This dissertation objective to analyze the influence of wavelets families and their orders in the performance of a fault location algorithm through the traveling waves from two ends of an overhead transmission line. Then, a secondary objective is the modelling of an
electric power system (EPS) to obtaining of a fault’s universe to validate the algorithm. For that, part of an EPS of the Eletrobrás-Eletronorte on 500/230 kV was modeled in Alternative Transient Program (ATP) using real parameters. The Wavelet Transform, through multiresolution analysis (MRA), it’s used because its time domain localization characteristic to allow characterizations of electromagnetic transitory moments caused by contingencies. The waves generated by faults traveling towards the line’s terminals containing the wave propagation times from defect point at the terminals and these times can be conveniently extracted for such transformed. By means of the methodology adopted in the algorithm, the
difference between those times determines the fault occurrence place on the line with good accuracy. However, one of the error variant agents in the estimation it’s the selection of the wavelet used in MRA of the signals, therefore the evaluation of that choice upon those errors, the main objective of this work, it’s justified for the still inexistent scientific bases that assure
the choice of a great wavelet to a certain application. Among a diversity of discrete wavelets, appropriated results were obtained for 16 of them with maximum errors smaller than 250 meters, stipulated precision adopted. Two wavelets, Db15 and Sym17, were more exact than the others, respectively, 3.5 and 1.1 times. The methodology to get those results consists of
the: export of the fault data files from ATP for MATLAB®; application of the Clarke’s modal transformation of voltages; decomposition of alpha modes and synthesis of the levels 1 of details through AMR; calculation of their maximum magnitudes and time indexes determination; finally, the traveling waves theory is used to formulate the estimate equation of the contingency place on the line. All methodology was programmed in MATLAB and the errors’ statistical analysis was made in Microsoft Excell®. At the end, a GUI (Guide User Interface) for (Human Machine Interface) HMI of locator was elaborated, which is also a graphical analysis interface, no matter which the contingency applied to the EPS. The attained results demonstrate an improved performance in reason of the most appropriate wavelet selection to the algorithm and they suggest a practical application of the locator.
|
