京津冀能源需求预测分析及发展对策研究

发布时间：2018-01-06 04:35

本文关键词：京津冀能源需求预测分析及发展对策研究　出处：《天津理工大学》2017年硕士论文　论文类型：学位论文

【摘要】：能源是经济发展和社会进步不可或缺的重要物质基础,是人类各种活动顺利进行的保证。京津冀协同发展是国家重大发展战略,也是国家经济发展的重要核心区域。在经济快速发展的同时,能源的保证是必不可少的。能源需求预测是能源规划和政策制定的基础,分析京津冀区域能源需求有助于政府准确制定能源政策,保障京津冀经济可持续健康发展。能源需求预测方法很多,这些方法考虑问题角度不同,各有其优缺点,但它们之间不是相互排斥的,而是相互兼容、相互补充的。因此,Bates和Grange提出了组合预测思想。组合预测方法能够有效保留单项预测方法的有价值的数据信息,相比单一预测方法,具有更高的预测精度,并可以增强预测的稳定性。然而目前的能源需求组合预测文献中,对于如何选取单一预测方法,单一预测方法的数量没有依据,且组合的形式大都停留在线性组合或者是简单的非线性组合形式,且非线性组合的权系数确定困难,计算复杂。针对这些问题,本文做了以下研究工作:第一,本文系统研究了能源需求的影响因素,分析了经济增长、产业结构、人口与城市化、能源消费结构、技术进步、居民消费水平、能源价格及环境政策等影响因素与能源需求的相关关系。第二,本文提出了选取单一预测方法的基本原则,即主观与客观相结合、线性模型与非线性模型相结合、单一预测方法的数量要适度等原则。基于以上原则本文选取了ARIMA模型(客观线性模型)、灰色预测方法(客观非线性模型)、多元回归(主观线性模型)、二次非线性能源需求预测模型(主观非线性模型)及灰色神经网络模型(主观非线性模型)等5种预测方法,并对模型进行了改进。第三,本文提出了基于BP神经网络的非线性组合预测模型。BP神经网络是非线性映射模型,权重可以在网络训练中确定,避免了计算权重困难的问题。第四,针对神经网络收敛速度慢,易收敛到局部极值等问题,采用混沌遗传算法对神经网络性能进行优化。混沌遗传算法具有收敛速度快、全局搜索能力强、适合参数优化等特点,通过混沌遗传算法优化神经网络,能大幅提升网络性能,从而提高预测精度。最后,利用组合模型对京津冀能源需求进行预测,2015年至2020年京津冀能源需求总量分别为4.48、4.62、4.75、4.90、5.04、5.17亿吨标准煤,并给出了京津冀能源发展对策建议。
[Abstract]:Energy is an important indispensable material basis for economic development and social progress, is the guarantee for human activities. The Beijing Tianjin Hebei coordinated development is a major national development strategy, but also the national economic development an important core area. In the rapid economic development, energy guarantee is essential. The energy demand forecast is the basis of energy planning and policy the analysis of Beijing Tianjin Hebei regional energy demand will help the government to develop accurate energy policy, protection of Beijing Tianjin Hebei economic sustainable and healthy development. Many methods of energy demand forecasting, these methods consider the problem from different angles, each has its advantages and disadvantages, but they are not mutually exclusive, but are compatible with each other, complement each other. Therefore, Bates and Grange put forward the thought of combination forecasting. The combination forecasting method can effectively preserve the prediction method of single valuable data information, compared to a single The prediction method has higher prediction accuracy, and can enhance the prediction stability. However, the energy demand forecasting literature, how to choose a single prediction method, the number of single forecasting methods without the basis, and the combination of the form of stay in linear combination or nonlinear combination of simple form, and the weight coefficient of nonlinear the combination was difficult to determine, the complexity of the computation. To solve these problems, this paper does the following research work: first, this paper studies the factors affecting energy demand analysis, economic growth, industrial structure, population and city, energy consumption structure, technological progress, consumption level, energy prices and environmental policy related effects the factors and energy demand. In second, this paper puts forward the basic principles of selecting a single prediction method, which is a combination of subjective and objective, linear model and nonlinear model. With the number of single forecasting methods to moderate principles. These principles were selected based on the ARIMA model (objective linear model), grey prediction method (objective nonlinear model), multivariate regression (subjective linear model), two nonlinear energy demand forecast model (subjective nonlinear model) and grey neural network model (subjective nonlinear model 5) prediction method, and the model is improved. Third, this paper proposes a nonlinear BP neural network combination forecasting model based on.BP neural network is a nonlinear mapping model, the weights can be determined in network training, avoiding the problem of weight calculation difficult. Fourth, according to the neural network slow convergence speed, easy convergence to the local extremum problem, using chaos genetic algorithm to optimize the performance of neural network. The chaos genetic algorithm has fast convergence and global search ability, suitable for parameter optimization So, through the optimization of neural network chaotic genetic algorithm, can greatly enhance the performance of the network, so as to improve the prediction accuracy. Finally, forecast of the energy demand of Beijing Tianjin Hebei using the combination model, the total energy demand from 2015 to 2020 in Beijing Tianjin Hebei were 4.48,4.62,4.75,4.90,5.04,5.17 million tons of standard coal, and gives the Beijing Tianjin Hebei energy development countermeasures.

【学位授予单位】：天津理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F426.2

【相似文献】