Jump to content
Thank you for the great forum, Safe Driving over the weekend. Sincerely Jason
Pietpower

Solar dipping

Recommended Posts

Why does this happen?

When I run a high load the pv panels supply max power and the balance comes from the batteries.

When the high load stops it seems the pv panel output also dips instead of swinging over to charge the batteries again.
Especially if you look at the graph just before 1pm.  The solar did not recover and the batteries did not charge for about a half hour.

Inverter is a Goodwe hybrid and location is Gauteng with no cloud cover or shading or anything.

 

Purple line is the total load
Blue line below it is the solar supply
Green line is the battery SOC
Orange line is eskom/meter power use

Solar dip.JPG

Share this post


Link to post
Share on other sites
1 hour ago, GreenFields said:

Looks almost as though the charging cycle is interrupted each time the load picks up, and the solar generation begins tracking/prioritising the load.

That might be normal. As I recall the Goodwe has a high frequency design and the MPPT boosts directly to the high voltage DC bus. It runs a buck/boost pipeline from there down to the battery. One would expect it to pick up fairly quickly once the load drops, but there will most likely be some kind of software regulation that controls the dc/dc pipeline between the high voltage side and the low-voltage side.

Share this post


Link to post
Share on other sites
7 hours ago, DeepBass9 said:

How often are those points being sampled? It looks like a smoothed curve over fairly wide spaced data points.

The curve points are every five minutes.

What concerns me is that between 12pm to 1pm the solar never came back to what it could be and draw power from eskom which it should not.

Share this post


Link to post
Share on other sites
6 hours ago, GreenFields said:

Looks almost as though the charging cycle is interrupted each time the load picks up, and the solar generation begins tracking/prioritising the load.

Yes, charging cycle stops, the solar goes to the load.

Why is the charging cycle not restored when the load goes away?

Share this post


Link to post
Share on other sites
4 hours ago, plonkster said:

That might be normal. As I recall the Goodwe has a high frequency design and the MPPT boosts directly to the high voltage DC bus. It runs a buck/boost pipeline from there down to the battery. One would expect it to pick up fairly quickly once the load drops, but there will most likely be some kind of software regulation that controls the dc/dc pipeline between the high voltage side and the low-voltage side.

Will read this a few times to understand.  So is it a software glitch?

Share this post


Link to post
Share on other sites
8 hours ago, Pietpower said:

Yes, charging cycle stops, the solar goes to the load.

Why is the charging cycle not restored when the load goes away?

That's exactly what it's doing. The charging cycle is the light blue/turquoise line (Battery W). Before 12:00 it's above zero, so you're drawing up to 2000 Watts from the battery to help supply the 4000W load. Then straight after 12:00 when the load drops off, the battery watts go to minus 2000W, when you're starting to charge the batteries.

The slightly more complicated part is between 12:00 and 13:00. Because each time the load starts to go up and down (somewhere between 0-2000W, the battery charging starts going to zero, then begins charging again and reverses again. And at the same time the solar PV generation starts to match the load to some level below its maximum 2000W each time the load goes up. In fact, when the load spikes momentarily above the 2000W panel output at 13:00, you're drawing power from the battery again to help support the load. Just guessing, maybe the battery is not charging so that it can be on standby to help balance any fluctuations in PV vs load.

Lastly, after 13:00 when the load is dropping to zero again, all the 2000W PV is going into charging the batteries full steam again.

Share this post


Link to post
Share on other sites
17 hours ago, DeepBass9 said:

It looks like a smoothed curve over fairly wide spaced data points.

I agree with this. The upload frequency to the SEMS portal is very low (minutes!). To get a better picture what is going on in real time I use the PV Master app.  I have also observed @plonksterpoint of slow response to load changes. Especially slow is the change from battery discharge to charge status (and vice versa). My Goodwe ES inverter takes about 10 seconds to switch over. The reaction time is so slow, that a cyclic load like for example a washing machine is totally confusing the inverter.

Share this post


Link to post
Share on other sites

This is a thing. Here's a snip from my graph for yesterday. You can see about 9 load shoots up, then goes down, but the solar dips brieftly as well. Same round about 10:10 and 10:30.

image.png.a873ef19c3517230477d5f3bf2287f9e.png

I've been noticing this behaviour for a while, but never given it much thought. In the snip above, the solar generation peaks out at just over 2.8 kw at about 11:30

Share this post


Link to post
Share on other sites

Bobster, I think the resolution on your graph is too low and you are missing data.

See below from a parallel 5kW Axpert setup, 6x pylon 2Ks with around 5kW of solar on the roof. This is for a winters day with cloud and sun patches. 10s update intervals.

2x heatpumps that come online at 10:00 and 12:00 (looking at the graph now it seems like their clocks drifted a bit 😅 )

image.thumb.png.97ddc0ed17ecc301f12a2b51b1f6cfea.png

Notice its not a smooth graph at all.

Below is from a winters day with full sun.

image.thumb.png.1775a7bef172bcfab66fcff3809875d6.png

Share this post


Link to post
Share on other sites
38 minutes ago, root said:

Bobster, I think the resolution on your graph is too low and you are missing data.

I snipped what I could from the Goodwe portal. I can increase the horizontal resolution but not the vertical (or don't know how to). 

Share this post


Link to post
Share on other sites

OK I kept an eye on PV master today.

Figured out the load in my case is ironing which switches on and off a few times a minute and is misrepresented on the sems data.

When the iron switch off the solar dips. It then pics up again but seems to take about 20-30 seconds or so. By then the iron kicks in again and use the power only to switch off again repeating the process.  This seem to block the solar from charging the batteries.

Need to get a battery operated iron to smooth out the power consumption. ;) 

Share this post


Link to post
Share on other sites

SEMS portal samples every 5 minutes over the internet.

Is there another way to sample for a graph direct without going over the internet?

Connecting a tablet direct to the inverter?

Share this post


Link to post
Share on other sites
17 hours ago, Pietpower said:

SEMS portal samples every 5 minutes over the internet.

Indeed. Not ideal and it's delayed, so can't be used for real time monitoring. But still, it may smooth out graphs, but it won't show a dip that never happened.

The annoying (I'm joking) thing is that it doesn't happen all the time. Today the portal shows load suddenly dropping off and solar continuing on an upwards path...
image.png.32f52438073d172458297f86cd36619e.png

 

Share this post


Link to post
Share on other sites
20 hours ago, Pietpower said:

This seem to block the solar from charging the batteries

that's it. Exactly my experience as well. In my case the worst offender is the washing machine. The motor starting and stopping every few seconds. I wonder if there is any technology to mitigate this behavior of the inverter. @plonkstercan you explain why it is not possible to write software capable of reacting much faster to load changes? 

Share this post


Link to post
Share on other sites
17 minutes ago, Fuenkli said:

possible

Oh it is possible... sometimes. Fronius inverter are very fast in their adjustment. But the moment you involve batteries it gets more difficult, and there's simply latency in everything and by the time a measurement arrives from whatever device you're using it is already a couple milliseconds out of date. What is more, the system is inherently "closed loop", as in whatever adjustment you make immediately affect another measurement down the line, and if you do it wrong it becomes unstable and swings out of control. So the smaller the system (eg only a PV inverter vs batteries) the easier it might be to do that, but all systems will have a some kind of damping built in to smooth it out.

In this case there is a DC/DC converter from the battery, and you might think of it as a pipe carrying water where one moment the water has to go the one way, and the next moment it has to change direction and go the other way. It's always going to need time to do that. Normally one would expect this time to be around the 5-10 second mark and not the 1-2 minute mark though.

Share this post


Link to post
Share on other sites
14 hours ago, Fuenkli said:

can you explain why

I can have a go at integral windup. System control is an entire engineering discipline, with all manner of surprising behaviour, so it's not surprising that the cheaper models don't perform as well as those from manufacturers (like Fronius) who rely on their name and reputation more than their low price.

The obvious control algorithm is proportional control. You figure out where your system is at, and where it should be, and the difference is the error. You apply a correction proportional to the error, and that tends to get the system back to where it should be. If that means you run into a limit, e.g. the battery is low and you have maxed out your charger current, there is no problem. But there is a choice here: how much correction do you apply? Usually it's a constant proportion of the error, often called Kp (the "p" should be a subscript). If you make Kp too low, the system is sluggish, and responds slowly to a change in conditions. If you make Kp too high, the system can become unstable, usually oscillating. For example, you detect that you need more charge current, so you increase it, then detect that the battery voltage is too high, so you decrease it, but by too much so now the battery voltage is too low, so you increase it, and repeat the cycle. Under certain conditions, the amplitude of the swings can increase until the system is hitting both limits (in this case, maximum charge current, and zero charge current). Somewhere in between will be the best value for Kp, though you might want to back it off from that value, to account for variation in performance between individual models, and changes as they age. It turns out that delays are the enemy here, as @plonkster has alluded. So a more expensive processor that can measure and correct faster will be able to be stable with a larger value of Kp, and so will perform better.

Believe it or not, that's the simple part. The problem with proportional control is that there is always a residual error; you can never eliminate it. In fact, it needs some error to make anything happen. With a proportional gain of say Kp=10, you might find that the system always goes to 90% of the desired value, or 110% if coming from the opposite direction, and won't go all the way to 100%.

So now we consider integral control. This is a simple idea (like proportional control); instead of acting on the error itself multiplied by a constant, you act on a different constant (Ki) multiplied by the integral of the error. This simply means you add the past errors and call that a rough integral of the error. So now with the 10% error, the effect of the integral term will increase until the integral control forces the error to zero. Once the error term reaches zero, the integral of the error also goes to zero, and the system is happy to sit at the desired value. All it good, so far, although now we have two constants (Kp and Ki) to tune, and it happens that the values of these constants interact with each other slightly. But that's not too hard to handle.

The problem is when you have an integral term, and your system hits a limit. For example, when charging the battery, you reach rated current for the charger, and can't go higher. Or the battery voltage is too high now, and you reduce the charge current, but it can't go below zero. What happens to the error? It stays constant. That's fine for the proportional system, but for the integral system, the number representing the integral of the error just keeps increasing (or decreasing, depending on the sign of the error, positive or negative). You might think that' not a problem: the system is limited, and can't charge any faster or any less, so just keep doing what it's doing. But now (finally!), what happens if the system changes and we have to go in the opposite direction? The error value changes, and the proportional part of the control works just fine. But the integral system still has a massive backlog of error, so it might take a very long time for that integral value to start heading in the right direction. In the mean time, the control system, which should have changed direction, is still continuing in the wrong direction, waiting for that integral term to change. This is integral windup, and it causes over- and under-shoots.

When you're aware of the problem, it's not too hard to take measure to mitigate it. You can detect when your system is limiting, and stop adding the error to your integral term under those conditions. Or you can detect this overshoot / undershoot condition, and set the integral value to zero temporarily. There are other simple measures. But my suspicion is that the Voltronic engineers aren't aware of it, or have forgotten or ignored it. After all, ignoring integral windup works fine in most conditions, it's only when conditions change that the problem might show up.

Control systems are everywhere. Your car's cruise control is one example, and one where a fixed error is not desirable, so it definitely has integral as well as proportional control. If your car continually overshot the set speed, or undershot it when you set the speed down, you'd really notice, and you'd say the cruise control was poor. Most cars seem to get it pretty right, although you can always grumble about how long it takes to react to a hill, or how it uses too much fuel or braking to get back to the set speed, etc.

In an inverter, there are several control systems. One affects charge current, as I've been using in my examples. But others regulate the inverter output voltage, the bus voltage (the ~400 VDC bus that powers the inverter), one that handles phase synchronisation of the inverter and AC-in, and so on.

There is also derivative control. That's where (usually in addition to proportional and integral control), you have a term that depends on the rate of change of the error. This is usually something reserved for high performance control systems, and has more tendency to produce instability than the other control algorithms. Thus there are pure P systems (proportional only), PI systems (the most common, proportional and integral), and PID systems (all three).

Edited by Coulomb

Share this post


Link to post
Share on other sites

I may be wrong here, but I think the usual PID control is not the right thing to use to approach power levels in a home. In industrial control there's actually three terms, as @Coulomb said, the proportional part, the integral part, but it also has a third corrective part called the derivative part (in systems without that, you'd say the Kx for that part is simply zero).

The cruise control analogy is a very good one. One can intuitively feel that you should ease off the throttle as you get closer to the target speed, but if you start to back off too soon it will take a long time to get there, and if you don't back of soon enough you will end up just overspeeding a bit. So now the rate at which you back off (the first derivative in other words) becomes proportional to the error (the closer I am to the target speed, the faster I need to back off the loud pedal).

PID control is brilliant for working with physical stuff with continuous properties. If you want to control to concentration of a salt mixture for example. I am not convinced it is much good for residential power control, where things pretty much come and go at the whim of more-or-less autonomous agents (aka human beings). It's a bit like attempting to adjust your cruise control to stop-go-rush-hour traffic. It can be done to some extent*, but it's messy and in Cape Town someone will just push into that "regulation gap" you left for yourself 🙂

So in my experience, simple proportional control with a Kp of 0.6 is good enough. Even adjusting once a second with such a system gets you to within 99.5% of the target value within 10 seconds 🙂

* I've driven manual transmission vehicles my whole life, and riding the clutch is just something I was taught not to do, so in stop-go traffic I don't close the gap immediately: I wait for a bit of a gap to build up (keeping other impatient drivers in the mirrors), and then try to hold a constant speed just off idle in first gear for as long as possible before having to stop again, essentially trying to regulate my speed to the average speed of the rest of the stop-goers 🙂

Share this post


Link to post
Share on other sites
13 hours ago, Coulomb said:

I can have a go at integral windup

@Coulomband @plonksterthank you very much for the detailed explanation regarding the complexity of system control. I learned a lot and my admiration for the guys writing this kind of software grew immensely. What about the hardware in the inverter? Is there any time limits the control software has to stay within to prevent damage?@plonksteryour analogy with water flowing through a hose suggests that an abrupt change of a value could cause problems. What kind of problems? This is difficult for me to understand because electricity has no mass. 

Share this post


Link to post
Share on other sites
10 hours ago, Fuenkli said:

Is there any time limits the control software has to stay within to prevent damage?

Sort of. Let's say the SCC "looks like" 12Ω to the panels, to extract 300 W @ 60 V from a cloudy sky. Then suddenly the cloud passes, and the panel can momentarily provide 120 V into 12Ω. P = E²/R = 120²/12 = 1200 W. Suppose that this SCC is a small one, rated at only 1000 W; it now has a 20% overload till the SCC reacts and pulls back the power.

Fortunately, the nature of MOSFETs and heatsinks is such that a set of MOSFETs can usually handle a short term overload (of the order of 1 minute or less) without damage. So while there is a time factor, it's usually not a problem.

Share this post


Link to post
Share on other sites

I have to confess at this point that I'm not too familiar with the workings of a buck/boost pipeline and how you control that with a microprocessor, it's possible that the software is more of a limiting factor than the hardware. It will probably use some kind of state machine  so it knows in what mode it is, and there will be reasons not to flop around between modes. I'm totally guessing though.

Share this post


Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


Topics

Contact Us - Power Forum South Africa

×
×
  • Create New...