DC fault ride-through capability and STATCOM operation of a HVDC hybrid voltage source converter

: High voltage direct current (HVDC) transmission systems are becoming increasingly popular when compared to conventional AC transmission methods. HVDC voltage source converters (VSC) can offer advantages over traditional HVDC current source converter topologies; as such, it is expected that HVDC-VSCs will be further exploited with the growth of HVDC transmission. This study presents the DC fault ride through capability and new static synchronous compensator (STATCOM) modes of operation for the recently published alternate arm converter, intended for the HVDC market. Operation and fault ride through of the converter during a local terminal to terminal short circuit of the DC-link is demonstrated; during the fault STATCOM operation is also demonstrated.


Introduction
Integrated gate bipolar transistor (IGBT)-based voltage source converters (VSC) are being actively developed for HVDC applications. These systems offer advantages over, the now mature, line commutated conversion methods used in classical HVDC; these advantages include improved AC side power quality performance without the need for large harmonic filters, ability to exchange both leading and lagging reactive power with the connected AC network, a non-reversing DC cable voltage, ability to feed into a dead load and smaller physical footprint [1,2].
Successive generations of HVDC-VSC schemes have been developed, these include hard switched series strings of IGBTs operated in pulse-width modulation (PWM) mode [3,4]. In addition, modular multi-level converter M 2 C topologies are being commercially developed and finding initial application [5][6][7][8]. Furthermore, hybrid M 2 C topologies are being actively developed and promise additional benefits, such as increased efficiency, decreased number of sub-modules and lower sub-module capacitor rating [9][10][11][12][13]. This paper describes DC fault ride through capability and new concepts of STATCOM mode of operation, that exploit the hybrid topology of a recently published VSC topology [14] intended for HVDC power transmission.
Traditional VSC topologies suffer from poor DC fault performance that can result in the complete disconnection of the converter by opening the AC side circuit breaker [9,15]. The topology presented offers a significant improvement, enabling the converter to remain connected to both the DC and AC networks even in the presence of a zero impedance DC fault. During the fault the converter is additionally able to exchange reactive power with the AC network. These factors not only help to stabilise the AC network, but significantly reduce the time needed to restore normal power transmission capability after the DC fault has been cleared. Fig. 1 shows the alternate arm converter (AAC) published in [14]; the converter is formed from six limbs each comprising a director switch, a limb inductor and a cascaded string of H-Bridge sub-modules. The sub-module arrangement is shown in more detail in Fig. 2. The director switches, shown as single IGBTs, would likely be formed from a series string of zero current soft switched devices; this would provide the necessary voltage rating required for HVDC applications. Unlike the standard M 2 C each limb of the AAC must be able to produce a negative voltage; hence the requirement for H-Bridge sub-modules. The H-Bridge sub-modules also provide this converter with a DC fault current blocking capability; this is achieved in the same way that the M 2 C converter with H-Bridge sub-modules is able to extinguish DC faults [16,17]. This fault blocking capability is true even in the event of a complete short circuit across the DC-link, which would typically result in large AC and DC fault currents and normally requires disconnection of the converter by the opening of the AC side circuit breaker. If a major DC voltage depression occurs on the DC bus or there is a complete DC network short circuit, the converter is able to continue to operate as a STATCOM providing voltage support to the AC network www.ietdl.org whereas AC/DC real power transmission is disabled, real power transmission can be restarted after the fault has been cleared or isolated from a multi-terminal DC grid.

DC fault ride through and STATCOM operation
A unique feature of this topology is that it offers multiple STATCOM modes during fault ride through. These STATCOM modes allow the converter to operate without supplying current into the fault site and therefore avoid the need for any physical disconnection.

Normal operation
Under normal operating conditions the converter is controlled using a method presented in [14]. Using this method the six director switches, shown in Fig. 1, open and close synchronised with respect to the line voltages; such that the upper and lower limbs in any given phase alternately conduct. Arbitrary voltage waveforms may be produced at the point of common coupling by appropriate modulation of the cascaded H-Bridge sub-modules in a conducting limb. Multi-level modulation schemes are well documented and any of a number of strategies can either be used directly or adapted for use with this converter [18,19]. During normal operation the non-conducting limbs also act to produce the target waveforms, in this manner it is ensured that the director switches do not become reverse biased and the voltage stress on the director switches is minimised.
In this work during normal operation, each phase of the converter is operated such that the commencement of the conduction period of one limb is advanced with respect to the ending of the conduction period of the opposing limb, thus there are two periods every fundamental cycle when both the upper and lower limbs conduct. This period, where both limbs in a phase conduct, is referred to here as the overlap period; its introduction allows the introduction of a circulating current, I circ , which conducts through the DC-link and the conducting pair of limbs during the overlap period. The introduction of a circulating current permits a variable ratio between the AC converter voltage, v C , and the mean DC-link voltage v DC [14].
It is assumed that the aim of the converter is to synthesise sinusoidal target waveforms on the AC network side of the transformer as in the following equation If the voltage drop across the limb inductor, L L , is ignored then the voltage the upper cascaded sub-modules must produce, in order to synthesise the target waveform, is given by (2); similarly the lower cascaded limbs must produce the voltage is given by (3). It should be noted that these are the required voltages for both conducting and non-conducting periods. This ensures the director switch diodes are reverse biased during the non-conduction period, as stated in the introduction The AC side line currents are assumed to be sinusoidal with an arbitrary phase shift α as given in the following equation During the overlap period an additional voltage is added to both the upper and lower limbs, such that a net voltage appears across the limb inductors which does not disturb the target AC voltage. This net voltage can be added or subtracted as required such that the DC circulating current, I circ , is controlled. The steady-state value required for this circulating current can be calculated by considering the change of energy stored in the converter over a fundamental period. If one considers phase x of the converter the change in the converter stored energy, W x C , over a fundamental period can be expressed as in (5). In steady state this must be forced to equal zero; assuming the overlap periods are centred around the zero crossing of the converter voltage,  the required circulating current can be expressed as in (6). This equation demonstrates that the introduction of a circulating current permits the additional degree of freedom that allows a variable ratio between the converter voltage magnitude and the DC-link voltage. As a result the converter is able to operate around a PQ envelope with independent control of both real and reactive power where φ OL is the overlap angle

STATCOM operation
A special feature of this converter is that it can be operated in multiple STATCOM modes. The first of these modes shall be labelled STATCOM 'A', in this mode the converter operates using the normal operation mode described in the previous section; however the real power demand is set to zero. A drawback of this mode of operation is, current must flow in the DC-link capacitor and therefore may spill into the parallel fault site. This can be explained if one considers the converter conduction path during normal operation or STATCOM 'A' mode when there is a DC-link fault. If the conduction path when both the upper limbs from phase A and C and the lower limb from phase B is conducting is considered, then the conduction path between phases A, B and C is given, that is, shown in Fig. 3. When operating in these STATCOM modes, although only either the upper or lower set of limbs is conducting at one time, the non-conducting set of limbs must be appropriately modulated to ensure that the diodes in the director switches remain reverse biased. This ensures no conduction path from the top to bottom limbs.
As the converter can be operated as two separate STATCOMs it is useful to transition from one STATCOM state to the other, that is, from STATCOM 'B1' mode to STATCOM 'B2' mode. This enables the distribution of the conduction time and associated losses to be shared equally in the converter, whereas also enabling the capacitor voltages in

Control strategy
In normal operation and when operating in the STATCOM modes previously described, the converter is able to produce arbitrary voltage waveforms on the AC side. This allows the line current to be controlled using traditional VSC methodologies. For this work a dual sequence control scheme has been implemented. This scheme provides independent control of both the positive and negative sequence line currents in separate dq reference frames [20,21]. The implementation of the control scheme is shown in Fig. 6; the superscript dqp is used to denote positive sequence dq components and the superscript dqn negative sequence components.
During normal operation to ensure net AC/DC power transfer is maintained the DC-link current, i DC in Fig. 1,i s regulated by the addition of a circulating current as discussed in Section 2.1. Any transient imbalances in AC/ DC power transmission will give rise to a change in the net energy stored in the sub-module capacitors. To ensure this energy is regulated correctly, the instantaneous AC power demand is trimmed to regulate the net limb capacitor energy.

Simulation results
Simulation results have been obtained using MatLab Simulink with the PLECS blockset; simulations are based on ratings for a 20 MW 11 kV low-power demonstrator. The model of the converter has been implemented using ten H-Bridge sub-modules per limb; naturally for systems rated at transmission level powers many more sub-modules would be used, resulting in high quality waveforms. The VSC is operated as to maintain a charge on each H-Bridge sub-module capacitor of 1750 V; this is representative of the voltage limitations of presently available IGBT modules. A carrier-based modulation strategy, adapted for the HVDC-VSC, has been used to modulate the H-Bridge sub-modules [19]. In order that the capacitor voltages remain balanced, a simple algorithm has been used to determine the firing order of the sub-modules in each limb. This is not discussed further here, since techniques that can be used directly or adapted have been previously published [10,11].
Results are presented demonstrating fault ride through in the event of a local DC terminal to terminal short circuit fault. To assist in modelling of the fault, a stiff supply is used to represent the fault voltage.
Prior to the fault real power is being exported to the AC network from the converter, whereas reactive power is exchanged with the network. During the fault real power transfer is disabled and the converter operates as a STATCOM. Results are presented for STATCOM 'A' and 'B' modes discussed in Section 2. Table 1 lists the main parameters used for the simulation.  consequently, demanding the opening of the AC side breaker. It will be shown in the simulation results, that there is no large spike in the AC currents and after a short transient sinusoidal line currents are restored. Additionally, the reactive power demand is increased to provide further support to the AC network. Fig. 7 shows the DC side voltages and currents during the transition from normal operation to a terminal to terminal short circuit fault. The top plot shows the voltage across the DC-link capacitors and the voltage at the fault location; v DC and v cab , respectively, as labelled in Fig. 1. The fault voltage applied is a step at time 0, from the rated DC-link voltage of 20 kV down to 0 kV. The voltage response across the DC-link capacitors is dominated by the filter components and not by the response of the converter, therefore only one set of voltage results is presented, shown here is the case of transitioning from normal operation to STATCOM 'A' fault ride through. The middle plot and bottom plot show the DC converter current and the DC cable current, i DC and i cab , respectively. The middle plot shows the response when transitioning from normal operation to STATCOM 'A' mode, whereas the bottom plot shows the transition from normal operation to STATCOM 'B' mode for fault ride through.
When the fault occurs there is a large reversal in the cable current, i cab . This response is a function of the filter and determined by the filter parameters. The filter output is short circuited during the fault; thus, the rated DC-link voltage of 20 kV is applied across the filter inductor causing the reversal in the current. The current freewheels through the fault site until all the energy stored is dissipated in the damping resistor. It should be noted that in neither modes of operation is this current seen by the semiconductor devices. This can be attributed to the series combination of director switches and cascaded H-Bridge sub-modules. When the DC-link capacitor voltage, v DC , becomes negative the capacitors in the H-Bridge sub-modules naturally oppose the negative DC-link capacitor voltage and prevent the diodes, labelled d 1 and d 2 in Fig. 8, from conducting. Ensuring that control of the converter is maintained during fault ride through. This further illustrates the benefits of this arrangement.
When operating in STATCOM 'A' mode it is seen that the DC converter current, i DC , has a large ripple approximately 800 A Pk-to-Pk, most of which is absorbed by the filter; however, the current which is not absorbed by the filter capacitors, i cab , spills into the fault site. This is seen to be ∼ 300 A Pk-to-Pk. By inspection of the bottom plot in Fig. 7 it is observed that after a short transient of half a  www.ietdl.org fundamental period the transition to STATCOM 'B' mode is made. It is further observed that there is zero DC-link converter current and consequently no current spills into the fault site. Fig. 9 shows the AC side response to the local DC fault, the same time scale as Fig. 7 has been used. The top plot shows the AC supply voltage, this is modelled as a stiff source and so is unaffected by the DC fault. The middle plot shows the line current when transitioning from normal operation to STATCOM 'A' mode, whereas the bottom plot shows the line current under the same fault conditions when transitioning from normal operation to STATCOM 'B' mode. Inspection of the results reveals that in the event of a DC-link short circuit there is no spike in the line current. Thus using either method of fault ride through there is no need to open the AC side breaker. Furthermore, it is shown that after a short transient of approximately one fundamental period, sinusoidal line currents are restored. After a second fundamental period the line current is increased to rated current to demonstrate delivery of maximum reactive power; although any reactive power operating point can be chosen, provided rated conditions are not exceeded. Thus, not only is there no need to open the AC side breaker, but further to this the AC network can be supported by operating as a STATCOM. This is a clear advantage of this topology over traditional HVDC-VSC converters. Fig. 10 shows the average capacitor voltages in phase A when transitioning from normal operation to STATCOM 'A' and 'B' modes. The figure shows the average capacitor voltage for the upper and lower limbs, in the interests of clarity only phase A is shown. It should be noted that this plot uses a different time axis to the previous plots to highlight the longer response time associated with controlling the capacitor voltage. The longer time is also required in order that the switching from STATCOM modes 'B1' to 'B2' be shown.
When the converter first enters STATCOM 'B' mode immediately after the fault, only the upper limbs conduct thus the stored energy in the bottom limbs remains the same. Twelve fundamental periods after the fault, the converter transitions from STATCOM 'B1' to STATCOM 'B2' mode, the lower limbs begin to conduct and the upper limbs stop conducting. Thus the capacitor energy remains the same in the upper limbs and the bottom limbs exchange reactive power with the network. It can be seen that there is little disturbance in the capacitor voltages in the transition between STATCOM modes.

Conclusions
New modes of STATCOM operation have been identified for a recently published and practical VSC topology for HVDC power transmission. These new modes allow the converter to remain connected to the AC network and additionally operate as a STATCOM in the event of a DC network fault. Thus enabling DC fault-ride through without the need to open the AC side circuit breakers, a problem traditionally associated with HVDC-VSCs. This results in the elimination of post-fault down time, reduction in AC and DC side fault currents and AC network voltage stabilisation during the disturbance.
Simulation results have shown the fault ride through capability of the converter based on 20 MW demonstration equipment. Additionally, simulation results have verified the new concepts in STATCOM mode of operation presented here. Successful transition between STATCOM modes has been further demonstrated; this has been done whereas still maintaining independent control of the bulk energy stored in each of the capacitor limbs.