Method for establishing DO prediction model based on optimized BP neural network

Abstract

The invention discloses a method for establishing a DO prediction model based on an optimized BP neural network. The method comprises following steps: variables are selected on the basis of predicting aeration tank DO; data is pretreated; an initial prediction model of the aeration tank DO is established; and modeling data is used for training network models to obtain an initially built DO prediction model and then the model prediction effect is analyzed; parameter and structure optimizations on the initially built DO prediction model are performed one by one and then the initial weight and threshold of BP network are optimized by means of the mind evolutionary algorithm; after the optimized DO prediction model is obtained, the prediction precision and prediction performance of the optimized model are evaluated. The method improves the prediction accuracy of the time sequence data, improves the model performance to a certain extent and improves the prediction accuracy, and facilitates the implementation of the precise control of the aeration tank aeration in the sewage treatment plant.

Description

Technical field [0001] The invention relates to the technical field of sewage treatment, in particular to a method for establishing a DO prediction model based on an optimized BP neural network. Background technique [0002] Dissolved Oxygen (DO) plays a vital role in the wastewater treatment process of A2O process. The concentration of DO determines the degree of biochemical reaction in sewage treatment: DO is not sufficient, aerobic microorganisms cannot survive and play their role in oxidative decomposition; if the concentration of DO is too high, a large amount of DO will flow back to the anoxic tank when the sludge is returned, causing organic matter The increase in oxidation rate causes the denitrification and denitrification process to fail due to the lack of carbon source. Therefore, DO must be controlled within an appropriate range to achieve the effective goal of achieving sewage purification and reducing energy consumption. [0003] There are many factors that affect t...

AI Technical Summary

Problems solved by technology

However, when the problems to be solved are relatively complex, basic artificial neural networks (ANN) generally cannot meet the needs of practical applications, and even small network behaviors are not easy to understand, let alone large-scale multi- Layer nonlinear network structure, and currently there are no strict design rules and forming theories to follow

In the forecasting, these shortcomings will inevitably lead to a relatively large deviation between the forecasted results and the actual output, which cannot meet the expected forecasting requirements

Images