The smart grid uses data communications techniques to gather and transfer information for scheduling and decision making so that the electric power can be used more efficiently and economically. To reduce the cost of data communications in smart grid, in this paper, we assume that cognitive radio technique is used to transmit the meter data from smart meters to the data aggregator unit (DAU). Smart meters are deployed in the houses representing consumption nodes to measure and report power demand. Also, smart meters are used by generators of the community renewable power generation facility (CRPGF) which is one of the distributed energy resources (DERs), to collect and estimate renewable power production capacity. However, the transmission of meter data must be performed within a limited period of time (i.e., deadline). By using absorbing Markov chain, we analyze the average packet loss probability of meter data, and study the impact of packet loss on the power supply cost optimization made by the meter data management system (MDMS).