Application Of System Identification On Water-level/outflow Relationship

In hydrology, one of the emphases on codifying outflow datum is how to correctly establish water-level/outflow relationship. As we known, the traditional method solving this problem is Manning 's equation, which is deduced from wide shallow artificially channels. Because the situation of an actual river is so complex (not as artificially channels) that using Manning's equation will lead to error inevitably. Besides, as it is very difficult in using Manning's equation to compute directly, traditional method made some approximation in application. It generated errors too. For resolving these problems existed in traditional method, scholars have put forward many new methods and some improvements such as GA, multinomial-fitting method have been brought forward.In this paper, we analyzed water-level/outflow relationship using system identification. System identification is an algorithm with feedback. It can modify model parameters step by step according to concrete situation.Because it bases on difference equation that signifies now time system output with the weighted sum of last time output and now time input. It can modify part error in some degree. Considering the uncertainty of an actual river, we could use system identification to deal with water-level/outflow relationship. For using system identification, we regarded an actual river as a SISO system. Given a water level data, we could derive an outflow data. Based on this, we used system identification as a tool to dispose the relationship. This paper is divided into five parts:Chapter I: The introduction and analysis of the traditional method.Chapter II: The analyses of existing methods such as GA, zero-point error modified method, multinomial-fitting method and so on.Chapter III: The introduction of the principia of system identification, including how to use F order probation identify the order of system and use Newton method identify the parameters of system.Chapter IV: Using the datum from Sichuan province Dujiang weir, we attained the efforts of all methods that we have known. By analyzing these results, we have derived the goodness and shortcoming of every method and reasoned the cause.Simulated results manifested that the traditional method may cause great error; the method of zero-point error modified method and multinomial-fitting method can great improve precision. Besides, it could notably improve fit precision using system identification to deal with water-level/outflow. Relationship.