Abstract: In order to solve the problem that the existing non-stable and non-linear residual water level prediction models are less and low accuracy,a combined residual water level prediction model based on MEEMD algorithm and genetic optimization BP neural network is studied.Based on the time series data of residual water level obtained from four long-term tidal stations in Hawaii island,the genetic algorithm MEEMD is firstly used to process and analyze the time series data of residual water level,and a relatively stable IMF component of residual water level is obtained.Then,the stable IMF components decomposed by genetic algorithm optimization are taken as the input variables of BP neural network prediction model,and the prediction models of BP neural network optimized by MEEMD genetic algorithm for 12,24 and 48 h short-term residual water levels are established respectively.By comparing with the results of the non-optimal BP neural network prediction model,the results show that the deviation of the root mean square error before and after optimization is up to 2.03 cm,which verifies that the short-term residual water level within 24 h is still maintained its relevant characteristics.The combined prediction model is of great significance to the analysis of the variation law of residual water level,the accuracy of tide prediction and the correction of water level.

Key words: modified ensemble empirical mode decomposition;genetic algorithm;residual water level;BP neural network