Abstract
With the increasingly serious pollution of water resources, China's awareness of water resources protection is becoming stronger and stronger, and a series of effective measures have been taken. The water quality monitoring technology can detect and analyze the water quality timely, so as to obtain accurate detection results. However, due to the limitations of the current measurement methods and measuring instruments, some parameters are difficult to be measured in time, thus affecting the control effect of water quality parameters in wastewater treatment. To solve this problem, this paper tries to use the strong nonlinear identification ability of neural network to realize soft sensing of chemical oxygen demand. Firstly, the self- adaption method is used to change the learning rate of BP algorithm to control the gradient descent speed of BP neural network in learning, and then to improve the convergence characteristics of standard BP algorithm. Learning rate adjustment process is based on certain principles, so it can ensure that the learning rate can maintain a larger value, but also ensure the stability in the learning process. In order to find out the water quality parameters which have great influence on the treatment effect, 8 water quality parameters are selected as the auxiliary variables, and the chemical oxygen demand (COD) parameters are treated as the main variables after the treatment of the sewage treatment system. The results show that the BP neural network model has high prediction accuracy, fast convergence speed, good generalization ability. It can accurately predict the COD value in the sewage treatment process, and better reflect the change law of COD and influence factors.
