On-Line Process Parameter Identification for Control of Variable Rate Nitrogen Fertilization

Variable rate nitrogen fertilization (VRNF) is a technique being developed for agricultural applications in recent years. In designing a controller for VRNF system, it is hoped that the process parameters, which are difficult to measure, could be identified on-line so that the model-based digital controllers or adaptive controllers could be used to achieve better system performance.

In this paper, the problems associated with the on-line process parameter identification for the control of variable rate nitrogen fertilization are analyzed. Simulation studies and experimental tests were conducted to determine a suitable parameter estimation algorithm and the influence of pseudo random binary signal and external torque disturbance on the parameter identification accuracy. The results from the simulation and experimental studies revealed that the recursive least square algorithm is most suitable for parameter estimation. The parameter identification accuracy decreases with increasing torque disturbance level. The pseudo random binary signal should have a magnitude which is 20 percent of the set point input, and a switching interval close to the dominant time constant of the process.