摘要
针对火电机组故障预测准确率低的问题,提出基于BERT-STAM1DCNN的火电机组故障预测模型。在预处理阶段,利用随机森林算法对原始数据进行处理,筛选重要的特征,提高故障预测准确率;利用BERT(bidirectional encoder representations from transformers)模型对特征数据进行训练,并提出一种融合二次加权时空注意力机制的一维卷积神经网络(STAM1DCNN)模型,提高关键信息对预测结果的影响程度。以火电机组实际运行数据作为数据集,与其他模型相比,所提出的模型具有良好的性能和稳定性。
To solve the problem of low fault prediction accuracy of thermal power units,a fault prediction model of thermal power units based on BERT-STAM1DCNN is proposed.In the pre-processing stage,random forest algorithm is used to process the original data and screen out important features to improve the accuracy of fault prediction.Bidirectional encoder representations from transformers(BERT)model is used to train the feature data,and a 1D convolutional neural network integrating spatial-temporal attention mechanism with quadratic weighting(STAM1DCNN)model is proposed to improve the degree of influence of key information on the prediction results.The actual operating data of thermal power units is taken as the data set.Compared with other models,the proposed model has good performance and stability.
作者
钱晓峰
潘泉洪
田冰
吴超
QIAN Xiaofeng;PAN Quanhong;TIAN Bing;WU Chao(Zhejiang Zheneng Jiahua Power Generation Co.,Ltd.,Hangzhou 310000,China)
出处
《微型电脑应用》
2026年第1期196-199,共4页
Microcomputer Applications
