With the evolution of wireless communication networks, the access to the internet became much faster for mobile devices. This enables new distribution strategies like offloading of complex tasks to a remote server. Nevertheless energy resources on mobile devices is limited by the battery capacity. In order to decide if calculations should be offloaded to a server ot executed on a mobile device, a prediction model for energy consumption is needed. This thesis proposes energy prediction models for wireless communication on mobile battery-powered devices. The prediction models cover the network types LTE and WiFi. The energy consumptions predicted by combining several linear regressions alongside statistical prediction intervals. The regression is based on energy measurements on a mobile device of 288 client-server communications. The prediction results can compared to other data to decide on the usage of the wireless network. Evaluating the prediction model with an additional set of measurements shows a relative error of 1.21% in case of LTE and 26.73% in case of WiFi.