基于LSTM预测反馈的垃圾焚烧烟气湿法脱酸控制优化
Optimization of wet deacidification control of waste incineration flue gas based on LSTM prediction feedback
罗国鹏 1胡思捷 2孙殿伟 3黄群星 2高峰 3朱燕华 3时丕伟 3王松 3汪守康2
作者信息
- 1. 光大环保(中国)有限公司,广东 深圳 518000
- 2. 浙江大学热能工程研究所,浙江 杭州 310027
- 3. 光大环境能源(杭州富阳)有限公司,浙江 杭州 311400
- 折叠
摘要
吸收液流量控制是生活垃圾焚烧烟气湿法脱酸系统控制的关键,针对现有控制系统存在的pH值反馈延迟长导致流量控制不稳定的问题,提出基于LSTM神经网络算法的吸收液流量预测控制方法,利用脱酸系统进出口SO2浓度值等系统信息,构建吸收液流量控制的前馈信号.通过对实际运行数据的建模分析,证明了LSTM模型在系统负荷变动平稳、入口二氧化硫浓度稳定的工况下,预测的吸收液流量与实际值在各个负荷段的预测误差为4.5%～6.3%,吸收液流量预测值与原始流量接近,满足出口SO2气体浓度稳定排放的条件.在系统负荷变动和入口烟气浓度波动的前提下,基于该预测模型调节吸收液流量,相较原先的pH值控制手段,吸收液流量相较原始流量平均增加1.7%,出口二氧化硫浓度值的波动范围缩小23.4%,出口二氧化硫排放总量下降8.77%,SO2排放的稳定性与效率都得到了明显的提高.
Abstract
Absorbent flow control is key part in wet deacidification system control.A predictive method on the control of absorption liquid flow based on LSTM neural network algorithm was proposed to suppress the unstableness in flow control caused by long pH feedback delay in existing control systems.The concentrations of SO2 at the inlet and outlet of deacidification system were used as feedforward signal for absorption liquid flow control.By modeling and analysis of actual operation data,the error between predicted flow rate of absorption liquid and actual value under stable load change and the concentration of sulfur dioxide are within 4.5%～6.3%,and the predicted flow of absorption liquid is close to original flow,which meets the demand of stable discharge of outlet SO2 gas concentration.Under unstable load conditions,the absorption liquid flow rate increased by an average of 1.7%compared with original flow control method,the fluctuation range of outlet sulfur dioxide concentration value decreased by 23.4%,the total sulfur dioxide emission decreased by 8.77%and the stability and efficiency of desulfurization were significantly improved.
关键词
湿法脱酸/吸收液流量/LSTM网络/预测/控制Key words
wet deacidification/absorption fluid flow/LSTM network/forecast/control引用本文复制引用
基金项目
&&(FYNY-HT-20220314)
中央高校基本科研业务费专项(2022ZFJH04)
出版年
2024
国家科技期刊平台
NSTL
万方数据