A Rainfall Estimation Method based on RBFNN

摘要

Doppler Weather Radar (DWR) plays an important role in short-term forecast and now-casting, and the weather radar data could be used for rainfall estimation of a certain region or other further processing. We improved the algorithm of rainfall estimation based on Radical Basis Function Neural Network (RBFNN), and optimized the key parameters of RBFNN such as centers, length and link weights of the neural network with Cuckoo Search (CS). In order to realize the match of the weather radar data and the rain gauge data in time, we proposed the Time Interpolation Method (TIM). This rainfall estimation model from weather radar data was called CS-RBFNN model, which was employed in a rainfall estimation algorithm of DWR. The experimental results show that the estimated results are more close to the measured rainfall with the CS-RBFNN model.