I'm building a grid-tie inverter of my own. I have yet to prototype the unit but if the specs of the components I'm using on my design are any indicator, my unit shall have the following specs...
Voltage = 120VAC
Peak Apparent Power = 2400VA
Peak Current = 20A
Duty = 100%
I'm quite proud of the design. I've adapted the diode-clamped multilevel design for use on 120VAC systems and use fewer components than prescribed in the literature. The topology is a multilevel H-bridge hybrid. The beauty of it is that like a multilevel diode-clamped inverter it also supplies VARS but with FAR fewer parts. Remarkably the bill of material is close to $150 without the need for large transformers, excessive heating of parts or injecting loads of harmonics back onto the grid.
I can't seem to add pictures here so I'll describe the circuit as best as I can. There are several stages to the design. The circuitry is far simpler to grasp in schematic form, believe me.
1) Boost Converter DC-DC. Whatever the DC source you're using, you need to boost the signal up to peak dc grid voltage (170Vpeak = 120Vrms) and beyond. You will need to control the converter so that output power (to the next stage) is maximized depending on the available power at the input (wind, water, solar etc.)
2) Multi-Buck DC-DC converter. You convert the approximately 170VDC into several smaller DC voltage levels with buck converters. The more levels you use, the higher the current rating you will achieve overall and the less the THD (total harmonic distortion) on the output. I'm using 8 levels but the minimum you would want to use is 3 to significantly reduce the harmonics you would get with a PWM. The additional cost per level is about $10. Each voltage level is separated by a large 1000uF capacitor to form a ladder.
3) Switches. The output voltage of each level (in stage 2) connects to a common bus (stage 4) via 2 opposing N-channel mosfets (per level) with integral clamping diodes. This is CRUCIAL since this will allow you to supply and consume VARS which you need to do if you want to run motors, inductive loads and electronic power supplies. To save the cost of driving each mosfet you connect the mosfets such that you can use a high-side mosfet driver to turn both of them on simultaneously regardless of which side of the switches has the higher voltage.
4) Bus. The outputs of each pair of switches in stage 3 are shorted together, filtered with a capacitor (more on that in a minute) and then fed into an h-bridge using higher rated mosfets.
Principle of operation:
You convert your raw input power into a stepped up DC voltage. Use this voltage to both transmit your power to your inverter (since your copper losses will be significantly reduced compared to if you send 12V power to your inverter) and modulate the inverter power going back to the utility. The voltage of each stage 2 element is proporational to the secondary voltage of stage 1. A controller circuit (easlily done) would turn the switches of each successive level on and then off again forming a staircase waveform on the bus with as many "steps" as there are levels. This approach has the distinct advantage of reducing commutation loss that high frequency switching PWM inverters use which inevitably over-heat your switch and inject unneccessary harmonics into the grid. Several switches are used, switch on and off at the grid frequency (60 Hz where I live),share the I2Rdson losses evenly between eachother and eliminate the need for output filters. It is important to ensure that only one level's switch be on at a time to avoid a multilevel short circuit. A smoothing capacitor is placed on the bus to provide intermediate power during commutation. The H-bridge commutates when the step waveform reaches the crossover voltage and the cycle continues.
The strategy I'm using for synchronizing the output voltage waveform is by means of a microcontroller measuring the time between a half-cycle for each half-cycle and then starting the staircase waveform at a zero-crossover of the output/grid waveform (triggered by an op/amp to interrupt a microcontroller) to last as long as the previous half cycle. This way the circuit is immune to the smallest frequency variation on the grid's waveform, is easy to implement and gives total control of the phase angle. This is done with a tiny transformer (a spare ac adapter sitting around the house will do) between the line and neutral with the secondary connected to control voltage circuitry, being cleaned up with op-amps and fed into a microcontroller.
I realize this sounds complex but if you saw the circuit in schematic form, it is fairly elegant and simple considering what the circuit is being asked to do. I would not even think of using an isolation transformer as sourcing one is a big problem for most people. PWM in my mind is out of the question as it makes the output power very "dirty" and forces all the output power through a single switch, forcing you to pay for higher rated components whereas with the above approach you would buy more components but with lower required ratings.
If anyone is interested please post and I'll send a schematic.