A Method of Building Neural Network Model Using Field Data

Abstract

The invention discloses a method for establishing a neural network model using on-site data. The neural network model is established by using a large amount of DCS data operated on the site, which overcomes many disadvantages of traditional mechanism modeling. In order to solve the problem of poor generalization ability of the neural network model, the pruning algorithm is applied to the RBF neural network, and the hidden nodes and input nodes of the network are pruned, which not only improves the generalization ability of the network, but also determines the accuracy of the model. Order. In order to avoid deleting the input node by mistake, when pruning the input node, a separate pruning strategy is adopted, that is, the process input and output in the input node are pruned separately, so that it is possible to avoid all the input of the process in the input node The case of deletion.

Description

technical field [0001] The invention relates to the field of thermal process control, in particular to a thermal process identification method. Background technique [0002] The thermal process has the characteristics of nonlinearity, time delay, uncertainty, and correlation between variables, and it is difficult to establish an accurate mathematical model, which brings great difficulty to thermal control. There is a certain gap between the performance of the model obtained by traditional mechanism modeling and that of field equipment, because the performance of most of the field equipment is nonlinear, and it is impossible to obtain an accurate model through mathematical formulas. A modern power plant is equipped with a huge DCS system, and a large amount of operating data has been accumulated over time. How to extract the rules between variables from the data containing a large amount of information has become a current research hotspot. [0003] RBF neural network has th...

AI Technical Summary

Problems solved by technology

However, there has been no fixed formula for the determination of the neural network structure. It can only be determined by experience, and there is no basis to follow. Its size is directly related to the generalization ability of the neural network.

Images