The increasing penetration of renewable power generators has brought a great challenge to develop an appropriate energy dispatch scheme in a microgrid system. This paper presents a hierarchical energy management scheme by integrating renewable energy forecast results and distributed consensus algorithm. A multiple aggregated prediction algorithm (MAPA) is implemented based on satellite weather forecast data to obtain a short-term local solar radiance forecast curve, which outperforms the multiple linear regression model. A distributed consensus algorithm is then incorporated into the HVAC (heating ventilation air conditioning) units as the adjustable loads in order to dynamically regulate power consumption of each HVAC unit, based on solar power forecast in a day. The scheme aims to alleviate the local supply-demand power mismatch by varying demand response of the HVAC units. Two case studies are performed to demonstrate the feasibility and robustness of the algorithms.