我国能源需求预测中的动态模糊系统建模方法研究

项目名称： 我国能源需求预测中的动态模糊系统建模方法研究

项目编号： No.71201019

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 管理科学与工程

项目作者： 宋雯彦

作者单位： 东北财经大学

项目金额： 19万元

中文摘要： 能源需求的预测对我国社会经济的发展具有重要意义。现有的建模方法多需依赖较大规模的历史信息来确定模型参数，或对已有数据的统计特征做出某些预设,而这在实际的应用研究中常常得不到满足。本项目首先提出一种新的基于模糊系统理论的边缘线性化建模方法，从已知数据中提炼模糊推理规则，建立动态系统的非线性变参数模型；其次，分别就所得的连续型状态空间微分方程模型和离散型非线性时间序列模型，实现模型的结构辨识和参数估计，进而研究模型的逼近性能；第三，在建模数据充足或数据得以更新的情况下，利用神经网络进一步训练优化所建模型的结构和参数，从而提高模型的预测精度；最后，结合模糊聚类和因果检验，将该建模方法运用于我国能源需求预测的实证分析中。本课题的研究丰富了动态非线性系统的建模和分析预测途径，其在能源需求预测中的应用结果，有望为我国能源发展战略的调整转型提供一定的参考。

中文关键词： 动态系统；模糊推理；神经网络；能源需求；

英文摘要： Energy demand forecasting is essential for social and economic development in our country. There are many modeling methods to handle practical estimation problems, but most of them need a large amount of information in advance to determine parameters in the models, or require that available data should be in agreement with some statistic assumption. Based on fuzzy system theory a novel marginal linearization modeling method is studied in this project. First, fuzzy inference rules can be extracted from known data, and a kind of non-linear models with variable parameters are obtained for dynamic systems. Second, for state-space differential equation models in continuous situation and for time series models in discrete situation respectively, the procedure of structure identification and parameter estimation are given. Third, when modeling data are sufficient enough or can be refreshed in real time, neural networks are exploited to train and adjust the structure or parameter in the models to increase the accuracy of forecasting precision. In the end, this integrated modeling method, combined with clustering method and causality estimation technique, is used in the empirical analysis for energy demand in China. The modeling method discussed in this project enriches modeling and forecasting methods. Further, the impl

英文关键词： dynamic system；fuzzy inference；neural network；energy demand；