Grid operators live and die by demand forecasts. A miscall of even 2–3 GW on a summer afternoon or a bitter cold snap can force emergency interconnector purchases at punishing prices, trigger rolling brownouts, or push ageing plant into unsafe territory. Conventional numerical weather models give broad temperature outlooks, but they miss the granular urban heat island effects and soil moisture states that drive localised demand surges — the very surges that knock distribution circuits off balance first.
Satellite thermal infrared sensors resolve land surface temperature (LST) at 30–100 m, exposing heat island cores, cool parks and industrial heat sources that NWP grids at 3–10 km resolution simply smear together. Combined with microwave sounding of atmospheric moisture profiles and vegetation moisture indices from multispectral sensors, operators gain a physically grounded picture of where cooling loads will spike or where heating demand will ramp faster than expected. These inputs feed short-range (6–72 h) machine-learning demand models that are demonstrably sharper than purely ground-based equivalents, especially in rapidly urbanising territories where historical consumption patterns are already obsolete.
A sovereign constellation closes the loop operationally. The grid operator schedules reserve capacity, activates demand-response contracts, gates interconnector nominations and queues fast-start peakers against a forecast they control end-to-end — no commercial API terms, no degraded service on the days it matters most. Countries with significant seasonal extremes, whether monsoon heat or Arctic cold, can tune the satellite revisit schedule and ML inference pipeline to their own climate signature rather than relying on a vendor whose priority algorithms are calibrated to temperate European or North American baselines.
Frequently asked
Why can't we just use commercial weather services like IBM Environmental Intelligence Suite or Copernicus Climate Change Service? Why do we need our own satellite?
Commercial services aggregate and resell data derived from sensors owned by the US, EU, and a handful of allied nations. The revisit schedule, spectral bands, and spatial resolution are optimised for global median use cases, not your grid's specific load zones. A sovereign constellation lets your transmission system operator commission targeted overpasses during critical forecast windows—morning ramp and evening peak—over exactly the substations that matter. You also retain the raw data permanently; commercial subscriptions can be terminated, repriced, or export-controlled.
What is land surface temperature (LST), and how does it improve demand forecasting versus air temperature from weather stations?
LST is the radiative temperature of the Earth's surface as measured by a thermal infrared sensor; it reflects the thermal mass of rooftops, pavements, and vegetation that directly drives building cooling and heating loads. Air temperature from a weather station at 2 m height can lag surface heating by 1–3 hours, whereas LST captured by satellite shows the thermal state of the city in near-real time. Feeding LST fields into a load forecasting model alongside NWP air temperature has been shown in peer-reviewed studies to reduce mean absolute percentage error by 8–14% at the substation level.
How many satellites does a useful constellation require for this application?
For a single mid-sized nation (roughly 500,000–2,000,000 km² of load area), a minimum of 8 microsatellites carrying thermal IR payloads achieves a 2-hour average revisit at mid-latitudes, sufficient for 24-hour ahead demand forecasting. Scaling to 16 satellites cuts revisit below 60 minutes and enables same-day intraday dispatch optimisation. Starting with a pathfinder pair and expanding incrementally is a proven approach used by several national space agencies in the 2020–2026 period.
What ground infrastructure does a grid operator need to actually consume this data?
At minimum, the operator needs an energy management system (EMS) with an API-accessible demand forecasting module, a data pipeline from the national ground station to that EMS, and a retrieval algorithm for converting raw satellite radiance to LST and then to degree-day inputs. IEC 61968 standards define the common information model for integrating external data feeds into EMS platforms. The satellite data itself arrives as raster files (GeoTIFF or HDF5); most modern forecasting platforms ingest these natively.
Is this application proven, or still experimental?
The core physics—thermal infrared remote sensing for surface temperature—has been operational since NOAA's AVHRR in the 1980s. Applying high-resolution nanosatellite LST specifically to short-term electricity demand forecasting is a more recent integration, with documented pilots by utilities in the EU, South Korea, and Australia from 2020 onward. The application is tagged 'live' on this platform because operational products exist; the question for any new adopter is integration maturity with their specific EMS vendor, not scientific validity.
How does this capability interact with renewable energy forecasting on our grid?
Heat and cold forecasting and renewable energy forecasting are two sides of the same dispatch equation: the former tells the operator what load will be, the latter tells them what solar and wind supply will be. Both inputs feed the same unit commitment and economic dispatch model. A satellite constellation optimised for thermal band imaging can often carry secondary payloads for cloud-cover and irradiance estimation at marginal additional cost, making the asset dual-purpose and improving the return-on-investment calculation significantly.
What happens to forecast accuracy during extreme heat events—precisely when reliability matters most?
Extreme heat events often arrive with clear skies, which is actually favourable for thermal IR sensing; cloud gaps are minimal and LST retrievals are most accurate. The challenge is that statistical load models trained on historical data underperform during record-breaking temperatures because consumer behaviour shifts nonlinearly—demand for cooling can exceed model bounds by 10–20% during multi-day heatwaves. Sovereign operators should supplement satellite LST with smart meter data under IEC 61968-9 to capture this nonlinear response in near-real time.
What does this cost compared to buying a commercial demand forecasting subscription?
A national subscription to a tier-1 commercial weather and demand forecasting service typically costs $2–8M per year depending on coverage and resolution, and provides no raw data ownership. A sovereign 8-satellite nanosatellite constellation with a 10-year life spans approximately $120–180M in total (build, launch, operations), equating to $12–18M per year—a 2x to 4x premium over the subscription. However, the sovereign asset also serves maritime monitoring, disaster response, agricultural forecasting, and emissions oversight simultaneously; the demand-forecasting use case is one of several that collectively justify the investment.