为了提高油中溶解气体体积分数的预测精度,进而为变压器早期故障预警和制定维修方案提供理论依据,提出了一种基于多分解策略和贝叶斯优化—长短期记忆神经网络的预测模型。首先,采用改进的完全自适应噪声集合经验模态分解对原始油中溶解气体时间序列进行一次分解,得到一系列不同时间尺度下的子序列分量;其次,利用变分模态分解对预测难度最高的高频IMF1分量进行二次分解,进一步降低IMF1分量的非线性和非平稳性;然后采用贝叶斯优化的长短期记忆神经网络对二次分解后得到的所有子序列分量进行时序建模,并进行单步预测和递归多步预测 ...
陈志勇, 杜江
doaj
对变压器油中溶解气体浓度进行预测分析可为其运行状态评估提供重要依据,从而有效掌握设备状态发展趋势,为预警和检修提供参考。鉴于现有算法难以实现长期精确预测,文中首次提出一种基于策略梯度(policy gradient,PG)算法优化的变压器油中溶解气体体积分数预测模型。首先以门控循环单元(gate recurrent unit,GRU)为基础,引入编码器—解码器结构搭建Sequence to Sequence(Seq2Seq)网络模型,并且结合注意力机制和Scheduled Sampling算法 ...
汤健 +5 more
doaj
基于改进箱线图和ISMAHHO-KELM的变压器油中溶解气体体积分数预测
油中溶解气体分析能够有效地揭示变压器内部的运行状况,在评估变压器运行状态和预测潜在故障方面具有重要作用。为此文中提出一种基于改进箱线图和ISMAHHO-KELM的变压器油中溶解气体体积分数预测方法。首先,为提高数据质量,利用基于中位偏差系数的改进箱线图对原始时间序列进行离群值检测和校正;然后采用变分模态分解对校正过的时间序列进行分解,得到多个子序列,以削弱时间序列的非平稳性;其次,通过核极限学习机模型预测子序列,同时提出改进黏菌—哈里斯鹰算法优化其超参数;最后,重构各子序列的预测值,得到最终预测结果 ...
付文龙 +3 more
doaj
[Interpretation of the "Expert consensus on the management of neonatal parenteral nutrition (2025)"]. [PDF]
Lin XZ +8 more
europepmc +1 more source
[Determination of 30 homologues of phosphatidylcholines and lysophosphatidylcholines in human serum by liquid chromatography-tandem mass spectrometry and their correlation analysis with coronary artery disease]. [PDF]
Li WY +6 more
europepmc +1 more source
[Lipid metabolomics-based biomarker analysis of neonatal sepsis in serum and cerebrospinal fluid]. [PDF]
Wu WX +5 more
europepmc +1 more source
[Metabolomic alterations in preterm infants with bronchopulmonary dysplasia]. [PDF]
Wu YY +10 more
europepmc +1 more source
[Shihu Mixture alleviates diabetic cardiomyopathy in rats by Sirt3-mediated upregulation of myocardial mitochondrial mitophagy pathway]. [PDF]
Lin X, He Y, Shi H, Liu J, Hu H.
europepmc +1 more source
[Establishment and application of a knowledge-directed pseudo-targeted analytical method for diabetic retinopathy]. [PDF]
Hou YQ +8 more
europepmc +1 more source
[Medical treatment standards for emergency burn wounded (2025 edition)]. [PDF]
National Health Commission Of The People's Republic Of China.
europepmc +1 more source

