A study of industrial electricity consumption based on partial Granger causality network

作者：Yao, Can-Zhong^*; Lin, Qing-Wen; Lin, Ji-Nan

来源：Physica A: Statistical Mechanics and Its Applications , 2016, 461: 629-646.

DOI：10.1016/j.physa.2016.06.072

• Granger causality network
• Partial Granger causality network
• Bootstrap
• Complementary network
• Industrial influence

The paper studies the industrial energy transferring paths among the industries of China by distinguishing direct causality from the indirect. With complementary graphs, we propose that industrial causal relationship can be heterogeneous, and provide insights for refining robust industrial causality framework.
First, by analyzing the in-weight and out-weight of the industries in Granger causality networks we find that public utilities have significant causality with other industries, and the industries with higher degree value tend to have stronger causality with others.
Further, we eliminate the exogenous links by partial Granger causality model and find both Granger and partial Granger networks have consistent hub industries while some outliers emerge in partial Granger causality networks. Besides, compared with GX, GZ, HN and YN, the correlation between the volume of electricity consumption and the weight of each industry is more significant in the networks of GD and NF. By studying the characteristics of complementary graphs, we show that the industrial energy transferring paths in GD are more multidimensional, and the corresponding interdependent relationship among industries is more robust.
Finally, using bootstrap method we verify the reliability of each industrial relationship network. Results exhibit that GD, GX and NF have more reliable causal relationship networks than other provinces, revealing their industrial structure to be more stable.

• 出版日期2016-11-1
• 单位华南理工大学