Study On Precipitation Prediction Model Of Extreme Learning Machine Based On Intelligent Optimization Algorithm

Due to the non-linear and high fluctuation characteristics of precipitation time series,there is no possible to predict the rainfall accurately with a single model.For this reason,this work proposes a combination model made up of ensemble empirical mode decomposition(EEMD)and improved extreme learning machine(ELM)to predict the rainfall.With the experimental simulation,we find the combination model is efficient.1.In view of the rainfall time series by external factors so as to make the original signal in a lot of noise.This work gets characteristics of precipitation with the EEMD method,and successfully reduces the noise of the original precipitation signal as well as stays original rules.Meanwhile,we compare the EMD method with the EEMD method,find that the EEMD method solved the mode mixing problem subtly.2.We make prediction models of precipitation with BP neural network and extreme learning machine respectively.We do the numerical experiments and find that the result of extreme learning machine is better than BP neural network in forecast data,forecast trend and error index.Therefore,we take the extreme learning machine as the main predict model.3.Improving the parameter of ELM with particle swarm optimization(PSO)and chaos particle swarm optimization(CPSO)algorithm respectively.The result of ELM-CPSO is better than ELM-PSO by the experiments their error index.We make a combination model made up of EEMD and improved ELM to predict and analysis the precipitation.The results show that the EEMD-ELM-CPSO model is higher than single model such as BP neural network and ELM,ELM-PSO model,ELM-CPSO model for11.9%,9.3%,6.6%,4.1%.The prediction results are obviously better than others.The study shows that this model has a good performance in precipitation prediction.