Jeffery C. Chan

Pattern recognition techniques and their applications for automatic classification of artificial partial discharge sources

Partial discharge (PD) source classification aims to identify the types of defects causing discharges in high voltage (HV) equipment. This paper presents a comprehensive study of applying pattern recognition techniques to automatic PD source classification. Three challenging issues are investigated in this paper. The first issue is the feature extraction for obtaining representative attributes from the original PD measurement data. Several approaches including stochastic neighbour embedding (SNE), principal component analysis (PCA), kernel principal component analysis (KPCA), discrete wavelet transform (DWT), and conventional statistic operators are adopted for feature extraction. The second issue is the pattern recognition algorithms for identifying various types of PD sources. A novel fuzzy support vector machine (FSVM) and a variety of artificial neural networks (ANNs) are applied in the paper. The third issue is the identification of multiple PD sources, which may occur in HV equipment simultaneously. Two approaches are proposed to address this issue. To evaluate the performance of various algorithms in this paper, extensive laboratory experiments on a number of artificial PD models are conducted. The classification results reveal that FSVM significantly outperforms a number of ANN algorithms. The practical PD sources classification for HV equipment is a considerable complicated problem. Therefore, this paper also discusses some issues of meaningful application of the above proposed pattern recognition techniques for practical PD sources classification of HV equipment.

Self-adaptive partial discharge signal de-noising based on ensemble empirical mode decomposition and automatic morphological thresholding

This paper proposes a self-adaptive technique for partial discharge (PD) signal denoising with automatic threshold determination based on ensemble empirical mode decomposition (EEMD) and mathematical morphology. By introducing extra noise in the decomposition process, EEMD can effectively separate the original signal into different intrinsic mode functions (IMFs) with distinctive frequency scales. Through the kurtosis-based selection criterion, the IMFs embedded with PD impulses can be extracted for reconstruction. On the basis of mathematical morphology, an automatic morphological thresholding (AMT) technique is developed to form upper and lower thresholds for automatically eliminating the residual noise while maintaining the PD signals. The results on both simulated and real PD signals show that the above PD denoising technique is superior to wavelet transform (WT) and conventional EMD-based PD de-noising techniques.

Automatic Blind Equalization and Thresholding for Partial Discharge Measurement in Power Transformer

Partial discharge (PD) signals acquired from on-line measurements of power transformers are easily overwhelmed by various interference and noise. This paper proposes an automatic blind equalization (BE) and morphological thresholding method for PD signal de-noising. Firstly, BE automatically selects an equalized signal that reveals PD impulses from an acquired noise-corrupted signal. Then, automatic morphological thresholding (AMT) is adopted for determining thresholds on the equalized signal. After de-noising with BE and AMT, phase-resolved pulse sequence (PRPS) is constructed and used for analyzing the types of insulation defects that cause discharges. To verify the proposed method, PD measurements on experimental PD models and a distribution transformer have been conducted. The results show that PD impulses can be extracted from severely noise-corrupted signals by using the proposed method. Also, PRPS constructed from de-noised signals can achieve consistency in revealing the types of insulation defects even different types of PD sensors and measurement systems are used.

Partial discharge pattern recognition using multiscale feature extraction and support vector machine

An accurate interpretation of partial discharge (PD) signals in high voltage (HV) equipment provides crucial information for assessing the insulation conditions. To automate the interpretation process, feature extraction of PD signals and pattern recognition using the extracted features are required. This paper adopts discrete wavelet transform (DWT) and empirical mode decomposition (EMD) for signal decomposition and feature extraction on the PD signals obtained from different insulation defects. Support vector machine (SVM) is then used for classifying the features. Results indicate that features extracted from decomposed signals provide higher classification accuracy when compared with the conventional method that the features are extracted from original PD signals.

Stochastic noise removal on partial discharge measurement for transformer insulation diagnosis

Measurement of partial discharge (PD) paves a way for transformer insulation diagnosis. However, noise always interferes with PD signals and can jeopardize the diagnostic reliability. Therefore, it is necessary to adopt signal processing techniques to remove noise from collected signals. Among various types of noise, stochastic noise is considerably difficult to remove due to its similarity with PD signals. This paper proposes an effective method, which adopts fractal dimension and entropy analyses to remove stochastic noise. To verify the proposed method, PD measurements have been performed on a number of experimental models and a substation transformer. Results prove that PD signals can be extracted while the noise can be eliminated from collected noise-corrupted signals by using the proposed method. A comparison with a wavelet transform-based noise removal method has also been made in the paper.

A comparison of Gumbel and Weibull statistical models to estimate wind speed for wind power generation

Wind energy is becoming more common, especially as costs are falling. In Australias National Electricity Market, the total available generation is managed by the Australian Energy Market Operator (AEMO). One of its tasks is to forecast the availability of generation twenty four months into the future, to ensure that the predicted customer load requirements are met. A challenge, however, is to accurately forecast the contribution of wind energy to the market on this time frame. Since the energy of wind is a function of its speed, it is common to use climate data to estimate the wind speed into the future using statistical distributions. In this analysis measurements on power generation from a South Australian wind farm and on wind speed from a weather station were compared. Statistical techniques were applied to monthly data samples. The power generated from a wind turbine is generally highest at the tail end of the wind speed distribution. Thus, the accuracy of two distributions to model wind speed, the Weibull and the Gumbel, was investigated to see which gave better fits. The Gumbel distribution was found to estimate wind speed more accurately than the Weibull model, not only at the tail end of the distribution, but also at lower levels.

Partial discharge sources classification of power transformer using pattern recognition techniques

Continuous Partial discharge (PD) monitoring can help assess the integrity of transformer insulation system. Over the past few decades, various aspects of PD techniques have been investigated. Current research of PD focuses on multiple PD sources classification, which aims to identify the types of several defects that may coexist in a transformer and cause discharge. This paper develops a hybrid discrete wavelet transform (DWT) and support vector machine (SVM) algorithm targeting multiple PD sources classification. To evaluate the performance of this algorithm, experiments on a number of artificial PD models and transformers are conducted in the paper.

A case study into improving the 24-month mid-term forecasting of wind energy by combining with PVs

Forecasting the output of wind energy plants over many months is problematic because of the unpredictability of the weather. Usually, a large proportion of conventional generation must be standing by as reserve capacity in case the weather is not conducive to that type of renewable technology. A possible solution is to operate in conjunction different types of renewables. For instance, during stormy weather the output of a PV plant might be low, however, the output of a wind farm will be high. To investigate the appropriateness of this aggregation, data from a South Australian wind farm was studies, along with a hypothetical PV plant, to determine whether the output of the conjoined plant was more reliable than that of the wind farm on its own.

Advanced signal processing techniques for transformer condition assessment

Partial discharge (PD) measurement has been widely adopted for condition assessment of transformers. The major tasks include effective extraction of PD signals from measured signals, accurate representation of PD signals, explicit multiple PD source separation, and PD source classification. This paper applies empirical mode decomposition (EMD) and mathematical morphology (MM) for extracting PD signals from noise-corrupted measured signals and representing PD signals on a joint time-frequency (TF) map, which is used for separating multiple PD sources. A Support Vector Machine (SVM) algorithm is then adopted for classifying each PD source. Case studies are provided to demonstrate the applicability of the two techniques in analyzing PD signals obtained from online PD measurement of field transformer. Comparisons between the two techniques and conventional wavelet transform-based techniques are also provided in the paper.

Bayesian neural network and discrete wavelet transform for partial discharge pattern classification in high voltage equipment

Partial discharge (PD) pattern recognition has been applied for identifying the types of insulation defects in high voltage (HV) equipment. This paper proposes a novel Bayesian neural network (BNN) and discrete wavelet transform (DWT) hybrid algorithm for PD pattern recognition. Laboratory experiments on a number of PD models have been conducted for evaluating the performance of the proposed algorithm.