Page 137 - LAS CIENCIAS AMBIENTALES EN EL ANTROPOCENO - DIGITAL(1)
P. 137

algoritmo de aprendizaje programado   medibles de toma de decisión, pudiendo
              funcionó correctamente, encontrando   integrar una interacción del usuario con
              una política de acciones óptimas. Es   el sistema, de forma tal, que se pueda
              importante destacar que es la primera   aumentar la sensación de confort de
              aplicación que se ha publicado, según   una manera holística y no solo el confort
              el mejor conocimiento de los autores,   térmico, como se hizo en este trabajo.
              del uso de la API de EnergyPlus, y no
              se tuvieron errores. En cuanto a las
              métricas de evaluación, de consumo de
              energía y de horas de confort, el modelo
              propuesto no logró mejorar con respecto
              a los sistemas de control con los cuales,
              se lo comparó; sin embargo, no se
              exploraron todos los hiper parámetros
              del algoritmo, como así tampoco otras
              alternativas de aprendizaje. Además,
              una ventaja clara de esta clase de control
              es que permitiría ampliar las variables












              SUMMARY




                   he  residential  sector  is  a  large   alternative to avoid these problems   to tackle complex problems such as the
                   consumer of energy, representing   is the automation of the passive   one presented here. At present, models
              T25% of the demand in Argentina,   strategies implemented, thus obtaining   have been developed that allow us to
              which together with the industry and   new mechanisms that, although they   optimally solve highly complex problems.
              transportation sector are the main   consume energy to actuate, have   That is why in this work a reinforcement
              pillars of the demand for primary   greater energy savings. Many control   learning algorithm is implemented,
              and  secondary energy.  In  order  to   systems for these mechanisms have   specifically Q-learning, in such a way
              reduce energy consumption, passive   been developed, however, they have   that optimal policy of the passive and
              air conditioning strategies can be   the disadvantage of not considering the   active systems of a home can be carried
              implemented in home design. However,   particular preferences of the inhabitants   out and that considers the comfort of
              these require active users, which   due to the complexity that this implies.   the inhabitants. The main objective of
              entails higher energy consumption   On the other hand, the development of   this work is to present the advances
              than expected due to laziness or   artificial intelligence has been growing   in the research line that is being
              lack of  knowledge  of  the users. An   in recent decades, making it possible   approached, finding the case study



                                                                                                     128
   132   133   134   135   136   137   138   139   140   141   142