Skip to content

fix(ev): wait up to 15 minutes while the charger's load balancing holds a start - #2050

Open
frahlg wants to merge 1 commit into
masterfrom
easee-equalizer-wait
Open

frahlg wants to merge 1 commit into
masterfrom
easee-equalizer-wait

Conversation

@frahlg

@frahlg frahlg commented Nov 18, 2026 •

Copy link
Copy Markdown
Member

Problem

Göran runs an Easee Cloud charger with an Easee Equalizer and a Pixii battery on a 25 A three-phase fuse. In #1948 some starts waited 10–14 minutes: FTW ordered a positive current, the charger read it back and acknowledged three resumes, but the pilot stayed at 0 A and the car drew nothing. Easee support, in his translation, says the Equalizer lowers the current at once and raises it again slowly, up to 15 minutes after household use falls, and that a three-phase start needs at least 6 A on each phase.

During that wait Core said:

  • "Not following" (no_power_response, warning) after two minutes, with Easee's reason code as the only hint;
  • on the Charge now line, "Charger offers 16 A but the car is not drawing", which blames the car;
  • in the log, "loadpoint resume acknowledged but no charge observed" after the third resume.

Change

When the driver reports that the charger's own load balancing lets the car take less than 6 A, the least a car charges on (current_limited_by: "load_balancer", current_limit_a < 6), Core waits up to 15 minutes before it calls the start failed:

  • Verdict. New reason load_balancer_wait: status waiting, severity info. It counts from the first zero reading under positive orders; a pause ends that run. After 15 minutes the verdict is no_power_response as today. Safety pauses, a charger that reads back a lower order, a pilot of 6 A or more, and a run without that clock that is older than 15 minutes keep their own verdicts.
  • Charge now. The hold stays accepted for up to 15 minutes instead of turning not_drawing after two. manual.load_balancer_a carries the allowed current.
  • Resume retries. They still send three resumes. The failed-start warning waits for the same 15 minutes, or comes at once when the driver stops reporting the hold. Both log lines carry load_balancer_hold.
  • Words. "Waiting for load balancing: Easee's load balancing allows 0 A, too little for the car to start. FTW asked for 16 A." Next step: "It raises the current slowly once the home uses less, which can take up to 15 minutes. FTW reports a problem if the car has not started by then." After the wait, the no_power_response next step says "Easee's load balancing still allows only 0 A. Compare its settings with FTW's main fuse setting." Charge now says the same in its own line.

The 15 minutes and the 6 A live in go/internal/loadpoint/load_balancer.go; a web test reads them so the words cannot drift.

Request, acceptance, device response and measured effect stay apart: the readback, the allowed current and the measured 0 W each show in the numbers.

Where the evidence comes from

Core claims the load balancing only on the driver's word. Today's bundled easee_cloud (1.4.1) reports no allocation, and srcfl/device-drivers#194 reports it only while the car charges. srcfl/device-drivers#236, stacked on #194, adds the start hold: an allocation on all three phases below 6 A, or else a reason in which Easee names the Equalizer (10 or 28). Reason 78 is not evidence, since FTW writes that setting too. Until a pin moves to a driver with that change, this PR changes nothing on a box.

Evidence

  • make verify passed; node --test 'web/**/*.test.mjs': 945 passed.
  • New TestLoadBalancingHoldsAStartForFifteenMinutes replays the 12 November trace (12 A ordered and read back, 0 A allocated, 0 W): waiting within the wait, no_power_response warning after 15 minutes; a start at 11 minutes follows; a pause starts a new wait; the hold ending without a start fails at once.
  • TestResumeWarningWaitsOutALoadBalancerHold, TestManualHoldWaitsOutALoadBalancerHold and TestEVObservationReadsALoadBalancerHold cover the retries, Charge now and the parse. The verdict and retry tests fail on master.
  • Browser review on a local rig: real Core, the Ferroamp simulator as site meter, and a test Lua driver that reads back the order, reports 0 A from load balancing and draws nothing. After Charge now at 16 A the EV sheet showed "Waiting for load balancing" and the Charge now line below it, at 35 s and at 3 minutes. At 15 minutes (18:07:26 → 18:22:39) the sheet turned to "Not following" with the step "Easee’s load balancing still allows only 0 A. Compare its settings with FTW’s main fuse setting.", and the Charge now line said the same. The log had three resumes in the first minute and the failed-start warning at 18:22:34 with load_balancer_hold=true. Screenshots are with the owner.
  • Not seen on hardware. Göran's collector on the Pi records allocation, pilot and draw around the next delayed start.

What the apps owe

  • ftw-webapp: srcfl/ftw-webapp#131 (paired) adds the words for load_balancer_wait, the step after the wait and the Charge now lines.
  • iOS and Android: both owe a title, sentence and next step for load_balancer_wait (today they show it as unknown), the load-balancing step on no_power_response, and Charge now words when manual.load_balancer_a is present. Today EV.swift and EV.kt say "The car is not taking power. Check the car's charge limit." for not_drawing. With this PR the state stays accepted ("The charger is starting…") during the wait, so the apps stop blaming the car without a change.

Refs #1948

🤖 Generated with Claude Code

…ds a start

Easee support says its Equalizer lowers the charging current at once and
raises it again up to 15 minutes after household use falls (#1948).
During that wait the charger reads back FTW's order, signals 0 A and the car
draws nothing. Core showed only Easee's reason code, called it "Not
following" after two minutes, Charge now said the car was not drawing, and
the resume retries logged a failed start.

When the driver reports that the charger's own load balancing lets the car
take less than the 6 A a car starts on (current_limited_by "load_balancer"),
Core now:

- reports waiting/load_balancer_wait for up to 15 minutes from the first
  zero reading under positive orders, then no_power_response as before;
- keeps Charge now in "accepted" for that long, with load_balancer_a;
- logs the failed start after the same wait.

Safety pauses, a lower readback, a pilot a car could start on and a car
that stopped asking keep their own verdicts. The words name the load
balancing during and after the wait.

Refs #1948

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Signed-off-by: Fredrik Ahlgren <[email protected]>
@chatgpt-codex-connector

chatgpt-codex-connector Bot commented Nov 18, 2026 •

Copy link
Copy Markdown
Codex Review Summary: ✅ Completed

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

Review Status Commit Review trigger
📝 Code Review ✅ Completed 2026-11-18T16:15:29.667748Z 9ecb5f0 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.

@chatgpt-codex-connector chatgpt-codex-connector Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 9ecb5f0c74

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".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

limitMatches := m.ChargerLimitKnown && m.CommandedA >= 0 && math.Abs(ch.LimitA-m.CommandedA) < 1
// The charger holds the order, but its own load balancing does not let
// the car start yet. That is not a car that declines, for a while.
balancerWait := ch.BalancerA != nil && *ch.BalancerA+0.5 < m.CommandedA && elapsed < LoadBalancerRaiseWait

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Time the manual wait from the start of zero delivery

If Charge now has already been charging for more than 15 minutes and the charger's load balancing subsequently reduces its allocation below 6 A, elapsed still counts from the original hold/order rather than the new period without delivery. Consequently, balancerWait is immediately false and the manual status becomes not_drawing, even though classifyControlFeedback grants the same interruption a fresh 15-minute load_balancer_wait. This also makes the Charge now text prematurely claim that the car has failed to start. Track the continuous zero-delivery interval for the manual grace period instead of using the age of the unchanged command.

Useful? React with 👍 / 👎.

@miravoss26 miravoss26 left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

What it does: when the driver reports the charger's own load balancing (`current_limited_by=load_balancer`) allows <6 A, Core now waits up to 15 min (`LoadBalancerRaiseWait`) before saying "not following" / logging a failed resume. During the wait the UI shows `waiting / load_balancer_wait` instead of blaming the car (#1948).

Read: the guard in `classifyControlFeedback` is tight (accepted command, fresh site meter, no fuse cooldown, pilot <6 A, offered ≥ requested, clock reset by a pause), the hold clock in `recoverOfferedCharge` resets when the hold ends, and the synctest scenarios cover the bounded wait, a start in time, a pause restarting the wait, and the hold ending early. No new deps, hosts, secrets or authz changes.

  • nit: the 15 min figure comes from Easee support via the issue thread. Worth a one-line note that other chargers reporting `load_balancer` will get the same wait.
  • I read the Go/web changes but did not run the tests; CI shows a skipped job, so check that the Go suite actually ran.

Safe to merge from my read.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants