A DIY solar power station—is it really possible? Absolutely!

A while ago, I posted an announcement on Instagram about building a solar power station. Well, not only has it been installed, it’s already been generating power for three weeks, and I’ve finally gotten around to writing up how it was built.

As usual, a reminder: to keep up with new posts on the blog, I recommend following me on Instagram. I’ll post notifications about every new topic there, and might also write about something interesting from time to time — link to my account

As with the sprayer, I’m writing this article after the work was completed. I hadn’t originally planned to write it, but I have some material, and, more importantly, practical knowledge and experience from the project, so I think it’s worth sharing.

So, given what’s been happening lately, the idea came up of generating electricity without relying on the grid, power lines with their falling poles, and so on. A friend also has a gas-fired combi boiler with a pump that won’t run without electricity, plus a fridge full of food. In short, a typical private home—I imagine plenty of people live in similar circumstances.

I’ll say right away that this system is designed specifically to provide power independently. We’re not talking about saving money or getting a return on the investment, since it’s not being installed somewhere sunny in the south. The output won’t be what it could be if we were in Egypt, for example.

We discussed the requirements first, and the initial spec looked roughly like this:
1. Panel capacity: 1 kW
2. Battery capacity: 1 kWh
3. Inverter power: 1–1.5 kVA.

A classmate of mine has a similar system, so we already had a real-world example to go by. By the way, he’s building a more powerful system, with a higher-power inverter and a larger battery based on LTO cells.

First, a bit about the house—or, more specifically, its roof. This is very important, and this detail can kill the whole idea before it gets off the ground.
It’s a gable roof, but the slope on the sunny side is shallower, so the roof is asymmetrical. That makes the installation easier, but increases the cost; I’ll explain why later.
The roof pitch is 22 degrees.

One option would be to install the panels on the ground, on brackets, or on the roof of an outbuilding. The first option was ruled out because the lot is already quite small; the second, because those roofs face the wrong way. Turning the building around would be, well… difficult.


The house actually faces not south, but roughly south-southwest. The photo was taken at 13:30, around noon; the house is directly in front of me, and the post’s shadow points a little to the left. At a guess, the azimuth is about 20 degrees, though in this case it’s not that critical. As the friend who advised us said, this is even better than if the house faced south-southeast.


While comparing panels, installation options, inverters, and so on, we decided to install 1,600 W of panels rather than 1,000–1,200 W. We found 405 W panels locally, and four of them were a convenient fit. They were also relatively inexpensive, and adding one more panel didn’t increase the budget by much.

We enter our details into the calculator: choose the location on the map, the panels’ power and quantity, their tilt angle relative to the horizon, and their azimuth relative to south.
We chose a 45-degree tilt as a compromise: to boost output in winter without losing too much in summer. The company we bought the panels from suggested installing them without any additional tilt, using only the roof’s 22-degree pitch. But the calculation showed that we’d lose a lot in winter, and winter was exactly when we needed more power. We’d have a surplus in summer anyway.


According to the calculations, the system should produce this much power. December is the worst month, of course, at around 50–52 kWh. Increasing the tilt to 60 degrees raises that to a maximum of 56 kWh, while at 22 degrees the calculated output is 39 kWh.
There’s also hope that the panels will be less likely to get covered in snow in winter than if they were lying flat on the roof.


Before moving on, I’ll briefly explain the differences between system types. Broadly speaking, they can be divided into a few configurations:
1. With or without a battery
2. Grid-tied or off-grid.

We never planned to feed power back into the grid: the equipment is more expensive, replacing the meter is complicated, and overall there’s no point.
But we did want a battery, so the house could run on it at night, and because it would provide at least some backup power even without the panels.
In fact, the system ran like that for several days—as a big UPS.

In any case, besides the panels themselves, you’ll need an inverter to convert their DC output to AC for powering appliances. You could, of course, rework the wiring and appliances so they could run on low voltage, but I can tell you the gains are tiny, while the complexity and potential problems are considerable. Any savings from avoiding conversion would be offset by:
1. Larger wire cross-sections or losses in the wiring.
2. installation of a second, low-voltage network
3. modifying appliances to run on low voltage
4. reduced flexibility, since, for example, converting a TV or refrigerator is much more difficult.

We initially wanted to get a LogicPower inverter, but it turned out they had been discontinued, and we also wanted a higher power rating. In the end, we went with a relatively budget-friendly Axioma 3000 VA.
One more digression: inverters come in different types, PWM and MPPT. The latter is preferable because it makes better use of the panels' output, but it also costs a little more.
We bought the inverter in Lviv for UAH 14,200, and the ordering process was incredibly quick. We picked a model, found a good price through Hotline, called, discussed it, placed the order, and picked it up from the post office the next day.
Besides the inverter, the box included cable holders, a 200 A fuse (a useful thing), a software disc, and cables for connecting to a computer. The RS232 cable connects via an RJ45 connector.


The design is pretty familiar: a box that mounts on the wall, with a display and control buttons on the front.


1. Connectors for connecting to a computer on the left
2. Power switch on the right, although it works a little strangely because it doesn't turn everything off. The warranty seal is also located here.
3. Mounting holes on top, plus a couple of seals covering unused holes :)
4. Display and buttons. The controls are reasonably convenient and the indicators are clear, but in practice I recommend controlling it from a PC.


Unfortunately, the warranty seal stopped me from opening the inverter to inspect and show its internals. The inverter isn't mine; I think I would have ignored the warranty on my own. You can see a little of the components through the openings, and from what I could make out, it looks like a decent mid-range unit.


Here you can see the mistake we made because we didn't know how inverters for solar systems work.
The inverter can charge the battery at up to 40 A from the panels, up to 25 A from the grid, or up to 60 A when charging from both the grid and panels at the same time. These settings are configurable.
However, this inverter model can't use more than 1000 W from the panels, regardless of whether the power goes to charging or supplying the load.
For spring, fall, and winter, this isn't a problem in our setup, but in summer we won't be able to use all the energy from the panels.
There's a version of this inverter marked “Plus” that can use up to 1500 W from the panels, but we couldn't find it for sale.

The inverter can deliver up to 3000 VA, but in that mode it will draw 1000 W from the panels even during the day and get the rest from the batteries.

I think this is because the inverter is in the budget class, so some compromises are inevitable.


There are lots of terminals on the bottom for connecting the grid (input and output), battery, and panels. I have no complaints about most of them, but the battery terminals look a little flimsy for the maximum rated current of 140 A.
That said, the wires clamp down firmly enough, but given the design, I recommend using a screwdriver with an insulated shaft.


While we waited for the other components, we went to the store and bought electrical cable, trunking, and one very important component: a live-and-neutral changeover switch.


This switch looks like a regular double-pole circuit breaker, but unlike one, it has three positions and six contacts instead of four, and no overload trip mechanism.


It's used to bypass the inverter manually, so that if the inverter fails or needs maintenance, you can connect the load without going through it.


In the simplified version (the full version switches both live and neutral), it looks like this:
The inverter is powered through a C25 breaker, so it can be completely disconnected if needed.
A C16 breaker provides bypass power when working on the inverter.
The 32 A switch lets you connect the load either to the inverter output or to the bypass line, or disconnect it from everything (middle position).

In principle, I could have installed a C16 breaker after the switch, but I don't remember why I didn't; there was some safety-related reason.


Now let's move on to the batteries.
First, a quick explanation of the battery types used in systems like this.
1. Lead-acid: the classic option, but a dull and pointless one, since these batteries aren't really designed for this kind of use, and the ones that are cost a lot more.
2. LiFePO4 is a compromise: a service life of up to 5,000–7,000 cycles and the ability to handle high charging and discharging currents.
3. LTO, or lithium titanate, is for those with money to spare who want a system that will last. It has plenty of advantages, but one drawback: the price.
4. “Conventional” lithium-ion batteries are pretty good and efficient, but they require more sophisticated charge control because they’re less safe than the types listed above.

We chose LiFePO4 batteries because they’re more reliable and efficient than lead-acid batteries, and less expensive than lithium titanate.

Actually, we ordered them before the inverter, but they took noticeably longer to arrive, so we got them later.
I ordered them from a company I work with. At first I wanted to buy new batteries and build a 50 Ah, 24 V battery pack, but they offered me an alternative: used batteries with a capacity of 170 Ah. As I understand it, they were originally installed in a UPS and had a capacity of 180 Ah; their remaining capacity is about 170 Ah.

The complete battery weighs about 45 kg. It’s fairly compact, although there are more compact alternatives, such as other LiFePO4 or lithium-ion batteries, which offer a better weight/size/capacity ratio.


The batteries had different charge levels: some were clearly closer to 100%, while others were probably around 30%. Their impedance was more or less similar, although a couple of them showed a noticeable difference.


The capacity of every battery was measured, and before I placed the order they asked whether to select for maximum capacity or the closest match. Since they would all be connected in series, I chose the second option.
So the lowest capacity is 169.92 Ah and the highest is 173 Ah. Without an active balancer, the battery pack’s capacity will be limited by its weakest battery—roughly 170 Ah.
With a LiFePO4 cell voltage of 3.2 V, the battery pack’s capacity will be 3.2 × 170 × 8 = 4,352 Wh, or 4.35 kWh.

The usable capacity will be lower because I set conservative charging and discharging limits. I’m expecting closer to 4 kWh, which is still four times more than what we originally planned to build.
This battery pack cost about 19,000 UAH. They gave me a discount as an old friend, so we went with this option.


But before assembling the battery pack, you need to balance it—that is, equalize the charge across all the cells. To do this, connect all the cells in parallel. Note that the wires are connected diagonally; this improves charging speed and uniformity.

WARNING: BE EXTREMELY CAREFUL AT THIS STAGE. IF YOU SHORT IT, THE 2.5 mm WIRE WILL SIMPLY BURN UP, AND YOU COULD GET BURNED!
While taking it apart, I shorted one cell. The wire didn’t have time to burn through, but it was an unpleasant experience, and I had to balance the cells all over again.

Anyway, I connected everything, hooked it up to a power supply set to a maximum current of 24 A, and started charging.


To speed things up, I initially set the charge termination voltage to 3.7 V. After about a day, the charging current had dropped almost to zero, so I lowered the voltage to 3.65 V, and the current went to zero. By the way, the voltage drops very slowly, which feels unusual for LiFePO4.
After that, I stopped charging, waited a few hours, then started again. The current initially rose to 10–15 A, but quickly dropped to 100 mA.

The full balancing cycle took more than a day, but since we were waiting for the panels anyway, it didn’t hold us up. After finishing and assembling the battery pack, I noticed that the voltage on two cells was still lower than on the others after a while. When I tried charging them, their voltage immediately rose to the same level as the rest, so this is simply a characteristic of those cells.


While we were working on the battery pack, the special mounting rails arrived. They’re 4,200 mm long, and we ordered four.


We also ordered various fasteners right away, plus some offcuts of the rails, since we planned to mount the panels on a separate structure. We also used eight mounting points per panel instead of four for better wind resistance. Special thanks to Marina, who works at the company and coordinated everything, both with the panels and the mounting hardware.

One item worth mentioning is the special studs used to fasten the structure to the roof beams and adjust it all into a single plane. The studs have rubber seals, are made of stainless steel, and the screw tips are clearly hardened, most likely by induction. All in all, they’re definitely worth the money. We saw something similar at a fastener shop, but only similar: the quality was completely different, and there was no rubber seal.


Special inserts are used to secure the panels to the rails. In addition to their complex shape, they have a built-in spring-loaded ball that makes installation easier.


While waiting for the panels, we decided to redo the wiring and install the inverter and battery. We also ran the cables up to the roof at this stage. The cables inside the house are routed through trunking. The run up to the roof is in a reinforced hose, and the exit is sealed with silicone.

The battery is connected with a hybrid cable: each terminal has 3x4 mm² ШВВП plus 6 mm² solar cable. The first cable was left over from the grid connection; the second was bought specifically, with some extra length.
The inverter’s grid input is connected with 3x4 mm² cable, and its output with 2x2.5 mm² cable. The ground is common anyway and comes from the same panel, while 2.5 mm² is sufficient for the stated 3000 VA.

There’s also a fuse on the battery, supplied with the inverter. I mounted it on a separate pair of terminals to keep mechanical strain off it, which could happen if the output cables were connected directly to it.

The inverter is installed on the second floor of the house to improve the battery’s thermal conditions. That revealed another miscalculation: it’s great for the battery, of course, but the inverter is quite noisy. What’s more, its fan switches on based on a power threshold, not temperature.


We considered several ways to assemble the battery, but we needed the terminals closer to the front left edge. The second option worked best, and it also made the assembly more compact.

I mentioned above that the cells vary in capacity. In an assembly, I recommend placing cells with the greatest capacity differences next to each other. For example, if you have four cells rated at 170, 171, 172, and 173 Ah, it’s better to arrange them as 170, 173, 171, 172 Ah. Why? It comes down to the balancer: it’s better at equalizing differences between neighboring cells than at shuttling energy across the entire battery from the weakest cell to the one with the highest capacity.


We used terminals for 6 and 10 mm² wire, but since we didn’t have a crimping tool, we had to fill them with solder instead.
The battery interconnects were a little trickier. I didn’t like the ones supplied with the kit, but a friend had a 1000x20x2 mm busbar made of anodized aluminum.
We cut it into strips and drilled holes, including a pair of threaded holes for connecting the balancers.

Everything seemed fine, except anodizing doesn’t conduct electricity, so we had to sand it all off. The result was a set of neat busbars with a cross-section of 40 mm².


Speaking of cable, we also bought it from the same company that supplied the panels and mounting hardware. It was advertised as a special cable made in Germany. I liked it, although I forgot to measure its actual resistance, and I can’t do that just yet. It has double, tough insulation, stranded conductors, and tinned wires—in short, I like it. Before ordering, I read the datasheet; it really is suitable for outdoor use, and it costs no more than ordinary cable.



After connecting the inverter and battery, we started the system without the panels, so it was working as an uninterruptible power supply.
Here’s the program menu. Sorry these aren’t screenshots; I just took photos for my own reference.


In this mode, the system is only useful for things like those “flying aces” who crash into utility poles and cut power to nearby houses. The photo shows a real example of this system coming in handy: my friend may not even have realized right away that the street had lost power.


Some time later, the panels arrived—four of them, at about $166 each. We chose a local company for two reasons:
1. Local warranty service
2. Free delivery to the house.

The only downside was having to wait a little while for them to gather enough orders and add our four panels to the shipment.


There’s a datasheet for them, but all the information we need is on the nameplate anyway.
At first, they sent us a datasheet for a slightly different model with the same power rating, but dimensions of about 2000x1000 mm. Later, they wrote that we’d be getting another model, measuring closer to 1750x1100 mm, which is a little more convenient.

When choosing a panel-and-inverter setup, you need to consider the panels’ output voltage and the inverter’s input voltage.
For our inverter, the operating range is 30–80 V, with up to 100 V open-circuit voltage.
The first panels they offered had an operating voltage of 40.55 V and an open-circuit voltage of 48.5 V.
So with this inverter, those panels can only be connected in parallel, since two in series would give an operating voltage of 81 V.

The second option has an operating voltage of 34.64 V and an open-circuit voltage of 41.6 V, so the panels can be connected in pairs in series, giving about 70 V operating voltage and 83 V open-circuit voltage.
This nearly halves the current in the cables, which means power losses in them will be reduced by a factor of four! By my estimates, losses in the first configuration were about 55 W; with the 2S2P setup, losses were only 11 W.

There was also a question about which is better, polycrystalline or monocrystalline. As I understand it, the former is better and more durable, but more expensive and harder to find, so in the end we chose the latter.


Screenshot from the datasheet for the first panel option, which was mistakenly recommended to us


And the second option, which is the one we installed.
I should point out that both datasheets are confusing: the first lists cable lengths of 270 and 100 mm, while the second lists 350 mm for both. In reality, the wires were long enough to connect the panels in series without extending them.


We planned to attach the whole setup to the roof beams, initially using three beams. But after measuring everything and drawing up a plan, we realized we would need to attach it at four points, which meant using four brackets per pair of panels. That actually turned out to be better.
We drew the plan to scale, which helped us a lot later.


As I wrote, we wanted to set the panels at a 45-degree angle. The roof is pitched at 22 degrees, so we needed to make brackets with a 33-degree angle. I'm not very good at geometry, so I simply launched Sprint Layout and drew what I needed, with the right angles, then measured the bracket dimensions :)
We also planned to add a cross brace between the central brackets to make the structure more rigid. It was made, but during installation, “something went wrong,” and it needs to be redone. Actually, I had originally planned to fit it on site, but the person who helped make the brackets built it as a single rigid piece. In some ways, that was good—even excellent—but in this case, the spacing between the brackets was determined by the shape of the roof, so it didn't quite match the specified dimensions.


The brackets arrived last, and as soon as we got them, we started drilling and getting everything ready. We wanted to finish in a day, but even with two friends helping with the prep work and speeding things up, we still couldn't get everything done as planned.
By the end of the first day—or rather, half a day—we had prepared the brackets and installed the first pair. We even managed to level them, although we had to adjust them again later.
Honestly, the idea of using brackets was a good one, but it was complicated and added extra expense. Maybe we were being overly cautious and could have made do with some supports from metal profile and skipped welding the whole structure, but we did it this way, so that's that.


In the same Sprint Layout project, I modeled how the sun would shine in winter, then calculated the spacing between the rows so the panels wouldn't shade one another. We also left room for a third row of panels in case we upgrade the system.
We also left some clearance beneath the panels so snow could slide underneath instead of piling up. We hope the 45-degree angle will also help keep snow from accumulating on the panels themselves.


On the second day, we thought we'd just, uh... get everything finished quickly. Yeah, right...
We did finish, but not quickly—in fact, it was already late in the evening.
Installing the brackets caused the most trouble. Now I realize we could have done it not only faster and more conveniently, but more securely too. We should have attached the brackets wherever it was convenient, rather than adjusting their height relative to the roof, and then used threaded rods or long bolts to adjust the rails relative to the brackets. Especially since we were given special plates for this.

We finished with the brackets at around 18:30.


Then the three of us—me, a friend, and a neighbor (special thanks to him for the useful advice and help fastening the brackets)—hauled the panels up onto the roof, and the two of us got to work installing them.
At around 10 p.m., we finished the rough assembly and even connected the wires. The instructions say not to connect or disconnect the wires under load, but since it was already evening, the current was low, which actually made the job a little easier for us :)


We used standard connectors to connect the panels—two sets, each consisting of a male and female connector. The panels were connected to one another using their original cables and connectors.
We routed the wires into junction boxes, soldered on terminals, and bolted them together. We installed two boxes, one for the positive wire and one for the negative, and screwed the boxes to the brackets.


A few words about the wiring. The panels are connected in a 2S2P configuration, but we left the option to upgrade to 2S3P. The diagram shows both the installation layout and the wiring configuration (top right). At the top is one way to connect two additional panels.

As for upgrading the system, to make it work efficiently the additional panels need to be either exactly the same or have the same output voltage. This is usually determined by how the cells on the panel are wired. For example, the first panels we were considering had a 6x12+6x12 configuration, while the second ones had 5x12+5x12 and a lower output voltage.


The very next day, we climbed up and tidied up the whole setup. Just to be safe, we didn't use plastic cable ties and secured the wires with copper wire instead.


And here it is, almost finished. “Almost” because we still need to install the crosspiece and seal one hole in the roof that turned out to be unnecessary... By the way, rubber washers on the threaded rods don't eliminate the need for sealant—unless, of course, you want your attic to stay dry.


Now let's see how it turned out.
First, here's the program's main window. It's 10:30, but the panels are already delivering more than 500 W.


Interestingly, after connecting the panels, information about a second CPU appeared at the bottom of the window. It wasn't there while the unit was operating as a UPS.


The settings menu lets you configure the operating modes, voltage thresholds, charging current for different modes, and so on.
Since the system is slightly underloaded in practice, by midday the battery would be fully charged, the load would be low, and the sun would still be shining, causing the panel voltage to rise to almost 80 V. To address this, I decided to extend the charging time by limiting the current to 20 A.
Charge/discharge voltage thresholds:
Charge complete — 28.8 V (3.6 V per cell)
Float mode, entered after a full charge — 27.6 V (3.45 V per cell)
Over-discharge — 22.4 V (2.8 V per cell)

Operating mode: SBU — Solar, Battery, Utility. In other words, as long as there's sun, it powers the load from the panels; when there's no sun, it switches to the battery; and if the battery runs down, it switches to utility power. Later, I also changed the utility mode to UPS. That sets the low utility-voltage threshold for switching to battery at 175 V, compared with about 100 V in the regular mode, which is critical for a refrigerator.


While monitoring the system, I saw the panels delivering 831 W. But note that this isn't a limit imposed by the panels: the combined charging power and load consumption simply weren't very high, so the remaining power went unused. As I mentioned, though, the inverter itself can't use more than 1000 W from the panels.


Here are a few graphs saved over 10 days.

Input power from the panels peaked at 943 W. That's already close to the inverter's maximum input, but even under these conditions the limiting factor was the load, not the panel output—they can deliver more.


Panel voltage peaked at 80.8 V, which shows that the panels are underloaded and still have some headroom. That extra capacity will come in handy in autumn and spring.


Load power also shows why the panels are underloaded: there are only brief peaks at high power, while the rest of the time it hovers around 250–300 W. Add roughly 550 W for battery charging, and that gives us a total of 800–850 W.
The graph shows both apparent and active power.


Battery voltage: the lowest recorded value was 25.93 V.
By the way, we didn't install a protection board, although that was the original plan. The thing is, the inverter won't work without a battery, and if the battery protection board disconnects it due to overcharging or over-discharging, getting the system running again will be difficult, since the inverter won't know what happened.
We decided to make do with a fuse on the battery, the inverter's settings, and two balancers: a "coarse" one rated for 10 A and a "fine" one rated for 1.2 A. They're on order and still on their way.


I could leave it there, but I wanted to say that I really enjoy climbing around on rooftops. Maybe deep down I'm a cat :)


And here are a couple of photos from near this spot. If you didn't know where this was, you might not believe there could be such overgrown vegetation here. But I think it's beautiful and unusual in its own way :)



And here's our chief assistant, keeping the cats from stealing the nuts and bolts :)


That's everything. While we're at it, here are answers to some questions you might have:
1. What's the payback period?
There isn't one. This system won't pay for itself in this configuration, at least not unless electricity rates go up significantly.
2. How much did it all cost?
About $1,900, not including the balancers. We originally planned to keep it within $1,000–1,200, but our ambitions grew, so although it ended up costing one and a half times as much, we got:
An inverter output rating of 3,000 VA instead of 1,500 VA
1,620 W of panel capacity (claimed), instead of 1,000–1,200 W
A 4 kWh battery instead of 1 kWh
We hadn't accounted for the mounting brackets at first.
So I think it turned out pretty well, even taking our mistakes into account.
3. What mistakes or shortcomings were there?
Well, sometimes they weren't really mistakes, just a lack of information. In one case, even the datasheet didn't help. Still, there are a few things I'd do differently now, for example:
I'd choose a different inverter, or at least the same model from the "Plus" series. Unfortunately, they're not available for sale.
I'd change the design of the panel mounts and their adjustment mechanism.
I'd add an RS232-WiFi module. That's still on the to-do list, just for later.
I'd look for a different place to install the inverter. During the day, this thing is as loud as a plane taking off. It has two fans, one for the charger and one for the inverter, and both are noisy during the day...
4. Why add more panels?
In winter, output drops sharply, and there'll only just be enough energy for the fridge and boiler (if there's no grid power at all). Adding panels would let us run the lights too.
5. How much electricity does the system generate per day?
A better question is how much it can generate, because right now it produces exactly as much as the house consumes, plus the inverter and battery-charging losses. That's about 3.5–4 kWh. All the lights in the house, TVs, computers, the fridge, and the boiler are connected to it.
6. Can you do it?
Of course you can. As a famous actor said in an equally famous movie:
What can you do for money?
For money? I can do... anything :)

That's about it. I'd be glad to hear your questions, advice, personal experience building similar systems, and practical experience using batteries.