Prediction of equivalent salt deposit density of contaminated glass plates using artificial neural networks

Abstract

Contaminants are deposited on the outdoor insulator surface due to environmental conditions. Conductivity or Equivalent Salt Deposit Density (ESDD) normally expresses this contamination on the insulator surface. In the laboratory, NaCl, tap water and distilled water are used for measuring conductivity and ESDD. In addition, different sizes of glass plates are used as an insulating medium. A conductivity-measuring instrument (Cond 300i) is used to measure the conductivity of the salt-solution. Based on the experimental data, the relationship between the different variables (temperature, salinity, salt, type of water, plate size and sigma) and the ESDD are modeled using artificial neural networks. The developed model showed a good predictive success with R 2 value above 0.98. This value indicates high accuracy for both model development and the model generalization capability. The meteorological variables (temperature, salinity, salt, type of water, plate size, etc.) with the greatest influence on ESDD are also identified using the weight partitioning method. It is found that glass plate size is the variable that has the greatest effect on the prediction of the ESDD since it has a contribution of 47%. The volume conductivity at different degrees had a contribution between 12.38% and 12.87%, while the type of water, the salt quantity, the salinity, and the temperature used had a contribution percentage of 7.92, 7.92, 7.43, and 4.46, respectively. The investigation indicated that the ANN models are well-suited for predicting the contamination level to prevent flashover on the insulator surface and for analyzing the contribution of the different factors affecting this contamination level that are represented either by the ESDD or the conductivity. Additionally, the ANN models can be extended for other applications in which nonlinear relationships are observed.