基于ARIMA-DBN的水质参数预测模型研究

发布时间：2018-08-11 21:33

【摘要】：“养鱼先养水”,养殖环境的好坏直接关系到养殖品种的发育和生长,因为养殖水域是养殖品种的生活环境,从而决定着养殖品种产量和质量的高低。近年来,随着水产养殖种类的不断增多,规模化、集约化程度的不断提高以及养殖密度的增加,养殖水域病害发生率越来越高,水质环境也日趋恶化,由此容易引发水产品质量安全问题。因此,为了能够及时掌握和估计养殖品种水质环境状况,需要迫切地构建一种水产养殖水质环境监测和预测系统,从而采取有效的措施调控水质,达到高效和安全生产,并保障水产品质量安全。结合实际情况来看,采用现有的水质预测监测方法存在不及时性、缺乏可靠性、成本高等问题,已远远不能实现上述需求。以规模化、科学化为主要方式是未来水产养殖的发展趋势,从而实现水产品的高成活率和高质量。针对上述问题,本文结合物联网技术,研究了水产养殖水质、水产品流通、销售环节环境参数的自动监测及预测系统,完成了水产养殖水质、物流和销售环境数据和监测点位置信息数据的采集、存储、处理和转发,实现养殖环境实时监控,并根据传感器监测的实时水质参数进行预测预警。本文主要工作归纳如下:(1)通过传感器得到某一固定时间段的水质参数数据,针对数据流传递的水产养殖水质参数可能存在的数据质量问题,提出了基于滑动窗口的水质参数异常检测算法,对原始数据进行异常点排除。(2)深入分析了单整自回归移动平均(ARIMA)模型和深度信念网络(DBN)模型特点的基础上,建立了一种适用于水质参数预测的组合预测模型ARIMA-DBN,并以溶解氧为例与单一的ARIMA模型和DBN分别进行了比较分析,验证了组合预测模型ARIMA-DBN的准确性和有效性。(3)在上述组合预测模型研究基础上,设计并实现了水产养殖、水产品流通及销售环节环境参数的自动监测及预警系统。该系统可以对水产养殖及流通全过程的主要环境因子实施有效监测,并利用组合预测模型ARIMA-DBN对关键环境因子进行预测,若预测结果发现异常,则提前预警。
[Abstract]:The quality of the culture environment is directly related to the development and growth of the breed, because the aquaculture water is the living environment of the breed, which determines the yield and quality of the breed. In recent years, with the increase of aquaculture species, scale, intensive degree and density of aquaculture, the incidence of diseases in aquaculture waters is becoming higher and higher, and the water quality environment is deteriorating day by day. This will easily lead to aquatic product quality and safety problems. Therefore, in order to grasp and estimate the water quality environment of aquaculture varieties in time, it is necessary to establish an environmental monitoring and forecasting system for aquaculture water quality urgently, so as to take effective measures to regulate water quality and achieve high efficiency and safety in production. And to ensure the quality and safety of aquatic products. According to the actual situation, the existing methods of water quality prediction and monitoring are not timely, lack of reliability, high cost and so on. In order to realize the high survival rate and high quality of aquatic products, it is the development trend of aquaculture in the future to take the scale and science as the main way. In view of the above problems, this paper studies the automatic monitoring and forecasting system for the environmental parameters of aquaculture water quality, aquatic product circulation and sales link, combining with the technology of Internet of things, and accomplishes the aquiculture water quality. The collection, storage, processing and forwarding of logistics and sales environment data and location information data of monitoring points are carried out to realize the real-time monitoring of aquaculture environment, and the real-time water quality parameters monitored by sensors are used to predict and early warning. The main work of this paper is summarized as follows: (1) the data quality problem of aquiculture water quality parameters transmitted by data stream can be solved by obtaining the water quality parameter data of a fixed time period through the sensor. An anomaly detection algorithm for water quality parameters based on sliding window is proposed. (2) the characteristics of single integral autoregressive moving average (ARIMA) model and depth belief network (DBN) model are deeply analyzed. A combined forecasting model, ARIMA-DBN, which is suitable for water quality parameter prediction, is established and compared with a single ARIMA model and a DBN model, taking dissolved oxygen as an example. The accuracy and validity of the combined prediction model ARIMA-DBN are verified. (3) based on the research of the combined prediction model, the automatic monitoring and warning system of the environmental parameters in aquaculture, aquatic product circulation and marketing is designed and implemented. The system can effectively monitor the main environmental factors in the whole process of aquaculture and circulation, and predict the key environmental factors by using the combined forecasting model (ARIMA-DBN). If the prediction results are abnormal, early warning will be given.
【学位授予单位】：上海海洋大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S959;TP274

【参考文献】