Resultados estadísticos del inventario forestal

An important step in PD pattern recognition is to quantitatively [40] describe the ongoing degradation process within the insulating media. To this end, the mea- sured data are either directly used or manipulated to form representative measures for an ongoing process. In this work, the representative quantities are classied in two main groups namely basic quantities and derived quantities [42]. The basic quantities involve information that directly can be extracted from the recorded data. The derived quantities require post-processing the data and may include choices for their representation. In the following subsections, description of dis- charge related parameters are introduced.

4.2. PD RELATED PARAMETERS 39

Basic quantities

ˆ PD magnitude: PD magnitude (q) is one of the indicators accompanying aging phenomena in cable insulation. The PD magnitude depends, among other factors, on the size of the discharging source and its location in the insulation [52]. This parameter is usually expressed in picocoulombs [pC] or nanocoulumbs [nC]. The PD magnitude is symptom of the existence of a defect in the insulation and its level is related to the extend of the damage [67] in the insulation. Besides, due to the discharge event taking place in the insulation, the source can grow both over the length and depth of the insulation for instance in the form of electrical treeing. Such growth, both in length and volume, results in changes of the PD pulse magnitude [51]. This can be understood according to the ratio of the defect size to the healthy section of the insulating material. The growth of defect results in increased PD magnitude. Moreover, during the growth of the defect, new small PD sites are created and therefore, new small PDs are initiated during the degra- dation of the insulation. However, not always the PD magnitude increases while degradation is progressing. For certain defects, it is observed that the PD activity drops or even can cease during the aging time. This behavior can be attributed to changing of the chemical composition of the insulation near the defect changing the electrical characteristics, e.g. carbonization causing a conductive path preventing further discharge activity. Therefore, continu- ous monitoring of the PD magnitude helps rstly in revealing the existence of the defect and secondly in indicating the state of the insulation and de- fect growth rate. However, as mentioned earlier, the level of the measured value is inuenced by the location of the defect. Basically, PD pulses get attenuated depending on insulating material and length of the cable as well as equipment present in the propagation path [13] while traveling toward the detection sensors. Ring main units in the cable connection reduce sig- nal amplitude especially for frequency content above 1-2 MHz. Substations with huge rail structures have a high impedance and partly block PD signal transfer (For details see [13]).

ˆ PD number: Number of PDs (n) measured during a selected time interval is also indicative of a damage and/or deterioration process in the insulation. This value in combination with PD magnitude provides valuable information to be used to identify the state of the insulation as well as the type of the discharging source. The number of PD pulses may increase with the progress of the deterioration. This phenomenon can be ascribed to the fact that the the probability of generation of an initial electron needed to trigger a PD event increases in the deteriorated insulation [68, 69]. Also, the PD inception voltage can drop. Therefore, PD occurrence increases which can lead to faster degradation of the insulation.

Derived quantities

ˆ PD Charge density: According to the IEC 60270 standard [70], PD charge density is dened as the summation of absolute values of individual apparent charge magnitudes q from a chosen eective length Lef f and during eective

measuring time Tef f and normalized both on Lef f and Tef f. For distributed

discharge regions this is a proper quantity. However, when the PDs arise from a concentrated origin with size less than the chosen eective length, the charge density becomes dependent on the arbitrarily chosen value of Lef f. In

fact, in the practice of power cable diagnostics, this is usually the case since localized cable joints are the most common origins of the PD activity (see e.g. Figure 4.9 related to mapping diagram shown in Figure 4.1). Therefore, in this thesis the summation is made over a relative length (Lef f/Lcable) which

is taken as 1‡ of the full cable length and the normalization on length is omitted. Note that this fraction is about a factor ten below the typical overall location accuracy, which is limited mainly by PD signal dispersion during propagation along the cable [13]. However, the reproducibility of the pulse location is higher when PDs arise from a localized defect with magnitude well above the noise level. As mentioned in Chapter 2, one measurement corresponds to the duration of a power cycle, i.e. 20 ms. The eective measuring time is equal to the number of combined measurements multiplied by the duration of a single measurement. The optimal value of Tef f is

a compromise between the statistical signicance of the accumulated data, and the time scale over which one would like to observe varying PD activity. Longer averaging time obviously results in smoother statistics (see Figure 4.10). Too long averaging time, however, will disguise temporarily activity occurring during e.g. an overvoltage situation. At present, measurements taken over one hour are combined (60 power frequency cycles when a record is taken every minute). An additional advantage of this time scale is that it corresponds to a typical thermal response time of cables upon changing load conditions. Varying PD activity related to load cycling can therefore still be recognized and distinguished from trends due to aging. Another advantage of hourly based parameters is that it allows cable network owners to perform corrective actions in case there is a high risk PD activity found. Trends can be observed in PD activity after already one or a few days of monitoring. The density is taken at position l which is mÖLef f (with m an integer ranging

from 1 to 1000) and time t is discretized according to nÖTef f. For each

location lmand time tm, the PD density is given by:

P Ddens. (lm, tn) = X j P i qi, j Tef f, j/Tcycle (4.1)

where qi, jis the discharge magnitude with i indicating the discharges within

the selected length and j the discharges within the selected time range. Usu- ally, each record is taken just over the duration of one cycle since it needs

4.3. PD RELATED PATTERNS 41