---
id: VOLT-HOME-WP-002
title: "Can daylight-saving time break an EV or heat-pump schedule?"
slug: can-daylight-saving-time-break-an-ev-or-heat-pump-schedule
description: "An interval-duration audit of complete local delivery days and the scheduling risk created by assuming every day has the same length."
published: 2026-08-30
cluster: "Market intervals, clocks, and data integrity"
status: measured
evidence_url: /research-data/home-papers/can-daylight-saving-time-break-an-ev-or-heat-pump-schedule.json
figure_url: /research-media/home-papers/can-daylight-saving-time-break-an-ev-or-heat-pump-schedule.webp
figure_alt: "Chart for Can daylight-saving time break an EV or heat-pump schedule?: share of complete local delivery days not equal to 24 hours, shown as 23 h, 24 h, 25 h."
source_ids:
  - google-scaled-content
  - entsoe-sdac
  - ec-sdac-15m
  - acer-sdac-products
  - volt-architecture
peer_reviewed: false
---

# Can daylight-saving time break an EV or heat-pump schedule?

## Abstract

A home-energy controller can fail even when every price record is present if it assumes that each local delivery day has the same duration. This paper audits complete local delivery days using weighted interval duration rather than a fixed interval count. The evidence defines a day as daylight-saving-time affected when its weighted duration differs from 24 hours. Across 3,340 zone-days, 948 are recorded as affected, and the primary result reports 28.38323353293413% of complete local delivery days as not equal to 24 hours. The figure counts 26 23-hour, 2,392 24-hour, and 922 25-hour zone-days.

The regenerated evidence reports no interval for this estimand: `bootstrap_95_interval` is null and `interval_method` is `not reported for this estimand`. The finding is therefore a descriptive clock-integrity result: legitimate 23-hour and 25-hour days can be mis-indexed by fixed-shape controllers. It does not measure device failures in occupied homes, retail bills, comfort, or savings. Analysis and drafting were model-assisted. This working paper is not peer reviewed and is not trading advice.

## Plain-language answer

Yes. A schedule can break at a daylight-saving transition if the software treats “tomorrow” as a fixed list sized for a normal day. The market curve and the household deadline are both tied to time, but local civil time does not always advance in the same way. A complete local delivery day can be 23 hours, 24 hours, or 25 hours. A controller that hard-codes the normal shape may skip a valid interval, repeat another, attach a price to the wrong physical delivery period, or stop before an EV or heat pump has received the intended energy.

The evidence does not say that 28.38323353293413% of all household days experience a civil-clock change. Its unit is zone-days in the detailed sample, and the measured flag is based on weighted duration. The aggregate reflects how the sampled complete local delivery days were represented in the source data and method. It should be read as a warning about data handling, not as a population failure rate.

The safe rule is simple: schedule against timezone-aware interval boundaries and actual duration, not array position. “Hour ending” labels, local date strings, and assumed row counts are display conveniences, not reliable physical identities. An EV departure at a local clock time and a heat-pump comfort deadline should be translated into explicit instants using the bidding zone's timezone and then matched to intervals by start and end timestamps.

## Research question

The research question is whether complete local delivery days in the detailed market sample can differ from a fixed 24-hour shape, and therefore whether a home controller that assumes that shape is exposed to indexing error. The primary metric is the share of complete local delivery days whose weighted interval duration is not 24 hours.

This is not a study of whether daylight saving is socially beneficial, whether one timezone policy is preferable, or whether a particular home platform contains a bug. It isolates the precondition for failure: legitimate variation in local-day duration. Once that variation exists, correctness depends on the controller's clock model. A robust implementation may handle all affected days without incident; a fixed-position implementation may not.

## Data and provenance

The evidence artifact reports daily prices from 2021-01-01 through 2026-08-29, detailed intervals from 2025-10-01 through 2026-08-29, and a long historical window from 2015-01-01 through 2026-08-29. The detailed interval analysis is limited to ten representative European bidding zones beginning on 2025-10-01. The result therefore describes the registered sample and must not be generalized to every zone, device, tariff, or software stack.

The source contract comprises `zones`, `day_ahead_prices`, `grid_revisions`, `auction_publications`, and `ingestion_runs`. These tables provide zone clock identity, price intervals, revision lineage, publication records, and ingestion state. The analysis used a SELECT-only transaction, identified in the evidence as read only, with a 180-second statement timeout and a publication cutoff of 2026-08-30T00:00:00Z.

The frozen snapshot SHA-256 is `7e97489fc8528c8cc8c38830e05b48d949ce1f67b98575dff26f5d7c321e4c67`. The analysis-code SHA-256 is `57c57de79cdab2b5b6d6c54c485cb5162598c5ba0b0bfe995da40d75e6c52ba9`; the protocol SHA-256 is `adb36bf6b447af9f96339249b8becaefc20422499cca1977242866347a97bd4b`; the paper-registry SHA-256 is `7bcb91d7476d0a69fe9fa75a5c7782f8117e0153f82f9112b7e1d307d3943717`; and the source-registry SHA-256 is `07949550ac443ff673fda5c0209b99f137544f3ffecf6775f109bb9d09663bd6`. The evidence manifest records evidence SHA-256 `99698587bf2e062690bbb0614d6317d3c3eedebc46a95690d55bd49d6a9ffbf9` and figure SHA-256 `5d7d7991b83eec01919dc18511cf8ec8112cc79f137206e489f16c22483056e1`. These identifiers make the public aggregate traceable to frozen research objects. They do not substitute for reviewing the interval logic.

## Method

The method is a duration audit within the registered family of paired interval reconstruction, clock-safe counterfactuals, and day-block bootstrap. Instead of declaring a day complete because it has a conventional number of rows, the analysis sums interval duration and evaluates the local delivery-day boundary. The sole explicit assumption is that a day is daylight-saving-time affected when weighted interval duration differs from 24 hours.

This distinction is important for quarter-hour markets. Counting rows can work on an ordinary day and still fail when the local clock jumps forward or repeats an hour. Duration and contiguous timestamp identity preserve the physical sequence. A clock-safe reconstruction starts from timezone-aware instants, associates them with the correct zone, and only then presents local labels. It does not generate timestamps by adding a fixed number of positions to local midnight.

At the 15-minute market time unit used in the detailed window from delivery date 2025-10-01, complete 23-, 24-, and 25-hour local days correspond to 92, 96, and 100 quarter-hour delivery intervals after any source-curve expansion. Those counts describe physical duration, not a reason to manufacture or delete a row. Before the MTU seam, an hourly curve has a different native shape and must remain explicitly hourly rather than being expanded as if it were native quarter-hour data.

Eligible complete zone-days form the denominator. Days whose weighted duration differs from the ordinary 24-hour case form the numerator. The evidence separately reports 948 affected zone-days and a total primary sample size of 3,340. The primary statistic is descriptive. No uncertainty interval is reported for this estimand, so the result is not presented with a confidence range.

The evidence states that within-family Holm control applies to inferential claims and that this result makes no unadjusted significance claim. That is appropriate because the paper establishes the frequency of a structural condition in the sample; it does not test a causal model of household failures.

## Results

The recorded result is 28.38323353293413% of 3,340 complete local delivery zone-days not equal to 24 hours. The secondary result records 948 affected zone-days. The figure labels are `23 h`, `24 h`, and `25 h`, with values 26, 2,392, and 922 zone-days respectively; the first and third values sum exactly to the affected count.

The evidence stores a null interval and labels the interval method `not reported for this estimand`. Accordingly, the percentage and counts are reported as descriptive aggregates only. No old fractional endpoints, converted range, or confidence interpretation applies to the regenerated artifact.

The substantive result remains narrower than a software defect claim. A non-24-hour day creates a hazardous input shape. It does not prove that any named EV charger, heat pump, tariff integration, or household controller failed. That requires implementation-specific tests.

## Robustness and placebo checks

Using weighted duration is the principal robustness check against naïve row counting. A day can have a valid shape that differs from the ordinary shape; duration-aware reconstruction asks whether the physical delivery coverage is coherent. Clock-safe interval identity also prevents repeated local labels from being mistaken for duplicate physical intervals.

The result is descriptive and is not elevated by an unadjusted significance threshold. Within-family Holm control remains the stated policy for any inferential claims across the family, but this paper reports no interval or adjusted test for the duration-share estimand.

The 24-hour category serves as the ordinary-day comparison. The 23-hour and 25-hour categories are not treated as corrupted records merely because their shape differs. That is an important negative control: a detector that marks every nonstandard count as missing would fail this clock test by construction.

The null interval is handled conservatively. This paper does not manufacture an interval from the zone-day counts or use an unsupported uncertainty calculation to strengthen the conclusion. Reproduction should match the descriptive value and exact figure counts.

## Limitations

The detailed sample covers ten representative European bidding zones from 2025-10-01. It is not a census of all European clock regimes or household products. Some zones share time rules, and the aggregate unit is a zone-day, not a unique civil date or household.

The definition of an affected day is methodological: weighted duration differs from 24 hours. That can identify the relevant shape but does not diagnose why a particular upstream record has that duration. Correct interpretation still depends on zone metadata and contiguous timestamps.

No device commands, temperatures, EV state of charge, household occupancy, comfort outcomes, or retail charges are observed in this paper. Consequently, it is not a household bill study and cannot estimate a financial or comfort loss caused by a bug. It tests the clock condition under which a bug is possible.

The evidence reports no interval for this estimand. The measured status rests on the primary aggregate and affected count, not on an uncertainty range. Software should display no confidence interval unless a future governed artifact supplies one.

## Practical implication

Controllers should model schedules as sequences of explicit `[start, end)` intervals with a zone-aware timezone, actual duration, and immutable physical instant. Local labels are needed for human deadlines, but they should be derived from those instants. A scheduler should never assume that tomorrow can be allocated by copying the ordinary day's array shape.

For an EV, the energy constraint should be integrated over the intervals that fall between arrival and departure instants. For a heat pump, thermal and comfort constraints should be evaluated across the actual timeline. Repeated local labels need an offset or instant to remain distinct; skipped labels must not cause a synthetic gap to be filled with an invented price.

Operationally, DST transition cases belong in acceptance tests. The controller should validate total duration, contiguous boundaries, energy conservation, and deadline mapping before issuing commands. If the curve is incomplete or ambiguous after those checks, the safe action is a declared fallback, not a guessed index shift.

## Reproducibility

The evidence URL in frontmatter points to the public JSON containing the assumptions, windows, source tables, primary result, secondary result, limitation, disclosure, and provenance hashes. A reproduction should first verify the snapshot, protocol, registry, and code identities. It should preserve the publication cutoff and use a read-only query path.

The reconstruction should group intervals by bidding-zone local delivery date only after retaining UTC or equivalent unambiguous instants. It should calculate weighted duration from explicit boundaries, verify contiguity, classify the ordinary and transition-day shapes, and aggregate at zone-day level. Bootstrap resampling should be performed by day blocks, not by independent quarter hours.

Reviewers should also reproduce the point estimate, the 948 affected count, and the figure vector `[26, 2392, 922]`. The null interval and `not reported for this estimand` method label must remain null rather than being backfilled from an earlier artifact or inferred from the chart.

## Disclosure

Analysis and drafting were model-assisted; sources, code, assumptions, and evidence hashes are disclosed. This paper is not peer reviewed, is not a household bill study, and is not trading advice. Volt has zero live traders and zero live trading capital; C0R is the only paper strategy. Production weather forecasting is a non-trading service. Nothing in this clock audit authorizes an order, capital, or a live-trading interpretation.

## References

- [Spam Policies for Google Web Search — Scaled content abuse](https://developers.google.com/search/docs/essentials/spam-policies) — Google Search Central.
- [Single Day-ahead Coupling (SDAC)](https://www.entsoe.eu/network_codes/cacm/implementation/sdac/) — ENTSO-E.
- [EU electricity trading in the day-ahead markets becomes more dynamic](https://energy.ec.europa.eu/news/eu-electricity-trading-day-ahead-markets-becomes-more-dynamic-2025-10-01_en) — European Commission.
- [ACER Decision 13-2024 on SDAC Products](https://www.nemo-committee.eu/assets/files/ACER%20Decision%2013-2024%20on%20SDAC%20Products-702e63479704f5a1b75b75aa71c45ec8.pdf) — Agency for the Cooperation of Energy Regulators.
- [Voltcast Architecture](https://github.com/ossedk/voltcast/blob/main/docs/voltcast/ARCHITECTURE.md) — Voltcast.