Time series analysis and forecasting of China’s energy production during Covid-19: statistical models vs machine learning models

Abstract

COVID-19 is a huge catastrophe of global proportions, and this catastrophe has had far-reaching effects on energy production worldwide. In this paper, we build traditional statistical models and machine learning models to forecast energy production series in the post-pandemic period based on Chinese energy production data and COVID-19 Chinese epidemic data from 2018 to 2021. The experimental results showed that the optimal models in this study outperformed the baseline models on each series, with MAPE values less than 10. Further studies found that the LightGBM, NNAT and LSTM machine learning models worked better in unstable energy series, while the ARIMA statistical model still had an advantage in stable energy time series. Overall, the machine learning models outperformed the traditional models during COVID-19 in terms of prediction. Our findings provide an important reference for energy research in public health emergencies, as well as a theoretical basis for factories to adjust their production plans and governments to adjust their energy decisions during COVID-19.

Deep Learning Model Effectiveness in Forecasting Limited-Size Aboveground Vegetation Biomass Time Series: Kenyan Grasslands Case Study

“…For instance, [19] found that basic statistical models performed better than machine learning (ML) models and indicated that complex models are not necessarily superior to simple forecasting models [19]. Other research has demonstrated the superiority in predictive capabilities of statistical models over ML models with respect to energy production [38], financial market data [57], health, and cryptocurrency [58].…”

“…The disappointing performances of the DL models might be related to specific characteristics of time series data, such as periodicity, seasonality, and noise [19,21,38], along with factors such as the dataset size [19,37] and model parameters [58]. Since ML models involve multiple nonlinear transformations, it would seem that they should prove superior to statistical models; however, evidence from previous studies has indicated that ML models cannot be generalized from small datasets-a limitation relative to traditional statistics.…”

“…Widely used as benchmarks [14,31], these models enable meaningful comparisons. Focused on a specific problem domain with proven effectiveness [22,37,38], the selected DL models, including MLP for simplicity, LSTM and GRU for sequential modeling, and CNN for spatial dependencies, offer a comprehensive evaluation. The hybrid CNN-LSTM model integrates spatial and temporal characteristics for a robust assessment [31].…”