Transformer Differential Protection Based on Wavelet and Neural Network - Report Example

Transformer Differential Protection Based on Wavelet and Neural Network Student’s Name Student ID Date Supervisor Abstract In power systems, transformer protection is crucial and is dependent on precise and quick discrimination of internal fault current from magnetizing inrush current. Previously, differential protection has been the most preferred technique of protecting transformers but with its tendency to falsify faults, this technique has been coupled with wavelet transform and artificial neural networks to reinforce classical protection principles while facilitating faster, secure and reliable differential protection for power transformer. The role of the wavelet transform is to extract components of transient signal while neural network will be trained to extract the components of the transient signal and precisely differentiate between internal faults and inrush current. Three phase transformer rating 315 MVA, 220/400kV, and 50Hz is modelled in PSCAD/EMTDC software. For the resulting algorithm, simulation is done using MATLAB. The results will indicate that the conventional differential protection was less accurate, less reliable and slower compared to the proposed differential transformer. 1. Introduction Transformers are classified as stationary electrical devices. They are used in electrical power systems to transmit power-linking circuits using electromagnetic induction or ordinary magnetic field. The functioning of transformers occurs when alternating current runs through a conductor and creates a magnetic field around itself. By introducing a second conductor in the resulting field influx lines connect the subsequent conductor. In this case, the latter conductor also has voltage induced into it through the principle of transformers. There are different forms of transformers all of which operate under the principle of transformers despite their differences in sizes and their design for use in different electric and electronic applications. Some types of power systems transformers include three or two winding electrical power, earthing transformers, regulating transformers, and rectifier transformers. For all these transformers, protection is crucial as a way of controlling or avoiding damage arising from abnormal voltages/currents. Transformer protection scheme ensures that the overcurrent and overvoltage withstand limits are not surpassed. Different transformer protection schemes exist for different transformers. However, the choice of protection scheme depends on factors such as normal service condition, type of transformer faults, tap changing scheme, and the extent of present overload among others. According to Rockerfeller [Roc07], the common influencers of the kind of transformer scheme to use are transformer’s MVA and primary kV as well as issues regarding personal safety issues and the extent of risk reduction linked to the chosen scheme. In addition, the capability of a given protection scheme to reduce the occurrence of a given catastrophe, and the general economic impact of transformer fault and the measures needed to combat such risk like financial impact of repair and maintenance, time needed for repair, and accessibility to backup power. Given the costs associated to dealing with catastrophes or incidents of transformer fault, this paper proposes proactive differential transformer protection as a means of promoting personal safety, reducing dangers to nearby structures, and as a mean of reducing costs incurred as a result of transformer failure. 1 Background Traditionally, power transformers have been protected using differential protection techniques. These schemes use current and voltage to sense any abnormalities in the differential zone. Consequently, initiating a trip required there to be a short circuit or high level current. However, when transformers required overloading for emergency stipulations mitigation, these protection schemes were not effective. However, in order to overcome this drawback, transformers needed adjustments on their thermal ratings such that they exceed standard requirements but within safe operations. 1.1. Motivation The motivation of proposing transformer protection is the critical nature of the balance between transformer protection scheme and the costs of arising from the consequences of depending on other forms protection or forfeiting the transformer. The engineer has to understand the different forms of transformer failure classifications and focus on proactively implementing transformer protection. Winding to core transformer faults result from weakened insulation resulting to short-circuiting. These faults occur on transformer terminals[Pai10]. On-load tap changing or OLTC gear is also type of transformer fault. In addition, it is common for transformers to suffer from enormous inrush currents rich in harmonic content during switching particularly if transformer is unloaded. In addition, transformers experience hot spots within windings due to inter turn faults, or experience over fluxing because of operating at rated voltage. Another transformer fault is oil leakage due to transformer immersion into oil. There are numerous forms of transformer protection like protection against overheating and over fluxing, over current protection, restricted earth fault protection, and voltage differential protection. These protection measures are crucial especially in ensuring that any fault condition is cleared within the fastest possible time to facilitate harmonization between shielding devices upstream and downstream the apparatus being protected. The implication here is that despite there being numerous protecting devices, any occurring fault detection has to include only one must detect the problem. The fault detection devices may be relays or fuses[Sch131]. Relays react on the circuit-breaker coil indirectly while fuses offer direct faulty circuit clearance. Additionally, fuses work in conjunction with mechanical tripping attachment to open linked three-phase load-break switch. 1.2. Objectives The research focuses on transformer protection particularly dealing with transient inrush current when the generator is energized. The inrush is current is uniquely characterized by a huge second harmonic and may as well occur during transformer faults that most digital differential protection transformers do not effectively contain given that it is expressed as a ration of fundamental component of differential current. Consequently, mal-operation occurs because of to second harmonic component of inrush current not being distinguished from internal faults. Objective To investigate the role of d1 level wavelet signal co-efficient as an input to artificial neural network that establishes a novel approach to online detection method that discriminates magnetizing inrush current and inter-turn fault, and fault location. 1.3. Significance Although differential relay has been used as the principal protection of power systems, its drawback in distinguishing inrush and faults has led to the proposition of numerous methods to analyse and recognize internal faults and inrush current[Par11]. In order to overcome the difficulty of distinguishing inrush currents and internal faults, artificial neural network offers a solution. Bakshi & Bakshi [Bak10] reveals differential protection as based on the occurrence of differences between current entering and current leaving electrical equipment because of the presence of a fault. The differential protection method is attractive especially since the in and out currents can be compared in terms of their magnitude or in phase or both[Pai10]. In case the difference exceeds the predefined set value, a trip output is delivered. Additionally, differential protection is founded on the assumption that both ends are located close to each like in the case of a busbar, or generator while the distance between ends in transmission lines is very close making it hard to apply differential relaying directly. However, with continuous power transformer monitoring, the likelihood of reducing catastrophic destruction and obtaining early electric failure warning is improved. This way, there is a reduction of damages while providing uninterrupted power supply. Consequently, the protective relays on the power transformer experience high expectations. These expectations include security or no false tripping, dependability or missing operations, and operation speed or short fault clearing time. With huge alterations in input terminal transformer voltage, huge current is drawn from supply by the transformer to the extent of about 10 times that of full load current[Dom69]. Such current can result from recovery from external fault, or switching-in. Such current persists temporarily prior to rapid decay and since it has very high magnitude, the result is that it causes the relay to operate falsely causing magnetic inrush. Inrush is perceived as an internal fault current by the differential relay hence ends up as a spill current that causes the relay to mal-operate[Ahm13]. The distinction between internal faults from inrush current is very challenging and this proposal focuses on the use of a new algorithm to discriminate between inter-turn fault and magnetizing inrush current[Wat03]. 2. Proposed Approach This research proposes using Artificial Neural Network (ANN) approach for transformer protection. ANN approach can detect ordinary, magnetizing inrush and over excitation currents. The ANN concept in the detections is found on wave shape recognition, which is possible through differentiating false differential currents from internal fault current wave shapes[Mor00]. The result of the detection process is practicing restraint und normal conditions of inrush and over excitation and a trip signal just for the internal fault conditions. False differential currents are categorized as inrush currents, over excitation conditions, and current transformer saturation. Inrush currents causes include external fault occurrence, voltage recovery, synchronization out-of-phase, and alterations in character fault. The main elements of inrush currents dc offset, even and odd harmonics; unipolar and bipolar harmonics; and initially low value second harmonics that increases with decrease in inrush current. Overexcitement currents results from the separation f unit-connected generation plant is isolated during VARs export. Over excitation exhibits itself in odd harmonics production. However, the fifth harmonic is preferred for restraining given that the third harmonic or any other triple cancels in transformer windings in order to distinguish between faulty state and over excitation[Ozg05]. Wavelet transform According to Okan, Onbilgin, & Cagrikocaman [Oka04], wavelet transform is fundamental and powerful tool used in the analysis of transient occurrence in power transformers. This is because wavelet transform has the capacity to obtain information from the transient simultaneously in both the time and frequency domain compared the Fourier transform that does give wave information only in the frequency domain. The discrete wavelet transform (DWT) is useful for signal processing, de-noising, compression, and estimation among others. There are numerous forms of wavelet expansion functions or mother wavelets and the choice depends on the application. ANN approach In order to implement ANN, formulation of the problem is done by identifying input and output set where the inputs are the samples of differential current comprising of two currents namely current entering and current leaving the transformer, and 8 consecutive samples used to represent each thereby summing into 16 inputs. The output will be four varying instances in power transformer: over excitation, magnetising inrush, normal and internal fault conditions[Mor00]. The next step after input and output definition is incorporation of hidden layers in the network. The figures below indicate expected waveforms forms for normal operation, magnetizing inrush, and internal fault (see Fig 1). a. Differential current waveform for normal operation b. Differential current waveform for internal fault c. Differential current waveform for magnetizing inrush Figure 1: Diagrams a, b and c indicates anticipated differential current waveforms Implementing combined wavelet and neural networks For the neural network, the input data set is the pattern waveform of wavelet energy. First, the wavelet transform is applied to decompose the transformer’s differential current. The result is a series of detailed wavelet elements whose spectral energies are calculated. The input signal comprises of both high and low frequency components using high pass h(k) and low pass g(k) filters. The spectral energies are then fed into OFFBNN or Optimal Feed Forward Back Propagation Neural Network. The power transformer to use will be 315 MVA 220/400kV, 50Hz and modelled in PSCAD/EMTDC software and MATLAB used to evaluate the resulting algorithm[Tri14]. The functional diagram of from input to display is shown in the diagram below (see Fig 2) Figure 2: Functional diagram of Alternative Transient Program-Electromagnetic T Transient Program package Source: [Nay13] 3. Timeline The preparation of the entire research is anticipated to take 8-10 weeks. Below is the timeline showing activities to be completed. Task/Week 1 2 3 4 5 6 7 8 Topic identification X Research proposal draft X Research proposal final draft X Supervisor feedback and approval X Dissertation introduction X X Dissertation Literature review X X Dissertation methodology X X X Dissertation discussion and conclusion X X X Figure 3: timeline for research paper preparation 4. Risk Assessment During transformer operations, the differential current relays are affected by transformer tap changer operation resulting into falsified response causing mal-operation due to On Load Tap Changes spill current[Kho07]. One way of maintaining security and avoiding unnecessary operations by during differential protection, the research proposes an increment of the relay’s least sensitivity level to a value above 20% of the rated current. Other risks assessment techniques involve regular commissioning and inspection to ensure that there are not overheating, abnormal vibrations, oil leaks, malfunction, and noise. In addition, the devices used to inspect the transformers must operate well and be calibrated properly[Fac14]. In addition, the transformer should not be overloaded while any need for overloading is implemented after proper evaluation of life loss, environmental, and bubbling risks linked with overloading. Additionally, overloading should utilize dynamic transformer control system and monitoring. In addition, the transformer should not have on-line insulation fluid processing done routinely and without evaluating the involved consequences. 5. Progress to Date So far, the research proposal draft and research proposal are completed. Task/Week 1 2 Proposal draft submission X Final proposal writing X X 6. Conclusion The proposal presents a power transformer protection scheme using a combination of wavelet transform and artificial Neural Networks for faster and stable results. The wavelet transform is used for feature extraction from the differential relaying signal given its inherent capacity property of time frequency localization employed to obtain discriminating features from the differential current. DFT wavelet ensures better extraction of fundamental and other components of harmonics, while using the second harmonic to restrain and the fifth harmonic to block through differential protection to distinguish between normal and faulty state power transformer. 7. References Roc07: , (2007), Pai10: , (Paithankar & Bhide, 2010), Sch131: , (Schenider Electric, 2013), Par11: , (Paraskar & Dhole, 2011), Bak10: , (2010), Dom69: , (Dommel, 1969), Ahm13: , (Ahmadipour, 2013), Wat03: , (Watson & Arrillaga, 2003), Mor00: , (Moravej, Vishwakarma, & Singh, 2000), Ozg05: , (Ozgonenel & Akuner, 2005), Oka04: , (2004), Tri14: , (Tripathy, Prakash, & Nirala, 2014), Nay13: , (Nayir, 2013), Kho07: , (Khorashadi-Zadeh & Li, 2007), Fac14: , (Factory Manual Global, 2014), Read More

Some types of power systems transformers include three or two winding electrical power, earthing transformers, regulating transformers, and rectifier transformers. For all these transformers, protection is crucial as a way of controlling or avoiding damage arising from abnormal voltages/currents. Transformer protection scheme ensures that the overcurrent and overvoltage withstand limits are not surpassed. Different transformer protection schemes exist for different transformers. However, the choice of protection scheme depends on factors such as normal service condition, type of transformer faults, tap changing scheme, and the extent of present overload among others.

According to Rockerfeller [Roc07], the common influencers of the kind of transformer scheme to use are transformer’s MVA and primary kV as well as issues regarding personal safety issues and the extent of risk reduction linked to the chosen scheme. In addition, the capability of a given protection scheme to reduce the occurrence of a given catastrophe, and the general economic impact of transformer fault and the measures needed to combat such risk like financial impact of repair and maintenance, time needed for repair, and accessibility to backup power.

Given the costs associated to dealing with catastrophes or incidents of transformer fault, this paper proposes proactive differential transformer protection as a means of promoting personal safety, reducing dangers to nearby structures, and as a mean of reducing costs incurred as a result of transformer failure. 1 Background Traditionally, power transformers have been protected using differential protection techniques. These schemes use current and voltage to sense any abnormalities in the differential zone.

Consequently, initiating a trip required there to be a short circuit or high level current. However, when transformers required overloading for emergency stipulations mitigation, these protection schemes were not effective. However, in order to overcome this drawback, transformers needed adjustments on their thermal ratings such that they exceed standard requirements but within safe operations. 1.1. Motivation The motivation of proposing transformer protection is the critical nature of the balance between transformer protection scheme and the costs of arising from the consequences of depending on other forms protection or forfeiting the transformer.

The engineer has to understand the different forms of transformer failure classifications and focus on proactively implementing transformer protection. Winding to core transformer faults result from weakened insulation resulting to short-circuiting. These faults occur on transformer terminals[Pai10]. On-load tap changing or OLTC gear is also type of transformer fault. In addition, it is common for transformers to suffer from enormous inrush currents rich in harmonic content during switching particularly if transformer is unloaded.

In addition, transformers experience hot spots within windings due to inter turn faults, or experience over fluxing because of operating at rated voltage. Another transformer fault is oil leakage due to transformer immersion into oil. There are numerous forms of transformer protection like protection against overheating and over fluxing, over current protection, restricted earth fault protection, and voltage differential protection. These protection measures are crucial especially in ensuring that any fault condition is cleared within the fastest possible time to facilitate harmonization between shielding devices upstream and downstream the apparatus being protected.

The implication here is that despite there being numerous protecting devices, any occurring fault detection has to include only one must detect the problem. The fault detection devices may be relays or fuses[Sch131]. Relays react on the circuit-breaker coil indirectly while fuses offer direct faulty circuit clearance. Additionally, fuses work in conjunction with mechanical tripping attachment to open linked three-phase load-break switch. 1.2.

Read More