Neighbour discovery is a process that new devices gather the information from their neighbours when they initially join in a wireless sensor network. During this process, energy utilisation efficiency serves as an essential evaluated factor to concern. This paper proposes a novel asynchronous multi–channel neighbour discovery method for the energy–sensitive time division multiple access (TDMA) wireless sensor network. Two aspects are designed to enhance the efficiency of the energy utilisation, including minimising the listening time of the first frame, which is formulated as a linear programming problem and solved using the CPLEX tool, as well as predicting the time slot and position of the beacon frame based on the attained information from the frames and the process dormancy when a beacon frame is impossible to come out. Our simulation results show that the proposed methods effectively reduce the energy consumption during neighbour discovery process in a TDMA–based wireless sensor network.