Department of Systems and Control, Jozef Stefan Institute. Jamova 39, SI-1000 Ljubljana, Slovenia.
Water Sci Technol. 2011;64(5):1115-21. doi: 10.2166/wst.2011.477.
In this paper a model predictive controller (MPC) for ammonia nitrogen is presented and evaluated in a real activated sludge process. A reduced nonlinear mathematical model based on mass balances is used to model the ammonia nitrogen in the activated sludge plant. An MPC algorithm that minimises only the control error at the end of the prediction interval is applied. The results of the ammonia MPC were compared with the results of the ammonia feedforward-PI and ammonia PI controllers from our previous study. The ammonia MPC and ammonia feedforward-PI controller give better results in terms of ammonia removal and aeration energy consumption than the ammonia PI controller because of the measurable disturbances used. On the other hand, with the ammonia MPC, comparable or even slightly poorer results than with the ammonia feedforward-PI controller are obtained. Further improvements to the MPC could be possible with an improved accuracy of the nonlinear reduced model of the ammonia nitrogen, more sophisticated control criteria used inside the controller and the extension of the problem from univariable ammonia to multivariable total nitrogen control.
本文提出并评估了一种用于氨氮的模型预测控制器（MPC）在实际活性污泥工艺中的应用。该模型基于质量平衡，使用简化的非线性数学模型来模拟活性污泥厂中的氨氮。应用了一种仅在预测间隔结束时最小化控制误差的 MPC 算法。将氨氮 MPC 的结果与我们之前研究中的氨氮前馈-PI 和氨氮 PI 控制器的结果进行了比较。由于使用了可测量的干扰，氨氮 MPC 和氨氮前馈-PI 控制器在氨氮去除和曝气能耗方面的效果优于氨氮 PI 控制器。另一方面，由于氨氮 MPC 的非线性简化模型的精度提高，以及控制器内部使用更复杂的控制标准，以及将问题从单变量氨氮扩展到多变量总氮控制，因此可能会进一步改进 MPC。
