The growing boom in smart grids and home automation makes possible to obtain information of household energy consumption. In this work, we study if entropy is a good mechanism to detect anomalies in household energy consumption traces. We propose an entropy algorithm based on windowing the temporal series of energy consumption. We select a trace with a duration of 3 months from the REFIT project household energy consumption data set, available open access. Entropy can adapt to changes in consumption in this trace, by learning and forgetting patterns dynamically. Although entropy is a promising technique and it has many advantages, as the traces in this data set are not sufficiently labeled to check the correct functioning of the algorithms, we propose to further validate the results using synthetic traces.