Abstract: Traditional GM(1,1) forecasting model has such deficiencies as single optimization method and poor adaptability in fitting and estimating dam displacement, which affect the estimation result. In this paper, an improved non-equidistance GM(1,1) forecasting model is proposed by using Suffled Frog Leaping Algorithm to optimize background value and smoothing coefficient, to search for optimal definite condition and correct residual error. The fitting results and prediction accuracy of traditional model and the proposed model are compared through an engineering example application. Results suggest that the improved GM(1,1) model could enhance the prediction accuracy effectively, hence can be used in the monitoring and prediction of dam displacement.

Key words: dam displacement, forecasting model, improved GM(1,1), shuffled frog leaping algorithm(SFLA), optimization of residual error