Application of wireless technologies to forward predict crop yields in the poultry production chain
Keywords:Environmental control, Productivity, Average bird weight, Wireless agricultural sensors, cloud computing
Average bird weight is the primary measure of crop yield and is the basis for calculating payment for the grower by the wholesaler. Furthermore the profit per bird is very small. Thus very tight control of growing process is essential to ensure average bird weight is maximised. The important factors (air temperature, air humidity, Carbon Dioxide concentration and Ammonia concentration) that affect the intake of feed and water must be kept at their optimum during the progress of the growing cycle. These factors can be influenced by activating burners and opening the vents on walls of the growing house. It then follows that the burning and venting strategy will be influential on the average bird weight of the crop.
Currently the burning and venting strategy is based on notional ideal levels and data from wall mounted sensors. This suffers from two fundamental problems; firstly the strategy is determined by ideals that may not be suitable for all growing houses and secondly the data is not measured from the chickens own airspace. Thus the management strategy is based on a model that may not reflect reality and on data that may not reflect reality
The “BOSCA” project addresses these problems by placing wireless environmental sensors into the chickens own airspace. This provides for direct measurement of the air experienced by the chickens and reports the recorded data in near real-time to a cloud based data management system. The sensor data is merged with the data from the growing house weighing scales in the cloud repository so a predictive model of average bird weight from the measured environmental data can be calibrated and validated. Furthermore, a timeshift can be applied to the environmental data during model calibration and validation so the average bird weight can be forward predicted by 72 hours (r2 up to 0.89 with neural networks). This gives the grower advance notice of a deviation from ideal feeding and watering conditions and the likely consequences of failing to take remedial action such as turning on the burners or venting the house.