Satellite and space-power catalogue#
The catalogue follows the Energy Data Insights pattern: reviewable JSONL records, stable identities, field-level evidence and reproducible typed Parquet snapshots. It lives entirely in OrbitalPV. The initial batch contains 27 spacecraft and 20 components from agency, mission-team and manufacturer sources. It is deliberately an initial research sample, not a global completeness claim.
Explore the viewer, executable examples and public API:
Install and query#
These additions are in the source tree until a release includes them. From a checkout, install the analytical extra:
python -m pip install ".[catalogue,examples]"
python -m orbitalpv.catalogue validate
python -m orbitalpv.catalogue build --output outputs/satellite-snapshot
Each build requires a new output directory. Canonical records are included in the wheel, so basic queries work offline after installation without a repository path:
import pandas as pd
from orbitalpv import catalogue
satellites = pd.DataFrame(catalogue.load_records("satellites"))
components = pd.DataFrame(catalogue.load_records("components"))
print(satellites[["name", "pv_power_w", "power_basis", "power_conditions"]])
print(catalogue.coverage())
# Flat typed columns; nested evidence remains serialized JSON in Parquet.
frame = pd.read_parquet("outputs/satellite-snapshot/satellites.parquet")
print(frame[["name", "pv_power_w", "reference_distance_au"]])
Inputs, methods and outputs#
Inputs: selected public agency mission pages and exact OEM datasheets, retained
in the source registry with access dates and revision information where available.
Sources were read online; sha256=null means raw document bytes were not archived,
not that a document checksum was verified.
Methods: explicit unit conversions, identity checks, source-to-field mapping, qualified rating validation, conflict-preserving merges and completeness counts. No capacity is estimated from satellite mass, photos, cell efficiency, orbital elements or a constellation-wide total. No internet call occurs during validation, Python queries, tests or documentation builds.
Outputs: satellites.parquet, components.parquet, browser JSON projections
read back from those Parquet files, sources.json, schema.json, attribution
notice and a manifest with hashes and field-completeness counts. Parquet floats,
64-bit integers and lists have explicit types, including all-null columns. Nested
*_json fields use canonical JSON strings in Parquet and objects in browser JSON.
The manifest pins the canonical content and every generated file, and records PyArrow’s version. Identical inputs and exporter version reproduce the same bytes. Browser checksums detect mismatched assets, not a maliciously replaced manifest.
Power semantics#
pv_power_w is a preferred, directly supported array-output assertion with its
qualifier, life basis and conditions. It can be approximate, a lower bound or an
upper bound; it is never automatically a measured maximum-power-point value.
power_ratings_json keeps alternate conditions, ranges and conflicting assertions.
When no preferred scalar is justified, it stays null even if alternate ratings exist.
BOL/EOL mean beginning/end of life only when the source says so.
Mission reference preserves the mission phase or operating condition. It does not assert an explicit BOL/EOL qualification.
reference_distance_auis the reference for a power statement, not the spacecraft’s present location. Temperature and irradiance remain nullable.reported_spacecraft_power_wholds ambiguous spacecraft/bus power separately. Payload consumption, EPS throughput and battery energy are never PV watts.cell_efficiency_pcttravels with its conditions and whether the value is a minimum average, production average or typical value. Bare cells and CICs with coverglass are distinct configurations. AM0 irradiance conventions differ.Batteries retain Wh and Ah separately. Array area is not inferred from panel envelope dimensions or silently replaced by active cell/coverglass area.
There is no global PV-wattage total. For example, Rosetta’s 850 W at 3.4 AU and 395 W at 5.25 AU describe the same arrays in different conditions. Psyche’s Earth and asteroid ratings must not be counted as separate installed systems. Design specifications do not establish present spacecraft health or live generation.
Connected catalogues#
The satellites table covers an individual spacecraft or independently documented
module, including Earth satellites, planetary orbiters and deep-space probes.
No parent mission is counted again when a module is catalogued. The component table
covers exact OEM models or explicitly documented configurations of solar cells,
CICs, panels, arrays, EPS electronics and batteries.
component_ids records only direct installation/configuration evidence. Current
examples are DART/ROSA, Europa Clipper/Airbus arrays and SPACE HAUC’s documented
EnduroSat/GomSpace hardware. Matching manufacturers is not proof of installation.
No exact AZUR cell model is assigned to Europa Clipper without evidence.
Identity and update contract#
The canonical directory is src/orbitalpv/catalogue/data. Packaging the single
source of truth here makes the same records available in wheels and source checkouts;
there is no second editable copy under catalogues/ or in the viewer.
Records use UUID5 over an immutable identity key. Preserve an existing identity when adding a NORAD/COSPAR identifier later. Exact OEM model case and punctuation matter. One JSONL line is one record, sorted by ID in a two-character identity shard. Creation, content-update and source-verification timestamps are separate UTC fields.
Register evidence in sources.json first; add exact URLs, publisher, revision and
reuse notes. Every populated factual field needs source IDs in
field_sources_json. Every cited source needs a concise locator and evidence
summary. Unknown data stays absent/null, never zero or a guessed nominal value.
merge_records(existing, incoming, now=...) is a pure conservative merge helper.
It fills unknown fields. Any conflicting existing assertion keeps the entire record
unchanged and emits the incoming assertion to a review list. It does not overwrite
nested ratings, old evidence or timestamps on identical observations. Validate the
complete proposed catalogue before writing canonical JSONL in a reviewed branch.
It is not a multi-writer database transaction.
The schema is versioned in schema.json. Add new nullable columns through a
reviewed minor schema revision; changes to meaning, units, removals or renames need
a major revision and migration. Preserve source-specific fields in namespaced
extensions_json until a common column is justified.
Orbital identifiers and future expansion#
stage_omm accepts an already downloaded CCSDS OMM/CelesTrak JSON list and stages
identity candidates with NORAD/COSPAR IDs, including six-digit NORAD numbers:
python -m orbitalpv.catalogue stage-omm \
--input downloaded-omm.json --source-id reviewed_celestrak_snapshot \
--now 2026-11-02T23:00:00Z --output candidates.jsonl
It neither downloads nor admits records automatically. Before admission, register the exact source, hash the downloaded bytes, respect provider terms, and check aliases against existing mission identities. OMM lacks PV capacity and does not prove current operating status. Orbital tracking stays in the existing orbit/connectors APIs.
The first batch does not provide a full satellite identity inventory, live positions, constellation-wide PV totals or comprehensive operator/manufacturer geography. Its open backlog prioritizes Starlink, OneWeb, Kuiper, navigation, GEO telecom, Chinese missions, additional JAXA/ISRO missions and qualified EOL/temperature data. OEM model and spacecraft linkage needs independent evidence; constellation-level ratings must never be copied indiscriminately to individual satellites.