摘要：随着社会的进步发展，传统的农业生产模式愈发不能满足人们的需求。新型农业设施—温室大棚，有着不限于时间空间的优点，广泛的应用于各种环境下的农业生产。而相较于传统温室，自动化温室有着无可比拟的优势。自动化温室的传统的环境因素调控方式是通过复杂的线性函数模型，而采用神经网络对于各个环境因素实现预估从而实现调控相较于传统方式更加方便快捷。神经网络中的递归神经网络在时间序列的预测上有着优异的效果。普通的递归神经网络对于长期依赖的信息没有良好的学习效果，LSTM网络是一种改进的递归神经网络模型，能有效的解决这一问题。在自动化温室的复杂环境下，采用LSTM能够有较好的预测效果。本文工作主要包括一下三个方面：（1） 本文开始是数据的特征处理,以及数据的不同处理方式。（2）其次对与网络模型的搭建，通过对温室环境因素的学习预测，测试模型中不同的参数，使之对比调整使之达到最佳的学习效果。（3）最后是图形界面的编写，实现一个可视化界面，更加便于展示操作。34610
毕业论文关键词：自动化温室；递归神经网络；长短记忆网络；时间序列
Based on the recursive neural network prediction of short series for greenhouse
Abstract： With the progress of the society development, the traditional mode of agricultural production can't meet the needs of people. New type of facilities agricultural—greenhouses, has a merit of that is not limited to time or space. It is widely used in agricultural production in various kinds of environments. Compared with traditional greenhouse, automated greenhouse has incomparable advantages. Tranditional automated greenhouse controls the environmental factors by complex linear function model. Neural network estimates all of the environmental factors, then realizes the control. So neural network is easier and faster than the tranditional way. Recursive neural network has excellent effect in time series prediction in the numerous neural network. While the ordinary recursive neural network is not good in learning effect for long-term dependence on information, LSTM - an improved recursion network model, can effectively solve the problem. So in case of the automated greenhouse under the complex environment factors, adopt LSTM can have better prediction effect.This paper mainly includes three aspects:(1) Firstly, it’s the characteristics of the data processing and different approaches to process data.(2) Secondly, constructs the network model and test different parameters in the model based on the learning and calculating to the environmental factors of the greenhouse. And makes a contrast adjustment to achieve the best learning effect. (3) Finally, writes a graphical interface, makes it easy for operation by a visual interface.
Key words: Automated Greenhouse ;Recurrent Neural Network ;Long Short-Term Memory ;Time Series
目 录
摘要：    1
关键词：    1
Abstract：    1
Key words:    1
1 绪论    2
1.1自动化温室研究背景及现状    2
1.2递归神经网络研究背景及现状    2
1.3 研究内容和技术路线    3
1.3.1研究内容    3
1.3.2 技术路线    4
2 递归神经网络结构及相关函数介绍    4
2.1 RNN    4
2.2 LSTM    7
2.3 激活函数    9
2.3.1 Sigmoid 函数    9
2.3.1 Tanh 函数    10
2.3.2 ReLu 函数    11
2.4 优化函数    11
2.4.1 SGD    12
2.4.2 RMSprop    12
2.4.3 Adam    12 java递归神经网络大棚温室短序列预测+源代码:http://www.youerw.com/jisuanji/lunwen_32168.html