Grid-scale battery storage facility construction and commissioning tracking
Utility-scale battery storage sites are compact and fast-built, making them easy to miss in coarse imagery. Very-high-resolution optical tasking from Planet SkySat and Airbus Pléiades Neo can track container installation, transformer pads, and grid connection works at the site level.
Sensors
- Planet SkySat: 0.5 m native resolution (50 cm pan, resampled to 50 cm colour product), tasked on demand with same-day or next-day revisit to a target site. Sufficient to distinguish individual container-format battery blocks (typically 6 m × 2.4 m footprint) and resolve transformer pad poured concrete from bare ground.
- Airbus Pléiades Neo: 0.3 m native panchromatic, 1.2 m multispectral, four-band plus SWIR variant. Stereo and tri-stereo tasking enables a surface model from which pad height and grading works can be inferred. Revisit to a given point is typically 1 to 2 days with the two-satellite constellation.
- Planet PlanetScope: 3 m resolution, daily global revisit. Insufficient to resolve individual battery containers, but useful for detecting site clearance, gravel spreading, and the appearance of large transformer units as change signals that cue SkySat tasking.
- Sentinel-2 MSI: 10 m resolution in visible and near-infrared bands, 5-day revisit at mid-latitudes. Useful for monitoring site-perimeter changes, access road construction, and vegetation clearance over the weeks before container delivery. Free and openly archived from 2015.
Why battery sites are hard to count from space
A 100 MW / 200 MWh battery energy storage system built from lithium iron phosphate containers occupies roughly 0.5 to 1.5 hectares, depending on the vendor's form factor and the associated power electronics. That is a fraction of the footprint of a comparable gas peaker plant or solar farm. At 10 m resolution, a freshly installed block of containers can register as little more than a slight brightness change on a gravel pad. The site disappears into the noise.
The problem compounds because construction timelines are short. Container-format systems can go from cleared ground to energised in under six months. A monthly or even fortnightly Sentinel-2 pass may catch only the beginning and end states, missing the installation sequence entirely. For analysts trying to build a real-time census of grid storage capacity additions in a target market, that gap is commercially significant.
What the imagery actually shows, and what it does not
At 0.5 m (SkySat) or 0.3 m (Pléiades Neo), individual ISO-standard container units become resolvable objects. A standard 20-foot battery container is approximately 6 m long and 2.4 m wide. At 0.5 m ground sample distance, that is a 12 × 5 pixel object with a distinctive aspect ratio. Rows of containers cast measurable shadows whose length is a function of solar elevation angle and container height, typically around 2.4 m. Shadow geometry can confirm that objects are raised above grade rather than flat markings.
Transformer pads are the other reliable signal. A grid-scale battery system requires one or more step-up transformers to connect to the medium- or high-voltage network. These are large, heavy objects sitting on poured concrete pads, often surrounded by oil-containment bunds. At sub-metre resolution, the transformer body, bushings, and bund walls are all distinguishable. The presence of a transformer pad with associated cabling trenches is a strong indicator that a site is approaching commissioning rather than still in civil works.
What imagery cannot tell you is battery chemistry or nameplate capacity. A site built from NMC containers and one built from LFP containers look identical from orbit. Capacity estimation requires external data: planning consents, grid connection applications, or operator disclosures. Imagery confirms presence, installation progress, and approximate container count. It does not replace the document record.
A detection workflow from coarse to fine
The practical approach runs in two stages. PlanetScope or Sentinel-2 time series cover a broad target region daily or every few days, with a change-detection algorithm flagging sites where surface reflectance or texture shifts in a pattern consistent with industrial ground preparation: vegetation removal, gravel spreading, the appearance of rectangular high-reflectance objects. These flags are cheap to generate at scale across an entire country or grid region.
Confirmed or candidate sites then receive tasked SkySat or Pléiades Neo collections. A single tasked collect costs a fraction of the analytic value of knowing, a quarter earlier than a competitor, that a 200 MWh project has reached energisation. The high-resolution image is processed for object detection: container count, transformer presence, cable-trench routing, and perimeter fencing completion. Each of these is a proxy for construction stage. The output is a site-level status record updated on each new collect.
Honest limits of the method
Cloud cover is the most persistent constraint. In tropical and maritime climates, a site may be obscured for days or weeks at a time. Optical methods have no answer to this beyond increasing tasking frequency and accepting that some collects will be unusable. SAR could in principle detect large metallic objects through cloud, but at currently available commercial resolutions (1 m for Capella Space, ICEYE), the container-level discrimination that makes optical analysis useful is harder to achieve reliably.
Sites inside existing industrial compounds or co-located with solar farms present an occlusion problem. Containers stored temporarily before installation are indistinguishable from installed containers in a single image. A time series resolves the ambiguity: containers that do not move over successive collects, and that are subsequently joined by transformer equipment, are almost certainly installed. A single snapshot is not enough.
Capacity inference from container count carries real uncertainty. Container sizes vary by manufacturer. Some sites use larger 40-foot containers; others use proprietary enclosures. Without ground truth from at least one site in a market, a per-container capacity assumption can introduce a 20 to 40 percent error in aggregate capacity estimates.
Building a market-level census
The value of site-level tracking compounds when aggregated. A country-level or regional census of battery storage construction, updated monthly, gives grid operators, commodity traders, and project developers a picture of storage capacity additions that is independent of self-reported data. In markets where grid connection queues are long and project timelines are opaque, imagery-derived construction status is often more current than any regulatory disclosure.
Satellize runs this kind of structured optical-change programme for sovereign and commercial clients, combining open-data baselines with commercial tasking on client licence. The Tonga crop-estimation programme demonstrated the same underlying logic at a different scale: systematic, repeatable observation of a defined asset class across a defined geography, updated on a schedule that matches the decision cycle of the buyer.
A practical census programme for a single national grid market typically requires monthly PlanetScope screening across the full territory, SkySat or Pléiades Neo tasking of 20 to 60 confirmed or candidate sites per quarter, and a structured site database delivered as a GIS layer or structured feed. The archive depth of Sentinel-2 (from 2015) and PlanetScope (from approximately 2016 for most markets) allows retrospective baselining of sites that are already operational.
From site count to grid intelligence
A container count is not a capacity figure, but it is a leading indicator. Sites that reach transformer installation are typically within eight to twelve weeks of energisation, based on publicly documented project timelines for utility-scale storage in the United States, United Kingdom, and Australia. That window is actionable for traders pricing frequency response or capacity market products, and for developers assessing competitive saturation in a target region.
The method is most powerful when combined with planning consent data and grid connection registers, which are public in most OECD markets. Imagery confirms which consented projects are actually being built, and at what pace. The gap between consented capacity and imagery-confirmed construction is itself a signal worth tracking.
Typical figures
| Best spatial resolution (tasked) | 0.3 m panchromatic (Pléiades Neo); 0.5 m pan (SkySat) |
| Screening resolution | 3 m (PlanetScope daily); 10 m (Sentinel-2, 5-day revisit) |
| Tasked revisit to a single site | Same-day or next-day for SkySat and Pléiades Neo under clear-sky conditions |
| Latency from collect to analysis | Typically 24 to 48 hours after image delivery from operator |
| Minimum resolvable object | Individual 20-foot battery container (6 m × 2.4 m) at 0.5 m GSD; transformer pad at 3 m GSD |
| Spectral bands used | Panchromatic and visible RGB for object detection; SWIR (Pléiades Neo) for material discrimination |
| Archive depth for screening | Sentinel-2 from 2015; PlanetScope from approximately 2016 for most markets |
| Coverage per tasking pass | SkySat: up to 100 km² per collect; Pléiades Neo: up to 400 km² per strip |
| Capacity inference accuracy | Container count reliable to ±1 unit at 0.5 m GSD; nameplate capacity not derivable from imagery alone |
| Delivery formats | GeoTIFF imagery, GeoJSON or Shapefile site database, structured CSV status feed, PDF site report |
Analytics Satellize can run
| Site discovery and candidate list | Spectral change detection and texture analysis on PlanetScope and Sentinel-2 time series; rule-based filtering for industrial ground preparation signatures | Monthly GeoJSON layer of new and candidate battery storage sites across target geography |
| Construction stage classification | Multi-date object detection on SkySat or Pléiades Neo collects; stage taxonomy: site clearance, civil works, container installation, transformer installation, perimeter complete | Per-site status record updated on each new high-resolution collect, delivered as structured feed or GIS attribute table |
| Container count and array geometry | Object detection and manual QA on sub-metre imagery; shadow-length confirmation of container height | Per-site container count with confidence interval, array layout diagram, and change log between collects |
| Transformer pad detection and commissioning proximity estimate | Object recognition for transformer bodies and oil-containment bunds at sub-metre resolution; cross-referenced against cabling trench presence | Binary transformer-present flag per site, with estimated weeks-to-energisation range based on publicly documented project timeline distributions |
| Regional capacity-addition census | Aggregation of site-level container counts with externally sourced planning consent data; gap analysis between consented and imagery-confirmed construction | Quarterly market report with capacity-addition timeline chart, site-level appendix, and data export |
| Retrospective baseline for operational sites | Archive Sentinel-2 and PlanetScope time series analysis to reconstruct first-installation date and construction duration for existing sites | Historical site database with estimated commissioning dates, delivered as GIS layer with attribute table |
Who does the work
We can get this done for you. Satellize runs its own analyst desk and a strong science team. You do not buy a data feed and work out what it means; our people source the imagery, run the analysis described on this page, and hand you the answer with its confidence limits stated. Discuss this requirement.