Software-defined wireless sensor networks (SDWSNs) are a great step forward to centralize and facilitate the management of low-power networks. However, the continuous sending of local information from the sensor nodes to the controller causes excessive energy loss and increases the network overhead. This dilemma has become one of the hot topics of current research to make SDWSNs more efficient. This paper has proposed an efficient energy predictor for SDWSNs (E²-SDWSN) to improve the performance of the sensor network. The proposed approach reduces the cost of packet delivery and energy consumption using the well-known MapReduce framework with an intelligent selection of the reducers. Using MapReduce in the SDWSN by in-network processing significantly reduces energy consumption and balances traffic. Furthermore, the controller using an energy prediction model predicts the residual energy level of the nodes. Therefore, assuming the position of the nodes is fixed, the need to sequential send local packets is reduced. Reducing the number of local and energy information packets of the network decreases the network traffic. This consequently improves the quality of service and reduces energy consumption. Therefore, the proposed model increases the lifetime of the network. The results also confirm that the proposed E²-SDWSN significantly improves the energy throughput, the packet delivery ratio, and the delay through the sensor network.