Power Cable Fault Recognition Based On IPSO-SVM Algorithm

In recent years, with the development of our country’s modernization and the progress ofscience and technology of the power industry, electric power cables are getting more andmore widely used in the power system. With the long-term utilization, various safe leakswould exist with the power cable. Therefore, how to detect and recognize the faults accuratelyand quickly is a very important and difficult task for the researchers.To solve the parameter optimization problem of fault types identification model, basedon the particle swarm optimization algorithm, nonlinear inertial particle swarm optimizationalgorithm is proposed in this paper. The algorithm combines the convergence factor and theinertia factor particle swarm, and uses the nonlinear inertia factor,which better improves theconvergence of the traditional particle swarm.To overcome the shortcomings of the common intelligent approaches that require a largenumber of samples to construct the model, this paper uses support vector machine,which isbased on the structural risk minimization principle, as an original model of the fault typesidentification. But for different penalty factors and kernel width parameters, which also affectthe classification results and generalization ability of the classification system. Aimed at thisproblem, the paper establishes a support vector machine model of the optimized kernelparameters and optimized penalty factors.In order to optimize the parameters of support vector machine, this paper uses thenonlinear inertia particle swarm algorithm to optimize the penalty factors and kernel widthparameters, based on which,then establishes the support vector machine model of the faulttypes identification.Finally, the proposed algorithm and models are applied for the recognition of cablenormal and fault conditions (phase short circuit, three-phase short circuit). Compared with the traditional support vector machine model, the presented algorithm in this paper has a highrecognition accuracy.