基于RSM-GA的稻谷含水率预测模型设计与试验

doi:10.13304/j.nykjdb.2024.0318

摘要/Abstract

摘要：

为解决稻谷在干燥过程中含水率预测不准确、滞后性较大以及决定系数不高等问题，采用响应面（response surface methodology，RSM）法分析干燥过程中影响稻谷含水率的主要因素，利用遗传算法（genetic algorithm，GA）对传统BP（back propagation）神经网络模型进行优化，欲建立一种基于RSM-GA的稻谷含水率预测模型。结果表明，热风温度、粮层温度、环境相对湿度在干燥过程中对稻谷含水率影响显著。以热风温度、粮层温度、环境相对湿度为预测模型的输入层，稻谷含水率为输出层，通过经验公式确定预测模型的最优中间隐含层数为10，由此建立预测模型的神经元结构为3-10-1。在进行模型训练时，最优性能表现在第15次，最小均方根误差为0.621 84×10^-3，得到最优的Matlab仿真试验设置参数，当迭代至200代时，适应度值在0.019 5时趋于稳定。经过遗传算法优化后的预测模型决定系数为0.980，较传统模型提高5%；均方根误差为0.009，降低17%。综上，优化后的神经网络模型性能提高，为后续控制策略研究提供参考。

关键词: 稻谷, 响应面法, 含水率预测, 遗传算法, BP神经网络

Abstract:

In order to solve the problems of inaccurate prediction of rice moisture content， large hysteresis and low coefficient of determination during the drying process， the response surface methodology （RSM） was used to analyze the main factors affecting the rice moisture content during drying. Genetic algorithm （GA） was used to optimize the traditional BP （back propagation） neural network model， and a prediction model was established based on RSM-GA， which could accurately predict the moisture content of grain during drying. The results showed that the hot air temperature， grain layer temperature and ambient relative humidity had significant effects on the change of moisture content during drying process of rice. The hot air temperature， grain layer temperature and ambient relative humidity were used as the input layers of the prediction model， the rice moisture content was the output layer. The optimal number of intermediate hidden layers of the prediction model was determined to be 10 through empirical formulas， and the neuron structure of the prediction model established was 3-10-1. During model training， the optimal performance was at the 15 th time， the minimum root mean square error was 0.621 84×10^-3， and the optimal Matlab simulation test setting parameters were obtained. When the iterating was to 200 generations， the fitness value was stabilized at 0.019 5. After optimized by genetic algorithm， the coefficient of determination of the prediction model was 0.980， which was 5% higher than the traditional model； the root mean square error was 0.009， which was 17% lower than the traditional model. In summary， the performance of optimized neural network model was improved， which provided a reference for subsequent control strategy research.

Key words: rice, response surface method, moisture content prediction, genetic algorithm, BP neural network

中图分类号:

S126

潘衡, 万霖, 车刚, 王思佳, 郑宇, 张强. 基于RSM-GA的稻谷含水率预测模型设计与试验[J]. 中国农业科技导报, 2025, 27(10): 95-104.

Heng PAN, Lin WAN, Gang CHE, Sijia WANG, Yu ZHENG, Qiang ZHANG. Design and Experiment of Rice Moisture Content Prediction Model Based on RSM-GA[J]. Journal of Agricultural Science and Technology, 2025, 27(10): 95-104.

图/表 15

图1 混流式稻谷干燥试验装置注：1—温湿度传感器；2—混流式稻谷干燥机；3—进风管道；4—排粮电机；5—卸粮装置；6—控制系统柜；7—混配管道；8—混配装置；9—热风管道；10—电加热控制柜；11—电加热装置。

Fig. 1 Mixed flow rice drying experimental deviceNote：1—Temperature and humidity sensor； 2—Mixed flow rice dryer； 3—Air inlet pipeline； 4—Grain discharge motor； 5—Unloading device； 6—Control system cabinet； 7—Mixing pipeline； 8—Mixing device； 9—Hot air pipeline； 10—Electric heating control cabinet； 11—Electric heating device.

表1 检测仪器参数

Table 1 Parameters of testing instrument

仪器名称

Instrument name

数量

Quantity

生产公司

Production company

PM-8188New型谷物水分检测仪

PM-8188New type grain moisture detector

日本凯特公司

Japan Kett Company

温湿度传感器

Temperature and humidity sensor

北京昆仑远洋仪表科技有限公司

Beijing Kunlun Ocean Instrument Technology Co.， Ltd.

AMI300多功能测量仪

AMI300 multifunctional measuring instrument

KIMO公司

KIMO Company

管道式风速传感器

Duct type wind speed sensor

建大仁科

Kenda People Soft

表2 正交试验因素编码

Table 2 Encoding of orthogonal experimental factors

序号 Serial number	编码值 Encoding value	A：热风温度 Hot air temperature/℃	B：粮层温度 Grain layer temperature/℃	C：环境相对湿度 Ambient relative humidity/%
+γ	+1.682	55	35	90
+1	+1	51	33	82
0	0	45	30	70
-1	-1	39	27	58
-γ	-1.682	35	25	50

表3 不同隐含层节点数对应的均方根误差与决定系数

Table 3 RMSE and R2 to different number of hidden layer nodes

隐含层节点数 Number of hidden layers	均方根误差 RMSE	决定系数 R²
3	0.781	0.83
4	0.640	0.87
5	0.585	0.88
6	0.662	0.82
7	0.533	0.84
8	0.328	0.86
9	0.191	0.90
10	0.026	0.93
11	0.257	0.91
12	0.719	0.81
13	0.794	0.79
14	0.867	0.77
15	0.618	0.85

图2 BP神经网络结构

Fig. 2 BP neural network structure

图3 遗传算法流程

Fig. 3 Genetic algorithm flow

图4 不同环境相对湿度、粮层温度、热风温度处理下的稻谷含水率A：环境相对湿度；B：粮层温度；C：热风温度

Fig. 4 Moisture content of rice grains under different treatments hot air temperature， ambient relative humidity and grain layer temperatureA： Ambient relative humidity； B： Grain layer temperature； C： Hot air temperature

表4 二次回归正交旋转组合试验及结果

Table 4 Quadratic regression orthogonal rotation combination experiment and results

编号 Number	A：热风温度 Hot air temperature/℃	B：粮层温度 Grain layer temperature/℃	C：环境相对湿度 Ambient relative humidity/%	Y：稻谷含水率 Rice moisture content/%
1	39	27	58	16.6
2	51	27	58	18.2
3	39	33	58	17.1
4	51	33	58	16.9
5	39	27	82	17.4
6	51	27	82	20.5
7	39	33	82	21.5
8	51	33	82	17.2
9	35	0	70	16.3
10	55	0	70	16.3
11	45	25	70	17.8
12	45	35	70	19.7
13	45	30	50	16.7
14	45	30	90	18.8
15	45	30	70	18.5
16	45	30	70	18.3
17	45	30	70	18.9
18	45	30	70	16.9
19	45	30	70	17.3
20	45	30	70	17.3
21	45	30	70	16.3
22	45	30	70	16.2
23	45	30	70	17.4

表5 稻谷含水率回归模型系数显著性检验及方差分析

Table 5 Significance test and analysis of variance of regression model coefficients for rice moisture content

参数 Parameter	稻谷含水率Rice moisture content
参数 Parameter	系数Coefficient	F值F value	P值P value
模型Model	34.480 0	5.930 0	0.002 2
A：热风温度Hot air temperature	6.070 0	9.400 0	0.009 0
B：粮层温度Grain layer temperature	4.370 0	6.760 0	0.022 0
C：环境相对湿度Ambient relative humidity	14.320 0	22.170 0	0.000 4
AB	0.031 3	0.048 4	0.829 3
AC	0.911 2	1.410 0	0.256 2
BC	1.360 0	2.110 0	0.170 3
A²	0.070 8	0.109 5	0.745 9
B²	7.260 0	11.230 0	0.005 2
C²	0.088 6	0.137 2	0.717 0
残差平方和Sum of squared residuals	0.916 3
纯误差平方和Sum of squares of pure errors	0.476 9
决定系数R²	0.804 2
校正决定系数R_Adj²	0.668 6
预测决定系数R_Pre²	0.072 0
变异系数Coefficient of variation	4.530 0

图5 环境相对湿度、粮层温度、热风温度对稻谷含水率的交互影响

Fig. 5 Interactive effects of environmental relative humidity， grain layer temperature， and hot air temperature on rice moisture content

图6 均方误差曲线

Fig. 6 Mean square error curve

图7 BP神经网络模型预测结果

Fig. 7 BP neural network model prediction result

图8 遗传算法-BP神经网络预测模型结果

Fig. 8 Genetic algorithm BP neural network prediction model result

图9 遗传算法-BP神经网络适应度值

Fig. 9 Genetic algorithm BP neural network fitness value

图10 遗传算法-BP神经网络回归图

Fig. 10 Genetic algorithm-BP neural network regression graph

参考文献 24

[1]	国家统计局.国家统计局关于2023年粮食产量数据的公告[N].中国信息报,2023-12-12(001).
[2]	彭冬根, 聂江涛, 陈文华.基于分层模型的溶液除湿谷物就仓干燥系统性能及能耗[J].农业工程学报,2021,37(14):274-282.
	PENG D G, NIE J T, CHEN W H. Performance and energy consumption of liquid desiccant dehumidification grain in-Bin drying system via layered model [J]. Trans. Chin. Soc. Agric. Eng., 2021,37(14): 274-282.
[3]	MUJUMDARA S. Handbook of Industrial Drying [M].New York: Marcel dekker,1987: 21-25.
[4]	徐永阳,邓安,邱伟强,等.不同含水率稻谷电学物理量变化规律 [J].食品工业,2024,45(2):114-120.
	XU Y Y, DENG A, QIU W Q, et al.. The change law of electrical physical quantity of rice with different moisture [J]. Food Ind., 2024, 45(2):114-120
[5]	王赫,刘国辉,林琳,等.基于神经网络的玉米干燥过程预测控制模型建立[J].粮食与油脂,2021,34(12):37-40.
	WANG H, LIU G H, LIN L, et al.. Predictive control model of corn drying process based on neural network [J]. Cereals Oils, 2021, 34(12): 37-40.
[6]	车刚,高瑞丽,万霖,等.水稻负压混流干燥室结构优化设计与试验[J].农业工程学报,2021,37(4):87-96.
	CHE G, GAO R L, WAN L, et al.. Optimized design and experimental study on rice negative pressure mixed flow drying chamber [J]. Trans. Chin. Soc. Agric. Eng., 2021, 37(4):87-96.
[7]	赵丽清,段东瑶,殷元元,等.基于PSO-Elman算法的茶叶烘干含水率预测[J].农业工程学报,2021,37(19):284-292.
	ZHAO L Q, DUAN D Y, YIN Y Y, et al.. Prediction of tea drying moisture content based on PSO-Elman algorithm [J].Trans. Chin. Soc. Agric. Eng., 2021, 37(19):284-292.
[8]	谢辉煌,金毅,张忠杰,等.基于LSTM网络的粮食干燥机水分预测与优化[J].中国粮油学报,2023,38(11):196-204.
	XIE H H, JIN Y, ZHANG Z J, et al.. Prediction and optimization of grain dryer outlet moisture content based on LSTM [J]. J. Chin. Cereals Oils Assoc., 2023, 38(11):196-204.
[9]	罗建军,杨红云,路艳,等.基于遗传算法优化的BP神经网络进行水稻氮素营养诊断[J].中国农业科技导报,2020,22(8):83-92.
	LUO J J, YANG H Y, LU Y, et al.. Identification of nitrogen nutrition in rice based on BP neural network optimized by genetic algorithms [J]. J. Agric. Sci. Technol., 2020, 22(8):83-92.
[10]	白竣文,周存山,蔡健荣,等.南瓜片真空脉动干燥特性及含水率预测[J].农业工程学报,2017,33(17):290-297.
	BAI J W, ZHOU C S, CAI J R, et al.. Vacuum pulse drying characteristics and moisture content prediction of pumpkin slices [J]. Trans. Chin. Soc. Agric. Eng., 2017, 33(17):290-297.
[11]	代爱妮,周晓光,刘相东,等.基于BP神经网络的旁热式辐射与对流粮食干燥过程模型[J].农业机械学报,2017,48(3):351-360.
	DAI A N, ZHOU X G, LIU X D, et al.. Model of drying process for combined side-heat infrared radiation and convection grain dryer based on BP neural network [J]. Trans. Chin. Soc. Agric.Mach., 2017, 48(3):351-360.
[12]	朱文学,孙淑红,陈鹏涛,等.基于BP神经网络的牡丹花热风干燥含水率预测模型[J].农业机械学报,2011,42(8):128-130, 137.
	ZHU W X, SUN S H, CHEN P T, et al.. Moisture content prediction modeling of hot-air drying for pressed peony based on BP neural network [J]. Trans. Chin. Soc. Agric. Mach., 2011,42(8):128-130, 137.
[13]	陈正发,车刚,万霖,等.稻谷四向通风混流干燥段数值模拟与试验[J].农业工程学报,2022,38(24):237-247.
	CHEN Z F, CHE G, WAN L, et al.. Numerical simulation and experiment of four-way ventilation mixed flow drying section for rice [J]. Trans. Chin. Soc. Agric. Eng., 2022,38(24):237-247.
[14]	卓金武. MATLAB在数学建模中的应用[M].2版.北京:北京航空航天大学出版社,2014:122-123.
[15]	赵辰,南星恒.基于MEA-BP神经网络的财务危机预警研究[J]. 财会通讯,2016(1):43-46.
[16]	刘艺梦,丁小明,王会强,等.基于蜣螂算法优化BP的冬夏生菜根区温度预测模型[J].农业工程学报 2024,40(5):231-238.
	LIU Y M, DING X M, WANG H Q, et al.. Prediction model for winter and summer lettuce root zone temperature based on dung beetle algorithm to optimize BP [J]. Trans. Chin. Soc. Agric. Eng., 2024, 40(5): 231-238.
[17]	周勇,胡中功.改进的快速遗传算法在函数优化中的应用[J].现代电子技术,2018,41(17):153-157.
	ZHOU Y, HU Z G. Application of improved fast-convergent genetic algorithm in function optimization [J]. Mod. Electron.Tech., 2018, 41(17):153-157.
[18]	史峰,王辉,郁磊,等.MATLAB智能算法30个案例分析[M].北京:北京航空航天大学出版社,2011:27-35.
[19]	弭宝福,徐梅,王福林,等.交叉概率对遗传算法运算速度影响的测试与分析[J].数学的实践与认识,2013,43(22):180-183.
	MI B F, XU M, WANG F L, et al.. The testing and analysis of the effects of crossover probability on genetic algorithms’ calculating speed [J]. J. Math. Pract. Theory, 2013, 43(22):180-183.
[20]	张学锋,朱梅,汤亚玲,等.基于GA-BP的烧结风箱阀门预测模型[J/OL].西华大学学报(自然科学版),2024:1-7[2024-04-07]..
	ZHANG X F, ZHU M, TANG Y L, et al.. GA-BP Based Sintering Airbox Valve Prediction Model [J/OL]. J. Xihua Univ. (Nat. Sci.), 2024:1-7 [2024-04-07]. .
[21]	王佳楠,王玉莹,何淑林,等.基于改进遗传算法优化BP神经网络的土壤湿度预测模型[J].计算机系统应用,2022,31(2):273-278.
	WANG J N, WANG Y Y, HE S L, et al.. Optimized BP neural network model based on lmproved genetic algorithm for soil moisture prediction [J]. Comput. Syst. Appl., 2022, 31(2):273-278.
[22]	葛继科,邱玉辉,吴春明,等.遗传算法研究综述[J].计算机应用研究, 2008, 25(10): 2911-2916.
	GE J K, QIU Y H, WU C M, et al.. Summary of genetic algorithms research [J]. Appl.Res.Comput., 2008, 25(10):2911-2916.
[23]	邱进栋,王同勋,万芳新.基于改进BP神经网络的香菇真空远红外辐射干燥含水率预测[J].林业机械与木工设备,2023,51(10):32-38.
	QIU J D, WANG T X, WAN F X.Prediction of moisture content in vacuum far-infrared radiation drying of shiitake mushroom based on improved BP neural network [J]. For. Mach. Woodwork. Equip., 2023, 51(10):32-38.
[24]	梁兆超,郭显炜,宋艳娟,等.响应面法优化双孢蘑菇多糖提取工艺及其体外抗氧化活性研究[J].中国农业科技导报,2019,21(8):161-168.
	LIANG Z C, GUO X W, SONG Y J, et al.. Extraction process of polysaccharide in Agaricus bisporus optimized by response surface method and its antioxidant activity in vitro [J]. J. Agric. Sci. Technol., 2019, 21(8):161-168.