摘要
风机作为火力发电的重要辅机,对其进行及时高效的故障诊断,可有效减少停机损失,提高火力发电效率。k近邻(KNN)对非平稳数据样本有良好的分类能力。为了改进传统KNN算法存在的缺陷,构建投票加权网格搜索-k近邻算法(投票加权GS-KNN)故障诊断模型,利用网格搜索完成k值的选取,基于前k个近邻构建与距离值呈负相关的权值投票公式,依据投票得分情况进行故障诊断。使用投票加权GS-KNN模型对离心风机常见的9种运行状态进行故障诊断,拟合k值与准确率的关系,诊断准确率可达到100%。
As a critical auxiliary component in thermal power generation,the efficient and timely diagnosis of faults in fans can significantly reduce downtime losses and enhance the overall efficiency of thermal power generation.The knearest neighbors(KNN)algorithm demonstrates strong classification capabilities for non-stationary data samples.To address the limitations of traditional KNN algorithms,this study proposes a vote weighted grid search k-nearest neighbors algorithm(vote weighted GS-KNN)for fault diagnosis.The algorithm utilizes grid search to select the optimal k value,establishes a weighted voting formula based on the negative correlation between distance values and the proximity of the top k neighbors,and performs fault diagnosis according to the voting scores.The vote weighted GS-KNN model is applied to diagnose nine common operational states of centrifugal fans,and the relationship between the fitted k values and diagnostic accuracy is explored.The diagnostic accuracy of the proposed model reaches 100%.
作者
曾学文
陈高超
付名江
邵峰
伍仁杰
ZENG Xue-wen;CHEN Gao-chao;FU Ming-jiang;SHAO Feng;WU Ren-jie
出处
《节能》
2024年第1期47-50,共4页
Energy Conservation
关键词
故障诊断
火力发电
网格搜索
K近邻算法
投票加权
fault diagnosis
thermal power generation
grid search
k-nearest neighbor algorithm
vote weighting
