It is projected that communication among cognitive machine-to-machine (M2M) in smart grid networks is necessary and vital in forthcoming years. The objective of this paper is to bring out the significance of cognitive medium access control (MAC) protocol focusing on the distinctive characteristics of the devices in M2M and requirements of the smart grid communication. Therefore, developing a cognitive radio system which supports dynamic access to the spectrum available that is recently considered as a novel solution for extending wireless systems. In this paper, MAC layer sensing approaches in cognitive radio networks are investigated by considering both proactive and reactive sensing. To evaluate the performance, analytical modeling and simulation studies are carried out by comparing with state of art methods. As a result, the proposed HCM2M-MACachieves 687.3 kbps of blocking probability, 368.6kbps of dropping probability, 96.4% bandwidth utilization, 67.3% end to end delay and 89.3% of throughput.