能源是人类社会赖以生存和发展的物质基础，在国民经济中具有重要的战略地位。随着社会经济的发展，能源的需求越来越大，因此需要对能源的使用进行有效的管理。能源消费预测是能源供需管理的理论前提，建立可靠的能源消费预测模型显得尤为重要。目前，已有的能源消费量预测模型主要包括单一模型和混合模型两大类。本研究提出了基于数据分组处理（group method of data handling，GMDH）的混合预测模型GHFM。该模型首先使用基于GMDH的自回归模型在原始能源消费时间序列上建模，预测其线性趋势，并得到残差序列（非线性子序列）。考虑到非线性子序列预测的复杂性，分别建立BP神经网络、支持向量回归机、遗传规划和RBF神经网络模型，再运用GMDH在非线性子序列上建立选择性组合预测模型，得到非线性子序列的组合预测值。最后，将两个部分的预测值进行整合得到总的能源消费量预测值。选取中国统计年鉴2014能源统计数据中的中国能源消费总量和石油消费总量数据进行实证分析，结果表明，GHFM模型与其他模型相比具有更好的预测效果。最后，给出了使用GHFM模型对2015-2020年中国能源消费总量的样本外预测值。
Energy is the material basis for the survival and development of human society and it plays an important strategic role in national economy.With the development of society, it shows an increasing demand of energy, so effective management is important for energy use. Energy consumption prediction is the theory premise of energy supply and demand management, therefore, establishing a reliable energy consumption forceasting model is particularly significant. At present, the existing models mainly includes two types:single prediction models and hybrid prediction models.In this study' a group method of data handling (GMDH) based hybrid forecasting model (GHFM) is proposed for China's energy consumption prediction. In this model, GMDH based auto-regression is first constructed in the original energy consumption time series, the linear trend of time series is predicted, and residual series (i.e., non-linear sub-series) is obtained. Considering the complexity of non-linear sub-series prediction, BP neural network, support vector regression, genetic programming and RBF neural network are constructed respectively, and then combined forecasting model is condueted selectively by GMDH in non-linear sub-series and the combination prediction are obtained for the non-linear sub-series. At last, the total energy consumption forecast values are integrated from the two parts above. Empirical analysis is conducted on the total energy consumption and oil consumption time series from China Statistical Yearbook of energy statistics data (2014) and the results show that the forecast performance of GHFM model is better than other models. In addition, the out-of-sample forecasts of China's total energy consumption from 2015 to 2020 is given based on GHFM model. The proposed model can be used to improve the accuracy of energy consumption forecasting. The model can also be applied to other time series forecasting problems, including container throughput forecasting, crude oil price forecasting, electricity load forecasting, stock price forecasting, et al.
[1] Zhu Z X. China statistical yearbook[M]. Beijing:China Statistics Press,2013.
[2] Ahmad A, Hassan M Y, Abdullah M P, et al. A review on applications of ANN and SVM for building electrical energy consumption forecasting[J]. Renewable and Sustainable Energy Reviews, 2014, 33(5):102-109.
[3] Yu Shiwei, Wei Yiming, Wang Ke. China's primary energy demands in 2020:Predictions from an MPSO-RBF estimation model[J]. Energy Conversion and Management, 2012, 61:59-66.
[4] 杨波, 郭剑川, 谭章禄. 基于国民生产总值增长率微调制的国家能源年度消费总量Logistic修正模型研究[J]. 中国管理科学, 2017,25(6):32-38.
[5] Zeng Chunlei, Wu Changchun, Zuo Lili, et al., Predicting energy consumption of multiproduct pipeline using artificial neural networks[J]. Energy, 2014,66:791-798.
[6] Pindyck R S, Rubinfeld D L. Econometric models and economic forecasts[M]. Boston:McGraw-Hill Boston, 1998.
[7] 曾波, 刘思峰, 曲学鑫. 一种强兼容性的灰色通用预测模型及其性质研究[J]. 中国管理科学, 2017, 25(5):150-156.
[8] 杨保华, 赵金帅. 优化离散灰色幂模型及其应用[J]. 中国管理科学, 2016, 24(2):162-168.
[9] Erdogdu E. Electricity demand analysis using cointegration and ARIMA modelling:A case study of Turkey[J]. Energy Policy, 2007, 35(2):1129-1146.
[10] Nilsson N J. Principles of artificial intelligence[M]. San Francisceo:Morgan Kaufmann, 2014.
[11] 胡雪棉, 赵国浩. 基于Matlab的BP神经网络煤炭需求预测模型[J]. 中国管理科学, 2008, 16(S1):512-525.
[12] 卫敏, 余乐安. 具有最优学习率的RBF神经网络及其应用[J]. 管理科学学报, 2012, 15(4):50-57.
[13] 张冬青, 马宏伟, 宁宣熙. 基于结构可变的RBF神经网络的时间序列预测[J]. 中国管理科学, 2010, 18(3):83-89.
[14] 彭建良, 李新建. 能源消费量模拟分析和预测的神经网络方法[J]. 系统工程理论与实践, 1998, 18(7):76-83.
[15] Lu C J, Lee T S, Chiu C C. Financial time series forecasting using independent component analysis and support vector regression[J]. Decision Support Systems, 2009, 47(2):115-125.
[16] Brereton R G,Lloyd G R. Support vector machines for classification and regression[J]. The Analyst,2009,135(3):230-287.
[17] Kavaklioglu K. Modeling and prediction of Turkey's electricity consumption using support vector regression[J]. Applied Energy, 2011, 88(1):368-375.
[18] 陈荣, 梁昌勇, 谢福伟,等. 基于自适应GA-SVR的旅游景区日客流量预测[J]. 中国管理科学, 2012,20(S1):61-66.
[19] Kova?i? M, Šarler B. Genetic programming prediction of the natural gas consumption in a steel plant[J]. Energy, 2014, 66:273-284.
[20] Lee D G, Lee B W, Chang S H. Genetic programming model for long-term forecasting of electric power demand[J]. Electric Power SystemsResearch, 1997, 40(1):17-22.
[21] Wang Shuai, Yu Lean, Tang Ling, et al. A novel seasonal decomposition based least squares support vector regression ensemble learning approach for hydropower consumption forecasting in China[J]. Energy, 2011, 36(11):6542-6554.
[22] Ivakhnenko A G. Polynomial theory of complex systems[J]. IEEE transactions on Systems, Man and Cybernetics, 1971, 1(4):364-378.
[23] Xiao Jin, Xie Ling,He Changzheng,et al. Dynamic classifier ensemble model for customer classification with imbalanced class distribution[J]. Expert Systems with Applications, 2012, 39(3):3668-3675.
[24] Xiao Jin, Xiao Yi,Huang Anqiang, et al. Feature-selection-based dynamic transfer ensemble model for customer churn prediction[J]. Knowledge and Information Systems, 2015, 43(1):29-51.
[25] Xiao Jin,Jiang Xiaoyi,He Changzheng, et al. Churn prediction in customer relationship management via GMDH-based multiple classifiers ensemble[J]. IEEE Intelligent Systems, 2016, 31(2):37-44.
[26] Xiao Jin,He Changzheng,Jiang Xiangyi, et al. A dynamic classifier ensemble selection approach for noise data[J]. Information Sciences, 2010, 180(18):3402-3421.
[27] Xiao Jin, He Changzheng, Jiang Xiaoyi. Structure identification of bayesian classifiers based on GMDH[J]. Knowledge-Based Systems, 2009. 22(6):461-470.
[28] Xiao Jin, Cao Hanwen,Jiang Xiaoyi, et al. GMDH-based semi-supervised feature selection for customer classification[J]. Knowledge-Based Systems, 2017, 132(9):236-248.
[29] 贺昌政. 自组织数据挖掘与经济预测[M]. 北京:科学出版社, 2005.
[30] Xiao Jin, Sun Haiyan,HuYi, et al. GMDH based auto-regressive model for China's energy consumption prediction[C]//Proceedings of 2015 International Conference on Logistics, Informatics and Service Sciences,Barelona,Sipain,July,27-29. 2015.
