Rainfall Prediction Based On Deep Learning Convolution Neural Network

Rainfall plays an important role in human life in various weather events.Accurate rainfall information is very important to the planning and management of water resources,and it is also the key to reservoir operation and flood control.In addition,rainfall has a significant impact on traffic,sewer systems and other human activities in urban areas.However,due to the complexity,diversity and instability of climatic conditions,and influenced by many factors,it is difficult to predict the fuzziness and uncertainty of weather forecast.The accuracy of prediction using traditional statistical methods is not high.In recent years,neural network has been developed and matured,because of its high degree of nonlinearity,flexibility and data-driven learning ability,it can depict the fuzziness and uncertainty of weather prediction,so that it has been widely used in the field of meteorology.Based on the radar echo intensity data(100 x 100)from January to October 2016 in Zhejiang Province,this paper introduces the BP neural network model and convolution neural network into the rainfall forecasting system of Zhejiang Province and compares them with the stochastic forest model.Experimental results show that the prediction effect of stochastic forest model is lower,the rainfall intensity is more easily underestimated,the prediction effect of BP neural network and convolution neural network is better than that of stochastic forest network.