In order to develop reliable on-site partial discharge (PD) pattern recognition algorithm, the fuzzy inference-based polynomial network pattern classifier (PNC), one of the fuzzy-neural networks, was investigated. This algorithm was designed and tested using PD data measured from laboratory defect models. Considering the on-site situation in which it is difficult to obtain voltage phases in PRPDA (phase resolved partial discharge analysis), the measured PD data were artificially changed with shifted voltage phases for the test of the proposed algorithm. In the viewpoint of linguistic analysis, the proposed classifier was expressed as a collection of "if-then" fuzzy rules. Its parameters, such as the learning rate, momentum coefficient and fuzzification coefficient, were optimized by means of particle swarm optimization (PSO). The proposed algorithm showed an excellent result, up to a 100% recognition rate for the measured PD data.