Scatec's 569 MW Egypt Smelter Solar+BESS: What 100 MWh Actually Buys

Scatec has signed US$217 million in non-recourse loans with the EBRD, AfDB and EIB for phase one of its Dandara project in Egypt and has started construction. The first phase is 569 MW of solar PV paired with a 100 MWh battery energy storage system, against a capital cost of US$271 million, supplying state-owned Egypt Aluminium under a 25-year power purchase agreement. Commercial operation is targeted for the first half of 2028.

One number in that paragraph deserves more attention than it usually gets. 100 MWh against 569 MW is 0.18 hours of storage — under eleven minutes at full plant output. On a battery blog that reads like the storage got cut from the budget. For a smelter supply contract it reads like something else entirely: a design decision about which problem the battery is being paid to solve.

The Deal: What Phase One Actually Contains

According to Energy-Pedia's coverage of Scatec's 28 October announcement, phase one comprises the 569 MW solar plant and the 100 MWh BESS under the 25-year PPA with Egypt Aluminium. Scatec's own company release confirms construction has begun. Reuters reported that Egypt signed a preliminary US$650 million financing agreement for the full Dandara project the same week, and pv magazine notes that construction of phase two is expected to start in the first half of 2027.

The ownership structure is worth noting for anyone modelling merchant or corporate PPAs: Scatec holds 56%, with National Bank of Egypt joining at 20%. Scatec is EPC contractor and will provide asset management and O&M. The plant is expected to generate approximately 1.4 TWh a year and avoid more than 550,000 tonnes of CO2 annually.

The capital cost of US$271 million against a nameplate of 569 MW plus 100 MWh works out to roughly US$476/kW of solar capacity — but that aggregate is misleading, because it also carries the 25-year PPA's development costs, grid connection and the BESS. Split out properly, this is a competitive utility-scale solar build in a market where cell prices below US$45/kWh have made generation cheap and grid access expensive.

0.18 Hours: Reading the Ratio Correctly

Industry convention labels a system by its duration — energy divided by power. For a co-located plant, that convention quietly assumes the battery is charged by the PV plant at its own rate, so "2 hours" means "two hours of the solar array's output." Apply it here and Dandara looks like a rounding error.

But duration is only meaningful once you know what the battery is discharging into. Compare the two reference frames:

The last row is the honest one. A battery of this size cannot carry a smelter. What it can do is make the solar plant a better-behaved generator on the utility's side of the meter — smoothing the 1-minute and 10-minute ramps that a 569 MW PV plant produces as cumulus fields pass over the site, and providing the plant's own reactive support and frequency response where the grid code demands it.

That is a real service, and it is priced differently from energy arbitrage. A ramp-rate-controlled PV plant avoids the curtailment penalties and interconnection conditions that a raw inverter fleet attracts. In an islanding or weak-grid context, it is also what keeps the plant from tripping offline during a voltage excursion.

The Load Being Served

Aluminium smelting is the least forgiving industrial load in the power system. A modern potline runs at 400–600 kA DC, with cells connected in series at 4.0–4.3 V each; a single smelter commonly draws several hundred megawatts continuously, 24 hours a day, 365 days a year. Cutting power to a potline is not a load-shed event — it is a capital event. Cells cool, electrolyte freezes, and restarting can take weeks and nine figures of remediation.

That physics defines the contract. The PPA cannot be a "solar-shaped" supply agreement, because the load has no solar shape. It must be a firm supply arrangement in which the PV plant's output is a cost-reduction instrument against a grid-delivered base, plus a renewable attribute that satisfies the smelter's trade-exposed environmental requirements.

pv magazine reports that Scatec framed the project explicitly as helping Egypt Aluminium meet evolving trade-related environmental standards — the EU's carbon border adjustment mechanism being the obvious driver for an export-oriented smelter. That is the commercial logic for building PV in the first place. It is not a logic that requires a six-hour battery.

Sizing to a PPA, Not to a Dispatch Model

This is where a lot of storage analysis goes wrong. A tool optimising against spot prices will size storage where the arbitrage spread is widest. A project with a 25-year PPA against a single industrial offtaker is not arbitraging anything — it is managing three risks:

• Ramp compliance. The PPA and grid code set a maximum rate at which the plant may change its delivered output. The battery absorbs the difference between the PV plant's raw output derivative and the permitted ramp.
• Firmness shortfalls. Where the contract specifies a minimum hourly delivery, the battery bridges short gaps rather than the developer paying a shortfall penalty or buying expensive grid energy.
• Curtailment recovery. When the grid operator constrains the plant during high-solar, low-demand hours, stored energy lets the asset recover some of the constrained MWh later in the day instead of losing it.

All three are power-limited rather than energy-limited problems. They are solved by megawatts of inverter and a modest energy buffer — the exact shape of 100 MWh behind 569 MW. A sensitivity run on those three constraints typically converges on a duration in the 15–60 minute band, and the chosen 100 MWh sits inside it.

By contrast, the projects that get deep batteries are the ones selling into evening peaks or providing capacity adequacy. Egypt has been running that playbook separately: in January 2026 Scatec signed a PPA for a 1.95 GW solar plus 3.9 GWh battery project in the country, a system with roughly two hours of storage at full array output. If you want to compare storage business models, those two Scatec projects are a cleaner pair than any cross-country benchmark.

How to Model It in 8760 Hours

If you are sizing a smelter solar-plus-storage build, the simulation has to be run in a different order from a merchant project. The load is not a variable to be shaped — it is a hard constraint. My sequence in Energy Optima:

1. Fix the load first. Build the smelter's 8760-hour profile as a constant band with explicit potline blocks, and record the shed-ability of each block. This is the constraint set, not a forecast.
2. Size PV to the daytime grid import it displaces. The objective is maximum self-consumption MWh per dollar, not maximum generation. Curtailment above the smelter's instantaneous demand is close to worthless unless the grid pays for it.
3. Then size the battery to the ramp constraint. Set the PPA's ramp limit as a hard rule, run the dispatch, and read off the peak exceedance the battery must absorb. The energy buffer that follows is small — often 20–30 minutes.
4. Check the grid-import cap. If the interconnection agreement limits import, the constraint may bind in the evening and shift the optimum to longer duration. In Dandara's case the grid connection is the base supply, so this constraint is handled contractually.
5. Run degradation over 25 years, not one. A 100 MWh LFP system cycling shallowly and irregularly behaves very differently from one cycling twice a day. Use a degradation-aware capacity model rather than a flat annual fade, and check the year-25 SOH against the contract's firmness requirement — not against nameplate.
6. Stress the cloud field. Aggregate hourly data hides 10-minute ramps. Use sub-hourly irradiance where the grid code penalises ramps, because that is the only condition under which a 0.18-hour battery looks right rather than undersized.

That last point is the one that changes answers. A model run on hourly PVGIS TMY data will conclude that 100 MWh is pointless and a 2-hour system is obvious. A model run on 1-minute irradiance with a ramp-rate rule will conclude the opposite. Same project, same resource — different constraint, different answer.

Egypt's Wider Storage Pipeline

Dandara is one deal in a rapidly expanding Egyptian programme, and it is instructive that it is not the deep-storage one. Reuters and pv magazine both place the financing in a broader national build: Egypt aims to source 42% of its electricity from renewables by 2030, and the electricity ministry said in February 2026 that it planned to add 3 GW of solar and 600 MW of battery storage during the year. Scatec alone has already completed its 1.1 GW Obelisk solar-plus-storage plant in Egypt and signed the 1.95 GW / 3.9 GWh PPA in January.

The pattern that emerges is a portfolio split between two very different storage roles: deep storage on projects that sell firm capacity to the utility, and thin storage on captive industrial supply where the battery's job is power quality and contract compliance. Both are legitimate. Both need different models. Treating the second as a compromised version of the first is the analytical error to avoid.

What the Thin Battery Will Not Do

Three honest limits on the Dandara design, for anyone taking it as a template:

• No evening firmness. Post-sunset, the plant delivers essentially nothing. Any PPA term that requires firm evening delivery is satisfied by the grid, not by this asset.
• Minimal curtailment recovery. With 0.18 hours of buffering, a multi-hour midday constraint event still loses nearly all of the constrained energy. If curtailment becomes a material revenue risk — as it has in Spain, where curtailment reached 2.5 TWh in the first half of 2026 — the economics of a deeper battery change quickly.
• No black-start or islanding capability. A system this size cannot energise the smelter. Islanding a potline requires a much larger, grid-forming architecture with a dedicated microgrid controller — a different product at a different price.

None of those are deficiencies in a contract where they were never asked for. They are the boundaries inside which this design is correct, and they are precisely what a properly constrained 8760-hour simulation will surface before the steel arrives rather than after.

Sources