Simulation of Water Demand for Farmland Subarea Irrigation Based on Bayesian Neural Network
-
摘要:
针对农田分区灌溉需水量模拟过程中普遍存在的求解过程易陷入局部最小化、出现过度拟合,以及过度依赖历史用水数据,导致最终模拟结果存在显著误差的问题,研究基于贝叶斯神经网络的农田分区灌溉需水量模拟分析方法。以前一周需水量、年内月需水量占比、日内温度上限值及日降雨量为指标,通过聚类分析获取指标数据均值,对农田分区灌溉历史用水的样本数据进行聚类分析。构建贝叶斯神经网络模型,将指标数据均值输入模型,根据BP神经网络原理与贝叶斯规则训练指标数据,然后输出农田分区灌溉需水量模拟结果。试验结果显示数据聚类结果中数据间关联度高于95%,数据拟合效果较好,模拟需水量时具有更高的精度与稳定性。
Abstract:Aiming at problem that common solution process in simulation process of farmland partitioned irrigation water demand was prone to fall into local minimization, excessive fitting, and excessive dependence on historical water use data, resulting in significant errors in final simulation results, a simulation analysis method of farmland partitioned irrigation water demand based on Bayesian neural network was studied. Water demand in previous week, proportion of monthly water demand in the year, upper limit of daily temperature and daily rainfall were taken as indicators, and average value of indicator data was obtained through cluster analysis, and sample data of historical water use for farmland subarea irrigation were clustered. Bayesian neural network model was built. Average value of index data was input into model, and indicator data were trained according to BP neural network principle and Bayesian rule, at last, simulation results of irrigation water demand of farmland were output. Experimental results showed that correlation between data in data clustering results was higher than 95%, data fitting effect was good, and simulation of water demand had higher accuracy and stability.
-
表 1 需水量模拟结果与实际监测结果对比
Table 1. Comparison of water demand simulation results and actual monitoring results
日期 需水量监测值/m3 需水量模拟值/m3 基于BP神经网络的方法 基于DSSAT模型的方法 本文方法 2020年7月16日 1175348 1184933 1189512 1179212 2020年7月17日 1224679 1209534 1208416 1224079 2020年7月18日 1201622 1212200 1217056 1215846 2020年7月19日 1169951 1171534 1171469 1170268 2020年7月20日 1172316 1180275 1180719 1173498 2020年7月21日 1236301 1226513 1225474 1235495 2020年7月22日 1122917 1114925 1139815 1120687 2020年7月23日 721920 726874 730005 726541 2020年7月24日 690349 735870 736841 700042 2020年7月25日 614468 648412 641587 618714 2020年7月26日 1284574 1348874 1205879 1290475 2020年7月27日 987223 1025499 1026987 981274 2020年7月28日 1003856 1058624 1049856 1007486 2020年7月29日 1247372 1189565 1279521 1239577 2020年7月30日 594371 578784 625410 584265 表 2 数据聚类结果关联度
Table 2. Correlation degree of data clustering results
单位:% 指标 区域A 区域B 区域C 区域D Xn-1 97.91 95.74 97.98 96.46 Xn-2 98.26 96.05 96.82 97.21 Xn-3 96.04 98.23 97.00 95.84 Xn-4 98.73 96.47 96.59 95.63 Xn-5 95.07 97.36 97.86 96.84 Xn-6 96.95 96.88 95.81 96.28 Xn-7 97.22 98.29 96.42 97.06 Tn 95.99 96.07 97.50 98.44 Rn 96.54 96.85 96.23 96.51 Sn 97.30 98.06 95.89 97.92 -
[1] 程亮,胡霞,王宗志,等.考虑水文气象贡献的区域用水总量驱动因子识别[J].应用基础与工程科学学报,2020,28(6): 1350- 1360. doi: 10.16058/j.issn.1005-0930.2020.06.008CHENG Liang,HU Xia,WANG Zongzhi,et al.A regional total water use driver identification model considering the contribution of hydro-climatic variability[J].Journal of Basic Science and Engineering,2020,28(6): 1350- 1360. doi: 10.16058/j.issn.1005-0930.2020.06.008 [2] 魏童彤,彭致功,刘福胜,等.华北小麦-玉米轮作净灌溉需水量变化趋势:以大兴区为例[J].灌溉排水学报,2019,38(9):63-72.WEI Tongtong,PENG Zhigong,LIU Fusheng,et al.Trend of net irrigation water requirement of wheat-maize rotation in north China:a case study of Daxing District[J].Journal of Irrigation and Drainage,2019,38(9):63-72. [3] 马建花,曾润强,边世强,等.基于Visual Modflow的河流阶地灌溉区地下水模拟分析:以黑台为例[J].兰州大学学报(自然科学版),2020,56(5):580-586.MA Jianhua,ZENG Runqiang,BIAN Shiqiang,et al.Simulation and analysis of groundwater in irrigation in a river terrace based on the Visual Modflow:taking the Heitai platform as an example[J].Journal of Lanzhou University(Natural Sciences),2020,56(5):580-586. [4] 张晓斌,李抗彬,郝改瑞,等.基于BP神经网络的新安江模型初始土壤蓄水量计算研究[J].灌溉排水学报,2021,40(3):15-22. doi: 10.13522/j.cnki.ggps.2020324ZHANG Xiaobin,LI Kangbin,HAO Gairui,et al.Using BP network to estimate initial soil water storage in Xin'anjiang model[J].Journal of Irrigation and Drainage,2021,40(3):15-22. doi: 10.13522/j.cnki.ggps.2020324 [5] 郭兵托,孙素艳,张金萍,等.基于多尺度小波变换WNN的灌区灌溉水量研究[J].人民黄河,2019,41(11):154-158. doi: 10.3969/j.issn.1000-1379.2019.11.032GUO Bingtuo,SUN Suyan,ZHANG Jinping,et al.Study of irrigation water amount in irrigation area based on multi-scale wavelet transform WNN[J].Yellow River,2019,41(11):154-158. doi: 10.3969/j.issn.1000-1379.2019.11.032 [6] 刘影,关小康,杨明达,等.基于DSSAT模型对豫北地区夏玉米灌溉制度的优化模拟[J].生态学报,2019,39(14): 5348- 5358.LIU Ying,GUAN Xiaokang,YANG Mingda,et al.Optimizing the summer maize irrigation schedule in north Henan Province based on the DSSAT model[J].Acta Ecologica Sinica,2019,39(14): 5348- 5358. [7] 王会青,郭芷榕,白莹莹.基于BP和朴素贝叶斯的时间序列分类模型[J].计算机应用研究,2019,36(8): 2271- 2274,2278. doi: 10.19734/j.issn.1001-3695.2018.02.0081WANG Huiqing,GUO Zhirong,BAI Yingying.Time series classification model based on BP and naive Bayes[J].Application Research of Computers,2019,36(8): 2271- 2274,2278. doi: 10.19734/j.issn.1001-3695.2018.02.0081 [8] 杨珉,汪洁.解决深度探索问题的贝叶斯深度强化学习算法[J].计算机科学与探索,2020,14(2):307-316. doi: 10.3778/j.issn.1673-9418.1901020YANG Min,WANG Jie.Bayesian deep reinforcement learning algorithm for solving deep exploration problems[J].Journal of Frontiers of Computer Science and Technology,2020,14(2):307-316. doi: 10.3778/j.issn.1673-9418.1901020 [9] 张千,魏正英,张育斌,等.基于广义回归神经网络的灌溉系统首部多用户配水快速PID控制[J].农业工程学报,2020,36(11):103-109. doi: 10.11975/j.issn.1002-6819.2020.11.012ZHANG Qian,WEI Zhengying,ZHANG Yubin,et al.Rapid-response PID control technology based on generalized regression neural network for multi-user water distribution of irrigation system head[J].Transactions of the Chinese Society of Agricultural Engineering,2020,36(11):103-109. doi: 10.11975/j.issn.1002-6819.2020.11.012 [10] 陈森朋,吴佳,陈修云.基于强化学习的超参数优化方法[J].小型微型计算机系统,2020,41(4):679-684. doi: 10.3969/j.issn.1000-1220.2020.04.002CHEN Senpeng,WU Jia,CHEN Xiuyun.Hyperparameter optimization method based on reinforcement learning[J].Journal of Chinese Computer Systems,2020,41(4):679-684. doi: 10.3969/j.issn.1000-1220.2020.04.002 [11] 聂志刚,任新庄,李广,等.旱地小麦产量及其构成因素灌溉效应的模拟分析[J].干旱地区农业研究,2019,37(3):117-122. doi: 10.7606/j.issn.1000-7601.2019.03.15NIE Zhigang,REN Xinzhuang,LI Guang,et al.Simulation analysis of irrigation effect on yield and its constitute factors of dryland wheat[J].Agricultural Research in the Arid Areas,2019,37(3):117-122. doi: 10.7606/j.issn.1000-7601.2019.03.15 [12] 邓帅.基于改进贝叶斯优化算法的CNN超参数优化方法[J].计算机应用研究,2019,36(7): 1984- 1987. doi: 10.19734/j.issn.1001-3695.2018.01.0021DENG Shuai.Hyper-parameter optimization of CNN based on improved Bayesian optimization algorithm[J].Application Research of Computers,2019,36(7): 1984- 1987. doi: 10.19734/j.issn.1001-3695.2018.01.0021 [13] 常振,段先华,鲁文超,等.基于多尺度的贝叶斯模型显著性检测[J].计算机工程与应用,2020,56(11):207-213. doi: 10.3778/j.issn.1002-8331.1905-0169CHANG Zhen,DUAN Xianhua,LU Wenchao,et al.Multi-scale saliency detection based on Bayesian framework[J].Computer Engineering and Applications,2020,56(11):207-213. doi: 10.3778/j.issn.1002-8331.1905-0169 [14] 宋倩玉,徐征和,马蒙,等.基于耗水均衡的东北水田灌区地下水数值模拟与水源配置[J].灌溉排水学报,2021,40(6):140-146. doi: 10.13522/j.cnki.ggps.2021111SONG Qianyu,XU Zhenghe,MA Meng,et al.Groundwater numerical simulation and water source allocation of paddy field irrigation area in northeast China based on water consumption balance[J].Journal of Irrigation and Drainage,2021,40(6):140-146. doi: 10.13522/j.cnki.ggps.2021111 [15] 施海洋,罗格平,郑宏伟,等.基于“水−能源−食物−生态”纽带因果关系和贝叶斯网络的锡尔河流域用水分析[J].地理学报,2020,75(5): 1036- 1052. doi: 10.11821/dlxb202005011SHI Haiyang,LUO Geping,ZHENG Hongwei,et al.Water use analysis of Syr Darya River basin:based on "water-energy-food-ecology" nexus and Bayesian network[J].Acta Geographica Sinica,2020,75(5): 1036- 1052. doi: 10.11821/dlxb202005011