(much of this is a repost of the original fieldlines thread, but since the images in there are screwed up I decided to do a repost, and also include the latest comments. Original thread can be found here:
http://www.fieldlines.com/index.php/topic,130334.18.html)
Generally, I see using an MPPT is the easiest way of doing the most important aspect of a wind power system - load matching. Of course, good results can be also had with other methods of load matching(using capacitors, line resistance etc..), so MPPT is not really necessary, but it's easier and faster than the trial and erroring with other methods. If building multiple of similar machines, then it makes sense to use other methods of matching the load, as MPPT controller adds quite a bit to the cost and is another unreliability factor especially in lightning-prone sites. But for one off MPPT makes sense. Unless one has enough experience and doesn't need much fiddling to match the load without electronic devices
The main goal of the project was just in the fun of doing it - tinkering with power electronics. And at that time I saw the MPPT as the "holy grail" of diy electronics. My 3.2m axial turbine was also well suited for MPPT use, as I had engineered the blades to work at a design TSR of 7 through the whole length of the blade.. Not realizing at the time how badly an engineered blade would perform with direct battery charging. Unlike the popular design with angles and blade widths "off the cuff", the engineered blade will stall through the whole length at once. If you have a blade with improvised dimensions, or even a straight no taper no twist blade, there will be always some portions of the blade providing lift and giving much better overall performance with direct battery charging in mind. Before installing the MPPT, I logged a few weeks worth of power output and wind speed data. The graph I compiled from the data shows the blade performance dropping sharply after the blades hit stall - giving quite poor output in mid-range winds. Due to the low 12V cut-in speed of about 120rpm, the peak efficiency is obtained quite early, and by the mid winds the blades are stalled.
When studying various approaches to the MPPT controller I pretty soon decided to go with the most proven design, using rotor speed to determine the output power. This is not really MPPT in the literal sense, as the controller is not tracking the maximum power point, just meagerly doing what the rpm-to-outputpower algorithm determines. But with optimized loading curve it should very closely track the peak power, while avoiding all the problems of trying to implement a true MPPT on wind. Due to rotor intertia it would be very hard / impossible to achieve unless one had a very good wind site with very little sudden changes to wind speed.
Since the rotor on my wind turbine was loaded too much, and hence stalling, I needed a device that would allow the generator voltage to rise, in other words step down the voltage from the generator to the batteries. The simplest switched mode device for that would be a standard buck converter, so that's what I decided to use. At this stage I had all the bits together to make a general block diagram of the MPPT converter:
(bigger sized version of the drawing:
http://pics.ww.com/v/Janne/Electronics/schets.jpg.html)
The 5-phase output from the generator is first rectified(on the bottom of the tower), and one phase is brought to the conroller for determining the rotor speed from the AC frequency. Using a lookup table(that contains an output power vs. input rpm), the microcontroller generates a reference voltage with an external DAC. This reference voltage is then fed up to the TL494 pwm controller, which compares the reference voltage to the measured value of the controller's current output. The pulse width of the switching transistor is adjusted until the output current matches the desired value.
When I started designing the actual layout of the controller, I decided to split it into two main components, the bussed power stage and a separate controller board. Thinking of it now, it may had been a better idea to just build the power stage into a PCB as well. Using some application notes and information found on this site:
http://powerelectronics.com/power_systems/dc_dc_converters/power_buckconverter_design_demystified/ I decided to use 100kHz as the switching frequency, and then based on that calculated the values for the switcher components. Looking back now, it would have been better to use some lower f, maybe 20-30kHz, use bigger components but avoid a lot of problems caused by high frequency.
Ironing out all the problems indeed took a good couple of months, with leisure pace. I had the usual problems of blowing MOSFETs, and the isolating transformer wasn't also as easy to implement as it had sounded. Other bits of the controlling board were quite easy, I had previous experience with picaxe so I used that for the controller.
Testing the controller on a resistive load yielded a minimum efficiency of .8 up to output currents of 50A, and above .85 on output currents less than 35A. That seemed quite acceptable, considering the expected increase in rotor efficiency.
With the hardware problems sorted, I focused on the firmware. It is really quite simple, all the controller does is monitor the rotor speed, and use the programmed lookup-table to determine the desired output current. Rotor speed is derived by measuring the time between AC zero crossings. As a starting point for the lookup table, I just generated it based on the ideal power curve function.
I installed the controller to a steel enclosure. One reason for this was the electrical noise, the power stage was generating some amounts of RF noise. Wonder how that happened with such professional layout design?. With the steel case the problem was localized. The commissioning brought some additional problems still, more blown switching transistors for starters. But after the initial problems and fine adjusting the load curve I was actually quite happy with the thing. I managed to log a few days worth of data from the turbine after the installation(before the anemometer crapped), not enough to make decisive judgement, but it definitely showed improvement in mid-wind performance. Maybe next summer if I get around putting the anemometer & logger back to service I'll be able to get more real world data.
In the end i think it was a success. After installation a year + bit ago it has been working flawlessly. If I would now start from scratch again, I would do quite a bit of things differently, but the overall idea would be the same. It really came into value a few months ago, when after damaging 1 blade I had to cut down the blades, decreasing the swept area into 2/3 of what it used to be. Without the MPPT box and the shortened blades the load matching directly into battery would have been dysmal. Now, with a new load table, it has allowed extraction the maximum of what's available of that smaller rotor, until I get around of making a new set of blades. And then it will be again just a matter of loading a new load table.
And it sure provided much fun at tinkering with power electronics
.