Innovative applications of diagnosis, forecasting, pattern recognition and knowledge discovery in power systems

We present our experiences in five power systems domains where we applied diverse knowledge discovery - data mining techniques. The first domain is about electric generator diagnosis. The second one is related to flashover forecasting in high-voltage insulators. The third case is about obtaining expert knowledge, applying data mining techniques to hydroelectric and thermoelectric utilities databases. The next case approaches a pattern recognition problem to detect potential electric illicit users. The last case presents a fossil fuel power plant diagnosis system based on temporal probabilistic networks. We outline successful and bad practices, our contributions, and comment about possible solutions for future work that we think it has to be done to maximize the usefulness of knowledge discovery in the power industry.