首页|Comparative analysis of deep learning and classical time series methods to forecast natural gas demand during COVID-19 pandemic
Comparative analysis of deep learning and classical time series methods to forecast natural gas demand during COVID-19 pandemic
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
Taylor & Francis
The lockdown measures implemented to contain the COVID-19 pandemichave had a considerable effect on the consumption of natural gas, which isclosely linked to the economic growth of countries. Accurately forecastingnatural gas demand is critical for making informed decisions in unprecedentedand unexpected situations. This study aims to compare artificiallearning-based algorithms and classical statistical time series models in predictingnatural gas demand during the pandemic, using Turkey as a casestudy. Common time series prediction methods, including AutoregressiveIntegrated Moving Average (ARIMA), Nonlinear Autoregression NeuralNetwork (NARNN), Support Vector Regression (SVR), and Long Short-TermMemory (LSTM), were utilized for this purpose. The impact of the pandemicon natural gas demand was analyzed by including 2-year natural gas consumptiondata since its onset. Root mean square error (RMSE), correlationcoefficient (R), and mean absolute error (MAE) criteria were used as performanceevaluation metrics to select the best model. The results confirmedthat the deep-learning-based LSTM model provided better prediction accuracythan time-series benchmark models, with the lowest RMSE (9.442) andthe highest R (0.997) values in the test dataset. Furthermore, the results werevalidated by statistical analysis using the Diebold-Mariano and Nemenyitests.
COVID-19 pandemicnatural gas consumptionenergytime-seriesdeep learningprediction
Zeynep Ceylan
展开 >
Faculty of Engineering, Industrial Engineering Department, Samsun University, Samsun, Turkey
NETL
NSTL
Taylor & Francis
