The optimization of multi-parameter insulation diagnosis

On the basis of statistical theory, a method to quantitatively determine the reliability of multi-parameter diagnosis and to optimize the algorithm of multi-parameter diagnosis is put forward. This method is based on the statistical law, and concerns correlation between the different parameters. Moreover, as an example, the stator bar of generator (300 MW, 18 kV) is studied. 30 samples were selected from different parts of the practical stator bar of generators, and the parameters of dielectric loss, partial discharge and remaining breakdown voltage (BDV, 50 Hz) of each sample were measured. This method is applied to the estimation BDV of generator bars and the optimized multi-parameter diagnosis algorithm is determined on the basis of actual data. Sk+, tanδ and ΔC are selected as the optimal parameter group by this method. Comparing with experimental data, it shows that the quantity of parameters is not the more the better. To choose appropriate parameters for assessing insulation condition is important. In order to choose appropriate parameters, two following principles are suggested. 1) It is efficient to select the parameters which have significant correlation with estimation object, as well as to obviate those of limited correlation. 2) Correlation between parameters should be taken into account. Generally speaking, the degree of correlation between parameters is the lower the better.