Research On Weather Forecasting Based On SCA And LSSVM

As an important predictive service in people's daily life,meteorological forecast is of tremendous instructional significance in production and life style of mankind since it first appeared.With the gradual improvement of related equipment in weather stations around China,meteorological data went through explosive accumulation,which raises higher demand for the analysis and process of meteorological data.Currently,the refinement of meteorological forecast has become the latest trend in global range while it is still relatively lag behind domestically.How to improve the refinement of meteorological service in China has become a hot issue in fields involved.This paper states the meaning of meteorological forecast and the trend of meteorological forecast refinement by analyzing the traditional technology of meteorological forecast and major researches done by researchers from related fields as well as creating targeted model for different prediction objective.The main content consists following aspects:To find optimum parameters of Least Squares Support Vector Machine(LSSVM)with the help of Sine Cosine Algorithm(SCA).So the SCA-LSSVM model shall be used as primary prediction tool.In the prediction of relative humidity,the range of relative humidity could have strong influence on the regression forecasting.Therefore,in an effort to increase the accuracy of the prediction,this article will use Empirical Mode Decomposition(EMD)on relative humidity to get several components.Then,in accordance to different combinations,the study will apply SCA-LSSVM model to make a prediction after analyzing components.The final prediction result will be acquired by component predicted value fitting.The experiment shows that to combine the components according to its correlation before prediction will increase the accuracy of prediction and reduce calculation.This model will enjoy a more optimized performance,higher forecast accuracy and better generalization capability compared with other models.In the prediction of atmospheric temperature,prediction modules are divided into two parts: one based on multidimensional meteorological factor and the other based on the time sequence of atmospheric temperature.SCA-LSSVM model is applied in both modules as prediction tool.Through the combination of Artificial Neural Network(ANN)and predicted value,dual SCA-LSSVM model is created.The experiment indicates that the various indicators of SCA-LSSVM model has proved to be excel than unitary prediction model and in result has a wider application prospect.