Sizing an Off-Grid Solar System for Always-On Homelab Loads

Solar for a homelab is a harder problem than solar for a house, because the load never stops. A residential system can lean on the grid overnight and shrug at a cloudy week. Equipment that must stay up has to survive the worst stretch of weather your location produces, which means the design starts at the load and ends at the panels, never the other way around.

Measure the load before buying anything

Put a meter inline and record actual consumption over several days, including idle periods and whatever spikes happen during backups or heavy jobs. Nameplate ratings on power supplies describe a maximum the equipment will probably never reach, so sizing from them wastes money. What you want is average watts across a full day, which converts directly to watt hours, and a separate figure for peak draw, which matters only for inverter selection.

Once that daily watt-hour figure exists, the solar and battery sizer will turn it into a first-pass array wattage and bank capacity in a few seconds, which is enough to decide whether the project is viable before any of the detail below matters.

This is also the cheapest point to reduce the problem. Every watt removed from a continuous load removes it from the battery bank, the array and the charge controller simultaneously. Consolidating onto fewer, more efficient machines usually costs less than the solar capacity needed to feed inefficient ones.

One line item is routinely missed at this stage: the inverter’s own idle consumption. An inverter that draws 25 W simply being switched on adds 600 Wh to every single day, whether or not anything is plugged into it. Against a 100 W continuous load that is a quarter of the total, paid every day for the life of the system. Check the no-load draw on the datasheet before the model is chosen, and consider whether any of the equipment can be fed from the DC bus directly, since a DC-DC converter avoids the inverter round trip entirely for the loads that can take it.

A worked example to anchor the numbers

Take a homelab drawing a steady 100 W. The arithmetic runs in one direction, load outward:

• Daily energy at the load: 100 W × 24 h = 2.4 kWh/day.
• Daily energy off the battery: at roughly 90 % inverter efficiency, about 2.67 kWh/day.
• Bank size for two days of autonomy: 2.67 × 2 = 5.33 kWh delivered. Keeping a lithium iron phosphate bank inside about 80 % depth of discharge for cycle life means a nameplate capacity near 6.7 kWh, which at a 48 V nominal bank is roughly 140 Ah.
• Array size: replacing 2.67 kWh plus charging losses is about 2.9 kWh a day. In a worst month delivering three peak sun hours, and derating by about 25 % for temperature, soiling, mismatch and controller losses, that needs roughly 1.3 kW of panels.
• Charge controller: 1.3 kW into a 48 V bank is about 27 A of charge current, so a 30 A or larger MPPT unit with headroom.

Change the autonomy figure and watch what happens: three days instead of two takes the bank past 10 kWh nameplate without touching the array. That is the sensitivity that makes autonomy the most expensive decision in the design, and it is why the honest answer for many sites is two days of storage plus a generator or a grid tie for the genuinely bad weeks, rather than a bank sized for the worst week of the decade.

Battery bank sizing and chemistry

The bank has to carry the load through darkness plus a margin for consecutive poor solar days. Decide how many days of autonomy you actually need, because that number drives what is typically the largest single cost in the system.

Lithium iron phosphate has largely displaced lead acid for this role. It tolerates much deeper discharge without damage, delivers far more cycles, holds voltage more steadily under load, and does not need the periodic absorption behavior lead acid demands. Two constraints matter. It requires a battery management system, which is not optional and should not be improvised. And charging below freezing damages the cells, so any bank in an unheated space needs either low temperature charge cutoff, heating, or both. Discharging in the cold is far less harmful than charging in it.

Higher bank voltage reduces current for the same power, which means thinner cable and smaller losses. That is why larger systems move away from low voltage banks. The same 1 kW of load draws about 83 A at 12 V and about 21 A at 48 V, and cable cost and voltage drop both scale with that current. Anything much above a few hundred watts of continuous draw is easier and cheaper at 24 V or 48 V than at 12 V, and the equipment ecosystem above 1 kW largely assumes 48 V.

Buy the bank once, and buy it whole. Cells age together, and a bank assembled from a mixture of ages and states of health is limited by its worst member, so adding capacity later by bolting a new string onto an old one rarely delivers what the arithmetic promises. It is also worth fitting a shunt-based battery monitor at the same time as the bank. Inferring state of charge from resting voltage is unreliable on lithium iron phosphate specifically, because its discharge curve is deliberately flat: the difference between comfortably charged and nearly empty can be a fraction of a volt, which is exactly the situation where a coulomb-counting monitor earns its price.

When a bank that was sized correctly stops accepting charge, the cause is almost always a protection device, a set point or a connection rather than a failed cell. The order to work through is in solar batteries not charging: how to diagnose it.

Charge controller and array

An MPPT controller continuously finds the panel array’s maximum power point and converts the excess voltage into usable current. A PWM controller simply connects the array to the battery and discards the difference. MPPT costs more and is worth it in almost any system large enough to run equipment, particularly in cold and low light conditions where panel voltage rises. The full comparison, including the module compatibility constraint that usually decides the choice before efficiency does, is in MPPT vs PWM charge controllers for homelab solar.

Size the array against the worst solar month at your latitude, not the annual average, and account for panel output falling as cells heat up. The arithmetic, the derate factors and worked panel counts for a range of locations are set out in solar panels for a homelab: how many do you need?. Wiring panels in series raises voltage and keeps current low, which suits long runs, but the total open circuit voltage on the coldest expected morning must stay under the controller’s limit. Cold weather raises panel voltage, and that is how controllers get destroyed.

Inverter and the wiring nobody photographs

Choose the inverter for continuous load with headroom for surge, since motors and some power supplies draw far more at startup than they do running. Split phase output matters only if you have loads that need it.

The unglamorous parts are the ones that fail dangerously. Batteries can deliver enormous fault current, so a correctly rated fuse or breaker belongs as close to the battery terminal as possible. Size DC conductors for voltage drop, not just ampacity. Do not mix aged cells with new ones in the same bank. Follow local electrical code, and get an inspection where one is required.

Plan the failure, not just the sunshine

A system sized for a continuous load will still run out of energy eventually, and the question is what happens when it does. Left to itself, the battery management system disconnects at its low-voltage cutoff and everything loses power at once, mid-write, with no warning. That is a worse outcome than a planned shutdown.

The fix is to treat state of charge as a signal rather than a statistic. Most monitors and inverters expose it over a serial link or a network API, which is enough to trigger a graceful shutdown of non-essential machines at one threshold and the remaining ones at a lower threshold, then bring them back once charging has recovered a margin. Set those thresholds above the BMS cutoff, not at it: the cutoff is a protection device, not an operating limit, and a bank that reaches it regularly is a bank that will not last its rated cycles.

Two more things belong in the same plan. Decide in advance which loads are genuinely non-negotiable, because a bad week is much cheaper to survive at 40 W than at 100 W. And log the daily harvest from the first week, since a month of real data from the actual site beats any estimate — including the one above — and it is the only way to find out that a tree shades the array for two hours every afternoon in winter.

Where to go next

• Solar panels for a homelab: how many do you need? turns a daily watt-hour figure into a panel count with the derate factors worked through.
• MPPT vs PWM charge controllers for homelab solar covers the controller choice and the module compatibility trap behind it.
• Solar batteries not charging: how to diagnose it is the fault-finding order for the day the bank stops taking charge.
• The solar and battery sizer runs the arithmetic above against your own numbers.