基于MIC特征提取与ICEEMD-RIME-DHKELM的建筑业碳排放预测模型

张新生; 聂达文; 陈章政

doi:10.13205/j.hjgc.202504005

基于MIC特征提取与ICEEMD-RIME-DHKELM的建筑业碳排放预测模型

doi: 10.13205/j.hjgc.202504005

1. 西安建筑科技大学管理学院,西安 710055;
2. 陕西省新型城镇化和人居环境研究院,西安 710055

基金项目:

陕西省教育厅科学研究计划项目“陕西省新型城镇化与低碳发展的耦合协调研究”（23JT018）

详细信息

作者简介:
张新生（1978-），男，教授，主要研究方向为城镇化与碳排放、机器学习与应用。zhangxs@xauat.edu.cn

通讯作者:
张新生（1978-），男，教授，主要研究方向为城镇化与碳排放、机器学习与应用。zhangxs@xauat.edu.cn

计量
- 文章访问数: 6
- HTML全文浏览量: 3
- PDF下载量: 0
- 被引次数: 0
出版历程
- 收稿日期: 2024-04-29
- 录用日期: 2024-07-17
- 修回日期: 2024-07-02
- 刊出日期: 2025-04-01

Research on a prediction model for carbon emissions of construction industry based on MIC feature extraction and ICEEMD-RIME-DHKELM

1. College of Management, Xi'an University of Architecture and Technology, Xi'an 710055, China, China;
2. Shaanxi Institute of New Urbanization and Human Settlement Environment Research, Xi'an 710055, China

摘要

摘要: 为解决建筑业碳排放研究中影响因素选取局限性、数据预处理不足、碳排放复杂动态变化及非线性问题，提出了一种基于最大信息系数（MIC）特征提取、改进互补集合经验模态分解（ICEEMD）、雾凇优化算法（RIME）与深度混合核极限学习机（DHKELM）的建筑业碳排放量预测模型。首先，根据IPCC计算方法，从直接和间接两个方面测算 1992—2021年我国建筑业碳排放量，基于STIRPAT模型选取年末总人口数、国内生产总值、建筑业房屋竣工面积和能源结构等17个影响建筑业碳排放量的因素，然后利用灰色关联分析和MIC方法两阶段筛选出12个关键影响因素；其次，使用ICEEMD将建筑业碳排放量分解为多个平稳序列和一个残差项，并将其分别代入RIME算法优化关键参数后的DHKELM模型中。最后，将各分解序列的预测结果相加获得建筑业碳排放预测值，并对比分析多种基准模型的预测结果。结果显示：MIC-ICEEMD-RIME-DHKELM模型的预测性能最优，其均方根误差、平均绝对误差、平均绝对百分比误差和绝对相关系数分别为0.2782亿t、0.2672亿t、1.3783%和0.9576，均优于其他模型，证明该模型适用于建筑业碳排放量的预测。该研究成果为建筑业的低碳发展提供理论支持和技术参考。
- 建筑业 /
- 碳排放 /
- 最大信息系数 /
- 改进互补集合经验模态分解 /
- 雾凇优化算法 /
- 深度混合核极限学习机
Abstract: As one of the important sources of global carbon emissions， the carbon emission prediction of the construction industry is crucial to promote the low-carbon transition and formulate effective carbon emission reduction policies. However， the existing carbon emission prediction models are limited in many aspects， especially in the accuracy of influencing factors selection， the integrity of data preprocessing， the complex dynamic changes of carbon emission data， the processing of nonlinear characteristics and so on. To solve these problems， this paper proposed a carbon emission prediction model for the construction industry based on maximum information coefficient （MIC） feature extraction， improved complementary ensemble empirical mode decomposition （ICEEMD）， RIME optimization algorithm， and deep hybrid kernel extreme learning machine （DHKELM）. First， according to the calculation method provided by the IPCC， this paper calculatesd the carbon emissions of China’s construction industry from 1992 to 2021， covering multiple factors of direct and indirect emissions. Combined with the STIRPAT model， 17 potential influencing factors were identified， including year-end total population， gross domestic product， construction area completed and energy mix. These factors fully reflected the driving mechanism of carbon emissions in the construction industry. Through grey correlation analysis and MIC method in two stages， this paper selected 12 key factors that had a significant impact on the carbon emissions of the construction industry. In the data preprocessing stage， the improved ICEEMD method was used to decompose the original carbon emission data， and several stationary sequences and one residual term were obtained. This decomposition process effectively reduced the noise in the data and enhanced the stationarity of the time series， thus providing more reliable data input for the subsequent modeling. Next， the RIME optimization algorithm was used to optimize the key parameters of the DHKELM model to further improve the model’s predictive power. DHKELM， as an effective deep learning algorithm， can adaptively capture the nonlinear characteristics of carbon emission data through the built-in hybrid kernel function. In the process of model construction and training， each decomposed sequence was input into the optimized DHKELM model for training and prediction. Finally， based on the forecast results of each decomposition sequence， the predicted value of carbon emissions of the construction industry as a whole was obtained. In order to evaluate the predictive performance of the model， the results of the model were compared with those of various benchmark models. The experimental results showed that the MIC-ICEEMD-RIME-DHKELM model was superior to other models in many performance indexes. Specifically， the model had a root mean square error of 0.2782 million tons， a mean absolute error of 0.2672 million tons， a mean absolute percentage error of 1.3783%， and an absolute correlation coefficient of 0.9576， showing its superior forecasting ability. Through the above analysis， the proposed MIC-ICEEMD-RIME-DHKELM model can effectively predict the carbon emission of the construction industry， and provide important theoretical support and practical reference for the carbon emission monitoring and policy formulation of the construction industry.
- construction industry /
- carbon emissions /
- maximum information coefficient /
- improved complete ensemble empirical mode decomposition /
- rime optimization algorithm /
- deep hybrid kernel extreme learning machine

HTML全文

1 软雾凇粒子运动图

1. Soft-rime particle motion diagram

下载: 全尺寸图片幻灯片

2 硬雾凇粒子穿刺图

2. Hard-rime particle puncture diagram

下载: 全尺寸图片幻灯片

3 DHKELM模型结构

3. DHKELM model structure diagram

下载: 全尺寸图片幻灯片

4 基于MIC特征提取与ICEEMD-RIME-DHKELM模型预测流程

4. Flowchart of prediction based on MIC feature extraction and ICEEMD-RIME-DHKELM model

下载: 全尺寸图片幻灯片

5 1992—2021年中国建筑业碳排放趋势

5. China construction industry carbon emissions trends from 1992 to 2021

下载: 全尺寸图片幻灯片

6 建筑业碳排放影响因素MIC系数热力图

6. Heatmap of MIC coefficients of influencing factors of construction industry carbon emissions

下载: 全尺寸图片幻灯片

7 ICEEMD分解后的建筑业碳排放量序列图

年份

7. Decomposed construction industry carbon emissions series diagram after ICEEMD analysis

下载: 全尺寸图片幻灯片

8 分解后各imf _j 和res预测结果对比

8. Comparisons of prediction results for each decomposed imf _j and res

下载: 全尺寸图片幻灯片

9 各模型预测结果对比

9. Comparison chart of prediction results from various models

下载: 全尺寸图片幻灯片

10 各模型相对误差对比

10. Comparisons of relative errors for each model

下载: 全尺寸图片幻灯片

11 各模型平均相对误差雷达图

11. Radar chart of average relative errors for each model

下载: 全尺寸图片幻灯片

1 能源碳排放系数^［22］

1. Energy carbon emission coefficients^［22］

能源种类	碳排放系数	能源种类	碳排放系数
原煤	0.7143 kg C/kg	燃料油	1.4286 kg C/kg
焦炭	0.9714 kg C/kg	石油沥青	1.3307 kg C/kg
原油	1.4286 kg C/kg	液化石油气	1.7143 kg C/kg
汽油	1.4714 kg C/kg	天然气	1.33 kg C/kg
煤油	1.4714 kg C/kg	热力	34.12 kg C/GJ
柴油	1.4571 kg C/kg	电力	0.345 kg C/（kW·h）

下载: 导出CSV

2 建筑材料碳排放系数^{［23，24］}

2. Carbon emission coefficients of building materials^{［23，24］}

建筑材料种类	钢材	木材	水泥	平板玻璃	铝材
碳排放系数	1.789 kg C/kg	-842.8 kg C/m³	0.815 kg C/kg	0.966 kg C/kg	2.6 kg C/kg

下载: 导出CSV

3 积极贪婪选择机制伪代码^［28］

3. Pseudo-code of the hard-rime puncture mechanism

Initialize population R
Get the current optimal agent and optimal fitness
Whilet ≤ T
Coefficient of adherence E = sqrt(t/T)
Ifr₂ < E
Update rime location by the soft-rime search strategy
End If
Ifr₃ < Normalize fitness of S_i
Cross updating between agents by the hard-rime puncture mechanism
End If
If F(R_i^new ) < F(R_i )
Select the best solution over the inferior one using positive greedy selection
End If
t = t + 1
End while

下载: 导出CSV

4 建筑业碳排放影响因素

4. Factors influencing construction industry carbon emissions

结构	影响因素	序号
人口	年末总人口数	x1
	人口密度	x2
	城镇化率	x3
财富	国内生产总值	x4
	第三产业增加值	x5
	全社会固定资产投资	x6
	居民人均可支配收入	x7
	建筑业劳动生产率	x8
	建筑业生产总值	x9
技术	建筑业房屋竣工面积	x10
	单位GDP能源消耗	x11
	一次能源生产总量	x12
	建筑业企业从业人员	x13
	能源结构	x14
	能源强度	x15
	能源加工转换效率	x16
	投入产出率	x17

下载: 导出CSV

5 灰色关联分析排名及MIC系数

5. Grey relational analysis ranking and MIC coefficients

影响因素	灰色关联度	排名	MIC系数
建筑业劳动生产率	0.929	1	1
建筑业房屋竣工面积	0.923	2	1
居民人均可支配收入	0.918	3	1
能源强度	0.913	4	1
建筑业生产总值	0.906	5	1
人口密度	0.904	6	1
第三产业增加值	0.899	7	1
建筑业企业从业人员	0.86	8	1
一次能源生产总量	0.854	9	1
城镇化率	0.814	10	1
能源加工转换效率	0.79	11	1
年末总人口数	0.77	12	1
全社会固定资产投资	0.77	12	0.92
国内生产总值	0.766	14	0.5
能源结构	0.746	15	0.72
投入产出率	0.736	16	0.53
单位GDP能源消耗	0.682	17	0.53

下载: 导出CSV

6 各IMF序列的皮尔逊相关系数

6. Pearson correlation coefficient of each IMF series

IMF序列	imf₁	imf₂	imf₃	res
皮尔逊相关系数	0.25864	0.293667	0.352066	0.925142

下载: 导出CSV

7 模型相关参数及对应数值

7. Model-related parameters and corresponding values

DHKELM模型基础参数	对应数值	RIME算法优化相关参数	对应数值
输入层神经元数	12	L2正则化参数	864.1517
输出层神经元数	1	顶层KELM惩罚参数	0.001
隐含层数量	3	核函数1的权重	0.365
激活函数	Sigmoid	隐含层节点数	9
核函数	RBF		99
核函数	poly		17

下载: 导出CSV

8 各模型预测性能评价指标对比

8. Comparison of prediction performance evaluation indicators for each model

模型	MAE/10⁸ t	RMSE/10⁸ t	MAPE/%	R²
MIC-ICEEMD-RIME-DHKELM	0.2672	0.2782	1.3783	0.9576
ICEEMD-RIME-DHKELM	0.2134	0.4481	2.3100	0.8831
MIC-ICEEMD-WOA-DHKELM	0.3862	0.4191	1.9817	0.9038
MIC-ICEEMD-SSA-DHKELM	0.3903	0.4122	2.0167	0.9070
MIC-RIME-DHKELM	0.3483	0.3662	1.7800	0.9266
MIC-ICEEMD-DHKELM	0.4716	0.4960	2.4383	0.8654
MIC-ICEEMD-KELM	0.5378	0.5559	2.7800	0.8308

下载: 导出CSV

参考文献(37)

[1]	LIU Y H，LI Y H，WANG W T. Challenges，opportunities and actions for china to achieve the targets of carbon peak and carbon neutrality［J］. China Population，Resources and Environment，2021，31（9）:1-5. 刘燕华，李宇航，王文涛. 中国实现“双碳”目标的挑战、机遇与行动［J］. 中国人口·资源与环境，2021，31（9）:1-5.
[2]	WEI G P，KANG Y，FAN H W，et al. Accounting and prediction on construction carbon emissions in heavy industry city［J］. Ecological Economy，2022，38（9）:43-48. 魏光普，康瑜，范浩文，等. 重工业城市建筑业碳排放核算与预测研究［J］. 生态经济，2022，38（09）:43-48.
[3]	VITO A，UMBERTO B. Green buildings and organizational changes in italian case studies［J］. Business Strategy and the Environment，2012，21（6）:387-400.
[4]	XI Z，ZHENG L，LIN W M，et al. Analyzing carbon emissions embodied in construction services:a dynamic hybrid input-output model with structural decomposition analysis［J］. Energies，2019，12（8）:1456.
[5]	WANG Z Q，LI K R，REN J G，et al. Influential factors and scenario forecast of carbon emissions of construction industry in shandong province based on LMDI-SD model［J］. Environmental Engineering，2023，41（10）:108-116. 王志强，李可慧，任金哥，等. 基于LMDI-SD模型的山东省建筑业碳排放影响因素与情景预测［J］. 环境工程，2023，41（10）:108-116.
[6]	LIU X H，LIAO C P，HUANG Y，et al. Analysis of factors affecting building carbon emissions and emission reduction measures in guangzhou based on STIRPAT model［J］. Renewable Energy Resources，2019，37（5）:769-775. 刘兴华，廖翠萍，黄莹，等. 基于STIRPAT模型的广州市建筑碳排放影响因素及减排措施分析［J］. 可再生能源，2019，37（5）:769-775.
[7]	NUŢĂ F M，NUŢĂ A C，ZAMFIR C G，et al. National carbon accounting—analyzing the impact of urbanization and energy-related factors upon cosub2/ sub emissions in central-eastern european countries by using machine learning algorithms and panel data analysis［J］. Energies，2021，14（10）:2775-2775.
[8]	ZHANG J X，ZHANG H，WANG R，et al. Measuring the critical influence factors for predicting carbon dioxide emissions of expanding megacities by XGBoost［J］. Atmosphere，2022，13（4）:599-599.
[9]	SHANG C J，CAI J，LIU Y R，et al. Study on analysis and forecast of carbon dioxide emissions accounting of construction industry in hainan province［J］. Environmental Engineering，2016，34（04）:161-165. 尚春静，蔡晋，刘艳荣，等. 海南省建筑业碳排放核算分析及预测研究［J］. 环境工程，2016，34（04）:161-165.
[10]	CHEN Y X，CHEN J G，WANG X Q，et al. Research on the prediction of carbon emission and emission reduction strategies in construction industry［J］. Construction Economy，2016，37（10）:14-18. 陈雨欣，陈建国，王雪青，等. 建筑业碳排放预测与减排策略研究［J］. 建筑经济，2016，37（10）:14-18.
[11]	CHANG L，MOHSIN M，HASNAOUI A，et al. Exploring carbon dioxide emissions forecasting in china:a policy-oriented perspective using projection pursuit regression and machine learning models［J］. Technological Forecasting Social Change，2023，197:122872.
[12]	ZHAO J H，LI J S，WANG P L，et al. A study on carbon peaking paths in henan，china based on lasso regression-BP neural network model［J］. Environmental Engineering，2022，40（12）:151-156+164. 赵金辉，李景顺，王潘乐，等. 基于Lasso-BP神经网络模型的河南省碳达峰路径研究［J］. 环境工程，2022，40（12）:151-156+164.
[13]	ZHANG X S，WEI Z Z，CHEN Z Z，et al. Prediction of industrial carbon emissions in Shaanxi province based on LASSO-GWO-KELM model［J］. Environmental Engineering，2023，41（10）:141-149. 张新生，魏志臻，陈章政，等. 基于LASSO-GWO-KELM的工业碳排放预测方法研究［J］. 环境工程，2023，41（10）:141-149.
[14]	CHEN Z，GUO Y L，GUO C，et al. Greenhouse gas prediction using whale optimization algorithm-deep extreme learning machine in Chengdu metro construction stage［J］. Tunnel Construction，2022，42（12）:2048-2063. 陈政，郭亚林，郭春. 基于WOA-DELM的成都地铁建设阶段温室气体预测［J］. 隧道建设（中英文），2022，42（12）:2048-2063.
[15]	ZHANG Z H，LIU R J. Carbon emissions in the construction sector based on input-output analyses［J］. Journal of Tsinghua University（Science and Technology），2013，53（1）:53-57. 张智慧，刘睿劼. 基于投入产出分析的建筑业碳排放核算［J］. 清华大学学报（自然科学版），2013，53（1）:53-57.
[16]	DANESHGAR S，AMERLINCK Y，AMARAL A，et al. An innovative model-based protocol for minimisation of greenhouse gas（GHG）emissions in WRRFs［J］. Chemical Engineering Journal，2024，483148327.
[17]	SIMON E，LEANDRO B，KUNIHISA M，et al. 2006 IPCC Guidelines for national greenhouse gas inventories volume 1:general guidance and reporting［J］. IGES，2006.
[18]	FENG B，WANG X Q，LIU B S. Provincial variation in energy efficiency across china’s construction industry with carbon emission considered［J］. Resources Science，2014，36（6）:1256-1266. 冯博，王雪青，刘炳胜. 考虑碳排放的中国建筑业能源效率省际差异分析［J］. 资源科学，2014，36（06）:1256-1266.
[19]	CHEN W L，YANG S Y，ZHANG X Z，et al. Embodied energy and carbon emissions of building materials in China［J］. Building and Environment，2022，207（PA）:108434.
[20]	LI Z J. Study on the life cycle consumption of energy and resource of air conditioning in urban residential buildings in China［D］. Beijing:Tsinghua University，2007. 李兆坚. 我国城镇住宅空调生命周期能耗与资源消耗研究［D］. 北京:清华大学，2007.
[21]	NIU H L，LIU Z Y. Research on influencing factors of carbon emission of construction industry in China:based on dynamic spatial durbin panel model［J］. Ecological Economy，2017，33（8）:74-80. 牛鸿蕾，刘志勇. 基于动态空间杜宾面板模型中国建筑业碳排放的影响因素研究［J］. 生态经济，2017，33（8）:74-80.
[22]	SU Y X，CHEN X Z，YE Y Y，et al. The characteristics and mechanisms of carbon emissions from energy consumption in china using DMSP/OLS night light imageries［J］. Acta Geographica Sinica，2013，68（11）:1513-1526. 苏泳娴，陈修治，叶玉瑶，等. 基于夜间灯光数据的中国能源消费碳排放特征及机理［J］. 地理学报，2013，68（11）:1513-1526.
[23]	XU Y G，SONG J X. Carbon emission prediction of construction industry based on FCS-SVM［J］. Ecological Economy，2019，35（11）:37-41. 徐勇戈，宋伟雪. 基于FCS-SVM的建筑业碳排放预测研究［J］. 生态经济，2019，35（11）:37-41.
[24]	ZHANG X S，REN M Y，CHEN Z Z. Research on carbon emission prediction of construction industry based on CEEMD-SSA-ELM method［J］. Ecological Economy，2023，39（10）:33-39+88. 张新生，任明月，陈章政. 基于CEEMD-SSA-ELM方法的建筑业碳排放预测研究［J］. 生态经济，2023，39（10）:33-39+88.
[25]	DAVID N R，YAKIR A R，HILARY K F，et al. Detecting novel associations in large data sets［J］. Science，2011，334（6062）:1518-1524.
[26]	MARCELO A C，GASTÓN S，MARÍA E T. Improved complete ensemble EMD:a suitable tool for biomedical signal processing［J］. Biomedical Signal Processing and Control，2014，1419-1429.
[27]	SU H，ZHAO D，HEIDARI A A，et al. RIME:A physics-based optimization［J］. Neurocomputing，2023，532183-532214.
[28]	YANG B，WANG J R，SU S，et al. Mismatch losses mitigation of PV-TEG hybrid system via improved RIME algorithm:design and hardware validation［J］. Journal of Cleaner Production，2024，434139957.
[29]	SUN W，HUANG C C. Predictions of carbon emission intensity based on factor analysis and an improved extreme learning machine from the perspective of carbon emission efficiency［J］. Journal of Cleaner Production，2022，338:130414.
[30]	LIU J，XU H Y，PENG X Y，et al. Reliable composite fault diagnosis of hydraulic systems based on linear discriminant analysis and multi-output hybrid kernel extreme learning machine［J］. Reliability Engineering and System Safety，2023，234:109178.
[31]	WANG R，XU X C，LU J. Short-term wind power prediction based on SSA optimized variational mode decomposition and hybrid kernel extreme learning machine［J］. Information and Control，2023，52（4）:444-454. 王瑞，徐新超，逯静. 基于麻雀搜索算法优化变分模态分解和混合核极限学习机的短期风电功率预测［J］. 信息与控制，2023，52（4）:444-454.
[32]	WANG C，LIN H，PANG X H. Short-termphotovoltaic power combination prediction based on HPO-VMD and MISMA-DHKELM［J］. Acta Energiae Solaris Sinica，2023，44（12）:65-73. 王超，蔺红，庞晓虹. 基于HPO-VMD和MISMA-DHKELM的短期光伏功率组合预测［J］. 太阳能学报，2023，44（12）:65-73.
[33]	BRANT L，SIDNEY L. Age-structure，urbanization，and climate change in developed countries:revisiting STIRPAT for disaggregated population and consumption-related environmental impacts［J］. Population and Environment，2010，31（5）:317-343.
[34]	JIA W，WEI S，JING R R，et al. Exploring the driving factors and their spatial effects on carbon emissions in the building sector［J］. Energies，2023，16（7）:3094.
[35]	SUN Y H，HAO S Y，LONG X F. A study on the measurement and influencing factors of carbon emissions in China's construction sector［J］. Building and Environment，2023，229:109912.
[36]	LIU Y，WANG X，ZHU L. Analyzing carbon emissions in the Yangtze River Delta’s construction industry:spatiotemporal characteristics and influencing factors［J］. China Environmental Science，2023，43（12）:6677-6688. 刘颖，王远，朱琳. 长三角地区建筑业碳排放变化的时空特征及影响因素分析［J］. 中国环境科学，2023，43（12）:6677-6688.
[37]	XU T S，LI X W，DONG X. Spatial characteristics and influencing factors of carbon emission intensity of construction industry in China［J］. Science& Technology Review，2024，42（6）:103-111. 徐水太，李晞薇，董信. 中国建筑业碳排放强度的空间特征与影响因素分析［J］. 科技导报，2024，42（6）:103-111.