fix(mpc): give Energyplan 1.5 s when a car is plugged in

[frahlg]

Problem

Core gives the Energyplan planner 0.5 s unless the model is large, demand charges apply or PV curtailment is active. Since #1440 most homes with one car fall under 0.5 s. On the home Pi 4 that sits just below the point where the planner can improve on its first plan for the car.

Two plans with the car on the home box (v0.139.3-dev test build with Energyplan 0.5.0, 3 Oct, 112-113 slots) came back about 11% above the proven HiGHS optimum: 6 800.21 against 6 115.16 öre and 6 770.17 against 6 098.44 öre. Replaying the same requests on a laptop, the planner reaches the optimum at about 30-35 ms there, which is about 0.6-0.7 s on the Pi. Below that it returns its first plan.

Change

Requests with an active loadpoint get 1.5 s (energyplanEVBudget). Battery-only requests keep 0.5 s; large fleet, demand-charge and curtailment requests keep 5 s. The remaining time of the first slot still caps every budget, as before.

Evidence

• The 116 real requests with the car from the box's 0.5 s class (1-2 Oct), run with Energyplan 0.5.0 at the laptop equivalent of 0.5 s and of 1.5 s on the Pi: the summed gap to the proven optimum falls from 7 676 to 3 249 öre, plans within 0.5% of it rise from 111 to 114, 4 requests get cheaper and none get dearer. Short-charge deficits do not change.
• On the box with this branch, the first plan with the car used the 1.5 s budget (1 516 ms).

Known limitation

The same morning request showed a case where more time does not lower the bill: a car 253 Wh short of its target, less than one five-minute run at the lowest step. With more time the planner proves that a plan without a short run exists and then requires one, and the search found such a plan at 924 öre where the optimum is -116.89 öre. With less time it returned -110.8 öre with a 191 s short run. The charging-period preference ranks above cost by design, and v0.4.10 behaves the same. Whether a need smaller than one minimum run should count as met is an open product decision for Energyplan.

Verification

• TestEVGetsLongerBudgetThanBatteryAlone (new) and TestBatterylessEVBudgetDoesNotInventStorage (updated) fail without the change and pass with it.
• make native-solver-test and make verify pass.
• Installed and running on the home Pi 4 box as v0.139.3-dev.c33c2c4a (same change on top of chore(optimizer): bundle Energyplan 0.5.0 #1512).

Checklist

• The change follows VISION.md and one selected scope.
• I checked overlapping PRs and coordinated shared files/contracts. No open PR touches go/internal/mpc/energyplan.go.
• Relevant checks cover the changed behaviour and failure paths.
• A human reviewed changed UI in a browser, or no UI changed. No UI changed.
• A Changeset is included (patch), or the change is exempt.
• Every commit has a DCO sign-off.

🤖 Generated with Claude Code

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-04T03:19:10.871105Z	f594f1f	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.