Paper: arXiv 2511.04361
Authors: Dennis Thumm
Abstract
Energy markets exhibit complex causal relationships between weather patterns, generation technologies, and price formation, with regime changes occurring continuously rather than at discrete break points. Current approaches model electricity prices without explicit causal interpretation or counterfactual reasoning capabilities. We introduce Augmented Time Series Causal Models (ATSCM) for energy markets, extending counterfactual reasoning frameworks to multivariate temporal data with learned causal structure. Our approach models energy systems through interpretable factors (weather, generation mix, demand patterns), rich grid dynamics, and observable market variables. We integrate neural causal discovery to learn time-varying causal graphs without requiring ground truth DAGs. Applied to real-world electricity price data, ATSCM enables novel counterfactual queries such as “What would prices be under different renewable generation scenarios?”.
Complexity vs Empirical Score
- Math Complexity: 8.0/10
- Empirical Rigor: 4.0/10
- Quadrant: Lab Rats — high math complexity, limited empirical testing
Why this score: The paper employs advanced causal inference theory, neural architectures, and time-varying DAG learning with complex mathematical formalisms, but lacks detailed empirical validation, backtesting results, or specific datasets in the provided excerpt, focusing more on theoretical framework.
Research Flowchart
flowchart TD A["Research Goal:<br>Causal Regime Detection in Energy Markets"] --> B["Data: Multivariate Time Series<br>(Weather, Generation, Price, Demand)"] B --> C["Methodology:<br>Augmented Time Series Causal Models<br>+ Neural Causal Discovery"] C --> D["Process: Learn Time-Varying<br>Causal Graphs & Regime Changes"] D --> E["Process: Counterfactual Reasoning<br>(e.g., 'Prices under different<br>renewable generation?')"] E --> F["Outcome: Interpretable<br>Energy System Model"] F --> G["Outcomes:<br>Causal Regime Detection<br>+ Counterfactual Queries<br>(Energy Markets)"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes Augmented Time Series Causal Models (ATSCM), a framework that extends counterfactual reasoning to multivariate temporal energy market data with learned, time‑varying causal structure.
Source sentence
We introduce Augmented Time Series Causal Models (ATSCM) for energy markets, extending counterfactual reasoning frameworks to multivariate temporal data with learned causal structure. We bridge this gap by introducing Augmented Time Series Causal Models (ATSCM) for energy markets, enabling counterfactual reasoning about alternative weather patterns, generation scenarios, and policy interventions.
Assumes 3
- Electricity cannot be stored economically, creating immediate causal dependencies between supply, demand, and pricing. market
Source sentence
Unlike traditional financial markets, electricity cannot be stored economically, creating immediate causal dependencies between supply, demand, and pricing [11].
- Causal discovery can be performed without ground‑truth DAGs, learning time‑varying causal graphs from data. model
Source sentence
We integrate neural causal discovery to learn time-varying causal graphs without requiring ground truth DAGs. However, these frameworks assume static causal structures and cannot handle the temporal dependencies and regime changes inherent in energy systems.
- Energy market regimes change continuously rather than at discrete break points. other
Source sentence
Energy markets exhibit complex causal relationships between weather patterns, generation technologies, and price formation, with regime changes occurring continuously rather than at discrete break points. Causal Regime Detection in Energy Markets With Augmented Time Series Structural Causal Models Dennis Thumm National University of Singapore Singapore [email protected] Abstract Energy markets exhibit complex
Data
- real electricity market data market-prices unknown
Source sentence
Applied to real-world electricity price data, ATSCM enables novel counterfactual queries such as "What would prices be under different renewable generation scenarios?". Our key contributions include: • A theoretical framework extending ASCM to multivariate temporal data with learned causal structure • A neural architecture modeling interpretable energy factors, complex grid dynamics, and market observations •
Costs
Not applicable
Finds 3
- ATSCM enables novel counterfactual queries such as estimating prices under alternative renewable generation scenarios. other
Source sentence
Applied to real-world electricity price data, ATSCM enables novel counterfactual queries such as "What would prices be under different renewable generation scenarios?".
- The proposed approach achieves competitive forecasting performance while modeling complex energy systems through interpretable factors. performance
Source sentence
Our approach successfully models complex energy systems through interpretable factors while achieving competitive forecasting perfor-mance.
- ATSCM provides interpretable counterfactual capabilities on real electricity market data. theoretical
Source sentence
Our key contributions include: • A theoretical framework extending ASCM to multivariate temporal data with learned causal structure • A neural architecture modeling interpretable energy factors, complex grid dynamics, and market observations • Integration of causal discovery for time-varying DAG learning without ground truth require-ments • Empirical validation on real electricity market data with interpretable
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 12 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (134)
- Causal data science for financial stress testing (2018)
- Navigating Market Turbulence: Insights from Causal Network Contagion Value at Risk — → 2402.06032
- Causal Change Point Detection and Localization (2024) — link
- Counterfactual Image Editing (2024) — link
- Counterfactual Image Editing with Disentangled Causal Latent Space (2025) — link
- Neural Causal Models for Counterfactual Identification and Estimation (2023) — link
- Towards Causal Market Simulators — → 2511.04469
- CausalFormer: An Interpretable Transformer for Temporal Causal Discovery (2024)
and 126 more
- Towards Dynamic Structure Changes Detection in Financial Series via Causal Analysis (2021)
- From Identifiable Causal Representations to Controllable Counterfactual Generation: A Survey on Causal Generative Modeling (2024) — link
- Continuous inspection schemes (1954)
- Point estimation of the parameters of piecewise regression models (1976)
- Detecting “disorder” in multidimensional random processes (1981)
- A new approach to the economic analysis of nonstationary time series and the business cycle (1989)
- On likelihood ratio tests for threshold autoregression (1990)
- Bayesian online changepoint detection (2007) — link
- Optimal detection of changepoints with a linear computational cost (2012)
- Wild binary segmentation for multiple change-point detection (2014)
- Selective review of offline change point detection methods (2020)
- A survey of methods for time series change point detection (2017)
- Automatic change-point detection in time series via deep learning (2024)
- Change-point detection with deep learning: A review (2025)
- Anomaly Detection for Time Series Using VAE-LSTM Hybrid Model (2020)
- STL: A seasonal-trend decomposition (1990)
- Forecasting, structural time series models and the Kalman filter (1990)
- Going Beyond Static: Understanding Shifts with Time-Series Attribution (2025)
- Causal inference on time series using restricted structural equation models (2013)
- Toward causal representation learning (2021)
- Invariant causal prediction for sequential data (2019)
- Partial Disentanglement via Mechanism Sparsity (2022)
- Marrying causal representation learning with dynamical systems for science (2024)
- Instance-based counterfactual explanations for time series classification (2021)
- CoMTE: Counterfactual explanations for multivariate time series (2021)
- A kernel two-sample test (2012)
- D’ya like DAGs? A survey on structure learning and causal discovery (2022)
- Causal-oriented representation learning for time-series forecasting based on the spatiotemporal information transformation (2025)
- Amortized causal discovery: Learning to infer causal graphs from time-series data (2022)
- Causality (2009)
- The Book of Why: The New Science of Cause and Effect (2018)
- Elements of causal inference: foundations and learning algorithms (2017)
- Time Series Domain Adaptation via Latent Invariant Causal Mechanism (2025) — link
- Change-point detection with deep learning: A review (2025)
- Improvement and application of back random response detection: Based on cumulative sum and change point analysis (2024)
- Causal-oriented representation learning for time-series forecasting based on the spatiotemporal information transformation (2025)
- ADIA Lab Structural Break Challenge (2025)
- QRT Electricity Price Challenge Dataset (2021)
- Time- Causal VAE : Robust Financial Time Series Generator — → 2411.02947
- Causal-oriented representation learning for time-series forecasting based on the spatiotemporal information transformation (2025) — link
- Bayesian Online Changepoint Detection
- Towards Dynamic Structure Changes Detection in Financial Series via Causal Analysis (2021) — link
- Learning Structural Causal Models through Deep Generative Models : Methods , Guarantees , and Challenges (2024) — link
- Causal Inference and Data Fusion in Econometrics (2023) — link
- Day-ahead electricity price forecasting with high-dimensional structures: Univariate vs. multivariate modeling frameworks (2018) — link
- Forecasting day-ahead electricity prices: A review of state-of-the-art algorithms, best practices and an open-access benchmark (2021) — link
- Crafting Papers on Machine Learning (2000)
- The Need for Biases in Learning Generalizations (1980)
- Computational Complexity of Machine Learning (1989)
- Machine Learning: An Artificial Intelligence Approach, Vol. I (1983)
- Pattern Classification (2000)
- Suppressed for Anonymity (2021)
- Mechanisms of Skill Acquisition and the Law of Practice (1981)
- Some Studies in Machine Learning Using the Game of Checkers (1959)
- Vector autoregressions (2001)
- Time series analysis by state space methods (2012)
- Time series deconfounder: Estimating treatment effects over time in the presence of hidden confounders (2020)
- Causal discovery with attention-based convolutional neural networks (2019)
- Time-series generative adversarial networks (2019)
- Time-series generation by contrastive imitation (2021)
- The sensitivity of counterfactual fairness to unmeasured confounding (2020)
- Reintroducing Dynamics into Static Causal Models (2019)
- Causal Models for Dynamical Systems (2022) — link
- Dynamic Structural Causal Models (2024)
- A Dynamical View of the Question of Why (2024) — link
- Algorithmic Recourse: from Counterfactual Explanations to Interventions (2021) — link
- Optimal transport and Wasserstein distances for causal models (2024) — link
- On Pearl’s Hierarchy and the Foundations of Causal Inference (2022) — link
- Learning Structural Causal Models through Deep Generative Models: Methods, Guarantees, and Challenges (2024) — link
- A Survey on Causal Discovery Methods for I.I.D. and Time Series Data (2023) — link
- A logic for causal inference in time series with discrete and continuous variables (2011)
- Continuous-Time Modeling of Counterfactual Outcomes Using Neural Controlled Differential Equations (2022)
- Partial counterfactual identification of continuous outcomes with a curvature sensitivity model (2023)
- On Pearl’s Hierarchy and the Foundations of Causal Inference (2022) — link
- Auto-encoding variational bayes (2013) — link
- Intervening and letting go: On the adequacy of equilibrium causal models (2021)
- Causal inference for time series (2023)
- Causal discovery in a complex industrial system: A time series benchmark (2024)
- Density estimation using Real NVP (2017) — link
- Causal machine learning for predicting treatment outcomes (2024)
- Data Generating Process to Evaluate Causal Discovery Techniques for Time Series Data (2020)
- The causal-neural connection: Expressiveness, learnability, and inference (2021)
- Temporally Disentangled Representation Learning (2022)
- Causal Representation Learning for Instantaneous and Temporal Effects in Interactive Systems (2023)
- On the Identifiability of Switching Dynamical Systems (2024) — link
- CausalTime: Realistically Generated Time-series for Benchmarking of Causal Discovery (2024) — link
- Dy CAST : Learning Dynamic Causal Structure from Time Series (2025) — link
- Cot-gan: Generating sequential data via causal optimal transport (2020)
- From optimal transport to generative modeling: the VEGAN cookbook (2017) — link
- Forecasting treatment responses over time using recurrent marginal structural networks (2018)
- Estimating counterfactual treatment outcomes over time through adversarially balanced representations (2020) — link
- Tts-gan: A transformer-based time-series generative adversarial network (2022)
- CausalVAE: Disentangled Representation Learning via Neural Structural Causal Models (2021)
- Weakly supervised disentangled generative causal representation learning (2022)
- Causal inference meets deep learning: A comprehensive survey (2024)
- Deep Structural Causal Models for Tractable Counterfactual Inference (2020)
- Partial counterfactual identification from observational and experimental data (2022)
- Causal machine learning for cost-effective allocation of development aid (2024)
- Estimating the effects of continuous-valued interventions using generative adversarial networks (2020)
- Causal transformer for estimating counterfactual outcomes (2022)
- Time Series Foundation Models for Multivariate Financial Time Series Forecasting — → 2507.07296
- AI Challenges in Mathematical Investing (2025) — link
- Causal discovery from temporally aggregated time series (2017)
- Quantitative Financial Models with Scenarios from LLM : Temporal Fusion Transformers as Alternative Monte - Carlo (2024) — link
- Quantitative Models with Qualitative Scenarios: Simulation with Transformer and Large Language Model (2025)
- Generative Models in Finance: Market Generators, a Paradigm Shift in Financial Modeling (2025)
- Generative Models in Finance : Market Generators , a Paradigm Shift in Financial Modeling (2025) — link
- Toward Automating Causal Discovery in Financial Markets and Beyond (2025)
- Causality-Inspired Models for Financial Time Series Forecasting — → 2408.09960
- Temporally Disentangled Representation Learning under Unknown Nonstationarity (2024) — link
- The market generator (2019)
- Empirical properties of asset returns: stylized facts and statistical issues (2001)
- Theoretical foundations of deep selective state-space models (2024)
- Quant GANs: deep generation of financial time series (2020)
- Generating realistic stock market order streams (2020)
- Backtesting (2015)
- Time-Causal VAE: Robust Financial Time Series Generator — → 2411.02947
- Long short-term memory (1997)
- On the Properties of Neural Machine Translation: Encoder–Decoder Approaches (2014)
- Optimal transport and Wasserstein distances for causal models (2025)
- The Brownian movement and stochastic equations (1942)
- Arbitrage theory in continuous time (2009)
- Diffusion Factor Models: Generating High-Dimensional Returns with Factor Structure — → 2504.06566
- Event-Based Limit Order Book Simulation under a Neural Hawkes Process: Application in Market-Making — → 2502.17417
- Predicting Liquidity-Aware Bond Yields using Causal GANs and Deep Reinforcement Learning with LLM Evaluation — → 2502.17011
- Do- PFN : In-Context Learning for Causal Effect Estimation (2025) — link
9 of 134 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.