David Zapata Gonzalez
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
David Zapata Gonzalez has a total of 7 co-authors. Here are the top co-authors by number of shared publications:
- Oliver Müller
Paderborn University, Paderborn, Germany
3 shared publications - Marcel Meyer
Data analytics group Paderborn University
2 shared publications Florian Schlosser
1 shared publication- Katharina Brennig
Paderborn University
1 shared publication - Kevin Zalipski
Data analytics group Paderborn University
1 shared publication - Sascha Kaltenpoth
Data analytics group Paderborn University
1 shared publication
Latest Publications
- Open Access
- 24 September 2026
- Published by Elsevier in SSRN Electronic Journal
Process and Energy Digital Twin: Modeling and Simulating Industrial Processes and their Dynamic Energy Profiles with Process Mining and Machine Learning
Improving industrial energy efficiency requires models of how industrial processes generate their energy demand. Manually constructing process and energy models is costly and time-consuming. Existing data-driven approaches focus on industrial processes and reduce energy to aggregated values, leaving the continuous energy demand and its process causes unmodeled. This paper proposes a framework for automatically constructing a Process and Energy Digital Twin (ProEn-DT) from historical event and sensor data. Process Mining discovers the activity-level process structure underlying the energy demand, while Machine Learning predicts activity and case durations. The energy demand is represented as a continuous profile per activity, with its shape learned from historical profiles and adapted to activity duration, product, and external conditions. Coupling these models translates production plans into detailed process execution and energy profiles. We evaluate the framework on six industrial processes: one synthetic and five from real food-production facilities. Compared with benchmark methods from the literature, the framework reduces energy-profile generation error by approximately 30% and joint process and energy simulation error by 45%. Among the evaluated approaches, only those deriving the energy demand from simulated process execution reproduce the timing and shape of observed energy profiles. By linking energy demand to activities and their execution context, the Digital Twin helps practitioners understand how energy patterns arise and assess the energy consequences of new production plans, supporting process-and energy-aware decisions and more sustainable industrial operations.
- 24 September 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessProcess and Energy Digital Twin: Modeling and Simulating Industrial Processes and their Dynamic Energy Profiles with Process Mining and Machine Learning
Improving industrial energy efficiency requires models of how industrial processes generate their energy demand. Manually constructing process and energy models is costly and time-consuming. Existing data-driven approaches focus on industrial processes and reduce energy to aggregated values, leaving the continuous energy demand and its process causes unmodeled. This paper proposes a framework for automatically constructing a Process and Energy Digital Twin (ProEn-DT) from historical event and sensor data. Process Mining discovers the activity-level process structure underlying the energy demand, while Machine Learning predicts activity and case durations. The energy demand is represented as a continuous profile per activity, with its shape learned from historical profiles and adapted to activity duration, product, and external conditions. Coupling these models translates production plans into detailed process execution and energy profiles. We evaluate the framework on six industrial processes: one synthetic and five from real food-production facilities. Compared with benchmark methods from the literature, the framework reduces energy-profile generation error by approximately 30% and joint process and energy simulation error by 45%. Among the evaluated approaches, only those deriving the energy demand from simulated process execution reproduce the timing and shape of observed energy profiles. By linking energy demand to activities and their execution context, the Digital Twin helps practitioners understand how energy patterns arise and assess the energy consequences of new production plans, supporting process-and energy-aware decisions and more sustainable industrial operations.
- Open Access
- 8 May 2026
- Published by Elsevier in Applied Energy
- Vol. 417
Temporal causal feature selection for robust machine learning modelling of data center operations
- 8 May 2026
- Published by Elsevier in Applied Energy
- Vol. 417
Open AccessTemporal causal feature selection for robust machine learning modelling of data center operations
- Open Access
- 24 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 218141-218153
Benchmarking Time Series Foundation Models for Short-Term Household Electricity Load Forecasting
Accurate household electricity short-term load forecasting (STLF) is key to future and sustainable energy systems. While various studies have analyzed statistical, machine learning, or deep learning approaches for household electricity STLF, recently proposed time series foundation models such as Chronos, TimesFM or Time-MoE promise a new approach for household electricity STLF. These models are trained on a vast amount of time series data and are able to forecast time series without explicit task-specific training (zero-shot learning). In this study, we benchmark the forecasting capabilities of time series foundation models compared to Trained-from-Scratch (TFS) Transformer-based approaches. Our results suggest that foundation models perform comparably to TFS Transformer models, while certain time series foundation models outperform all TFS models when the input size increases. At the same time, they require less effort, as they need no domain-specific training and only limited contextual data for inference.
- 24 December 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 218141-218153
Open AccessBenchmarking Time Series Foundation Models for Short-Term Household Electricity Load Forecasting
Accurate household electricity short-term load forecasting (STLF) is key to future and sustainable energy systems. While various studies have analyzed statistical, machine learning, or deep learning approaches for household electricity STLF, recently proposed time series foundation models such as Chronos, TimesFM or Time-MoE promise a new approach for household electricity STLF. These models are trained on a vast amount of time series data and are able to forecast time series without explicit task-specific training (zero-shot learning). In this study, we benchmark the forecasting capabilities of time series foundation models compared to Trained-from-Scratch (TFS) Transformer-based approaches. Our results suggest that foundation models perform comparably to TFS Transformer models, while certain time series foundation models outperform all TFS models when the input size increases. At the same time, they require less effort, as they need no domain-specific training and only limited contextual data for inference.