The battery that charges from the grid: a real month at an industrial site
An industrial site with 169 kWp of PV and a 672 kWh battery paid PLN 5,513 less in July 2026 than it would have without the battery. Not a simulation - a month that actually happened. Almost 40% of the saving comes from the capacity fee, and the battery charged 85% from the grid.
An industrial site with 168.82 kWp of PV and a 672 kWh / 100 kW battery paid PLN 5,513 net less in July 2026 than it would have without the battery, and PLN 4,544 less than with the same battery on plain self-consumption control. This is not a simulation: it is a month that actually happened, and the headline figures come from the same billing the client sees in their own panel.
A month that actually happened
On this blog we have shown simulations before: what a battery would save under AI control. This time there is nothing to assume. The site below has been running under EnergyPal control for months, so we can simply open one full month of meter data - July 2026 - and settle the bill three ways.
- a 168.82 kWp PV installation in zero-export mode,
- a 672 kWh / 100 kW battery,
- the B23 distribution tariff, contracted capacity 194.75 kW,
- energy bought at dynamic hourly day-ahead prices with a trading margin.
The month in numbers: the plant consumed 63.2 MWh, PV produced 25.1 MWh (149 kWh/kWp in July), grid import was 40.7 MWh, and the battery worked through 11.7 equivalent full cycles. Peak grid draw reached 189.8 kW, just under the 194.75 kW contracted limit - that ceiling is part of every decision the algorithm makes.
The paradox: 99.8% self-consumption and a battery that charges from the grid
Here is the number that surprises everyone: this site consumes 99.8% of its PV production on the spot. Over the whole month only 47 kWh left the site. PV covers about 36% of the plant's consumption, so there is practically no solar surplus to store.
In the popular imagination a battery is a jar for your own solar energy. At this site that jar would stay empty. So where did the 9,696 kWh that flowed into the battery come from?
85% from the grid (8,212 kWh, versus 1,484 kWh from PV). The algorithm buys energy in the cheapest hours of the day, stores it, and the battery then covers 7,824 kWh of the plant's consumption in the expensive hours. The battery is not a solar jar here. It is a tool for playing the price curve and shaping the load profile - and that second job turns out to matter even more than the first.
Where the PLN 5,513 came from
We settled the same month three ways, with an identical pricing formula: the site as it actually ran under EnergyPal control, the site with no battery at all, and the site with the same battery driven by a plain self-consumption rule.
Active energy + variable distribution + capacity fee + battery wear. Scenarios compared on common metering slots.
| Scenario | Month cost | vs EnergyPal |
|---|---|---|
| No battery | PLN 28,545 | +PLN 5,513 |
| Battery on self-consumption | PLN 27,576 | +PLN 4,544 |
| Battery under EnergyPal control | PLN 23,032 | baseline |
The PLN 5,513 saving versus no battery breaks down into: PLN 3,874 on active energy, PLN 2,182 on the capacity fee, PLN 398 on variable distribution, minus PLN 941 of battery cell wear, which we always subtract - 11.7 full cycles a month is real work and it has a real cost. On average the control saved PLN 178 net per day.
Almost 40% of the saving has nothing to do with energy prices
The second-largest line is the capacity fee: PLN 2,182 of the PLN 5,513 saving. The fee is charged on every kWh imported on working days between 7:00 and 22:00, and its rate depends on the shape of the daily load profile: each day is assigned a K1-K4 band by comparing peak-hour draw with off-peak draw. Flat profile, K1, you pay 17% of the full rate; peaky profile, K4, you pay 100%.
Without a battery this site is a textbook K4: its peak-hour draw runs about 49% above off-peak. Under EnergyPal control the battery deliberately shifts import out of the peak window, and the profile flattens to a delta of about 2.1% - K1 instead of K4 on most days. Interestingly, the self-consumption variant also stays in K4 (46%): following the sun does nothing to flatten a working plant's evening draw. We wrote a separate guide to how the capacity fee works; this month is that guide happening in production.
The same battery on self-consumption barely works
The most instructive column of the comparison is the middle one. Give this site the exact same battery, but drive it with the classic rule - store PV surplus, discharge when PV falls short - and it saves PLN 969 a month. EnergyPal control saves PLN 5,513, which is PLN 4,544 more.
The wear line tells you why: PLN 120 versus PLN 941. Battery wear is proportional to the energy cycled, so on self-consumption the battery cycled roughly eight times less - at a site with no PV surplus it simply stands idle. Over 80% of the benefit of this installation comes from the control, not from owning the hardware. The same steel and lithium, a different brain, a PLN 4,544 monthly difference.
The days the algorithm deliberately paid extra
A month of control is not thirty-one identical wins. The daily view shows 3 days in the red - days when the site paid more than it would have without the battery:
Daily scenario pricing (energy + distribution + capacity fee, before battery wear) - it shows the shape of the month; the headline figures come from the monthly settlement.
Look at Sunday 12.07: the system ended it PLN 70 behind. Sunday electricity is cheap, so the algorithm spent the day buying and storing it. On Monday 13.07 that stored energy carried the plant through the expensive hours: PLN 446 ahead, one of the best days of the month. A rule that judges itself day by day would never make that trade. The algorithm plays for the month, not for the day.
The best single day, 29.07, brought PLN 618. And 3.07 saw negative day-ahead prices: the system charged the battery at full power, pushing grid draw close to the contracted limit, while the market paid for consumption. Honestly, at this site that spectacular-sounding day added a modest PLN 45 - the quiet, repeatable money in this profile is in flattening the peaks, not in chasing price anomalies.
How we count (and what we do not)
- the headline figures come from the same settlement that powers the client's dashboard - the article and the client's own panel show one number, not two,
- the no-battery and self-consumption variants are the same month re-settled with an identical pricing formula, on the same metering slots,
- the capacity fee is computed with the daily K1-K4 band formula for every scenario separately,
- battery cell wear is subtracted from the saving, not hidden,
- all amounts are net (before VAT).
What this means for your site
If your plant runs on B23 with a high self-consumption PV installation, the takeaways travel well: a battery does not need solar surplus to save; the capacity fee is often the single easiest line to cut, because it rewards the profile shape, not the energy amount; and the gap between a simple rule and forecast-driven control is not a few percent - in this month it was the difference between PLN 969 and PLN 5,513. We report a month like this to every client, every month, with the same honesty: wear subtracted, red days included. Get in touch if you want to see what your month would look like.
