This paper proposes a bi-level optimal integration scheme for buildings' space heating loads in the integrated community energy system (ICES). The optimal integration scheme consists of an efficient energy management method and a heating pricing method for the ICES with buildings. At the upper level, the ICES operator optimizes the schedules of energy generation and supply, and the heating prices to buildings to maximize its profit. At the lower level, consumers in buildings optimize the water flow rates in the radiators to minimize their heating costs. The thermal dynamics of the building with controllable indoor radiators is modeled using the model of Resistor-Capacitor thermal network. Moreover, the model predictive control (MPC) is integrated with the bi-level optimization to achieve economic and reliable scheduling of the ICES and buildings in the presence of uncertainties. The bi-level MPC optimization is reformulated as an MPC based mixed-integer linear program using the Karush-Kuhn-Tucker optimality conditions and several linearization techniques. Numerical studies show that the bi-level MPC method can obtain a balanced scheduling scheme between the energy costs of consumers in buildings and the ICES operator's profits. The MPC method can ensure higher profits of the ICES operator and simultaneously, lower energy costs of consumers in buildings.