MAGRES - Aprendizagem automática para uma utilização efetiva da resposta de grupos de consumidores em edifícios em suporte de integração de produção de origem renovável

MAGRES proposes methods to determine the contextual use of different DR programs according to the context, considering that some programs are faster and that some programs have more reduction capacity, both of them with different uncertainty of response. Dynamic clustering is used to group different consumers and distributed generation units according to the context, to support the definition of groups to be activated in DR events. Dynamic clustering methods adapt the factors that drive the cluster formation to the actual context characteristics. Uniform versus cluster differentiated methods will be addressed as options for the consumers and aggregators. MAGRES defines an efficient approach in the activation of DR programs, by issuing multiple requests to consumers according to the actual response, until the desired total amount is reached. Also, different DR deployers are considered, i.e. entities that activate DR, namely building manager, aggregator, DSO, and TSO. Multiple DR opportunities can be activated to achieve the targeted load reduction. The hierarchical DR activation is based on the actual cost of each DR contract. Since the response depends in a large extent on the remuneration or other incentives or real-time prices, and too high remuneration can make DR costly, MAGRESS will develop game theory based approaches in order to achieve optimized contracts between consumers, distributed generation owners, aggregators, and other deployers. Regarding distributed renewable based generation, MAGRES proposed that those are managed at the building level by the building energy management optimization. The exceeding amount of generation can than be used for the DR deployer convenience, according to the established contracts. MAGRES will conceive, develop, and make available new contextual DR programs in the MAGRES platform, together with an Automated Consumption Management tool. MAGRES platform enables the DR deployer to manage the complete process, from the context identification to the hierarchical DR activation, including the calculation of the due remuneration, the determination of contracted remuneration, and the learning of the consumers behavior in the response to a specific event. While the state of the art in buildings automation give users permission to define rules, which are not always easy for the user to quantify, in MAGRES, decision trees will be developed in sequence of the learning obtained from the users behavior in response to DR events, so automated decisions can be easily defined as a set of rules for the actual implementation of DR in buildings. MAGRES models and methods will be validated using patented MARTINE smart grid real-time simulation.