基于支持向量机和DGA的变压器状态评估方法

Transformer Condition Assessment Based on Support Vector Machine and DGA

电力变压器老化、故障机理复杂,具有不确定性,难以进行准确的状态评估,故提出了一种基于支持向量机的二叉树多级分类器变压器状态评估方法,该模型以变压器油中溶解气体的含气量和产气速率为评价指标,结合《电力设备预防性试验规程》和《变压器油中溶解气体分析和判断导则》制定了半梯形百分制评分模型对选定的评价指标进行评分;将变压器状态分为良好、一般、注意、较差4种状态,利用从变压器历史试验数据库中归纳整理的样本分别对三级支持向量机分类器进行训练,经过训练的分类器能够正确判断出变压器所处的状态。实例分析结果表明该方法的有效性和实用性。

Aiming at the problem that power transformer ageing, fault mechanism are complex and their conditions are difficult to evaluate accurately, this paper presents a method for evaluating the condition of the oil-filled transformer based on support vector machine and data of dissolved gasses aralysis（DGA）. This model takes the gas content and gas production rate in transformer oil dissolved gas as the evaluation indicators. Combined with《Preventive test electrical equipment order》and 《Transformer oil dissolved gas analysis and judgment guide》semi-trapezoid percentile score model is developed for the evaluation of selected indicators. With this model the transformer condition is into four grades, fine, general, pay attention to and relatively poor. the support vector machine （ SVM ） multi-classifer is trained by samples classfied from the historical experiment database of transformers. The trained SVM multi-classifier can correctly determine the current state of a transformer. The results of practical case verify the accuracy and the practicability of the proposed idea.