Conference Agenda

Power System Optimization Models are tools policymakers and practitioners use to evaluate and plan such systems' short, medium, and long-term evolution. The size and complexity of these models have evolved alongside their real-world counterparts to the point that they pose tractability problems that hinder their usefulness and suitability for extracting actionable insights. One of the main challenges in these models arises from their temporal structure, which makes them harder to solve as it exponentially increases the number of variables in the model; to overcome this, researchers developed temporal aggregation techniques which simplify the temporal structure of the model like representative periods and temporal downsampling, thus increasing the model's tractability. These techniques, however, come at the expense of losing sight of the interaction between short and long-term dynamics that play a critical role in the real world, like ramping or the intra-day behavior of renewable energy sources. In this work, we extend the Basis-Oriented Time Series Aggregation procedure to network flow problems and show how it greatly aggregates the model while maintaining its objective function value.

This study is in the context of the design of multi-energy microgrids under uncertainty, where the objective is to determine optimal sizes for renewable generators and flexibility assets considering stochastic parameters, such as energy demand and renewable energy sources availability.

In particular, we compare existing models to generate synthetic scenarios of these parameters leveraging historical data, e.g. using Markov Chains, probability distributions, and time series analysis. The assessment focuses on their ability to generate diverse scenarios that capture key characteristics of the original data. Additionally, we conduct an application-specific assessment to examine the impact of different scenario generation methods on design optimization. This evaluation utilizes a two-stage stochastic programming approach and a three-year dataset to evaluate performance on unseen scenarios.

Robust design of microgrids is a complex optimization process requiring multiple simulations in order to integrate uncertainty variables associated with the system environment or design models. In this context, having sufficiently accurate models that are compatible with the optimization algorithms and associated computational costs represents a real challenge. In this paper, we illustrate this through the robust design of a simple microgrid with electrochemical storage. Based on battery models that couple energy efficiency and aging, we develop an approach for choosing the right level of precision to match the microgrid's optimization criteria or constraints.