An improved particle filtering for time-varying nonlinear prediction of biomass and grain protein content

作者：Mansouri Majdi^*; Destain Marie France

来源：Computers and Electronics in Agriculture, 2015, 114: 145-153.

DOI：10.1016/j.compag.2015.04.006

• Crop model
• Particle filter
• Prediction
• Time-varying

This paper addresses the problem of time-varying nonlinear prediction of biomass and grain protein content. The objectives of this paper are threefold. The first objective is to use an improved particle filter with better proposal distribution for nonlinear prediction. The second objective is to extend the state and parameter estimation techniques (i.e., particle filter (PF) and improved particle filter (1PF)) to better handle nonlinear and non-Gaussian processes without a priori state information, by utilizing a time-varying assumption of statistical parameters. The third objective is to apply the state estimation techniques PF and IPF for predicting and modeling biomass and grain protein content. In a first step, we present an application of PF and IPF to a simple dynamic crop model with the aim to predict a single state variable, namely winter wheat biomass. In a second step, we apply PF and IPF for updating predictions of complex nonlinear crop models in order to predict protein grain content.
The comparative analysis is conducted to study the effects of two practical challenges (measurement noise, and the number of states and parameters to be estimated) on the estimation performances of PF, and IPF. To study the effect of measurement noise on the estimation performances, several measurement noise contributions are considered. Then, the estimation performances of PF and IPF are compared for different noise levels. Similarly, to investigate the effect of the number of states and parameters to be estimated on the estimation performances of PF and IPF, the estimation performance is analyzed for different numbers of estimated states and parameters.
The simulation results of both comparative studies show that the IPF provides a significant improvement over the PF. This is because, unlike the PF which depends on the choice of sampling distribution used to estimate the posterior distribution, the IPF yields an optimum choice of this distribution, which also utilizes the observed data. The results of the comparative studies show also that, for all the techniques, estimating more model parameters affects the estimation accuracy as well as the convergence of the estimated states and parameters. The performance of the proposed method is evaluated on a synthetic example in terms of estimation accuracy, root mean square error and execution times.

• 出版日期2015-6