Developing a digital twin model for improved pasture management at sheep farms

Abstract

Sheep farming plays an integral part in the South African livestock sector through the production of meat and wool. In sheep production, a pasture is of utmost importance. A pasture is deemed as a natural source of nutrients for animals, especially to small scale farmers who cannot afford to buy artificial feed for sheep. Pastures therefore need to be maintained to ensure good quality and quantity feed to get the best produce from the animals. The process of maintaining pastures is referred to as pasture management, and pasture management strategies require a farmer to be aware of factors such as climate and soil. Climate change is reducing pasture production which lowers livestock productivity and, as a result, small scale livestock farmers suffer great losses. Pastures are dependent on climatic conditions, therefore, climate change has a great impact on pasture management. With the progression of climate change, farming seasons have changed, and it has become a challenge for sheep farmers to implement proper pasture management techniques due to changing weather patterns. As a result, farmers struggle to keep their sheep alive due to unavailability of animal feed caused by poor pasture production due to climate change. The use of technology in livestock farming introduced the concept of smart farming which has made pasture management easier and more cost effective. Smart farming integrates advanced technological methods that include connectivity, Internet of Things (IoT), and data analytics. Thus, to alleviate the challenge of changing weather patterns that affect pasture management, digital twin technology is proposed as solution in this study. Digital twin development includes smart farming technologies, and the literature shows that digital twins help in identifying and anticipating issues before they happen. Hence, a digital twin is developed to improve pasture management by predicting pasture height to determine the anticipated amount of pasture and secure enough feed for the sheep for sustainable production. The system is comprised of the physical system and its digital twin. The physical system is set up using the IoT sensors and devices. These include soil moisture sensor that measures soil moisture content, ESP32 CAM that captures the pasture images, raspberry pi that processes captured images and calculates pasture height, weather station which measures rainfall and temperature and the modem that provides wireless connectivity. The digital twin is developed using both historical and the real-time data from the sensors. Sensors data was streamed in ThingSpeak® cloud in real-time. Transfer of data from the sensors is the link between the physical system and the digital twin. The data analysis is performed in MATLAB® using an Artificial Neural Network (ANN) machine learning algorithm. The prediction of the height of the pasture is then modelled using the SIMULINK® platform. The analysis of the digital twin results shows that the digital twin can accurately predict the pasture height based on historical and current data. Prediction of pasture height relies predicted temperature, rainfall and soil moisture. The results presented temperature and soil moisture prediction model to have a good performance with prediction error of ±1.62 and -1.7, respectively. Rainfall prediction model had a slightly higher prediction error of ±6.03. The outcome on the rainfall prediction model is the evidence of the impact of varying weather patterns. Nonetheless, based on the predicted temperature, rainfall, and soil moisture, the digital twin predicted pasture height to be 52 cm and observed pasture height was 56 cm, with the prediction error of -4. The difference on the pasture height predictions is 4cm which reflects a good prediction. Therefore, the digital twin was able to reflect good dependency amongst the models, on that account, the digital twin can serve to enhance pasture management through its capabilities to monitor pastures in real-time and perform pasture height predictions for the future that assist farmers in decision making.