Uncertainty quantification plays a significant role in computational energy research. For instance, physical models of new energy storage are helpful tools in electromobility applications but may suffer from random inputs. Other relevant examples are renewable energy units and reliable energy network systems under volatile sources. The minisymposium covers the broad field of computational methods for energy and power systems with a particular focus on efficient methods for uncertainty quantification and sensitivity analysis. The primary purpose is to identify common methodologies and interfaces. Contributions will address applications of current interest as well as efficient algorithms and their mathematical background.
08:30
Robust Intrusive Uncertainty quantification for coupled flow physics
Ryan McClarren | University of Notre Dame | United States
Show details
Author:
Ryan McClarren | University of Notre Dame | United States
Hyberbolic systems are common occurrences in the simulation of energy systems. In this work we consider how instructive uncertainty quantification methods that utilize filtered stochastic expansions can give robust uncertainty quantification. In particular, we consider the challenging problem of radiation hydrodynamics, a flow problem important in inertial confinement fusion and other high-temperature applications. We demonstrate how standard intrusive methods are either fragile, in the case of global polynomial expansions, or expensive, in the case of the intrusive polynomial moment (IPM) method. Our results show that we can have the speed of polynomial expansions with the robustness of IPM. We also indicate how our method could apply to other reactive flow problems.
09:00
Probabilistic Forecasting and Stochastic Programming for Optimal Bidding in Energy Markets
Alexander Dowling | University of Notre Dame | United States
Show details
Authors:
Xian Gao | University of Notre Dame | United States
Alexander Dowling | University of Notre Dame | United States
Undoubtedly, evolving wholesale electricity markets continue to provide new revenue opportunities for diverse generation, energy storage, and flexible demand technologies. In this paper, we quantitatively explore how uncertainty impacts optimal market participation strategies and resulting revenues. Specifically, we benchmark 2-stage stochastic programming formulations for self-schedule and bidding market participation modes in a receding horizon model predictive control framework. To generate probabilistic price forecasts, we propose an autoregressive Gaussian process regression model and compare three sampling strategies. As an illustrative example, we study price-taker thermal generators and hybrid energy systems using historical price data from CAISO (California market). We show that self-schedule is sensitive to the error in the forecast mean, whereas bidding requires price forecasts that cover extreme events (e.g., tails of the distribution). When paired with the best forecasting technique, we find optimization under uncertainty can realize 85% to 95% of perfect information revenues.
09:30
Global Sensitivity Analysis of the Transient Thermal Impedance of Power Semiconductor Heat Sinks
Dimitrios Loukrezis | Technische Universität Darmstadt | Germany
Show details
Authors:
Dimitrios Loukrezis | Technische Universität Darmstadt | Germany
Robin Scheich | Technische Universität Darmstadt | Germany
Herbert De Gersem | Technische Universität Darmstadt | Germany
Power semiconductor devices are crucial parts of power electronic applications, employed to control or convert electric power. We consider the problem of modeling the behavior of a heat sink, i.e., a passive heat exchanger that dissipates the heat due to Joule losses during the operation of the semiconductor device. In particular, we are interested in the transient thermal impedance of the heat sink for varying operation and design parameters. The common approach of characterizing the thermal impedance via an expensive CFD simulation becomes unaffordable when a large number of parameter realizations must be examined. As an alternative, we use a small number of CFD simulations to construct a sufficiently accurate surrogate model. Moreover, we reduce the parameter space by performing a surrogate-based sensitivity analysis (SA) and discarding non-influential parameters. The SA also reveals varying parameter influences depending on the time point of the device’s transient operation.
10:00
- CANCELED - Statistical inference and stochastic simulation for uncertainty quantification with imperfect dataset
Pengfei Wei | Northwestern Polytechnical University | China
Show details
Authors:
Pengfei Wei | Northwestern Polytechnical University | China
Matteo Broggi | Leibniz University Hannover, Institute for Risk and Reliability | Germany
Michael Beer | Leibniz University Hannover, Institute for Risk and Reliability | Germany
Imprecise Probability models have been widely accepted as a natural extension of the probability theory for treating uncertainty while the available dataset is imperfect (small, imprecise and incomplete). With imperfect datasets, two kinds of uncertainties are introduced, i.e., the aleatory uncertainty and epistemic uncertainty, where the former one is due to the intrinsic randomness, and cannot be reduced, while the latter is caused by lack of information, and can be reduced by collecting more data or improving the data quality. The most appealing advantage of imprecise probability models is the utilization of a hierarchical model structure, which enables separating the two kinds of uncertainties. However, most of the current studies are based on distributional imprecise probability models, which leads to a biased estimation of probability bounds while the assumed distribution type is not consistent with the real one. For dealing with this problem, we firstly develop a statistical inference technique for generating a distribution-free p-box model, and then propose a statistical hypothesis test method for judging the fitness of the distributional p-box model to the dataset. The target is to make sure that the generated probability bounds of any event should, definitely or with a high confidence, include the real probability value. We also develop an efficient simulation method for propagating the imprecise probability models.