Conditional density forecast of China’s energy demand via QRNN model

ABSTRACT

We propose a new method for conditional density forecast of China’s energy demand through quantile regression neural network (QRNN). This method has at least two advantages. First, it is flexible to explore the true nonlinearity in the energy demand system via neural network structure. Second, it is able to describe the whole conditional distribution of energy demand via quantile regression. In the empirical study on China’s energy demand, QRNN outperforms several classical methods in terms of forecast accuracy both in-sample and out-of-sample. Considering China’s economic and social environment, we set a scenario for predictors and forecast the conditional density of China’s energy demand from 2015 to 2020. The empirical results show that the conditional density curve moves to right and its dispersion increases over time, which indicates that the energy demand in China will keep growing with an average annual rate of 9.672% and its uncertainty is enlarged with 42.210%.

KEYWORDS:

• Energy demand
• neural network
• quantile regression
• QRNN
• conditional density forecast

JEL CLASSIFICATION:

• C45
• C53

Acknowledgements

The authors would like to thank the editor and anonymous reviewers for their helpful recommendation and comments. The second author also would like to thank Prof. Cannon for sharing their R codes of the QRNN approach.

Disclosure statement

No potential conflict of interest was reported by the authors.