基于BayesianOpt-XGBoost的煤电机组碳排放因子预测

Prediction of coal-fired power units carbon emission factor based on BayesianOpt-XGBoost

以贝叶斯参数优化的XGBoost算法为基础,基于机组特征和煤炭特性建立BayesianOpt-XGBoost预测模型,其发电、供热碳排放因子预测的相关系数R^(2)分别为0.91和0.87,绝对误差百分比为2.51%和2.91%.进一步,通过特征标准化方法减少对煤炭特性的依赖,模型预测R2分别为0.79和0.77,绝对误差百分比为3.94%和2.75%,精度仍可得到保障.基于该模型分析全国各省区煤电机组碳排放因子并与公布数据进行比较,证明了该模型的有效性.对机组预测结果的分析表明对现存的低容量机组进行改造、对新建造电机组采用大容量高参数可以减少碳排放强度.

A Bayesian-Opt-XGBoost model was established on the basis of the features of power generation units and coals,in which the parameters were optimized with Bayesian.The prediction of the carbon emission factors of power and heat generation of coal-fired power plants had coefficients of(R^(2))of 0.91 and 0.87,respectively,the corresponding mean absolute errors are 2.51%and 2.91%.Normalization methods were used to get rid of the dependence on coal's features,the corresponding R2 values were 0.79 and 0.77 respectively,and the mean absolute errors were 3.94%and 2.75%,the accuracy can still be acceptable.With the model,the carbon emission factors of coal power units in different provinces of China were estimated and compared with the published data,which proved the valid of this model.The analysis of the above estimated results shown that the carbon emission intensity of coal-fired power industry can be reduced by reforming the existing low-capacity units and building large capacity and high parameters units for newly plants.