---
id: VOLT-HOME-WP-029
title: "What is a seven-day price forecast worth for EV energy planning?"
slug: what-is-a-seven-day-price-forecast-worth-for-ev-energy-planning
description: "A sensitivity translation from lineage-matched forecast MAE improvement to an 18 kWh EV planning-error wholesale scenario scale."
published: 2026-08-30
cluster: "Smart EV charging"
status: measured
evidence_url: /research-data/home-papers/what-is-a-seven-day-price-forecast-worth-for-ev-energy-planning.json
figure_url: /research-media/home-papers/what-is-a-seven-day-price-forecast-worth-for-ev-energy-planning.webp
figure_alt: "Chart for What is a seven-day price forecast worth for EV energy planning?: MAE improvement translated to an 18 kWh planning-error scale, shown as model MAE, baseline MAE."
source_ids:
  - nature-v1g-v2g-2026
  - applied-energy-smart-charging
  - acer-retail-2025
  - iea-demand-flexibility
  - ec-sdac-15m
peer_reviewed: false
---

## Abstract

Forecast accuracy has no automatic monetary interpretation for EV charging. This paper translates the difference between mean baseline MAE and mean model MAE onto an 18 kWh planning-error scale. The frozen scenario assumes wholesale energy only, perfect charger efficiency, and local availability from 17:00 to 07:00. Across 17,750 lineage-matched forecast-accuracy observations, the translated wholesale scenario value is €0.01370521267605659 per event-equivalent. The evidence separately reports mean lineage-matched MAE of 87.14137855774648 EUR/MWh. The translation is explicitly a sensitivity bound, not guaranteed charging value. The regenerated evidence reports no interval for this estimand. The frozen calculation also does not isolate exact seven-day horizons, so the title’s seven-day interpretation is not identified by this aggregate alone.

## Plain-language answer

The evidence produces a wholesale scenario value of €0.01370521267605659 per 18 kWh event-equivalent when average forecast MAE improvement over its recorded baseline is mechanically translated to an energy scale. It does not show that an EV optimizer would realize that amount.

Forecast MAE measures average absolute price error. Charging value depends on whether errors change the ranking of intervals that fit the vehicle’s constraints. An accuracy improvement can occur in intervals the schedule never uses, while a small error around two nearly tied intervals can change a plan. The translation is therefore a screening scale, not a dispatch backtest.

The word “seven-day” requires exact-horizon evidence. The frozen branch pools lineage-matched forecast-accuracy rows with available model and baseline MAE; it does not filter them to a seven-day horizon. The honest answer is that this evidence gives a pooled sensitivity value, not a separately identified seven-day EV value.

## Research question

The intended question is: what wholesale scenario value can be associated with better price-forecast MAE when planning an 18 kWh EV energy requirement over a forecast horizon extending to seven days?

The implemented estimand is narrower and more mechanical. It takes mean baseline MAE minus mean model MAE across qualifying lineage-matched rows, multiplies the improvement by 18 kWh, and converts from EUR/MWh to EUR per event-equivalent. It does not run an EV schedule, compare chosen intervals, or retain an exact horizon in the outcome.

This paper therefore reports the implemented metric while refusing to overstate the horizon or decision interpretation. A direct answer would require exact seven-day lineage, a declared multi-day availability and energy problem, and paired forecast-versus-realized dispatch.

## Data and provenance

The registered source contract includes `day_ahead_prices`, `forecasts`, `forecast_accuracy`, `generation_mix`, and `zones`. The primary calculation uses forecast-accuracy rows with model MAE, baseline MAE, and a lineage-matched flag. The evidence does not expose customer or charger data.

Declared windows are daily prices from 2021-01-01 through 2026-08-29, detailed intervals from 2025-10-01 through 2026-08-29, and long history from 2015-01-01 through 2026-08-29. Publication cutoff is 2026-08-30T00:00:00Z. Extraction was read-only with a 180-second timeout.

Snapshot SHA-256 is `7e97489fc8528c8cc8c38830e05b48d949ce1f67b98575dff26f5d7c321e4c67`. Protocol, registry, source-registry, and code SHA-256 values are `adb36bf6b447af9f96339249b8becaefc20422499cca1977242866347a97bd4b`, `7bcb91d7476d0a69fe9fa75a5c7782f8117e0153f82f9112b7e1d307d3943717`, `07949550ac443ff673fda5c0209b99f137544f3ffecf6775f109bb9d09663bd6`, and `57c57de79cdab2b5b6d6c54c485cb5162598c5ba0b0bfe995da40d75e6c52ba9`.

## Method

The analysis selects forecast-accuracy rows for which model MAE is present and lineage matching is true. It separately selects lineage-matched rows for which both model MAE and baseline MAE are present. The mean model MAE is subtracted from the mean baseline MAE.

That MAE improvement is expressed in EUR/MWh. The frozen translation multiplies it by 18 kWh and applies the MWh conversion, producing EUR per event-equivalent. This is dimensional scaling: it assumes the full 18 kWh is exposed to the average improvement. It does not model which forecasted prices the charger selects.

The evidence’s secondary result is mean lineage-matched model MAE of 87.14137855774648 EUR/MWh. The primary transformed wholesale scenario value is €0.01370521267605659. The figure publishes model MAE = 87.14137855774648 and baseline MAE = 87.90277926197184 EUR/MWh.

The method family is constraint-aware charging simulation with paired schedule regret and scenario sensitivity. Within-family Holm control applies to inferential claims. This descriptive sensitivity result makes no unadjusted significance claim.

## Results

Across 17,750 observations, the translated wholesale scenario value is €0.01370521267605659 per 18 kWh event-equivalent. Mean lineage-matched model MAE is 87.14137855774648 EUR/MWh.

The result is positive because the mean baseline-model difference used by the frozen calculation is positive. It does not establish that each forecast row beats baseline, that every horizon improves, or that a charging schedule captures the translated difference.

The JSON reports `bootstrap_95_interval: null` and no interval method for this estimand. No exact seven-day filter appears in the frozen outcome. Both the missing estimator-valid uncertainty and the pooled horizon are material to the headline question.

## Robustness and placebo checks

Lineage matching is the main evidence control. It reduces the risk of combining accuracy values with unrelated forecast identities. Requiring baseline MAE for the comparison also avoids silently treating a missing baseline as zero.

Dimensional consistency is transparent: EUR/MWh is translated through the declared 18 kWh energy scale. A zero mean improvement would produce zero translated wholesale scenario value. That arithmetic property is a useful implementation check but not a dispatch placebo.

No exact-horizon stratification, schedule replay, alternative energy requirement, rank-based accuracy measure, or forecast-vintage placebo appears in the evidence. No uncertainty interval is reported for the transformed metric. No significance claim is made, and Holm control remains a family-level guardrail.

The recorded baseline is the strongest available comparison because the transformed value is based on baseline MAE minus model MAE rather than model MAE alone. Even so, averaging each series before subtraction can hide heterogeneous wins and losses. The JSON does not publish paired row-level differences, so the paper does not claim that improvement is broad across zones, models, or horizons.

A decision-specific placebo would perturb prices without changing their rank inside the feasible EV window. Such a perturbation can increase MAE while leaving the chosen schedule unchanged. The current evidence does not run that test, which is why the arithmetic translation is described as an upper-level sensitivity rather than schedule value.

## Limitations

The frozen metric is not a realized schedule comparison. MAE treats errors at all included points, while charging responds only to feasible candidate intervals and their ranking. The full 18 kWh scaling is a sensitivity assumption.

The calculation pools qualifying lineage-matched rows without demonstrating that they are exact seven-day forecasts. The paper therefore cannot isolate seven-day forecast worth. A title registered before outcomes does not override the implemented data selection.

Every monetary amount is a wholesale scenario value. Taxes, network charges, supplier margin, charger losses, battery losses, and tariff rules are excluded. Perfect efficiency and 17:00–07:00 availability are declared but not operationally used in the arithmetic translation.

The evidence publishes the baseline MAE only as a figure value; it does not publish horizon composition, zone composition, or model-version composition. It reports no interval for the primary metric.

The word “worth” can also imply a decision comparison that the evidence does not contain. No forecast-driven charging cost is compared with realized perfect-foresight cost, immediate charging, or a habitual schedule. The event-equivalent unit is dimensional, not behavioral.

Serial and cross-sectional dependence are not summarized in the public aggregate. The 17,750 observations may include related accuracy cells, so the count is not a count of independent EV planning experiments. No interval is reported to resolve that interpretive gap for the transformed result.

## Practical implication

Forecast providers and charging controllers should not convert MAE improvement directly into promised economic value. A defensible valuation should replay a fixed feasible scheduler against forecast vintages and realized prices, then compare its wholesale scenario cost with a declared baseline schedule.

Exact horizon identity is essential. A seven-day product should be evaluated on exact seven-day decisions rather than pooled accuracy. Lineage matching is necessary but does not replace horizon matching.

The €0.01370521267605659 result is useful as a reproducible order-of-magnitude sensitivity under an 18 kWh scale. It is not a tariff, purchasing, or charging recommendation.

For a future exact seven-day study, the charging problem must also be defined. An overnight event normally consumes only part of one day’s curve, whereas seven-day energy planning may allocate several events or choose among travel-dependent opportunities. Energy requirements, arrival windows, completion constraints, and forecast issuance clocks would need to be frozen before measuring value.

Reporting should keep forecast accuracy and dispatch economics side by side. MAE can diagnose forecast quality; paired realized wholesale scenario cost can diagnose decision quality. Neither metric should be silently substituted for the other.

## Reproducibility

Verify title, slug, measured status, assumptions, windows, primary and secondary results, sample size, figure fields, and hashes against the public JSON. Match all five citations to exact source-registry metadata.

Using snapshot `7e97489fc8528c8cc8c38830e05b48d949ce1f67b98575dff26f5d7c321e4c67` and code `57c57de79cdab2b5b6d6c54c485cb5162598c5ba0b0bfe995da40d75e6c52ba9`, select lineage-matched model MAE rows and the subset with baseline MAE. Subtract mean model MAE from mean baseline MAE, translate with 18 kWh, and preserve units.

The output should use 17,750 observations, report 87.14137855774648 EUR/MWh model MAE and 87.90277926197184 EUR/MWh baseline MAE in the figure, and return €0.01370521267605659 per event-equivalent. Report no interval for this estimand and do not claim an exact seven-day result. The current evidence and figure SHA-256 hashes are `b1c2b823305e9a7ed4e6d7ee62e6ceeb98c4b3b613f52422d5cccbb12a9301f2` and `7f5aeed0b912e4ada24d4e40ee35cbfd4a088c33af9f7cf7a2b2d2dc127d2667`.

## Disclosure

Analysis and drafting were model-assisted. This public working paper is not peer reviewed. Evidence, assumptions, hashes, code identity, and source metadata are disclosed.

It is not trading, investment, tariff, or purchasing advice. Monetary amounts are wholesale scenario values. No external finding was invented.

## References

- `nature-v1g-v2g-2026` — Nature Energy. [Coordinated planning of European charging infrastructure and energy system for optimal V1G and V2G deployment](https://doi.org/10.1038/s41560-026-02107-5). Kind: peer-reviewed.
- `applied-energy-smart-charging` — Applied Energy. [The value of smart charging at home and its impact on EV market shares](https://doi.org/10.1016/j.apenergy.2024.124997). Kind: peer-reviewed.
- `acer-retail-2025` — ACER and CEER. [Rewarding flexibility: How retail contract choice can help unlock consumer flexibility](https://www.ceer.eu/wp-content/uploads/2025/11/ACER-CEER-2025-Retail-monitoring.pdf). Kind: official.
- `iea-demand-flexibility` — International Energy Agency. [Scaling Up Demand Flexibility](https://www.iea.org/reports/scaling-up-demand-flexibility). Kind: official.
- `ec-sdac-15m` — European Commission. [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). Kind: official.