摘要
火力发电的主要能源来自煤炭,而由于燃煤发电过程中产生的煤尘扩散是引起火电厂粉尘爆炸风险和尘肺职业病的主要根源之一,因此必须进行实时有效的检测和控制。当前粉尘检测方法仅使用单一的粉尘质量浓度指标来评估粉尘污染整体状况,缺乏对多种复合因素影响的考量,依靠单一阈值设定进行报警,易出现误报、漏报等现象,以及忽略粉尘爆炸这一重要事故场景,不能建立粉尘污染全面客观的评价方法。研究建立了一种粉尘质量浓度预测模型,基于金豺优化算法对极限学习机的最优初始权重进行寻优,再使用极限学习机对样本数据进行训练学习,提高神经网络模型的精度,可较为准确地预测30 min以内任意时间间隔的粉尘质量浓度,并将现场数据及模拟仿真数据与建立的粉尘质量浓度预测模型进行对比分析。结果显示:建立的粉尘质量浓度预测模型准确度良好,与现场数据及模拟仿真数据对比误差分别为0.72%和2.1%,可加强对火电厂粉尘环境进行预测预警,从而及时采取合理的粉尘控制策略,确保火电厂的生产安全并降低粉尘对作业人员的职业危害。
In this paper,data fusion and optimization are performed on the collected raw field data from a thermal power plant,and the collected dust samples are analyzed for particle size using an LS 909 laser particle size analyzer.This paper describes the process of establishing a dataset for a thermal power plant dust environment,with a capacity of 28780.The weights of the dataset are then optimized using the golden jackal optimization algorithm.The optimal initial weights for the limit learning machine are determined based on the solution results.The limit learning machine is then trained using the sample data to create a dust concentration prediction model.Compared with the field data collected,the model shows an error rate of 0.72%.The dust concentration prediction algorithm can predict the dust concentration after 1 minute per iteration.By repeating the iteration,the dust concentration after a longer time interval can be obtained.However,the longer the time interval,the lower the accuracy of the prediction results.The model can more accurately predict the dust concentration within 30 minutes.In addition,the Ansys Fluent platform is used to simulate the movement and distribution of dust in the thermal power plant.The results show a high risk of dust explosion in the area downwind of the dust leakage source,as well as a high concentration of dust in areas where personnel are present.The simulation results are extracted from the CFD-Post software,and relevant data,such as the concentration,movement velocity,and distribution height of dust particles,are derived from 30 randomly selected positions in the flow field.These results are then compared and analyzed with the dust concentration prediction model established in this paper.The validation results show that the dust concentration prediction model has good accuracy,with a relative error of 2.1%.It can effectively predict the dust concentration in areas downwind of the leakage source,as well as areas with a high concentration of dust.Overall,the model can accurately predict the dust environment in the thermal power plant.
作者
王博
商宇航
姚立超
蒋永清
WANG Bo;SHANG Yuhang;YAO Lichao;JIANG Yongqing(School of Measurement and Control Technology and Communication Engineering,Harbin University of Science and Technology,Harbin 150080,China)
出处
《安全与环境学报》
CAS
CSCD
北大核心
2024年第5期1768-1777,共10页
Journal of Safety and Environment
基金
黑龙江省“双一流”学科协同创新成果项目(LJGXCG2022-068)。
关键词
安全工程
粉尘防爆
粉尘危害
粉尘质量浓度在线检测
工厂环境
劳动者安全
危险预知
safety engineering
protection against dust explosion
dust hazards
on-line detection of dust mass concentration
factory environment
worker safety
hazard prediction