Day-Ahead Energy Price Forecasting with Machine Learning: Role of Endogenous Predictors DOI Creative Commons
Chibuike Chiedozie Ibebuchi

Forecasting, Год журнала: 2025, Номер 7(2), С. 18 - 18

Опубликована: Апрель 9, 2025

Accurate Day-Ahead Energy Price (DAEP) forecasting is essential for optimizing energy market operations. This study introduces a machine learning framework to predict the DAEP with 24 h lead time, leveraging historical data and forecasts available at prediction time. Hourly from California Independent System Operator (January 2017 July 2023) were integrated exogenous engineered endogenous features. A custom rolling window cross-validation, validation blocks sliding daily across 2372 folds, evaluates an Extreme Gradient Boosting (XGBoost) model’s performance under diverse conditions, achieving median mean absolute error of 6.26 USD/MWh root squared 8.27 USD/MWh, variability reflecting volatility. The feature importance analysis using Shapley additive explanations highlighted dominance features in driving time relatively stable conditions. Forecasting runtime 10 AM on prior day was used assess model uncertainty. involved training random forest, support vector regression, XGBoost, feed forward neural network models, followed by stacking voting ensembles. results indicate need ensemble evaluation beyond static train–test split ensure practical utility varied dynamics. Finally, operationalizing forecast bidding decisions real-time prices presented discussed.

Язык: Английский

Technological innovation, trade openness, natural resources, clean energy on environmental sustainably: a competitive assessment between CO2 emission, ecological footprint, load capacity factor and inverted load capacity factor in BRICS+T DOI Creative Commons
Jie Sun,

Md. Qamruzzaman

Frontiers in Environmental Science, Год журнала: 2025, Номер 12

Опубликована: Фев. 4, 2025

The study investigates the relationship between technological innovation, clean energy, trade openness, and natural resource rents on environmental sustainability within BRICS + T nations. Motivated by urgent need to address escalating CO2 emissions—reaching 36.4 billion metric tons in 2022—the research aims understand how these factors influence emissions, ecological footprint, load capacity factor, its inverse, contributing Sustainable Development Goals (SDGs). uses panel data from countries spanning period 1990 2022. Employing advanced econometric techniques such as Dynamic Seemingly Unrelated Regression (DSUR), Cross-Sectionally Augmented Panel Unit Root (CUP-FM, CUP-BC), nonlinear autoregressive distributed lag (ARDL) models, tests Environmental Kuznets Curve (EKC) hypothesis evaluates asymmetric effects of variables. Key findings indicate that innovation consistently reduces emissions footprints, reinforcing role promoting through cleaner technologies more efficient industrial processes. Clean energy adoption has also been shown be a significant driver reducing degradation, with consistent negative while improving factor. However, openness exhibits dual effect. While it enhances use efficiency, simultaneously increases likely due heightened activity. Natural display mixed results: some cases, they exacerbate others, contribute funding eco-friendly initiatives. recommends nations prioritize investments green technologies, strengthen regulations, enhance international collaboration accelerate transition renewable energy. Policymakers should balance benefits stricter standards mitigate adverse sustainability. These integrated strategies are essential for achieving targets outlined SDGs.

Язык: Английский

Процитировано

4
Day-Ahead Energy Price Forecasting with Machine Learning: Role of Endogenous Predictors DOI Creative Commons
Chibuike Chiedozie Ibebuchi

Forecasting, Год журнала: 2025, Номер 7(2), С. 18 - 18

Опубликована: Апрель 9, 2025

Accurate Day-Ahead Energy Price (DAEP) forecasting is essential for optimizing energy market operations. This study introduces a machine learning framework to predict the DAEP with 24 h lead time, leveraging historical data and forecasts available at prediction time. Hourly from California Independent System Operator (January 2017 July 2023) were integrated exogenous engineered endogenous features. A custom rolling window cross-validation, validation blocks sliding daily across 2372 folds, evaluates an Extreme Gradient Boosting (XGBoost) model’s performance under diverse conditions, achieving median mean absolute error of 6.26 USD/MWh root squared 8.27 USD/MWh, variability reflecting volatility. The feature importance analysis using Shapley additive explanations highlighted dominance features in driving time relatively stable conditions. Forecasting runtime 10 AM on prior day was used assess model uncertainty. involved training random forest, support vector regression, XGBoost, feed forward neural network models, followed by stacking voting ensembles. results indicate need ensemble evaluation beyond static train–test split ensure practical utility varied dynamics. Finally, operationalizing forecast bidding decisions real-time prices presented discussed.

Язык: Английский

Процитировано

1