Skip to Main content Skip to Navigation
New interface

Estimation et prédiction des productions d’énergies renouvelables et des consommations d’un réseau de distribution d’électricité

Abstract : In recent years, electricity distribution system operators (DSOs) have seen a significant integration of renewable energy production. In addition,with the development of new energy uses, such as electric mobility, microgrids and storage technologies, grid management has become more and more complex, a complexity that will increase in the years to come. In this context, SRD, DSO in the department of Vienne in France,has funded research in the field of smartgrids, notably on the dynamic optimization of the operating scheme of its high and medium voltage electricity distribution network, and this thesis on the forecasting, prediction and estimation of the values of the energy flows circulating on this network. In the first phase, the thesis proposes an approach to select the most relevant energy consumption data and study their influence on the efficiency of the optimization. For this purpose, a data dimensionality reduction approach is proposed. It uses techniques from the unsupervised machine learning and temporal data analysis fields. This methodology allows to detect similarities in the data in order to group them in homogeneous groups. The second phase develops a methodology for estimating the energy production of photovoltaic installations distributed over the SRD distribution network using spatial interpolation methods. In fact, most of the generators with low voltage production capacities are not instrumented for real time measurement. On the other hand, medium and high voltage generators are instrumented and allow fine-grained time series measurements. The goal of this study is to estimate the production of thousands of small generators by exploiting the data of medium and large generators of reference equipped with communicating meters. Finally, the problem of forecasting photovoltaic production is considered. The goal of this study is to develop a short-term point forecast with a horizon of one hour to manage the intermittency of the solar production and a long-term probabilistic forecast to schedule and optimize the network over a horizon of one to three months. We show that machine learning algorithms with a global approach improve the forecasts provided by naive methods. The pertinence of the obtained forecasts to the application framework has been validated using a loadflow estimator to quantify the differences in losses, voltage drops and voltage rises between a forecasted state and an actual state.
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Monday, December 13, 2021 - 3:05:21 PM
Last modification on : Thursday, September 8, 2022 - 3:59:50 AM
Long-term archiving on: : Monday, March 14, 2022 - 7:11:56 PM


Version validated by the jury (STAR)


  • HAL Id : tel-03477632, version 1



Mohamed Tribak. Estimation et prédiction des productions d’énergies renouvelables et des consommations d’un réseau de distribution d’électricité. Autre [cs.OH]. ISAE-ENSMA Ecole Nationale Supérieure de Mécanique et d'Aérotechique - Poitiers, 2021. Français. ⟨NNT : 2021ESMA0012⟩. ⟨tel-03477632⟩



Record views


Files downloads