1998 Winter IEEE/PES

STS Panel Session Paper

STS Design Requirements & Considerations

by

Colin E.J. Bowler, ABB - US Power Systems Inc.

Introduction

Much has been written about the STS application relative to requirements and value, however, very little has been written or said about what is required to achieve a workable STS, considering the following important areas of design:

1. Control Signal Measurement
2. Non-Sinusoidal Load Current Operation
3. Transfer Operation
4. Up-Stream/ Down-Stream Fault Discrimination
5. Protection Functions
6. Monitoring Functions
7. Cooling Methods
8. Main Circuit Design.

These areas concern some important specification requirements that are not easily met, and require a control capability far in excess of the more common voltage control devices such as the SVC. Establishing and meeting these requirements is the subject matter of this paper.

The most important reason for the complexity of control is the fact that the STS is applied in series with the power system elements, using a discontinuous switching device like a thyristor, and to create from it a behavior similar to a perfect conductor in the on-state, and a high impedance off-state. Because the STS is in series with the load, the STS must be able to perform these functions for arbitrary current level, current shape, and power-factor; from load-transformer low load pick-up, and up to full short circuit current, without introducing any appreciable current or voltage discontinuity of its own making, such as might be caused by delayed gate signal application. These facts lead us to the following discussions.

Control Signal Measurement

As a power system controlled device, the STS is required to measure its applied voltages and currents, act on the embedded information in these signals and exercise control of a thyristor switch as determined by its control functions, and also respond to operator inputs.

The most demanding requirement for measurement is the dynamic range of current that must be measured. In the case of cold load pick-up the load current may only be a few amperes such as for a power transformer excitation current, or just feeder capacitive reactive loading. In the case of short circuit operation the current may be hundreds of thousands of amperes. For a typical 15 kV feeder we may see a range of 50,000 :1 for instantaneous current level maxima.

Under this range of current conditions the control is required to track current zero points to affect thyristor switching current polarity reversal, at the identical instants as if the switch were replaced by a normal conductor. The measurement requirements to deal with this situation are very demanding because the signal-to-noise ratio is highly variable unless special means are taken to control this factor. Noise in the analog signal is inevitable, especially because of the need to isolate the signal source galvanically, to meet surge requirements, and to isolate high common-mode voltages from the control electronics. This noise source generally precludes using a high order (16 bit) converter to full advantage, because if we set gains for low level signals, the high level signals will be clipped (distorted) and this is not allowed (see below); while if we set-up the measurement for distortion-free high-current measurements, the noise will approach the signal level at low current, and consequently current zero switching will not be possible (also required, see below).

One method to achieve the required signal-to-noise over the wide range of operating currents is to use automatic gain control of the signal ahead of the signal isolation device. One approach would use two digital control bits for a gain controlled instrumentation amplifier to achieve four input gain settings (e.g. 1, 10, 100, 1000) that are automatically switched by the control according to the instantaneous current level being measured. The control is in charge of the gain settings and obviously knows the effective gain to use in calibrating the signal information to actual engineering significance for use in the firing control algorithm. Even with this programmable gain amplifier (PGA) capability, it is still useful to utilize a wide range 16 bit analog-to-digital converter in order to deal with the remaining range of signal-to-noise at each gain setting, and to reduce internally generated noise, such as quantization, and a host of other sampling induced noise effects.

The second important measurement requirement is the need for minimum phase delay, input to output, so that firing signals going to the thyristor switches are also not delayed. This requirement sets conditions on the control algorithm for detecting current zeros, namely phase compensation, which are affected by both the accuracy of signal measurement, and the sample rate of data acquisition. Signal accuracy and sample rate are tied together in a digital control because of signal aliasing, the creation of in-band signals due to too low a sample rate. To avoid aliasing requires the sample rate to exceed the bandwidth of the signal information by a wide margin (typically 3:1 minimum). This sets the parameters of signal filtering in terms of roll-off rate and bandwidth. A suitable data acquisition arrangement that meets these requirements would have an 8 pole linear phase anti-aliasing filter, sampling at 120 samples per power frequency cycle. Linear phase filtering is a requirement to eliminate signal distortion, and to allow simple digital correction of phase delay in the firing control, even under transient conditions. With these measurement provisions, it is reasonable to proceed with firing control design requirements evaluation.

Non-Sinusoidal Load Current Operation (Firing Control)

Firing control of a thyristor switch at prospective current zeros points, is well understood for situations when the currents are near sinusoids, such as for SVC type equipment. The SVC is a shunt connected device that is of known impedance, where the voltage applied varies over a small range. In contrast, the STS is series connected and the impedance of the load is always unknown. What is required for firing control generally, is a synchronous phase-locked sample data collection system that tracks the system frequency variations. One approach for this will sample voltage and currents synchronously based on phase-locking the sampler to a system voltage, using a digital phase-locked loop (PLL) control algorithm. The computation of the time to current zero, the firing point of the next thyristor switching, is determined from a phasor calculation of the individual phase voltage and current sinusoids relative to the phase-locked reference signal.

In the case of an STS, the control has no information about the load current expectation, other than it can be almost arbitrary in level and harmonic content, and will generally exhibit transient phenomena on odd occasions, including long-term dc-offsets (transformer inrush current). Tracking current zeros in this environment is very difficult, but is required of the STS if the control is to avoid introducing changes in the current due to firing activity. To do otherwise will create chaos of the load voltage and as a result produce power quality problems rather than eliminating them.

	Bus Voltage
	Line Current
	Firing pulse +
	Firing pulse -
	Mode Selector

Figure 1, Firing Mode Tracking Control

It is possible to track current zeros for arbitrary current shapes with time, by special interpolation algorithms that can be executed in modern digital signal processor type cpu’s, in real-time. Such a control is illustrated in Figure 1, demonstrating a three mode control, for tracking changes in transient, harmonic, and harmonic free sinusoids. Mode changing is seen to be almost instantaneous with the change in the signal quality as illustrated by the firing point tracking of the current zeros, and the mode predictor output.

Transfer Operation and Up-Stream/Down-Stream Fault Discrimination

The objective of the STS is to switch (transfer) the power to the load from a preferred source to an alternate source automatically and rapidly when a reduced voltage (imperiling power quality at the load) is established in the preferred source, and while the alternate source meets certain quality criteria. Usually the need to transfer power supplies is because of an up-stream fault that may reduce the preferred source voltage transiently.

Two algorithms are required to pull this off namely:

· High speed fault direction indication

· High speed transfer algorithm

In the case of the up-stream fault the transfer operation must be accomplished rapidly in the faulted phase and in a manner that does not transfer the fault to the alternate source even for a fraction of time, as may happen if the transfer algorithm is not very intelligent. Figure 2, indicates that for the condition shown, operating the S2+ thyristor gate, on fault detection, at the instant when current would want to flow in this direction from the preferred source will transfer the fault to the preferred source until the next current zero of S1. The transfer algorithm must then be intelligent enough to pick the right polarity of the alternate source switch, relative to the preferred source current polarity, in order to switch at the earliest instant in time to affect source commutation without fault transfer, and at the same time minimizing voltage disturbance at the load. The corollary situation is to prevent transfer during a down-stream fault incident, because this would transfer the fault to an otherwise healthy feeder, which is not too smart considering the additional number of customers that would be affect by this act.

Figure 2, Avoidance of Fault Transfer

We note that the detection of the up-stream fault, and the transfer action must be very fast in order to minimize power quality issues. This means that the detection of a down-stream fault must be equally fast in order to inhibit transfer. Ideally, the fault direction detector and the transfer algorithm must be essentially instantaneous in order for the transfer switch to operate in a sub-cycle time window.

These requirements represent a very tall order for the control, compared with algorithms that are typically used for this in modern digital protective relay applications, where at least a half-cycle is used to make a determination of magnitude, and much longer is required to determine phase information accurately. A relatively high sample rate, and very intelligent signal processing make this possible in the STS control.

The reason that an STS can operate in sub-cycle time at all, using thyristor based switches, is based on their rapid forced commutation resulting from the under-voltage on the preferred source, and the much higher voltage on the alternate source forcing open the preferred source thyristor. Figure 3, illustrates a typical forced commutation switching operation. The expedient of transferring the gate signal at the correct instant, allows the force commutation of the thyristor switch to take place naturally. This kind of switching is called make-before-break switching, because during the transfer, current flows briefly (recovery current) from the alternate source to the commutating thyristor in the preferred source. For phases in the switch that are not affected by low voltage, the transfer switching is called break-before-make. This latter kind of switching is normal commutation at the next available

	Bus Voltage
	Source Current
	Load Current

Figure 3, High Speed Transfer Operation

current zero after thyristor gate signals have been removed. There may be a slight delay in this type commutation to allow the control to be sure that current zero has been achieved in the preferred source, prior to gating the alternate source. Manual switch operation is also of the break-before-make variety.

Figure 4 illustrates an incorrect high speed transfer that resulted in fault transfer. Here, the delayed fault current zero, due to dc-current offset, also played a role causing the fault to be fed from both sources for two cycles.

	Bus Voltage
	Source Current
	Load Current

Figure 4, Incorrect Transfer Operation

Monitoring & Protection Functions

It takes a very large number of power and signal electronic components to construct an STS system. This is so, because many identical power electronic devices are required in series to handle the voltage, and parallel branches may also be required to handle the highest current levels. As a consequence of this, there are many potential failure modes that may cause switch malfunction. Actual failure modes are minimized by careful component selection for reliability in the operating environment. However, the control system must still provide for protection from equipment failure to:

1. protect from producing power quality problems, and 2) to protect the valuable equipment from additional failure.

An important quality in design of the STS is self protection, i.e. fail-to-safe-condition, from any major failure in the firing control computer, such as software error for example. While every effort is made to make the software and hardware reliable, care is required in design to allow the switch to turn-off naturally, as soon as practical, in the event of failures of this type. This condition can be achieved, in this case, by using a high sample rate for data collection, and requiring up-date of the firing time every sample, so that if up-date does not occur due to some failure, the timer will never re-fire the thyristor switch.

Because we are dealing with a thyristor switch, many hardware failure modes may result in serious consequences for the load current.

For instance a failure to gate a thyristor on one switch polarity side, or a delay in firing will cause dc and harmonics to be present in the load current and voltage, induced by this failure. This leads to a requirement for harmonic and misfire protection to be built into the control protection algorithms.

Protection Type	Indication	Action
Unbalance Current	Exceeds specification	Transfer
Over-temperature	Excessive RMS current	Turn on more fans or transfer or by-pass & trip
Transient over-temperature	Excessive peak current	Transfer
Misfire	Protect load	Transfer or Trip & by-pass
Excess harmonics	Higher than specification	Transfer
Over-voltage	Greater than twice rated	Transfer
Gate Failure	Stack stops conducting	Transfer
Thyristor failure	More than 2 in one phase	Transfer
	One failure in one phase	Alarm
Power Failure		Alarm Bypass Trip

Table 1, STS Protection Functions

The preferred list of protection functions and actions is given in table I.

Many of the protection functions are similar to those used in other electrical apparatus, the main differences relate to the special considerations required in thyristor applications. Excessive peak current is one of these, which relates to the local temperature of the semi-conductor junction which has a short thermal time constant on the order of milliseconds. The sensitivity here requires a thermal analog algorithm executing in the control computer, to predict the temperature based on the thyristor volt-ampere relationship including its variability with temperature. Other protections in this category include loss of the gate signal, and the protection from a shorted thyristor element.

Typically a thyristor switch for 15 kV application may contain eight or so thyristor elements in series in order to withstand the voltage applied when not conducting (blocked). Obviously the loss of a gate signal to one level will cause an over-voltage on that level when the other thyristor elements in series start conducting. The failure mode in this case is likely a shorted thyristor. The gate failure monitoring provides a rapid means of detecting this condition, for which the protection function can operate and save the thyristor from total failure by transferring the switch to the alternate source.

It is important to notice that the transfer option for protective action, is utilizing the alternate source as a means to maintain supply reliability to the load for internal switch failures as well as for external event power quality issues. This is important, and is a contrast with other competing power quality device mitigation solutions that do not have this natural redundancy as part of the design, such as in the case of an SVC or other voltage regulating device. This feature should be evaluated carefully because it makes the STS intrinsically more reliable than some of its competing alternatives.

Monitoring Functions

There are a number of monitoring functions of value in the STS to provide inputs to the protection system for the thyristor elements. These are:

· Gate Reception Detection

· Heat Sink Temperature Measurement

· Thyristor Failure Detection

These monitor functions occur on the high voltage elements of the switch, and the information detected must be communicated, via a galvanic isolation barrier, to the control for protective action to take place.

Gate reception monitoring is important to provide early warning of a failure of the control to deliver a switching impulse to the gate of the thyristor. This function is both a protective system input as well as a monitoring of the health of the switch. A failure to gate a thyristor on time will assure its failure from over voltage. There are many systems required to be working in order for the thyristor gate to be energized including: analog signal measurement, the firing time computer, the pulse output timers, the gate energy delivery system, and the gate information delivery system. The gate reception monitoring system, working on every level of the switch, together with the thyristor failure detection system and thyristor temperature monitoring serves to narrow down the causes of thyristor failure.

Thyristor temperature monitoring systems are useful to protect the STS from cooling system malfunction, as well as to protect from operation outside of the switch design specification. The objective of the temperature monitoring system should be to predict the temperature of the thyristor junction. This element of the switch is the smallest in terms of thermal inertia, and is small enough to be affected by the cycle by cycle values of peak current. The thermal analog energy storage vs cooling temperature model must calculate the heat generation of the thyristor, which is nonlinear with temperature, and the cooling effect of the heat sink effects in the solution of the heat flow equation. Ultimately the switch can be transferred to the alternate source in order to allow for cooling, but in worst cases would provide a trip as the protective means for the complete series thyristor string.

While the cause of thyristor failure is required to be known, we also must detect the fact of the actual thyristor failure when it occurs. Failure to detect will result in operation of the switch with a higher voltage per thyristor level. Ultimately the number of failures will increase until the string of thyristor element is too short to support voltage, resulting in complete failure of all of the thyristor in the series string. Detection of a single thyristor failure is then vital so that maintenance of the thyristor string can be accomplished before such complete failure occurs.

Cooling Methods

The type of an electrical apparatus is indicated by its ability to work at its rating at the highest expected ambient temperature. Thus the equipment rating is primarily controlled by the ability to cool the apparatus. In the case of a thyristor based transfer switch we have the option of air or water cooling based on existing technology. For transmission equipment thyristor application (HVDC, SVC etc), water cooling has generally been used to maximize equipment utilization to allow very high ratings. Industrial applications of thyristor elements, such as drive systems, have traditionally been air cooled. This is more for convenience, recognizing the lower required equipment ratings and the ability to use air as an option. The use of air cooling allows the choice of free or forced convection. Forced convection provides for a more compact equipment as fans take up little space while providing an order of magnitude increase in heat transfer capability.

A major consideration in air cooling, especially for an outdoor or indoor electrical apparatus is the question of audible noise. The control of noise begins with attention to pressure drop and flow rate. The higher these figures the higher the potential noise induced by flow buffeting. Once flow is established, acoustic noise may be amplified by resonance with the mechanical structure.

Air cooled apparatus must also be designed to handle air-borne pollution, especially particulate matter that may block air passages over time. For outdoor application there is also the need to consider moisture, which increases the collection of particulate matter in a condensing atmosphere. One approach, to cater for outdoor switch application with forced air cooling, is to provide separate internal and external air flows using an air-to-air heat exchanger, or air-to-water heat exchanger. This arrangement allows the internal air to be maintained none condensing by the expedient of heating the air above the dew point of the internal air. Under normal loading there is enough local heat to perform this function. For the occasions when current is low, and with low ambient temperature, heaters may be employed in the internal air path to stay in the non-condensing region.

The use of forced air cooling allows the possibility of regulating temperature by fan control. Where dual redundant fans are available, each may be operated separately, so that in cold weather no fans will run. As ambient temperature increases, the single fan will regulate an acceptable temperature for operation up to a 40^oC ambient. For extreme conditions or if for some reason air flow is impeded, the second set of redundant fans will be called into operation to maintain temperature to acceptable limits. Above some maximum operating temperature the switch may be tripped to protect from over-heating.

In addition to main circuit cooling, the cooling and management of humidity is also required for the control hardware, for switch environments where climate control is not supplied, such as outdoor application. This requirement leads to a totally enclosed control, with an air-to-air heat exchanger of its own. There is added value from total enclosure of the control by increasing noise immunity of the control from outside sources, and reduction of noise emanating from within the control enclosure.

Figure 5, Pad Mount STS

Main Circuit Design

The main circuit of the STS comprises the power electronic switches themselves, as well as the supporting breakers, and isolation switches, and arrestor type over-voltage surge protection. The arrangement of the supporting breakers, and interrupters is application dependent, and may require a complete set of bypass and protective breakers, as well as isolators to allow maintenance on portions of the switch, while other components are live. It is also important that the switch, breakers and isolators be easily integrated into exiting equipment line-ups including the required cooling facilities as well as supporting the facility for outdoor application such as dead-front pad-mounting. Figure 5 shows a view a pad mount STS application for operation adjacent to a 69 kV substation feeding several 12.6 kV circuits. This outdoor application has facilities for complete bypass, isolation, and breaker protection in standard size 15 kV switch-gear line-up equipment. There is nothing very special about this arrangement, other than the provision of PT, CT and CPT equipment to make the STS completely self contained. Other than communication signals, the STS is completely self contained, obtaining power directly from its 15 kV sources for operation of all controls. The CPT power is redundant and obtained from each source with provision to transfer control power on loss of feeder voltage.

The power electronic components are of very special design. The switch design must be carefully considered in order to meet the voltages that will be applied both from internal and external effects of operation, and for normal as well as abnormal service. The current rating of the switch together with the design of the snubber, defines the voltage environment for the switch for internal effects. It is essential that the switch be designed to meet the worst case voltage resulting from such service. Normally, when either switch is properly conducting, and carrying up to rated current (with no delay in firing), the switch voltage drop should be purely the sum of diode drops associated with the number of thyristor elements in series; essentially one volt per diode drop. This is true even on the non-conducting switch if the phase angle between the sources is small. On switch blocking, from an initial conducting state, the current in the switch will attempt to continue to flow as the current in a circuit cannot be interrupted instantaneously. Also the thyristor cannot establish a blocking state without the flow of reverse recovery current. The transient process of reaching a blocking state requires the switch to go from essentially zero voltage up-to rated voltage with essentially equal division between levels. The peak voltage reached in this process leads to transient over-voltage before returning to a fully blocking state. The maximum voltage on each level is controlled by the snubber design and the number of thyristor in series, and by the effect of any varister elements that might be in parallel with each level or in parallel with the switch.

The worst case external event of over-voltage of an RLC circuit is typically twice rated voltage for a single event. This situation is also true of the STS, except for special circumstances such as ferro-resonance which may occur because of the effect of PT and CPT magnetizing reactance and saturation, and snubber capacitive reactance. Ferro-resonance of this kind may result when the switch is unloaded, such as a line-drop on the alternate source, with the snubber and PT being driven from the voltage of the preferred source. A more typical situation will be a line down on the preferred source, causing transfer due to low voltage, resulting in resonance driven from the alternate source. While over-voltage from this source cannot be totally avoided, the persistent effect of resonant over-voltage can be avoided by choosing the PT design carefully. Standard rated CT and PT arrangements cannot generally be used in this application because of this fact.

STS Operation

This STS is connected to two 12.6 kV feeders fed from a common 69 kV bus via two 69/12.6 kV transformers. The load side of the STS is fed through a 12.6 kV to 480 volt Y/Y transformer to two 500 kW load banks in parallel, each switched in 25 kW increments. The 12.6 kV side is also connected to a utility load in parallel with the load bank transformer. During initial commissioning, this arrangement provided the means to transfer to the utility load with current already flowing in the switch. This expedient allows the load bank to be removed incrementally to be sure that there are no start-up disturbances in this environment when there are only manually operated by-pass switches available.

The switch has operated down to 8 amperes load current, with more than adequate signal to noise in the control input signals for current tracking firing control purposes, and has been transferred manually and automatically many times.

The operation of the switch, to demonstrate automatic transfers has been affected so far by the expedient of removing a voltage signal from the operating side of the switch, to induce the control into thinking an actual low-voltage condition exists on this phase. The switch is then called to perform a transfer operation. Because there is no actual low voltage condition, the transfer proceeds in the fashion of break-before-make.

Figure 6, Automatic Transfer Test Operation

Figure 6, provides a plot taken from a power quality meter of the sources and load phase ‘a’ voltages and currents during a synthetic transfer event. The load voltage is taken from the low side of the 12.6 kV/415 v transformer. It can be seen in this plot, there is a slight delay in the start-up of the second source onto the load, following current zero in the source going out. This delay can be minimized, but is required, to properly distinguish the transfer point when it is initiated close to a current zero, and especially when there is a phase angle difference between source voltages.

The performance of the STS for loss of voltage is demonstrated in the plot of Figure 7, for down-stream three-phase bus voltage and phase ‘a’ current. Here the preferred source was tripped by opening an 12.6 kV oil circuit breaker in the supply sub-station. The reduction in voltage was sensed by the control which initiated the make-before-break transfer sequence. The initiating event as seen by the PQ meter output, for the protected downstream 125 volt circuit, is seen to be the reduction of the ‘b’ phase voltage. The event disturbance is seen to be very small and the complete transfer, of all three phases, is over in about sixty degrees of the fundamental wave-form, with essentially no disturbance in subsequent fundamental wave-form. The harmonic content of the voltage wave-form immediately after switching is a result of the lost volt-seconds (flux) in all of the down-stream supply transformers, due to the voltage disturbance. Such voltage disturbance after switching is normal and exists transiently until the transformer flux is reestablished. The STS minimizes this effect by providing rapid transfer.

Figure 7, Automatic Transfer on loss of Preferred Source

Commissioning a complex electronic system like the STS is a challenge because all of the sub-systems must be working correctly at the point of full voltage application. Reduced voltage application is not practical, and failure to work correctly due to misalignment of controls can mean expensive component failures.

The first commissioning of the control was performed in a novel manner that proved the functionality of all of the controls and monitoring before the switch actually carried current. The method was to apply the actual switch voltages and control power, but have the control also use the voltage signals as a substitute for current signals. This expedient allowed the control to think that it was in actual operation, and allows the alignment of the firing pulses to current zero, the phase-out of the switch, simulate transfer operation, and check for correct operation of all of the monitoring for gate reception and temperature measurement, all before seeing normal voltage for the first time. With this expedient, the first operation was carried out without incident.

Conclusions

The STS is seen to be a relatively complex device from a control and monitoring point-of-view. This complexity is offset by the fact that there are many, many years of experience with thyristor application where the reliability issues have needed to be addressed in order to have a viable product, not-with-standing complexity and the large number of components.

The main contribution in the STS, to thyristor application technology, are the algorithms that make firing control possible without delays at current zero crossings, and the near instantaneous detection of fault direction to inhibit transfer of down stream faults. The enabling technology for this is the modern floating point digital signal processor, especially those that can be worked in parallel to share the volume of calculations. The other valuable feature of the DSP is the ability to replace much of the analog signal electronics of yesterday with digital computation. This effect is a major help in increasing overall reliability of the application by reducing the number of discrete components.

References

"Application Considerations for Power Quality Equipment & Solutions for Distribution Networks", Colin EJ Bowler, Vinod Bapat, Robert Hirt, Power Systems World 1997, Baltimore MD September 6-12 1997.