基于决策树集成算法的电力变压器状态评估

Power Transformer State Assessment Based on Decision Tree Integration Algorithm

针对变压器故障诊断困难和单个决策树分类精度低的问题,提出了基于决策树集成算法的电力变压器状态评估方法以进行辅助诊断。该系统首先使用多目标二进制编码遗传算法从变压器溶解气体的14种特征中选择有效的分类特征,然后使用这些特征训练一系列基决策树分类器,并使用遗传算法选择集成部分精度较高的决策树构造强分类器提升分类性能,最后使用D-S融合规则融合基分类器的分类结果得到最终的诊断结果。仿真结果表明,该算法不仅提升了决策树的分类性能,且能够提高电力变压器的诊断精度,具有良好的实用性。

Aiming at the difficuhies of h-ansformer fauh diagnosis and low accuracy of single decision tree classification, this paper proposes a transformer state evaluation method based on the decision tree integration algorithm for auxiliary diagnosis. The system first uses the muhi-ohjeetive binary coding genetic algorithm to select effective classification features from 14 features of the dissoved gas in transformer, and then uses these features to train a series basis decision tree classifiers, and uses the genetic algorithm to select the decision tree of higher preeisiun to form a strong classifier to improve the classification perfurmance, and finally the system uses the D-S fusion rules In fuse the basis classifiers＇ resuhs to get the final diagnosis. The simulation resuhs show that the proposed algorithm not only improves the classification performance of the decision tree, but also improves the diagnostic accuracy of power transformer failures and has good praelicability.