‫ كلية الهندسة رقدالين‬- ‫جامعة صبراتة‬

Fuzzy Logic-Based
Relaying for Large Power
Transformer Protection
‫ أحمد الحاجي‬: ‫إعداد الطالب‬

2024
 INTRODUCTION
Power system protective relaying is designed to promptly remove abnormal conditions from the
power system. Since the introduction of microprocessors in the mid-1970s, digital protective
relaying has gained significant attention. One crucial element of the power system is the power
transformer, whose electrical protection is based on percentage differential relaying to
differentiate between transient magnetizing inrush and internal faults.
The harmonic-restrained differential relay relies on the large second harmonic component in the
magnetizing inrush current. However, this technique has limitations because harmonics can also
occur during normal operations, and improvements in core steel have reduced the second
harmonic component during inrush conditions. This results in difficulties in distinguishing
between faults and inrush conditions based on differential currents alone. Thus, a more reliable
relaying technique is needed to adapt to changing power system conditions.
Recently, new AI-based techniques for protective relaying have been developed. These
include algorithms based on artificial neural networks, fuzzy logic, and wavelet-based
methods. However, these methods can sometimes misoperate when magnetizing inrush
currents have a low second harmonic component, or when internal faults have a high second
harmonic component.
The paper focuses on a fuzzy logic-based relaying method for power transformer protection,
which improves fault discrimination between magnetizing inrush and internal faults. To
enhance the sensitivity of traditional percentage differential current relaying algorithms, fuzzy
logic is used. The input variables for the proposed fuzzy-based relaying include the flux-
differential current derivative curve, second harmonic restraint, and percentage differential
characteristic curve. The performance of this relaying method is evaluated using transformer
inrush currents, external fault currents, and internal fault signals, sampled at 720 Hz per cycle
from EMTP simulation software.
Flux-Differential Current Slope Method

This method, which uses the slope of the flux-differential current curve, is not affected by
remnant flux. This makes it more effective than previous flux-current methods. The flux-
differential current slope is calculated using the following formula:

Here, the subscripts ppp and sss represent the primary and secondary sides of the power
transformer, Δt\Delta tΔt is the sampling interval, IdI_dId​is the differential current, and
LlL_lLl​is the leakage inductance.
To differentiate between various transformer phenomena, the flux-differential currents are
computed and plotted against sample points. The results show that:
• In the steady state, the computed values approach zero.
• During transformer energization, the values fluctuate between 0 and around 15.
• Under internal fault conditions, the values remain around 15.
 Fuzzification
The fuzzification process in the fuzzy information approach quantifies the uncertainty of
input relaying signals and ensures that all data from these signals are acquired without loss.
The effectiveness of this quantification and data acquisition depends on the defined input and
output fuzzy sets.
The proposed fuzzy-based relaying uses three fuzzy inputs defined by specific formulas. These
inputs are derived from the analysis of flux-differential currents and are used to create
membership functions for each fuzzy set.
Membership Functions
Membership functions for the input and output fuzzy sets
• FI1: Composed of three fuzzy ranges:
• FI1_S (Small): Represents the steady state.
• FI1_M (Medium): Represents the energized state.
• FI1_L (Large): Represents internal faults.
• FI2_S (Small): Represents the steady state.
• FI2: Composed of two fuzzy ranges:
• FI2_L (Large): Represents internal faults.

• The range for fuzzification of the three input variables is set between 0 and 1, with
membership values ranging from 0 to 1 at intervals of 0.01. These membership functions
are crucial for accurately representing different operational states of the power
transformer, including steady state, energization, and internal faults.
• faults, respectively. FI2 consists of input fuzzy set FI2_S, FI2_L. FI3 consists of input fuzzy set
FI3_S, FI3_M, FI3_L. FI3_S, FI3_M, FI3_L support steady state, energized, and internal faults,
respectively. Fig. 2(d) shows that fuzzy output set is established in range from 1 to 9 at intervals
of 0.2. There is just one output fuzzy set for fault trip. Range of output fuzzy set is established
the defuzzified value to be around 3 under no faults and the defuzzified value to be around 6
under faults.
 B. Inference Method
• The proposed relaying scheme employs 19 fuzzy inference rules, each consisting of two
antecedents (inputs) and one consequent (output). These rules are categorized into three groups
based on the combinations of input variables. An example rule is:
• Rule: If input1 is X and input2 is Y, then FO_FUZZYSET is DEFINITELY FO_STEADY.
The compositional fuzzy inference method is used, specifically the Max-Min method, to perform
the necessary mathematical operations.
C. Defuzzification
The output of the fuzzy-based relaying is the FO_FUZZYSET. For defuzzification, the center
of area method is used, as defined by:

In this formula, each point on the domain of the final output fuzzy set is considered along
with its membership value.
System Modeling
The study used signals of transformer inrush currents, external fault currents, and internal fault
signals, sampled at 720 Hz per cycle, obtained from EMTP simulation. A 3-phase, 45/60 MVA,
154/23-kV transformer was simulated using the saturable transformer model and the BCTRAN
routine. Figure 4 illustrates the power system model selected for this study.
 Relay Performance Summary
• Fig. 5(a): Shows primary and secondary currents in steady state .
• Fig. 5(b): Illustrates the process of renewing the computed output fuzzy set. It includes the domain
of the output fuzzy set, sample point (defuzzification value), and the degree of output fuzzy set
(membership value). The defuzzified values are used to identify internal faults by comparing them
with a predefined threshold value. In this case, the defuzzified value remains constant, indicating
no trip command under trip conditions.
• Fig. 6(a): Depicts relaying signals when energizing at approximately 0.0444 seconds (32nd sample).
• Fig. 6(b): Shows the computed output fuzzy set using the proposed fuzzy relaying method.
.
.
 Test Results:
1. The fuzzy-based relaying algorithm effectively prevents relay trip
maloperation in cases of magnetizing inrush with a low second
harmonic component and internal faults with a high second
harmonic component.
2. The flexible nature of the fuzzy inputs and inference rules
allows the proposed relaying method to be applied to various
transformers and conditions.

Conclusions
This paper presents a digital protective relaying algorithm for
power transformers using fuzzy logic. The key aspects of the
proposed method include:
• Utilization of flux-differential current derivative curves, harmonic
restraints, and percentage differential characteristic curves to
address the limitations of conventional relaying methods.
• Evaluation of the algorithm was conducted using relaying signals
from EMTP simulations.