中国非居民用户用气量的行业异质性动态分析及预测——基于一种新的综合集成分析预测模型框架

Dynamic Analysis and Forecast of Industry Heterogeneity of Natural Gas Consumption by Chinese Non-Resident Users——Based on a New Integrated Analysis and Forecasting Model Framework

能源革命背景下的能源转型是一项长期战略任务,在化石能源短期内仍占主导地位的情景下,“化石能源清洁化”将是中国“十四五”期间能源系统转型的重点.清洁、高效的天然气是调整化石能源消费结构的关键,在保证居民用气的基础上,主要非居民用气行业的燃气消费是天然气市场的重点,但不同行业的天然气消费状况具有阶段性和异质性等特点.文章采用一种新的综合集成分析-预测思路框架,针对中国主要3个非居民用气行业-天然气发电、天然气化工、工业燃料分别进行研究.首先,采用PPM-KM模型和马尔科夫区制转移模型(MS-AR)对各行业天然气消费序列的突变特征和阶段特征进行处理和刻画,发现中国各行业的天然气消费量均存在“低速增长”和“高速增长”两种不同的波动状态,且在同一时间段内各行业天然气消费的波动状态及状态转移特征均存在明显的异质性.其次,采用贝叶斯模型平均(BMA)分行业、分区制对天然气消费的影响因素进行筛选,结果表明不同行业间、同一行业不同波动状态间的天然气消费影响因素存在显著差异.接下来,基于MS-AR模型的状态划分以及BMA模型的影响因素筛选结果,采用情景分析法对各行业的天然气消费进行预测,结果表明未来各行业的天然气消费预测结果虽有差异,但均呈现出明显的季节性和稳定上升的发展态势.文章针对不同行业的天然气消费所具有的异质性、阶段性的特征,提供了一种精细化的分析和预测方法,是对以往的天然气消费分析和预测方法的一种改进.

"Clean fossil energy" is the focus of Chinese energy system transformation.Clean and efficient natural gas is the key to adjust the fossil energy consumption structure.Accurate consumption analysis and prediction will provide an important reference for China’s energy transformation.This paper reviews the domestic and foreign literature on natural gas consumption forecast and the influencing factors of natural gas consumption.In recent years,more and more studies have adopted multiple methods for integrated prediction.In addition,a large number of literatures point out that the factors influencing natural gas consumption are mainly related to economic development level,energy consumption structure,natural gas price and alternative energy price.In this paper,a new integrated analysis-prediction framework is employed to study three non-residential gas industries in China:Natural gas power generation,natural gas chemical industry and gas for industrial fuel.First,this paper uses a product partition-Kmeans model(PPM-KM) to identify and eliminate the change points of natural gas consumption sequence in different industries to eliminate obvious abnormal points.Second,we employ Markov switching model(MS)to examine the regime switching of natural gas consumption in different industries.Third,the Bayesian model average(BMA) method is employed to determine the core determinants of natural gas consumption in different industries under sub-regimes and full sample.Finally,based on the above analysis results,the scenario analysis method is used to predict the natural gas consumption.The results show that there are obvious heterogeneity in state transfer characteristics of natural gas consumption in different industries of China,and there are significant differences in the influencing factors under different regimes and periods.The predicted value of natural gas consumption in each industry shows an obvious upward trend and seasonality,and the predicted results also have industry heterogeneity.