A Novel Robust Grey Model and Its Application

Abstract

Previous studies paid attention to improving the predicted capability of the classical grey model, but its robustness is still unclear and not explored, in particular when these exhibit outliers in the time series, which is due to measurement error, uncorrected record, and censored date. In this study, we proposed a novel robust grey model. The novel robust grey model adopts the median regression method to address these problems caused by outliers, which provides the robust parameters. The analytical expression for the time response function and the forecasting values is derived by the grey system technique and mathematical tool. With annual observational data of Chinese electricity demand, we examine the fitness capability of the novel robust grey model, by comparing it with the classical grey model. Also, we adopt the bootstrapping test to further illustrate the sensitivity for the new robust grey model when there are outliers in the time series. To our knowledge, it is the first to introduce the bootstrapping test to the literature related to the grey model and to focus on the robustness of the grey model. The computational results suggest that the new robust grey model has higher precision than the classical grey model, but it is also very robust to outliers, whose accuracy and robustness are better than the classical grey model. Finally, we apply the novel grey model to forecast the future values in Chinese electricity demand during the year 2022 to 2025. This new model proposed in this study estimates that the Chinese electricity demand would continue to increase after the year of 2022, arriving at 10.446×105 million kW·h in the year of 2025 approximately.