Projects per year
Abstract
Energy generation from wind and sun is increasing rapidly in many parts of the world. This presents new challenges on how to integrate this uncertain, intermittent and non-dispatchable energy source. This thesis deals with forecasting and decision making in energy systems with a large proportion of renewable energy generation. Particularly we focus on producing forecasting models that can predict renewable energy generation, single user demand, and provide advanced forecast products that are needed for an efficient integration of renewable energy into the power generation mix. Such forecasts can be useful on all levels of the energy systems, ranging from the highest level, where the transmission system operator is concerned with minimizing system failures and is aided by wind power forecasts, to the end user of energy where power price forecasts are useful for users with flexible power demand.
The main contributions of this thesis lie in the realm of using gray box models to produce forecasts for energy systems. Gray box models can be defined as a crossover between physical models (or white box models), that base their model on a physical understanding of the system at hand, and data driven models (or black box models) that focus on accurately describing the data without considering physical limitations of the system. Integrating these physical structures into a data driven approach allows for producing better forecasts with more accurate predictions. In this thesis we have developed and applied methodologies for gray box modeling to produce forecasts for vehicle driving patterns, solar irradiance, wind speeds, wind power, and solar power. The model for driving patterns has subsequently been used as input into an optimization algorithm for charging a single electric vehicle. In a subsequent study the behavior of a fleet of electric vehicles has been studied.
In the thesis we go through various examples of forecasts products and their applications. We emphasize that forecasting can not stand alone and should be complimented by optimization and decision making tools for an efficient integration of renewable energy.Thus forecast products should be developed in unison with the decision making tool as they are two sides of the same overall challenge.
The main contributions of this thesis lie in the realm of using gray box models to produce forecasts for energy systems. Gray box models can be defined as a crossover between physical models (or white box models), that base their model on a physical understanding of the system at hand, and data driven models (or black box models) that focus on accurately describing the data without considering physical limitations of the system. Integrating these physical structures into a data driven approach allows for producing better forecasts with more accurate predictions. In this thesis we have developed and applied methodologies for gray box modeling to produce forecasts for vehicle driving patterns, solar irradiance, wind speeds, wind power, and solar power. The model for driving patterns has subsequently been used as input into an optimization algorithm for charging a single electric vehicle. In a subsequent study the behavior of a fleet of electric vehicles has been studied.
In the thesis we go through various examples of forecasts products and their applications. We emphasize that forecasting can not stand alone and should be complimented by optimization and decision making tools for an efficient integration of renewable energy.Thus forecast products should be developed in unison with the decision making tool as they are two sides of the same overall challenge.
Original language | English |
---|
Place of Publication | Kgs. Lyngby |
---|---|
Publisher | Technical University of Denmark |
Number of pages | 175 |
Publication status | Published - 2016 |
Series | DTU Compute PHD-2015 |
---|---|
Number | 363 |
ISSN | 0909-3192 |
Fingerprint
Dive into the research topics of 'Probabilistic Approaches to Energy Systems'. Together they form a unique fingerprint.Projects
- 1 Finished
-
Multivariate Probabilistic Forecasting for Energy Systems
Iversen, J. E. B. (PhD Student), Madsen, H. (Main Supervisor), Morales González, J. M. (Supervisor), Møller, J. K. (Supervisor), Pinson, P. (Examiner), Dent, C. (Examiner) & Lindström, E. (Examiner)
Technical University of Denmark
01/10/2011 → 21/09/2015
Project: PhD