How to efficiently collect sensory data for supporting energy-efficient operation of buildings has become a great challenge, especially for large-scale networks in buildings. In this paper, a spatio-temporal compression-based optimized clustering scheme is proposed for energy-efficient environmental data collection in buildings. In the scheme, first, according to the establishment of a cluster, an adaptive dynamic cluster head selection method for prolonging the lifetime of sensory nodes in buildings is developed. Meanwhile, to further reduce energy consumption, we construct the dynamic optimal quantity of the cluster heads solution method to achieve minimum energy consumption. Moreover, the spatio-temporal correlations of sensory data are explored in the data sampling and transmission processes, which can decrease the amount of data transmission and further extend the lifetime of the entire data collection network. The proposed scheme provides sensing data with high quality for environmental quality management in buildings and supports energy-efficient building operation. Finally, the simulation results show that the proposed scheme obtains obvious advantages in energy consumption compared with other related schemes, and the lifetime of the proposed scheme is also longer than others when it maintains high reconstruction precision.