5)
Line Load (L5) Network Load (Feeder 1) Sum of all Withdrawals from Loads and Network Losses (Feeder 1)

Figure 13. Correlation of the load of line segment 5 with the network load (left) and the sum of all withdrawals from loads
and network losses (right) from Feeder 1.
Energies 2021, 14, 494 21 of 26

To determine the conductor temperature of the selected PILC cable type according
to the load of line segment 2, the methodology presented in Section 3 is used. Within this
approach all heat sources are considered as Joule losses or losses converted therein. This
allows a simplified modelling with one heat source. As a further simplification, the dry-
ing-out of the soil is ignored. However, in Section 2, formulas are shown to account for
this effect. In addition, a constant ground temperature of 10 °C is assumed. This corre-
sponds roughly to the annual mean value in Germany at a depth of 2 m, which has risen
from approx. 10 to 12 °C between 1900 and 2020 [55]. A more precise temperature deter-
mination could be made by varying the ground temperature during the year, e.g., using
Formula (16).
The calculation of the conductor temperature was carried out using both a dynamic
and a static calculation method. The dynamic calculation is made according to Formula
(14). Thereby a thermal time constant of 2 h (cf. [27]) is taken into account. If lower thermal
time constants would apply, e.g., 30 min, only a small deviation from a static analysis
would occur (approx. 86% steady-state temperature after ∆𝑡𝑡 = 2 ∙ 𝜏𝜏), when hourly load
fluctuations are considered. For comparison with the dynamic computation, a static cal-
culation is also performed according to Formula (13). The conductor temperatures result-
ing from both calculation methods with an exemplary course of the cable load is shown
in Figure 14. The figure illustrates the significant effect when thermal time constants are
respected.

Temperature and Load Curves of Line Segment 2

100 100
Static Temperature (°C)
90 90
Dynamic Temperature (°C)
80 80
Load (%)
70 70
Temperature (°C)

60 60

Load (%)
50 50
40 40
30 30
20 20
10 10
0 0
08:00
08:20
08:40
09:00
09:20
09:40
10:00
10:20
10:40
11:00
11:20
11:40
12:00
12:20
12:40
13:00
13:20
13:40
14:00
14:20
14:40
15:00
15:20
15:40
16:00
16:20
16:40
17:00
17:20
17:40
18:00

Figure 14. Comparison of the static and dynamic conductor temperature calculation method.

Based on the conductor temperature, the service life of the PILC cable is determined
using the Montsinger rule according to Formula (19) and the reference data from [45]. As
described in Section 3, there is an urgent need for solid reference data for aging models of
PILC cables.
Figure 15 presents the different resulting lifetimes for the three analyzed load scenar-
ios if a maximum utilization of 120% is allowed, whereby a distinction is made between
static and dynamic temperature calculation. In addition, the service life calculated with
the average annual load instead of time series values is included in the figure. Currently,
this approach is commonly used by DSO to estimate the degeneration of cables. As Figure
15 shows, this methodology leads to approximately identical lifetimes compared to a time
series analysis for a load scenario without DER. Also, in this scenario only slight differ-
ences result to the lifetimes with static compared to dynamic temperature calculation. The
small differences can be explained essentially by the relatively flat annual load duration
curve.
Energies 2021, 14, 494 22 of 26

In principle, however, the thermal aging in the scenario without DER is negligible so
that no relevant reduction in service life due to the load is to be expected. Other aging
mechanisms determine the lifetime of the cable in this case. If only thermal aging is exam-
ined, like it is done at this point, very high lifetimes result. Typically, PILC cables are
replaced after 50 years and after 80 years at the latest [3].

100
Lifetime with Regard to Thermal Ageing [a]
90
80
70 Typical Age
for
60 Replacement
50
40
30
20
10
0
Static Dynamic Static Dynamic Static Dynamic
Without DER Wind Power-Dominated PV-Dominated
Calculated Lifetime with Time Series
214.11 216.57 34.35 41.57 9.90 25.98
Analysis and Montsinger Rule (a)
Calculated Lifetime with Average Annual
229.45 229.45 179.08 179.08 238.41 238.41
Load (a)
Average Annual Load (%) 26.79 26.79 34.10 34.10 25.46 25.46

Figure 15. Calculated lifetime for the exemplary PILC cable of line segment 2 regarding different reference data and aging
models if a maximum utilization of 120% is allowed (typical age for replacement from [3]).

For load scenarios with DER, the service life of the PILC cable differs tremendously
between an assessment according to the average annual load and time series analysis. The
annual average utilization even decreases due to the integration of DER in the PV domi-
nated scenario. In contrast, the peak load increases clearly. The lifetime values in Figure
15 indicates that the high peak loads accelerate the degeneration of PILC cables signifi-
cantly, even if these are only temporarily occurring. This is due to the fact that the thermal
aging increases exponentially with load. The service life is particularly low in the PV-
dominated scenario. Here, peak loads occur more frequently than in the wind power-
dominated scenario. Furthermore, substantial discrepancies between the static and dy-
namic temperature calculation method can be seen. The results show that thermal time
constants urgently need to be taken into account in the lifetime assessment of cables with
regard to thermal aging. Otherwise the service life would be assessed to low.
When evaluating the results presented in this section, it should be noted that the life-
time data represent extreme values. The load on the line segment under consideration
represents an extremum. DSO could reduce the feed-ins during operation to avoid peak
loads. This is often permissible to the extent that only a certain proportion of the energy
volume generated annually by decentralized generation plants is allowed to be reduced,
e.g., 3%. In this context, an additional lifetime assessment was performed under the con-
dition that a DSO allows a maximum utilization of 100% of the cable under consideration.
In order to ensure this, less than 3% of the annual amount of energy generated by DER
must be reduced. The results are shown in Figure 16. They indicate that the service life of
the cable can be significantly increased by avoiding overloads. The service life is approx.
45 years in both DER scenarios, covering nearly the typical service life of PILC cables cur-
rently in use with the previous load without DER.
Energies 2021, 14, 494 23 of 26

If overloading is prevented, the maximum temperature of the conductor is always

below 65 °C. If an overload of 20% is allowed, the maximum temperature is approx. 78 °C
in the wind power-dominated and approx. 85 °C in the PV-dominated scenario. The tem-
peratures are additionally about 7 °C higher, if a static calculation without thermal time
constants is performed.
For all results of the lifetime assessment it has to be considered that an additional
consideration of the fluctuating soil temperatures during the course of the year and the
drying-out of the soil could lead to higher peak conductor or insulation temperatures and
thus to a higher degradation. Furthermore, the results of the lifetime estimation presented
in this section should only be considered as an approximation if the selected reference
data from [45] for the aging model are actually applicable.

100
Lifetime with Regard to Thermal Ageing [a]

90
80
70 Typical Age
for
60 Replacement
50
40
30
20
10
0
Static Dynamic Static Dynamic Static Dynamic
Without DER Wind Power-Dominated PV-Dominated
Calculated Lifetime with Time Series
214.11 216.57 39.80 45.11 23.48 45.44
Analysis and Montsinger Rule (a)
Calculated Lifetime with Average Annual
229.45 229.45 179.17 179.17 239.37 239.37
Load (a)
Averaged Annual Load (%) 26.79 26.79 34.09 34.09 25.32 25.32

Figure 16. Calculated Lifetime for the exemplary PILC cable of line segment 2 regarding different reference data and aging
models if a maximum utilization of 100% is allowed (typical age for replacement from [3]).

7. Discussion and Outlook

In this paper, the thermal aging of PILC cables under the influence of DER is consid-
ered for extreme scenarios. In networks with DER short-term peak loads on electrical
equipment can occur, even if the average load does not necessarily rise. For PILC cables
this could lead to significantly increased thermal aging. This is due to the fact that thermal
aging increases exponentially with utilization of PILC cables. Excessive aging of PILC ca-
bles due to the influence of DER could already be reduced by preventing overloads in a
few hours of a year by means of generation or load management. In order to achieve this
in a cost-optimized way, a compromise between the aging costs and the costs of the energy
management should be considered. The maximum permissible load on the electrical
equipment should be limited so that the costs of the corresponding accelerated aging do
not exceed the benefit of the higher transmission capacity. The results of the lifetime esti-
mation presented in this paper should only be considered as an approximation if the se-
lected reference data from the literature for the aging model are actually applicable.
Usually, lifetime assessment of PILC cables is performed based on the average annual
load. It is shown in this paper that this approach is only suitable for load situations with-
out DER, and that a time series analysis can overcome the shortcomings in situation with
decentralized and fluctuating feed-ins from wind power or PV plants. The determination
Energies 2021, 14, 494 24 of 26

of thermal aging is based on the calculation of the conductor or insulation temperature.

The consideration of thermal time constants has a significant influence on the determina-
tion of the temperature and the corresponding thermal aging. If the thermal dynamics are
not taken into account, the estimated service life turns out to be considerably too low.
Beyond the methodology used in this paper, the accuracy of the lifetime assessment
can be further improved by taking into account the fluctuating soil temperatures over the
course of the year and the drying-out of the soil. An approach to how this can be achieved
by using correction factors is also shown in this paper. If these influences are taken into
account, this could lead to higher resulting peak temperatures and greater degeneration
than without them. A possible approach to dealing with varying soil temperatures
throughout the year could be to allow different maximum cable loads depending on the
season.
In addition to the influence of DER, the methodology described in this paper could
also be used to determine the influence of consumers with high power requirements, e.g.,
electric vehicle charging stations, on the thermal aging or service life of PILC cables.

Author Contributions: Methodology, M.Z.; validation, M.Z., A.-C.M., T.B., H.P., I.M. and C.W.;
formal analysis, M.Z., A.-C.M., T.B., H.P., I.M. and C.W.; investigation, M.Z., A.-C.M. and T.B.;
resources, M.Z.; writing—original draft preparation, M.Z. All authors have read and agreed to the
published version of the manuscript.
Funding: The APC was funded by the Bavarian Academic Forum (BayWiss).
Informed Consent Statement: Not applicable.
Data Availability Statement: Publicly available datasets were analyzed in this study as the basis
for calculations. References to the data are made at the appropriate places in the text. Calculation
results are shown in figures. Further detailed data in this respect are available on request.
Acknowledgments: The authors would like to thank the editors and reviewers for critically re-
viewing the manuscript.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Chmura, L.A. Life-Cycle Assessment of High-Voltage Assets Using Statistical Tools. Ph.D. Thesis, Delft University of Technol-
ogy, Delft, The Netherlands, 2014; ISBN 9461823967.
2. FGH. Forschungsgemeinschaft für Elektrische Anlagen und Stromwirtschaft e.V. Zustandsdiagnose von Papiermasse-Kabel-
anlagen im Verteilungsnetzen. In Technischer Bericht 300; FGH: Mannheim, Germany, 2006.
3. Conseil International des Grands Réseaux Électriques. Remaining Life Management of existing AC Underground Lines. Technical
Brochures; CIGRÉ: Paris, Franch, 2008.
4. Künstl, B. Stromtragfähigkeitsgrenzen von Elektrischen Leitungssystemen. Diplomarbeit; Technische Universität Graz; Institut für
Hochspannungstechnik und Systemmanagement: Graz, Austria, 2011.
5. Adam, R.; Großmann, S.; Haller, R. Beitrag zur Thermischen Dimensionierung von Niederspannungs-Schaltgerätekombinatio-
nen. Ph.D. Thesis, Technische Universität Dresden, Dresden, Germany, 2019; ISBN 3000637443.
6. Oeding, D.; Oswald, B.R. Elektrische Kraftwerke und Netze, 8. Auflage; Springer Vieweg: Berlin, Germany, 2016; ISBN
9783662527030.
7. Brüggmann, J. Wechselspannungstechnologiebasierte Bipolare Mehrphasensysteme. Ph.D. Thesis, Universität Duisburg-Essen,
Essen, Germany, 2012.
8. Linenbrink, T.L. The history and mystery of the Neher-McGrath formula. Consult.-Specif. Eng. 2014, July, 46–52.
9. Schermeyer, H. Netzengpassmanagement in Regenerativ Geprägten Energiesystemen. Untersuchungen zur Abregelung Erneu-
erbarer Energien und zur Sektorenkopplung in Einem Deutschen Verteilnetz. Ph.D. Thesis, Karlsruher Instituts für Technologie
(KIT), Karlsruhe, Germany, 2018.
10. Schäfer, K.F. Netzberechnung: Verfahren zur Berechnung elektrischer Energieversorgungsnetze, 1st ed.; Springer Vieweg: Wiesbaden,
Germany, 2020; ISBN 365826733X.
11. Sillaber, A. Leitfaden zur Verteilnetzplanung und Systemgestaltung. Entwicklung dezentraler Elektrizitätssysteme; Springer Vieweg:
Wiesbaden, Germany, 2016; ISBN 9783658147136.
12. Woschitz, R.; Schichler, U.; Pirker, A.; Komar, G. FEM-Simulation des Thermischen Langzeitverhaltens von Hochspannungs-Kabel-
anlagen bei Laständerungen; Symposium Energieinnovation: Graz, Austria, 2016.
Energies 2021, 14, 494 25 of 26

13. Czaap, S. Effect of soil moisture on current-carrying capacity of low-voltage power cables. Prz. Elektrotech. 2019, 1, 156–161,
doi:10.15199/48.2019.06.29.
14. Lindström, L. Evaluating Impact on Ampacity According to IEC-60287 Regarding Thermally Unfavourable Placement of Power Cables;
KTH Electrical Engineering: Stockholm, Sweden, 2011.
15. Cichowski, R.R.; Kliesch, M. Kabelhandbuch; EW Medien und Kongresse: Frankfurt, Germany, 2012; ISBN 9783802210563.
16. Schlabbach, J. Elektroenergieversorgung. Betriebsmittel, Netze, Kennzahlen und Auswirkungen der Elektrischen Energieversorgung, 3rd
ed.; VDE-Verl.: Berlin, Germany, 2009; ISBN 9783800731084.
17. Scheffler, J. Verteilnetze auf dem Weg zum Flächenkraftwerk. Rechtlicher Rahmen, Erzeuger, Netze; Springer: Berlin/Heidelberg, Ger-
many, 2016; ISBN 9783642552960.
18. Heinhold, L. Kabel und Leitungen für Starkstrom, 4th ed.; Siemens AG: Berlin, Germany, 1989; ISBN 3800915243.
19. Buhari, M. Reliability Assessment of Ageing Distribution Cable for Replacement in Smart Distribution Systems; University of Manches-
ter: Manchester, UK, 2016.
20. Tang, W.H.; Wu, Q.H. Condition Monitoring and Assessment of Power Transformers Using Computational Intelligence; Springer: Lon-
don, UK, 2011; ISBN 9780857290519.
21. Heizmann, T. Berechnungsmethoden für Auslegung, Betrieb und Sicherheit von Elektrischen Energieversorgungssystemen. Thermische
Berechnung von Kabelanlagen; FKH-/VSE—Fachtagung: Zürich, Switzerland, 2011.
22. Hruška, M.; Clauser, C.; De Doncker, R.W. The Effect of Drying around Power Cables on the Vadose Zone Temperature. Vad.
Zone J. 2018, 17, 180105, doi:10.2136/vzj2018.05.0105.
23. Rerak, M.; Ocłoń, P.; Taler, J.; Węglowski, B.; Sobota, T. The effect of soil and cable backfill thermal conductivity on the temper-
ature distribution in underground cable system. E3S Web Conf. 2017, 13, 2004, doi:10.1051/e3sconf/20171302004.
24. Rummich, E. Energiespeicher: Grundlagen, Komponenten, Systeme und Anwendungen, 2nd ed.; Expert Verlag: Renningen, Germany,
2015; ISBN 9783816932970.
25. Schröder, D. Elektrische Antriebe—Grundlagen, 6th ed.; Springer Viewg: Berlin, Germany, 2017; ISBN 9783662554487.
26. Specovius, J. Grundkurs Leistungselektronik. Bauelemente, Schaltungen und Systeme, 8, Erweiterte und Aktualisierte Auflage; Springer
Vieweg: Wiesbaden, Germany, 2017; ISBN 9783658169114.
27. Wieben, E. Multivariates Zeitreihenmodell des Aggregierten Elektrischen Leistungsbedarfes von Standardverbrauchern für die
Probabilistische Lastflussberechnung. Ph.D. Thesis, Technische Universität Clausthal, Clausthal-Zellerfeld, Germany, 2008.
28. Sami, T.; Gholami, A.; Shahrtash, S.M. Effect of electro-thermal, load cycling and thermal transients stresses on high voltage
XLPE and EPR power cable life. In Proceedings of the 8th International Conference on Advances in Power System Control,
Operation and Management (APSCOM 2009), Hong Kong, China, 8–11 November 2009; pp. 1–6, doi:10.1049/cp.2009.1818.
29. Ma, K.; He, N.; Liserre, M.; Blaabjerg, F. Frequency-Domain Thermal Modeling and Characterization of Power Semiconductor
Devices. IEEE Trans. Power Electron. 2016, 31, 7183–7193, doi:10.1109/TPEL.2015.2509506.
30. Kerber, G. Aufnahmefähigkeit von Niederspannungsverteilnetzen für die Einspeisung aus Photovoltaikkleinanlagen. Ph.D.
Thesis, Technische Universität München, München, Germany, 2011.
31. Recknagel, H.; Schramek, E.-R. Taschenbuch für Heizung und Klimatechnik. Einschließlich Warmwasser- und Kältetechnik, 73rd ed.;
Oldenbourg: München, Germany, 2007; ISBN 9783835631045.
32. Müller, A.-C.; Weindl, C.; Linossier, J.P.; Schramm, J.; Benkert, L. Requirements of a system for the artificial ageing of paper
insulated lead covered medium voltage cables. In 2018 International Conference on Diagnostics in Electrical Engineering (Di-
agnostika), Pilsen, Czech Republic, 4–7 September 2018; pp 1–4; ISBN 978-1-5386-4423-2.
33. Stötzel, M. Strategische Ressourcendimensionierung von Netzleitstellen in Verteilungsnetzen, 1st ed.; Gerettete Schriften-Verl. epubli
GmbH: Wuppertal, Germany, 2014; ISBN 9783844278262.
34. Balzer, G.; Schorn, C. Asset Management für Infrastrukturanlagen—Energie und Wasser, 3rd ed.; Springer: Berlin, Germany, 2020;
ISBN 9783662615263.
35. Willis, H.L.; Welch, G.V.; Schrieber, R.R. Aging Power Delivery Infrastructures; CRC Press: New York, NY, USA, 2001; ISBN
0824705394.
36. Chmura, L.; Morshuis, P.H.F.; Mehairjan, R.P.Y.; Smit, J.J. Review of the residual life assessment of high-voltage cables by means
of bottom-up and top-down analysis—Dutch experience. Gaodianya Jishu 2015, 41, 1114–1124.
37. Feilat, E.A. Lifetime Assessment of Electrical Insulation. In Electric Field; Kandelousi, M.S., Ed.; IntechOpen: London, UK, 2018;
ISBN 978-1-78923-186-1.
38. Chmura, L.; Jin, H.; Cichecki, P.; Smit, J.; Gulski, E.; Vries, F. Use of dissipation factor for life consumption assessment and
future life modeling of oil-filled high-voltage power cables. IEEE Electr. Insul. Mag. 2012, 28, 27–37,
doi:10.1109/MEI.2012.6130529.
39. Mazzanti, G. The combination of electro-thermal stress, load cycling and thermal transients and its effects on the life of high
voltage ac cables. IEEE Trans. Dielect. Electr. Insul. 2009, 16, 1168–1179, doi:10.1109/TDEI.2009.5211872.
40. Mazzanti, G. Analysis of the Combined Effects of Load Cycling, Thermal Transients, and Electrothermal Stress on Life Expec-
tancy of High-Voltage AC Cables. IEEE Trans. Power Deliv. 2007, 22, 2000–2009, doi:10.1109/TPWRD.2007.905547.
41. Houtepen, R.; Chmura, L.; Smit, J.J.; Quak, B.; Seitz, P.P.; Gulski, E. Estimation of dielectric loss using damped AC voltages.
IEEE Electr. Insul. Mag. 2011, 27, 20–25, doi:10.1109/MEI.2011.5871365.
42. Densley, J. Ageing mechanisms and diagnostics for power cables—An overview. IEEE Electr. Insul. Mag. 2001, 17, 14–22,
doi:10.1109/57.901613.
Energies 2021, 14, 494 26 of 26

43. Buhari, M.; Levi, V.; Awadallah, S.K.E. Modelling of Ageing Distribution Cable for Replacement Planning. IEEE Trans. Power
Syst. 2016, 31, 3996–4004, doi:10.1109/TPWRS.2015.2499269.
44. Institute of Electrical and Electronics Engineers. IEEE Guide for Loading Mineral.-Oil-Immersed Transformers and Step-Voltage Reg-
ulators; C57.91-2011 (Revision of IEEE Std C57.91-1995); IEEE: New York, NY, USA, 2012; ISBN 78-0-7381-7195-1.
45. Huifei, J. Application of Dielectric Loss Measurements for Life Consumption and Future Life Estimation Modeling of Oil-Im-
pregnated Paper Insulation in HV Power Cables. Master’s Thesis, Delft University of Technology, Delft, The Netherlands, 2010.
46. Hayder, T.; Radakovic, Z.; Schiel, L.; Feser, K. Einfluss der Kurzschlussdauer auf die Alterung eines Transformators; Elektrie: Berlin,
Germany, 2003; Volume 57.
47. Weindl, C. Verfahren zur Bestimmung des Alterungsverhaltens und zur Diagnose von Betriebsmitteln der Elektrischen Energiever-
sorgung. Habilitation; Friedrich-Alexander Universität: Erlangen-Nürnberg, Germany, 2012.
48. Conseil international des grands réseaux électriques. Benchmark Systems for Network Integration of Renewable and Distributed En-
ergy Resources; CIGRÉ: Paris, Franch, 2014; ISBN 2858732701.
49. Büchner, J.; Katzfey, J.; Flörcken, O.; Moser, A.; Schuster, H.; Dierkes, S.; van Leeuwen, T.; Verheggen, L.; Uslar, M.; van
Amelsvoort, M. Moderne Verteilernetze für Deutschland (Verteilernetzstudie): Abschlussbericht. Available online:
https://www.bmwi.de/Redaktion/DE/Publikationen/Studien/verteilernetzstudie.pdf?__blob=publicationFile&v=5 (accessed on
23 October 2019).
50. Ecofys; Fraunhofer IWES. Smart-Market-Design in Deutschen Verteilnetzen: Entwicklung und Bewertung von Smart Markets
und Ableitung einer Regulatory Roadmap. Available online: https://www.agora-energiewende.de/fileadmin2/Pro-
jekte/2016/Smart_Markets/Agora_Smart-Market-Design_WEB.pdf (accessed on 23 October 2019).
51. BDEW. Bundesverband der Energie- und Wasserwirtschaft e.V. Repräsentative VDEW-Lastprofile: VDEW-Materialie M-32/99.
Available online: https://www.bdew.de/media/documents/1999_Repraesentative-VDEW-Lastprofile.pdf (accessed on 10 Feb-
ruary 2020).
52. BDEW. Bundesverband der Energie- und Wasserwirtschaft e.V. Anwendung der Repräsentativen VDEW-Lastprofile Step by
step: VDEW-Materialie M-05/2000. Available online: https://www.bdew.de/media/documents/2000131_Anwendung-reprae-
sentativen_Lastprofile-Step-by-step.pdf (accessed on 10 February 2020).
53. Koch, M.; Tambke, J. Erstellung Generischer EE-Strom-Einspeisezeitreihen mit Unterschiedlichem Grad an Fluktuierendem
Stromangebot. Available online: https://www.oeko.de/aktuelles/2016/daten-zur-einspeisung-erneuerbarer-energien (accessed
on 21 September 2020).
54. Zapf, M.; Weindl, C.; Pengg, H.; German, R. Specific Grid Charges for Controllable Loads in Smart Grids: A Proposal for a
Reform of the Grid Charges in Germany. In Proceedings of the NEIS 2018 Conference on Sustainable Energy Supply and Energy
Storage Systems, Hamburg, Germany, 20–21 September 2018; pp 237–242; ISBN 978-3-8007-4822-8.
55. Potsdam Institute for Climate Impact Research (PIK) e. V. Bodentemperatur: Jahresmittelwerte. Available online:
https://www.pik-potsdam.de/de/produkte/klima-wetter-potsdam/klimazeitreihen/bodentemperatur (accessed on 16 Septem-
ber 2020).