Random geographical networks are realistic models for wireless sensor networks which are used in many applications. Achieving average consensus is very important in sensor networks and the faster the consensus is, the durable the sensors’ life, and thus, the better the performance of the network. In this paper we compared the performance of a number of linear consensus algorithms with application to distributed averaging in random geographical networks. Interestingly, the simplest algorithm – where only the degree of receiving nodes is needed for the averaging – had the best performance in terms of the consensus time. Furthermore, we proved that the network has guaranteed convergence with this simple algorithm.