基于输入-输出修正的水轮机特性曲线精细化模型

目前常用的水轮机特性曲线模型包括线性模型和非线性模型,它们都是根据描述水轮机内外特性的方程或试验数据得到的。然而在建模的过程中,各种误差难以避免,如模型误差、测量误差和读数误差等,可能造成预测值与实际值不相符,影响模型的精度。同时,有关水轮机特性曲线模型修正的文献较少,缺乏行之有效的方法或准则,不利于相关研究成果的实际应用。本文以水轮机神经网络模型为例,提出了基于输入-输出修正的水轮机特性曲线精细化建模方法,并验证了该方法的有效性。首先根据模型综合特性曲线和边界条件,分别建立混流式和轴流式两种机组的非线性特性曲线模型。其次利用粒子群优化算法和二次多项式逼近原理,依次对模型输入和输出参数进行修正,得到最优修正系数。最后整合原始模型和修正部分,得到水轮机特性曲线的精细化模型。试验结果表明该方法能够有效提高模型的仿真精度,对研究真实机组的非线性特性具有重要意义。

Refined hydro-turbine characteristic curve model based on input-output correction

At present, the common characteristic curve models of hydro-turbines include linear models and nonlinear models, which are derived from equations or experimental data describing the internal or external characteristics of a hydro-turbine. However, in the process of modeling, various errors are difficult to avoid, such as model error, measurement error and reading error, which may cause the predicted value to be inconsistent with the actual value,affecting the accuracy of the model. At the same time,there are few literatures on the model correction of hydro-turbines,and lack of effective methods or guidelines,which is not conducive to the practical application of relevant studies. Taking the neural network model of hydro-tur-bines as an example, this paper proposes a refined model based on input and output correction, and veri-fies the effectiveness of the proposed method. Firstly, the nonlinear simulation models of the two types of Francis and Kaplan turbines are established respectively based on the comprehensive characteristic curve and boundary conditions of the model. Secondly, the model input and output parameters are corrected in turn by using the particle swarm optimization algorithm and the quadratic polynomial approximation princi-ple, so as to obtain the optimal correction coefficients. Finally, the original model and the correction part are integrated to obtain a refined model of the turbine. The experimental results show that the proposed method can effectively improve the simulation accuracy of the model and is of great significance for explor-ing the nonlinear characteristics of real units.