Grid-Tied vs. Hybrid Solar Systems: Which Works Better for Businesses

Your tariff clock decides which system pays. On a 1 MW array, grid-tied returns 572,000 yuan a year with no battery to buy. A 2.2 million yuan battery lifts that to 881,000 yuan on one daily cycle, or 1.22 million on two. Clear a 0.6 yuan/kWh spread with two cycles a day and hybrid wins; under 0.43 yuan with no room for a second cycle, grid-tied does.

Every number in that table comes off the same roof. 1 MW of modules at 550 W each comes to 1,818 units, or 4,690 sq m of module area once the racking is counted. Add walkways and shading gaps and you're at about 9,000 sq m of roof, the footprint that drives what a 1 MW build costs.

The site gets 1,100 full-load hours a year. A 2 MWh battery at 90% depth of discharge only gives you 1,800 kWh to work with, so a single cycle drains it. Charging back up at 1 MW takes 2.0 hours.

The revised Distributed Generation Hosting Capacity Assessment Guide took effect on 18 June and scrapped the old 80% cap on reverse load at the transformer. In its place, the operator scores each distribution area green, yellow or red.

Yellow and red want storage before the utility approves a hookup. For a plant sitting in one of those two zones, the battery is now a permit condition rather than an optional extra.

Nationally, new-type storage hit 153 GW / 396 GWh by the end of June 2026, up 61% year on year^[1]. Solar connections reached 1,272 GW, with 71.77 GW added in the first half of 2026^[2].

Distributed projects make up 45.3% of that installed base, and China still makes around 80% of the world's modules^[3]. The IEA puts global renewable capacity growth at 4,600 GW through 2030, almost 80% of it solar PV, with China supplying nearly 60%^[4].

Cost and lifespan

Two-hour storage systems cleared at 602.1 yuan/kWh on average in the first half of 2026. Two-hour EPC landed at 1,040.4 yuan/kWh, up 8.8% and 3.2% from a year earlier. Every calculation below uses 1,100 yuan/kWh.

A commercial cabinet costs 1,176 yuan/kWh on an EPC basis, against 1,040.4 yuan/kWh for a container EPC at the same two-hour duration, a 13% premium. Hardware alone carries a wider gap: 756 yuan/kWh for the cabinet against 602.1 yuan/kWh for the container system, or 26%.

Installation, civil work and grid hookup make up the remaining 420 yuan of the cabinet price, or 36% of the bill.

Cells eat the biggest slice of the hardware. 314 Ah LFP cells went from 0.31 yuan/Wh in late 2025 to 0.37–0.40 yuan/Wh by June 2026, a 19% to 29% climb. At 0.37 yuan, 2,000 kWh of them run 740,000 yuan.

Cells alone account for 49% of the 1.512 million yuan hardware bill.

Two certificates matter at purchase, one mandatory and one voluntary. GB 44240-2024 took effect on 1 August 2025 and is the first mandatory national standard for storage, covering everything from cell up to pack^[5].

GB/T 36276-2023 came in on 1 July 2024. Cycle testing runs at 45°C, and capacity retention of 80% is where the test stops^[6].

Battery warranties usually promise 10 years or 6,000 cycles, whichever comes first. Warranties on modules built for commercial and industrial rooftops promise 84.8% of initial output at year 25, or 2% degradation in year one and 0.55% a year after. Both come off the manufacturer's own paperwork, so read the cycle count, the degradation curve and the temperature range line by line before you sign anything.

The yearly revenue gap

The array is 1 MW and makes 1.1 million kWh a year, the same scale as a commercial rooftop already running in Switzerland. The plant soaks up 60% of that during the day, so 660 MWh. The other 440 MWh goes to the grid at 0.25 yuan/kWh. Power you don't have to buy is worth 0.7 yuan/kWh.

These prices are province-level ballparks and can swing 20% to 30% between regions. The 0.7 yuan/kWh self-use figure sits in the common band for commercial tariffs. The 0.25 yuan valley and 1.0 yuan peak come from a 4:1 spread. The 42 yuan/kW per month demand charge matches a typical 10 kV two-part tariff. Pull your own 12 months of bills before you commit to anything.

Grid-tied, all year

660 MWh at 0.7 yuan gives you 462,000 yuan. 440 MWh at 0.25 yuan gives you 110,000 yuan. The two together come to 572,000 yuan a year, and no cycle limit applies.

Over 25 years, allowing for 2% first-year degradation and 0.55% a year after, that roof produces 25,140 MWh and 13.08 million yuan in revenue.

Modules last 25 years, inverters 10, batteries 10. Inverters get swapped twice in 25 years. A battery running one cycle a day lasts 18.2 years and gets swapped once; at two cycles a day it is spent in 9.1 years, so 25 years takes two swaps.

What the battery adds

Power you used to sell is the biggest of the three streams. Those 440 MWh earned 110,000 yuan at the 0.25 yuan export rate. Through the battery, 88% round-trip efficiency turns them into 387 MWh displacing 0.7 yuan power. That comes to 271,000 yuan. Net gain: 161,000 yuan.

Arbitrage puts the same battery to work a second time. Charge at the 0.25 yuan valley, discharge at the 1.0 yuan peak, and one cycle a day brings in 449 yuan, or 148,000 yuan a year.

Two-part tariff customers choose between billing on transformer size and on peak demand, and that choice opens a third stream. Cutting monthly peak from 800 kW to 600 kW at 42 yuan/kW takes 25% off the bill, or 101,000 yuan a year.

All three wins share one battery, so they fight each other. Shaving 200 kW of peak for two hours eats 400 kWh of discharge. The valley arbitrage cycle eats another 627 kWh. Together that is 1,027 kWh.

The squeeze takes 57% of the 1,800 kWh the battery can deliver in a day. Over 300 working days, peak shaving claims 120 MWh on its own, against the battery's 594 MWh of annual output spread across 330 days.

If the sharp peak falls inside the two discharge hours, one pass earns both the arbitrage and the demand cut. A sharp peak landing in a flat period leaves the two earnings squeezing each other.

Eight hours overnight is enough for four cycles, though you'll realistically use one or two. A 2 MWh battery discharging over 330 days puts out 594 MWh, or 54% of what the array makes. Running two cycles doubles that to 1,188 MWh, and no piece of hardware changes.

Charging soaks up 2,050 kWh a day, or 677 MWh over 330 days. Discharge is 594 MWh. The 83 MWh gap between them is your 12% round-trip loss.

With storage in the picture, self-consumption climbs to 1,047 MWh, or 95.2% of annual generation. One daily cycle yields 309,000 yuan, or 309 yuan per kW of battery per year against the 2,200 yuan/kW the battery costs. That is a 14% gross return.

Surplus shifting brings 161,000 yuan of that, or 52%. Valley arbitrage brings 148,000, or 48%.

A second pass charges at 0.55 yuan in the midday flat period and discharges at 1.2 yuan at the sharp peak, bringing in another 1,033 yuan a day, or 341,000 yuan a year.

Total climbs to 650,000 yuan, or 650 yuan per kW a year, a 29.5% gross return on the same 2,200 yuan/kW. The second cycle's spread is 0.65 yuan, a full 0.1 yuan below the first cycle's 0.75 yuan, yet it earns 1.1 times as much.

Those extra 330 cycles net 1,033 yuan each. Surplus shifting drops to 25% of the total, and the two arbitrage cycles take 75%.

Levelized cost sets the bar

A single daily cycle for 330 days uses only 3,300 of your 6,000 cycles over a 10-year warranty and delivers 5,940 MWh. Add 2.2 million yuan of hardware to 330,000 yuan of O&M and the levelized cost lands at 0.43 yuan/kWh.

The two-cycle case burns all 6,000 in 9 years for 10,690 MWh and lands at 0.23 yuan/kWh.

That 0.2 yuan gap is worth 1.19 million yuan across the 5,940 MWh one cycle a day delivers over ten years, and 2.14 million across the 10,690 MWh two cycles deliver over nine.

O&M at 1.5% of capex runs 33,000 yuan a year, 330,000 over ten years, or 15% of the hardware bill. Two cycles a day pays the same 33,000 yuan a year as one.

A one-cycle schedule nets 276,000 yuan a year, 2.76 million over ten. Take out the 2.2 million in hardware and 560,000 yuan stays. A two-cycle schedule nets 617,000 a year, 5.55 million over nine, and leaves 3.05 million after the 2.497 million comes back out.

The profit differs by 5.5 times on identical equipment.

Net of O&M, the same 2.2 million yuan returns 12.5% a year on one cycle and 28% on two. The 8.0-year payback leaves two years of profit inside the 10-year warranty, and the 3.6-year payback leaves 5.4 years inside the 9-year cycle life.

Levelized cost at 0.43 yuan runs 0.18 above the 0.25 yuan export rate and 0.27 below the 0.7 yuan self-use rate. That leaves 0.02 yuan of cushion inside a 0.45 yuan spread, and one battery swap in year 10 wipes it out.

The single-cycle case needs a 0.43 yuan/kWh spread to come out ahead. The two-cycle case needs only 0.23 yuan.

At a 0.6 yuan spread you've got 0.17 yuan of headroom on the first case and 0.37 on the second, off identical equipment.

Running two cycles a day eats 6,000 cycles in 9.1 years, so the first battery replacement lands in year 10. At today's prices that is another 2.2 million yuan, and it turns 3.05 million of ten-year profit into 850,000.

The US Department of Energy's Long-Duration Storage Shot targets a 90% cut in grid-scale storage costs within the decade^[7]. Today's prices are the conservative case for that swap.

The floor price is written into NDRC Document 1093 of 2021. Wherever the forecast peak-valley gap tops 40%, the spread must be at least 4:1. Everywhere else, 3:1. Sharp-peak rates have to run at least 20% above peak rates^[8].

Net of the 0.23 yuan levelized cost, the valley-to-peak pass clears 0.52 yuan/kWh and the flat-to-sharp-peak pass 0.42 yuan.

IRENA's 2025 figures put global PV LCOE at USD 44/MWh and Chinese centralized PV at USD 36/MWh. A PV-plus-four-hour-storage project hitting 95% reliability comes in under USD 85/MWh^[9]. That USD 49/MWh gap is what your battery has to claw back out of the tariff spread.

Grid rules and power prices

Chinese PV generated 1.17 trillion kWh in 2025 at a 95% utilization rate^[10]. The first half of 2026 came in at 655.5 billion kWh and 91.4%, down 3.6 points in six months, and most of that curtailment lands at midday.

The IEA lists China among the markets where curtailment is rising and negative-price hours are surging, both concentrated at peak solar output^[4].

DL/T 2041-2025 was issued on 30 December 2025 and replaced the 2019 version. The operator now works out hosting capacity county by county and substation by substation.

Green means connect on request. The two restricted zones need storage or a control device before approval, and a grid-tied project with no battery won't get a connection plan there.

Shanxi, Shandong, Anhui and Hebei have already published their rules. In a red zone with zero hosting capacity you can still build, as long as you add storage and keep every kWh on site.

After Document 136, all exported PV goes to market. Revenue comes from a mechanism price plus a settlement spread. Shandong's January 2026 reference price was 0.1448 yuan/kWh^[11]. The IEA lists China's move from fixed tariffs to auctions among the policy changes weighing on renewable project economics^[4].

At that rate those 440 MWh bring 64,000 yuan instead of 110,000, so 46,000 yuan goes missing. A hybrid system pushes surplus shifting from 161,000 to 207,000 yuan, absorbing 46,000 yuan of it.

The Distributed PV Development and Construction Measures require self-consumption projects to sit inside the same land boundary as the user^[12].

Multi-user green power direct connection, under NDRC Document 688 of 2026, sets floor self-consumption at 60% of available generation and 30% of total demand. Exports are capped at 20% of available generation^[13].

When to add storage

The roof and the 2.2 million yuan are identical either way. The battery brings 650,000 yuan a year on two cycles and 309,000 on one, so payback moves 4.4 years on nothing but the tariff structure.

State Grid territory ran new-type storage for 573 equivalent full-load hours in the first half of 2026. China Southern Grid ran 626, up 9 and 41 hours respectively. The south sits 53 hours ahead of the north^[3].

Transformer capacity is the other hard stop. Behind-the-meter storage carries three rules: no reverse flow into the grid, discharge power never above what the plant is drawing, and nothing above the contracted capacity. Miss one and the whole system gets rejected.

When to skip storage

The average Chinese customer lost 2.25 hours of power last year. Cities saw 0.55 hours and rural areas 2.53, 4.6 times worse^[14].

A 500 kW plant on the city figure loses 0.55 hours, or 275 kWh. Even at a generous 20 yuan/kWh, that is 5,500 yuan a year.

A 2 MW load on the national 2.25-hour average loses 4,500 kWh, worth 90,000 yuan at 20 yuan/kWh. The battery has to earn 220,000 yuan a year just to amortize 2.2 million over ten years.

Outage avoidance alone would need 11 hours of downtime at a 1 MW load to pay for that, and the national average is 2.25.

A 1 MW roof making 1.1 million kWh covers only 22% to 37% of a plant burning 3 to 5 million kWh a year. The battery's 594 MWh covers 12% to 20%. Sizing an array starts from the load profile, not the roof area.

At 30% self-consumption the numbers sag. 330 MWh used on site earns 231,000 yuan, and the other 770 MWh exports at 0.25 yuan for 192,500. That totals 423,500 yuan, 148,500 short of the 60% case.

Modules take 4,690 sq m of net roof, about 9,000 sq m once you leave the row clearance that keeps modules out of each other's shade. The 2 MWh cabinet takes 30 sq m and stores 67 kWh per sq m, 0.3% of the module footprint. Fire separation and structural loading are the constraints.

Scale shows up hard in the price. US solar plants built in 2024 between 1 and 5 MW averaged $8,062/kW. Projects above 100 MW averaged $1,496/kW^[15], a 5.4 times gap in unit cost.

What to check before signing

1. How much hosting capacity is left on your transformer, and whether your site is green, yellow or red.
2. Your peak, flat and valley load profile over the last 12 months. That tells you whether two cycles fit.
3. How your capacity charge is billed, on transformer size or on peak demand, and how long a switch takes.
4. The three hookup rules, checked one by one: no reverse flow, discharge under plant draw, discharge under contract capacity.
5. GB 44240-2024 certification, plus which of the two expiry dates lands first, the 6,000-cycle limit or the 10-year warranty. Both are printed on the product specification sheet.
6. Whether the existing string inverters can go hybrid, or a separate PCS has to go in alongside them.

Common questions

Does storage raise my export price?

No. Document 136 sends all exported PV to market, and Shandong's January 2026 reference price of 0.1448 yuan/kWh holds whether or not you own a battery. Storage moves self-consumption, from 60% to 95.2%.

How long should the battery run?

Two hours is the commercial-roof standard, meaning 2 kWh of battery for every 1 kW of solar. Four-hour systems cleared at 541.3 yuan/kWh in the first half of 2026, 60.8 yuan below the two-hour price, and that saving costs double the floor space.

What about three-shift plants?

Peak demand is the only lever left. Run the same load around the clock and there's no spread to arbitrage and little surplus to shift. The 200 kW peak cut is worth 101,000 yuan a year.

How long can it ride through an outage?

At 500 kW discharge a 2 MWh battery gives you 4 hours flat, or 3.6 hours at 90% depth of discharge. At 1 MW discharge you're down to 1.8 hours.

Pick your row

• Spread under 0.43 yuan/kWh, or no room for two cycles? Go grid-tied. Zero extra capex, and no battery warranty to track.
• Spread between 0.43 and 0.6, one cycle a day? Add 1 MW / 2 MWh and run the EMS on a single-cycle schedule.
• At a spread above 0.6 with two cycles a day, fill both slots: charge in the midday flat and discharge at the sharp peak.
• A site in a yellow or red zone needs storage before it gets a connection plan. Meet the utility's ratio first, then run the 0.43 or 0.23 yuan threshold.

Peak kWh times the peak spread, plus valley kWh times the valley spread, is the annual total. Run it against your last 12 months of bills. If the total covers the 220,000 yuan a year it takes to amortize 2.2 million over ten, the numbers work.