【作者】 梁健；

【导师】 鲍文；

【作者基本信息】 哈尔滨工业大学 ， 动力机械及工程， 2007， 硕士

【摘要】 随着工业技术的发展,电厂对于汽轮机运行过程中状态监测的要求也越来越高,虽然目前电厂监测设备与水平已有长足进步,但是仍然有很多问题趋待解决。其中汽轮机漏汽现象就是一个尚未解决的关键问题,发生于汽轮机的漏汽不仅对汽轮机安全运行构成潜在威胁,影响汽轮机的启动时间,而且会影响汽轮机组的热经济性,导致汽轮机组相对内效率下降,从而使机组的热效率下降,做功能力降低,影响机组的经济性。由于汽轮机状态多变,工作环境复杂,一直以来对于汽轮机漏汽情况没有很好的监测手段,使现场技术人员无法实时了解漏汽的变化趋势,那么一旦出现严重的漏汽问题,就只能在机组大修时进行处理和解决,所以及时了解漏汽的变化情况是必要的。如何解决这一问题,已经越来越成为人们关注的热点。本文针对汽轮机级间漏汽难以直接测量这一问题,充分借助数学模型这一重要的分析问题手段,对漏汽问题进行了系统的研究。具体内容如下:1、建立汽轮机漏汽模型并进行仿真研究数学模型是针对复杂对象的一种行之有效的分析手段,而实际运行中的汽轮机结构复杂,状态多变,现场进行漏汽分析困难。本文从控制角度出发,以哈尔滨汽轮机厂生产的200MW汽轮机为原型,分析了简化的汽轮机结构,详细解释了各部分的工作机理及数学模型,从而建立了汽轮机动态分析模型。进而对汽轮机级间漏汽现象进行了详细的机理分析,指出了漏汽产生的原因,引入了表征漏汽变化趋势的特征指标——漏汽系数,该系数同时也刻画了级内漏汽面积的变化。根据漏汽机理,内部漏汽必然会使级后参数发生变化,在对级间漏汽现象进行了合理假设后,结合之前的动态分析模型建立了汽轮机漏汽仿真模型。对于漏汽现象进行了针对性的仿真研究,特别是考察了在汽轮机变工况运行中随着漏汽系数的变化给级后参数所带来的影响,为接下来的漏汽估计模型提供了大量的仿真数据。2、基于最小二乘的漏汽估计模型完成仿真研究之后,接下来要考虑的就是如何建立漏汽估计模型?从工程应用角度考虑,现场数据由于通过传感器采集,不可避免地存在各种干扰成分,这些成分会影响估计模型的精度与泛化能力,因此在建模之前首先要理解数据的干扰项有哪些?本章对一些典型的干扰成分进行了分析,指出了温度的“时滞”效应与传感器的噪声来源,利用最小二乘的抑噪能力建立了漏汽估计模型,并比较了温度“时滞”对模型的影响,表明模型能够很好的抑制各种干扰成分,具备较强的容错能力。3、基于人工神经网络的漏汽估计模型近年来作为软测量领域研究热点的人工神经网络为级间漏汽估计提供了另外一种思路。人工神经网络是软测量技术中极其重要的一种建模手段,因其优秀的非线性处理能力与容错能力而具有广阔的应用领域。本文第四章采用神经网络建模方法,通过比较,选取了目前应用广泛的RBF网络结构建立漏汽估计模型,并且引入了关于对象的先验知识以改进模型的精度与泛化能力,通过验证表明模型具有较高的精度与良好的泛化能力。最后比较了神经网络模型与最小二乘模型各自的特点,指出了各自的优劣,从两个不同的角度描述了针对漏汽的估计。更多还原

【Abstract】 With the development of industrial technology, the requirements for steam turbine operating condition monitoring requirements are increasing, Although the level of monitoring has been made substantial progress, there are still many problems to be resolved. Steam leakage occurred in the steam turbine stages not only influences the safe of steam turbine, it would also affect the steam turbine heat economy. Since the steam turbine working environment is complex, there are not good tools for steam leakage monitoring. This lets the staff unable to catch the steam leakage trend. If serious steam leakage problems happened, it can be addressed only on the overhaul. So finding the steam leakage changes in a timely manner is essential.To counter the problem, it is proposed in this paper a new measure of steam leakage with soft-sensor. The main work contains the following aspects:1. Establishing the steam turbine leakage simulation modelThe mathematical model is one effective analysis method, for complex systems. We build a simplified steam turbine model, explain each part of work mechanism and the mathematical model in detail. Then we analyze the steam leakage phenomenon in detail. The leakage of steam is characterized with steam leak factor. The coefficient also depicts a steam leakage level within the leaking area change. We study the steam leakage phenomenon for a specific simulation. The simulation provides a large number of data for the steam leakage prediction.2. Steam leakage prediction model based on the least-squaresFrom engineering perspective, there is some noise in the data acquired from sensors. The noise will affect the accuracy of model estimates. Before modeling we should discuss the noise in data. Chapter III analyzes the influence of some typical classes of noise. Then we use the least-squares to establish the steam leakage estimation model. The experiments indicate that the model can restrain the noise.3. Steam leakage estimation model based on Artificial Neural NetworkIn recent years as a view hot topic of, artificial neural networks provide another kind of view to the steam leakage estimation. Chapter IV utilizes neural networks modeling methods. To construct predicting model for leakage, we introduce a priori knowledge of the target to improve model accuracy and generalization, the test shows that the model has good generalization ability. Finally we compare the neural networks model with the least-squares model, we discuss their merits and disadvantages.更多还原