GSCOP-RUB-GCSP

Séminaire GCSP - Mardi 18 novembre 2014 - 13h15 - C319 - Présentation de Mahendra Pratap SINGH

Mardi 18 novembre 2014 à 13h15 en salle  C319, Bât C, Site Viallet Grenoble INP

Nous aurons le plaisir d'écouter Mahendra Pratap SINGH, doctorant en 2A, sous la direction de Stéphane Ploix. Il nous présentera ses travaux effectués dans le cadre de sa thèse.

Résumé :
Nowadays, buildings account for 40% energy consumption of total available energy resources. Several methods are proposed to avoid such kind of immense consumptions. Different anticipative approaches are proposed for one day energy management that considers one hour as a sampling period. Now the reactive problem begins when the predictions are very far from the reality. An important question arises: why there are differences between predictions and measurements and how we can cope with the situation. Till now it is difficult to define absolutely why there is a disparity between the anticipated plan and current measurement. In order to solve the problem, a reactive optimization technique is proposed which include the cause detection and correction mechanism also. It takes into account shorter time samples and faster dynamics of user behavior as well. A two layer energy management architecture is considered which are anticipative and reactive. In this architecture, anticipative layer is responsible to anticipate the future plan form by using the weather information and past energy consumption history of the occupants. Reactive layer detects the conflicts between anticipative plan and occupants expectations. It also identifies the necessary corrective actions in the anticipative plan, such that, the occupants comfort and total available energy cost constraints are not get violated. The key objective of the reactive layer is to achieve long term goals and to preserve the integrity of previously achieved goals from the higher layer i.e. anticipative layer. Model used for reactive layer will be different from the anticipative layer. It includes faster response for the occupants behavior and physics of the surroundings in short time intervals. For the practical implementation, PREDIS-­‐MHI  platform is considered, located in ENSE3 Grenoble-­‐INP.  This platform is equipped with around 160 sensors to monitor the occupant presence, temperature, CO2 concentration, and is also able to measure all kind of energy consumptions. Within the PREDIS-­‐MHI,  reactive layer will decide, when should specific appliances be turn on or off? Turning on/off appliances will include both continuous and discrete phenomena. This will result the efficient energy management. In this work I will present the a optimization based reactive energy management which will improve the the performance of BEMS.