Author(s): Hiroki Tsujikura; Kohji Tanaka; Kenji Miyamoto
Linked Author(s): Kohji Tanaka
Keywords: Particle Filter; State Variable; Water Level Prediction
Abstract: This paper describes a method to solve the problems involved in applying a particle filter for water level prediction. A water level prediction model at the Yodo River has been constructed using a particle filter with state variables such as upstream and lateral discharge. From the results of water level prediction for two floods, Typhoon No. 18 (2013) and Typhoon No. 11 (2014), we found that the prediction model needs improvement. Besides the upstream and lateral discharges, we added the Manning’s coefficient to the state variables. Then, we found the degree of improvement in the estimation accuracy for the upstream and lateral discharges. Moreover, we found that an increase in the number of state variables does not assure the accuracy improvement of the estimation of the boundary discharges, even if the number of particles is increased. In a real-time water level prediction, it is important to consider the balance between the setting of state variables and the number of particles.
Year: 2018