Abstract: Aiming at the problems of strong nonlinear relationship in dam deformation monitoring data and low precision of traditional dam prediction model, a dam deformation prediction model based on improved bat algorithm (IBA) and optimized extreme learning machine (ELM) is proposed by using the improved bat algorithm to select the optimal parameters as the connection weights and thresholds of the extreme learning machine(IBA-ELM). The IBA-ELM model is applied to an engineering case, and the accuracy of IBA-ELM, BA-ELM and GA-ELM model prediction results are evaluated through the prediction and analysis of dam monitoring data of a reservoir. The average absolute errors between predicted values and measured values are 1.178 3, 0.459 8 and 0.335 6, IBA-The prediction accuracy of ELM model is higher than the other two models, which indicates that IBA-ELM model can effectively improve the prediction ability of dam deformation and has certain application value.

Key words: dam deformation, extreme learning machine, bat algorithm, genetic algorithm, variation factor