2017 IEEE PES-IAS PowerAfrica
Fault Diagnosis in an Extra-High Voltage Power Line Oluleke Babayomi
Peter Oluseyi
Cent. for Space Transport & Prop. Nat. Space Res. & Dev. Agency Nigeria
[email protected]
Dept. of Elect. & Elect. Engineering University of Lagos Nigeria
[email protected]
Nkemdilim A. Ofodile
Godbless Keku
Nig. Air Force Res. & Dev. Cent. Nigerian Air Force Nigeria
[email protected]
Plan., Res. & Data Mgt. Services Nig. Mar. Admin & Safety Agency Nigeria
[email protected]
Abstract - The use of fuzzy and adaptive neuro-fuzzy technique for fault detection, classification and location is presented in this study. Ten different types of electrical faults in a transmission line were investigated. The results obtained show that a high degree of accuracy was recorded for detection, classification and location of electrical faults in the extra-high voltage line. Ongoing studies on the subject are expected to further improve levels of accuracy in fault classification and location especially. Keywords - three phase faults; adaptive neuro-fuzzy inference system; transmission line; fault classification, fault location, fuzzy-logic.
I. INTRODUCTION Due to the importance of accurate fault detection, classification and location in transmission lines, a vast amount of studies have been carried out on the subject. An accurate fault distance and direction estimation based on application of artificial neural networks for protection of doubly fed transmission lines is presented in this paper. [1] applies the measured voltage and current at the local end of the transmission line to accurately estimate fault distance and direction. More features about the method employed stand out, viz: it requires less than one and half cycles and adjusts automatically to fault location, fault resistance and inception angle. In [2], support vector regression method was applied to analyze post-fault voltage waveforms decomposed with wavelet packet decomposition within a half cycle after the fault incipience. Stationary wavelet transform was applied to extract decaying DC offset of current and voltage after noise was filtered out. This was done with samples taken over a quarter of a cycle after the fault occurred. Adaptive neuro-fuzzy inference system (ANFIS) is a technique that has been used extensively because of its abilities to learn from historical data. It was applied in [34] for fault location with relatively high accurate results. Artificial neural networks (ANN) methods have also been widely applied due to their self-learning abilities and high processing speed. Coefficients of current and voltage 978-1-5090-4746-8/17/$31.00 ©2017 IEEE
signals of a double circuit transmission line between 0500Hz were filtered out using digital wavelet transform and applied to train an ANN in [5] . The use of discrete Fourier transform to extract the fundamental components of current and voltage waveforms was reported in [6]. Subsequently, the ANN was used to train the ANN. Furthermore, stationary wavelet transforms and apace vector machines were also reportedly employed for fault location in [7-9]. When combined with the unique fault signatures extracted by time, frequency or wavelet transform methods, artificial intelligent fault location techniques including support vector machines, artificial neural network (ANN), neuro-fuzzy methods produce close results. However, the adaptive neuro-fuzzy inference system (ANFIS) has the following advantages when compared with other artificial intelligence algorithms. It gives a higher precision of accuracy than ANN. It also has a better optimized learning process than ANN and expert systems. Thus ANFIS was employed in this study. II. COMPUTATIONAL MODEL A. Fault Analysis Four different types of faults could occur randomly in a transmission network, namely: line-to-ground, double-line-to-ground, line-to-line and three phase faults. Fault analysis is done via the use of sequence components and equations (1) to (4) respectively give the fault current of all afore-mentioned fault types. Zf represents the fault ( ) ( ) ( ) the impedance at bus, while , , impedance, are the zero, positive and negative sequence fault currents in phase A respectively and is the per unit voltage at the fault point. The extra high voltage circuit analyzed is illustrated in Fig. 1. ( )
( )
=
( )
=
= ( )
(1)
( )
=
( )
( )
(2)
( )
311
2017 IEEE PES-IAS PowerAfrica
TABLE I: FUZZY INFERENCE RULES FOR 6-INPUT FAULT SYSTEM
Rule 1 2 3 4 5 6 7 8 9 10 11 to 64 20kV G1
A
Ia HI LW LW HI LW HI HI LW HI HI
C
330KV
T1
Ib Ic Va Vb LW LW LW HI HI LW HI LW LW HI HI HI HI LW HI HI HI HI HI HI LW HI HI HI HI LW LW LW HI HI HI LW LW HI LW HI HI HI LW LW ALL OTHER INPUT COMBINATIONS
330KV
T2
22.22kV G2
B
F1
F2
Figure 1: Extra high voltage power circuit.
M1 OUTPUT
X1 INPUT
M2
Y Neuron Addition
M3 X2 Fuzzy Rules Normalization Membership Defuzzification Functions (Fuzzification)
M4
Figure 2: Basic ANFIS model for a two-input-one-output system. ( )
= ( )
( )
=
( )
=
( )
( )
( )
( )
=−
=−
( )
+
( )
(3)
( )
( )
( ) ( )
( )
( )
( )
+
+3
( )
;
+3
( )
+3
+
( )
+3
;
Vc HI HI LW HI HI HI HI LW LW LW
Output 1 2 3 4 5 6 7 8 9 10 0
Fault Type LG fault - Phase A LG fault - Phase B LG fault - Phase C LL fault - Phases AB LL fault - Phases BC LL fault - Phases AC LLG fault - Phases AB LLG fault - Phases BC LLG fault - Phases AC 3-phase fault N/A
B. Fuzzy Inference System As opposed to binary logic that maps data binary numbers namely 0 and 1, fuzzy logic maps signals to values of 0 through 1 by employing three steps, viz: fuzzification, fuzzy processing and defuzzification. As shown in Fig. 2 two inputs X1 and X2 are mapped to from 0 through 1 by fuzzification. By the application of fuzzy rules, the inputs are combined to give normalized outputs. 1) Fuzzification: During the fuzzification process, input data is interpreted by the fuzzy controller. This is done with the aid of membership functions. 2) Membership Functions: Membership functions are charts that are selected by the user and are used to analyze the input data into the fuzzy system. Membership functions are defined according to shapes and the common shapes include S, Z, and triangle. The most important criterion that regulates the choice of a membership function is that it must vary between 0 and 1. If X is the universe of discourse and its elements are denoted by x, then a fuzzy set B in X is defined as a set of ordered pairs. (5) = , ( )| ∈ }. where ( ) is called the membership function of in . The membership function maps all the elements in to values ranging from 0 to 1 inclusive. These membership functions can be combined by a binary mapping T, which aggregates two membership functions as shown below: ( ), ( )) (6) . ( ) = ( A T-norm operator (a fuzzy intersection operator) is a binary mapping T(.,.) that satisfies the following conditions: • Boundary condition: T(0,0)=0,T(a,1)=T(1,a)=a • Monotonicity condition: T(a,b)
