A scalable PQ event identification system

A scalable event identification system for power quality events is proposed. Unlike ANN-based approaches where the system is not scalable and not "debug-able" without retraining, the proposed approach is particularly advantageous compared to those of ANN's since it is scalable, debug-able and easily modified. This approach is adopted from artificial intelligence's rule-based approach and attempts to mimic power engineers thought process in identifying PQ events. This paper describes prerequisites in constructing such a scalable system. Examples of rules to identify power quality event are also presented. The prototype of the system is built and tested using 770 field-measured voltage waveforms which covers ten types of PQ events. The accuracy rate is nearly 95% with less than 6% of rejection rate. Potential applications of the proposed system in PQ community are also described.