The ONLY thing that matters for how much charge you put on your batteries is how much current you drive through them.

(Things like pulse charging at high voltages have to do with breaking down sulfation on batteries that have partially sulfated without overheating them by driving the current continuously, or driving an equlizing charge through the fully charged cells to top off the undercharged ones during normal cycling without taking the whole bank up to an overvoltage.)

If you haven't let your batteries go down to dead-flat there is no load on the mill until it is up to the speed where it generates the voltage the batteries are sitting at.  (The "cutin speed".)  Once your capacitors have an initial charge your system would present this same load to the mill when each capacitor switched in - but if the mill had been stopped long enough for the capacitors to self-discharge it would present a load at lower speeds or at stop - exactly what you seem to be trying to avoid.

You lose efficiency because you have to charge your capacitors at a higher voltage than the battery voltage.  The extra power goes into heating your wiring - particularly the initial contact spot on your relay contacts when they close to connect the capacitors to the batteries.  You'll quickly weld your relays in that position - assuming they didn't weld in the other position from the inductive kick when they tried to disconnect the capacitors from the mill, or the similar inrush surge when they reconnect the partially discharged capacitor to the mill and try to suddenly slow it down to a speed corresponding to the voltage on the capacitor.  (Of course you'll probably never get that far because that latter inrush will pop your diodes.)

Now there IS an approach that will somewhat related to this that can get more out of your mill by playing with switched reactances.  It's called a "maximum power point controler".  It involves switching an INDUCTOR around, letting current build up in it at one voltage, then switching it to deliver the energy as a different amount of current into a different voltage, with a "free-wheeling" diode to provide extra current from ground if you're configured for a current boost, or to block back-current if you're cofigured for a voltage boost.  A capacitor is used on the input side to accumulate power while the inductor is in the delivery part of the cycle.  (A capacitor is also used to store power on the output side for devlivery while the inductor is in the acceptance part of the cyce if you're providing a regulated voltage to a load.  But this is unnecessary for battery charging operation  - except to reduce radio interference from the current switching by reducing the voltage swing rate.)  Think of it as an electronic "transformer" for DC, swapping current for voltage with the current-voltage product unchanged (except for ineficieny losses).

You COULD do this with relays in cycles with timescales in seconds.  But you'd need honiking big capacitors and inductors.  Instead it's usually done at a high audio or ultrasonic rate using transistors for the switches, because the size of the inductors and capacitors are proportional to the inverse of the frequency.  Running at 10 kilohertz requires inductors and capacitors 1/10,000th as large as running on one-second cycles.

The voltage ratio and current draw of this "DC transformer" can be adjusted by adjusting the on and off times - both their size and their ratio.  So a smart controller can observe the current and voltage (or current and frequency) from the mill, deduce the wind, compute the ideal operating load for maximum power, and adjust the switcher to let the mill run at that speed and load, transforming the voltage to what's necessary for charging the battery and getting as much charging current as the available energy will produce.

You should try this with a small, low voltage/low wattage windmill, if you really want to experiment. Overvoltage, unloaded runaways, etc..., are things to think about, as has been said. You'd probably be better off simply leaving the generator connected to the capacitor the whole time and just limit the duration of the discharge pulse into the batteries.

I think you'll find that microwave capacitors have too little capacitance to be of much use.

A large electrolytic (something larger than 100,000uf) will work better.

A 1 Farad stereo capacitor would cost you about $80, and they're readily available.

Just make sure your power source doesn't exceed the voltage ratings for the capacitor you're using.

Make all the connections between the capacitor and the batteries with short lengths of #4 battery cable, or heavier. The discharge contacts, or heavy solid-state switch will be dumping that capacitor in a split-second. I've done this with copper contacts made from hammered-flat water pipe. You'll need to replace the commutator contact and the POSITIVE discharge contact periodically. The negative contact doesn't get too chewed up, even with extended run-time.

Watch across the capacitor with a scope. You should see a slow charging ramp, then a nearly vertical drop as the cap discharges into the batteries.

You'll also see the battery voltage start "sloshing" up and down in a wavy fashion that isn't necessarily indicative of the frequency of the input pulses. Neat.

Yes, the battery charges very nicely, as will be seen when you run a load from the battery and see the work being done. Use a controlled charging source for the capacitor for your first tests. That way you can calculate the joules used for charging and the joules used in the known test load. This will give you your overall charging efficiency.

Once you've watched one of these setups charging a battery bank, you might begin to ask yourself how the bank can possibly be charging (voltage stays high) between pulses?

Good question. Wish I had a good answer. Since it works just fine on new batteries, over and over, I don't believe that it's all about desulphation.

Capacitors that store one Joule of energy are considered large.  A joule is one watt-second; by comparison deep-cycle batteries are commonly around 1 KWH (3.6 million Joules).  

So, while it's possible to do what you propose, the capacitor storage is so small that it would need to be filled and transferred at a very high frequency.  Of course, when you use a capactitor to "smooth" a rectified waveform, this is essentially what you are doing.

Pulse charging has been mentioned as a solution and Jerry has published a NE555- driven pulser circuit running at one Hz (from memory) with a huge capacitor. This charging solution is also recommended for batteries, which have been standing around for a while.

There is a lot of interest in Pulse Width Modulation circuits, MPPT, pulse charging, etc. on this board, but one does not see too many circuits. Glenn Littleford has published a PC/Visual Basic-based circuit for his Picaxe-based MPPT controller.