Telemetry and Performance Analytics#
The core analytics principle is to compare measured performance with expected performance after deterministic orbital and thermal effects have been removed.
Measured power#
For voltage/current telemetry:
A physics-aware residual is then:
Loss accounting uses the opposite sign, so a positive value is a shortfall:
opv.power_loss_breakdown reconciles this gap into named, evidence-labelled
stages and reporting categories. opv.summarize_power_losses integrates
the ledger over common observed intervals and reports the uncovered portion
of each recorded time span. Gains remain negative losses and unsupported
causes remain unexplained. See PV loss budgets and operating diagnostics for the full workflow.
Case 0 baseline#
Before a full orbit/attitude model is available, OrbitalPV includes a telemetry-only baseline based on a robust upper envelope. This is useful for testing ingestion, event plumbing, and persistent-loss detection.
The baseline deliberately does not claim physical normalization.
Health progression#
The modelling hierarchy is:
telemetry-only upper-envelope health;
orbit/Sun normalized power;
temperature-corrected expected power;
radiation-aware expected power;
topology/EPS-aware residuals;
uncertainty-aware degradation and remaining-power forecasts.
The sequential PV model and loss ledger support supplied geometry, temperature, degradation and electrical/controller states. Parameter identification, causal diagnosis, calibrated radiation inference and uncertainty-aware forecasts are still separate work; a loss bucket is not evidence that its physical cause has been established.
Benchmarking#
Time-series evaluation should split by time and, where possible, spacecraft. Random row splits can leak orbital cycles and make anomaly performance look better than it is operationally.