The BATNRJ project, managed by Pyrescom, focuses on improving energy efficiency while preserving comfort in tertiary buildings. To this end, an open and cost-friendly monitoring solution based on instrumentation as well as analysis and control tools is being developed. Solar radiation and indoor temperature being key parameters, the present paper deals with estimating missing data in case of sensor failures. First, solar radiation is interpolated using as a basis the Gaussian or the Cosine function. Mean relative error is about 10%. Then, based on the concept of time series, feedforward artificial neural networks are used to estimate up to the next 24 hours missing data about indoor temperature. We obtained accurate results, especially for failures limited to 3 hours. The mean relative error does not exceed 6%, even in case of long sensor failures.