Living with Solar: part 3
Once my solar PV system was up and running, I was able to watch how it performed.
This system has a 6kW hybrid inverter from Ginlong Solis — a mid-range, work-horse Chinese brand — and includes a wifi module that connects to the Internet. It uploads data to the “Solis Cloud” every 5 minutes, where it can be inspected using an app (iOS/Android) or a web browser.
Since it’s a hybrid system, all the information about both solar and battery state is available in a single interface.
The most important information is shown in two graphs: generation and consumption. Here are my graphs for the first full day after installation:
The generation graph has one line showing the overall solar power generated by the panels (in yellow). It then has three more lines showing where this power has been sent to:
- “Self-use”, i.e. directly powering the house (orange)
- Battery charging (light green)
- Export to grid (blue)
The sum of these three lines equals the yellow one¹.
In the top graph, you can see that the solar generation was mostly used for battery charging (the green area) until 14:00 UTC, or 3pm local summer time, when the battery was full. After that, self-use jumped up (the orange area): this was my iBoost kicking in, using all the spare electricity for heating the water tank. If I didn’t have the iBoost then it would have been exported to the grid instead.
The consumption graph has one line for the overall power consumed by the house (dark green). It has three other lines for where this power has come from:
- “Self-use”, i.e. direct from solar panels (orange; this line is identical to the one in the generation graph)
- Battery discharging (purple)
- Import from grid (red)
Again, the sum of these three lines equals the dark green one.
In the lower graph, you can see consumption spikes at various times of day, most likely for boiling the kettle. And happily, almost nothing was coming from the grid.
This all makes sense, once it clicks. It’s obvious really that the power you use in the house at any particular instant is always going to be different from the amount that the panels are generating at that time.
Therefore, if the panels are providing more power than you are consuming, the extra goes either to battery charging or to the grid. That’s what the top graph shows. On the other hand, if the panels are generating less power than you are consuming, the shortfall has to be made up from either battery discharging or by drawing from the grid. The bottom graph shows this happening.
This system has two Pylontech US3000c batteries, with a total storage capacity of 7kWh. I wanted to see how the energy stored in the battery varied over the day — also called its State Of Charge (SOC).
This isn’t available in the default “plant overview” screen, but can be found by switching to “device” view. In fact, there are very many inverter parameters available here, and you can combine multiple ones onto the same graph.
Here is the power generation graph (orange), combined with the battery state of charge (grey), for the same day as above:
The battery SOC fell to 21% overnight (having reached 79% the previous day), but climbed right up to 100% by mid-afternoon.
The graph has finer resolution in this view, and you can more clearly see the sun going in and out of clouds for much of the day.
An unexpected problem
This was all very exciting. It was early May, the weather was good, I could see the battery charging up and discharging as expected — although some days the battery didn’t manage to fill to 100%, I was not using grid power.
But then there was one perfect, cloudless sunny day. This let me see the system as it should be performing at its absolute best. Here is the generation graph for that day, with all lines turned off except for the total power:
The maximum value was 2.09kW, at 13:00 UTC (2pm), as expected for a south-westerly facing system. You can see a sharp drop-off starting 17:00 UTC, when shade falls across the roof due to tall trees nearby.
But hang on. My system has 14 panels, with a design capacity of 5.18kWp. Even in these perfect conditions, I wasn’t getting anywhere near that!
Time to dig down further.
Again by switching from the “plant overview” to the “device” view in Solis Cloud, there are many more measurements to inspect.
My system was designed with two strings, of 6 and 8 panels, connected to two separate DC inputs on the inverter, PV1 and PV2. I could check the voltages on both these strings independently:
PV1 has a healthy 200V DC, staying mostly flat thanks to the optimizers. But PV2 has zero volts! It’s clear that the whole string is not working.
According to the panel specs (JA Solar), they have an optimum operating voltage of 34V. Seeing 200V suggests that there are six panels in the working string (200/34 = 5.9).
These are 370Wp panels, so six panels are 370x6 = 2.22kWp, and I was achieving very close to that.
Further observations on another sunny day confirmed this theory. When shade from nearby trees fell across the rear set of 8 panels, the power output did not drop. But as soon as shade fell across the front 6 panels, it dropped.
It was time to call back the installers to fix the system. Maybe they hadn’t plugged a cable in? Maybe one of the panels or optimizers was bad? Maybe the inverter itself was bad? It was their job, not mine, to sort it out.
Unfortunately I had to go away, and it was nearly two weeks before they were able to come.
After the fix
When the electrician arrived, it took only a few minutes to find and fix the problem. A cable had not been crimped properly.
Here’s the graph for the first few hours, showing the power from PV1 and PV2 separately, plus the battery charge. You can see exactly when it was fixed.
It was a miserable rainy morning. But still, I was getting about 150W combined from the panels, which was enough to reduce the battery drain to a trickle — and sometimes up to 400W.
It remained overcast, but as the rain cleared, generation briefly spiked at 1kW, and then at 2.5kW as the cloud thinned, and then to an astonishing 6.3kW as the sun broke through:
I had read about this phenomenon before. When the sun lights up a nearby area of thin cloud, it can briefly result in more light reaching the panels than from direct illumination alone. The thirsty batteries were lapping it up as fast as they could², and there was even excess available for the iBoost.
Fortunately this didn’t last long, as the inverter is only rated for 6kW, although it allows short excursions above this (and can protect itself).
Here’s the complete power and battery graph for the whole of that day. Since we spiked above the inverter’s rated capacity 6kW, the software has added warning lines for rated power and 1.2 times rated power (ignore these for the battery charge, which has its own axis).
Despite raining heavily for the first half of the day, and then mostly grey with a few bursts of sunshine, the battery managed to charge up to 100% (just).
The next morning was sunny with light cloud, turning to heavier scattered cloud in the afternoon. The battery had dropped to 59% overnight, but by 10.45am (local time) it was fully charged and the iBoost kicked in. By 12:40pm the tank was hot, and for most of the rest of the day we were exporting substantial amounts of power³. The graphs show all this:
(The sharp drop in battery charge from 99% to 81%, starting around 6.15pm local time, is due to using both the oven and hob for cooking. The time we cook is usually just after the sun has gone behind the trees, so the battery really helps out with this)
Suffice to say, I’m extremely happy with the results. Of course, this is late May; the days are bright and long, and getting longer. No doubt things will be very different in December. But it’s good to see that the full system can generate useful amounts of electricity in cloudy conditions, and even a little when it’s raining.
Living with mini-solar
An unexpected benefit of this episode is that I was able to experience what it was like to live with a much smaller, 6-panel system. This gives me a hint as to how things may be when winter approaches.
Perhaps surprisingly in retrospect, on most days the battery managed to charge up fully. However on 10th May it only reached 90%. On the next day, here’s what happened:
When the battery discharged to 20%, we were switched to grid power for two hours in the early morning. The charge crept up only to 57%, so it was back down to 20% later that evening. We were then on grid throughout the night, until the next morning:
Notice how the battery discharges very slowly once it hits 20%. It was supplying only about 35 watts — the rest came from the grid⁴. In the day though, intermittent sun was enough to charge it back to 90%.
The worst day of all was this one:
The vertical power scale has changed. The peak output was less than 600 watts, and only 3 units of electricity were generated all day. There was just enough to stop the battery discharging for part of the day, and after that it soon drained away, leaving us on grid overnight again.
But despite this: over a 15-day period, and using readings from my “real” utility electricity meter, I can see that I used only 18 units in total, whereas normally I would expect to use about 150.
That’s a short period of time, and I don’t know if it was unseasonably sunny. Certainly there were some scorching days.
I think this experience shows that even a small solar PV system can produce beneficial results, at the right time of year. However, the panels themselves are only a small fraction of the cost. Put another way: a 6-panel system costs almost as much as a 14-panel system, and yet will have much less benefit, particularly when conditions are poor.
That leads me on to the fourth and final part of this series, where I’ll look at the available choices, costs and benefits.
¹Maybe it would be clearer if it were shown as three bands of colour stacked on top of each other, rather than side by side. If this is already possible, I couldn’t see how to do it.
²Each battery has a maximum sustained charge and discharge rate of about 1.8kW, so it’s 3.6kW when there are two. They can burst higher for very short periods.
³Unfortunately I’m not being paid for this yet, until the DNO has registered my system and I get a smart meter. However, afternoons are a good time to be exporting power, since demand is high. A variable export tariff like Agile Octopus pays high rates from about 4pm to 7pm.
⁴35W is about right for the the router and two NUC servers I have connected to the backup power output on the inverter. That is: I believe the inverter may be choosing to power these from the battery rather than from the grid, even when the rest of the house is on grid.