Hierarchical distribution system optimization under uncertainty

Details

Supervisors

Papavasiliou

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en mathématiques appliquées

Abstract

The integration of renewable energy resources (RES) leads to a significant change in the current state of the energy market in recent years. Although the high-voltage transmission system has been optimized with the integration of RES, the low-voltage distribution system has largely undeveloped due to its complexity. The difficulty of the distribution network operations can be characterized mainly by the following three parts: (i) Uncertainty, (ii) Scale, and (iii) Non-linearity of power flow constraints, and we tackle the first two challenges in this thesis. We propose a hierarchical approach which focuses on the radial structure of the distribution network that we can divide the network into small sub-networks (layers). In such a network, interactions of layers are weak in the sense that they exchange power only on a few interfaces. This fact implies that the possibility of solving a problem of each layer of the hierarchy, independently. Thus, the approach can be applicable to systems of arbitrary size. In the first proposed approach, we decompose the network and implement the stochastic dual dynamic programming (SDDP) at each layer in order to obtain the value functions which contain the information of future costs at a given stage. In the second approach, we combine the idea of the previous algorithm and Model Predictive Control (MPC) which is known as an iterative optimization algorithm with a feedback. We test our policies on a small-scale and a large-scale problem with several distribution system models as case studies. We demonstrate the superiority of proposed policies in terms of average performance and computation time. This study can be an important starting point for considering fully distributed systems or future electric market designs.