变压器分接开关内酚醛压板的神经网络全局优化设计

Global Optimization of Structure Design for Phenolic Laminate in Transformer Tap Switch Using Neural Network Method

为改善变压器分接开关电场环境,笔者重点研究了变压器分接开关内酚醛压板的电场分布,利用有限元法计算了这种绝缘结构的电场,提出了结构设计的神经网络最优化方法。在神经网络最优化方式中,构建酚醛压板绝缘结构参数——电场分布的多输入—多约束输出神经网络模型,利用最优化方法构建优化算法,真正实现了酚醛压板绝缘结构的全局优化设计。结果表明,该方法对于研究多输入—多约束输出模型的全局优化设计具有普遍的指导意义。

To improve the electric field environment of a transformer tap switch,this study focuses on the electric field distribution on the phenolic laminate in the transformer tap switch.The electric field of this insulation material was calculated based on finite element method.An optimization method of structure design was proposed by using neural network.In this optimization method,a multi-input and multi-constrained output model was establish through artificial neural network for the phenolic laminate insulation structure parameters-insulation electric field distribution.An optimization algorithm and programming languages were employed to construct the optimization method to achieve the global optimization of the phenolic laminate insulation structure.