【摘要】 由于锅炉过热系统具有强非线性、时变性的特点,采用常规方法建立其数学模型十分困难,因此提出了一种基于支持向量机和过程机理的过热系统建模方法。该方法利用机理模型产生的相关训练数据,对支持向量机网络进行训练,使之能够很好地逼近过热系统这一非线性模型,同时利用不相关的数据样本对其泛化性能进行验证。从仿真结果可以看出,采用内点法优化后的支持向量机网络,经3.18 s后收敛,学习的最大误差不超过0.035℃。因此,该方法可以有效地对系统进行建模,仿真精度高,并且不仅仅是对过热系统,同时也适用于整个锅炉系统的建模。更多还原

【Abstract】 Due to such features as a strong non-linearity and time-variation etc.specific to a boiler superheating system,it is very difficult to establish a mathematical model for the latter by using a conventional method.Hence,the authors have proposed a method for modeling a boiler superheating system based on a supportive vector machine and process mechanism.By making use of the relevant training data produced by a mechanism model,one can use the proposed method to train a network of the supportive vector machine,enabling the network to very well approximate to the non-linear model of the superheating system and in the meantime to utilize irrelevant data samples to verify its generalized performance.It can be seen from the simulation results that the network of the supportive vector machine will converge after 3.18 seconds when it has been optimized by using an inner point method.The maximal error of the learning process will not exceed 0.035 ℃.Consequently,the method can be used to effectively build a model for a system with a high simulation accuracy and is suited for modeling not only a superheating system but a whole boiler system.更多还原