作 者：
尹航;吕佳威;陈耀聪;岑红蕾;李景彬;刘双印
单 位：
石河子大学机械电气工程学院;北京市农林科学院信息技术研究中心;仲恺农业工程学院广州市农产品质量安全溯源信息技术重点实验室;仲恺农业工程学院信息科学与技术学院;广东省农产品安全大数据工程技术研究中心
关键词：
羊舍;集约化养殖;CO_2质量浓度预测;极限学习机;麻雀搜索算法;分布式梯度提升框架
摘 要：
为减少肉羊集约化养殖过程中因环境恶化产生的应激反应,精准调控CO_2质量浓度,提出了基于分布式梯度提升框架(LightGBM)、麻雀搜索算法(SSA)融合极限学习机(ELM)的CO_2质量浓度预测模型。首先利用LightGBM筛选出与CO_2质量浓度相关的重要特征,降低预测模型的输入维度;然后选择Sigmoid为激活函数,使用具有较强非线性处理能力的单隐含层ELM神经网络算法构建CO_2质量浓度预测模型;最后通过麻雀智能优化算法对ELM模型中所需要的超参数进行优化,并将优化后模型应用于新疆玛纳斯集约化肉羊养殖基地。试验结果表明,该模型预测均方根误差(RMSE)、平均绝对误差(MAE)和决定系数(R~2)分别为0.021 3 mg/L、0.013 6 mg/L和0.988 6,综合性能指标优于支持向量回归(SVR)、反向传播神经网络(BPNN)、长短记忆神经网络(LSTM)、门限循环单元(GRU)和LightGBM等;CO_2质量浓度预测曲线贴近真实曲线,具有良好的预测效果,能有效满足集约化肉羊养殖过程中CO_2质量浓度精准预测及调控要求。
译 名：
Prediction of CO_2 Concentration in Xinjiang Breeding Environment of Mutton Sheep Based on LightGBM-SSA-ELM
作 者：
YIN Hang;Lü Jiawei;CHEN Yaocong;CEN Honglei;LI Jingbin;LIU Shuangyin;College of Information Science and Technology, Zhongkai University of Agriculture and Engineering;Information Technology Research Center, Beijing Academy of Agricultural and Forestry Sciences;Guangdong Provincial Agricultural Products Safety Big Data Engineering Technology Research Center;College of Mechanical and Electric Engineering, Shihezi University;Guangzhou Key Laboratory of Agricultural Products Quality and Safety Traceability Information Technology,Zhongkai University of Agriculture and Engineering;
关键词：
sheep house;;intensive culture;;CO_2 concentration prediction;;extreme learning machine;;sparrow search algorithm;;light gradient boosting machine
摘 要：
Air quality plays an important role in mutton sheep breeding environment, in order to reduce the stress response of CO_2 to the growth of large-scale mutton sheep and ensure the healthy growth of mutton sheep in the appropriate environment, the key is to accurately control the CO_2 in the mutton sheep breeding environment. A CO_2 prediction model of mutton sheep breeding environment was proposed based on light gradient boosting machine(LightGBM), sparrow search algorithm(SSA) and extreme learning machine(ELM). Firstly, LightGBM was used to screen out the important characteristics of carbon dioxide concentration and reduce the input dimension of the prediction model. Then, ELM neural network algorithm with single hidden layer with strong nonlinear processing ability was used to build the CO_2 prediction model. Finally, through the sparrow intelligent optimization algorithm, the super parameters needed in ELM model were optimized to obtain the best prediction model. The prediction model was applied to a large-scale mutton sheep breeding base in Manas County, Changji Hui Autonomous Prefecture, Xinjiang Uygur Autonomous Region, and good prediction results were obtained. The experimental results showed that the prediction model had good prediction effect, and the root mean square error(RMSE) of ELM was higher than that of SVR, BPNN, LSTM, GRU and LightGBM. The RMSE, mean absolute error(MAE) and R~2 were 0.021 3 mg/L, 0.013 6 mg/L and 0.988 6, respectively. The results showed that the combined model can not only achieve accurate control of carbon dioxide in sheep house, but also meet the needs of fine decision-making for mutton sheep breeding. It also can help farmers make decisions and reduce farming risks.
