Prediction model for China's monthly LNG ex-factory prices based on BP-ARIMA
Objective China is the world's largest importer of natural gas, with LNG playing a crucial role in natural gas supply. However, fluctuations in LNG ex-factory prices introduce significant uncertainty. Accurately predicting these price trends is vital for optimizing the LNG industry chain layout...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | zho |
| Published: |
Editorial Office of Oil & Gas Storage and Transportation
2024-10-01
|
| Series: | You-qi chuyun |
| Subjects: | |
| Online Access: | https://yqcy.pipechina.com.cn/cn/article/doi/10.6047/j.issn.1000-8241.2024.10.010 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Objective China is the world's largest importer of natural gas, with LNG playing a crucial role in natural gas supply. However, fluctuations in LNG ex-factory prices introduce significant uncertainty. Accurately predicting these price trends is vital for optimizing the LNG industry chain layout and enhancing the economic efficiency of the natural gas supply chain. Methods Historical data on China's LNG ex-factory prices were collected and analyzed, identifying key factors that influence price changes through grey relational analysis. On this basis, BP neural network prediction model and ARIMA time series prediction model were established, respectively. According to the variable weight theory, a new variable-weight BP-ARIMA combined prediction model was established through the weighted combination of the two individual prediction models. Case verification analysis was carried out for different prediction models based on China's actual LNG ex-factory prices. Results The variable-weight BP-ARIMA combined model integrated the advantages of the traditional BP neural network and ARIMA models. By dynamically adjusting the weight ratio between the two, it significantly enhanced the accuracy of LNG ex-factory price predictions. The mean absolute error (MAE), mean absolute percentage error (MAPE), and root-mean-square error (RMSE) were RMB 188.7/t, 4.1%, and RMB 280.5/t, respectively. Compared with the BP neural network model, ARIMA model, and equal-weight combined model, the MAE was reduced by 65.85%, 44.20% and 37.50%, the RMSE by 63.6%, 38.2% and 29.8%, and the MAPE by 63.7%, 42.3% and 36.9%, respectively. Conclusion The variable-weight BP-ARIMA combined prediction model proposed in this study offers an effective solution for predicting China's LNG ex-factory prices, facilitating the orderly development of the LNG market and providing support for decision-making in natural gas supply chain management. |
|---|---|
| ISSN: | 1000-8241 |