A Second-order Grey Model for Forecasting Interval Number Series

Abstract

Interval number contains both the average and change information. This paper presents two second-order grey models suitable for the forecasting of binary and ternary interval number, named BINGM (2, 1) and TINGM (2, 1). The overall development coefficients of two models are taken as an accurate number, and the grey inputs are taken as interval numbers. By this method of parameter setting, the models are suitable for the interval number series directly, and no need to transform the interval number series into the accurate number series. The overall development coefficients, grey inputs, and constants in the forecasting formulas of TINGM (2, 1) and BINGM (2, 1) are all determined by the new information priority principle (NIPCM), instead of the least square method (LSM). The wind power generating capacity and the generating capacity of China are predicted based on BINGM (2, 1) and TINGM (2, 1). The forecasting performance is better than the models based on GM (1, 1).