New Approaches To On-Line Power Quality Disturbances Monitoring

Industrial and commercial electricity consumers and the utility companies that serve them need access to power quality (PQ) information that is relevant and timely. Online monitoring of Power Quality disturbances (PQD) is a direct and effective way to obtain PQ information, and it has become a hot research subject in IEC technical fields. There exist two core issues on PQD online monitoring: First, the real-time disturbance detection; the accuracy and reliability of the monitoring system are directly determined by the real-time detection algorithm. Second, PQ information "mining"; during on-line PQ monitoring process, a lot of PQD waveform data will be collected. In order to extract "useful" PQ information from these massive PQ waveform Data, a fast and effective PQ data mining algorithm is needed.The main idea of this dissertation is as following sequence: disturbance signal denoising, transient harmonic measurement, disturbance online detection, and PQD data classification mining. Utilizing wavelet theory, Teager energy operator (TEO), and data mining technology, this dissertation presents a comprehensive and systematic study of PQD online detection and classification.(1) A distributed power quality monitoring system (PQMS) is presented. The presented PQMS allows for finding the quality indexes at different nodes of the power network. We integrated in the system the capability to measure different disturbances that is required for the continuous monitoring of power quality.(2) Two different kinds of PQD denoising algorithm are proposed. One is an adaptive threshold denoising algorithm based on cross-validation (CV-based adaptive algorithm), the other one a block-thresholding-based denoising algorithm (BT-based algorithm). The proposed CV-based adaptive algorithm can adapt to the changes in the measured signals through the steepest descent method to find an optimal denoising threshold. And the CV-based algorithm is compared with the VisuShrink and SureShrink algorithm which are widely used in PQMS. Simulation results illustrated the superiority of the proposed algorithm. In BT-base algorithm, the wavelet coefficients of various scales are divided into several blocks, and threshold values were determined for each block. Simulation and experimental results show that BT-base algorithm is better than CV-based adaptive algorithm and responses faster.(3) A transient harmonic measurement algorithm based on wavelet packet transform (WPT) is proposed. The proposed algorithm overcomes the shortcomings of FFT-based method such as sensitive to noise and inaccurate performance in non-stationary environments. By choosing the appropriate sampling frequency and wavelet decomposition tree, the WPT-based algorithm can divide the frequency bands linearly with constant bandwidth to separate each harmonic. Based on simulation studies, performance of the method is presented and its accuracy is compared with FFT-based method. Simulation results indicate that the measurement accuracy of the WPT-based algorithm is superior to FFT.(4) Real-time and accurate detection of power quality (PQ) disturbances is a key issue to PQMS. By introducing Teager energy operator (TEO), a novel TEO-based detection algorithm is proposed. The proposed algorithm uses three adjacent samples of the signal and requires only three arithmetic operations per each time shift; it has excellent time resolution. Simulation and experimental results show that the algorithm can accurately and quickly detect the occurrence and location of power quality disturbances; and it is robust in the presence of white noise. The proposed algorithm is suitable for online real-time PQD detection.(5) The huge amount of PQD data collected and stored by the online power quality monitoring system pose a great challenge for data analysis and identification of the type of disturbance. And it has prompted the need for intelligent data analysis technology, which could discover useful PQ information from raw PQD data. In this dissertation, we investigate the potential of decision trees (DT) for PQ data mining in electric power systems. And a data mining approach to extract pertinent and distinctive information from each class of power quality disturbances is proposed. In the proposed approach, wavelet transform has been used to extract unique features of the various power quality disturbances. And feature vectors are used to train DT classifier. Simulation results show that the proposed approach is of high accuracy, and of fast response. It is suitable for power quality information extraction.All theoretical analysis, simulations and experiments carried out in this dissertation proved that the proposed algorithms will play an important role in the power quality disturbance monitoring system. The proposed algorithms provide a theoretical basis for the design and development of PQD online real-time monitoring system and can effectively improve the reliability of the PQMS.