Optimization Of Disc-type Spacer Electric Field Structure By Cloud-intelligent Algorithm

The traditional wavelet neural network all seeking a set of parameters to makethe energy function minimum for the suitable sample set. Generally it is throughdecreasing gradient iterative to realize the iterative optimization. But this kind ofoptimization method is easy to fall into local minimum. The combination betweencloud model and the traditional particle swarm algorithm or the traditional artificialfish-swarm, can avoid the defects of gradient algorithm.The paper based on the theoretical study of cloud model structure, improved therule generator of traditional cloud model, and formed a new nesting cloud rulegenerator. It used the output of the last Y-cloud to nest the input of the next X-cloud,which made the correction between the qualitative and quantitative more close andreduced the error of inverse cloud stage.The improvement of particle swarm algorithm is using the nested rule cloud toadaptive adjust the velocity formula in the algorithm process and the inertia factor,convergence factor, social and cognitive section in the position updating formula byitself. It realized the goal of dynamic search. And in the detailed algorithm process, italso adaptive adjusted the individual fitness and the population current position byitself. The paper combined each other to nested optimization, and realized theimprovement of the algorithm.The principle of improving the artificial fish swarm algorithm is similar to theparticle swarm algorithm, using the nested cloud qualitative rules to adaptive adjustthe forage, cluster and rear-end behavior of traditional algorithm by itself, based onthe visual field and the update of bulletin board.In the improved structure of wavelet network, input and output the data to cloudand inverse, and used the improved intelligent algorithm to optimize the weight andthreshold. Combined with the algorithm improvement above, the article takes the disc-typeinsulator of high voltage electrical apparatus for example. Through the normalizationof the collected fuzzy data, input it into the cloud layer to do cloud processing, theninput into the new network composited by the weights and thresholds of waveletnetwork optimized by the intelligent group algorithm which adaptive adjusted byitself through the nesting rule generator to do data processing. Then inverse cloudprocessing the random output data, finally output the clear structure optimizationdata.Through the simulation research of the electric field structure data of thedisc-type insulator, the two kinds of improved algorithm not only improved thenetwork global search ability, solved the chronic and stubborn disease that it is easy tofall into local optimum, but also improved the search accuracy of algorithm and theobject identification ability. Through comparing the two kinds of algorithm of searchaccuracy, we can conclude that the artificial fish swarm algorithm is relativelysuperiority to the particle swarm algorithm.It provides a new method for the modeling of nonlinear system problem, at thesame time provide useful reference for the complex nonlinear group system in theidentification ability. But when deal with a problem, there will appear the newproblem that the original advantage turn into the disadvantage. Then we need to studyhow to make the complementary advantages harmony with each other.