杭州市农田土壤重金属污染预警仿真研究-管理科学与工程专业论文.docxVIP

Dissertation Submitted to Hangzhou Dianzi University for the Degree of Master The Simulation Study of Heavy Metal Pollution in Farmlands in Hangzhou Candidate: Ding Shufang Supervisor: Prof. Xie Zhengmiao Prof. Weihong Wu Dec, 2013 杭州电子科技大学硕士学位论文 杭州电子科技大学硕士学位论文 I I 摘 要 随着工业化和城市化进程加快，土壤重金属污染给环境和人类的健康带来了 严重的威胁，开展土壤重金属污染方面的工作已经迫在眉睫。土壤重金属污染预 测、预警可以起到防患于未然的作用。如何建立一个适用性强、精确度高的土壤 重金属预测预警模型，是我们未来研究的方向。 土壤重金属污染系统是一个巨大的非线性系统，其浓度受多种因素影响，找 出影响因素和浓度数据之间的函数关系，把握规律进行预测预警。BP 神经网络 具有广泛的非线性影射能力、适应能力和学习能力，通过已有影响因素和浓度数 据之间关系的学习，用于土壤重金属污染预测预警具有一定的优势。 本文借鉴相关预警指标体系和对土壤重金属污染预测方法进行比较，以此选 择 BP 神经网络对土壤重金属污染变化趋势进行预测。BP 算法具有学习速度慢、 容易陷入局部极小值、预测精度不高等问题，采用比较普遍的自适应学习速率调 整法和附加动量法对其进行改进。以此建立基于 BP 神经网络的土壤重金属污染 预测预警模型，在 MATLAB 环境下对预测过程进行仿真实验。通过对杭州市农 田土壤重金属污染实证研究，结果发现 BP 神经网络的拟合精度达到 98%以上， 运用 BP 网络模型对农田土壤重金属污染进行预测，改进后的预测精度分别在 82%以上，明显高于未改进的 BP 网络模型 70%的预测精度。这证明了本文所构 建的指标及构建的预测模型是实用有效的。研究结果还表明未来杭州市农田平均 处于安全水平，对人体和环境不会造成威胁，但是 Pb 和 Cr 积累将会较严重。 关键词: 重金属；农田；预警；BP II II ABSTRACT With the acceleration of industrialization and urbanization, heavy metal pollution poses a serious threat to humans health and the environment. Prediction and early-warning of heavy metal pollution can play a proactive role preventing and controlling soil contamination. How to build a prediction and early warning model is an area that would be further investigated. Heavy metal pollution system is a huge nonlinear system and is influenced by many factors. Due to the the BP neural network has good non-linear mapping, adaption and learning ability, it not only can learn the the law of relationship between factors and content of heavy metal, but also predict and warn heavy metal pollution. The classification of early-warning index and compares various heavy metal forecasting methods was analyzed in this dissertations. BP neural network was used to forecast heavy metal pollution trend. There are some limitations of BP algorithms, such as the slow learning speed, local extremum and the low prediction precision, which was improved by use o