中国提出能源消费总量和强度双控制，优化能源结构具有非常紧迫的现实意义。现有省际能源消费的空间面板模型未考虑省际能源消费复杂的动态空间关联特征（动态空间权重）和明显的结构性差异特征（变系数）。本项目针对上述不足，将非参空间权重函数和变系数函数引入空间面板模型，构建变系数非参空间面板数据模型。首先，运用序列估计方法与局部多项式方法来转换空间权重函数和变系数函数，寻找工具变量与二阶矩条件，构建非参广义矩估计量推断空间相关系数和变系数；其次，构造广义似然比检验和LM检验相关统计量，分别检验空间相关性和变系数特征；再次，探讨估计量的大样本性质，通过蒙特卡洛模拟来检验其有限样本的性质；最后，应用该模型分析我国省际能源消费空间关联变动与结构差异。本项目在动态空间权重和变系数两方面拓展了现有空间面板模型，并且应用到我国省际能源消费现实问题中，为国家制定差异化省际能源发展规划提供理论依据和政策参考。

It has a very urgent practical significance that China put forward the implementation of the dual control of total and intensity energy consumption, optimizing of energy structure. The existing literature using spatial panel data model to study the inter-provincial energy consumption take into account neither the complex dynamic spatial correlation characteristics (dynamic spatial weight) of inter-provincial energy consumption, nor manifest structural differences (varying-coefficient). In this project, the non - parametric spatial weight function and varying coefficient function are introduced into the spatial panel data model for the above-mentioned problems of the existing literature, and the varying-coefficient nonparametric spatial panel data model is constructed. Firstly, we use the sequence approximation method and local polynomial method to transform the spatial weights function and varying-coefficient function. We find the instrumental variable and the second moment condition to construct the non-parametric generalized moment estimators to infer the spatial correlation coefficients and the varying-coefficient. Secondly, Generalized Likelihood Ratio test and LM test statistics are constructed, respectively, to test the spatial correlation and varying coefficient characteristics. Thirdly, we explore the asymptotic properties of the proposed estimators, and run Monte Carlo simulation to test the nature of its limited sample properties. Finally, application of the model analyzes the dynamic spatial correlations and structural differences of provincial energy consumption in china. In this project, the existing spatial panel model is extended in both dynamic spatial weight and varying-coefficient, and it has an empirical application of inter-provincial energy consumption problems in China. It provides scientific basis and policy reference for the government to make the inter-provincial energy development plan in the light of local conditions.