The design of Building Intelligence Energy Management Systems (BIEMS) in the air conditioning sector is a complicated multi-objective problem for large buildings with big engine rooms and scattered inputs that it must provide simultaneously maintaining system stability, adapting to time-varying conditions, reducing energy dissipation and maintaining proper air quality. To achieve these at times conflicting objectives, using a distributed control approach is appropriate for building energy management of these buildings. In this paper, a Cooperative Multi Agent-Building Management System (CMA-BMS) equipped with fuzzy controllers and virtual sensors is proposed, in which each of the agents controls one of the main components of the cooling engine room in the HVAC's primary circuit and contrary to many previous works, the topology of them is horizontal, and each agent operates independently and in parallel, which leads to faster performance, easier error detection, and more extensibility. For modeling, the virtual sensors implemented with Radial Base Function (RBF) neural networks are utilized, which also have the role of actual sensor backup. To evaluate the project, two BEMSs are designed and tested with a conventional centralized and the proposed strategies, where simulation results indicate that the proposed design is less processor and memory intensive, and the fuzzy controllers show a better performance in terms of energy consumption versus the non-fuzzy controllers