Effectiveness of Partition and Graph Theoretic Clustering Algorithms for Multiple Source Partial Discharge Pattern Classification Using Probabilistic Neural Network and Its Adaptive Version: A Critique Based on Experimental Studies

Table 5

Comparison on the role of variance parameter in classifying multiple PD sources.

Input feature vector		APNN; (without clustering)	OPNN with LVQ1	OPNN with LVQ2	OPNN with LVQ3
Φ-𝑞max-𝑛 (10°)	4 input types	𝜎^′1=0.097	𝜎^′1=0.018	𝜎^′1=0.019	𝜎^′1=0.033
		𝜎^′2=0.195	𝜎^′2=0.025	𝜎^′2=0.021	𝜎^′2=0.040
		𝜎3=0.058	𝜎3=0.070	𝜎3=0.073	𝜎3=0.062
		𝜎4=0.079	𝜎4=0.020	𝜎4=0.025	𝜎4=0.016
	5 input types	𝜎^′1=0.171	𝜎^′1=0.024	𝜎^′1=0.037	𝜎^′1=0.031
		𝜎^′2=0.241	𝜎^′2=0.025	𝜎^′2=0.008	𝜎^′2=0.008
		𝜎3=0.067	𝜎3=0.079	𝜎3=0.038	𝜎3=0.040
		𝜎4=0.078	𝜎4=0.032	𝜎4=0.028	𝜎4=0.029
		𝜎5=0.173	𝜎5=0.016	𝜎5=0.004	𝜎5=0.003
Φ-𝑞min-𝑛 (10°)	4 input types	𝜎^′1=0.138	𝜎^′1=0.044	𝜎^′1=0.038	𝜎^′1=0.054
		𝜎^′2=0.206	𝜎^′2=0.015	𝜎^′2=0.021	𝜎^′2=0.009
		𝜎3=0.071	𝜎3=0.035	𝜎3=0.043	𝜎3=0.028
		𝜎4=0.075	𝜎4=0.011	𝜎4=0.019	𝜎4=0.007
	5 input types	𝜎^′1=0.172	𝜎^′1=0.032	𝜎^′1=0.037	𝜎^′1=0.032
		𝜎^′2=0.258	𝜎^′2=0.006	𝜎^′2=0.007	𝜎^′2=0.007
		𝜎3=0.009	𝜎3=0.037	𝜎3=0.0307	𝜎3=0.039
		𝜎4=0.094	𝜎4=0.027	𝜎4=0.0277	𝜎4=0.028
		𝜎5=0.141	𝜎5=0.003	𝜎5=0.004	𝜎5=0.003