基于组合预测的变压器故障诊断模型

Transformer Fault Diagnosis Model Based on Combination Prediction

为了快速准确地预测出变压器的故障类型,及时做好维修工作,本文提出了一种基于非线性规划的组合预测模型.首先,利用改进的鲸鱼算法优化BP神经网络建立IWOA-BP预测模型;然后,在IWOA-BP预测模型和梯度提升树的基础上,利用非线性规划与遗传算法相结合的方法确定各算法的权系数,再将各算法的结果加权得出组合模型的最终预测结果.通过实例验证,IWOA-BP预测模型的变压器故障预测效果强于BP神经网络、随机森林等多种预测模型,并且利用IWOA-BP预测模型和梯度提升树建立的组合模型,其预测准确率高于组合中任意一种算法.

In order to accurately predict the type of transformer fault and do a good job of maintenance in time,a combined prediction model based on nonlinear programming genetic algorithm is proposed.In this paper,an IWOA-BP prediction model is established by Optimizing BP neural network with improved whale algorithm,and then a combined prediction model is established based on IWOA-BP and Gradient Boosting Decison Tree prediction model.Finally,the weight coefficients of each algorithm is determined by combining nonlinear programming with genetic algorithm.On this basis,an effective combination method for transformer fault detection is proposed.Through the example verification,the improved whale algorithm has a good optimization effect,and the prediction effect of IWOA-BP prediction model is better than that of BP neural network,random forest and other six prediction models.Finally,the IWOA-BP prediction model and the GDBT are used to build the combination model,and the prediction accuracy of the combination prediction algorithm is higher than that of any combination algorithm.