Package Overview#

OrbitalPV is a workspace rather than one large Python module. Each package has one primary responsibility.

Scientific packages#

space-pv-core

Stable domain contracts and physical constants.

space-pv-orbit

Orbit propagation adapters, attitude transforms, Sun geometry, incidence angle, and eclipse calculations.

space-pv-pvmodel

Space-specific PV modelling adapters. It reuses validated pvlib equations where appropriate without assuming a terrestrial PV system model is a spacecraft model.

space-pv-space-env

AM0 irradiance and the evolving space-environment/radiation layer.

space-pv-connectors

External data adapters for orbital elements, mission ephemerides, space weather, and open spacecraft telemetry.

space-pv-analytics

Signed expected-versus-actual loss breakdowns, evidence-labelled categories, gap-aware energy accounting, performance normalization, degradation, persistent-loss, residual, and health analytics.

Platform packages#

space-pv-storage

DuckDB analytics and PostgreSQL/PostGIS application persistence.

space-pv-orchestration

Celery task orchestration. Scientific packages do not depend on Celery.

Applications#

space-pv-cli

Command-line workflows and reproducible experiment entry points.

space-pv-api

FastAPI boundary for jobs, data products, and map-serving endpoints.

space-pv-web

React/Cesium frontend. This application is intentionally outside the Python workspace.

Design principles#

  1. Physics before machine learning where a physical model is available.

  2. Coordinate frame, time scale, unit, and source provenance are first-class.

  3. Raw source data and derived scientific products remain separate.

  4. Long-running orchestration stays outside scientific packages.

  5. Flight-data analytics should compare measured power with expected physical power rather than relying on raw power thresholds alone.