Xgboost算法在区域用电预测中的应用!
|
摘要
针对区域用户用电预测的影响因素众多、用户行为模式差异的特点,提出一种基于Xgboost算法的综合预测方法。该方法从周期、趋势、扰动三方面考虑用电的影响因素,搜集用户历史用电数据和气象数据,对数据中的缺失值进行补充,并对数据进行平滑化处理。将所得用户数据进行K-means聚类,从而得到不同类别的用户。运用最大信息系数(MIC)计算各影响因素与用户用电量的相关性大小,依据相关性大小对不同影响因素排序,进而提取主要影响因素。利用Xgboost算法构建不同用户类别的用电预测模型,对不同类别用户进行用电预测,进而获得区域整体用电情况。采用实际用电数据,调整算法参数,对综合方法进行仿真分析,并与其他方法进行对比。结果表明,该方法综合考虑了多方面影响因素,划分了不同的用户群落,具有更高的精度和可靠性。Xgboost算法在区域用电预测中具有较好的应用前景。
The forecasting of reginal power consumption is affected by many factors,and the behavior patterns of users are different,in accordance with these features,a comprehensive forecasting method based on Xgboost algorithm is proposed. With this method,the inflrencing factors of power consumption are considered from three aspects,i. e.,cycle,trend and disturbance,the historical power consumption data of users and the meteorological data are collected,then the missing values in these data are completmented,and smoothing process is conducted. The user data obtained are clustered by K-means to get the different categories of the users. Then,the maximal information coefficient( MIC) is used to calculate the correlation between each influencing factor and the electricity consumption of users. According to the magnitude of the correlation,the influencing factors are ranked,and then the main influencing factors are extracted. The Xgboost algorithm is used to build the electricity prediction model of different user categories,and to make the electricity prediction for different categories of users,so as to get the overall power consumption of the region. The actual power consumption data are used to adjust the algorithm parameters. The comprehensive method is simulated and compared with other methods. The results show that this method comprehensively considers many influencing factors,and divides different user categories and has higher accuracy and reliability. Xgboost algorithm has a good application prospect in regional electricity consumption forecast.
The forecasting of reginal power consumption is affected by many factors,and the behavior patterns of users are different,in accordance with these features,a comprehensive forecasting method based on Xgboost algorithm is proposed. With this method,the inflrencing factors of power consumption are considered from three aspects,i. e.,cycle,trend and disturbance,the historical power consumption data of users and the meteorological data are collected,then the missing values in these data are completmented,and smoothing process is conducted. The user data obtained are clustered by K-means to get the different categories of the users. Then,the maximal information coefficient( MIC) is used to calculate the correlation between each influencing factor and the electricity consumption of users. According to the magnitude of the correlation,the influencing factors are ranked,and then the main influencing factors are extracted. The Xgboost algorithm is used to build the electricity prediction model of different user categories,and to make the electricity prediction for different categories of users,so as to get the overall power consumption of the region. The actual power consumption data are used to adjust the algorithm parameters. The comprehensive method is simulated and compared with other methods. The results show that this method comprehensively considers many influencing factors,and divides different user categories and has higher accuracy and reliability. Xgboost algorithm has a good application prospect in regional electricity consumption forecast.
引文
[1]张东霞,苗新,刘丽平,等.智能电网大数据技术发展研究[J].中国电机工程学报,2015(1):2-12.
[2]曲朝阳,张率,刘洪涛.基于用电影响因素回归的小区用电预测模型[J].东北电力大学学报,2015(1):73-77.
[3]赵腾,王林童,张焰,等.采用互信息与随机森林算法的用户用电关联因素辨识及用电量预测方法[J].中国电机工程学报,2016,36(3):604-614.
[4]吴潇雨,和敬涵,张沛,等.基于灰色投影改进随机森林算法的电力系统短期负荷预测[J].电力系统自动化,2015,39(12):50-55.
[5]朱陶业,李应求,张颖,等.提高时间序列气象适应性的短期电力负荷预测算法[J].中国电机工程学报,2006,26(23):14-19.
[6]牛东晓,魏亚楠.基于FHNN相似日聚类自适应权重的短期电力负荷组合预测[J].电力系统自动化,2013,37(3):54-57.
[7]谷云东,张素杰,冯君淑,等.大用户电力负荷的多模型模糊综合预测[J].电工技术学报,2015,30(23):110-115.
[8]李媛,武岩岩,王思琪.基于混沌时间序列的Elman神经网络工业用电预测[J].沈阳工业大学学报,2016,38(2):196-200.
[9]石雪梅,葛斐,肖夕林.基于K-L信息量法的安徽省工业用电量预测[J].电网与清洁能源,2015,31(11):58-62.
[10]张淑清,师荣艳,张立国,等.混沌预测模型改进及在电力日负荷预测中的应用[J].仪器仪表学报,2016,37(1):208-214.
[11]HARTIGAN J A,WONG M A.A K-means clustering algorithm[J].Applied Statistics,1979,28(1):100-108.
[12]王天送,张杰,孙明明.拉伊达准则在交通调查数据处理中的应用[J].西部交通科技,2016(4):96-99.
[13]战泉茹.基于最大信息系数的人脸特征选择[D].长春:东北师范大学,2013.
[14]RESHEF D,RESSHEF Y,MITZENMACHER M,et al.Equitability analysis of the maximal information coefficient,with comparisons[J].Computer Science,2013(5):17-19.
[15]CHEN T,GUESTRIN C.XGBoost:a scalable tree boosting system[C]//ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.ACM,2016:785-794.
[16]张昊,纪宏超,张红宇.XGBoost算法在电子商务商品推荐中的应用[J].物联网技术,2017(3):102-104.
[2]曲朝阳,张率,刘洪涛.基于用电影响因素回归的小区用电预测模型[J].东北电力大学学报,2015(1):73-77.
[3]赵腾,王林童,张焰,等.采用互信息与随机森林算法的用户用电关联因素辨识及用电量预测方法[J].中国电机工程学报,2016,36(3):604-614.
[4]吴潇雨,和敬涵,张沛,等.基于灰色投影改进随机森林算法的电力系统短期负荷预测[J].电力系统自动化,2015,39(12):50-55.
[5]朱陶业,李应求,张颖,等.提高时间序列气象适应性的短期电力负荷预测算法[J].中国电机工程学报,2006,26(23):14-19.
[6]牛东晓,魏亚楠.基于FHNN相似日聚类自适应权重的短期电力负荷组合预测[J].电力系统自动化,2013,37(3):54-57.
[7]谷云东,张素杰,冯君淑,等.大用户电力负荷的多模型模糊综合预测[J].电工技术学报,2015,30(23):110-115.
[8]李媛,武岩岩,王思琪.基于混沌时间序列的Elman神经网络工业用电预测[J].沈阳工业大学学报,2016,38(2):196-200.
[9]石雪梅,葛斐,肖夕林.基于K-L信息量法的安徽省工业用电量预测[J].电网与清洁能源,2015,31(11):58-62.
[10]张淑清,师荣艳,张立国,等.混沌预测模型改进及在电力日负荷预测中的应用[J].仪器仪表学报,2016,37(1):208-214.
[11]HARTIGAN J A,WONG M A.A K-means clustering algorithm[J].Applied Statistics,1979,28(1):100-108.
[12]王天送,张杰,孙明明.拉伊达准则在交通调查数据处理中的应用[J].西部交通科技,2016(4):96-99.
[13]战泉茹.基于最大信息系数的人脸特征选择[D].长春:东北师范大学,2013.
[14]RESHEF D,RESSHEF Y,MITZENMACHER M,et al.Equitability analysis of the maximal information coefficient,with comparisons[J].Computer Science,2013(5):17-19.
[15]CHEN T,GUESTRIN C.XGBoost:a scalable tree boosting system[C]//ACM SIGKDD International Conference on Knowledge Discovery and Data Mining.ACM,2016:785-794.
[16]张昊,纪宏超,张红宇.XGBoost算法在电子商务商品推荐中的应用[J].物联网技术,2017(3):102-104.