论文部分内容阅读
针对目前接地网腐蚀预测中的涉及因素较多,提出了基于向量相似度与支持向量回归机(S-SVR)的综合的预测方法应用于变电站接地网腐蚀速率预测模型中.首先将影响接地网腐蚀速率的指标视为1个特征向量并进行无量纲化处理,其次计算各个训练站点的特征向量与实测站点指标向量的相似度;再次,在预测腐蚀速率时,针对传统线性贡献度平均法(LAM)描述非线性存在较大误差的缺陷,提出了先筛选相似度较高训练集再结合支持向量回归机训练模型.经验证,基于向量相似度与支持向量回归机(S-SVR)的综合预测方法预测能力较好.
In view of the many factors involved in the corrosion prediction of the grounding grid, a comprehensive prediction method based on vector similarity and support vector regression (S-SVR) is proposed in the prediction model of the corrosion rate of the grounding grid in the substation.First, The corrosion rate is regarded as one eigenvector and the dimensionless process is performed. Secondly, the similarity between the eigenvector of each training station and the index vector of the measured site is calculated. Thirdly, when the corrosion rate is predicted, LAM) to describe the shortcomings of large nonlinear errors, a training model with high similarity training set and support vector regression machine is firstly selected.Finally, based on the combination of vector similarity and support vector regression (S-SVR) Prediction method is better.