控制器嵌入Transformer模型的水厂加氯与余氯控制方法OA

[Objective]To ensure the stability of residual chlorine concentration in municipal water treatment plant(WTP)and to reduce the impact of water quality fluctuations on disinfection efficiency and chemical economics.This paper proposes an optimization method for drug addition based on intelligent control.[Methods]In response to the problems of traditional manual drug addition,such as response lag,the difficulty of directly constraining control targets with prediction-oriented models,and the complexity of model predictive control modeling,this paper designed a Transformer model with controller embedding.The model adopted an encoder-controller-decoder framework,where the controller directly generated future dosing sequences and the decoder evaluated the corresponding chlorine response.A composite loss function based on the target interval of residual chlorine was designed to enable end-to-end training.[Results]Validation was carried out using real operational data from a WTP.The proposed method significantly suppressed fluctuations in chlorine residual mass concentration,with a prediction standard deviation of 0.004 6 mg/L,and reduced chemical consumption by approximately 7.25％while maintaining water quality compliance,annual operating costs decreased by approximately 116 000 yuan.[Conclusion]The controller-embedded Transformer model improves the accuracy and stability of chlorine residual control and demonstrates good economic efficiency and robustness,providing a references and technical supports for intelligent disinfection control in municipal WTPs.