The main purpose of this thesis is to combine modified GML algorithm with 2-D simulated annealing for estimation of signal arrival time in the UWB systems.In a dense multipath environment, the generalized maximum-likelihood (GML) algorithm can be used for the time-of-arrival (TOA) estimation. Nevertheless, the GML algorithm usually takes a long period of time, and sometimes fails to converge. Hence, a modified GML (MGML) algorithm is investigated. Two threshold parameters need to be determined in using the estimation algorithm. One threshold is to decide the arrival time range of estimated path, and the other, an amplitude threshold, is to judge whether the estimated path is true. Generally, the decision rule of thresholds may be based on the minimum error probability, which is defined as the sum of false alarm probability and miss probability. To mitigate the effects from noise and dense multipath interference, and to reduce the computational complexity of the algorithm, a method of threshold settings based on the minimum root mean square error (RMSE) criteria is discussed. In this scheme, the RMSE value for each candidate threshold pair in an appropriate region is computed. Constructing an accurate RMSE table and performing a full-scale grid search of adequate threshold settings can be very time-consuming. A 2-D simulated annealing process is adopted for finding the best pair of thresholds for use in the modified GML algorithm. The simulated annealing, different from the gradient descent, can avoid trapping into a local minimum in finding the best threshold pair. The resulting threshold pair makes the modified GML algorithm become more efficient in estimating the signal arrival time with an automatic search manner. Simulation results show that the proposed scheme can achieve better performance than the grid search approaches in UWB environments.

The thesis discusses the design and experimental verification of a new method for detecting partial discharges in oil-cooled high-power transformers, emphasizing the origin of the measured signal from the outer and inner regions of the transformer tank. The detection and spatial localization of partial discharges utilizes UHF (Ultra High Frequency) electromagnetic signal measurement. The UHF partial discharge detection method, also referred as UHF PD, is one of the most advanced techniques for the accurate spatial localization of the signal source position. The focus of the thesis lies in the development of technical and software solutions for the separation of partial discharge signals originating from the internal space of the transformer and the spurious signals from the connection lines. The proposed technical and program solutions constitute a signal discriminatory method. The precise and repeatable diagnostics of the transformer state are guaranteed by observing the measurement methodology and special setting of the diagnostic system. The functionality of the signal discriminatory method was verified during real measurement of the oil-cooled power transformers at the Dukovany nuclear power plant in operation. The output of the designed methodology is to set up a calibration procedure and follow-up steps to ensure objective, repeatable, and comparable measurements for the purposes of regular predictive maintenance in transformers.