Zero-Carbon Industrial Parks: How Should Solar, Storage, and Energy Management Work Together

The energy platform decides whether this park pays. Solar sets the ceiling on green power, storage decides which hours that power lands in, and the platform decides how many cycles a day the battery actually runs.

A park drawing 590 GWh a year covers 5.3% of it with 30 MW on the roof. Reaching 33% needs a 155 MW direct-connect plant and 75 MWh of storage.

Storage costs 78.03 million yuan at 25 MW / 75 MWh, and a second daily cycle adds 7.39 million yuan a year without a single extra cell.

Two numbers decide whether that second cycle is worth running: the 0.65 yuan/kWh spread your load curve reaches at the sharp peak, and the 0.67 yuan/kWh levelized cost the battery carries on one cycle a day.

Clear both and the second cycle goes in. Miss them, and the roof and the direct-connect plant should be built out before any battery money is spent.

What each layer settles

The three layers share two numbers: the power ceiling where the park meets the grid, and how many cycles the battery runs a day. Transformer capacity and the declared interconnect capacity set the first. Your local time-of-use table sets the second, and this park lands on 70 MW and 2 cycles.

Write the relationship as one equation: park green power share = annual generation x self-consumption rate / annual demand. Storage and the platform never appear in the numerator. They change the self-consumption rate.

That rate is worth money. On the same direct-connect plant, lifting it from 68.0% to 81.2% moves 26.7 GWh from the export tariff to the self-use credit.

Sites that already run 56 PV-plus-storage and zero-carbon industrial parks publish the roof area, transformer headroom and declared capacity they settled on.

The compliance gate comes first

Three ministries issued the notice that defines the target. The test is how much carbon dioxide a park emits for every tonne of coal equivalent it consumes.^[1]

A zero-carbon park is one that, through planning, design, technology and management, brings the carbon dioxide from production and living activities inside its boundary down to a near-zero level, and is positioned to reach net zero.^[2]

The indicators split into one core number and five guiding ones. Parks consuming 200,000 to 1 million tonnes of coal equivalent a year must hold carbon intensity at or below 0.2 tonnes of CO2 per tonne of coal equivalent. Above 1 million tonnes, the ceiling is 0.3.^[3]

Chinese parks average about 2.1 tonnes of CO2 per tonne of coal equivalent. Reaching 0.2 means cutting fossil energy consumption by 90.5%. The guiding indicator asks for a 90% clean share, and that works out to 0.21 tonnes, still 0.01 above the line.

Electricity gets counted at its equivalent value, not its calorific value. One kWh enters the books as about 0.32 kg of coal equivalent, between 0.29 and 0.33 by province. That is 2.6 times the 0.1229 kg calorific figure, and this one convention sets the denominator for the core indicator.

The direct-connect route opened in May 2025. New energy skips the public grid, feeds a single user through a dedicated line, and keeps the origin of every kWh traceable.^[4]

A project's self-consumed renewable output must reach at least 60% of its available generation and at least 30% of its total electricity use, rising above 35% by 2030.^[4]

The export ceiling is set by each provincial energy authority and generally does not exceed 20% of available generation.

Four lines fix the floor and the ceiling: self-consumption at 60% of available generation and 30% of total demand, rising to 35% by 2030, with exports capped near 20% of available generation.

Three constraints get missed. Default supply through the grid company is off the table. Connection voltage stays at or below 220 kV, and 220 kV triggers a dedicated security review. During hours when renewables cannot be absorbed, the project stops pushing power back to the grid.^[4]

Parks with many tenants hit the definition. The notice specifies direct connection for a single power user, and direct supply across several users is left to separate rules from the national development and energy authorities.^[4]

So pick the filing entity before designing anything. Filing the park as one user and settling internally on sub-meters are two different routes, and the sub-metering has to cover generation, auxiliary load, self-consumption and storage.

Build PV large, not fragmented

By the end of July 2026, China's installed solar capacity reached 1,286 GW, passing coal at 1,285 GW and taking 31.5% of national generating capacity.^[5]

Solar produced 802.4 billion kWh in the first seven months of 2026, up 15.5% year on year, and covered 13% of national electricity demand. One in every eight kWh now comes from solar.^[5]

Curtailment follows scale. The IEA records rising curtailment in China, Germany, Brazil, Chile, the UK and Ireland, with negative-price hours clustering around the solar peak.^[6]

The same report puts the 2030 addition at 4,600 GW of renewable capacity worldwide, with solar near 80%, or about 3,680 GW. Green power supply grows on one line only.^[6]

US Energy Information Administration data for plants commissioned in 2024 shows 8,062 USD/kW at 1 to 5 MW against 1,496 USD/kW above 100 MW, a 5.4-fold gap.^[7]

The EIA figures exclude distributed generation, so the 1 to 5 MW bracket covers small ground-mount plants rather than rooftops. Read the four numbers as a pure scale effect.

Mounting comes second. Axis-based tracking costs 1,903 USD/kW and fixed tilt 2,529 USD/kW, a 32.9% penalty for fixed tilt that sits in steel and foundations.^[7]

The example park has 420,000 m² of usable roof, which at 14 m² per kW carries 30 MW. At 635 W per module, that is 47,244 units.

One module measures 2.382 by 1.134 m, or 2.702 m². That works out to 4.26 m² per kW, and the array covers 30.4% of the roof. The other 69.6% is walkway, service clearance and shading space.

Spacing is not cosmetic. Shade from the front row lands on the row behind and turns into a hot spot. Setting rows for three shade-free hours either side of solar noon puts row pitch at 1.5 to 2 times module height, and the row clearance that keeps modules out of each other's shade works through the whole calculation.

Rooftop output reaches 34.5 GWh a year, and 90% self-consumption keeps 31.05 GWh on site. That covers 5.3% of the park's 590 GWh.

The roof saves the retail tariff. A direct-connect plant saves wholesale, transmission, distribution and surcharges, and the 20% export ceiling applies to its available generation rather than to the roof.

The remaining 27.7 points come from off site, with 23.2 from the plant's direct consumption and 4.5 moved by the battery.

At 1,300 full-load hours and a 68% direct self-consumption rate, that calls for 155 MW, rounded. Annual output reaches 201.5 GWh across 258 to 413 hectares, in line with the self-consumption ratio that decides the development model for distributed and centralized builds alike.

The sizing sequence has a written method. Set the annual energy target from demand, size the array from full-load hours, then check the result against the self-consumption rate, as sizing an array against an annual demand target sets out step by step.

Size storage from the load curve

Storage power and storage energy come from two different inputs, and parks mix them up more than anything else. Power reads off the load curve, and energy reads off the clock.

A 1 MW / 1 MWh unit and a 1 MW / 4 MWh unit both get sold as "a 1 MW battery," at more than double the price, and how storage power and storage energy are sized separately is the split that explains why.

Shaving 900 kW for 15 minutes takes 225 kWh. Holding the same 900 kW for 3 hours takes 2,700 kWh. Identical power, 12 times the energy, and a quote sheet can call both of them a 900 kW battery.

The park peaks at 90 MW and the grid-side target is 70 MW, so storage has to deliver 20 MW. A 20% margin brings that to 24 MW, which lands on a standard 25 MW frame.

The margin is not padding. Inverters derate 5% to 10% at 40 °C and another 5% above 2,000 m, and reactive output plus station load consumes most of a 10% allowance.

Energy follows from a 2-hour overage. Twenty megawatts times 2 hours times 1.8 gives 72 MWh, rounded to 75 MWh. The 1.8 comes from three discounts multiplied together.

• AC round-trip efficiency at 85%, covering 98% to 99% at the inverter and 95% to 97% on the battery side;
• end-of-life capacity at 80%, because the system still has to hold the load on the last day of warranty;
• a usable state-of-charge window of 80%, leaving 20% reserved for outages.

Multiplied out that is 0.544, or 0.54 to two decimal places. A 75 MWh nameplate therefore delivers 40.5 MWh per cycle. Running 330 days a year on two cycles gives 26.7 GWh of discharge and burns 31.4 GWh on the charging side.

Two certificates matter at purchase. GB 44240-2024 is the first mandatory national standard for storage and runs from cell to pack. GB/T 36276-2023 tests cycles at 45 °C and stops at 80% capacity retention.^[8][9]

Warranties typically promise 10 years or 6,000 cycles, whichever comes first. Daily cycling over 330 days uses 3,300 cycles in a decade, while a twice-daily schedule pushes the count to 6,600 and runs the battery out around year nine.

Duration changes the price structure. Doubling it lifts cost per kW by 15% to 42% while cutting cost per kWh by 29% to 42%. That gap is why a 2-hour and a 4-hour quote for the same park sit so far apart.

What the energy platform runs

Policy places park energy and carbon management on a platform that monitors, forecasts and dispatches load. The work is a requirement rather than an option.^[1]

Internal resources have to be observable, measurable, adjustable and controllable, and the project reports to the dispatch authority under grid operation rules.^[4]

The project must adjust its internal generation and load so that exchange power with the public grid stays below the declared capacity, and it carries the consequences of any supply interruption it causes.^[4]

The platform's function list runs to five items, and each one maps to an acceptance clause.

1. Collect sub-metering from generation, auxiliary load, self-consumption and storage, with bidirectional meters at each gateway;
2. Hold exchange power with the public grid below the declared capacity at every moment, and keep its peak-to-valley ratio under the planned figure;
3. Forecast next-day solar output and load, then schedule battery charge and discharge at 15-minute resolution;
4. Block reverse flow during hours when renewables cannot be absorbed, curtailing generation where needed;
5. Report operating data to the dispatch authority and connect to the new-type power load management system or the dispatch automation system.

Sampling resolution sets what the algorithms can do. Fifteen-minute metering gives 35,040 points a year, enough to schedule charge and discharge, and it forms the basis for demand charges. Welding, crushing, injection moulding or large motor starts need 1-minute data at 525,600 points, because 15-minute averaging hides more than 20% of the spike.

Virtual power plants give the platform a second job. Aggregated distributed generation, adjustable load and storage can register as an independent market entity across medium-term, spot and ancillary markets. National flexibility targets run at 20 GW by 2027 and 50 GW by 2030.^[10]

This battery plus interruptible process load is already large enough for demand response and frequency regulation, and that revenue only arrives through the platform. When midday output exceeds what load and storage can absorb, the platform also has to hold the plant back instead of letting it push power back, losing generation but keeping the 20% line.

The arithmetic for one park

The park draws 590 GWh a year and consumes about 219,000 tonnes of coal equivalent. That sits in the 200,000 to 1 million bracket. Rooftop and direct-connect capacity together give 236 GWh of available generation.

Storage turns out to be a compliance device first. Without it, self-consumption reaches only 28.5% of demand and misses the 30% floor, while exports take 28.8% and break the 20% ceiling.

Priced at 1,040.4 yuan/kWh on a 2-hour basis, the storage costs 78.03 million yuan, or 3,121 yuan/kW. Its nameplate duration is 3 hours, so the 2-hour unit rate is the conservative read. O&M at 1.5% of capex runs 1.17 million yuan a year.

Three streams generate revenue. Shaving 20 MW off the peak at a 42 yuan/kW monthly demand charge saves 10.08 million yuan a year. Charging in the valley and discharging at peak gives 28,588 yuan a day. Charging in the midday flat period and discharging at the sharp peak gives 22,394 yuan a day.

Payback runs from 3.03 years to 9.4 years depending on whether the demand-charge stream is counted. That is not an accounting preference. It depends on the demand rate inside your two-part tariff and on the shape of your load curve.

The 0.67 yuan/kWh threshold comes from the same arithmetic. A one-cycle schedule delivers 133.65 GWh over ten years, and 78.03 million yuan of hardware plus 11.7 million yuan of O&M levels out at 0.67 yuan/kWh.

Running two cycles burns the 6,000-cycle warranty out in year nine and delivers 243 GWh, for 0.36 yuan/kWh. Cutting the levelized cost by 46% is where the whole 1.22 years comes from.

Run your own 12 months of bills before signing anything. Commercial tariffs swing 20% to 30% between provinces, so borrowing another site's numbers can flip the conclusion.

A second cycle earns 7.39 million yuan more a year and adds no hardware. It only takes the battery from one cycle a day to two, still inside the 6,000-cycle warranty.

One reconciliation has to be done by hand. Demand charges, valley arbitrage and surplus shifting share a single battery. Shaving 20 MW for 2 hours consumes 40 MWh, while the two daily cycles together offer only 81 MWh of discharge.

When the sharp peak lands inside the discharge window, one pass earns both the arbitrage and the demand reduction, as grid-tied and hybrid economies on one roof splits apart. A sharp peak in the flat period leaves the two streams competing for the same kWh.

EIA data puts solar cost flat beyond 1,496 USD/kW and battery storage at a 2024 weighted average of 1,469 USD/kW, so the two land in the same range.^[7]

The US Department of Energy targets a 90% cost cut within the decade for systems delivering 10 hours or more. That target does not cover the 2-hour systems parks actually buy, so do not count on it.^[11]

Battery fade belongs in the model. Year-10 capacity runs 20% to 30% below year one, and sizing energy on year-one capacity means peak shaving stops working by year eight and the demand charge comes back.

Three costs of the wrong order

Roofs first, storage second leaves no usable roof, so the battery crowds into the switch room or needs new land. A commercial cabinet lands at 1,176 yuan/kWh on an EPC basis against 1,040.4 yuan/kWh for a 2-hour container, a 13% premium for going indoors.

Platform first, hardware second means rework. Fifteen-minute and 1-minute collection are two different database structures with two different algorithms, and the equipment is inside warranty while the retrofit cost has nowhere to sit in the budget.

Supply agreement first, capacity second means rewriting the integrated plan. Direct connection sizes generation from load under the load-first principle, and changing capacity after filing redoes the system risk and power quality assessments.^[4]

Of the three, the last costs most. The first two cost money and the third costs time, because a wrong declared capacity restarts the negotiation over supply responsibility and charges with the grid.

Documents the acceptance review wants

Metering has to cover four gateways: generation, auxiliary load, self-consumption and storage. Each one needs a bidirectional meter that meets the standards and is recognized by the authorities.^[4]

Module data sets the green power ledger. A module built for industrial and commercial rooftops runs 615 to 635 W at 22.8% to 23.5% efficiency, with 132 cells in a 6 by 22 layout.

One unit measures 2,382 by 1,134 by 35 mm and weighs 29.0 kg, at a Pmax coefficient of −0.28% per °C and a 5,400 Pa snow load.

The degradation curve drives the 25-year production model. Year one loses 2%, then 0.55% a year, leaving 84.8% of initial output at year 25. That figure sits in the denominator of your self-consumption rate, and the module specification sheet carries the full curve.

Storage acceptance needs two standards and one warranty. Read the cycle count, the degradation curve and the temperature range line by line, because 10 years or 6,000 cycles decides when this system gets replaced.

Declared capacity and dispatch access close the list. Exchange power stays under the declared capacity and its peak-to-valley ratio under the planned figure, and both are operating constraints rather than design suggestions.^[4]

Existing projects fall into three categories with different steps. Parks running coal or gas self-supply plants must clear renewable development fund contributions in full, then cut plant output.

Export-oriented firms facing hard decarbonization requirements can run direct connection on nearby renewables. Projects held back by grid connection or absorption limits can switch after filing the change.^[4]

One agriculture-PV-storage project already operating runs at 300 MW and offers a load curve and storage ratio to check against, and industrial and commercial rooftops already built are listed in the project library.

For a rough first table, budget 14,000 m² of roof per MW and 1.7 to 2.7 hectares per MW on the ground. The line items budgets miss most often are site work, grid connection, taxes and contingency, and what a 1 MW build actually costs lists each of them.

Get the order right and the park's green power share comes down to available generation times the self-consumption rate, divided by annual demand. Everything else follows from those three numbers. Each extra point of self-consumption keeps 2.36 GWh a year inside the park, and that is what the 78.03 million yuan buys.

Park energy assessment

Three numbers decide this park: 236 GWh of available generation, a 20% ceiling on what may leave the site, and how many cycles a day your battery can actually run. Send us your 12 months of 15-minute meter data, the roof area and the transformer headroom you still have, and our engineers will run the same 25 MW / 75 MWh arithmetic against your own load curve before you commit to that 78.03 million yuan.

Request a park energy assessment