Hammerstein-Wiener model for the prediction of temperature variations inside silage stack-bales using wireless sensor networks

Abstract

Animal forage plays an important role in agriculture. To maintain the silage quality during storage, preserving it from decomposition by wrapping the harvested crop with plastic stretch film is crucial. Any failure in the preserving process permits the infusion of oxygen, which allows the growth of undesirable aerobic microorganisms and silage decomposition. To study the state of the silage, temperature inside the stack could be used as indicator of decomposition. However, no previous study has developed a model that can detect decomposition based on abnormal temperature variations inside a silage stack. The first objective of this paper was to develop a structured nonlinear model to estimate the dynamics of temperature variations measured by a wireless sensor network-based monitoring system (nRF95E) inside a silage bale and stack prior to the decomposition process. The requirement of a robust nonlinear model is that sensor nonlinearities that are inherent in nature are considered. The results showed that a Hammerstein-Wiener (HW) model with sigmoid network performed the best in terms of percentage of fit (91.70%) between the measured and simulated output. The second objective was to estimate the length of time required to detect the decomposition process after the airtight seal was punctured using the model. The error between the predicted model output and the sensor readings exceeded a preset threshold interval 17 days after the cover film was punctured and 9 days after the decomposition process was detected due to the odour released.