A novel forecasting approach based on multi-kernel nonlinear multivariable grey model: A case report

Carbon emissions are an important environmental problem. The objective and accurate prediction of carbon emissions can serve as a reference and advance indicator for the implementation of a government’s environmental strategy. In this paper, based on a Gaussian vector basis kernel function and a global polynomial kernel function combined with the characteristics of grey prediction models, a new multi-kernel GMC(1,N) model is established that is more comprehensive and suitable for nonlinearity. It can enable more flexible improvements in prediction modelling accuracy. In this study, this new model is used to simulate the carbon dioxide emissions in Chongqing, China (one of the four municipalities under the direct control of the central government), from 2009 to 2015. Raw coal and cleaned coal, which have the most significant impact on carbon emissions, are selected to analyse the effectiveness of the model. The results show that the proposed model offers better simulation and prediction accuracy than four other models considered for comparison. Moreover, the carbon dioxide emissions of Chongqing in 2016–2020 are expected to be similar to those of the previous years. Therefore, to prevent a strong rebound of carbon emissions, it will be necessary to increase energy conservation and emission reduction efforts and to reduce energy consumption, especially coal consumption. •Gaussian and polynomial kernel functions are used to address nonlinearity in the proposed model.•Multiple kernel functions are organically combined with a grey prediction model.•A multi-kernel grey prediction model is established.•The carbon dioxide emissions of Chongqing in the next five years are forecasted, and some suggestions are proposed.