Unscented Kalman Filterを適用した水位予測システムの高速化と精度評価

抄録

The prediction accuracy of the flood prediction system is better because of many state values is set in the data assimilation method such as the Particle Filter. On the other hand, nonlinear filter theorem such as the Particle Filter, the Ensemble Kalman Filter et al spends much time for calculating and is not the actual. The authors had developed the flood prediction system using the Particle Filter and secured enough accuracy of the water level prediction. So the Unscented Kalman Filter was applied for the flood prediction system of the Yodo River in order to secure the prediction accuracy of the water level under the few state values and to shorten the time that a calculation takes. It was concluded that the unscented Kalman Filter is effective and secure the prediction accuracy of the water level as the method for the Particle Filter.