摘要
对油中溶解气体浓度进行精确预测,可为变压器故障预警提供重要依据。现有的油中溶解气体预测模型主要基于单一或少数状态参量,而状态参量变化并非独立过程。利用Apriori算法挖掘变压器油中溶解气体间的关联规则,在此基础上提出了一种基于小波分解和长短期记忆网络的变压器油中溶解气体浓度预测方法。Apriori算法可挖掘变压器油中溶解气体间的关联规则,以此确定预测模型的输入矩阵。通过小波变换提取出参量序列中的低频趋势分量和高频波动分量,运用长短期记忆神经网络在不同分量上分别进行预测,并重构得到各参量的预测结果。算例结果表明,所提方法能更好地追踪油中溶解气体的浓度变化趋势,具有更高的预测精度。
Accurate prediction of dissolved gas concentration in transformer oil can provide an important basis for early warning of transformer failure.Existing prediction models for dissolved gas in oil are mainly based on a single parameter or a few parameters,while the change of state parameters is not an independent process.Apriori algorithm is employed to mine association rules between dissolved gases in transformer oil.Based on this,a method for predicting dissolved gas concentration in transformer oil based on wavelet decomposition and long short-term memory network is proposed.The Apriori algorithm can mine the association rules between dissolved gases in transformer oil to determine the input matrix of the prediction model.The low-frequency trend component and the high-frequency wave component in the parameter sequence are extracted by wavelet transform,and LSTM neural network is used to predict each component separately.Then the prediction results of each parameter are reconstructed.The case studies show that the proposed method can better track the concentration trend of dissolved gases in oil and has higher prediction accuracy.
作者
王兴
荣海娜
王健
张葛祥
WANG Xing;RONG Haina;WANG Jian;ZHANG Gexiang(School of Electrical Engineering,Southwest Jiaotong University,Chengdu 611756,China)
出处
《电工技术》
2020年第9期24-29,33,共7页
Electric Engineering
基金
国家自然科学基金项目(编号61702428,61972324,61672437)
四川省科技计划(重点研发或重大科技专项)项目(编号2018GZDZX0044,2018GZ0086,2018GZ0185)。
关键词
变压器油中溶解气体
关联规则
小波分解
长短期记忆神经网络
浓度预测
dissolved gas in transformer oil
association rules
wavelet decomposition
long short-term memory neural network
concentration prediction
