We used real plant data obtained from a water treatment plant in Peninsular Malaysia to develop a model to predict fluoride concentration in treated water. We validated the model with real plant data and demonstrated that the model yields good prediction, with mean absolute percentage error close to 10% or less. The validated model was then used to design an internal model control (IMC) tuned feedback controller, combined with a time-varying feedforward control approach by exploiting the predictable and repetitive pattern of the inlet water flow rate. Simulation studies using random and real disturbance data show that the proposed control has successfully prevented over fluoridation in treated water. The validated model, due to its simplicity, is suitable for future online implementation. It can also be readily applied to model-based control approaches (e.g., model predictive control) in future studies.

中文翻译：

---

饮用水处理厂的建模和控制：使用时变前馈控制的案例研究

我们使用从马来西亚半岛一家水处理厂获得的真实工厂数据来开发一个模型来预测处理过的水中的氟化物浓度。我们使用真实工厂数据验证了该模型，并证明该模型产生了良好的预测，平均绝对百分比误差接近 10% 或更低。然后使用经过验证的模型来设计内部模型控制 (IMC) 调谐反馈控制器，并通过利用进水流速的可预测和重复模式结合时变前馈控制方法。使用随机和真实干扰数据的模拟研究表明，所提出的控制已成功地防止了处理水中的过度氟化。经验证的模型由于其简单性，适用于未来的在线实施。