This project polls SolarEdge inverters over Modbus, combines the readings with SolarEdge cloud data, evaluates system health, and pushes notifications via Pushover/Healthchecks. The CLI also supports simulated runs for testing and a SQLite-backed state/retention layer. This can be useful to get quick notification if an inverter or optimizer has a fault.

• Modbus health checks: Reads configured inverters, evaluates per-inverter rules plus peer comparisons, and reports alerts.
• SolarEdge API integration: Pulls status/optimizer counts from the cloud when enabled, enriching alerts and summaries.
• Weather context (optional): Fetches Open-Meteo irradiance/temp/cloud cover to show per-inverter expected output alongside real readings and applies weather-aware alert suppression (irradiance floor + precip gate).
• Daylight-aware polling: A daylight policy decides whether to skip Modbus/cloud calls at night and when summaries should run.
• Simulation mode: Use [simulation] config sections or simulate --scenario NAME to test alert logic with synthetic data and timestamps.
• SQLite history: Per-run snapshots, optimizer counts, and site summaries are stored in ~/.solaredge_monitor_state.db (or the path you specify).
• Retention/maintenance: Configurable pruning and a maintain-db CLI command keep the database from growing indefinitely.

Edit solaredge_monitor.conf to define your environment:

• [daylight]: Timezone plus optional coordinates/sunrise/sunset windows.
• [modbus] & [inverter:NAME]: Global Modbus settings and per-inverter host/port/unit.
• [pushover], [healthchecks]: Enable flags and credentials for each notifier.
• [health]: Thresholds for peer comparison, low PAC/Vdc checks, etc. PAC/peer thresholds are configured as % of AC capacity (e.g., low_pac_threshold = 0.5 → 0.5% of capacity). Includes consecutive_health_alerts (require N consecutive bad checks before alerting; the Healthchecks failure ping follows the same gate) and consecutive_recovery_samples (require N healthy checks before sending recovery), reminder controls identical_alert_gate_minutes and repeat_alert_interval_minutes for persistent identical faults, min_alert_sun_el_deg (suppress low-PAC/Vdc/status alerts when sun elevation is below this angle), alert_irradiance_floor_wm2 (suppresses PAC/sleep/low-Vdc alerts when either GHI or POA is at or below this value — using the minimum of the two prevents stale hourly GHI from masking near-sunset low-light conditions), and weather gates (precip_weather_codes, precip_cloud_cover_pct) that suppress PAC alerts when it’s 100% cloudy with precipitation.
• [solaredge_api]: Enable flag, API key/site ID, and optional night skipping.
• [weather]: Optional Open-Meteo settings (enable flag, coordinates or fallback to [daylight], tilt/azimuth/albedo, array kW DC, AC capacity, derate, NOCT, temp coefficient) to show expected per-inverter output vs. weather and optionally append JSONL rows (log_path) for tuning.
• [logging]: Console log level/quiet/debug module overrides plus optional structured JSONL logging (structured_enabled + structured_path). Set log_path to write a self-rotating log file independent of stdout redirect; log_max_bytes (default 10 MB) and log_backup_count (default 5) control rotation. Nighttime skip messages and routine "no alerts" lines are logged at DEBUG and will not appear in the file at INFO level.
• [state]: Path to the SQLite database (~/.solaredge_monitor_state.db by default).
• [simulation] and [simulation:scenario]: Lists of inverters plus per-field overrides (PAC/Vdc/total_wh/optimizers). Include simulated_time to force a specific timestamp.
• [retention]: snapshot_days, summary_days, and vacuum_after_prune control how maintain-db prunes the DB.

See CHANGELOG.md for recent changes and defaults.

Run from the repo root:

• python -m solaredge_monitor.main --config solaredge_monitor.conf health: One-shot real run (default).
• python -m solaredge_monitor.main --config ... simulate --scenario fault: Run using [simulation:fault] data.
• python -m solaredge_monitor.main --config ... notify-test --mode fault: Send test alerts.
• python -m solaredge_monitor.main --config ... maintain-db: Prune historical data per [retention]. Override with --snapshot-days, --summary-days, --no-vacuum as needed.

Use --debug for verbose logs, --json to print Modbus snapshots as JSON, and --quiet to suppress stdout output.

If [logging] structured_enabled = true, each run appends a JSON object to structured_path with:

• timestamp, daylight_phase, daylight_context (is_daylight, grace flags, sunrise/sunset timestamps, skip_modbus/cloud, production_day_over)
• inverter_snapshots (per inverter: pac_w, vdc_v, idc_a, total_wh, status/vendor_status, serial/model, timestamp)
• weather_snapshot (timestamp, source_series_time used from feed, cloud_cover_pct, temp_c, wind_mps, ghi_wm2, dni_wm2, diffuse_wm2, weather_code, sun_azimuth_deg, sun_elevation_deg, provider, source_latitude/longitude, error)
• weather_expectations (per inverter: expected_ac_kw, expected_dc_kw, poa_wm2, cos_incidence, module_temp_c_est, temp_factor, array/ac capacity, derate, tilt/azimuth, albedo, NOCT, temp_coeff_per_c)
• residuals (per inverter: pac_w, expected_ac_w, residual_w, ratio)
• health (system_ok plus per-inverter inverter_ok/reason/reading)
• alerts (alerts generated this run)
• cloud_inventory (SolarEdge API inverter details, with raw payload)
• optimizer_counts (optimizer counts keyed by serial)