<p>Properly adjusting provincial fossil energy consumption standards is an effective way to promote high-quality economic development in China and achieve improvement of the ecological environment. Existing centralized DEA methods available to address this problem suffer from inaccurate characterization of bad outputs and lacking consideration of heterogeneity in regional productivity performance in future periods. This paper proposes an improved centralized DEA, which includes three steps: productivity performance evaluation, time series forecasting and resource allocation. First, a globally benchmarked directional distance function was used to evaluate the productivity performance of each province from 2005 to 2021. Secondly, several time series forecasting models were compared to obtain the best prediction values for the productivity performance of each province in 2022. Finally, the optimal allocation strategy for provincial fossil energy consumption in 2022 was calculated based on the predicted productivity performance and the weak disposability of bad outputs. We found that China’s provincial productivity performance is showing a polarized trend, with the productivity performance of central provinces gradually deteriorating. Fossil energy should be tilted towards some eastern regions with scarce energy reserves and the vast majority of economically weak central, western and northeastern provinces. The optimal allocation strategy may slow down the real GDP growth rate from 1.55% to 1.07%, but in return it will reduce carbon dioxide emissions by 131,193.92 million tons and sulfur dioxide emissions by 184,164.45 tons. Finally, policy implications are given based on the differentiated motive of fossil energy consumption demand in each province.</p>