Modelling of the Birth-Reversal Process Based on the SAE Stack Self-Coding Neural Network

Wastewater treatment plant as a key link in the chain of water environmental protection, its safe and stable operation is a prerequisite to ensure the quality of the effluent. In the wastewater treatment process, the quality of the biological reaction process directly determines the quality of the effluent water quality, the process of many predictable indicators, which reflect important information about the wastewater treatment process, but most of them are difficult to measure or not easy to measure, the maintenance cost of high response time delay, usually not suitable for real-time monitoring. In order to handle the online real-time acquisition of these variables, soft measurement techniques provide a feasible and effective solution, and the modelling of data-driven soft measurement techniques for municipal wastewater treatment plants is reviewed and discussed. A deep neural network soft measurement model (SAE, Stacked Auto-encoder) is proposed. Shallow neural networks tend to perform poorly in the face of complex industrial conditions such as wastewater treatment, so deep learning neural networks with multi-layer structures stand out as a more popular solution. The intermediate process of wastewater treatment has a lot of data that need to be measured, so the soft measurement modelling for wastewater treatment is of great industrial importance. Facing this situation, this paper constructs a deep learning model: stack self-coding, on the basis of which it is adjusted and changed to a deep neural network soft measurement model more suitable for wastewater treatment. Finally, the validity of the proposed model is verified by a simulation platform case study.