Artificial intelligence-enabled predictive planning for sewage treatment based on improved DNN and LSTM

• A framework for sewage treatment is proposed to reduce cost and increase efficiency. • Correlation analysis based on a combined calculation is used for data processing. • A combination of improved DNN and LSTM is created to predict critical data. • Results reveal that the accuracy of the data in predictive planning is 92.7 %. Predictive planning is essential for sewage treatment, which ensures water security. In the context of Industry 4.0, new sensor technologies are generating large amounts of heterogeneous data from multiple sources in increasingly complex sewage treatment processes. This complexity renders traditional methods inadequate for the accurate and timely prognostication of data essential for predictive planning. To solve the challenge, this study proposes an architecture of artificial intelligence-enabled predictive planning to reduce cost and increase efficiency for sewage treatment. Within this architecture, a combination of a sparsely connected deep neural network model based on combined correlation analysis and an improved long short-term memory model based on periodicity is used to predict critical data for sewage treatment. Then, the proposed architecture is applied by using production data from the high-density pool unit of a sewage treatment plant. Results reveal that the accuracy of the data in predictive planning is 92.7 % compared with the actual data. Establishing this architecture for predictive planning provides a practical basis for the digital transformation of sewage treatment plants to automate processes, improve decision-making, reduce costs and increase operational efficiency.