Abstract: To address the issue that the traditional grey model is difficult to predict nonlinear data series with multiple mutations caused by policy evolution, a time-delayed grey Bernoulli model with dummy variables (abbreviated as DTD-NGBM(1,1)) is proposed by introducing policy effects as dummy variables. Next, the solution of the time response function of DTD-NGBM(1,1) is discussed, and the optimal values of the time-delayed term and the nonlinear parameter are determined using the debugging method and the genetic algorithm, respectively. Finally, the validity and superiority of the proposed model are verified by taking solar power generation volume forecasting in China and the U.S. as examples. The results show that the proposed model can accurately describe the trend of the data series under the influence of dummy variables.

Key words: Grey Forecasting, Bernoulli Model, Dummy Variables, Time-delay, Genetic Algorithm