Chance-Constrained Optimization for Flexibility Provision From the Local Energy Communities Using Logit-Based Regression

Abstract

Local energy communities (LECs) represent a paradigm shift toward decentralized power management, facilitating self-consumption and efficient resource sharing. In addition, battery energy storage systems can empower LECs for flexibility provision (FP). However, the FP performance of LECs relies on the prediction capability, which unfortunately suffers from the uncertain nature of distributed energy resources, and electric vehicle’s charging demand. While (profile-based) stochastic optimization (SO) can partly handle forecast errors, integrating it with the network-aware model is computationally demanding in a rolling horizon optimization framework. To overcome this computational burden, this paper proposes a novel method based on chance-constrained optimization (CCO) in convex form by leveraging logit-based regression. By deriving a closed-form expression for probabilistic constraints, this approach correlates forecast errors with network issues, such as congestion and voltage violations. Numerical simulations are conducted on the modified IEEE 33-bus network connected to two LECs in the Bunnik campus, the Netherlands to demonstrate the method’s effectiveness. The proposed CCO method outperforms profile-based SO and network-aware SO by factors of 8 and 165, respectively, in terms of reducing computational time. Additionally, it limits voltage violation risks to below 5% compared to 20% in profile-based SO along with a 2% reduction in operations cost.

Keywords

• Chance-constrained optimization
• flexibility provision
• logit-based regression
• second-order cone programming
• rolling horizon