摘要
为了提前预防箱式变压器高压套管过热和爆炸等事故发生,提高变压器运行的安全稳定性,提出一种基于变分模态分解与改进门控循环单元神经网络的变压器高压套管温度预测方法。首先,运用变分模态分解将箱式变压器高压套管温度分解为具有不同特征的子序列分量,减少不同趋势信息对预测精确度的影响;然后,提出改进门控循环单元神经网络MGRU,针对分解后各子序列分别建立基于MGRU的时间序列预测模型;最后,叠加各子序列预测结果,得到高压套管温度最终预测值。结合某小区箱式变压器套管在线监测平台实际算例,仿真结果表明,相较于传统预测算法,所提方法在单步和多步预测中都能更好地预测箱式变压器高压套管温度,具备更优良的预测性能和更好的泛化能力。
In order to prevent accidents in time such as overheat and explosion of the transformer's high-voltage bushing and improve the safety and stability of box-type transformer operation,variational modal decomposition(VMD)and improved gated recurrent unit neural network(MGRU)are introduced into the prediction method of transformer high-voltage bushing temperature.Firstly,the transformer high-voltage bushing temperature was decomposed into a set of different characteristics sub-sequence components by using VMD to reduce the influence of different trend information on the prediction accuracy.Then,MGRU neural network was proposed,and forecasting models based on the MGRU neural network were constructed respectively for each subsequence.Finally,the prediction results of each sub-sequence are superimposed to obtain the final predicted value of the transformer high-voltage bushing temperature.Combined with an actual example of an online monitoring platform for transformer high-voltage bushing,the simulation results show that the proposed method can better predict the temperature of transformer high-voltage bushing in single-step and multi-step predictions compared with traditional prediction algorithms,and the proposed method has higher prediction accuracy and stronger generalization ability.
作者
赵洪山
王奎
王震
刘秉聪
彭轶灏
ZHAO Hong-shan;WANG Kui;WANG Zhen;LIU Bing-cong;PENG Yi-hao(School of Electrical and Electronic Engineering,North China Electric Power University,Baoding 071003,China)
出处
《电机与控制学报》
EI
CSCD
北大核心
2021年第8期18-28,共11页
Electric Machines and Control
基金
国家自然科学基金(51807063)。
关键词
箱式变压器高压套管
温度预测
变分模态分解
改进门控循环单位神经网络
在线监测平台
多步预测
box-type transformer high-voltage bushing
temperature prediction
variational modal decomposition
improved gated recurrent unit neural network
online monitoring platform
multi-step prediction