在污水处理过程中，重要的环节之一是对污水水质进行预测，以便了解污水水质变化情况。为了获得更加准确 的水质预测结果，首先对采集到的某污水处理厂的历史污水数据进行归一化、缺失值填充等处理，选取ORP(Oxidation-Reduction Potential)、DO(Dissolved Oxygen)、pH值等七个指标作为输入变量；然后利用DBSCAN算法得到经过聚类之后的污水 数据，使相近的污水数据归为一类；最后利用 LSTM 神经网络模型对化学需氧量(Chemical Oxygen Demand, COD)、氨氮 (NH4N)、总磷(Total Phosphorus, TP)三个污水水质参数值进行预测，从而得知水质变化情况。实验结果表明，水质预测结果的 均方根误差(Root Mean Square Error, RMSE)和平均绝对误差(Mean Absolute Error, MAE)两个值均较小，预测值与实际值较为 接近，预测结果更加准确。

: In the sewage treatment process, one of the important steps is to predict the sewage quality, so that we can understand the changes of the sewage quality. In order to obtain more accurate water quality prediction results, the collected historical sewage data of a sewage treatment plant is first normalized and missing values is filled, etc., and select seven indicators such as ORP (OxidationReduction Potential), DO (Dissolved Oxygen), pH as input variables; then use the DBSCAN(Density-Based Spatial Clustering of Applications with Noise) algorithm to obtain the clustered sewage data, thus the similar sewage data are grouped into one category; finally, the LSTM neural network model is used to predict the values of the Chemical Oxygen Demand (COD), Ammonia Nitrogen (NH4N) and Total Phosphorus (TP), so as to know the changes of water quality. The experimental results show that the Root Mean Square Error (RMSE) value and Mean Absolute Error (MAE) value of the prediction results are both small, and the predicted values are close to the actual values, the prediction results are more accurate.