Section: New Results

Applications to Energy

Participant : Giovanni Neglia.

Smart grids

Balancing energy demand and production is becoming a more and more challenging task for energy utilities because of the larger penetration of renewable energies, more difficult to predict and control. While the traditional solution is to dynamically adapt energy production to follow the time-varying demand, a new trend is to drive the demand itself. We have first considered the direct control of inelastic home appliances, whose energy consumption cannot be shaped, but simply deferred. Our solution does not suppose any particular intelligence at the appliances, the actuators are rather smart plugs, simple devices with communication capabilities that can be inserted between appliances' plugs and power sockets and are able to interrupt/reactivate power flow. During previous years we have considered both closed-loop and open-loop control of such devices in order to satisfy a probabilistic bound on the aggregated power consumption. Recently, G. Neglia, together with L. Giarré (Univ. di Modena e Reggio Emilia, Italy), I. Tinnirello and G. Di Bella (Univ. di Palermo, Italy) have considered a mixed approach [16]. They have been able to quantify the trade-off between the amount of controlled power and delays experienced by the users to evaluate to which scale this solution should be deployed.

We have also looked at Demand-Response (DR) programs, whereby users of an electricity network are encouraged by economic incentives to re-arrange their consumption in order to reduce production costs. Several recent works proposed DR mechanisms relying on a macroscopic description of the population that does not model individual choices of users. In [8], G. Neglia, together with A. Benegiamo (EURECOM /Inria) and P. Loiseau (EURECOM ) has shown that these macroscopic models hide important assumptions that can jeopardize the mechanisms' implementation (such as the ability to make personalized offers and to perfectly estimate the demand that is moved from a timeslot to another). Then, starting from a microscopic description that explicitly models each user's decision, they have introduced new DR mechanisms with various assumptions on the provider's capabilities. Contrarily to previous studies, they have found that 1) the resulting optimization problems are complex and can be solved numerically only through heuristics, 2) the savings from DR mechanisms are significantly lower than those suggested by previous studies.