基于L-M优化算法的猪舍氨气浓度预测模型研究  

Prediction model for piggery ammonia concentration based on L-M optimal algorithm

作　　者：谢秋菊[1,2] 苏中滨[1] 刘佳荟[3] 郑萍[1] 马铁民[2] 王雪[2] 

机构地区：[1]东北农业大学电气与信息学院,哈尔滨150030 [2]黑龙江八一农垦大学信息技术学院,黑龙江大庆163319 [3]索尼信息系统(大连)有限公司,辽宁大连116085

出　　处：《东北农业大学学报》2014年第10期74-79,共6页Journal of Northeast Agricultural University

基　　金：国家"863"项目(2012AA101905);黑龙江省青年科学基金项目(QC2013C065);黑龙江省教育厅科学技术研究项目(12531465);黑龙江省畜牧兽医行业公共基础数据平台关键技术研究与建立(GC10B501)

摘　　要：在规模化养殖中,猪舍环境直接影响猪健康水平及生产能力。针对猪舍环境因素(包括温度、湿度、风速和氨气浓度)进行数据采集,选取具有代表性30 d数据,建立基于L-M优化算法的3-7-1三层结构的BP神经网络模型,对猪舍环氨气浓度进行预测。结果表明,预测模型经过90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1.72%,与线性预测方法相比较可提高猪舍氨气浓度预测的准确性与及时性,为猪舍环境预警及控制提供支持,也为其他行业预测模型建立提供参考。In the large-scale farming, piggery environment impacts on the health of swine and production capacity directly. Piggery environmental factors mainly include wind speed, temperature, humidity and ammonia concentration, the representational data during 30 continuous days were selected. The 3-7-1 BP neural network model of the three-layer structure based on L-M optimal algorithm was built to predict the piggery ammonia concentration. It is shown in the experiment that network reaches the target error after 90 steps, the model has the characteristics of fast network convergence and high efficiency, and the biggest relative error between predicted and measured values is only 1.72%, the accuracy and timeliness of the piggery environmental prediction is greatly improved. The prediction model established in the paper can provide support for the piggery environment early warning and control, and also can provide a viable idea for other industries to establish prediction model.

关 键 词：L-M优化算法 BP神经网络 预测模型 猪舍氨气浓度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]