Skip to content

Rate grid fixes for multi-inverter and EMS sites, battery temperature sensors averaged

Latest

Choose a tag to compare

@springfall2008 springfall2008 released this 16 Nov 09:46
2ba60c2

A patch release for the rate grids introduced in v9.8.0. It fixes how they are learned on sites with more than one inverter, how long the learned evidence is kept, and how a list of battery temperature sensors is read.

What's Changed

Bug fixes

  • fix(curves): learn site-wide battery_power against the whole fleet; save rate grid evidence with an expiry; average battery_temperature_history by @springfall2008 in #6227
    • Site-wide battery_power (#6226). On installs where one sensor reports the battery power for the whole site (battery_power: [sensor, 0]), for example a GivEnergy EMS in front of several inverters, the rate grids were learned against the first inverter alone. That made learned rates about N times too high, and they were clipped flat. The grids are now learned against the combined battery, the same way the plan models it: power and rated rates are totals, SoC is the capacity-weighted average, and every inverter is included. The site-wide form is now documented under Power Data.
    • Rate grid evidence kept (#6225). The learned evidence is saved with an expiry past its 365-day retention, and re-saved at least daily, so a storage backend's default lifetime can no longer drop it.
    • battery_temperature_history as a list. With one sensor per inverter, the plan used to add the temperatures together (two batteries at 15 °C were planned as 30 °C, so the cold-weather rate cap never applied). The rate grid scan failed every plan run with unhashable type: 'list'. Both now use the per-minute average of the sensors, leaving out a sensor that reads exactly 0 while another has a real reading.
  • fix(battery): read battery_temperature_history per inverter in the rate grid scan by @tico24 in #6222 (merged into #6227)
  • fix(web): Curves page text wraps by @springfall2008 in #6217

Tests and docs

  • test(plan): pin that a Freeze Export with no solar is pruned to Demand (#4124) by @springfall2008 in #6196
  • docs(debug-journal): fold the queue slice and correct the Solcast, Hanchu and plan-card entries by @springfall2008 in #6219

Full Changelog: v9.8.0...v9.8.1