基于RBF网络的覆冰绝缘子闪络电压预测模型

Flashover Voltage Forecasting Mode of Iced Insulator Based on RBF Network

绝缘子覆冰是威胁电网安全的主要因素之一。在对LXZP-160绝缘子试验研究的基础上,提出了一种基于径向基神经网络(RBF)网络的覆冰绝缘子闪络电压的预测模型。预测模型以绝缘子串长、污秽度、气压、覆冰水电导率等因素为输入变量,以覆冰绝缘子的最低闪络电压为输出变量,网络隐含层单元个数和中心向量采用正交最小二乘法(OLS)算法确定,从隐层到输出层的权值采用伪逆法确定。预测的覆冰绝缘子闪络电压平均误差<1%,优于传统的BP网络,且与数据具有良好的一致性。试验和理论分析表明,该模型能反映覆冰绝缘子闪络电压与绝缘子串长、污秽度、气压、覆冰水电导率等因素间的非线性关系,这对于我国预防冰灾具有一定的参考价值。

The icing insulator is one of key problems for the security of power transmission line. Consequently, based on the test results on XZP-300 insulator in the large artifical climate, a flashover voltage forecasting model based on a radial basis function network is put forward for UHV DC insulators under complex Circumstances. The inputs of the model are environment conditions such as insulator string length ,atmosphere pressure, pollution degree and iceing water conductivity, the output of the model is fashover volatage of iced DC insulator. The orthogonal least squares （OLS） algorithm is used to select the right hide layer neurons and center vectors,and the weight from hide layer neurons to output layer neurons are determined by the pseud inverse method. The average percentage error of forcasting flashove voltage does not exceed 1 % ,which is in accordance with the test result and better than BP network,it is verified that the model is feasible as a new way to forecast the flashover voltage of icing insulator under complex circumstance conditions, which is of great value for providing technical references to the external insulation design and anti icing and de-icing in China.