Time difference model-based soft measurement modeling method

Abstract

The invention discloses a time difference model-based soft measurement modeling method. The method can be used for complicated industrial processes with nonlinearity, time lag, time variance, multi-variable coupling and other characteristics. The method comprises the following steps of: training and learning time differences of input data and output data so as to obtain difference recombination data, and carrying out data prediction by taking the difference recombination data as an input of a soft measurement model; and constructing a time difference-based soft measurement model to predict a response variable, and evaluating the prediction ability of the model. Through carrying out modeling prediction on practical sewage treatment data, a time difference model is capable of well processingdrift of input and output variables in the process so as to obtain implicit variable time lag information and dynamic information in the data; and the time difference-based soft measurement model iscapable of improving the prediction ability and generalization ability of the model, and is more suitable for complicated and variant wastewater treatment processes.

Description

technical field [0001] The invention relates to a soft-sensing modeling method for effluent indicators in the sewage treatment process, in particular to a soft-sensing modeling method based on a time difference correlation vector machine and typical correlation analysis. technical background [0002] Different industries have different attitudes and motives for control. In recent years, with the enhancement of people's awareness of environmental protection, the sewage treatment industry has the same advanced control as other heavy industries. And due to the unique characteristics of sewage treatment itself, for example, compared with most industrial processes, the daily water quality and quantity of sewage treatment plants fluctuate greatly; wastewater must be treated and discharged up to standard, unlike other industrial production that can be "returned to the factory" for treatment; because The content of pollutants in sewage may be very small, and the sensor is difficult ...

AI Technical Summary

Problems solved by technology

However, most methods do not consider the time-lag of the modeling data when establishing the soft sensor model.

In fact, because the process variables obtained by online instruments usually imply a long measurement cycle and analysis time, and the aging of the measuring instruments leads to problems such as drift and offset of the measured variable data, which makes the soft sensor modeling we use The data are affected so that the built model cannot accurately explain the process characteristics of the variable

In addition, for the parameter model, as time goes by, the process variable data will change dynamically. If the parameters cannot be adjusted accordingly, the model will not be applied normally, but it will take a lot of time in the process of parameter adjustment. and reduce plant operating efficiency

Images