Study Of Energy Prices' Fluctuation And Spillover Mechanism Based On Time Series Network

The world energy price system is a complex system involving diversified participants.Different regions and different types of energy market prices have different behavior characteristics,and energy prices interact with each other to form a variety of spillover relationships,thus making the fluctuation mechanism of energy prices more complex.Based on the modeling of time series network and the analysis of network topology,the behavior characteristics of energy price fluctuations and the correlation mechanism of volatility spillovers can be clearly characterized and revealed,which is hardly achieved by ordinary mathematical statistics and qualitative methods.Therefore,this paper attempts to construct the improved visibility graph model,the partial Granger causality network,the multi-scale grey correlational patterns network and the timevarying Bayesian patterns networks.The main purpose is to reveal the fluctuation characteristics and evolution rules of crude oil,natural gas,renewable energy and other different energy prices and seek the spillover mechanism and linkage characteristics between energy price fluctuations,so as to promote the diversification and low-carbon development of energy structure from the perspective of energy price fluctuation mechanism.The main research contents and innovative research results of this paper are mainly reflected in the following four aspects:(1)As a kind of clean fossil energy,natural gas is becoming more and more important in the energy consumption structure of various countries.Influenced by resource endowments and storage and transportation facilities,the world natural gas price system has obvious regional characteristics.In this paper,based on the existing visibility graph model,the parametric modified limited penetrable visibility graph model is constructed to preserve the dynamics of the sequence to the greatest extent.Then,the empirical analysis is carried on the price series of Japan LNG,Europe natural gas and North America Henry Hub respectively.Through the evaluation and comparison of the average degree,the average linkage length,the cumulative degree distribution and other network indicators,combined with the analysis of resource endowments,pricing mechanism and pipe network facilities,the characteristics of the three representative natural gas prices are revealed and identified.(2)As natural gas has a strong substitution effect on crude oil,a lot of empirical research has been carried out on the relationship between the prices of natural gas and crude oil.However,due to the different data selection and research methods,the research results have obvious differences.In this paper,the well-known Henry Hub spot price of natural gas and WTI spot price of crude oil are selected as sample data,and a multi-time scale grey correlation model network is proposed.In order to reveal the volatility spillover mechanism of natural gas price and crude oil price from three aspects: short-term market supply and demand imbalance,major events and policy factors,and long-term trend,the model first decomposes the data,and the gray correlation series is used to describe the dynamic correlation between natural gas and crude oil prices in different time scales,and then the gray-correlation patterns network is constructed by coarse grain method.Through the node's strength centrality,edges' conversion ability coefficient and pattern's self-transformation ability coefficient,this paper has identified the important correlation patterns,the crucial conversion between patterns and the patterns with larger self-transition capability,which provides a key reference for decision makers to grasp the relationship between natural gas and crude oil market and make investment portfolio decisions.(3)Fossil energy is exhausting and highly polluting.Although natural gas is an important resource in the transition of energy structure,renewable energy is the core of energy transformation in the world.In recent years,the world's demand for renewable energy has continued to grow.There are many renewable energy listed companies in the stock market,and the stock prices of these companies are interrelated each other,forming a complex stock price network.This paper selected 79 China's PV listed companies and intercepted the stock closing price of 2007.10.02-2016.10.03 as sample data.The dependence between the two companies' stock prices is described by Granger causality and partial Granger causality respectively,then Granger causality network model and partial Granger causality network model are constructed.The empirical results show that,since partial Granger causality can eliminate the influence of other companies in the network and reflect the direct dependence between companies,the partial granger network model can more truly reflect the characteristics and evolution rules of the interdependence between the stock prices of China's PV enterprises.Based on the analysis of network indicators such as strength,betweenness,clustering coefficient and cumulative strength distribution,it identifies and reveals the most important enterprises,the evolution between corporates' influence and conduction,and the transmission effect of stock price fluctuations in China's PV industry.In the process of analysis,factors such as PV support policies,industrial chain structure and geographical distribution are combined.(4)As an effective tool of energy saving and emission reduction,carbon market is emerging in more and more countries and regions.Since fossil energy is the main source of carbon emissions,there are complex transmission and spillover effects between carbon price and fossil energy prices.In this paper,four futures prices are selected,namely EUA,Brent crude oil,natural gas of IPE and Rotterdam coal,and the sampling period is 2009.06.02-2018.09.18.On the one hand,in order to reveal the causality between the prices of carbon and the fossil energy in EUETS II and III,this paper applies Bayesian network and the cross-correlation coefficient to identify the Bayesian causality and the lead/lag relationship between the prices.On the other hand,in order to reveal the dynamic evolution characteristics and rules of the causality between three fossil energy prices and carbon prices over time,a time-varying Bayesian pattern network model is constructed,based on the analysis of nodes' weighted degree and the accumulative distribution of weighted degree,the important Bayesian patterns and the dynamic evolution process of the causal relationship between four prices are identified.