For the prediction of interval grey numbers, the prediction model based on the kernel sequence of interval grey number is constructed, and the idea of information domain is expanded based on residual corrections in this paper. To be specific, the information domain is divided into two parts and processed by the improved function transformation to strengthen the fitting effects of the trends of the upper and lower bounds in interval grey numbers before establishing prediction models respectively. By combination of the forecasting models of the kernel sequence and the processed information domains, the prediction results for the interval grey numbers are optimized and the principle of “full usage of information” is reflected during the modeling process of the interval grey numbers. Through discussing the case of the per capita industrial wastewater discharge in the Yangtze River Delta, the results of this method is verified by compared with traditional grey prediction methods of interval grey numbers, which shows its effectiveness and practicability. The proposed method provides another feasible forecasting method for the interval grey number prediction. The clear principle and modeling mechanism of this method make it possible to be applied in every field where interval grey numbers exists.