1 引 言
2 蛋鸡设施环境质量评价预测模型构建方法
2.1 基于CS-BP预测模型构建
2.2 基于改进CS-BP预测模型构建
1:begin 2:目标函数f(x), x = (x1, ... , xd)T 初始化种群的n个鸟巢 xi (i = 1, 2, ..., n) 3:while (t <最大迭代次数) or (算法终止规则) 4: for 劣解种群中的解do 5: 对xi执行Lévy飞行并产生新解 xnew,i 6: xi→xnew, i 7: fi→fnew, i 8: end for 9: for 优解种群中的解 do 10: 对xj执行梯度下降并产生新解 xnew,j 11: xj→xnew, j 12:fj→fnew, j 13: end for 部分劣解 ( ) 被丢弃并产生新解 寻找种群中最优解 14: end while 输出最优解 将最优解做为BP神经网络初始权值阈值 输出网络模型net及误差值 15:end |
3 CS-BP预测模型性能试验设计
3.1 试验数据采集
3.2 监测点布置
3.3 环境质量等级划分
表1 蛋鸡设施养殖环境质量等级划分表Table 1 Classification of environmental suitability of laying hens facility breeding environment |
质量评价等级 | 温度/℃ | 湿度/% | 光照强度/lx | NH3浓度/(mg‧m-3) |
---|---|---|---|---|
5级(优) | 20~25 | 60~70 | 28~30 | <15 |
4级(良) | 18~20或25~26 | 55~60或70~75 | 20~28或30~35 | 15~17 |
3级(一般) | 14~18或26~27 | 50~55或75~78 | 15~20或35~40 | 17~18 |
2级(差) | 10~14或27~30 | 40~50或78~80 | 10~15或40~50 | 18~25 |
1级(极差) | <10或>30 | <40或>80 | <10或>50 | >25 |
表2 部分试验数据Table 2 Part of experimental data |
温度/℃ | 湿度/% | 光照强度/lx | NH3浓度/(mg‧m-3) | 质量评价等级 |
---|---|---|---|---|
14.3 | 54 | 19 | 18 | 3 |
9.2 | 39 | 63 | 26 | 1 |
9.8 | 37 | 51 | 28 | 1 |
18.8 | 58 | 27 | 15 | 4 |
19.5 | 58 | 28 | 16 | 4 |
22.3 | 61 | 29 | 10 | 5 |
19.4 | 58 | 21 | 16 | 4 |
14.7 | 50 | 19 | 17 | 3 |
19.6 | 58 | 29 | 16 | 4 |
18.3 | 57 | 22 | 17 | 4 |
3.4 评价模型对比
4 预测模型性能参数对比测试
4.1 基于改进CS-BP的预测模型与其他预测模型性能参数对比
4.1.1 基于CS-BP的预测模型
表3 基于CS-BP的预测模型的试验结果Table 3 Experimental results of prediction model based on CS-BP neural network |
试验号 | MAE | MAPE | R 2 |
---|---|---|---|
1 | 0.9113 | 0.0181 | 0.8116 |
2 | 0.9120 | 0.0175 | 0.8324 |
3 | 0.9131 | 0.0178 | 0.8231 |
4 | 0.9253 | 0.0188 | 0.7980 |
5 | 0.8929 | 0.0172 | 0.8449 |
6 | 0.9011 | 0.0177 | 0.8246 |
7 | 0.8744 | 0.0169 | 0.8564 |
8 | 0.8913 | 0.0171 | 0.8392 |
9 | 0.7913 | 0.0153 | 0.8793 |
10 | 0.9103 | 0.0186 | 0.8160 |
平均值 | 0.8923 | 0.0175 | 0.8326 |
表4 基于改进CS-BP的预测模型的实验结果Table 4 Experimental results of prediction model based on improved CS-BP neural network |
试验号 | MAE | MAPE | R 2 |
---|---|---|---|
1 | 0.9394 | 0.0184 | 0.8300 |
2 | 0.8091 | 0.0159 | 0.8616 |
3 | 0.8714 | 0.0171 | 0.8559 |
4 | 0.7825 | 0.0155 | 0.8869 |
5 | 0.9020 | 0.0177 | 0.8261 |
6 | 0.8913 | 0.0170 | 0.8416 |
7 | 0.9124 | 0.0175 | 0.8430 |
8 | 0.9129 | 0.0179 | 0.8364 |
9 | 0.9031 | 0.0180 | 0.8551 |
10 | 0.7710 | 0.0151 | 0.8796 |
平均值 | 0.8695 | 0.0170 | 0.8512 |
0.1849 0.9729 -1.6614 2.2703
4.1.2 基于BP、GA-BP、PSO-BP的预测模型
图4 基于BP的蛋鸡设施养殖环境质量预测评价结果输出Fig.4 Prediction and evaluation results of environmental suitability of laying hens facility breeding environment based on BP |
图5 基于GA-BP的蛋鸡设施养殖环境质量预测评价结果输出Fig. 5 Prediction and evaluation results of environmental suitability of laying hens facility breeding environment based on GA-BP |
图6 基于PSO-BP的蛋鸡设施养殖环境质量预测评价结果输出Fig.6 Prediction and evaluation results of environmental suitability of laying hens facility breeding environment based on PSO-BP |
表5 四种蛋鸡设施养殖环境质量评价预测模型性能分析Table 5 Performance analysis results of the four hens breeding facility environmental quality evaluation prediction model |
指标 | BP | GA-BP | PSO-BP | 改进CS-BP |
---|---|---|---|---|
MAE | 0.1701 | 0.1100 | 0.0913 | 0.0865 |
MAPE | 0.0421 | 0.0301 | 0.0165 | 0.0159 |
R 2 | 0.7125 | 0.7911 | 0.8316 | 0.8569 |
4.2 基于改进CS-BP的预测模型的分类评价准确率测试
表6 蛋鸡设施养殖环境质量评价预测模型测试结果Table 6 Test results of environmental suitability prediction model of laying hens facility breeding environment |
环境质量等级 | 样本总数/个 | 分类正确样本数/个 | 分类错误样本数/个 | 准确率 |
---|---|---|---|---|
1 | 60 | 57 | 3 | 0.9500 |
2 | 60 | 56 | 4 | 0.9333 |
3 | 60 | 57 | 3 | 0.9500 |
4 | 60 | 58 | 2 | 0.9667 |
5 | 60 | 56 | 4 | 0.9333 |