一种发电厂气体污染监测与预警模型设计  

Design of a gas pollution monitoring and early warning model for power plants

作　　者：付康民 邱建新[1] 解标 FU Kangmin;QIU Jianxin;XIE Biao(The Production and Technology Department of Guoneng(Suizhong)Power Generation Co.,Ltd.,Liaoning Huludao 125222,China;Anhui Deyuan Environmental Technology Co.,Ltd.,Anhui Bengbu 233000,China)

机构地区：[1]国能(绥中)发电有限责任公司生产技术部,辽宁葫芦岛125222 [2]安徽德源环境科技有限公司,安徽蚌埠233000

出　　处：《工业仪表与自动化装置》2025年第2期21-25,共5页Industrial Instrumentation & Automation

摘　　要：针对电厂气体污染监测的需要,该文提出了一种融合多源无线传感器网络(Wireless Sensor Networks,WSN)与改进长短期记忆网络(Long Short Term Memory,LSTM)的气体污染监测与预警模型。通过设计多源WSN污染气体采集框架,实现了对电厂多种污染气体的高效采集,为预测模型提供高质量的数据输入。所设计的改进LSTM模型融合了模拟退火算法(Simulated Annealing,SA)和SVM模块,其中SA用于优化LSTM的超参数,SVM则作为分类器有效避免过拟合问题,使改进后的LSTM能够充分提取多源传感器数据的特征并进行准确预测。实验结果表明,结合SA和SVM模块有效提升了LSTM的预测性能,与其他分类算法相比表现出了明显的优势,准确率高达97.83%,相比于对比算法中表现最佳的BiLSTM提高了9.64%。In response to the need for gas pollution detection in power plants,this paper proposes a gas pollution monitoring and early warning model that integrates multi-source Wireless Sensor Networks(WSN)and improved Long Short Term Memory(LSTM)networks.By designing a multi-source WSN pollution gas collection framework,efficient collection of various pollution gases from power plants has been achieved,providing high-quality data input for prediction models.The improved LSTM model designed integrates Simulated Annealing(SA)algorithm and SVM module,where SA is used to optimize the hyperparameters of LSTM and SVM acts as a classifier to effectively avoid overfitting problems,enabling the improved LSTM to fully extract the features of multi-source sensor data and make accurate predictions.The experimental results show that the combination of SA and SVM modules effectively improves the predictive performance of LSTM,showing significant advantages compared to other classification algorithms,with an accuracy rate of up to 97.83%,which is 9.64% higher than the best performing BiLSTM among the comparison algorithms.

关 键 词：多源数据融合 WSN 污染气体检测 LSTM 模拟退火算法 SVM 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]