Slow SMA AC coupled system charging ramp up rate.

[brum]

The system is grid tie with two inverters - SMA Sunny Boy 3.6 and SMA Sunny Island 4.4. The problem is the Sunny Island ramp up speed when charging. If there is drop in the sun (short cloudy condition, i.e. 30 seconds) and then full sun the system does not get to full speed in matter of seconds, but in matter of minutes. In some cases this takes over 10 minutes.

Here is an example:




Green is the PV power produced by the Sunny Boy. Yellow is the battery charging power. The first vertical line at 10:19 is the moment when 2kW load was turned on. The second one that is 20 seconds later - when it was turned off. The PV power dip is obvious. It took about 5 minutes to ramp up the PV production to full speed.


How this is a problem - the family is at home full time. Me working from home + wife + 3 kids. These dips in the produced PV power are quite often. As result I'm loosing quite a few kWh each day and often the battery does not fully charge. An example is yesterday - sunny with no clouds, calculated max produced power is close to 28kWh, actual produced power is 20kWh and the battery was left at 80% in the evening. And it is a lot worse in the winters. Worst case scenario is when the oven is turned on. Once it reach the required temperature it switched between on and off cycles with 2.5kW load. This cripples the PV production severely. Not with 0.5kW like in the example above, but with close to 2kW for quite some time.

I suspect that this is caused by the huge grid supply spike when the load is removed. The system throttles the produced power and this impacts all systems. So I suspect that the cause is the algorithm in the SMA Home Manager 2.0 that is throttling the power production to avoid grid feedback.

Any ideas how this can be improved? Without switching from AC coupled to DC coupled system .

Note: The post was updated with more details and improved issue explanation. The old version was quite hard to understand.

 

[brum]

Some more research confirms that this is caused by the SMA Home Manager and its algorithm for limiting feedback. During trottled periods devices state that they are limited by external PV system control.

Now I'm considering turning off the devices' external control and setting them to some modifiable fixed setpoints. I'll likely need to experiment a bit with the setpoints. The idea is to set the PV inverter setpoint to -10W (10W feedback) and the battery inverter setpoint to 10W (10W consumption). This can be further optimized with some dynamic control over the setpoints through the ModBus over TCP/IP exposed by them. The documentation is public and I'm already talking to the inverters for collecting telemetry.

And one more example of how much PV energy has not been utilized. The gap in the top chart between the yellow and the red line shows that energy. The bottom chart shows the voltage of both PV strings. If voltages are equal and the yellow line is below the red line in the top chart, then PV strings are not fully loaded. The last one is a bit complex to explain, but if somebody is interested in it I'll give the details. The gaps between 10:30 and 12:30 are due to slow system ramp up rates.

 

[Kuma]

SMA SB has a Shadefix evaluation time. I recently watched a video and the default "sweep time" to evaluate shade/clouds is every 6 mins. The presenter did say you can turn off shadefix or change sweep time min to 3 mins.

 

[brum]

Yep, I'm aware of it. But when throttled the PV inverter says that the external system is dictating the limit, not that the max produced power has been reached. The PV inverter actually catches up quite fast. The Home Manager 2.0 that controls the zero export is likely the cause. But I'll be able to confirm that once I switch from external control to internal control based on the grid energy flow data. The home manager will become just a meter in that case. A power meter that is just too expensive...

 

[brum]

So I think I found a solution. Still testing it, but the initial run for several hours shows very promising results.

How it works - constantly monitor the PV inverter. Use some greedy logic to determine if more power can be produced and if yes - force the battery inverter to increase the charging power.

How do I know there is more PV power to be produced - the PV inverter output is 200W below the theoretical maximum output and its two MPPT trackers are working at voltages that seem to be above the Vmp. How do I force the inverter to charge - calculate how much more energy the PVs should be able to produce and increase the inverter charging power with that. This is done by writing several Modbus registers using the TCP Modbus interface of the inverter.

Sounds simple, but this ended up as more than 300 lines of Python code.

The results can be observed in the attached chart. After 2 PM the script was constantly running (prior to that it was running intermittently while being created). Usually, I get pictures like the one above - with a lot of gaps between the red and the yellow lines. Obviously, after 2 PM the gaps were severely reduced.

And last but not least - the cost of this. The grid power consumption will increase by around 0.3 to 0.5kWh per day if there are a lot of load spikes. There will be no impact if there are no short load spikes that are responsible for the reduced charging speed.