Design of Smart Cooling System for Power

Transformer
Waheed Aftab Khan Manzoor Ellahi Bilal Masood
FET, Superior Group of Colleges FET, Superior Group of Colleges FET, Superior Group of Colleges
Lahore, Pakistan. Lahore, Pakistan. Lahore, Pakistan.
[email protected] [email protected] [email protected]

Fuad Usman Ghulam Abbass,

FET, Superior Group of Colleges The University of Lahore, Pakistan.
[email protected] [email protected]

Abstract—The flexibility of a transformer enables it to have a inside the transformer. This emphasizes the use of techniques
wide range of applications. One of the most important to reduce the transformer’s temperature. With the purpose of
parameters that has an influence on the transformer insulation is enhancing the transformer life that gets shorter due to the
the temperature. Hence, the parameter to be controlled is the fan excessive heat accumulated inside the transformer. In electrical
speed and flow rate of oil (using pumps) that varies invariantly power transformer we use external transformer cooling system
with the load. The conventional control technique which is to accelerate the dissipation rate of heat of transformer. There
winding Temperature Indicator (WTI) fails to achieve this task. is different transformer cooling methods available for
The mechanism for Cooling of power transformer is designed on
transformer [2].
the process of measuring top-oil temperature along-with a
simulation used to find the point of highest temperature in the Different transformer cooling methods are:
winding. WTI has some short-comings and researchers now x Natural oil and air cooling system
prefer the presentation of fiber optic based temperature sensors
for this critical function. Fiber optic sensors have been selected x Natural oil and air forced cooling system
for the control of the cooling banks. The long-term performance x Forced oil and forced air cooling system
of transformer cooling can be better achieved with these more x Forced oil and forced water cooling system
accurate and modern fiber optic sensors. This paper presents a
mechanism to control the temperature of power transformer x Directed oil and forced air cooling system
using smart cooling system which includes Forced Oil (FO) WTIs come in several variations. The classical instrument
pumps and fans. A comparative analysis is also performed involves a bulb inserted in a thermos well surrounded by
between the WTI system and the proposed smart cooling system. insulating oil. In addition many modern techniques along-with
Flow control of these cooling devices is controlled and monitored conventional mechanisms are deployed at various points for
by the use of PID-controller via traditional Ziegler-Nichols thermal and insulation monitoring of transformer that are
algorithm and advanced Ziegler-Nichols algorithm organized discussed in [3-4] for any rise in temperature of winding.
with fresh developments in sensor technologies. This regulator Digital equipment has a feature of sharing the values of top oil
pushes the designed system to operate in flow and temperature temperature and load current to a microprocessor that perform
regimes which limit the problem of minimizing thermal calculations through a designed algorithm and the hot-spot
degradation of the insulation. winding temperature is found from the top oil temperature plus
an increment for winding hot spot rise [5]. Advance
Keywords—WTI; Transformer insulation; Comparative
terminologies have been developed that prefer testing process
analysis; Power Transformer
with bottom oil temperature, and the results obtained are more
I. INTRODUCTION precise. But, this technique is complicated as it requires use of
12 to 15 parameters that are determined accurately for each
Transformers are required throughout modern cooling stage [6]. The use of fiber optic sensors provide a swift
interconnected power systems. The in-service failure of a control approach by bypassing the uncertainties in calculations,
transformer is very hazardous to utility personnel through they also provide more authentic information about winding
explosions and fire. Mainly the failure of transformer is due to temperature under different loading values, with special
high rise in temperature which may cause excessive over consideration to the transformer temperature control. [7].
loading, copper losses, inefficient cooling system and improper
monitoring. To avert catastrophes and to preserve transformers The system proposed in the paper works on monitoring and
with healthy operating condition it is imperative to make controlling the temperature of power transformer by using
available better monitoring of temperature with smart cooling smart cooling system (i.e. oil pumps and fans). Flow control of
mechanism. In this way we can maintain the integrity of power these cooling devices is monitored by the use of PID-controller
system. With the efforts being done to make the implemented via traditional Ziegler-Nichols algorithm and advanced Ziegler-
power network more reliable, mechanisms of temperature Nichols algorithm together with recent advances in sensor
testing has gained importance in power system protection and technologies and comparative analysis is performed to find out
transformer engineering. Available theoretical techniques will which system is more efficient. This controller drives the
not suffice; hence the development of thermal monitoring cooling system to operate in flow and temperature regimes
technique has to be a solution in this direction [1]. which limit the problem of minimizing thermal degradation of
the insulation.
The temperature of the transformer keeps rising resulting in
the damage of paper insulation and liquid insulation deployed
The paper comprises of six sections. Section 1 consists of
introduction, literature review, and problem statement. Section Distribution
Pump 1 Pump 2 Pump 3
2 consists of design of the system to get cooled. Section 3 Pump 1 Pump 2 Pump 3
Cooling
Transformer
(Step-down),
Cooling
involves the methodology used that is followed to analyze Mechanism 2
PID Controller
Mechanism 2 11KV/ 440V

implementation of the PID controller for control purpose. Power Transmission Line
Power
Transf
Section 4 presents the comparative analysis of the Traditional Transform ormer Distribution Line
2
L
o
er 1
and Advance Ziegler-Nichols Algorithms for the different
AC Source
(Step- a
11 KV (Step-Up),
down), d
cooling equipment deployed at transformer. Section 5 11KV/ 132
KV
132KV
Fiber Optic Sensor / 11
describes the comparison between WTI & the proposed SCFO KV

cooling schemes. In the end, section 6 provides the conclusion Cooling

Mechanism 1
PID Controller Cooling Mechanism 1

followed by the references. Fan 1 Fan 1

Fan Fan Fan 2 Fan 3 Fan 4
Fan
2 3 4
II. POWER SYSTEM WITH TRANSFORMERS TO BE Fig. 1. Power system interconnection of different components.
COOLED
The power system dynamics are constituted through the The system works with a very simple pattern, it keeps on
interconnection of different components. Starting from monitoring the oil pressure and temperature of the system and
generator that gives electric power generation in the level of 11 as soon as the temperature rise goes beyond prescribed limits
kV to 15 kV that is stepped up to the adequate voltage level as of the system the fans are switched on with flow rate as per
per requirement of that transmission that is 132 kV, 220 kV or temperature rise via pattern of PID-controller. Same
500 kV and in certain areas of world even at 765 kV using mechanism is adopted for the case where problem in oil
large power transformers. This electric power is than pressure rises. These controlling mechanisms are provided with
transmitted to different load centers using transmission lines, the auxiliary transformer at the EHV end of the system to keep
the level at which voltage is stepped up depends upon the the heart of the power system in normal state to keep the
distance to which the load is to be supplied. This voltage is electricity running throughout the system. As in case of any
than again stepped down to the level as per requirement of the fault the first impact is of drastic rise in current flowing
load where it is distributed to different categories of load like; through the system so the transformer temperature also rises
domestic, commercial, industrial and agricultural. This supply accordingly.
is than utilized by the consumers according to the requirement As soon as temperature exceeds safe limit fan-1 starts
of the customer. working as per temperature value following the Ziegler-
The load variation or any abnormality in the working of the Nichols algorithm. Fig. 1 shows operational mechanism and
system directly impacts the heart of the power system i.e. sequence of cooling system (fans and oil pumps). If the system
transformer. Any fault in the system results in raising the still doesn’t achieve the safer limit of the temperature the fan-1
temperature of the transformer, especially in developing gets instructed to work at its full flow rate. When need of the
countries like Pakistan where the system is in a continuous fan-2 arises the PID-controller issues signal to the oil pump to
overload condition, the transformers gets heated up resulting in start the flow rate as per cooling requirement. This sequence
damage if appropriate action to control the temperature gets continues as long as temperature does not come down to the
delayed. acceptable temperature limit (i.e. switching on of fan-3 and oil
pump 2). The whole mechanism depends upon the instructions
Fig. 1 shows the initial portion of simulation where each issued through PID-controller as presented in Fig. 2.
block is assigned a specified purpose and a reference number is
used to call that particular function at a certain position or Motor
location where it is required to operate. These reference blocks Starter
are then connected to the reference buses where a complete set DC Motor
of reference devices is established. These complete reference Current
5 HP,
220V, 18
sets are then connected to the components of the proposed 10 A, 1225
Kohm Load RPM
system which are to work for the essential condition to keep
the system in normal condition. All the operation is carried out Controlled X
using traditional and/or advanced Ziegler-Nichols algorithm. Voltage Product
Both of these two methods are compared on the basis of output Source
reaction, minimum resolving time, and minimum overpass for + - Scope
speed demand solicitation of DC motor. Fig. 1 contains EHV, 1
HV, and distribution portion of the system.
PID Scope
1 + Controller 1
-
Power IN

Fig. 2. Simulink model of speed control of DC motor using PID controller.

 III. METHODOLOGY desired performance and the efficient compensator
PID tuning rules employed to control the controlling structure as of Smith predictor [9].
parameters are summarized in Table 1. 2. Internal Model Control (IMC) and Genetic
Programming (GP) are the classical techniques for
TABLE I. PID TUNING RULES
point-based tuning (PBT). On same grounds, ZN and
§ 1 · § Tdcs  1 · CC methods are the ones that provide the most
K cp ¨1  ¸¨ ¸
Contro © Tics ¹ © 0.1Tdcs  1 ¹ optimum load-based tuning (LBT) [10].
ller
Kcp Tic Tdc 3. The tuning mechanism adopted should be robust and
T
to achieve it, a compromise must be established
1.08433

0.65 § W ·
1.04432
W §W · between set-point tuning (SPT) method and LBT. For
IAE- 0.50814T ¨ ¸
¨ ¸ 0.9895  .09539 ©T ¹
setpoint k ©T ¹ T the purpose it was gained that the PID controllers
tuned by the SPT approaches are generally more
0.76167 1.05211
0.980890 § W · T §W · 0.89819
vigorous than those tuned by the LBT as shown in
¨ ¸ ¨ ¸ §W ·
IAEL k ©T ¹ 0.91032 © T ¹ 0.59974 ¨ ¸
©T ¹ [11].
0.80368 T
4. Another problem arises in the comparison of SPT and
1.12762 § W · 1.0081
LBT methods as mentioned in [11]. For the purpose
ITAE- ¨ ¸ W §W ·
k ©T ¹ 0.99783  .02860 0.42844T ¨ ¸
setpoint T ©T ¹ of balanced, desired outcome a compromise is
1.06401 0.70949 1.03826
required to make the system more reliable.
0.77902 § W · T §W · §W · 5. The SPT methods (except ITAE-set point) have
¨ ¸ ¨ ¸ 0.57137T ¨ ¸
ITEAL k ©T ¹ 1.14311 © T ¹ ©T ¹
reduced sturdiness, so they are improved than LBT
approaches for developments having huge interval
§ 1 ·
K P ¨1   Td s ¸ [12].
Contro © Ti s ¹
ller 6. Among LBT techniques, only ZN and ITAEL
KP Ti Td
methods are acceptable. ISEL, ISTEL, IAEL and CC
ISE- 1.048 § W ·
0.897 T
§W ·
0.888 techniques should be avoided as the PID controllers
setpoint ¨ ¸ 1.195  0.368W / T 0.489T ¨ ¸
tuned by them are too hostile. So the optimum
k ©T ¹ ©T ¹
0.970 0.753 0.948
terminology should be to deploy the scheme with
1.473 § W · T §W · §W ·
ISEL ¨ ¸ ¨ ¸ 0.550T ¨ ¸ small delay [13].
k ©T ¹ 1.115 © T ¹ ©T ¹
7. From SPT’s, IMC and GP present similar trends for
0.897 T 0.906
ISTE- 1.042 § W · §W · robustness, but the GP shows better internal gain
setpoint ¨ ¸ 0.987  0.238W / T 0.385T ¨ ¸
k ©T ¹ ©T ¹ giving it slight edge over IMC. On contrary, IAE-
ISTE- 1.468 § W ·
0.970
T §W ·
0.725
§W ·
0.939
SPT’s have advantage of being robust but only at the
¨ ¸ ¨ ¸ 0.443T ¨ ¸
setpoint k ©T ¹ 0.942 © T ¹ ©T ¹ sacrifice of their internal gains. The ISE-SPT’s
ZN 0.6Ku 0.5Tu 0.125Tu provide with better internal gains in comparison to
T § 16T  3W · W (32  6W / T 4W IMC and GP at an expense on robustness. ISTE-
CC ¨ ¸
kW © 12T ¹ 13  8W / T 11  2W / T SPT’s has smaller internal gains than IMC and GP
2T  W W TW with a minor improvement in robustness.
IMC T
2k O  W 2 2T  W 8. Similar to ZN process, the sturdiness of the PID
tan I  b controllers tuned by the ITAE-SPT worsens with rise
,
GP mKu cosI DTd 2Zc in delay. Hence, it is effective with smaller delay
b 4 / D  tan 2I values [14].
Considering the mentioned points, following clusters are
Some comments on the tuning rules are: designed:
1. Controllers tuned by IMC, GP, ISTE-SPT and IAE-
1. The core factor for the proposed model is normalized
SPT methods have small integral gains and small
delay (τ/T) that presents actual time delay for the
robustness measures.
system under inspection [8]. The average value for
2. Controllers tuned by ZN, ITAE-SPT, ITAEL and
τ/T is mostly less than 1. This helps in having a
ISE-SPT procedures have medium gains and medium
shorter time-delay. This also helps in achieving the
sturdiness values.
 3. Controllers tuned by CC, ISEL, ISTEL and IAEL
methods have large integral gains and large
robustness measures.
It is observed that the integral gains and robustness
measures given by the first group are too small, thus the
closed-loop systems are very robust but the load rejection
performance can be further improved. The PID settings in this
group will be referred as conservative.
Frequency response of ZN tuning technique was initially
implemented closed-loop tuning that was used for manual
operation. But, for the aim of better and controlled results the
parameter values were increased gradually, only as long as the
loop oscillated for a sustained period [15].
On comparison it was observed that the third cluster
presented higher values of internal gain and robustness,
resulting in closed-loop response to be oscillatory, thus not
achieving the desired robustness from this design. The tuning
values for the PID schemes (traditional & advance) in this
cluster will be denoted as hostile.
The second cluster attains appropriate robustness and gain
measures. The ZN process thus shows to have the best
conciliation between sturdiness and performance for the
process considered. Due to above given reasons Ziegler-
Nichols method is used for the implementation in this paper.
IV. COMPARATIVE ANALYSIS OF OPERATING
ALGORITHM
The following graphs show the comparative analysis
Fig. 3. Variation in RPMs of fans with time as temperature rises.
performed on the algorithm techniques used for the controlling
of the smart cooling mechanism designed for the protection of B. Advanced Ziegler-Nichols Algorithm for Fans
power transformers. These graphs are taken between the Fig. 4 depicts the variation in RPMs of fans with time as
system parameters (RPM v/s time & RPM v/s temperature) temperature rise takes place. The graphs show that in case of
using traditional Ziegler-Nichols and advanced Ziegler-Nichols advanced Ziegler-Nichols transient stability of output reaction,
algorithms. minimum resolving time, and minimum overpass for speed
A. Traditional Ziegler-Nichols Algorithm for Fans demand solicitation of DC motor takes lesser time than
traditional method.
Fig. 3 depicts the variation in RPMs of fans with time as
temperature rise takes place. The graphs show that in case of
traditional Ziegler-Nichols, transient stability of output
reaction, minimum resolving time, and minimum overpass for
speed demand solicitation of DC motor takes much time.
 Fig. 5. The variation in flow rate of oil pumps with time as temperature
rise takes place.

D. Advanced Ziegler-Nichols Pump Controlling Algorithum

The graph depicted in Fig. 6 provides the variation in rate
of flow for oil pumps with time as temperature rise takes place.
The graphs show that in case of advanced Ziegler-Nichols
transient stability of output reaction, minimum resolving time,
and minimum overpass for speed demand solicitation of DC
Fig. 4. The variation in RPMs of fans with time as temperature rises. motor takes lesser time than traditional method.
C. Traditional Ziegler-Nichols Algorithm for Pumps
Fig. 5 depicts the variation in flow rate of oil pumps with
time as temperature rise takes place. The graphs show that in
case of advanced Ziegler-Nichols transient stability of output
reaction, minimum resolving time, and minimum overpass for
speed demand solicitation of DC motor takes much time.
 Fig. 7. Comparison of hot spot temperature for WTI and SCFO control

Fig. 8. Comparison of WTI and SCFO control response for a step overload
of short duration.

VI. Conclusion
The paper presents a mechanism to control the temperature
of power transformer by using smart cooling system (i.e.
forced oil pumps and fans) and comparative analysis shows
that the performance of proposed smart cooling system is better
than the WTI system. Flow control of the cooling devices is
controlled and monitored by the use of PID-controller via
traditional Ziegler-Nichols algorithm and advanced Ziegler-
Nichols algorithm together with recent advances in sensor
technologies (fiber optic sensors) and the results show that the
Fig. 6. The variation in flow rate of oil pumps with time as temperature
rise takes place.
advanced Ziegler-Nichols algorithm performs better than the
traditional Ziegler-Nichols algorithm.
V. Comparative Analysis of WTI Control and Smart Cooling The results also show that the when SCFO is employed for
Forced Oil (SCFO) Control the temperature control instead of WTI. This brings second
Figs. 7 and 8 show-case comparison for cooling control at phase of temperature control much earlier in SCFO that helps
elevated temperature (110 degrees). It is evident that as SCFO in reducing the temperature that ultimately improves the
initiates the next phase of cooling much earlier than WTI transformer performance along-with its life as well. The results
system. Hence, SCFO can save transformer life when observed also shows that system compatibility is not a concern
implemented instead of using WTI. when it comes to utilizing the smart cooling mechanism for
controlling the temperature of the system under over heating
conditions. Thus, increasing the life of the transformer and
allied components of the power system.
REFERENCES
[1] M.J. Melfi, R.T. Hart: "Considerations for the use of AC Induction
Motors on Variable Frequency Controllers in High Performance
Applications" - IEEE Textile Film & Fiber Conference, May 1992.
[2] H.E. Jordan: “Energy Efficient Electric Motors and their Application” -
Van Nostrand Reinhold Company Inc., 1983.
[3] Oleg Roizman, Valery Davydov, “Temperature Rise Tests: Centre for
Power Transformer Monitoring, Diagnostics and Life Management
(transformerLIFE)”, IEEE Transformers Committee Meeting Miami FL,
21 April 2009.
[4] “Testing and diagnostics of power transformers”, by OMICRON, 2012,
https://www.omicronenergy.com/.../4E9D31D1-FBB6-46CD-8DF1-
938BA0496CB2/
[5] National Electrical Manufacturer's Association: “Motor and Generator
Standards, MGI, MG10, MG13”.
[6] J. N. Bérubé, B. L. Broweleit and J. Aubin, “Optimum Transformer
Cooling Control with Fiber Optic Temperature Sensors”, Neoptix, Inc.,
October 2009.
[7] D.P. Connors, D.A, R.H. Daugherty: “Considerations in Applying
Induction Motors with Solid State Adjustable Frequency Controllers" -
IEEE Transactions on Industry Applications, page: 815-823, 1970.
[8] “Guide for The Interpretation of Gases in Oil-Immersed Transformer” -
ANSI/IEEE, C57, Page: 104, 2008.
[9] ; “Guide for Acceptance & Maintenance of Insulating Oil in Equipment”
- ANSI/IEEE, C57, Page: 106, 2006.
[10] A. C. Franklin and D. P. Franklin: “The J&P Transformer Book” -
Butterworth Publishing, 11th Edition.
[11] “Gold Book Recommended Practice for Design of Reliable Industrial
and Commercial Power Systems” - ANSI/IEEE, Std: 493, 2007.
[12] “Standard Test Procedure for Thermal Evaluation of Oil Immersed
Distribution Transformers” - ANSI/IEEE, C57, and Page: 100, 2007.
[13] ISA (Instrument Society of America): “Installation, Operation and
Maintenance of Combustible Gas Detection Instruments” - ANSI/ISA-
S12.13, Part II, 1987.
[14] “Fire Hazard Properties of Flammable Liquids, Gases, and Volatile
Solids” - NFPA 325M, 1984.
[15] Michael G. Zabetakis: “Flammability Characteristics of Combustible
Gases and Vapors” - Bureau of Mines, Bulletin 627.