基于BP神经网络的牛舍环境预测模型研究

Study on barn environmental prediction model based on BP neural network

在规模化养殖中,牛舍环境直接影响着牛的健康及生产能力。实验针对牛舍的主要环境因素——风速、温度、湿度和氨气质量分数进行了连续3天的数据采集,获得72组环境数据,建立了4-9-4三层结构的BP神经网络模型,对牛舍环境进行预测,采用L-M优化算法对前两天的48组环境数据进行训练,第三天的24组环境数据作为测试样本。仿真实验表明,经过12步达到目标误差,网络收敛速度快、效率高,预测值与实测值最大相对误差仅为4.62%,大大提高了牛舍环境预测的准确性与及时性。文章建立的环境预测模型可以为牛舍环境预警及控制提供支持,同时也可为其他行业预测模型的建立提供一种可行的思路。

In the large-scale farming, barn environment has a direct impact on the health of cattle and production capacity. Barn major environmental factors such as wind speed, temperature, humidity and ammonia concentration datawere collected for three consecutive days, 72 sets of environmental data were acquired. The 4-9-4 BP neural networkmodel of the three-tier structure was built to predict the barn environment. 48 environmental data of the first two dayswere trained based on L-M optimization algorithm. The 24 group environmental data of the third day were as a testsample. It is shown in simulation experiment that network reaches the target error after 12 steps, the model has thecharacteristics of fast network convergence and high efficiency, the biggest relative error between predicted andmeasured values is only 4.62%, the accuracy and timeliness of the barn environmental prediction is greatly improved.The environmental prediction model established in the paper can provide support for the barn environment early warning and control, and also can provide a viable idea for other industries to establish prediction model.