The impact of IT on the global energy consumption has frighteningly increased over the last years. One of the reasons for this is the demand for infrastructure to support the increasing number of online (24x7) services and data, followed by the popularisation of practices like Cloud Computing. From the infrastructure point of view, hardware throttling and server consolidation are techniques used to deal with energy efficiency. However, details about the application behavior are not visible from the infrastructure layer, which prevents a more complete energy-efficient treatment. This paper presents an approach for self-optimisation of the energy consumption at the application layer. We rely on Service-Oriented Architectures, since they allow rapid and seamless service composition and eases the application adaptation. The energy efficiency properties of services are defined by means of Quality of Service criteria and a set of event-condition-actions is defined to enable the application to react to environmental changes and optimise its energy consumption. As a proof of concept, we present a prototype for energy-aware self-adaptation in SOA-based applications as well as an example scenario that shows the practical usage of our approach.