Professional Documents
Culture Documents
Particle Swarm Optimization of Power Cable Performance in Complex Surroundings
Particle Swarm Optimization of Power Cable Performance in Complex Surroundings
net/publication/323518234
CITATIONS READS
5 93
1 author:
Mamdooh Al-Saud
King Saud University
28 PUBLICATIONS 156 CITATIONS
SEE PROFILE
All content following this page was uploaded by Mamdooh Al-Saud on 15 May 2019.
Research Article
Mamdooh S. Al-Saud1,2
1College of Engineering, King Saud University, Riyadh, Saudi Arabia
2SaudiElectricity Company Chair in Power System Reliability and Security, King Saud University, Riyadh, Saudi Arabia
E-mail: mamdooh@ksu.edu.sa
Abstract: Underground cable performance indices such as maximum cable temperature and ampacity are non-continuous
functions of the configuration parameters such as depth and width of various trench layers. In this respect, existing traditional
gradient-type methods cannot be used to optimise such performance indices. This study presents an efficient methodology for
optimising power cable thermal performance with respect to configuration parameters involving cable spacing, depth of burial
and size of backfill. The new methodology integrates the powerful features of the finite elements (FEs) technique coupled with
the flexibility and effectiveness of the particle swarm optimisation (PSO) algorithm in order to handle various geometrical
parameters in the complex surrounding operating environment. The introduced methodology is tested using a commercial FE
simulation package used in conjunction with developed PSO code. The integrated methodology can be employed to minimise
the maximum cable temperature, minimise installation cost or maximise cable ampacity. Practical applications are presented for
15 kV cables, which demonstrate the usefulness and versatility of the presented methodology. Notable improvements have
been achieved by optimising the cable trench configuration parameters. For example, the cable ampacity was maximised,
optimising the cable spacing, barrier depth and backfill thermal conductivity, which resulted in an appreciable increase of 4.5%.
The thermal field in the cable medium is governed by the pi = u1, u2, …, um (6)
differential equation of the heat conduction [23]
The cognitive component pBest is calculated as
∂T
∇ ⋅ k∇T = − Q + c (1)
∂t pBesti t if f pi t + 1 ≥ f pBesti t
pBesti t + 1 = (7)
where T denotes the temperature at any point, k and c represent, pi t + 1 if f pi t + 1 < f pBesti t
respectively, the thermal conductivity and capacity, Q is the heat
generation per unit of area and t denotes the time. where pi(t + 1) is the particle's new position, pBesti(t) is the current
In steady-state thermal analysis of two-dimensional media, (4.1) personal best, and pBesti(t + 1) is the new personal best position of
reduces to the particle. The value of gBest represents the best fit that any
particle of the swarm has ever achieved as shown in equation
∂ ∂T ∂ ∂T below:
k + k +Q=0 (2)
∂x ∂x ∂y ∂y
gBest t = argMinin= 1 f (pBesti t (8)
On the basis of the formulation of the FE method, the temperature
at any point in the region subjected to specified boundary where n is the total number of particles. Both pBest and gBest are
conditions can be found by minimising the function utilised to define the velocity of the particle which guides the
particle toward a better solution. The velocity of a particle is thus
∫∫ ∫ ∫ QTdx dy
2 2
1 ∂T ∂T calculated as
I= kx + ky dx dy − (3)
2 ∂x ∂y
V i t + 1 = ω V i t + q1 c1 pBesti t − pi t
2.2 Cable optimisation model (9)
+ q2 c2 gBesti t − pi t
The power cable thermal circuit includes various regions of
different thermal specifications and heat generations with various where Vi(t) represents the current velocity of the particle i, Vi(t + 1)
dimensional scenarios. The cable temperature depends on the represents the new velocity the particle in order to move from the
thermal properties of the surrounding soil, the ambient current position to the new position. The parameters q1 and q2 are
margin for improvement is attained in comparison of the existing circuit is shown in Fig. 2b, which depicts three parallel circuits
(nominal) configuration design. directly buried in the soil and spaced out within a backfill layer.
The nominal values of the cable parameters, as well as the soil
3.1 System description thermal specification, are described in Table 1. A drain pipe (hot
water) of a diameter of 0.5 m is positioned at the proximity of the
A cross-section of 15 kV, 3 × 300 mm2 CU/XLPE/SWA/PVC cable trench.
underground cable representing a three-phase system is shown in In the applications presented in this section, the cable
Fig. 2a. Both configurations of two parallel circuits (two cables) performance is optimised for the existing practical complex
and three parallel circuits (three cables) directly laid out in the scenario involving the influence of pipes carrying fluids (or steam)
trench are being used by SEC in practise. The layout of the thermal within the vicinity of the cable. As outlined earlier, the integrated
drain temperature) for both two and three parallel circuits are
shown in Figs. 3 and 4, respectively.
Fig. 3 FEs mesh The FE simulation model was checked and verified by performing
(a) Two parallel circuit, (b) Three parallel circuit the following two tasks:
Table 2 summarises for both two and three parallel circuits, the (i) Conduct mesh sensitivity analysis by enlarging the FE mesh
associated performance indices at the nominal case including dimensions Wx and Wy and ensuring that the obtained cable
maximum cable attained temperature, backfill installation cost and temperature value is stable (does not change significantly). In
cable maximum current capability. The sequential ‘Case-#’ in the addition, the normal temperature gradient (°C/m) at the far X, Y
first column is shown to provide convenient comparisons in ends of the FE mesh is calculated to ensure that its value is
subsequent discussion. sufficiently small.
The conductor current is fixed at the nominal value of 403.4 A. (ii) Compare the obtained cable temperature value from FE
Also, the backfill cost in SR per cubic metre (last column of the simulation model with the results of the well known ANSYS
table) was shown for later comparisons. It is clear that the presence commercial package [39] for the same cable data.
of the hot-water drain has a significant impact on the cable
temperature. For the two-cable system, the 90°C drain has caused Table 3 summarises the FE simulation model verification
the temperature to increase by almost 16°C (from 74 to 90.0°C) results for the case of three cables with a drain at 90°C.
while the increase was 12.7°C for the three-cable system (from The PSO optimisation model, on the other hand, was easily
83.5 to 96.2°C). The FEs grid and the temperature profile (at 90°C verified by perturbing the optimal parameter values from PSO and
re-calculating the objective function at the perturbed parameter
values, which – as is expected – was degraded from its optimal optimal cable spacing, namely 0.5 (in case #B4) and 0.431 (in case
value (cable maximum temperature increased and ampacity #B2).
decreased from the PSO optimal values). (iv) Comparing case #B5 with case #B6, it is noted that the impact
of varying the cable depth when minimising maximum cable
3.4 Minimisation of maximum cable temperature temperature for the three-cable system with respect to cable
spacing is observed from the results of cases #B5 and #B6. The
Using the developed integrated PSO-FE approach, the cable system maximum cable temperature has dropped from 95 (in case #B5) to
is optimised to attain a minimum value of the maximum cable 90.8°C (in case #B6) as the cable depth has decreased from 0.8 to
temperature using the previously defined objective function and 0.6 m. The optimal value of cable spacing has changed slightly
constraint formulas (10a). The optimised parameters associated from 0.45 to 0.43 m.
with each configuration are summarised in Table 4. These optimal (v) Considering the results of case #B7, it is noted that when the
values satisfy the defined constraints while minimising the cost is of no concern, minimising the maximum cable temperature
maximum cable temperature. Again, the conductor current is fixed with respect to cable spacing, cable depth and the thermal
at the nominal value of 403.4 A. conductivity of lower backfill for the three-cable system would
The results of Table 4 reveal the following points of interest: cause the cable temperature to drop further to 88.7°C. In this case,
both cable depth and thermal conductivity of lower backfill have
(i) Comparing case #B1 with case #B2, it is noted that the presence been pushed to their minimum and maximum allowable values,
of hot-water pipe (in case #B2) as compared with case #B1 when respectively. In this regard, since the thermal conductivity is driven
minimising maximum cable temperature for the three-cable system to its maximum value, it allows more heat dissipation away from
with respect to cable spacing and cable depth, has caused the the lower backfill to the surrounding layers, and therefore reduces
maximum cable temperature to increase significantly from 82.9 (in the effect of the mutual heating between the three cables, while on
case #B1) to 94.1°C (in case #B2). The optimal value of cable the other hand, increases the heating effect of the drain and,
spacing has also changed slightly from 0.50 (in case #B1) and 0.43 consequently, drives the cables to be in more close proximity in
(in case #B2). order to avoid the drain effect.
(ii) Comparing case #B2 with case #B3, it is noted that when the It is to be noted that when ANSYS was used to calculate the
thermal conductivities of 0.67 W/(°C m) for the mother soil and optimal maximum temperature evaluated from the PSO-FE scheme
1.25 W/(°C m) for the lower backfill (in case #B2) are exchanged (case #B7), it yielded the value of 89.4°C. In this context, the
to 1.25 W/(°C m) for the mother soil and 0.67 W/(°C m) for the following observations are to be made.
lower backfill (in case #B3) while minimising maximum cable
(a) The difference between the maximum temperatures obtained for
temperature for the three-cable system with respect to cable
the two configurations (nominal and optimal) using MATLAB and
spacing and cable depth, the maximum cable temperature has
ANSYS is in the range of 1–4%. This is considered a strong proof
increased from 94.1 (in case #B2) to 96.1°C (in case #B3). This
of the validity of the results generated separately and independently
indicates that the conductivity of the medium immediately
by the two codes.
surrounding the cable has more effect on the cable temperature for
the thermal circuit configuration considered. (b) The very slight difference can be attributed to the different
mesh densities used in each code. The computed domain has been
(iii) Considering cases #B2 and #B4, it is of interest to note that
meshed with about 48,500 elements in ANSYS while only 14,000
when the ambient temperature has increased from 35 (in case #B2)
elements have been used to discretise the same domain in
to 55°C (in case #B4) while minimising maximum cable
MATLAB. The models have been built and meshed independently
temperature for the three-cable system with respect to cable
for validation reasons. The lower number of elements used in
spacing and cable depth, the maximum cable temperature has
MATLAB programme is driven by the need of time efficiency
dropped from 105.9 (in case #B4) to 94.1°C (in case #B2). The
during PSO optimisation.
difference is about 11.8°C, which is less than the difference in the
ambient temperature. This is due to the presence of the hot-water (c) It is to be mentioned that in the nominal case (case #A6), the
pipe in the vicinity of the cable coupled with different values of drain temperature was 90°C and the distance between the cables
was of 0.3 m. This created high-temperature gradients and
interaction of the field of temperature surrounding the two cables case #C4) to 100°C (in case #C2) has caused the thermal
located in the proximity of the drain. The temperature field at this conductivity of the lower backfill to decrease from 1.352 (in
location had also interacted with the induction by the drain. In case #C4) to 0.799 W/(°C m) with the corresponding reduction
these circumstances and considering the difference in element in cost from 7.01 to only 0.98 SR/m3. However, with the
numbers used in each code the 4% difference on the maximum reduction in thermal conductivity (in case #C2), the cable
temperature captured is justified. spacing had to increase slightly to 0.45 m (in case #C2) from
(d) In the optimal case, the drain temperature is only 80°C and the 0.41 (in case #C4) in order to allow for better heat
distance between cables is close to 0.4 m which reduced the dissemination around the cable.
temperature field interaction between the different cables and with iii. Comparing case #C3 with case #C4, it is noted that for the
the drain. For this lower gradients, the two models have given two-cable system, there was less need to increase the thermal
closer matching (1%) validating the modelling of the physics by conductivity of the lower backfill to meet the maximum
either software. temperature constraint than in the three-cable system. In this
(e) Although commercial software such as ANSYS have different regard, the thermal conductivity of the lower backfill for the
optimisation techniques most of which are based on differential two-cable system was 0.762 W/(°C m) as compared with
evolution algorithms, the PSO algorithm is not yet implemented 1.352 W/(°C m) for the three-cable system.
directly in ANSYS though it has its advantages in solving complex
problems with increased reliability when used alone or coupled 3.6 Maximisation of cable ampacity
with the differential evolution techniques available in commercial
software such as ANSYS. For this reason, different authors [40– In this case, the cable ampacity is maximised using the developed
42] have demonstrated the importance of using the PSO technique integrated PSO-FE approach with the previously defined objective
in combination with FEA through MATLAB or using the FEA function and constraint formulas (10c). The optimised parameters
software design language such as APDL for ANSYS. In the associated with each configuration are summarised in Table 6.
presented paper and for simplicity, ANSYS has been used These optimal values satisfy the defined constraints while
separately to validate the results of the FE model built in maximising the cable ampacity.
MATLAB. Once this is done, the latter has been used for both It is to be noted that the cable ampacity is a very complex
optimisation and FE computation for simplicity and efficiency as function of the thermal circuit variables and is related to the
the two methods are integrated into the same software preventing geometrical parameters via the differential equations governing the
data transfer and communication. In the future, both codes can be thermal behaviour of the cable system. However, the use of the
involved in different combinations of optimisation techniques. combined PSO-FE scheme will have no difficulty in dealing with
such complexity when conducting function evaluations (whether an
3.5 Minimisation of backfill cost objective function or a non-linear constraint) from within the PSO
calls to the FE simulation module.
Here, the cost of the backfill (size and material) is minimised using From the results of Table 6, we observe the following:
the developed integrated PSO-FE approach with the previously
defined objective function and constraint formulas (10b). The i. Comparing case #D3 with case #C3 (of Table 5), it is noted
optimised parameters associated with each configuration are that the objective of minimising cost [in case #C3 (of Table 5)]
summarised in Table 5. These optimal values satisfy the defined has focused mostly on reducing the backfill conductivity (the
constraints while minimising the cost of backfill. Again, the most costly item) producing a value of 0.762 W/(°C m) for the
conductor current is fixed at the nominal value of 403.4 A. backfill conductivity which was barely sufficient to keep the
From the results of Table 5, we note the following: maximum cable temperature within the constraint value of
95°C, while at the same time forced the thickness of the
i. Comparing cases #C1 and #C3 for the two-cable system, it is backfill to its minimum limit. On the other hand, the objective
noted that relaxing the maximum temperature constraint from of maximising ampacity (in case #D3) has allowed the backfill
90 (in case #C1) to 95°C (in case #C3) has caused the thermal conductivity to reach the value of 2.398 W/(°C m) (since the
conductivity of the lower backfill to decrease to 0.762 W/(°C cost was of no concern in this case) while – at the same time –
m) (in case #C3) from 1.265 W/(°C m) (in case #C1) with meeting the maximum cable temperature constraint value of
corresponding reduction in cost from 5.83 to only 0.69 SR/m3. 95°C. The optimal cable spacing value has also changed from
ii. Comparing case #C2 with case #C4, it is noted that, the effect 0.50 in case #C3 (of Table 5) to 0.35 m in case #D3.
of relaxing the maximum temperature constraint from 95 (in