• No results found

Hybrid Methods: Proposed Schemes and their chal- lenges

LITERATURE REVIEW

2.4 Review of Fault Location Techniques in Ac- tive Distribution Systems

2.4.6 Hybrid Methods: Proposed Schemes and their chal- lenges

The hybrid methods for fault identification and location are adopted to develop a novel method which exploits the benefits provided by the incorporation of more than one fault location methods. The idea is to utilize the strengths of individual methods while restraining the shortcomings. Some techniques using hybrid methods for fault location in distribution network have been proposed till now [96–105]. These methods generally use artificial intelligence techniques along with advanced signal processing techniques to extract fault information from the recorded signals and locate fault. Work done in [96] proposes a method combining wavelet theory and

fuzzy logic for islanding detection and fault location in distribution systems with distributed generations, the method claims to be very fast and detects fault within 10 ms from its inception. A. Rafinia et. al., [97] presents a method combining artificial neural network (ANN) and fuzzy logic for fault location in three phase underground distribution system, wavelet is used for feature extraction from the simulated data. The method is tested under various conditions like load shedding, load increase and load unbalance, also the method is tested against the effect of DG addition to the network. Hybrid methods combining ANN and support vector mechanism (SVM) are proposed in [98, 99]. Authors in [98] use ANN and SVM for faulted section identification and fault location respectively. Feature extraction is done using measurements obtained at substation, relay status and at circuit breaker.

The data is analyzed using principal component analysis technique. This method robustness is tested under verity of conditions but the distribution system used does not contain any DG source. Whereas, the method proposed in [99] use ANN and SVM application for fault location in power transmission lines in presence of non- linear loads. In [100] a method based on wavelet transform and SVM is presented for locating fault in micro grid. In this the search space for faulted section identification is reduced by detecting the DG with the lowest harmonics and finally the faulted segment is determined using the wavelet and optimized multi-class support vector mechanism (OMSVM). Another method combining WT and SVM proposed in [101]

is used for fault location in distribution network in case of high impedance faults.

In [102] a hybrid method combining SVM and smart meters are proposed. Smart meters are used for providing data for training of SVM which are used for fault location purpose. A faulted section identification followed by fault location for distribution network with multiple branches based on WT and ANN is proposed in [103] in this paper, WT is used for faulted section identification and ANN is used for fault location. Fault location method using impedance based method and voltage sag matching algorithm in [104] is applicable for single phase faults only and the method proposed in [105] locates fault in an active distribution network using polarities of current transients. Table 2.5 presents comparative analysis on hybrid

methods.

Table 2.5: Comparative analysis of hybrid methods

Reference Network Model Fault Type Type of Diagnosis

M. Dehghani et al., [96]

20 kV 5 km pi section distribution line

Shunt faults

Fault location in distribution systems with DG based on combination of wavelet singular entropy and fuzzy logic

A. Rafinia et al., [97]

20 kV underground power distribution system

Shunt faults

Combination of ANN and FLS to classify and locate fault

D. Thukaram et al., [98]

52 bus distribution system

Shunt faults

Combination of ANN and SVM forfault location and classification

E. Koley et al., [99]

400 kV three phase double circuit transmission line

Shunt faults

(SVM), ANN

and Kalman filter based algorithm

S. Hamid Mortazavi et al., [101]

20-kV radial distribution feeder

High impedance fault

Stationary wavelet transform and SVM based fault location M. Daisy

et al., [104]

IEEE 34 node system

Single phase faults

Impedance based method combinedwith voltage sag matching

Amila Pathirana et al., [105]

230 kV transmission system and IEEE 34 node system

Not specified

Rogowski coil sensors based transient protection scheme

Some limitations associated with application of hybrid method in distribution system are as follows

• These methods generally use some AI technique therefore reliable modeling of the system is required to generate data for AI training.

• Application to distribution system is difficult because of network complexity, different conductor size and multiple branches.

• When two methods are combined decision regarding which relay output to be used for fault location requires careful decision making process.

Table 2.6: Comparative analysis of different methods

Techniques Merits Demerits

Adaptive methods

Provides correct settings for different operating conditions

Communication system requirements which increases system cost

Complex relay setting calculation in case of high DG penetration

Retains the original protection setup hence economical

Requirement of modification in FL algorithm following an alteration in the system

Impedance based

Communication channels not required in one ended technique and simple utilization into digital protection relay

Multiple fault positions in multilateral distribution network

More accurate results can be obtained with two ended method with communication channel

Assumes homogenous network, error due to system loading and fault impedance

Travelling wave based

Very fast and accurate Needs high frequency data acquisition Independent of many

system parameters

Requires very precise common time reference

Distributed device based

Provides continuous

monitoring of system Requires large investment Distributed architecture

of protection

Requirement of communication between the devices

Intelligent techniques

Possess characteristics such as fast learning rate, fault tolerant

Slow convergence, amount and quality of data provided for training

Able to deliver accurate output with partial input

Retraining of algorithm in case of system change

Hybrid methods

Exploits benefit of two or more methods

Requires careful decision making for relay output to be used

Reliable models to create training data