I’m not satisfied with the speed of venus os performing a ESS adjustment to reach the grid setpoint after a significant change of power-consumption.
ESS has a realtive high latency, but the reason is not my grid-meter. It give feedback within 1 sec, so venus knows that a significant power-adjustment has to be done.
I know that power-adjustments are limited to <= 500W per phase and second but most of the time, my ESS is performing significant less adjustment.
So I temporarly switched from ESS mode 1 to 3 (= external control). There, I was able to achieve a power-adjustment of ~500 W/s. This proved that ESS in standard-mode 1 does some kind of smoothing.
I don’t want to use ESS mode 3 in real-live, because I would have to script my own control loop.
My question: Is it possible to tune the smoothing of power-changes done by ESS in mode 1?
What is Settings/CGwacs/PvPowerSetpointOffset for?
I ask because I have run my 3-phase system both with the DNO’s grid meter which updates every 5 seconds, and with the Multis themselves as grid meter, which are much faster, but inaccurate.
I found that with the DNO meter I get to about 10Wh/hour of export with a grid setpoint of 10W, even though power can oscillate wildly when loads are switched on or off. With the Multis this climbs to about 30Wh/hour, but regulation is much faster.
So, if you’re after that smooth graph, keep on looking for an answer, but if you just want to avoid excessive grid import and export (on the meter, where you have to pay) there actually is no problem in my experience.
As an example, this is what happens when I switch on a 1200W vacuum cleaner with the 5-second interval grid meter:
With faster ESS-regulation on power-changes, I want to avoid buying expensive enery as well as avoiding unneccessary battery-discharging/energy selling.
A typical example is an electric hotplate which permanently causes big power changes by turning on an off.
My grid meter give feedback 2x/sec, and I can see contantly alternating between big positive and negative power values.
But in deed, I have never checked whether the energy counters (taken and given kWh) also increase with the permanent power-changes, or whether these counters have a certain latency. The last would be fine
Are your loads connected parallel (AC_in) or in line (AC_out)?
In other threads the opinion is that loads on AC_out are adjusted without limit of 400W/sec.
Currently my load is on AC_in with a 10 Hz grid meter VM-3P75CT. With my PID controlled espresso machine switching every 1-3 seconds about 1200W I see this overshooting.
Will check on AC_out as soon as more wires arrive…
My loads are on Ac_in.
I can confirm that power-variations on Ac_out are immidiatly adjusted in a way, a can’t see power-variations with my 2 Hz gridmeter.
Now I did precise testing/measurements.
My headplate switches on&off 17,5x/min. These are 72 alternatings within 4 minutes.
Within this time, my gridmeter reported minimum power of -560W, maximum of 720W. I don’t know the power-consumption of my heat-plate => let’s asume 700W. This makes 700W * 4/60h = 46,34 Wh.
Within these 4 mins, power consumption as well as power given, reported by my gridmeter, was 10,8 Wh each direction.
Conclussion: About 11/46 = 1/4th of the consumed power additionally had to be bought from grid, and another 1/4th was discharged from battery and bought to grid.
My gridmeter has no built-in latency which would equalize These positive/negative power-alternatings. This is disappointing because legal regulations forbid immediatly ESS-power-Adjustments on AC-in, which could be technical possible and are done on AC-out.
Next test:
Switching on+off exaclty one time a Boiler which consumes 2200W.
Here, unnecessary power give to and consumed from grid is 4 Wh each direction.
(Earlier I mentioned my gridmeter reporting data with 2 Hz. That was wrong - it’s 1x/sec.)
I’m using 2 MPs, one on L1 and one on L2. Each could adjust power with ~500W/s - I tested this with ESS on mode 3. This means my installation is capable of achieving 1000W/s power-adjusting.
Unfortunatly, ESS in mode 1 is much slower/more smoothed.
