Joint neural network model for water quality parameter prediction and training method thereof

Abstract

The invention discloses an RNN-based joint neural network model for water quality parameter prediction, and the model comprises a plurality of RNN neural networks which are respectively deployed at all sewage measurement points; the middle layer input of the RNN neural networks at the two ends is only connected with the middle layer output of one adjacent RNN; the middle layer input of any RNN neural network between the two ends is connected with the middle layer outputs of the two adjacent RNNs at the same time; a Gaussian function module and a multiplication module are arranged on one sidesof the RNN neural networks at the two ends; a Gaussian function module and a multiplication module are arranged the two sides of any RNN neural network between the two ends; the input of each Gaussianfunction module is connected with the input of the corresponding RNN, and the output of each Gaussian function module is connected with one input of the corresponding multiplication module; and the other input of each multiplication module is connected with the output of the middle layer of the corresponding RNN, and the output is connected with the input of the middle layer of the adjacent RNN.The joint neural network can fully extract historical information of a plurality of measurement points at the same time for parameter estimation, and the prediction accuracy of whole-process parameterestimation is improved.

Description

technical field [0001] The invention relates to the field of water quality parameter prediction, in particular to a joint neural network model and its training method for water quality parameter prediction. Background technique [0002] After years of development, domestic sewage treatment technology has made great progress. However, in the actual process control process, the sewage treatment industry still faces many challenges. For example, there are many parameters to be measured in the sewage treatment process, some of which can be measured in real time by sensors, and some of them need to be tested manually. It may even take several days to complete the measurement for some parameters that need to be tested, which greatly reduces the timeliness of the data. Therefore, in order to better collect water quality parameters, soft-sensing technology can be used to estimate some water quality parameters that require manual testing through many parameters that can be measured ...

AI Technical Summary

Problems solved by technology

[0005] Based on the problems existing in the prior art, the purpose of the present invention is to provide a joint neural network model and its training method for water quality parameter prediction, which can solve the problem of setting a RNN cycle neural network at each of the existing multiple measurement points for water quality parameter prediction. In prediction, the existing single RNN cyclic neural network can generally only extract the historical information of a single measurement point for parameter estimation, but it is difficult to fully extract the historical information of multiple measurement points for parameter estimation at the same time, and the accuracy of the predicted water quality parameters is not high. The problem

Images