What is shown here is the default behavior of Victron MPPT controllers. In my case a 150/35. Every 10 minutes it moves away from the maximum power point.
I assume it has been programmed to do that in order to be able to move away from a local maximum and find a global maximum power point.
While that might be helpful in some installations (with shading, or strings facing different directions), in many cases it is completely unnecessary. On each of my MPPTs all panels face in exactly the same direction and shading is zero.
It would be good to be able to turn this off completely, or at least significantly increase the interval from the 10 minutes.
The downsides of the current behavior:
Waste of energy. By my estimate close to 1% of PV energy is lost to this re-tracking. 20 seconds every 10 minutes with output reduced by about 1/3rd.
Unnecessary output fluctuations. My EV charger tries to track PV output. It needlessly adjusts charging speed quite frequently due to this re-tracking.
Around 9000 shadow edges were identified in measured irradiance data, and the electrical characteristics of 250â500 PV module arrays with different configurations were studied during the irradiance changes. It was found that most of the partial shading events caused by clouds do not cause multiple MPPs for PV arrays, even for a moment.
It makes me wonder if the various algorithms that do full sweeps are guilty of a type of prematurely optimising, where the benefit of the sweep is not realised, or not realised often enough to make it worthwhile. This is probably vastly different in a solar farm situation, where arrays might be 100s of meters long, the probability of cloud affecting part of the array is much much higher (or more correctly, occurs for longer) than in an array that is 10m long.
But ⌠if im reading the paper correctly, they are already talking about farms ; â250-500 module arraysâ
This is fundamentally how a mppt works. it canât be turned off. For small systems there is a PWM charger.
With the number of chargers in the field, if a longer interval was beneficial it would have been implemented.
The impact to yield produced will be nominal over a typical day, on a 40kWh yield that is 400Wh, way less than is consumed by the chargers, inverter etc.
Factor in the variation in irradiance, weather etc, and typical panel losses outside lab conditions of 1/3 it isnât worth affecting the majority of installations that it helps.
Every modern MPPT adjusts multiple times a second, in minute fluctuations of voltage and current that have very little impact on the wattage delivered, but are enough to keep up with changes of irradiation, temperature and cloud cover.
They can get away with those small changes by assuming that there is only one maximum power point and no other local maxima that they could get stuck in. For the vast majority of systems that will be the case 100% of the time.
These big sweeps every 10 minutes donât happen with MPPTs of other brands. Grid connected inverters, with inbuilt MPPT, deliver very much constant output when there are no changes to irradiation.
In any case, it should not take 20 seconds with several sweeps across a vast voltage range to escape a local maximum power point. A single sweep, taking perhaps 2 seconds would be enough. And a single sunny day of doing sweeps should be enough of an indication that no further sweeps are required for that installation everâŚ
Does anyone know if the Victron MPPTâs algorithm is P&O plus a ten minute sweep? Based on oscillations ive seen this seems the most likely.
It could be Incremental Conductance (i.e finding where the slope of the power curve is zero) , plus a ten minute sweep?
FOCV and FSCC seem unlikely, but it does seem plausible that the dips in the graph are not sweeps, but Voc or Isc tests?
My googling says that RCC, FLC, ANN, PSO etc are only found in research settings, but they also all seem to need a fair amount of CPU work to implement.
Thatâs probably true, but only because of a design decision. There are many examples of the programmers adding a UI element that makes a previously static variable into a user settable variable.
FOCV and FSCC seem unlikely, but it does seem plausible that the dips in the graph are not sweeps, but Voc or Isc tests?
This is the best resolution I can get, since my charts are based on the MQTT feed coming from the Cerbo. Doesnât look like it goes either open or short circuit. Just a large sweep. Well 3 in a row. Even just reducing that from 3 to 1 would be a significant benefit, without any downside I can see.
We did experiment with allowing more customisation of these parameters for the RS series, also where the processing power is much greater.
While the sweeps look dramatic in the fine grain high resolution data, the effect on power output was minimal. And the downside is that it creates a lot more questions, and people also changing them without understanding and then being worse off.
While we do allow a lot of scope for customisations with Victron, in this case it was decided that the default behaviour now is really working best for almost everyone, and the downside is not so bad in the real world for the very small number of people where it would be slightly better with more options.
@guystewart Does your comment âWe did experimentâ infer that the âPartial Shading Detectionâ on/off option on the Multi-RS;
will not be rolled out to other products?
and/or
will be possibly removed in updates to firmware of the Multi-RS?
The Multi-RS looks great, but seems like it needs a few more updates (like parallel support, ESS, 8:1 ratio, grid codes) before its something we would use often.
guystewart
(Guy Stewart (Victron Community Manager))
10
The status quo regarding the âpartial shading detectionâ feature of the RS series, and for the rest of the range is how it is now, is how it is planned to stay. That is there is no current active development going into this feature. Further experiments were above and beyond this, but results were the improvements were too marginal to justify the additional complexity for us and customers.
There are no plans to remove it, or add it, or extend it.
8:1 ratio is quite hard fixed into the current design and also unlikely to change.
Parallel, ESS, and grid codes are all actively being developed, and have been for some time already, already yielding positive results, and much more still to come.
