I’ll have a go at this.
With an optimisation problem such as mppt, there is a graph of the voltage vs current (ie power). On a sunny day, this graph is basically the top and right sides of a square that has been smoothed
- high voltage with no current (Voc) = no power (bottom right on the blue line)
- high current with no voltage (Isc) = no power (top left on the blue line)
- somewhere in the middle, the VxI produces the largest area under the graph, = max power.
The red curve is the power - this is the line the mppt controller tracks, trying to find the maximum point. At this point the derivation of this line (ie the slope) is zero - its flat, and moving left or right causes either no change, or less power. As clouds come over, or temperature changes, you keep adjusting left and right to “see” if you are still at mpp (max power point).
Lets say you start the point on the graph where the arrow from the letter “P” points.
Here the red line is steep, and negative, so you need to reduce the voltage to go ‘backwards’ on the graph because you know the mpp is behind you. You can do this in 0.00001 volt increments, or 0.1v, or 1v, but if you know the graph has only one maximum, why not halve the voltage with each step until the slope goes positive - then you will get there faster.
This is the key to a ‘fast tracking’ mppt controller - it finds the max in X step by
- being agressive in the steps
- having a very fast loop of move the voltage → measure the power → make a decision → repeat
If you have a slow loop, or take tiny steps, you will still get to the mpp but it takes longer, and if the mpp moves faster than your steps do, you can be chasing that mpp all day.
Even if your algorithm goes way past the mpp, it will step back half that distance on the next step, so you get to mpp. You can be aggressive, because you know there is only one mpp.
Look up optimisation algorithms - there are literally hundreds
Now on to a graph where there are possibly more maximums.
I couldn’t find one quickly for solar, but here is one with multiple maximums;
In this graph, the black maximum is the one we want, and so in a problem like this you have to make some sacrifices in your algo to avoid landing at the red maximums and staying there.
Two options are;
- periodically picking a random point on the graph, jumping there, and seeing if the max closest to that point is higher than where you were
- keep a table of previous graphs, and when you are at a max that you have seen before, look up the table and jump to the other max points to test them.
In any case, moving from the non-mpp point to the mpp point more slowly means a lost opportunity - maybe you were pulling 300w when you could have been pulling 302w, and it took you 10 seconds to get there, so that 20 watt-seconds lost.
Stepping away from the mpp to another point to ‘test’ how much power is at the other point will cost you again - maybe production drops from 300w to 280w for a second, then you know that you have gone somewhere worse, so you go back to your mpp - you have lost 20 watt-seconds, which is hardly anything, but can be significant. This is what the designers mean when they say an mppt is fast (gets to the mpp quickly) and accurate (actually lands on the mpp, not close to it)
So in answer to your question;
- I don’t know the programming of the algo in the Victron mppt, but i know its better than the Growatt and the Epever algo.
- The algo will have certain parameters that allow it to find the mpp fast, and track it accurately
- In order to not land on false maximums in a multi-max graph, you have to compromise those parameters which will worsen the performance of the algo
- My expectation would be that the following statement is the most descriptive explaination: “If your panels cannot experience shade due to your array placement and solar angles, the normal algo will perform the fastest. If you expect shade, Advanced Maximum Power Point Detection might be more reliable at finding the mpp at a small cost to speed or total captured energy.” So why pay the price of Advanced Maximum Power Point Detection if you already know your panels cannot experience shade?
