Bicycle rental systems, which allows a person to use a bike at any time of day is a complex behavioral system. This behavior can be influenced by different factors like the time of day or the origin. The study of different policies to increase the availability of the service allows analyzing different situations and their consequences for the load of the system. This paper proposes a comparison, through an agent-based model, of different policies to balance the occupancy of BiciMAD bike rental service stations and therefore to the availability of the service. Routes for users are obtained based on the probability of the real trips. The stations will act as passive and reactive agents. In addition, the distribution of the stations throughout the geography of Madrid will be analyzed, obtaining areas in need of policies to regulate their load. The result is a 32.2% improvement in the number of stations with low occupancy and 44.7% in heavily loaded stations. An increase in the number of trips made of 23.25% is also achieved. The simulations show saturation in the stations of the area adjoining Paseo de la Castellana and a lack of bikes in more distant stations.