{"paper_id":"2506.08113","title":"Benchmarking Pre-Trained Time Series Models for Electricity Price Forecasting","authors":["Timothée Hornek Amir Sartipi","Igor Tchappi","Gilbert Fridgen"],"paper_date":"2025-06-09","paper_type":"dataset-benchmark","paper_url":"https://arxiv.org/pdf/2506.08113v2","page_url":"https://thequant.space/flowcharts/benchmarking-pre-trained-time-series-models-for-electricity/","scores":{"empirical_rigor":7,"math_complexity":2,"quadrant":"Street Traders"},"proposes":{"status":"available","statement":"The paper benchmarks several state-of-the-art pretrained time series foundation models against established statistical and machine learning methods for electricity price forecasting.","quote":"we benchmark several state-of-the-art pretrained models--Chronos-Bolt, Chronos-T5, TimesFM, Moirai, Time-MoE, and TimeGPT--against established statistical and machine learning (ML) methods for EPF.","summary":"Accurate electricity price forecasting (EPF) is crucial for effective decision-making in power trading on the spot market. While recent advances in generative artificial intelligence (GenAI) and pre-trained large language models (LLMs) have inspired the development of numerous time series…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"Days with daylight saving time are transformed to a consistent 24‑hour format by interpolating or averaging hours.","type":"data","stated":"explicit","quote":"To simplify processing, we transform these days into 24 hour days: at the start of DST, we interpolate to insert an additional hour, while at the end of DST, we average to remove an hour."},{"assumption":"All models provide zero‑shot forecasting without fine‑tuning.","type":"model","stated":"explicit","quote":"All these models provide zero-shot forecasting (i.e., forecasting without fine-tuning or additional training) capabilities, enabling them to use their inference abilitie…"},{"assumption":"Statistical and ML models are trained on the previous 12 weeks of data with a one‑week input window.","type":"data","stated":"explicit","quote":"For statistical and ML models, we use 12 weeks of historical data for training, with an input size of 1 week."},{"assumption":"Diebold‑Mariano tests use a maximal p‑value of 0.1 as the significance threshold.","type":"statistical","stated":"explicit","quote":"We use a maximal p-value of 0.1 and reject the null hypothesis if the test statistic falls below the critical value."}]},"datasets":{"status":"available","items":[{"name":"DAA market data","provider":"ENTSO‑E Transparency Platform","kind":"market-prices","access":"public","period":"2024 and parts of 2023","universe":"Germany, France, Netherlands, Austria, Belgium (including Luxembourg)","quote":"We use publicly available DAA market data from multiple European countries, including Germany, France, the Netherlands, Austria, and Belgium, sourced from the ENTSO‑E Transparency Platform1 for 2024 and parts of 2023."}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"intraday-bars"},"transaction_costs":{"status":"not_applicable","modeled":"n/a","headline_net_of_costs":null,"detail":null,"quote":null,"source":"front_matter","rule":"R1","candidates":{"front_matter":"n/a","evidence_card":"unclear","evidence_card_quoted":false},"meaning":"the paper reports no trading or execution result"},"results":{"status":"available","evaluation_type":"forecast-accuracy","findings":[{"finding":"The biseasonal MSTL model consistently achieves the best performance across countries and metrics, and no TSFM statistically outperforms it.","kind":"performance","quote":"the biseasonal MSTL model, which captures daily and weekly seasonality, stands out for its consistent performance across countries and evaluation metrics, with no TSFM statistically outperforming it."},{"finding":"Chronos‑Bolt and Time‑MoE are the strongest TSFMs, performing on par with traditional models.","kind":"predictive-accuracy","quote":"Chronos‑Bolt and Time‑MoE emerge as the strongest among the TSFMs, performing on par with traditional models."},{"finding":"ElasticNet outperforms other ML models across all countries and metrics.","kind":"performance","quote":"Among the ML models, the ElasticNet model demonstrates superior performance across all countries and metrics."},{"finding":"MSTL statistically outperforms all Chronos‑T5 models in the Diebold‑Mariano test.","kind":"statistical","quote":"the MSTL model statistically outperforms all Chronos T5 models, but is not significantly more performant than the ElasticNet model."}],"reported_performance":null,"out_of_sample":{"method":"unclear","headline_out_of_sample":null}},"independent_evidence":{"status":"none_found","supports":[],"contradicts":[],"qualifies":[],"compared_with":48,"citations":{"cited_by_in_corpus":1,"by_independent_authors":1,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":0,"unclassified":1},"independent_papers":[],"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":"2026-11-14T17:36:03Z"},"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/"}