摘要
高频、高精度的碳排放指标监测与预测是燃煤电厂运行优化的重要支撑。针对传统排放因子法在精确度和时效性上的不足,基于煤电机组运行、碳排放在线监测、燃煤成分数据,采用极致梯度提升(XGBoost)等算法识别出机组碳排放强度变化的主要动因,并提出降碳的对策建议。以安徽省3座燃煤电厂中7台320 MW及以上机组的日度数据作为模型训练的输入数据,发现XGBoost预测精度最优(R^(2)=0.951),碳排放强度与空干基含碳量、全水分、主汽温度和机再热汽温度关系密切。对案例机组的偏依赖分析发现,负荷率及其他不同因素共同作用于碳排放强度,机组负荷率上升可降低碳排放强度,而在中等负荷率的状态下(负荷率处于52%~80%的区间时),适当提升主汽温度(562℃→569℃)和机再热汽温度(550℃→564℃),及降低燃煤全水分(11.0%→6.5%)、燃煤的含碳量(42%→37%),可降低碳排放强度(最大降幅可达12.2%),总体上,负荷率小于70%时降碳效果更好。结论可为燃煤电厂智能化、低碳化运营提供依据,为碳排放精细化管理提供支撑。
High-frequency and high-precision monitoring and prediction of carbon emissions are crucial for the optimization of coal-fired power plant operations.In response to the limitations of traditional emission factor methods in terms of accuracy and timeliness,machine learning algorithms,such as XGBoost were employed to identify key factors influencing carbon emission intensity based on unit operation,online emission monitoring,and fuel composition data.Using the daily data from seven units of 320 MW or above at three coal-fired power plants in Anhui Province as training inputs,the XGBoost achieved the highest predictive accuracy(R^(2)=0.951).Carbon emission intensity was found to be closely related to air-dried basis carbon content,total moisture,main steam temperature,and reheat steam temperature.Partial dependence analysis revealed that load rate interacted with other factors to jointly influence carbon emission intensity.Increasing the load rate significantly lowered carbon emission intensity.Under medium-load conditions(load rate 52%-80%),raising main steam temperature(562℃→569℃),reheat steam temperature(550℃→564℃),and reducing total moisture of coal(11.0%→6.5%)and carbon content(42%→37%),could reduce carbon emission intensity by up to 12.2%.Furthermore,the carbon reduction effect was better when the load rate was below 70%.These findings could provide a basis for the intelligent and low-carbon operation of power plants and support the refined management of carbon emissions.
作者
吴妍
马大卫
张更
计巧珍
宋纪双
赵常威
周仲康
曹菊林
WU Yan;MA Dawei;ZHANG Geng;JI Qiaozhen;SONG Jishuang;ZHAO Changwei;ZHOU Zhongkang;CAO Julin(Anhui Xinli Power Technology Co.,Ltd.,Hefei Anhui 230601;State Grid Anhui Electric Power Research Institute,Hefei Anhui 230601)
出处
《环境污染与防治》
北大核心
2025年第8期97-104,共8页
Environmental Pollution & Control
基金
安徽新力电业科技有限责任公司科技项目(No.2024ZI-KJ-03)
国家自然科学基金资助项目(No.62473125)。
关键词
碳排放强度
在线监测
机器学习
极致梯度提升
降碳策略
carbon emission intensity
online monitoring
machine learning
XGBoost
carbon reduction strategy
