马尔可夫链的组合模型在降雨量预测中的应用

发布时间：2019-06-16 17:36

【摘要】：由于气候条件的复杂性、多样性和变异性,使得天气预测的过程中存在着大量的模糊性和不确定性,从而导致中长期降雨量的预测成为了计算机预测和气象科学中的一个难点问题。虽然我国幅员辽阔,地形复杂,但降雨量的空间分布仍有一定规律。随着我国经济的快速发展,气象信息越来越受到广泛的关注,无论是人民的生活,还是经济的发展都与气象息息相关。天气预报基于卫星观测,可以较为精确地预测未来数天的风力、湿度、温度等天气状况,但是如何能够知道未来几个月甚至未来几年大概的降雨量变化状况,将是本文研究的重点。马尔可夫链适用于随机波动性大的预测问题,但它要求状态无后效性,对极值的预测不太理想,而组合模型可以通过结合其他算法来优化其缺陷,所以本文基于马尔可夫链的组合模型对降雨量进行预测,并提出了相应的两种预测方式,然后利用Arcgis组件设计与实现了城市降雨分布系统,主要研究工作如下:(1)本文提出了基于马尔可夫链和模糊集组合模型的预测算法。先将降雨量有序聚类,计算出状态转移矩阵,然后引入隶属度概念,求出每个状态对其他状态的影响值,再通过加权的方式算出预测的状态区间,最后通过模糊集公式算出具体的降雨预测值。算法取缔了传统上用的较多的平均分类,然后引入隶属度更详细的描述随机变量序列值分别属于所有状态的隶属程度,较好的减小了原来“非此即彼”思想带来的误差。(2)本文也提出了基于马尔可夫链和灰色理论组合模型的预测算法。该方法引入了滑动GM(1,1)模型,先通过最小二乘法求出预测函数,然后代入具体时间求出第一次预测值及其与实际值的误差,再通过误差进行分类,最后通过加权马尔科夫模型完成预测。该算法主要采用二次预测的方式,降低了偶然因素的影响,简化了建模的步骤,提高了预测的精度和速度。(3)本文查阅并参考国内外关于降雨分布图制作和降雨预测的研究技术,利用插值算法和ArcToolbox构建了等值模型。然后我们通过ArcGIS服务器发布地理信息服务,实现了城市降雨分布图制作的功能。最后通过加权马尔可夫算法,在系统中实现年降雨量预测并制图的功能,为人们的生产生活和有关部门的防灾减灾都提供了便利。通过以上研究,我们论文实现了对马尔可夫组合预测算法的优化,提高了预测的速度和精度,为长期降雨量的预测打下了坚实的基础。
[Abstract]:Due to the complexity, diversity and variability of climate conditions, there are a lot of fuzziness and uncertainty in the process of weather prediction, which leads to the prediction of medium and long term rainfall has become a difficult problem in computer forecasting and meteorological science. Although China has a vast territory and complex terrain, the spatial distribution of rainfall is still regular. With the rapid development of economy in our country, meteorological information has been paid more and more attention, both the life of the people and the development of economy are closely related to meteorology. Based on satellite observation, weather forecast can accurately predict wind, humidity, temperature and other weather conditions in the next few days, but how to know the approximate rainfall changes in the next few months and even in the next few years will be the focus of this paper. Markov chain is suitable for the prediction problem with large random volatility, but it requires that the state has no aftereffect and the prediction of extreme value is not ideal, and the combined model can optimize its defects by combining other algorithms. Therefore, this paper forecasts the rainfall based on the combined model of Markov chain, and puts forward two corresponding forecasting methods, and then designs and implements the urban rainfall distribution system by using Arcgis component. The main research work is as follows: (1) in this paper, a prediction algorithm based on Markov chain and fuzzy set combination model is proposed. Firstly, the rainfall clustering is ordered, the state transition matrix is calculated, then the concept of membership degree is introduced, the influence value of each state on other states is obtained, then the predicted state interval is calculated by weighted method, and finally the specific rainfall prediction value is calculated by fuzzy set formula. The algorithm outlaws the traditional average classification, and then introduces the membership degree to describe the membership degree of the random variable sequence value belonging to all states respectively, which reduces the error caused by the original idea of "either one or the other". (2) this paper also proposes a prediction algorithm based on Markov chain and grey theory combination model. In this method, the sliding GM (1, 1) model is introduced. The prediction function is obtained by the least square method, then the first predicted value and its error with the actual value are calculated by the specific time, then the error is classified by the error, and finally the prediction is completed by the weighted Markov model. The algorithm mainly adopts the method of secondary prediction, which reduces the influence of accidental factors, simplifies the modeling steps, and improves the accuracy and speed of prediction. (3) in this paper, the equivalent model is constructed by using interpolation algorithm and ArcToolbox with reference to the research technology of rainfall distribution map making and rainfall prediction at home and abroad. Then we publish geographic information service through ArcGIS server, and realize the function of making urban rainfall distribution map. Finally, the function of annual rainfall prediction and mapping is realized in the system by weighted Markov algorithm, which provides convenience for people's production and life and disaster prevention and mitigation in related departments. Through the above research, we realize the optimization of Markov combination forecasting algorithm, improve the speed and accuracy of prediction, and lay a solid foundation for long-term rainfall prediction.
【学位授予单位】：扬州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P457.6;O211.62

【相似文献】