Use the sun: why South Africa needs batteries now
By 2029, on a sunny spring day, more than half of South Africa's utility solar could be switched off. The System Operator is already throttling (curtailing) solar, and there are lurking threats of a slowdown in grid access for solar. What is to be done?
Solid = measured (Eskom data) · Hatched = modelled · Full scale from 0 GW
Methods, sources and assumptions
- Data: Eskom Data Portal, Station Build-Up (hourly), 29 Oct – 3 Nov 2026; Eskom renewable statistics (installed PV 2 780 MW; all-time PV record 2 546 MW on 23 Oct 2026).
- Grouping: coal floor = lowest coal output that week (15.15 GW), drawn on top of nuclear and imports; coal flex = coal above that floor; "pumped storage & CSP" = pumped-storage generation plus concentrated solar; peakers = OCGT, gas, hydro and interruptible load. Demand = generation minus pumping load. Rooftop and other embedded solar is not visible in Eskom's data and is not shown.
- Today's curtailment (step 2) is modelled: the gap between actual utility PV and an unconstrained envelope of about 95% of installed capacity, 09h–14h (≈1.2 GWh). Not confirmed by NTCSA.
- Solar pipeline (steps 3–5): The Energy Desk (TED), energydesk.africa – about 2.3 GW in 2027, 2.4 GW in 2028 and 1.8 GW in 2029 of utility solar. Year buckets are TED estimates and assume everything awarded is financed.
- 2029 model: new PV follows 30 Oct's utility PV profile; demand, wind, CSP, hydro, OCGT, nuclear and imports unchanged; pumped storage absorbs up to 2.7 GW / 12 GWh in daylight and returns 75% in the evening; batteries 88% round trip, 4-hour duration, discharging 17h–06h (first replacing peakers, then coal above its floor). With batteries, night-time pumping of surplus coal is no longer needed and is dropped. Once batteries have flattened coal onto its floor for all 24 hours, further storage lowers the floor (fewer coal units running) as far as batteries can still cover the evening and night. Coal units offline ≈ floor reduction ÷ 0.73, an assumed minimum stable load inferred from 30 Oct (15.2 GW floor vs 19.6 GW peak) – to be checked against Eskom unit data.
- Slider markers: "procured to date" ≈ the three BESIPPPP rounds (513 MW/2 052 MWh, 615 MW/2 460 MWh, 616 MW/2 464 MWh, ≈1.7 GW/7 GWh), not all yet in commercial operation; excludes private and Eskom-owned batteries. "New 4.6 GW round" = the 4.6 GW of battery storage announced in October 2026 as a first allocation under IRP 2025, assumed at 4 hours (≈18.4 GWh). The slider assumes 4-hour batteries throughout (power = energy ÷ 4).
- Step 6 – coal still running: coal units online estimated as the day's coal peak ÷ 0.95 (30 Oct) versus flat coal output ÷ 0.90 with batteries – an indicative range, not unit-by-unit dispatch, and for one spring day only. Retirement depends on winter peaks. IRP 2025 assumes about 8 GW of coal shut down in 2029–30 (contingent on 6 GW of new gas) and a further 15 GW in 2034–42; Eskom has said it plans to shut about 7.4 GW net by 2030. The coal fleet is roughly 38 GW installed.
- Not included: rooftop solar growth (which would deepen the midday dip and strengthen the case), Koeberg Unit 2, demand growth, winter conditions. A winter-day check is planned.
- Data questions: 29 Oct shows unusually low PV under clear skies (cause unconfirmed); 31 Oct and 1 Nov were partly cloudy.
Analysis by Frank Spencer, sbg.earth · All data stories. Figures are indicative and subject to review.