How does the alarm system work and what types of alarms can I set up?

The alarm system monitors your meter readings and notifies you when consumption deviates from expected patterns. You can create alarm rules that automatically evaluate against incoming data and trigger alerts when conditions are met.

Six alarm types are available

1. Threshold (alarm_type = "threshold"): The simplest type. You set a fixed value and a comparison operator (>, <, >=, <=, =). When a meter reading crosses that threshold, the alarm fires. Example: "Alert me when electricity exceeds 500 kWh/h."

2. Schedule (alarm_type = "schedule"): Monitors usage during specific time windows. You define which days (mon–sun) and a time range (e.g., 22:00–06:00), plus an expected usage level (zero, low, medium, high). Ideal for detecting after-hours consumption — e.g., "No electricity usage expected on weekends."

3. Pattern (alarm_type = "pattern"): Uses a learned consumption model built from your historical data. The system calculates hourly and daily patterns, seasonal factors, and statistical bounds. When a reading deviates beyond the configured threshold (default 25%), it triggers. You must first train a pattern model via Analytics → Alarms → Pattern Learning.

4. Trend (alarm_type = "trend"): Compares current consumption against a historical baseline — either year-over-year (YoY) or a rolling average (4–52 weeks). You set a deviation percentage and direction (above, below, or both). Example: "Alert when consumption is 20% above the 12-week rolling average."

5. AI-Learned (alarm_type = "ai_learned"): Similar to pattern alarms but generated from AI analysis of your data. The system suggests alarm rules with confidence scores based on detected anomalies.

6. Data Timeout (alarm_type = "data_timeout"): Watches for a silent feed rather than abnormal values. It fires when a meter has delivered no readings at all, or when the newest reading is older than the configured timeout_hours (1–8760 hours). The cooldown period limits re-firing while the feed stays dead, and triggers do not auto-resolve — acknowledge or resolve them once data flows again. Example: "Alert me when this meter has sent no data for 24 hours."

Where to set up alarms: Analytics → Alarms. The page has four tabs — Dynamic Threshold (for threshold/pattern alarms), Trend-Based (for trend alarms), Schedule (for schedule alarms), and Data Timeout (for data_timeout alarms, at /dashboard/analytics/alarms/data-timeout).