This thesis is focused on the problem associated with non-technical losses of electrical energy that are product of fraudulent connections. As a result of the research, a novel methodology is presented for the detection of users who are fraudulently connected to the electricity distribution network. The identification of users with anomalous behaviors is carried out through the use of a system based on computational intelligence techniques. The proposed intelligent system performs the analysis of the energy consumption behavior of the users through three stages. The first is a hybrid cluster between self-organizing maps and genetic algorithms that allows grouping users with similar consumption curves. The second models the users consumption profiles using ARMA/ARIMA models and intelligently corrects them through the use of neural networks, this stage allows predicting the future consumption of customers. The final stage is a classifier based on random forest, which receives the outputs of the previous stages and a set of characterization variables to determine if a user is fraudulent or not. The system was applied on a real case study and the results obtained were compared with proposals of other authors, as well as with the current detection process that is used on the users of the case study. It was found that the proposed system allows to obtain satisfactory results, placing itself in a good position within the reviewed works and significantly exceeding the process currently used on the users of the case study.