The evolution towards Smart Grids (SGs) represents an important opportunity for the energy industry. It is characterized by the integration of renewable and alternative energy resources into the existing power grids while ensuring a finegrained control for the different measuring points. Therefore, this evolution requires the ability to send a maximum of data over the network in real time while controlling the grid. A Wireless Sensor Network (WSN) deployed across the grid is a potent solution to achieve this task. However, sensor nodes have limited energy and computation resources especially the battery powered ones. For that, reducing transmission is an essential priority in order to increase the lifetime of the network. Data prediction is a widely used, yet effective, solution in literature to accomplish this task. In this paper, we propose a Quality of Service (QoS) aware algorithm based on time series prediction and linear regression for data prediction in WSN. We test our approach in a SG context on real data traces of photo-voltaic cells. Our algorithm takes into consideration the diversity of applications of SGs with different requirements while being energy efficient. Our results show that our proposal provides satisfactory results compared to literature solutions in terms of data reduction percentage, Root Mean Square Error (RMSE) and energy consumption.