涌潮非线性灰箱建模预报理论方法及在钱塘江的应用

The interaction of celestial bodies results in tidal phenomena on earth. It is difficult to forecast river tidal bore due to different geographical environments and climate changes, which causes causalities and property damage every year. Astronomy based forecasting principles in the literature are only suitable for sea tide prediction rather than tidal bore prediction, the latter of which is often inaccurate. With the consideration of tidal bore's uncertainties, the project puts forward nonlinear grey-box based forecast theory for rival tidal bore and applies it to Qiantang river. First, we study the mechanisms of tidal bore on the basis of the laws of astronomy and harmonic analysis methods, which is the problem of white-box modeling. Second, we define the problem of grey-box modeling for tidal bore. The theoretical model is parameterized and appended by a parameterized correction term. The unknown parameters are going to be identified, what is called the black-box modeling problem. Third, we design a field test to collect real data from Qiantang river. The information of acoustic emission, video images, weather conditions are acquired through wireless sensor network and processed through feature extraction techniques. Fourth, we apply nonlinear system identification methods, say, neural network, to solve grey-box modeling problem, using experimental data. Last, we propose a grey-box model based forecast theory for river tidal bore to enhance the accuracy of tidal bore forecasting.

天体相互作用引起地球潮汐现象，江河涌潮受地理环境与气象变化等不确定因素影响而难于预测，常导致人员伤亡和财产损失。文献中基于天文规律的预测方法仅适用于海潮预测，直接应用于涌潮预报不准确。针对具有不确定性特征的江河涌潮，项目结合钱塘江涌潮实例提出涌潮非线性灰箱建模预报理论，研究重点为：1）基于天文规律，借鉴海潮理论（如：调和分析方法），研究江河涌潮形成机理并给出机理模型，即：白箱建模问题；2）针对涌潮的不确定性影响因素，将理论模型进行参数化并给出修正项，所有待定参数与修正项的辨识属于黑箱建模问题，从而定义了江河涌潮的灰箱建模问题；3）基于音频、视频、气象等信息与无线传感网络技术，给出涌潮数据采集实验与特征提取方法，以钱塘江涌潮为实例采集现场试验数据；4）结合实验数据，采用非线性系统辨识方法（如：神经网络）解决灰箱建模问题；5）提出基于灰箱建模的涌潮预报理论，解决江河涌潮预测不准确的难题。

天体相互作用引起地球潮汐现象，江河涌潮受地理环境与气象变化等不确定因素影响而难于预测，常导致人员伤亡和财产损失。本项目针对江河涌潮不确定因素多、缺乏理论支撑、以及现有预测方法不准确的现状，结合海洋潮汐理论、涌潮经验方法、与系统辨识方法，提出了基于非线性灰箱建模的涌潮预测理论，并成功应用于钱塘江涌潮的预测。基于音频、视频、气象等信息与无线传感网络技术，给出涌潮数据采集实验与特征提取方法。并以钱塘江涌潮为实例采集了现场试验数据；结合实验数据，采用多种非线性系统辨识方法进行了灰箱建模，解决了江河涌潮预测不准确的难题。