Remote metering is a key task in smart grid to collect the power load information for the pricing in power market. A wireless communication infrastructure is assumed for the smart meter network. The dynamics of the power market are assumed to be a Markov decision process (MDP) by modeling the power load evolution as a two-state Markov chain. The fundamental elements of the MDP are discussed and the optimal strategy is obtained from dynamic programming. Due to the curse of dimensions, the myopic strategy is proposed to significantly simplify the algorithm. The validity of the proposed scheduling algorithm is demonstrated using numerical simulations. For example, for price insensitive power users, the average price and power generation errors are reduced by 60% and 40% using the myopic approach, compared with the simple round robin schedule, in certain circumstances.