Deep Learning for Energy Forecasting

From RNNs to Transformers: Building Production-Ready Forecasters
Erscheinungsjahr 2026
ISBN: 978-981-9238-99-6
Verlag: Springer

Buch, Englisch, 474 Seiten, Format (B × H): 155 mm x 235 mm

Reihe: Power Systems

This book provides an end-to-end, practice-oriented path from fundamental deep-learning concepts to state-of-the-art sequence models for time series, with a sustained focus on energy use cases. Readers learn how to formulate forecasting problems, engineer data pipelines, select and train neural architectures (RNNs, attention-based seq2seq, CNNs, and Transformers), and evaluate models with robust metrics and baselines. Dedicated chapters cover multivariate and hierarchical settings, probabilistic forecasting for uncertainty quantification, and domain-specific workflows for load and renewable generation forecasting. The final part turns models into usable systems, addressing hyperparameter optimization, reproducibility, deployment, monitoring, and practical failure modes. Primary audiences include graduate students, researchers, and practitioners who build forecasting models for electricity demand, renewable generation, and related energy time-series tasks.

Zielgruppe

Upper undergraduate

Autoren/Hrsg.

Liu, Xiufeng

Fachgebiete

• Mathematik | Informatik EDV | Informatik Informatik Künstliche Intelligenz Maschinelles Lernen
• Technische Wissenschaften Energietechnik | Elektrotechnik Energieverteilung, Stromnetze
• Technische Wissenschaften Energietechnik | Elektrotechnik Energietechnik & Elektrotechnik

Inhaltsverzeichnis

Preface.- Introduction Why Deep Learning for Energy.- Deep Learning Foundations for Time Series.- Recurrent Neural Networks.- Sequence to-Sequence Models with Attention.- Convolutional Neural Networks for Time Series.

Über Autor(innen)

Xiufeng Liu is a Senior Researcher (associate professor level) at the Technical University of Denmark (DTU), Department of Technology, Management and Economics. His research focuses on data systems and machine learning for energy and smart-city applications, with particular emphasis on time-series forecasting, anomaly detection, distributed/federated learning, and visual analytics. He received his Ph.D. in Computer Science from Aalborg University and has held research appointments at DTU and the University of Waterloo.

His academic work lies at the interface of methodological innovation and operational deployment: he develops forecasting models (from classical baselines to modern sequence architectures such as RNNs, CNNs, attention models, and Transformers) and integrates them into reproducible pipelines that address data quality, uncertainty quantification, evaluation, and production monitoring. He wrote this book to bridge fast-moving deep-learning research with the practical needs of energy forecasting practitioners and graduate learners, emphasizing clear modeling choices, rigorous evaluation, and deployment-ready engineering workflows. Contact: [e-mail];ORCID: 0000-0001-5133-6688.