针对因超临界火电机组协调系统建模精度不高，进而影响先进控制算法控制效果的问题，提出了一种分立-组合式建模方法，并将其应用于模型预测控制算法中。首先，利用子空间辨识方法分步求解分立的子系统模型，之后将分立模型进行组合得到机组的整体模型，以某600 MW超临界火电机组的多组实际数据为训练样本和检验样本，模型输出参量的拟合度可达到90%以上，从而解决了传统机理建模中假设条件较多、机组动态特性不足等问题；其次，以所建立的分立-组合模型为基础，分别应用于模型预测控制（Model Predictive Control，MPC）算法和常规PI解耦控制算法中，由于MPC拥有对多进多出(MultipleInput MultipleOutput，MIMO)系统更好的适用性，并可以在每个周期内求解带约束条件的最优解，使得控制量在超调量方面幅度下降了近20%，调节时间缩短30%以上，从而体现了所提改进算法在机组协调系统控制中的优越性。

Aiming at the problem that the modeling accuracy of the supercritical thermal power unit coordination system is not high,affecting the control effect of the advanced control algorithm,a discretecombined modeling method is proposed and applied to the model predictive control algorithm.The spatial identification method solves the discrete subsystem models step by step,and then combines the discrete models to obtain the overall model of the unit.The actual data of a 600 MW supercritical thermal power unit is used as the training sample and the test sample,and the fit of the model output parameters can reach more than 90%,which solves the problems of many assumptions in the traditional mechanism modeling and insufficient dynamic characteristics of the unit.Secondly,based on the established discretecombined model,the model predictive control (MPC) algorithm and the conventional PI decoupling control algorithm are applied.Because MPC has better applicability to the MIMO system,and can solve the optimal solution with constraints in each cycle,the control amount is reduced by nearly 20% in terms of overshoot,and the adjustment time is shortened by more than 30%.,indicating the superiority of the proposed improved algorithm in the control of the unit coordination system.

TK221

国家电网公司总部科技项目（521702150006）