汽车制造厂油漆车间能耗预测模型
|
摘要
随着国家和社会对制造企业节能减排要求的不断提高,汽车制造企业现有的粗放型能耗管理考核方法已无法满足精细化能源管理需求,在新电改直购电的背景下能耗建模与预测已成为企业节能的重要研究方向。为建立能耗与生产、天气、人员等因素之间的定量关系,帮助整车制造企业合理对标能源使用与生产信息,使用上海某汽车制造企业油漆车间2016年11月至2018年4月日能耗、产量、温湿度和设备运行等数据,综合考虑整车理论喷涂面积对能耗的影响,采用机器学习的方法建立了基于XGBoost的用能预测模型,在少量训练样本下也达到了良好预测效果。同时,重点分析和对比了4种特征工程方法对预测模型结果的影响,总结了能耗预测模型适宜的特征工程,对同类型的模型优化具有一定的指导意义。
With the continuous increase of energy conservation requirements from the government and the society, automotive OEM's current extensive energy management and assessment methods can no longer satisfy the company's need of precise energy management. In the case of new electric change plan, energy consumption modeling and forecasting have become an important research direction for enterprise energy conservation. In order to establish quantitative relationship among energy consumption, production, ambient condition, staff scheduling and other factors, and help automobile OEMs build reasonable and scientific links between energy use and production information, this article used daily data provided by paint shop of one Shanghai OEM factory to establish energy consumption models using XGBoost. The data include energy consumption, production, ambient temperature and humidity, and equipment status from November 2016 to April 2018, and the model considers the influence of theoretical spray area of each car model, and reaches a satisfying result with a small number of training samples. At the same time, this article focused on analyzing and comparing four different feature engineering methods and their effects on the prediction models, summarized the experience of feature engineering for energy consumption prediction, and provided suggestions for future model optimization.
With the continuous increase of energy conservation requirements from the government and the society, automotive OEM's current extensive energy management and assessment methods can no longer satisfy the company's need of precise energy management. In the case of new electric change plan, energy consumption modeling and forecasting have become an important research direction for enterprise energy conservation. In order to establish quantitative relationship among energy consumption, production, ambient condition, staff scheduling and other factors, and help automobile OEMs build reasonable and scientific links between energy use and production information, this article used daily data provided by paint shop of one Shanghai OEM factory to establish energy consumption models using XGBoost. The data include energy consumption, production, ambient temperature and humidity, and equipment status from November 2016 to April 2018, and the model considers the influence of theoretical spray area of each car model, and reaches a satisfying result with a small number of training samples. At the same time, this article focused on analyzing and comparing four different feature engineering methods and their effects on the prediction models, summarized the experience of feature engineering for energy consumption prediction, and provided suggestions for future model optimization.
引文
[1] 黄文燕,(再增加两名作者)等. 基于模拟退火PSO-BP算法的钢铁生产能耗预测研究[J]. 科学技术与工程, 2012(30): 7906-7910.
[2] 范伯良, 高峰,寇鹏. 在线Boosting回归算法及其在高耗能企业负荷预测中的应用[J]. 信息与控制, 2014(6): 750-756.
[3] 杨文人, 基于能耗预测模型的能源管理系统研究与实现[D]. 广州:华南理工大学, 2013.
[4] Chen, T. and C. Guestrin, XGBoost: A Scalable Tree Boosting System[J].. ACM,2016:785-794.
[5] 许裕栗,杨晶,李柠,等. Xgboost算法在区域用电预测中的应用[J]. 自动化仪表, 2018(7): 1-5.
[6] 夏博, 杨超, 郑凯文. 大数据平台下的电力负荷预测系统设计与实现[J]. 自动化仪表, 2018(6): 81-84.
[7] 杨修德, 王金梅, 张丽娜. XGBoost在超短期负荷预测中的应用[J]. 电气传动自动化, 2017(4): 21-25.
[8] 黄达文,方芃岚. 基于XGBoost算法的用电电量预测的实践应用[J]. 现代信息科技, 2017(4): 10-12.
[9] 李元诚, 方廷健, 于尔铿,等. 短期负荷预测的支持向量机方法研究[J]. 中国电机工程学报, 2003, 23(6):55-59.
[10] 李然然, 张永坚, 刘畅,等. 基于BP神经网络的建筑物用电能耗预测[J]. 山东建筑大学学报, 2011, 26(2):162-165.
[2] 范伯良, 高峰,寇鹏. 在线Boosting回归算法及其在高耗能企业负荷预测中的应用[J]. 信息与控制, 2014(6): 750-756.
[3] 杨文人, 基于能耗预测模型的能源管理系统研究与实现[D]. 广州:华南理工大学, 2013.
[4] Chen, T. and C. Guestrin, XGBoost: A Scalable Tree Boosting System[J].. ACM,2016:785-794.
[5] 许裕栗,杨晶,李柠,等. Xgboost算法在区域用电预测中的应用[J]. 自动化仪表, 2018(7): 1-5.
[6] 夏博, 杨超, 郑凯文. 大数据平台下的电力负荷预测系统设计与实现[J]. 自动化仪表, 2018(6): 81-84.
[7] 杨修德, 王金梅, 张丽娜. XGBoost在超短期负荷预测中的应用[J]. 电气传动自动化, 2017(4): 21-25.
[8] 黄达文,方芃岚. 基于XGBoost算法的用电电量预测的实践应用[J]. 现代信息科技, 2017(4): 10-12.
[9] 李元诚, 方廷健, 于尔铿,等. 短期负荷预测的支持向量机方法研究[J]. 中国电机工程学报, 2003, 23(6):55-59.
[10] 李然然, 张永坚, 刘畅,等. 基于BP神经网络的建筑物用电能耗预测[J]. 山东建筑大学学报, 2011, 26(2):162-165.