Evaluation Model Of Temporal And Spatial Distribution Of Feed Intake Of Sheep On Family Ranch

Grazing is the most direct and efficient use of grassland resources.Reasonable grazing can not only ensure the sound development of animal husbandry,but also have a positive impact on the sustainable utilization of grassland resources and the prevention of grassland degradation.The traditional grassland resource monitoring methods rely on the direct observation of grazing staff,which is time-consuming,laborious,poor real-time,and easy to be affected by the subjective consciousness of the observer,so it is unable to accurately obtain the actual situation of grazing.Therefore,in this thesis,computer information technology and global position system(GPS)are applied to the research of Grassland Science and animal husbandry,and a feed distribution monitoring method which can estimate the feed intake parameters of herds in different geographical locations in real time and generate the corresponding feed intake spatial-temporal distribution model combined with the spatial-temporal trajectory of shepherds is proposed,The purpose is to obtain the feeding intensity in different areas of the experimental pasture and the grazing law of experimental sheep through the temporal and spatial distribution of grazing intake,so as to provide a scientific basis for grazing decision-making.The main research contents are as follows:(1)Estimation of feed intake of grazing herd.In order to accurately estimate the feed intake information of herds at different time and different location points,this thesis proposes a feed intake estimation model of herds based on long short-term memory(LSTM)neural network optimized by genetic algorithm(GA).Firstly,the main influencing factors affecting feed intake data are obtained by Pearson correlation coefficient method,The estimation model of herd feed intake based on LSTM neural network algorithm is constructed,and the genetic algorithm is introduced to optimize the parameters of LSTM neural network model to increase the reliability of the model.Finally,the model is used to estimate the feed intake of herds.The results showed that the mean absolute error(MAE),mean absolute percentage error(MAPE)and root mean square error(RMSE)of each evaluation index of the feed intake estimation model were2.982%,9.85% and 6.108 respectively.Compared with the single LSTM neural network and GRU neural network model,they are better than other models;The model has good estimation performance and strong generalization ability,can provide scientific guidance for reasonable rotational grazing,and has certain application value for grassland protection.(2)Study on temporal and spatial distribution of feed intake of grazing sheep.Based on the above-mentioned estimation model of feed intake of grazing sheep,based on the spatiotemporal trajectory data of grazing sheep,combined with the analysis methods of spatiotemporal matching and superposition processing,the spatiotemporal distribution model of feed intake of grazing sheep is constructed;The feasibility of this method is verified by comparing with the method of buffer zone and grid analysis combined with simulated feeding method to obtain the temporal and spatial distribution of feed intake of grazing sheep.Taking the experimental pasture in the study area from June to December2020 as an example,the two temporal and spatial distribution models of feed intake reasonably and intuitively expressed the feeding intensity at different locations of the experimental pasture and the grazing law of the experimental individuals.(3)Study and Analysis on the correlation between grazing behavior of shepherds and micro factors of grassland.In this thesis,the laws of feeding behavior of shepherds were sorted and counted,and the correlation between feeding behavior of experimental sheep and micro factors of grassland was studied and analyzed.The results show that the research results of feeding behavior are consistent with the relevant research conclusions at home and abroad.(4)Study on the correlation of resting behavior.In this thesis,the improved density clustering(DBSCAN)algorithm is used to divide the "hot spot areas" when the shepherds lie down,and the geographical location of these "hot spot areas" in the experimental pasture is studied,so as to obtain the law when the shepherds choose the resting place.