Creation of a Digital Model of a Production Water Drainage System Using Artifi cial Neural Networks

The article considers a modern approach to managing wastewater disposal processes at industrial enterprises in order to reduce environmental risks and operating costs. The novelty of the proposed method lies in the use of recurrent neural networks to predict wastewater parameters taking into account their nonlinear dynamics and time dependencies. A practical example of implementing a digital twin of a pulp and paper mill wastewater disposal system, providing multi-variant forecasting of key wastewater quality indicators, is presented. The results of pilot operation demonstrate the high effi ciency of the proposed approach. The use of a digital twin made it possible to reduce the number of cases of exceeding the standards for chemical oxygen demand by 91%, reduce reagent consumption by 12–26%, reduce costs by 8.4 million rubles and prevent emergency situations with potential damage in excess of 10 million rubles.