Process control system and method based on neural network predictive control

Abstract

The invention discloses a process control system and a process control method based on neural network predictive control, comprising a physical controlled object, a DCS (distributed control system), an industrial control computer, a switch, a GPRS (general packet radio service) wireless communication module and a handheld client device. The physical controlled object is controlled by the DCS; the switch is connected with the DCS; the industrial control computer and the GPRS wireless communication module are connected with the switch; the handheld client device carries out wireless network communication through the GPRS wireless communication module, synchronizes with the industrial control computer and realizes remote and field monitoring. According to the process control method provided by the invention, the model prediction in predictive control is realized through a BP (back propagation) neutral network, the controller optimization in the predictive control is realized through an RBF (radial basis function) neutral network, and for a process control system with constraint conditions, the constraint conditions are taken into consideration as part of rolling optimization. In application, the process control system and the process control method based on the neural network predictive control has the following advantages: the response speed is high, the tracing performance is good, the robustness is good, and the anti-interference is strong.

Description

technical field [0001] The invention relates to a process control system and method, in particular to a multi-variable constrained industrial process system and method based on neural network predictive control, belonging to the technical field of industrial process control. Background technique [0002] At present, predictive control can achieve good results for controlling production processes or objects that change relatively slowly, but the algorithm of predictive control is proposed based on linear objects. Facing a large number of nonlinear and uncertain processes in industrial processes, the algorithm There is room for further improvement. Since the mid-1980s, artificial neural networks have attracted great attention due to their unique advantages. The neural network has a parallel mechanism, self-learning and self-adaptive capabilities, and can fully approach the complex nonlinear mapping relationship and the dynamic characteristics of the learning system. It is an ...

AI Technical Summary

Problems solved by technology

[0005] It is worth mentioning that the advent of touch-screen mobile phones, tablet computers and other products has also raised thoughts for the control industry. The traditional upper computer requires monitoring personnel to observe in front of the industrial computer, which is not conducive to liberating personnel and reducing costs, and is even more unfavorable. Engineers monitor in real time. If a handheld client software can be developed, it can be synchronized with the host computer through a wireless network connection, and the monitoring interface and alarm signals can be observed anytime and anywhere, which is convenient for production control

Images