MPPT using an inverting boost converter

I'm not sure I get it either,

but it may be that I-V slope on either side of the MPP with make the "hunting" go in opposite directions

i.e one step forward, two steps back on one side, two step forward one step back on the other

-Lasse

I think I get it. The curve is shaped such that while on a flatter portion of the curve, reducing one of the quantities will move a lot more toward the knee than reducing the other quantity will move away from the knee. When the dithering moves to an area where it moves about the same back and forth, it will have found the knee which seems to be the point of maximum power.

Consider that a curve of constant power is the hyperbola defined by I*V. In addition to this algorithm finding the region of maximum power there is very little power wasted as over a short range this region does not deviate far from the hyperbola.

You're overthinking it.

All you need to do is measure current and maximize that. The difference in battery voltage over the tiny change in current is negligible.

The more current you put into the battery, the faster it will charge. Actual voltage does not matter at all. Yes, the product defines power, but you don't care. Maximize the current. The voltage should be max at max current giving max power, if you care about that.

No over thinking at all.

Which current, the current to the battery or the current from the solar cell? Maximum current from the solar cell is at zero voltage and thus no power.

So please explain how you maximize the current to the battery? That is the entire point of the MPPT. I believe you are under thinking this.

On Wednesday, May 27, 2015 at 1:29:27 PM UTC-7, rickman wrote: ......

You adjust the PWM duty cycle to maximize the current into the battery usin g a hill-climbing strategy. The current drops monotonically as you go away from the optimum duty cycle.

The hill-climbing approach is a perturb and observe technique. Adjust the duty-cycle a small amount and observe whether the current into the battery increases or decreases. If it increases move a bit more in the same directi on or else go the other way. You may need to filter and constrain the oper ation to avoid falling into minima at the extremes.

I've seen designs that measure the power from the PV cell bat that seems mo re complicated and doesn't take into account the losses in the circuitry.

In many cases I suspect that the power required to run the converter will o utweigh the additional energy obtained.

kevin

Current into the battery. That's what you want to maximize. Solar cell current is irrelevant, just maximize the battery charge current.

perturb and observe. Measure the battery current. Increase the duty factor. If the current goes up, do it again. If the current goes down, decrease the duty factor. You'll dither around the peak BATTERY CHARGE current. We used the hall effect current sensors because they don't lose any voltage across them.

Your approach sounds complex to implement really. There is already a ripple in the current and voltage on the input. So why create more disturbances than are already present? Measure the current and voltage on the input and calculate the product at two points in the cycle. If it is higher in one phase than the other you know which way to adjust the switch time.

I don't think this is the optimal approach. For one it is slow and can't properly deal with changes in the curve. Why would the curve change??? Because illumination of the solar cell is not constant. If a shadow passes over the cell and causes a reduction in the illumination as you disturb and take your multiple measurements the result can be whacked pushing the circuit in the wrong direction.

Looking at each cycle of the switch allows the fastest measurement that will be much less subject to external influences. In essence it will ride the curve at all times maintaining the MPPT.

Make it as unnecessarily complicated as you want. If you live where clouds move so fast that you need to track 'em every 50uS, go for it. It don't matter to me. Publish the improvement you get over sampling 20,000 times slower.

Have you ever built one?

What is more complicated, measuring the variations that occur as part of operation or creating variations so you can observe them?

I like to measure the thing I want to control. I want maximum charge current. That's what I measure. Since I'm only looking for relative measurements, I don't need a real A/D. Charge a cap and measure the time. Doesn't need calibration. Works for whatever panel you put on it in any situation. Still works when one of the panels is covered with bird poop, or snow or shaded by a tree.

You may optimize for anything you like. I like stuff that just works even as the panels and batteries degrade.

Don't recall whether you mentioned that you'd actually built a MPPT controller and put it into service.

It's a well known way to control the MPPT, at least for space applications

When the ripple is 3% voltage and 3% current, the MPPT is excactly at the power point due to the characteristics of the panel

It also has the advantage, that it finds the powerpoint fast at startup. I did a speedup algoritm, so it moves faster if it had taken long time to find the power point last (during startup, satellite spinning etc)

Cheers

Klaus

Well, since the MPPT is running at 100Hz and 3% step in current, it takes only a fraction of a second before the power point is found. Do you have clouds moving that fast?

Yes, but that is susceptible to noise in the system. We used a superbuck converter, so we had very little current ripple in the panel

Cheers

Klaus

On Wednesday, May 27, 2015 at 8:08:20 PM UTC-7, rickman wrote: ...

The capacitance on the input will prevent that working - even the PV cells themselves have a very significant capacitance and the IV curve at microsec ond time scales will not match the long time scale parameters. You will al so need a large capacitor across the input to supply the ripple in the conv erter.

kevin

An elegant hardware implementation by a really great guy:

(He likes to use 4053's unconventionally--oscillator, inverter, and S/H!)

Cheers, James Arthur

Piston driven by boiling fluid.

a partially shaded reservoir fixed behind the west edge of panel, sun heats the reservoir flid boils, pressure purce liquid out the botton of the reservoir into the piston moves, panel moves westward until it shades the reservoir sufficinetly.

Are you interested in off-axis illumination or only that which comes directly from the sun. what do you actually want to know?

I want to be able to gauge how "sunny" (in some band) the day is/has been.

I know what the "shade patterns" are. These are related to the location of the Sun throughout the day. E.g., the rose bushes are only exposed to direct sunlight during the morning hours. The "brightness" of the sky has little effect on them after noon (because they are heavily shaded). OTOH, certain of the citrus trees are exposed to full Sun all day long -- others for a portion of the day (or, *reflected* Sun at other times).

Ditto with the house itself.

That is not a totally absurd approach. The sun only moves across the sky one way. So the tracker only has to track one way. When the sun is obscured it won't matter much that the panel tracks so sitting and waiting is an adequate response. When the sun is visible again the tracker resumes. Something similar can be done with a single solar cell which powers a motor with highly reduced gearing. When the sun moves a bit past the optimal point for the current panel position the solar cell is unshaded and powers the motor. When the cell becomes shaded again the friction is too much and the movement stops. But there has to be a return mechanism to reset the position of the panel overnight. With your piston it could be a spring that exerts a constant, slow restoring force.