Faults Signature Extraction in Wind Farm Integrated Transmission Line Topology

Received Jun 1, 2018 Revised Jul 10, 2018 Accepted Jul 25, 2018 The integration of Renewable Green Energy Sources (RGES) like Wind Farm Generators (WFG), and Photo Voltaic (PV) systems into convention power system as a future solution to the increase in global energy demands, generation cost reduction, and limited climate impact. The innovation introduced protection compromise challenges in power system due to infeeds fault current penetration from RGES on existing system, leading to an undesired trip of the healthy section of TL, equipment damages, and safety failure. A comparison study of extracted faults signature from two proposed Transmission Line (TL) network topologies with and without WFG integration, for onward fault identification, and classification model design. Descrete wavelet multiresolution Analysis (DWMRA) of extracted one-cycle fault signal signatures from 11 faults type’s scenarios in Matlab. Result demonstrated a unique fault signatures across all simulated faults scenarios harness for future work of an adaptive unit protection model for this new area of DG integration.


INTRODUCTION
The high cost of fossil fuel and global climate changes impact on nonrenewable energy (NRE) sources encouraging the integration of Renewable Green Energy Sources (RGES) on existing power system, considering their environmental friendly, elimination of waste management problem and cost-effectiveness in meeting the ever-increasing electrical energy demands as specified by the Global Wind Energy Council (GWEC) [1]. The integration of Wind Farm Generation (WFG) or Photo Voltaic (PV) RGES sources on existing Transmission Line (TL) network compromises existing protection scheme, leading to wrong trips of the healthy section of the lines and undesired damages to equipment and personnel safety [2]. The increase in WFG sources integrations on existing transmission system have been implemented in countries like the U.S with over 42% installed capacity, Europe, China, and Egypt has installed capacity of 36%. The penetrations introduced some system challenges in term of reliability, security, stability and power quality compromised [3]. The need to provide a more robust protection scheme that is adaptive in nature, towards meeting the current trend of RGES integration as new paradigm is not only necessary but eminent considering the limitations of both existing unit differentia and pilot protection for total TL protection coverage with high cost of communication channels [4]. Also, the non-unit distance relay protection with partial line protection coverage limitation [5]. Earlier TL protection schemes adopted phasor measurement unit and network system topology information for the protection scheme development for fault identification [6]. An adaptive backup scheme proposed for faulted line and location determination, using limited numbers of the phasor line signals measurements from backup protection zone [7]. Oscillogram data adopted for protection scheme development in high impedance fault detection, using mathematical model and power line communication approach for improved networks safety [8]. Nevertheless, the high computational rigour of these approaches necessitated an improved soft computational intelligent (SCI) approaches for network protection analyses with using advanced digital signal processing (DSP) in combination with Support Vector Machine (SVM) algorithm for ease of computation and accuracy. The cross-country earth fault identification on different TL phases in same circuit using one end signal statistical deviation data for the training of an artificial neuro network (ANN) detection model [9]. In addition, DSP and descrete wavelet transform (DWT) analysis of fault signals waveforms records at monitoring location of a multi-bus meshed network for useful feature information gathering adopted to detect and classify faults [10]. Protection fault analysis scheme on a series and shunt compensated TL on this network topology using half-cycle post-fault current signal for ANN training [11][12]. Other hybrid network topology protection scheme approach for fault identification, classification, and location in combined overhead TL and underground cables network topology using hybrid ANN-Fuzzy logic [13]. The robustness could not be guaranteed due to few numbers of faults scenario simulated in these research work. These reviewed literatures have not presented much works in the new frontier of RGES integration on TL protection scheme development as one challenge and limited to few scenerios studies as motivating factors for this research.
These necessitated the need for the design of a high-speed, low cost and reliable unit protection scheme model development for an integrated WFG-TL protection scheme with detail comparative assessment study of Discrete Wavelet Multiresolution Analysis (DWMRA) of one-cycle during-faults signals (voltage and current) signatures from two TL network topologies with and without WFG integration as case study.
The article organized with the introduction section illustrating the caparative advantages of the RGES against the traditionally existing energy generation sources with respect to cost, environmental assessments effects and negative impacts on existing protections scheme effectiveness. Methodology section divulges the proposed soft computational approach with the application of DWMRA. This is followed by the result and discussion section and finally, the implication of the result expressed in the Conclusion section.

RESEARCH METHOD
The CI approach for developing an improve protection classifier model that could be adopted for fault types identification and classificiation in TL protection relay. This will help in preventing and eliminating faults as fast as possible with high precision, selectivity, and reliability

Propose Unit Protection Classifier Model
This research proposed a Matlab Simulink model of a two-end power generation sources of 132kV, 50Hz, 200 km TL with integrated 9 MW RGES-WFG having six units of 1.5 MW on the common system bus of Figure 1    The signal is preprocessed by passing them through two set cascaded bands filters. The displayed MRA filter architecture of Figure 2 is made up of low pass filter h (k) (LPF) and high pass filters g (k) (HPF) for signal decomposition using DWMRA of extractes one-cycle noisy fault transient signals of voltage and current signatures from two propose networks topologies (with, and without integrated RGES-WFG). The LPF is realised by the scaling function (ɸ) of Equetion 1, the HPF is actualized with the application of the mother wavelet function (Ψ) of Equation 2.
Where n is the integer representing the number of samples, k is the translation integer and 2 is normalization factor.

DWMRA Application On Fault Signals
WT are small waveforms existing for a short time duration with an average value of zero and mostly adopted in advanced DSP of transient signals study. This is applied in MRA of transient faults signal into translated and scaled components of the applied mother wavelet function studies.The time-frequency localization benefits for little signal disturbances on TL along with the unique ability to extract, and analyze signal signature into various frequencies bands, while retaining the time function information as one of its unique benefits in fault transient studies.The LPF produced the approximate component Aj, while the HPF produces the details Dj component of decomposed signal.These extracted components are adopted for different fault identification and classification scheme of the protective scheme for both proposed TL network topologies. The approximate and detail coefficients relationship between two adjacent levels can be express mathematically by Equation 3 and 4.
Where j is the frequency band level. The first stage of the MRA is resolved into two halves of frequency component, while the second stage adopts the LPF output for decomposition into further two halves using another set of LPF and HPF. Subsequent stages further decomposed iteratively till 10th levels of Aj, is attaned using Daubechies (db4) wavelets as the mother wavelet to obtain the fundamental frequency of 50Hz based on a sampling frequency of 50 kHz.

RESULTS AND ANALYSIS
The decomposed approximate components of both extracted fault signals at every level of decompositions contain unique signatures based on the signals Standard deviation (STD) and entropy energy content that could be adopted for faults identification, and classification model development. The statistical analysis at 10th level min, max, and STD of the analyzed signal for discriminant application from proposed

Ground Fault
Present MRA result of Table 1-10 displayed unique values from both extracted faults signals from proposed network for all possible ground faults based on their min, max, and STD. these uniquesness were harness in the development of ground fault identification model building. The entropy energy and STD values from both signals shows an increase in the current entropy values for SLG faults with RGES-WFG integration as compared to that obtained without integration indicating and infeeding effect from the RGES in Table 1. This is supported by the higher STD values, which reduces with increase in fault distance for both network. Although the STD value is higher for current signal from WFG integrated topology when compared with other topology result in Table 2-4. The decomposed voltage entropy increases, while current the signal entropy reduces with increase in fault distance. Both signals energy contents from integrated WFG network topology are higher compared network without as observe in Table 2 Table 5. The three-phase to Ground (ABCG) fault displayed a lower values of STD and entropy energy content for both signasl in Table 6. Furthermore, the entropy energy for DLG faults increases with increase in fault distance for voltage signals but reduces for current signals with increase in fault distance as displayed in Table 7 (a) and 7 (b) respectively. The entropy energy contents obtained from integrated WFG network topology across all DGL is far higher when compared with another network.

Non-Ground Faults
The line to line faults result across both networks topologies indicated an increase in the voltage STD values but decreases for current as the fault distance increses for both proposed network topology in Table 8-11. The STD and entropy values from integrated WFG network topology is lower compared to architecture without integrated WFG. The STD and entropy values remains constant for both ABC and ABCG of Table 11. And Table 7 (a) and (b) respectively.

CONCLUSION
The individual uniques STD, and entropy energy signature values from the DWMRA of the fault signals across all 11 fault types will be adopted for the fault identification and classifier model building for an adaptive relay protection model building for an improved adaptive system model realization as proposed. The displayed result have help to substantiate the existing uniques in the protection impedances as a result of infeeds contribution into the faulted point from the integrated RGES which compromises the safety system when compared to same network topology without any integrations as demonstrated from the results. The discriminant features exixting between the ground fault and non ground fault analysis can also be adopted to build a fault type classifier model that could distinguish between gonrd and non-ground fault intelligent decision model.This research paper result have demonstrated unique fault signatures across all simulated faults scenarios that will be harness in future work for onward for the building of an adaptive intelligent unit protection relay model in this new area of DG integration protection on TL.The impact assessment study of the infeed effect of integrated RGES-WFG on TL established. Future progress will be achieved by integrating this unique extracted approximate coefficient for an onward building of an adaptive intelligent relay system.