On controlled switching of high voltage unloaded transmission lines

The /imitation of switching overvoltages on very high voltage transmission lines is traditionally carried out with the aid of circuit breakers - fitted with closing resistors. This widely used system has repeatedly shown itself to be highly efficient. Nevertheless, it is possible to envisage other methods, in particular, those giving control of the circuit breaker instant of closing. This paper examines the possibilities offered by controlled switching on UHV transmission lines and compares existing solutions. It takes into account the various network characteristics that may be encountered. It details the results in terms of overvoltage levels which could be attained, due to the performance of the circuit breaker. The paper also highlights the critical points relative to the application.

A sudden change in the configuration of the electrical trans mission network caused by the operation of a circuit breaker, or alternatively, by the appearance of a fault, can cause a rapid surge voltage known as a switching surge which propagates throughout the network. It is these switching oveNoltages which, in general, determine the insulating requirements of the electrical transmission lines. The limitation of these oveNolt ages is of extreme importance, because they have an important effect on both the cost of transmission lines, and on their rate of failure.
These switching oveNoltages can be manifest at both open ing and closing of circuit breakers. Given that the use of non-re striking circuit breakers is widespread, it is at the instant of clos ing, and above all, re-closing of a line which has held a trapped charge, that the strongest oveNoltages appear.
Faced with this problem, the options chosen by the various operators of very high voltage networks throughout the world are varied. Up until recently, there have been two major alter natives.
(1) The networks that required limitation by means of a closing resistor, because without this limitation the level of insulation required for the lines would reach, or even exceed, 4 pu, which would make them economically unacceptable. Typi cally, these are the types of networks on which the operator uses three phase re-closing operations following a single phase fault, which implies numerous re-closings on lines holding a trapped charge. Thanks to this closing resistor, the oveNoltage level is limited to a value of around 2 to 2.2 pu.
(2) The second alternative are networks that do not use a clos ing resistor, because the oveNoltage level does not exceed 3 pu. These are typically networks on which the operator uses single phase re-closing. In this latter case, the vast majority of closings are carried out on lines that do not hold a trapped charge. This is the reason why the overvoltage does not exceed 3 pu. The percentage of re-closings on a line holding a trapped charge which could give rise to oveNolt ages superior to 3 pu has been estimated at 1 % (for bi-phase faults or badly timed re-closings) and thus can be con sidered to be negligible (EDF-Direction).
For several years now, certain operators have had the ten dency to replace the limitation system using closing resistors to one using phase-earth surge arresters. In certain cases, it can be sufficient that a surge arrester is located at both ends of the line to avoid exceeding the line's insulation level. Nevertheless, in general, one can say that, to obtain the same results as achieved using pre-insertion resistors, it is not only necessary that surge arresters are fitted to both ends of the lines, but they also have to be installed along the length of the line (Ribeiro, McCallum, 1989;Bladow, Weaver, 1990;Leroux et al., 1996;Eriksson, Grandi, Knudsen, 1990 The reduction in overvoltages by using the controlled switching principle requires that the circuit breaker is re-closed at zero voltage across the circuit breaker terminals. Due to the various types of networks and operating methods, this condition can be easier, or harder, to satisfy.

Uncompensated lines
An uncompensated line is one without a shunt reactor. Thus, its particularity is that it remains charged after an opening. This trapped charge evolves according to atmospheric conditions (Beehler, 1964) and its real time measurement requires special voltage transformers, like mixed "resistor capacitor'' dividers. Such equipment are very costly and seldom used. Hence, they will not be considered in the course of this paper. There are sev eral other situations: 2.1.1 The line is fitted with Inductive Voltage Transformers on each phase. In this case, following an opening of the no-load line, the instrument transformer will rapidly discharge the line in less time than it takes for a re-closing (Marks, 1969). Therefore, all closings, or re-closings, will be on a discharged line. When using controlled switching, the closing target must be at zero of the source voltage, both for closings and re-closings. The sim plest case to lend itself to controlled switching 1s that of the con trolled switching of capacitor banks.

1.2
The line is fitted with Capacitive type Voltage Transform ers, conventional or not, which will not discharge the line.
2.1.2.1 If the operation of the line uses three-phase re-closing, this means that all single-phase faults, which represent more than 90 % of line faults, two phases will be re-closed onto trapped charges. The controlled operation should be dealt with differently depending on whether one is carrying out a closing operation on a discharged line or a re-closing. The trapped charge held by the line can evolve as a result of atmospheric conditions. As the capacitive voltage transformer cannot measure the evolution of the trapped charge, it is necessary to assess it. This assessment in some cases can be different from reality. It means that, in these cases, the overvoltages could be higher than calculated.
2.1.2.2 If the operation of the line uses single-phased re-clos ings, the number of cases in which the re-closing could be onto a trapped charge becomes extremely small and is limited to the cases of two-phase faults and untimely three phase re-closings. As stated in the second paragraph, if the insulation level of the line is 3 pu, the number of cases in which the overvoltages could exceed 3 pu dunng a two-phase fault or an untimely re closing is somewhere in the order of 1 % and thus, negligible. If one requires a lower level of insulation, in the order of 2 pu for example, the percentage should be recalculated in order to en sure that it remains negligible. Depending on the results ob tained, controlled switching on this type of line, using s1ngle phase re-closing, should be either very simple because it is li mited to controlled switching on a simple closing or identical to that dealt with in the previous paragraph (see 2.1.2.1 ).

Compensated lines
Shunt reactor compensated lines are particular in the fact that, after opening, there is an oscillation at a frequency somewhere in the order of 50 % to 90 % of the network frequency. The volt age appearing across the circuit breaker terminals shows vari ous degrees of fluctuation, depending on the degree of compen sation. The optimal switching moment for re-closing is a mini mum of the voltage beat.
The problematic network, that of single phase re-closing and three phase re-closing is the same as for the uncompensated lines. The maior difference between this case and the previous, is that this time, the line voltage can be measured. One is thus capatole, theoretically, of using the correct algorithms, of deter mining the ideal instant to close the circuit breaker. Neverthe less, quite apart from the writing of the algorithm, there are still several difficulties linked to the accuracy of the instrument trans formers when operating beyond their fundamental monitoring frequency. This can have an effect on the computational accu racy of the instant of circuit breaker re-closing, and thus, on the level of overvoltage attainable. With an efficient algorithm and good instrument transformers, the level of overvoltage will thus be governed by the closing, and not by the re-closing.
The level of shunt compensation for a line can vary from one instant to the next as a result of the power carried by the line. A line can thus be compensated at one instant but not at another, which causes a supplementary problem for the achievement of the controller.

Simulations
To study the various situations, numerous simulations have been run using EMTP/ATP network simulators.

General network circuit diagram
The simulated network is represented by the next Figures  1 (a, b):

Models used and hypotheses
The network models used are characterized by the following parameters: Rated network voltage: 500 kV Network frequency: 50 Hz Source impedance, Z1 = 0.5 + 25.1 ohm, chosen to produce a short circuit current of 10 kA.
The transmission line is a transposed three-phase line, 400 km long, with the following characteristics: Zero-phase sequence resistance: 0.075 ohm/km Direct resistance: Zero-phase sequence inductance: Direct inductance: 0.025 ohm/km 0.9 ohm/km 0.3 ohm/km Zero-phase sequence capacitance: 6 nF/km Direct capacitance: 14 nF/km To measure the line side voltage of a circuit breaker, an in strument transformer 1s fitted to each phase at the start of the line. Two types of transformers were modelled:  (1) The Inductive Voltage Transformer (IVT), which can dis charge the line, particularly in the case of an uncompensated no-load line.
(2) The Capacitive Voltage Transformer (CVT), which is more common with UHV circuits and which does not discharge the line and consists of a voltage divider in conjunction with an inductive transformer.
The study of compensated lines was carried out using two rates of compensation: 30 % and 80 %.
For certain simulations, a surge arrester with a rated voltage of 420 kV and a max. energy of 3.2 MJ was fitted to the end of the line. The highest overvoltages observed appear some three quarters of the way down the line.
In the event of a three-phase re-closing, one considers, that there is a permanent mono-phase fault between Phase A and earth, with a fault resistance of R1auit = 0 (short circuit). This fault is located at the start of the line.
In order to take into account the closing time, the circuit breaker is modelled using the statistical model for an discon-necting switch. The variation in the closing times of the poles of the circuit breaker is presumed to follow a gaussian distribution, with a standard deviation, which will vary in the simulations be tween 0.25 and 1 .5 ms. A standard deviation of 0.25 ms means that 98 % of the closing time is within a range of ± 0.6 ms of the mean value, which corresponds to a highly stable circuit breaker.
Within the simulations carried out, the circuit breaker as sumed has a rate of decrease of dielectric strength (RDDS) of 1 pu, thus covered by the pre-arcing times taken into account by these simulations.

Algorithms
Algorithms were developed for each network configuration de tailed in the previous paragraphs (see 3.1, 3.2).
In the case of an uncompensated line with an IVT, the line voltage is discharged, both for the healthy phases, as well as for the faulty phase. For the normally operating phases, the voltage transformer is the path by which the trapped currents escape. For the faulty phase, the presence of the short circuit means that there is an exponential voltage discharge via the fault 1m- pedance. As from the instant the line side voltage reaches zero for each phase, the optimum re-closing time is identical to that of an standard controlled closing, given a zero voltage on the source side.
In the case of an uncompensated line with a CVT, the faulty phase is always discharged, due to the presence of the fault, however, a trapped charge, assumed constant, can remain in the healthy phases. The optimum instant for re-closing is thus, for the latter examples, that of the minimum voltage across the circuit breaker terminals, which corresponds to the maximum voltage of same polarity on the source side (Fig. 2). This opti mum instant 1s expressed as a function of the network fre quency: In the case of a shunt compensated line with either an IVT or CVT, the line voltages across the healthy phases display, de pending on the degree of compensation, varying degrees of fluctuation (Fig. 4a, b). The optimal window for the closing cor responds to the minimum fluctuation level of the voltage across the circuit breaker terminals for the phases in question (Fig. 3a  Concerning the faulty phase, the line voltage is subjected to, as in the previous case, an exponential discharge (Fig. 5). The optimum closing instant is that of the source side zero-voltage.

Simulations carried out
The simulations were carried out for each of the configurations detailed in the previous paragraph, with the network parameters as shown in 3.2, and the aforementioned circuit breaker charac teristics.
In order to assess the influence on the accuracy of the cir cuit breaker closing time, 500 static tests were carried out on each of the optimum instants fixed by the algorithms. As well as that, the circuit breaker's pre-arcing time was taken into ac count.
These numerous simulations were carried out using an EMTP/ATP network simulator and an interface capable of gen erating the instant of closing automatically and then displaying the results obtained in terms of temporal voltage, current and overvoltage curves along the line.
The results of these simulations cannot be generalized, but nevertheless, show the order of magnitude expected from over voltages on a 400 km long line. Ta king into account the length chosen, one could consider that one is dealing with a difficult situation.
These simulations will essentially allow definition of the gen eral tendencies for controlled switching of lines.

1 Recapitulative tables and analyses
Tables 1 to 4 show the principle results of the simulations car ried out in terms of overvoltages obtained for various configur ations as a function of the standard mechanical closing time deviation of the circuit breaker.  The following conclusions may be drawn from these simu lations: (1) The highest overvoltage levels are obtained during three phase closings and not for three-phase re-closings in the presence of a single-phase fault (Tables 1, 2 and 3). This is due to the fact that the target window is much larger during a re-closing than during a simple closing. Moreover one sees that, in the case of re-closing in presence of single-phase fault, the maximum overvoltages often do not occur when re closing the breaker but just after the inception of the fault. These results can only be held as valid if the algorithms, updated to include the re-closing target, are reliable and 1f the line voltages have been correctly measured. Within our simulations, we have assumed that the uncompensated line remains charged at its initial value and the compensated line voltage measurements were not subject to error.
(2) The standard circuit breaker time variation is the most important variable parameter. The overvoltage level dep ends on it heavily if controlled switching 1s the only method used (no surge arresters) (Tables 1, 2 and 3).
(3) The influence of the rate of decrease of dielectric withstand (RODS) has not been covered by this article due to the fact that it is of less importance than mechanical dispersion.
(4) Amongst the various studies carried out, the closing on a line with 30 % compensation generated the highest levels of overvoltages (Table 3). A level of 2.3 pu is attained using a circuit breaker, the standard deviation of which is only 0.25 ms, which corresponds to an extremely accurate circuit breaker.
(5) The association of controlled switching and phase to earth surge arresters at the both ends of the line will limit the over voltage levels to a suitable level, which depends much less on the circuit breaker's mechanical dispersion time than in the solution of the controlled switching alone (Table 4). In this case, the maximum overvoltage is located at a point three quarters of the way along the line. As the mechanical disper sion of the circuit breaker increases, the energy absorbed by the surge arrester also increases. Nevertheless, for a stan dard deviation of 1.5 ms, which corresponds to a large dis persion value, the energy absorbed by the surge arrester, 420 kV-3.2 MJ, will never exceed 70 % of its rated value. This solution allows to combine the advantages of the con trolled switching and the protection by surge arresters. It drastically reduces the influence of unavoidable variations in the operating time of the circuit-breakers during all their life time.
4. The effect of controlled switching on equipment

Circuit breakers
The simulations detailed above show that the use of controlled switching as the sole method for limiting overvoltages during un loaded line closing requires the use of circuit breakers with ex tremely stable closing time.
On the other hand, the solution using a combination of con trolled switching and surge arresters should enable the use of standard circuit breakers.

Measuring devices in the substation
Controlled switching requires the installation of voltage trans formers on all three phases of the line output, in addition to the information on at least one phase of the source voltage. The Inductive Voltage Transformers (IVT) allow a rapid discharge of an uncompensated line or the accurate measurement of the fluctuations in a compensated line, due to their large band width.
Nevertheless, the price of IVTs becomes extremely high as the rated voltage increases, also the currently preferred trans formers, Capacitive Voltage Transformers (CVT) are far more economical. These CVTs nevertheless have the disadvantage of a narrow bandwidth, due to the presence of a network fre quency tuned circuit. As well as this, they do not allow trapped charges to escape from uncompensated lines, nor allow measurements to be taken of their level (information on the line side voltage is not indispensable however 1f one knows the exact instant of opening for the pole and the source voltage: This information 1s sufficient for the determination of the level of trapped charges). In the case of a compensated line, the preci sion of the CVTs decreases in line with the decrease in the level of line compensation. In fact, the frequency of the oscillating line load strays further from the network frequency as the degree of compensation decreases.
The use of CVTs with compensated lines should, in most cases, be preceded by a phase of simulations in order to verify the validity of the results obtained.
What can be expected from non-conventional voltage trans formers? Regardless of their technology, they neither discharge the line nor do they have significant influence on primary phe nomenon, due to their low input capacitance. However, associ ated electronics provides a relatively flat frequency response at the vicinity of the rated frequency. Compared to CVTs, whose frequency response presents a resonance at the rated fre quency, this is a significant advantage. Consequently, the tran sient response is more accurate and can be more easily mod elled, allowing, for example, the storage of the trapped charge polarity resulting from the trip operation.

Controlled switching devices
The controlled switching devices available today are in the form of autonomous electronic boxes (often in 19" rack format), in stalled in relay buildings and used as individual protection equipment. Used for simple closing, a single source of in formation on the voltage of one phase on the source side is often sufficient for the calculation of the instant of controlled switching. These are pre-determined by the type of circuit breaker and load, with a few simple corrections due to influenc ing factors (temperature, circuit breaker pressure etc.) and to wear in the circuit breaker (altering the operating time).
The controlled switching devices used for on-line operations should be a little more sophisticated: � acquisition of four sources of voltage information instead of one single source; � increased calculating power required in order to determine the instant of re-closing when dealing with compensated lines. In fact, for this application, 1t is no longer a question of taking into consideration pre-established controlled switch ing instants: Given that the degree of compensation varies with respect to the transitional power, the frequencies of oscillation on the line side can also change. Thus, one should always use algorithms designed to estimate the line side voltage sufficiently accurately and quickly in order that a command may be passed within the time limits laid down by the network control requirements (typically in the order of 300 ms). Ta king into account the mechanical operating time required by the circuit breakers, this would leave approxi mately 150 ms for the algorithm to estimate the frequency signals of between 25 Hz and 60 Hz. This time budget is more than enough to perform this assessment with the required accuracy. Therefore there is no need to reduce the mechanical operating time of present circuit-breakers, as that parameter appears to be non-critical. � Precise determination (with a level of accuracy below 0.5 ms) for the instant of opening of the three poles of the cir cuit breaker, in order to determine the levels of trapped charges in the case of uncompensated lines. � Distinction between simple closing and re-closing in the case of varying optimum instants of manoeuvre.
It is thus possible to envisage the evolution of such an on line controlled switching application into a single device, com municating digitally with remote acquisition units (for the acquisi tion of orders and voltages).

The computational algorithm
The applications for on-line re-closing need to be set up for the configuration of the network (levels of trapped charges and compensation). Apart from the problems due to the computing power, one should also pay particular attention to the sensitivity to "noise" of the algorithms in the estimation of the line side volt age.
The "noise" may come from various sources: harmonics, transients either coming from the network or a non-linearity of the instrument transformer's transfer function, perturbations in the secondary transformer circuit cables, and finally, the noise generated intrinsically within the controlled switching device data acquisition chain. It is essential that one has an algorithm capable of determining the evolution of the compensated line side voltage in 'real time', within a few cycles. Designing this al gorithm certainly represents the hardest part of controlled swit ching on no-load lines.

Conclusion
The wide variety of electrical networks and operational modes means that it is extremely difficult to generalize standardized conclusions. This is made even harder when one takes into con sideration the number of simulations carried out (closings on compensated and uncompensated lines, with or without trapped charges, with IVT/CVT). We cannot simulate all possible situ ations. Nevertheless, taking our simulations as a starting point, which corresponds to the most difficult network conditions, we think that we are capable of stating a few clear principles.
The networks considered correspond to networks in which the circuit breakers used are traditionally fitted with closing re sistors, which ,limit the overvoltages to approximately 2 to 2.2 pu.
One sees that the exclusive use of controlled switching, without other devices, does not always lead to the achievement of the .required overvoltage levels, even with extremely accurate circuit breakers. One could suppose that with such a solution, the operator would never be effected by the unavoidable vari ations in the Ollerating time of circuit breakers during their thirty year life expectancy and also the possibilities of inaccuracies in the algorithms.
As a result of this, it seems necessary to us that phase-earth surge arresters at the ends of the line are systematically associ-ated with circuit breaker controlled switching, as a safety measure. With a standard network surge arrester, it is possible to limit the overvoltage to 2.2 pu, even with a circuit breaker with a mechanical time deviation of up to ± 3 ms. Lower limitation le vels can be achieved using high-performance surge arresters.
The choice of solution to be adopted should be the result of a technical-economic study of the three solutions, which are: closing resistors, the conjunction of controlled switching with surge arresters and finally the exclusive use of surge arresters.
(1) Firstly, a study of the network should enable one to define the parameters of the devices required in order to achieve the required overvoltage levels.
(2) Secondly, it will be necessary to compare the costs of invest ments and operation of the various solutions.
The investment costs are directly drawn from the technical study and are easily detailed. On the other hand, an estimation of the operating costs is much harder to establish accurately due to its dependence on the reliability of the equipment in stalled. It will be all the more difficult to estimate, as there is little operational experience of the solution in question.