pti/blubgo
Watch
1
0
Fork
You've already forked blubgo
 0
Code Issues 7 Pull requests Projects Releases Packages Wiki Activity Actions

Learn the overrun in °C and pre-start before a schedule step #16

Merged
pti merged 18 commits from overrun-pre-start into main 2026-11-02 20:30:23 +01:00
Conversation 0 Commits 18 Files changed 37 +2712 -317
pti commented 2026-11-02 15:27:47 +01:00
Owner

Measured on blub, 2026-10-29 to 11-02 on the buros sensor, the overrun after switch-off is 0.63–0.82 °C whatever the rise rate, so the anticipator now learns the overrun as a temperature with EWMA 0.3 instead of a time, and before a schedule step up the thermostat pre-starts the warm-up so the room is halfway to the new target at the step: lead = learned dead time + half step / learned warm-up rate, latched until the step, controller untouched. Design: docs/superpowers/specs/2026-11-02-overrun-and-pre-start-design.md — plan: docs/superpowers/plans/2026-11-02-overrun-and-pre-start.md. Config: heating.anticipation.overrun / max_overrun replace overrun_time / max_overrun_time with no alias, and the new block heating.pre_start has enabled, warm_up_rate, dead_time, max_lead. Migrations: 007 converts coast_records and anticipation_samples to °C and drops the four *_time_seconds columns, values converted not lost — 008 adds warm_up_records and anticipation_samples.pre_start. API /api/history/anticipation gains warm_up_records and pre_start, the dashboard shows "pre-start for 09:00", heating history reads "Pre-start for 09:00" and freeze protection keeps its own reason. Deploy notes for blub: back up thermostat.db first, then update /etc/thermostat/config.yaml with overrun: 0.7, max_overrun: 2.0 and the pre_start block from configs/config.yaml.blub — old keys load silently as zero, and make deploy ships the example config with pre-start disabled. Verified: make fmt-check, make test, go vet, go test -race on application, domain and sqlite all green. 🤖 Generated with Claude Code, https://claude.com/claude-code

Measured on blub, 2026-10-29 to 11-02 on the buros sensor, the overrun after switch-off is 0.63–0.82 °C whatever the rise rate, so the anticipator now learns the overrun as a temperature with EWMA 0.3 instead of a time, and before a schedule step up the thermostat pre-starts the warm-up so the room is halfway to the new target at the step: lead = learned dead time + half step / learned warm-up rate, latched until the step, controller untouched. Design: docs/superpowers/specs/2026-11-02-overrun-and-pre-start-design.md — plan: docs/superpowers/plans/2026-11-02-overrun-and-pre-start.md. Config: heating.anticipation.overrun / max_overrun replace overrun_time / max_overrun_time with no alias, and the new block heating.pre_start has enabled, warm_up_rate, dead_time, max_lead. Migrations: 007 converts coast_records and anticipation_samples to °C and drops the four *_time_seconds columns, values converted not lost — 008 adds warm_up_records and anticipation_samples.pre_start. API /api/history/anticipation gains warm_up_records and pre_start, the dashboard shows "pre-start for 09:00", heating history reads "Pre-start for 09:00" and freeze protection keeps its own reason. Deploy notes for blub: back up thermostat.db first, then update /etc/thermostat/config.yaml with overrun: 0.7, max_overrun: 2.0 and the pre_start block from configs/config.yaml.blub — old keys load silently as zero, and make deploy ships the example config with pre-start disabled. Verified: make fmt-check, make test, go vet, go test -race on application, domain and sqlite all green. 🤖 Generated with Claude Code, https://claude.com/claude-code
pti added 17 commits 2026-11-02 15:27:47 +01:00 
Measured on blub 2026-10-29 to 2026-11-02: the overrun after switch-off is
about 0.7 °C whatever the rise rate, so the time-based model chases.
Pre-start begins the warm-up early so the room is halfway at the step.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The pre-start reason replaced only the band's switch-on reason, so a
freeze-protection switch-on is restored after a restart. A heating cycle
whose fit has no positive slope no longer teaches a 0.1 C/h warm-up rate.
Spec and CLAUDE.md amended.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
make deploy copied configs/config.yaml.example over /etc/thermostat/config.yaml,
which would have wiped blub's sensors, relay pin and MQTT settings. The config
is a manual step now. build-arm no longer fails where file(1) is missing.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
pti merged commit 26fa76ab6e into main 2026-11-02 20:30:23 +01:00
pti deleted branch overrun-pre-start 2026-11-02 20:30:32 +01:00
Sign in to join this conversation.
Reviewers
No reviewers
Labels
Clear labels
No items
No labels
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
pti/blubgo!16
No description provided.