Uncertainty-Aware Flexibility of Buildings: From Quantification to Provision

TL;DR

The paper tackles uncertainty in quantifying buildings' heating flexibility by introducing uncertainty-aware energy envelopes via a chance-constrained framework, augmented with affine feedback to model real-time baseline adaptation. It compares uncertainty-ignorant envelopes, uncertainty-aware envelopes, and fixed or optimal affine feedback policies, analyzing two operation modes: intra-day market participation and rebound adaptation in reserve markets. Through a case study on the UMAR building participating in the German aFRR market, the work demonstrates that accounting for forecast and model inaccuracies is necessary to avoid overestimation of potential, while affine feedback can expand the flexible horizon and improve revenues under comfort constraints. A comfort-cost approach is proposed to fairly compare formulations, highlighting the trade-off between financial gains and thermal discomfort, and showing broader practical implications for demand-side flexibility in future power systems.

Abstract

Buildings represent a promising flexibility source to support the integration of renewable energy sources, as they may shift their heating energy consumption over time without impacting users' comfort. However, a building's predicted flexibility potential is based on uncertain ambient weather forecasts and a typically inaccurate building thermal model. Hence, this paper presents an uncertainty-aware flexibility quantifier using a chance-constrained formulation. Because such a quantifier may be conservative, we additionally model real-time feedback in the quantification, in the form of affine feedback policies. Such adaptation can take the form of intra-day trades or rebound around the flexibility provision period. To assess the flexibility quantification formulations, we further assume that flexible buildings participate in secondary frequency control markets. The results show some increase in flexibility and revenues when introducing affine feedback policies. Additionally, it is demonstrated that accounting for uncertainties in the flexibility quantification is necessary, especially when intra-day trades are not available. Even though an uncertainty-ignorant potential may seem financially profitable in secondary frequency control markets, it comes at the cost of significant thermal discomfort for inhabitants. Hence, we suggest a comfort-preserving approach, aiming to truly reflect thermal discomfort on the economic flexibility revenue, to obtain a fairer comparison.

Figures (10)

• Figure 1: Flowchart describing the exchange of information between the stakeholders and the organization of the paper.
• Figure 2: Comparison between the measured model error and the state-space representation of the model error.
• Figure 3: Average flexibility envelope of UMAR in the set $\mathcal{S}_{\text{env}}^{20}$ for the different formulations over a horizon of a day, with $1-\epsilon_{\text{\tiny C}} = 80\%$ and $\mathbf{T}_{\text{max}} - \mathbf{T}_{\text{min}} = 2$° C.
• Figure 4: Average and maximum distance in $\mathcal{S}_{\text{env}}^{20}$ between the optimal affine policy $\mathbf{M}$ and the average one $\mathbf{M}^f_{\text{\scriptsize avg}}$, as a function of the number of samples used to compute $\mathbf{M}^f_{\text{\scriptsize avg}}$.
• Figure 5: Comparison between the uncertainty-aware and the uncertainty-aware with fixed feedback formulations, in $\mathcal{S}_{\text{env}}^{20}$.