Abstract Drinking water treatment plants (DWTPs) face changes in raw water quality, and treatment needs to be adjusted to produce the best water quality at the minimum environmental cost. An environmental decision support system (EDSS) was developed for aiding DWTP operators in choosing the adequate permanganate dosing rate in the pre-oxidation step. To this end, multiple linear regression (MLR) and multi-layer perceptron (MLP) models are compared for choosing the best predictive model. Besides, a case-based reasoning (CBR) model was approached to provide the user with a distribution of solutions given similar operating conditions in the past. The predictive model consisted of an MLP and has been validated against historical data with sufficient good accuracy for the utility needs (R2 = 0.76 and RSE = 0.13 mg·L−1). The integration of the predictive and the CBR models in an EDSS gives the user an augmented decision-making capacity of the process and has great potential for both assisting experienced users and for training new personnel in deciding the operational set-point of the process.

中文翻译：

---

在大型饮用水处理厂中实施用于控制预氧化步骤的环境决策支持系统

摘要饮用水处理厂 (DWTP) 面临着原水水质的变化，需要调整处理方式，以最小的环境成本产生最佳的水质。开发了环境决策支持系统 (EDSS)，以帮助 DWTP 操作员在预氧化步骤中选择适当的高锰酸盐剂量。为此，比较多元线性回归（MLR）和多层感知器（MLP）模型以选择最佳预测模型。此外，还采用基于案例的推理（CBR）模型，为用户提供过去类似操作条件下的解决方案分布。该预测模型由 MLP 组成，并已根据历史数据进行了验证，具有足以满足公用事业需求的良好准确性（R2 = 0.76 和 RSE = 0.13 mg·L−1）。EDSS 中预测模型和 CBR 模型的集成为用户提供了增强的过程决策能力，并且在帮助有经验的用户和培训新人员决定过程的操作设定点方面具有巨大潜力。