Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Predictions and Decision Making for Resilient Intelligent Sustainable Energy Systems

Martin Braun, Christian Gruhl, Christian A. Hans, Philipp Härtel, Christoph Scholz, Bernhard Sick, Malte Siefert, Florian Steinke, Olaf Stursberg, Sebastian Wende-von Berg

TL;DR

The paper addresses resilience in future energy systems subject to multiple uncertainties across timescales, proposing a framework that combines probabilistic uncertainty modeling with multi-level decision-making. It surveys recent trends in probabilistic forecasting and calibration, and analyzes decision schemes including MPC (and its robust, stochastic, and risk-averse variants), operational scheduling, and long-horizon expansion planning. Key contributions include synthesizing how forecast quality, uncertainty representations, and control/optimization methods interact to improve resilience, and highlighting gaps such as integration of probabilistic information into real-time decisions and adaptive planning under climate and market variability. The work underscores the practical impact of aligning forecasting, control, and planning with resilience objectives to enable secure, efficient, and sustainable energy systems.

Abstract

Future energy systems are subject to various uncertain influences. As resilient systems they should maintain a constantly high operational performance whatever happens. We explore different levels and time scales of decision making in energy systems, highlighting different uncertainty sources that are relevant in different domains. We discuss how the uncertainties can be represented and how one can react to them. The article closes by summarizing, which uncertainties are already well examined and which ones still need further scientific inquiry to obtain resilient energy systems.

Predictions and Decision Making for Resilient Intelligent Sustainable Energy Systems

TL;DR

The paper addresses resilience in future energy systems subject to multiple uncertainties across timescales, proposing a framework that combines probabilistic uncertainty modeling with multi-level decision-making. It surveys recent trends in probabilistic forecasting and calibration, and analyzes decision schemes including MPC (and its robust, stochastic, and risk-averse variants), operational scheduling, and long-horizon expansion planning. Key contributions include synthesizing how forecast quality, uncertainty representations, and control/optimization methods interact to improve resilience, and highlighting gaps such as integration of probabilistic information into real-time decisions and adaptive planning under climate and market variability. The work underscores the practical impact of aligning forecasting, control, and planning with resilience objectives to enable secure, efficient, and sustainable energy systems.

Abstract

Future energy systems are subject to various uncertain influences. As resilient systems they should maintain a constantly high operational performance whatever happens. We explore different levels and time scales of decision making in energy systems, highlighting different uncertainty sources that are relevant in different domains. We discuss how the uncertainties can be represented and how one can react to them. The article closes by summarizing, which uncertainties are already well examined and which ones still need further scientific inquiry to obtain resilient energy systems.
Paper Structure (8 sections, 3 equations, 3 figures, 1 table)

This paper contains 8 sections, 3 equations, 3 figures, 1 table.

Table of Contents

  1. Introduction
  2. Recent trends in uncertainty modeling
  3. Model predictive control
  4. Dealing with external uncertainties
  5. Dealing with uncertain system models
  6. Operational scheduling
  7. Expansion and investment planning
  8. Conclusions

Figures (3)

  • Figure 1: Time horizon and system boundaries of decision-making schemes.
  • Figure 2: The dashed line is the (unknown) ground truth. In contrast to the point forecast, the probabilistic forecasts reflect the variability in future outcomes. Quantile and density forecasts directly describe the uncertainty in the forecast, while ensemble forecasts consider discrete future outcomes. Data from Han2021.
  • Figure 3: The reliability diagram plots observed quantiles against predicted ones. Deviations from the diagonal indicate over- or underconfidence. A reliable predictor should be close to the diagonal. The pinball loss penalizes the deviation $\Delta$ between a predicted quantile and an observed value.