Over View and Technical Analysis of Multilevel Converters

1. Unified Converter Theory

In the preface of his book Switching Power Converters, Wood introduces the concept of a unified converter theory. There he states: “Most traditional views of the field have seemed somewhat disjointed; converters were largely regarded as related only because they all use semiconductor switches and have certain topological similarities. . . . the view expounded herein (is that) switching power converters are related by function and behavior; their basic characteristics do not in any way depend on the types of switches used, nor on the applications to which they are put, nor on the topologies in which they are realized.”. According to this unified theory, any power electronic converter can be viewed as a matrix of switches which connects its input nodes to its output nodes. These nodes may be either DC or AC, and either inductive or capacitive; and the power flow may be in either direction. Two obvious restrictions are enforced by some basic laws of electricity.

• If one set of nodes (input or output) is inductive, the other set must be capacitive, so as not to create a cut set of voltage or current sources when the switches are closed.

• The combination of open and closed switches should never open circuit an inductor, or short circuit a capacitor.

2. Inverter or Rectifier? Voltage or Current Source?

This unified set of converters is generally broken into a number of subsets. The term rectifier is used when the power flow is predominately from the AC port to the DC port and the term inverter is used when power flow is predominately from the DC port to the AC port. The term converter is used either when there is no predominant direction of power flow or as a general term to encompass both rectifiers and inverters. In a Voltage Source Converter (VSC), the DC port is the capacitive port and is voltage stiff (i.e. a large DC bus capacitor). The voltages in such a converter are well defined by this port and are generally considered independent of the converter’s operation. The value of the AC side inductance is comparatively small and modulation of the converter controls these AC side inductor currents. Should the voltage source converter be responsible for the control of the DC bus capacitor voltage, then this voltage is indirectly controlled by controlling the net current flow in the capacitor.

The switches in such a converter must block a unidirectional voltage, but be able to conduct current in either direction if bidirectional power flow is desired. The converse is true in a Current Source Converter (CSC) — the DC port is inductive and current stiff. The current in this port (and hence the converter) is well defined and slow to change. The voltage (particularly at the AC port) is considered the variable directly controlled by the converter modulation. Since the AC port usually has significant line or load inductance, line to line capacitors must be placed on the AC port. The switches must block either voltage polarity, but are only required to conduct current in one direction. This naturally suits thyristors and symmetrical GTOs.

Figure 2.1. A voltage source rectifier - inverter cascade (top) and a current source rectifier - inverter cascade

Since the AC line and AC motor loads are both inductive, Voltage Source Rectifier – Inverter cascades (Fig. 2.1) are usually used for small and now increasingly for large motor drives and similar applications, as GTOs and IGBTs have matured. Larger converters have traditionally been current source converters, both because this best suits the characteristics of the thyristors and because it requires a large DC bus inductor, which was preferred to a large capacitor. Some converters do not easily fall, or cannot be placed into either category. The matrix or Venturini converter [1] is one example (Fig. 2.2). Both input and output ports are AC, and the definition of voltage stiff or current stiff (and hence voltage or current source) becomes somewhat arbitrary. Both input and output ports are

Figure 2.2. The matrix converter, with one possible implementation of the bidirectional switches.

3.  The General Multilevel Converter

The next refinement is to define the meaning of multilevel. The following definition of a multilevel converter is offered:

A multilevel converter can switch either its input or output nodes (or both) between multiple (more than two) levels of voltage or current. The term “two-level” will be used where it is necessary to refer specifically to a converter which is not multilevel. This simple definition is deliberately quite broad and inclusive, in keeping with the spirit of the unified converter theory. For example, the multi-phase matrix converter (Fig. 2.2) is, strictly speaking, a multilevel converter, according to this definition. Consider the three phase to three phase matrix converter, with voltage source inputs and an inductive load. Any single output can be switched to one of three different voltage levels (the voltages of the three input phases) and similarly, any input can be switched to one of four current levels (including zero). In this preceding example, both the input and the output nodes are AC periodic varying quantities and so these levels can only be considered stationary for an interval much shorter than their AC period.

Figure 2.3. The current source converter (top right), voltage source converter (bottom left) and a simple three level voltage source converter (bottom right) can all be derived from the general topology of the matrix converter

                Both the voltage source and current source converters can be derived from the general matrix converter by setting one port to be either a two terminal DC voltage stiff or DC current stiff port [70, 30]. Retaining the third terminal leads to a simple and more conventional multilevel converter (Fig. 2.3). Note that now one of the ports has been made DC and voltage or current stiff, only one port will experience the multilevel stepped waveforms. The other will still have a continuous waveform similar to that of an equivalent two level converter.

For example, a converter with an appropriate structure may create a stepped multilevel voltage waveform at the inductive nodes, but will always have a continuous voltage waveform at its capacitive nodes. Similarly a different converter may create a stepped multilevel current waveform at its capacitive nodes, but must have a continuous current waveform at its inductive nodes.

4.  The Traditional Multilevel Converter

The traditional understanding of what constitutes a multilevel converter follows this more narrow definition. One of the ports has multiple (more than two) voltage or current stiff DC nodes or terminals, while the second port has a conventional single or three phase set of terminals which are switched to these multiple levels.

Most multilevel converters discussed in the literature step between multiple voltage levels. This is usually the most useful configuration for a high power converter, as reducing conduction losses in both converter and machines will always favour increasing the voltage rating rather than the current rating of the converter. Also as power levels increase, the input and output voltage levels presented to the converter increase. The structures of these multilevel converters place the switches in series to share the duty of blocking these higher voltages. Equally however, for high current applications, many switches can be placed in parallel, with their current summed by inductors. When switched separately, multilevel current waveforms result. As expected, multilevel converters can be DC-DC, DC-AC and as explained, in the broadest sense, even AC-AC.

5.  Multilevel Topologies

Generally multilevel topologies can be divided into two groups, although in some cases the dividing line is indistinct. The first approach relies on summing the outputs of a number of conventional two-level converters, to produce a resultant multilevel output. The second group replaces the two-level switch structure with a multilevel switch topology within an otherwise conventional converter. These two groups will be distinguished by the terms multi-bridge converter and multilevel converter respectively. Any of the basic DC-DC converters (buck, boost, buck-boost, Cuk) can be extended to a multilevel topology. Often these are not called or perhaps even recognized as multilevel converters, but rather simply described as, for example, paralleled converters with interleaved switching instants. Two recent examples cited are multilevel boost converters used for power factor correction. In both of these examples, the switches are effectively placed in parallel and their contributions summed by separate boost inductors. They present multilevel current waveforms to the input and reduced voltage ripple at the output. Multilevel DC-AC converters range from the simplest single phase, full bridge driven with unipolar voltage switching to complex multi-phase converters. These are the most commonly recognized and reported multilevel converters and will be further categorized and referenced in the next section. Even multilevel AC-AC matrix converters have been shown to be at least theoretically possible.

6.  Three Phase Multilevel Voltage Source Converters

At this point in the chapter, we will narrow the focus to that of three phase voltage source multilevel converters. Although this may seem somewhat limiting, it encompasses most of the higher power multilevel converters both in the published literature and in actual use. There are some examples of single phase converters functioning as AC-DC switching rectifiers, either in traction, computer or telecommunications power supplies. These Power Factor Correction rectifiers have lower inherent distortion and require less filtering because of their multilevel topology. There are four main voltage source DC-AC multilevel topologies which have been distinguished here and in the literature.

These are:

Ø       Multiple bridge using transformer or inductor summing;

Ø       Multiple bridge using direct series connection;

Ø       Multilevel diode-clamped converter; and

Ø       Multilevel flying capacitor converter.

Each of these will be examined in turn. Each of the diagrams presented are of a five-level converter, which can produce a nine-level phase to phase voltage waveform.

7. Transformer/Inductor summed Multiple Bridge Converter

As the title suggests, these multilevel converters are simply a number of conventional two-level bridges, whose inputs or outputs are summed using transformers or inductors. The multiple transformer secondary’s force voltage sharing between the switches (Fig. 2.4). The most common and well known example of a multi-bridge converter is the twelve pulse thyristor converter, well covered in most power electronic textbooks [49]. Harmonic cancellation in these converters is achieved through the phase displacement of the voltage waveforms of the star and delta transformer secondary’s.

Figure 2.4. A five-level Transformer coupled multiple bridges, which produces nine level phase-phase waveforms on the transformer primary.

This 30? phase shift between transformer secondaries allows identical secondary switching instants and current waveforms to appear interleaved on the transformer primary. A series connection is used for HVDC; a parallel connection for high current applications such as electrolysis and electro-plating. The technique can and is extended to many bridges each with a transformer secondary connection of the appropriate phase shift to achieve cancellation of the further low order harmonics in the primary. By clever connection of the transformer primaries, current as well as voltage sharing can be ensured.

A good example of the next degree of complexity and flexibility is seen in a 10 MW battery energy storage plant. The GTO converters operate in square wave mode and still rely on the transformer phasing for harmonic cancellation. However because forced commutation is used; now both the magnitude and the phase (real and reactive power) can be separately controlled. An extension of this approach to 48 pulse operation is achieved by eight GTO bridges operating in square wave mode, with reliance on the transformer for harmonic cancellation. The cancellation of switching harmonics can also be achieved by switching strategies, rather than relying on the transformer secondary’s for the necessary phase shifting. The simplest case — the series or parallel connection of two PWM bridges — has been investigated by a number of researchers. By the use of appropriate PWM modulation for each bridge, the odd multiples of the PWM carrier and sidebands, including the first cluster, were entirely removed from the output spectrum. This improvement is better than can be achieved by merely doubling the carrier frequency as the carrier which remains has lower amplitude. A particularly good example of a six bridge, transformer summed multilevel converter is used as an active filter for arc furnace static flicker compensation [71].

The AC connections of these bridges are summed by separate transformer secondaries, which allow either a series or parallel DC connection. Since the transformer no longer provides phase shifting, it may seem possible to remove the transformer entirely and place the converters directly in parallel (for a parallel connection). However, while no difference exists between the desired input and output components of the two converters, the undesired switching components are by definition exactly out of phase. Kirchhoff’s laws would be violated if the converters were directly connected.

The solution is to use inter-phase reactors (current sharing reactors) or interphase transformers on either the input or output of the converters. Although these reactors see the full combined converter current (and so have similar copper volume and copper losses), they only experience the difference in voltage between the converters. The volts-second component of this voltage is smaller and so the iron content of these reactors can be reduced in comparison to the transformers which would be required for full isolation. Normally the inductors are placed on the AC side, which is already the inductive port of a voltage source converter. Research on a five level three-phase motor drive which used this technique was conducted by Matsui et al . The outputs of two half bridge legs were summed with a current sharing reactor to form a three level intermediate output. This and another similarly formed three level output were summed by a third reactor to form the final five level phase output. One further solution is to sum the outputs of two converters across a bridge connected source or load. Both ends of the transformer or motor winding are brought out and the winding must be fully floating. One converter is driven with a phase inverted signal, so that twice the desired converter output is impressed across the floating load. If the carriers are appropriately phased, part of the undesired carrier component will appear as a common mode component to the load. Of course, this technique can only be applied for two converters.

To summarize, the transformer or inductor summed approach has the following advantages:

• The voltages within the individual converters and thus across the switches are well defined by the stiff voltage source output of the transformer secondaries.

• Should a converter module fail, or be removed for service, the converter may continue operating at full voltage, but at reduced current. • Other than the transformer (inductors), the structure is modular, which allows easier maintenance and reduced spares.

• Its mode of operation is easily understood and, again because of its modular structure, control is more easily applied. but also the following disadvantages:

• The transformer itself, if not needed for isolation, adds significantly to the cost of the converter and is one more item to maintain and potentially, to fail.

• The transformer requires multiple secondary windings, which must be isolated from one another and from ground. This is a significant problem at high voltages. This also increases the cost of the transformer.

8.  Series Connected Isolated Multiple Bridge Converter

A second topology, which is really only a variation on the first, is that of series connected bridge converters        (Fig. 2.5). Each phase leg consists of series connected single phase full bridges, the series connection being made directly (not by transformer as in the first case) on the AC side. A three phase converter can be constructed by connecting three of these single phase series strings to form a star or delta. Since this topology requires each full bridge to have an isolated DC bus, this connection has not been considered useful until recently re-examined. Now this topology is being considered for applications where no real power transfer is involved, such as for active power filtering and VAR correction. Then only a floating DC bus capacitor is required on each floating DC bus.

Some other sources of power which could easily be made modular and floating are batteries for battery energy storage systems (BESS) used for load leveling, or alternative energy sources such as solar panels. It is of course possible to power the isolated bridges from multiple isolated transformer secondaries, each with their own rectifier . By appropriate phase shifting of the transformer secondary windings, harmonic cancellation can be achieved on the primary side, as described previously, as well as at the multilevel output of the multi-bridge converter. However the disadvantages of a transformer with multiple isolated secondaries return. This multilevel converter structure has some very significant advantages, if its limitations are acceptable.

Its advantage is it has perhaps the simplest architecture and the lowest component count. No transformer is needed, so capital costs are low.

9. Applications of Multilevel Converters

At this point it should be clear that one of the major advantages of a multilevel converter, regardless of topology, is increased power rating. A converter need not be limited in size by the prevailing semiconductor technology, since a multilevel converter allows the voltage and/or the current to be shared among a number of switches. This advantage has traditionally justified the extra complexity of multilevel converters only at very high power levels, for large motor drives and utility applications. As the understanding and acceptance of multilevel converters has increased, these converters are being used at all power levels to extend the useful power range of semiconductor switches. For example, using multilevel topologies, IGBTs are challenging traditional GTO converters in motor drive and traction applications and MOSFETs are displacing IGBTs in some larger Switch Mode Power Supplies. The more stringent harmonic standards now being legislated also advantage multilevel converters, since they produce lower switching harmonic spectral components for a given switching frequency limit.

10.  Conclusion

The aim of this chapter has been to demonstrate the diversity of possible multilevel converter topologies. Each has its own mixture of advantages and disadvantages and for any one particular application, one topology will be more appropriate than the others. Often, topologies are chosen based on what has gone before, even if that topology may not be the best choice for the application. The advantages of the body of research and familiarity within the engineering community may outweigh other technical disadvantages. Despite the diversity, these different topologies contain common underlying links. Usually the modulation and, to a lesser extent, control strategies can be developed independently of the converter’s topology and then subsequently applied with little or no modification. In subsequent chapters, the simplest case of the transformer connected multi-bridge converter will be used as the implied default multilevel converter topology. Required variations on modulation and control strategies will be explained after the general technique has been presented.