基于Web技术的虫害预测系统的研究

发布时间：2018-06-24 22:02

本文选题：虫害预测系统 + PSO　；参考：《浙江理工大学》2017年硕士论文

【摘要】：我国作为农业大国,农作物的产量占据重要地位。虫害的发生对农作物的生产造成了严重的危害。因此,及时、准确的对虫害的发生进行预测预报,才能为虫害的防治工作提供基础,才能有利于农业的良好发展,减少损失。随着信息技术的发展,计算机网络,数据管理等技术早已被广泛应用到虫害的监测和预测等领域。为了适应农作物虫害资料的规范化、信息化、网络化的要求,以及相关人员日常工作的需求,本文构建了一个基于Web技术的虫害预测系统,该系统集虫害数据管理、查询、预测于一体。虫害的发生不仅受天气、农作物的生长情况等的影响,还具有地域上的差别,虫害数据自身又包含显著的动态时序特征,很难对其做到精准的预测。为了能及时和准确的对虫害的发生进行预测,本文主要研究内容为:(1)针对农业中虫害发生数据的特点,以及对相关基础理论的研究,以偏最小二乘支持向量机(LSSVM)为核心,建立了针对农业虫害发生的预测模型(PLS-PSO-LSSVM)。首先,针对LSSVM参数寻优的问题,提出用遗传算法(GA)和粒子群算法(PSO)对其参数寻优进行改进。其次,针对虫害发生影响因素之间的共线性问题,提出用偏最小二乘法(PLS)对虫害发生影响因素进行主成分提取。(2)基于区域的虫害发生量的预测。以稻飞虱为例,通过对其近30年发生情况的分析,稻飞虱的发生不仅受到温度、湿度、降雨、日照的影响,发生的情况还有地域上的差别。因此,本文以具体地区的虫害影响因素作为预测因子,结合本文建立的预测模型进行预测,并与BP神经网络、偏最小二乘(PLS)模型进行对比分析,得出本文建立的预测模型的预测精度更高。(3)根据基于Web技术的虫害预测系统的设计需求,对系统进行了总体设计和详细设计分析,并完成了系统的开发。实现了实时虫害数据监测、历史虫害数据查询、报表设计分析、虫害预测等功能模块。实现了数据的集中管理,及时将虫情发送到相应的用户手中。
[Abstract]:As a large agricultural country, the output of crops occupies an important position in China. The occurrence of insect pests has caused serious harm to the production of crops. Therefore, timely and accurate prediction of the occurrence of insect pests can provide a basis for pest control, can be conducive to the good development of agriculture and reduce losses. With the development of information technology, computer network, data management and other technologies have been widely used in pest monitoring and forecasting. In order to meet the requirements of standardization, information and networking of crop pest data and the daily work of related personnel, a pest prediction system based on Web technology is constructed in this paper. Prediction is one thing. The occurrence of insect pests is not only affected by the weather and crop growth, but also has regional differences. The pest data itself contains significant dynamic temporal characteristics, it is difficult to accurately predict them. In order to predict the occurrence of insect pests in time and accurately, the main contents of this paper are as follows: (1) according to the characteristics of pest occurrence data in agriculture and the research of related basic theories, partial least squares support vector machine (LSSVM) is the core. A prediction model for agricultural pest occurrence (PLS-PSO-LSSVM) was established. Firstly, to solve the problem of LSSVM parameter optimization, genetic algorithm (GA) and particle swarm optimization (PSO) are proposed to improve the optimization of LSSVM parameters. Secondly, aiming at the problem of collinearity between the influencing factors of pest occurrence, the partial least square (PLS) method is proposed to extract the principal components of the influencing factors of pest occurrence. (2) the prediction of pest occurrence quantity based on region. Taking rice planthopper as an example, the occurrence of rice planthopper is not only affected by temperature, humidity, rainfall and sunshine, but also by regional difference through the analysis of the occurrence of rice planthopper in recent 30 years. Therefore, this paper takes the pest influence factors of specific area as the prediction factor, combining the prediction model established in this paper, and carries on the comparison analysis with the BP neural network, partial least squares (PLS) model. It is concluded that the prediction model established in this paper has higher prediction accuracy. (3) according to the design requirements of the pest prediction system based on Web technology, the overall design and detailed design analysis of the system are carried out, and the development of the system is completed. Real-time pest data monitoring, historical pest data query, report design and analysis, pest prediction and other functional modules are realized. The centralized management of the data is realized, and the bug situation is sent to the corresponding users in time.
【学位授予单位】：浙江理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S431;TP393.09

【相似文献】