RBF神经网络与NSGA-II混合算法用于±1 100kV穿墙套管3维电场模拟及内屏蔽结构优化

Three-dimensional Electric Field Simulation and Inner Shielding Structure Optimization of ±1 100 kV Wall Bushing with RBF Neural Network and NSGA-II Algorithm

±1 100 kV特高压穿墙套管是直流输电系统中的重要设备,但目前对其3维静电场模拟和内屏蔽层结构智能优化鲜有报道。鉴于此,以特高压穿墙套管及其内屏蔽结构为研究对象,建立了3维有限元模型进行电场模拟。提出了应用径向基函数(RBF)神经网络与NSGA-II混合算法对套管内屏蔽结构进行多目标优化,并运用经典显示函数验证了该算法的有效性。在此基础上,建立了穿墙套管内屏蔽结构的多目标优化数学模型,结合RBF神经网络与NSGA-II混合算法对内屏蔽结构进行了优化设计,使套管内屏蔽各关键位置处电场强度(简称场强)均满足控制要求。研究表明:与自由网格划分相比,体旋转扫掠网格划分可使有限元模型生成的节点数量降低58.2%;墙体和均压环对套管复合外套有较好的屏蔽作用,且高场强区主要集中在内屏蔽表面,优化后最高场强降低14.5%。3维电场模拟结果可为穿墙套管的设计、制造和运行提供数据和理论依据,且所提算法能较好地解决大场域、多介质复杂模型结构优化耗时较多的问题。

The 3-D electrostatic field simulation and the intelligent optimization of inner shielding structure of ±1 100 kV UHVDC wall bushing are seldom reported. Thus, we established a 3D model of wall bushing to simulate the bushing and electric field of its inner shielding structure using the finite element method. We also proposed the multi-objective optimization of the inner shielding structure on the basis of the RBF neural network and NSGA-II hybrid algorithm, which were verified through some classical functions. Moreover, we established the mathematical model of the multi-objective optimization, and used it to optimize a wall bushing＇s inner shielding structure, where the electric field met the control requirements. The results show that, compared with the free meshing, the rotational sweeping meshing approach can effectively reduce the number of meshing nodes by 58.2%. The wall and grading rings have good shielding effects on the composite jackets of bushing. The optimization reduces the highest field intensity by 14.5%, and the intensive parts of the electric field are mainly on the surface of the shielding. The simulative results provide certain data and theoretical basis for the design, manufacture, and operation of wall bushing. Moreover, the proposed algorithm effectively solves the time-consuming problem of optimizing the complex large multi-dielectric models.