Wavelet-automated recognition system for power quality monitoring

Manufacturing industries are now expected to have substantial increases in flexibility, productivity and reliability as well as increasing quality and value of their products. Automatic Data Processing (ADP), sensitive microprocessor, and power electronic equipment are becoming an essential part to control and automate different assembly lines. However, due to the growing economic pressure, modern electrical equipments are designed to meet their operating limits. This fact means that different equipment manufacturers face a dual responsibility to both desensitize against power disturbances and protect their equipment from power faults. This incompatibility issue, between power system disturbance levels and immunity of equipment, results in a severe impact on the industrial processes, which is known as power quality problem. To control and improve electric power quality, the sources and causes of any disturbance must be determined. However in order to achieve this, monitoring devices must have the capability to detect, localize those disturbances and further classify and quantify different types of power quality problems for a proper mitigation method. Different monitoring devices and disturbance analyzers are available that can detect and collect large amount of power quality data. However, there are general problems that exist when dealing with these disturbance analyzers. Off-line analysis is always required. This is due to the design criteria for detection and classification the disturbance. If we utilize the point-by-point comparison technique it is often difficult to build automated recognition system that can on an on-line basis classify the power quality problems such as transient, oscillatory, or non-stationary disturbances. Using this monitoring strategy, one cannot monitor certain class of disturbances or distinguish among similar ones. Furthermore, the selected threshold values (high or low) to be used in detecting different disturbances, may lead to large dimensionality of stored data or undetected important disturbances. The limited capability of Fast Fourier Transform (FFT), while dealing with non-stationary disturbances, is another drawback in the existing monitoring devices. The goal of this thesis is to overcome the deficiencies that exist in monitoring devices and to design reliable, accurate and a wide-scale power quality monitoring system with superior characteristics. Some of the characteristics in the proposed technique are: (1) Fast detection and localization of disturbances that may overlap in time and frequency in a noisy environment. (2) On-line classification by extracting discriminative, translation invariant features with small dimensionality, which can represent efficiently the voluminous size of distorted data. (3) Analysis of different non-stationary disturbances and measure their indices. (4) De-noising ability and high efficiency in data compression and storing. A wavelet-based power quality automated recognition system is proposed in this thesis. This system will assist in the automated detecting, classifying, and measuring different power system disturbances. This system can overcome the drawback in the existing monitoring devices.