Inverters lighten up

An electric vehicle is like a competitive runner in one regard: Every pound it must carry is a pound that forces it to consume more energy. No surprise, then, that there is a concerted effort in industry today toward coming up with EV components and systems that are as light as possible.

The trend toward light components has even affected inverters. Many traditional inverter designs incorporate magnetic components such as transformers and line reactors that can weigh tens or even hundreds of pounds. Thus there is an incentive to create inverter designs that don’t need to use these heavy and bulky magnetics.

Researchers have devised several such inverters that eliminate the need for large energy storage components. One in particular that has transitioned from the research lab to commercial use has its first commercial application not in EVs but in connecting solar arrays to the grid. Today six 30 kW inverters installed at the University of Texas at San Antonio convert up to ± 450 V from solar arrays into three-phase ac at 480 V that feeds onto the power grid. The topology will also be used for other applications including wind converters, bidirectional battery / vehicle chargers, and variable frequency drives for induction motors. Each 30 kW solar inverter weighs in at just 94 lb compared with conventional solar inverters with traditional magnetics that can be 1,200 lb or even more.

This solar inverter uses a topology that is a compact alternative to existing methods. It also eliminates the need for a bulky and lossy transformer. It is called a buck-boost, indirect-transfer, pulsed current-sourced topology. It is unlike ordinary inverters which normally use an input stage to charge a capacitor -- often called a dc link -- that serves as a power source for an output stage that creates an ac waveform. Instead, the new design uses a pure ac link formed by an inductor-capacitor pair between the inverter input and output. The link charges from the input(s) and discharges into the output phases via a modulation scheme that results in high efficiency, unity power factor, and low harmonics.

It is useful to compare this new kind of inverter with traditional techniques. Most commercial inverters for photovoltaic applications include a transformer and several sections of power conversion. Without transformer isolation, there is a galvanic connection of grid and PV-generator and a leakage current may flow through the capacitance between PV-generator and ground. The magnitude of this capacitance depends on environmental influences. For example, it will be large when the PV-generator is covered with salty fog forming a conductive path to the grounded metallic frame of the PV modules.

Page 2 of 3

Inverters that incorporate a transformer either use a 60 or 50-Hz ac mains transformer or a medium-frequency 20 to 50 kHz transformer. One widely used topology consists of a boost dc-to dc converter that provides a dc link for the inverter. But in PV installations, the dc-to-dc converter is not electrically isolated. This necessitates precautions for safety and potentially produces EMI in the output resulting from switching harmonics on the input side.

One approach to avoid these problems is to use an isolation transformer. One typical method is a scheme where a dc-to-ac converter converts the power from the PV array to a low-voltage ac which is stepped up via the transformer. Another solution that has been proposed is a simple voltage-source inverter. However, this can only work if the input dc voltage is kept higher than the peak-to-peak grid voltage. At least in PV uses, this is difficult to realize because of the variable voltage coming from the PV array.

All of these solutions require expensive and lossy output filters. The inverters involved are complex and there is a need to simplify and make improvements in losses, costs, weight and size.

In contrast, the soft-switched, direct connect topology overcomes the shortcomings of the above-mentioned topologies. It’s a pulsed current-source dc-to-ac converter. Inputs never connect directly to the output; Hence, there is inherent isolation between the ac and dc stages, which allows neutral grounding of both the ac and the dc sides. This means there is no need to use an isolating transformer.

The input side of the converter uses four reverse blocking switches such as the series-diode/IGBTs shown in the accompanying figure, or even reverse-blocking IGBTs or GTOs. There are two fully bidirectional switches per leg of the output-side converter. The common element between the two sides of the inverter is an ac link formed by a low-reactive-rating inductor-capacitor pair. In a completely indirect energy transfer operation, the dc input from the PV panels charges the link inductor. This inductive energy then discharges to the output phases. The output current pulses are controlled precisely using an energy distribution scheme that ensures less than 3% harmonics below twice the link frequency.

It is possible to realize a link frequency of at least 5 kHz with the semiconductor power switches available today. In the 30 kW inverters installed at the University of Texas, the link frequency is 7 kHz at full power. It goes as high as 20 kHz as power drops to zero. The power cycle frequency is twice this, or 14 kHz at full power, because there are two power transfers per link cycle.

Page 3 of 3

All such devices in this inverter turn on at zero voltage. Turn-off losses are low because there is a capacitive buffer across each switch. The converter is essentially a PWM (pulse-width modulated) current source, although all link current flows are ac with no dc offset.

In operation, input switches are turned on to charge up the link. The link is then allowed to resonate partially, which lets it swing to the voltage of the output phase to which it will discharge.

The link nominally discharges to two output phase pairs. The sequence and the pairs are calculated so as to minimize the partial resonance times while meeting the desired harmonic levels.

The best way to visualize the converter operation is to follow what happens through its six operating modes. Mode 1: During this mode, depending on the polarity of the link current, S1i and S2i or S3i and S4i turn on charge the link. For the example in the nearby diagram, switches S1i and S2i are turned on. The link current rises to the amount needed to extract maximum power from the PV panel.

Mode 2: At the end of mode 1, all the switches are turned off and the converter enters its second mode. Here, the link resonates partially until its voltage swings to the first output phase pair. The link then discharges into the grid in two different modes to proportionately power the three output phases. The inverter uses the fact that the instantaneous sum of the three phases is zero.

We’ll give an example that simplifies things a bit by assuming that the output filter results in zero phase shift between voltage and current and that link frequency is much greater than the line frequency. Suppose instantaneous phase voltages are Van=100 V, Vbn=-70 V, Vcn=-30 V and the link charges to 10 A during mode 1. The link would then discharge to lines AB and AC to equivalently supply 7 and 3 A respectively. This makes the output currents in phase with the output voltages, resulting in unity power factor, thus extracting the maximum real power from the PV cells.

The output discharge happens in order of decreasing instantaneous line voltage to minimize the partial resonant periods. In real life, the actual algorithm removes the assumptions made to handle real scenarios.

Mode 3: Previously enabled output switches, corresponding to the selected phase pair, turn on at zero voltage as they become forward-biased by the rising link voltage. The link then discharges to the output until specific system-generated references are met. Again, the references are generated so as to give unity power factor at the output while maintaining strict harmonic levels. In the nearby example, switches S1o and S6o turn on to let the link discharge to phase BC.

Mode 4: Switches turn off to let the link resonate until its voltage equals that of the second output phase pair into which it will discharge. Mode 5: Switches become forward-biased to let the link discharge to the second phase pair. In our example, switches S4o and S5o are enabled and then turn on to let the link discharge to phase AC.

Mode 6: The link is allowed to partially resonate back to the input voltage to start the next charging cycle. Then Modes 7 through 10 are identical to modes 1 through 5, except the link current is in the reverse direction.

At full array voltage and power, the dc input voltage is nominally equal to the ac average output voltages. At reduced PV array voltage and power, as when clouds pass overhead, the converter boosts the array voltage to the required ac output voltage. The inverter can raise or lower the current from the array for MPPT (maximum power point tracking) by controlling the amount of current charge that the link receives from the input.

Evaluating the results
Simulation results for a 25-kW, 575-V utility grid converter show what the waveforms in various parts of the circuit look like. The simulation results shown here looked at the case where the PV array is 400 V and 200 V either side of ground. Simulations also assumed a grounded neutral bipolar array and a grounded neutral three-phase output. A careful look at the link voltage and currents shows the buck capability of the converter. After charging from the input capacitors, the voltage swings to a lower output phase pair to discharge.

Similarly, the boost capability of the converter is also evident. After charging from an input of ±200 V, the link swings to a much higher output voltage to discharge. The converter’s inherent buck-boost capability removes the need for a separate dc-dc stage that directly impacts efficiency. And, of course, the input never directly connects to the output, so there is no need for a transformer.

Finally, the output voltages and filtered output currents with a ±200 V input show some distortion at points of voltage crossover. We think slight imperfections in the switching control are responsible for these and that further work can mostly eliminate these low amplitude harmonics.EE&T