基于粒子群算法优化BP神经网络的土壤含水量短期预测模型

为提高土壤含水量预测精度,基于物联网监测数据,提出了粒子群算法(PSO)优化BP神经网络的土壤含水量预测方法。首先应用主成分分析法筛选出影响土壤含水量的关键影响因子,然后构建8-5-1的BP神经网络拓扑结构,应用粒子群算法优化BP神经网络的初始权值和阈值。结果表明:与传统BP神经网络相比,新模型优化了网络结构,避免了陷入局部最优解,具有良好的预测效果;模型的评价指标平均绝对误差、平均绝对百分误差、误差均方根分别为0.259 2、0.010 5和0.135 6,与单一BP神经网络相比,预测精度更高,可满足实际的土壤含水量预测的需要。     

基于分子生态学网络分析松嫩退化草地土壤微生物群落对施氮的响应

【目的】 氮素输入影响着全球草地生态系统的可持续性,关注施氮对土壤微生物群落的影响及其分子生态网络,为草地退化修复提供理论依据。【方法】 以松嫩退化羊草草地为研究对象,通过施氮和未施氮处理,利用高通量测序和随机矩阵网络构建理论构建土壤微生物群落分子生态网络。探讨氮素管理对退化羊草草地土壤微生物群落结构及网络的影响,氮添加条件下微生物网络结构中的关键微生物变化规律,以及该过程中微生物之间的互作关系,解析外源氮素添加条件下土壤细菌动态变化的关键结点和规律。【结果】 在门分类水平上施氮处理草地有细菌门22个,未施氮处理23个。7个菌门是施氮和未施氮处理草地的优势菌门,其中变形菌门（Proteobacteria）是含有OTU数量最多的门类,约占总序列的30.46%,酸杆菌门（Acidobacteria）是含有OTU数量次之的门类,约占总序列的30.15%,芽单胞菌门（Gemmatimonadetes）是含有OTU数量第3的门类,约占总序列的8.14%,放线菌（Actinomycete）约占总序列的6.15%,绿弯菌门（Chloroflexi）、拟杆菌门（Bacteroidetes）和硝化螺旋菌门（Nitrospirae）三者约占总序列的17.16%。施氮处理草地土壤微生物中的变形菌门、放线菌门、拟杆菌门的相对丰度均显著高于未施氮处理草地土壤（P<0.01）;未施氮草地土壤中绿弯菌门、酸杆菌门、芽单胞菌门相对丰度显著高于施氮草地土壤（P<0.01）,其他各门细菌施氮与未施氮处理间未呈现出显著差异。表征网络的正向连接比、平均路径长度、平均聚类系数和模块性均为施氮处理显著低于未施氮处理（P<0.001）。在土壤的分子生态网络中,未施氮处理有16个模块枢纽（Zi>2.5,Pi≤0.62）,施氮处理有6个模块枢纽,均属于酸杆菌门、芽单胞菌门和放线菌门。施氮导致土壤微生物种间关系改变,进而改变土壤整体生态网络。【结论】 施氮降低了退化草地土壤网络结构的复杂程度和紧密性;降低了退化草地土壤中的酸杆菌门和绿弯菌门相对丰度,提高了变形菌门、放线菌门和芽单胞菌门相对丰度。土壤中微生物关键物种（OTU）由16个（未施氮）减少为6个,且二者土壤中均没有重叠OTU,这表明施氮调控微生物群落网络的关键物种,进而改变其分子生态网络结构。     

基于PLSR-BP复合模型的红壤有机质含量反演研究

对红壤地区土壤有机质进行快速预测,以满足智慧农业与精准施肥的需要。以江西省奉新县北部为研究区域,采用1 km×1 km标准格网划分研究区进行采样,共得到红壤样本248个。对土壤光谱进行了包含分数阶导数在内的3种数学变换方法,将经过P=0.01显著性检验的波段用于模型的构建,选用偏最小二乘回归(PLSR)和BP神经网络建立土壤有机质含量预测模型。结果表明:当对红壤光谱数据进行1.5阶导数变换后再使用PLSR-BP复合模型对土壤有机质含量进行预测时的结果为最优,训练集R~2=0.89,RMSE=4.68g·kg~(-1),验证集R~2=0.87,RMSE=5.55g·kg~(-1),RPD=2.75。1.5阶导数对红壤光谱数据的变换能够更好地突出与有机质相关的特征信息,有助于其含量预测。PLSR-BP复合模型预测精度优于单一模型,能够较好地预测红壤有机质含量,为精准农业快速监测红壤有机质含量提供了新的途径。     

Dough and bread made from high- and low-protein flours by vacuum mixing: Part 1: Gluten network formation

The effects of reduced headspace pressure on the development of gluten network in doughs made from both high-protein flour (HPF) and low-protein flour (LPF) were investigated. The effect of vacuum mixing was found to be dependent on both flour-type and mixing-time. A significant increase in dough extensibility was observed when the HPF dough was mixed under moderate vacuum of −0.04 MPa for 3 min as compared to the one mixed under atmospheric pressure for the same duration. This was attributed to the formation of a more extensive gluten network associated with an increased disulphide bond density and a significantly higher β-sheet to β-turn ratio. On the other hand, over-mixing was observed in the LPF dough that was mixed for 5 min under atmospheric pressure. Applying moderate vacuum of −0.04 MPa allowed the LPF dough to withstand longer mixing time, as indicated by its increased disulphide bond density and biaxial extensibility compared to the control dough mixed under atmospheric pressure. Results of this study suggest the potential of applying vacuum to reduce the mixing time required for high protein flour and to prevent the over-mixing of low protein flour.     

Turbid water induces refuge behaviour of a commercially important ayu: A field experiment for interstream movement using multiple artificial streams

Riverine systems are inherently dendritic in nature, and turbid water may not necessarily flow equally throughout an entire catchment. Tributaries have the potential to serve as refugia from temporarily turbid water in the main stem. As ayu (Plecoglossus altivelis) are one of the important species for the inland fisheries industry and recreational fishing, their behaviour in a branched river network is a primary concern. We released ayu with radio‐transmitting tags into two experimental streams that joined at a downstream pool. Turbid water was released into one of the streams, and we identified whether avoidance movement of ayu between the two experimental streams was triggered by flowing turbid water. The suspended solid concentration elevated rapidly after adding the turbid water and was maintained at more than 400 mg/L for 3 hr. Avoidance movement of ayu was triggered when the suspended solid concentration exceeded approximately 200 mg/L. Then, most ayu moved from the turbid stream to the nonturbid one, leading to a difference in ayu density between the two streams. Therefore, we demonstrated that turbid water induced interstream movement of a commercially important ayu as refuge behaviour, suggesting that river connectivity is important for the persistence of the ayu.     

Artificial neural network modeling of microbial community structures in the Atlantic Forest of Brazil

Microbial communities vary across the landscape in forest soils, but prediction of their biomass and composition is a difficult challenge due to the large numbers of variables that influence their community structures. Here we examine the use of artificial neural network (ANN) models for extraction of patterns among soil chemical variables and microbial community structures in forest soils from three regions of the Atlantic Forest of Brazil. At each location, variations in soil chemical properties and FAME profiles of microbial community structures were mapped at 20 × 20 m intervals within 10 ha parcels. Geostatistical analyses showed that spatial variability in soil physical and chemical variables could be mapped at scale distances of 20 m, but that FAME profiles representing the microbial communities were highly variable and had no spatial dependence at the same scale in most cases. RDA analysis showed that FAME signatures representing different microbial groups were positively associated with soil pH, OM, P and base cations concentrations, whereas microbial biomass was negatively associated with the same environmental factors. In contrast, ANN models revealed clear relationships between microbial community structures at each parcel location, and generated verifiable predictions of variations in FAME profiles in relation to soil pH, texture, and the relative abundances of base cations. The results suggest that ANN modeling provides a useful approach for describing the relationships between microbial community structures and soil properties in tropical forest soils that were not able to be captured using geostatistical and RDA analyses.     

基于多能互补的光伏木耳大棚智慧控制系统设计

【目的】针对光伏木耳大棚的智能化管理和现代设施农业资源高效利用问题,设计光伏农业大棚供能系统和改造传统大棚控制方式。【方法】设计一种以光伏发电为主的光、水、热、储微能源网络和物联网智慧控制系统,建立数学模型分析能源需求和供应,同时利用控制系统为木耳生产、管理提供远程客户端服务。【结果】微能源网络系统中,光伏发电量为13 654 (kW·h)/a,占年总发电量的92%,直接用于需求侧负荷;水轮机发电量为1 108 (kW·h)/a,占年总发电量的8%,用于无光环境下的系统补能;盈余电能为5 109 (kW·h)/a,占年总发电量的34.6%,用于相变储热和蓄电池储能,或者送入配电网。控制系统采集点温度正常值设置为20～25℃,预警值设置为15～32℃,采集点空气相对湿度正常值设置为90%～95%,预警值设置为80%～98%,测试过程中通风系统、供热装置和报警系统正常工作。【结论】通过不同状态下微能源网络系统的功率调配,能够提高光伏大棚能源利用率,同时自动调控大棚环境参数,满足了木耳大棚远程遥控和预警要求。     

基于深度学习的大豆生长期叶片缺素症状检测方法

为了检测作物叶片缺素,提出了一种基于神经网络的大豆叶片缺素视觉检测方法。在对大豆缺素叶片进行特征分析后,采用深度学习技术,利用Mask R-CNN模型对固定摄像头采集的叶片图像进行分割,以去除背景特征,并利用VGG16模型进行缺素分类。首先通过摄像头采集水培大豆叶片图像,对大豆叶片图像进行人工标记,建立大豆叶片图像分割任务的训练集和测试集,通过预训练确定模型的初始参数,并使用较低的学习率训练Mask RCNN模型,训练后的模型在测试集上对背景遮挡的大豆单叶片和多叶片分割的马修斯相关系数分别达到了0.847和0.788。通过预训练确定模型的初始参数,使用训练全连接层的方法训练VGG16模型,训练的模型在测试集上的分类准确率为89.42%。通过将特征明显的叶片归类为两类缺氮特征和4类缺磷特征,分析讨论了模型的不足之处。本文算法检测一幅100万像素的图像平均运行时间为0.8 s,且对复杂背景下大豆叶片缺素分类有较好的检测效果,可为农业自动化生产中植株缺素情况估计提供技术支持。     

高校科研合作网络及演化研究——以京津冀地区211及省部共建高校为例

[目的/意义]京津冀协同发展是党中央作出的重大战略决策,科研合作是京津冀协同创新发展的重要方面。[方法/过程]笔者基于CNKI检索京津冀地区211及省部共建的26所高校在2008年—2017年发表的合著论文数,运用定量分析、社会网络分析方法,借助Ucinet绘制科研合作网络图谱,从而探讨京津冀地区高校科研合作网络现状及演化规律。[结果/结论]研究结果发现,高校科研实力强劲是形成科研合作的基础,清华大学、北京大学、中国人民大学、北京师范大学、北京科技大学、北京航空航天大学和北京交通大学在科研合作网络中处于核心地位;地域聚集特性是影响科研合作关系强弱的主要因素;各类大学归属于不同的块模型,占据不同位置和角色,具有不同的结构功能和作用;京津冀地区高校科研合作经过10年间的发展,已经形成一个动态演化的网络。     

基于WT-CNN-LSTM的溶解氧含量预测模型

溶解氧（Dissolved oxygen, DO）含量是影响水产养殖产量的重要因素之一，具有时序性、不稳定性和非线性等特点，且其影响因子过多、存在复杂的耦合关系，难以实现精准预测。针对传统长短时记忆神经网络（Long short term memory, LSTM）预测模型易引入冗余数据，且在训练过长序列时会出现梯度消失现象，从而不能捕捉因子间长期的依赖性问题，提出了基于小波变换（Wavelet transform, WT）、卷积神经网络（Convolutional neural network, CNN）和LSTM的溶解氧含量预测模型。首先，使用WT降低数据噪声；然后，使用CNN深度挖掘各变量之间的潜在关系；最后，利用LSTM的时序性预测2h后的水产养殖溶解氧含量。结果表明，本文提出的WT-CNN-LSTM模型预测效果良好，其平均绝对误差、均方根误差和决定系数分别为0.138、0.229和0.954，比传统LSTM模型分别优化了28.87%、21.03%和4.61%。