Recent advances in Micro-Electro-Mechanical Systems (MEMS) and low-power short-range radios have enabled rapid development of wireless sensor networks. Future sensor networks are anticipated to include hundreds or thousands of these devices in many applications, such as capturing multimedia content for surveillance, structural health monitoring, tracking of accidental chemical leaks, machine failures, earthquakes and intrusion detection. With the increase of sensor applications, a number of challenging problems related to the network protocol design has emerged - the most important ones relating to energy efficiency and lifetime maximisation. Techniques devised for sensor networks should deal with a large number of sensors distributed in the field. Wireless sensor nodes are deployed with limited energy reserves, so the networks should operate with minimum energy overhead. In fact, the network should take into account not only individual node's energy efficiency but also consider the global picture, because surviving nodes' energy reserves in a failed network are wasted energy. This thesis examines a node organisation technique to deal with the above challenges. The focus is on improving network lifetime via organising the nodes in a distributed and energy efficient manner. The main goal is lowering wasted energy via energy balancing and exploiting node redundancy in case of node failure. In particular, this thesis proposes Energy Balanced Clustering (EBC) method for node self-organisation where network tasks (such as data aggregation and data forwarding) are shifted to high-energy neighbours to reduce the energy consumption of low energy nodes. After showing how to extend network lifetime by energy balanced node organisation, the effect of redundant node deployments on network lifetime is addressed. Redundant nodes consume energy by performing unnecessary tasks so a method called Self-Calculated Redundancy Check (SCRC) is proposed to deactivate redundant nodes. A deactivated redundant node can be used as a replacement for a failed node. The Asynchronous Failed Sensor node Detection (AFSD) proposed in this thesis uses the data packets exchanged between neighbours to identify failed neighbours. To restore coverage for network holes caused by failed nodes, policies are given for re-activating redundant nodes. Detailed analytical analysis and simulation of the proposed methods demonstrate that by taking into account energy balancing, eliminating redundant tasks and replacing failed nodes sensor network lifetime can significantly be improved.