Voltage Variation Signals Source Identification System By Time-Frequency Analysis Technique
The quality of power supplies has become one of the issues concern of power supply utilities and users these days. Power quality (PQ) can bring severe problems such as processing interruption within the industries, malfunction and downtime of the equipment and impact in economic losses. Voltage vari...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English English |
| Published: |
2019
|
| Subjects: | |
| Online Access: | http://eprints.utem.edu.my/id/eprint/24695/1/Voltage%20Variation%20Signals%20Source%20Identification%20System%20By%20Time-Frequency%20Analysis%20Technique.pdf http://eprints.utem.edu.my/id/eprint/24695/2/Voltage%20Variation%20Signals%20Source%20Identification%20System%20By%20Time-Frequency%20Analysis%20Technique.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The quality of power supplies has become one of the issues concern of power supply utilities and users these days. Power quality (PQ) can bring severe problems such as processing interruption within the industries, malfunction and downtime of the equipment and impact in economic losses. Voltage variation signals are the common events in power systems that will affect the equipment or load within the premise. Thus, the source identification of the voltage variation is needed to reduce the impacts caused by the variation. This research presented the analysis and detection of voltage variation signals with time-frequency distributions (TFDs) which included spectrogram, Gabor transform and S-Transform. The voltage variation signals were generated in MATLAB/Simulink according to IEEE Standard 1159-2009. Parameters of the signals were calculated from the time-frequency representations (TFRs) and used for detection of the signals. The detection of the voltage variation signals were performed by k-Nearest Neighbors (kNN), Support Vector Machine (SVM) and rule-based classifiers in which both kNN and SVM gave 100% successful signals detection while rule-based gave successful detection above 97%. Based on the analysis, the best TFD was distinguished by comparing the performance analysis of the TFDs in terms of accuracy, memory size used and computational complexity of the signal analysis. Result showed that S-Transform was the best TFD to be used to analyze voltage variation signals among the three TFDs. In the source identification analysis, voltage variation signals from upstream, downstream and both uptream and downstream were simulated and analyzed by phase TFR. The phase angle of voltage and current variation signals calculated were similar to the input angle with precision above 0.95. The average impedance TFR phase power of the signals were calculated from the phase TFR of each variation signal at each problem source. The classification accuracy of source identification was performed by kNN, SVM and rule-based classifiers. SVM showed the best performance of 94.22% overall classification accuracy followed by kNN of 93.56% while rule-based showed the worst performance among the three classifiers with overall classification accuracy of 83.44%. |
|---|