{"paper_id":"2503.02518","title":"Extrapolating the long-term seasonal component of electricity prices for forecasting in the day-ahead market","authors":["Katarzyna Chęć","Bartosz Uniejewski","Rafał Weron"],"paper_date":"2025-03-04","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2503.02518v1","page_url":"https://thequant.space/flowcharts/extrapolating-the-long-term-seasonal-component-of-electricit/","scores":{"empirical_rigor":8.5,"math_complexity":7,"quadrant":"Holy Grail"},"proposes":{"status":"available","statement":"The paper proposes a novel approach that extracts the long‑term seasonal component by extrapolating the price series with forecasts for the next 24 hours and then combining it with ARX/LEAR models.","quote":"we introduce a novel approach which extracts the trend-seasonal pattern from a price series extrapolated using price forecasts for the next 24 hours.","summary":"Recent studies provide evidence that decomposing the electricity price into the long-term seasonal component (LTSC) and the remaining part, predicting both separately, and then combining their forecasts can bring significant accuracy gains in day-ahead electricity price forecasting. However, not…","strategy":"eSC-type models (eSCLEAR-MAS, eSCARX-MAS)"},"assumptions":{"status":"available","items":[{"assumption":"The loss differential series used in the Diebold‑Mariano test is assumed to be covariance stationary.","type":"statistical","stated":"explicit","quote":"Naturally, we assume that the loss differential series is covariance stationary."},{"assumption":"A 1.25 MWh battery cannot be discharged below 0.25 MWh and has charging/discharging efficiency of about 90% (≈80% per cycle).","type":"execution","stated":"explicit","quote":"we assume that we own a 1.25 MWh battery, which for technical reasons cannot be discharged below 0.25 MWh (or 20% of the nominal capacity) and its efficiency of charging as well as discharging is ca. 90% (which corresponds to a ca. 80% efficiency for a single charge-discharge cycle; Sikorski et al., 2019)."},{"assumption":"Missing hourly observations are replaced by the arithmetic mean of neighboring values; duplicate observations are replaced by their arithmetic mean.","type":"data","stated":"explicit","quote":"missing observations have been replaced by the arithmetic mean of two neighboring values, while duplicate ones have been replaced by their arithmetic mean"},{"assumption":"The price series is extrapolated for the next 24 hours using a naive forecast ˆPd,h = Pd−1,h only as a baseline, not in the main model.","type":"model","stated":"explicit","quote":"we also considered a simpler method to extrapolate the LTSC, where the price series is extended for the next 24 hours using a naive forecast, i.e., ˆPd,h = Pd−1,h, not an ARX or LEAR forecast as in steps 1-2 of the above algorithm."},{"assumption":"The datasets cover a 5‑year out‑of‑sample test period from 01‑01‑2019 to 31‑12‑2023.","type":"data","stated":"explicit","quote":"we use 5-year long test periods, i.e., from 1.01.2019 to 31.12.2023, for both datasets."},{"assumption":"Wavelet smoothing levels S_J correspond to time scales from 32 hours to 2048 hours.","type":"model","stated":"explicit","quote":"we consider five smoothing levels S_J with J ∈(5, 7, 9, 10, 11), corresponding to time scales ranging from 25 = 32 hours (or “1 day”) to 211 = 2048 hours (or “3 months”)."}]},"datasets":{"status":"available","items":[{"name":"German electricity market (EPEX)","provider":"ENTSO‑E transparency platform","kind":"market-prices","access":"public","period":"2019‑01‑01 to 2023‑12‑31","universe":"hourly day‑ahead electricity prices, load forecasts, RES generation forecasts","quote":"Both datasets are of hourly resolution and include day‑ahead electricity prices as well as forecasts of electricity consumption and generation from renewable energy sources (RES; here solar and wind) downloaded from the ENTSO‑E transparency platform (ENTSO‑E, 2024)."},{"name":"Spanish electricity market (OMIE)","provider":"ENTSO‑E transparency platform","kind":"market-prices","access":"public","period":"2019‑01‑01 to 2023‑12‑31","universe":"hourly day‑ahead electricity prices, load forecasts, RES generation forecasts","quote":"Both datasets are of hourly resolution and include day‑ahead electricity prices as well as forecasts of electricity consumption and generation from renewable energy sources (RES; here solar and wind) downloaded from the ENTSO‑E transparency platform (ENTSO‑E, 2024)."}],"known":[],"sample":{"start":"2019-01","end":"2023-12","market":"day-ahead electricity market (EPEX Germany, OMIE Spain)","asset_class":"commodities","frequency":"daily"},"data_resolution":"intraday-bars"},"transaction_costs":{"status":"not_extracted","modeled":"unclear","headline_net_of_costs":null,"detail":null,"quote":null,"source":"none","rule":"R0","candidates":{"front_matter":"unclear","evidence_card":"unclear","evidence_card_quoted":false},"meaning":"the classifier was not confident"},"results":{"status":"available","evaluation_type":"unclear","findings":[{"finding":"Extrapolated LTSC models improve RMSE by up to 15 % for the German market and up to 8 % for the Spanish market over a 5‑year period.","kind":"predictive-accuracy","quote":"improvements in predictive accuracy from using the proposed extrapolated SC‑type (i.e., eSC‑type) models compared to using a naive prediction of the LTSC can be as high as 15% for the German EPEX market (in terms of the RMSE; 14% in terms of the MAE)"},{"finding":"eSC‑type models significantly outperform their ARX and LEAR benchmarks according to the multivariate Diebold‑Mariano test.","kind":"statistical","quote":"the eSCARX‑based models significantly outperform the ARX, and similarly, the eSCLEAR‑based models outperform the LEAR benchmark."},{"finding":"Profit gains from the realistic day‑ahead bidding and battery storage strategy exceed 1 % for the eSC‑type models.","kind":"performance","quote":"... exceed 1% in terms of profits from a realistic trading strategy involving day‑ahead bidding and battery storage."},{"finding":"Naïve LTSC extrapolation (SCARX/SCLEAR) cannot cope with extreme price dynamics in Germany and is outperformed by classical ARX/LEAR benchmarks.","kind":"negative","quote":"the models that naively extrapolate the LTSC for the next day do not cope well with the extreme electricity price dynamics in Germany over the period studied and are outperformed by the classical ARX and LEAR benchmarks."}],"reported_performance":{"strategy":"eSC-type models (eSCLEAR-MAS, eSCARX-MAS)","sharpe":null,"sharpe_benchmark":null,"annual_return_pct":null,"total_return_pct":null,"max_drawdown_pct":null},"out_of_sample":{"method":"walk-forward","headline_out_of_sample":true}},"independent_evidence":{"status":"available","supports":[],"contradicts":[],"qualifies":[],"compared_with":33,"citations":{"cited_by_in_corpus":1,"by_independent_authors":1,"how_cited":{"used_as_baseline":0,"uses_method_or_data":1,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":0},"independent_papers":[{"paper":{"paper_id":"2604.19580","title":"Probabilistic Forecasting for Day-ahead Electricity Prices, Battery Trading Strategies and the Economic Evaluation of Predictive Accuracy","paper_date":"2026-04-21","authors":["Simon Hirsch","Florian Ziel"],"empirical_rigor":6,"page_url":"https://thequant.space/flowcharts/probabilistic-forecasting-for-day-ahead-electricity-prices-battery-trading/"},"how":"uses_method_or_data","quote":null}],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":1,"extracted_at":"2026-11-14T17:43:35Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}