Compositional data techniques for forecasting dynamic change in China's energy consumption structure by 2020 and 2030

China has adopted legislations for accelerating the non-fuel energy restructuring with explicit targets that need to be achieved by 2020 and 2030. The joint forecasting and systematic analysis of the integral energy consumption structure face great challenges due to the intrinsic features of a structure, namely the positive and constant-sum (e.g., 100% in percentage) constraints of the components. To examine the feasibility of the policy, this study constructs comprehensive time series forecasting models, where a novel data source, referred to as compositional data, is employed. Compositional root mean squared error (CRMSE) and mean absolute percentage error (CMAPE) are used to evaluate the forecasting accuracy of the model. Compared with the traditional forecasting method, the proposed compositional method indicates a better forecasting performance with lower CRMSE and CMAPE values. The results obtained by the best-performing model are compared with the plausible goals of the government. Based on the forecasted results, several different findings from the traditional forecasting models along with related policy recommendations are derived. The proposed compositional methods show different trajectories of China's energy consumption structure by 2030, indicating the sharper decrease of coal consumption and greater development potential of China's non-fossil energy consumption than the traditional forecasting models. The forecasting model based on compositional data possesses considerable potential for energy structure forecasting and can be considered as a viable alternative. Practicable and concrete recommendations are provided to ensure the feasibility of the policy.