Fiche individuelle

The energy sector is undergoing a profound transformation to tackle climate change issues rapidly and efficiently. The development of decentralized renewable energy and the electrification of the transport sector are two major aspects affecting electrical grids. This paper presents a novel energy management strategy that supervises controllable loads like electric vehicles and electric water heaters in a distribution grid to reduce energy transmission costs and promote the local consumption of renewable energy. The proposed strategy uses day-ahead forecasts of energy production and consumption to schedule future load profiles. It also considers potential grid constraints using a real-time co-simulation approach. The estimated constraints are mitigated by modifying the controllable loads’ dispatch or adjusting their power set-points. The system’s performance is finally evaluated on a real distribution grid model during days that present the most voltage constraints in the year.

In a context where smart grids are emerging, energy management strategies play a major role in tackling different challenges facing their development. This paper proposes a predictive supervision strategy to manage the activation of Electric Vehicles and Electric Water Heaters within a distribution grid. To do so, the energy production and consumption are forecasted using Artificial Neural Networks based on historical measurements and meteorological data. Then, these forecasts are used to determine controllable load profiles that minimize energy transmission costs. The system’s performance is finally tested through simulation using real-time measurements collected from a distribution substation, showing an important reduction in energy transmission costs and also an increase in local renewable energy consumption.

The growing integration of renewable energy production and important loads such as electric vehicles bring a lot of challenges to today’s electric systems. Demand-side management strategies are often presented as one of the solutions to these problems. In this study, a fuzzy logic supervision system is proposed to minimize energy transmission costs in distribution grids by controlling electric vehicle charging and electric water heater activation. The supervision system takes into account the production of renewable energy within the grid, the load consumption and other parameters including user’s constraints. A genetic algorithm is used to optimize the supervisor’s parameters. Finally, the system’s performance is confirmed through simulation by testing it on real distribution grid data.

Known for being a reliable alternative, microgrids have been widely deployed recently in power distribution field in order to guarantee a constant power supply especially in isolated zones. Moreover, microgrids have increasingly known the penetration of renewable energy as environmentally friendly energy sources. However, the intermittency of these sources oblige specialists to think about tools allowing determining their potential, over a predetermined time interval, in order to ensure energy security. For this reason, renewable energy forecasting is crucial. Thus, in this paper, a feed forward back-propagation neural network is used to forecast next 24 hours photovoltaic (PV) power of one of the catholic university buildings "îlot RIZOMM". The accuracy of the model built is evaluated with some performance metrics. Thereafter, prediction results of PV power are compared to those provided by SteadySat, an industrial solution developed by the company SteadySun. It is shown that the prediction Mean Absolute Errors (MAEs) of the model are of 3.05% in a clear sky day, 4.95% in a cloudy day and 5.98% in a partly cloudy day.

Les réseaux de distribution d'électricité sont depuis plusieurs années au cœur des évolutions énergétiques et environnementales qui touchent nos sociétés. Les clients délaissent progressivement les chauffe-eaux instantanés à gaz pour les ballons d’eau chaude électriques, ainsi que les voitures à carburants fossiles pour les véhicules électriques. D’autre part, la production d’énergie renouvelable commence à constituer une part importante du mix énergétique mais ne concorde pas nécessairement avec la consommation électrique des usagers. Elle peut donc être refoulée vers le reste du réseau de distribution, voire le réseau de transport. Toutes ces évolutions doivent être maitrisées pour éviter un renforcement conséquent des infrastructures du réseau, et limiter les répercussions économiques et environnementales de la transition énergétique. Dans ce sens, ces travaux de thèse portent sur la supervision énergétique des réseaux de distribution HTA, en particulier de type rural et péri-urbain, afin d’y encourager la consommation locale des énergies renouvelables et d’optimiser le coût d’acheminement d’énergie pour les gestionnaires du réseau. Les algorithmes développés visent principalement à piloter les charges tels que les véhicules électriques et les ballons d’eau chaude sanitaire raccordés au réseau de distribution. Le système de supervision agit au niveau d’un poste source HTB/HTA et utilise les mesures ainsi que les prévisions de production et de consommation d'énergie afin de piloter les différentes charges disponibles. La stratégie proposée prend aussi en compte les contraintes électriques qui peuvent apparaitre sur le réseau à travers une approche de co-simulation entre les logiciels Matlab/Simulink, où la supervision est implémentée, et DIgSILENT PowerFactory, où le réseau de distribution est finement modélisé. Les consignes de supervision sont donc testées et adaptées en temps-réel avant d'être envoyées aux différents utilisateurs. Enfin, le pilotage des producteurs décentralisés est aussi utilisé comme levier supplémentaire pour résoudre une partie des contraintes électriques.