Classification of disturbance records in power stations based on fuzzy reasoning

Abstract

Nowadays, it is a common practice in power generation utilities to monitor the generation units using Digital Fault Recorders (DFRs). In general, the disturbance records are stored at the utility central office or control center, leading to a substantial amount of data that in practice is not analysed in its totality. This paper describes a methodology to deal with this problem by proposing a fuzzy classification system. From the DFR phasor records, currents and voltages sampled signal are extracted. The data is processed in order to calculate some meaningful features that are applied to a fuzzy inference system. The fuzzyfied input variables are processed by fuzzy rules which emulate the engineers reasoning at the control center. The output of the fuzzy system indicate which kind of disturbance occurred and what is its degree of pertinence. The proposed methodology enables an automated pre-classification of the DFR data helping the engineers by focusing their attention to the most relevant occurrences. Related studies show that approximately 95% of the disturbance records can be automatically archived because they result from normal operational procedures. The results obtained by using real disturbance records show that the proposed scheme is able to correctly classify the occurrences and also to generalize the result from situations not directly represented in the rule set.