Texas is the largest state by area in the US after Alaska,and one of the top states in the production and consumption of electricity with many coal-fired plants.Coal-fired power plants emit greater than 70% of polluta...Texas is the largest state by area in the US after Alaska,and one of the top states in the production and consumption of electricity with many coal-fired plants.Coal-fired power plants emit greater than 70% of pollutants in the energy sector.When coal is burned to produce electricity,nitrogen oxides(NO_(x))are released into the air,one of the main pollutants that threaten human health and lead to a large number of premature deaths.The key to effective air quality management is the strict compliance of all plants with emission standards.However,not all Texas coal plants have the environmental equipment to lower pollutant emissions.Nitrogen dioxide(NO2)observations from the TROPOspheric Monitoring Instrument(TROPOMI)were used to evaluate the emissions for Texas power plants.Data from both the Emissions and Generation Resource Integrated Database(EGRID)and the Emissions Database for Global Atmospheric Research(EDGAR)were used to examine emissions.It was found that NOx emissions for Texas power plants range from 1.53 kt/year to 10.99 kt/year,with the Martin Lake,Limestone and Fayette Power Project stations being the top emitters.WA Parish and Martin Lake stations have the strongest NOx fluxes,with both exhibiting significant seasonal variability.Comparisons of bottom-up inventories for EDGAR and EGRID show a high correlation(r=0.956)and a low root mean square error(0.766).A more reasonable control policy would lead to much reduced NOx emissions.展开更多
锅炉燃烧优化在电厂锅炉经济稳定运行中起着重要作用,NO_(x)排放预测是其中的一个基本环节,因此提出了一种基于改进蜣螂优化算法优化卷积神经网络(convolutional neural network,CNN)与双向长短期记忆神经网络(long short term memory,L...锅炉燃烧优化在电厂锅炉经济稳定运行中起着重要作用,NO_(x)排放预测是其中的一个基本环节,因此提出了一种基于改进蜣螂优化算法优化卷积神经网络(convolutional neural network,CNN)与双向长短期记忆神经网络(long short term memory,LSTM)的组合模型超参数的超超临界锅炉NO_(x)排放预测的方法。首先通过Pearson相关性判定与NO_(x)排放相关的特征参数;其次建立CNN-LSTM预测模型,利用卷积神经网络CNN提取分层数据结构,长短期记忆网络挖掘长期依赖关系,然后结合佳点集、t分布变异策略对蜣螂算法进行改进,用改进后的算法对LSTM超参数进行优化得到最终预测模型;最后与其他神经网络模型进行对比验证。以某660 MW机组锅炉深度调峰实际数据进行预测,结果得到NO_(x)排放浓度实际值与预测值的平均绝对误差为3.3516,平均相对误差为2.4667,数据结果表明该预测模型具有更准确的预测效果。展开更多
基金This work was supported by the Basic Research Top Talent Plan of Lanzhou Jiaotong University(2022JC05).
文摘Texas is the largest state by area in the US after Alaska,and one of the top states in the production and consumption of electricity with many coal-fired plants.Coal-fired power plants emit greater than 70% of pollutants in the energy sector.When coal is burned to produce electricity,nitrogen oxides(NO_(x))are released into the air,one of the main pollutants that threaten human health and lead to a large number of premature deaths.The key to effective air quality management is the strict compliance of all plants with emission standards.However,not all Texas coal plants have the environmental equipment to lower pollutant emissions.Nitrogen dioxide(NO2)observations from the TROPOspheric Monitoring Instrument(TROPOMI)were used to evaluate the emissions for Texas power plants.Data from both the Emissions and Generation Resource Integrated Database(EGRID)and the Emissions Database for Global Atmospheric Research(EDGAR)were used to examine emissions.It was found that NOx emissions for Texas power plants range from 1.53 kt/year to 10.99 kt/year,with the Martin Lake,Limestone and Fayette Power Project stations being the top emitters.WA Parish and Martin Lake stations have the strongest NOx fluxes,with both exhibiting significant seasonal variability.Comparisons of bottom-up inventories for EDGAR and EGRID show a high correlation(r=0.956)and a low root mean square error(0.766).A more reasonable control policy would lead to much reduced NOx emissions.
文摘锅炉燃烧优化在电厂锅炉经济稳定运行中起着重要作用,NO_(x)排放预测是其中的一个基本环节,因此提出了一种基于改进蜣螂优化算法优化卷积神经网络(convolutional neural network,CNN)与双向长短期记忆神经网络(long short term memory,LSTM)的组合模型超参数的超超临界锅炉NO_(x)排放预测的方法。首先通过Pearson相关性判定与NO_(x)排放相关的特征参数;其次建立CNN-LSTM预测模型,利用卷积神经网络CNN提取分层数据结构,长短期记忆网络挖掘长期依赖关系,然后结合佳点集、t分布变异策略对蜣螂算法进行改进,用改进后的算法对LSTM超参数进行优化得到最终预测模型;最后与其他神经网络模型进行对比验证。以某660 MW机组锅炉深度调峰实际数据进行预测,结果得到NO_(x)排放浓度实际值与预测值的平均绝对误差为3.3516,平均相对误差为2.4667,数据结果表明该预测模型具有更准确的预测效果。
