【摘要】 针对工业废水-生活污水复合型污水厂由于突发水污染事件而导致污水厂工艺冲击负荷大、达标排放有风险等问题，该研究以湖北省某复合型污水处理厂为研究对象，利用自回归移动平均（ARIMA）算法构建进水水质预测及异常预警模型；结合自回归（ACF）、偏自回归（PACF）和增强迪基-富勒测试（ADF）分别确定化学需氧量、氨氮、总氮、总磷预测模型的最佳参数，优化水质预测模型，并探究不同监测频率和预测步长对模型的影响；在此基础上，选择自编码器（AE）和K-邻近值（KNN）算法构建水质预警模型，实现了水质的异常预警。该研究不仅为污水厂提前应对突发水污染事件提供科学依据，同时也为污水处理厂智能化转型奠定理论基础。更多还原

【Abstract】 Domestic-industrial integrated wastewater treatment plants often face significant operational challenges and risks of non-compliant discharge due to emergency pollution incidents. Focusing on a wastewater treatment plant located in a satellite city in Hubei Province, China, the autoregressive integrated moving average(ARIMA) algorithm was employed to develop a model for influent water quality prediction. The optimal parameters for predicting chemical oxygen demand, ammonia nitrogen, total nitrogen, and total phosphorus were determined using autocorrelation function(ACF), partial autocorrelation function(PACF), and augmented Dickey-Fuller(ADF) tests, leading to the optimization of the prediction model. Besides, the impact of different monitoring frequencies and prediction intervals were investigated. Additionally, autoencoder(AE) and K-nearest neighbors(KNN) algorithms were utilized to construct a water quality warning model, effectively enabling anomaly detection. This research not only provides a scientific basis to respond to emergency events but also lays a theoretical foundation for the intelligent transformation of traditional wastewater treatment plants.更多还原