Study of GA Assisted CSM Models Using Optimally Located Point Charges

The charge simulation method (CSM) due to its favorable characteristics is very commonly used technique for electric field analysis in high voltage engineering. In the conventional CSM location of these fictitious charges are predetermined by the programmer. In the present work charges are located optimally to increase accuracy. CSM models being case specific even while optimizing the model, the optimization parameters chosen play an important role. Here in the present work the models of high voltage sphere above the ground plane with and without another sphere below it are discussed. The study is done with optimally located symmetrically placed set of point charges. The infinite plane is simulated using image sphere(s). The paper compares the accuracies of the optimized CSM models and is an effort to evolve guide lines for CSM modeling which is a `user experience specific' field computation method. The genetic algorithm (GA) is used as the tool for optimization. Numerical experiments on the simulated models indicate that, even with optimized charge locations errors in simulating a ground potential (or lower potential) sphere in the vicinity of a high voltage sphere is always high. Also, if higher is the potential of the second sphere, lower is the CSM error. On reaching this limiting accuracy the options to improve the models accuracy further lie in selectively freeing the charges in optimization process (increasing the degree of freedom in optimization)