A 30-Day Observation-Level Telemetry Dataset with Minute-Level Tables for an Indoor Ice Rink Refrigeration System

This paper presents a 30-day indoor ice rink dataset comprising 1,305,260 observations from twenty signals, a full 43,200-minute calendar with acquisition gaps, and a 42,929-row reference preprocessed table produced by deterministic last-observation-carried-forward (LOCF) imputation and variable-specific plausibility screening. The release contains the three tabular stages, preprocessing code, signal mapping, measurement metadata, validation summaries, and per-value provenance. Applications include telemetry preprocessing, short-horizon forecasting with explicit observation-age controls, unsupervised anomaly screening, and refrigeration-monitoring prototype development.