Groove type reactor non-linear predication control method based on multi-kernel support vector machine

Abstract

The invention discloses a nonlinear predictive control method based on a multi-core support vector machine for a tank reactor, which belongs to the field of industrial automatic control. The control method mainly includes modeling and predictive control closed-loop design based on the multi-core support vector machine. The vector machine model is composed of the dynamic part based on the linear kernel support vector machine in series with the static part based on the spline kernel support vector machine; the multi-core support vector machine model is established according to the static and dynamic input and output data of the tank reactor. The future reference trajectory of the tank reactor temperature can be changed through the inverse based on the spline kernel support vector machine, and the nonlinear predictive control can be changed into a linear predictive control for the linear kernel support vector machine model, and then obtained according to the predictive control mechanism. The multi-step predictive control has an analytical solution of the optimal control law in a unified form, which acts on the reactor to make the temperature close to the set value to complete the control cycle.

Description

technical field [0001] The invention relates to the field of industrial automatic control, in particular to a non-linear predictive control method for a trough reactor based on a multi-core support vector machine. Background technique [0002] The trough reactor (CSTR) is a typical chemical process. Due to its inherent nonlinear characteristics, it is often used as a typical severe nonlinear object to test various control methods designed. For a schematic diagram of the principle of CSTR, see figure 1 . [0003] A single-stage irreversible exothermic reaction A→B (A represents the chemical species entering the reactor, B represents the product after the reaction) is carried out in the reactor, through the heat transfer fluid flowing through the cooling jacket (C represents the heat transfer fluid In, D represents the outflow of the heat transfer fluid) to control the characteristics of the entire chemical reaction. The whole process can be described by the following set o...

AI Technical Summary

Problems solved by technology

Fuzzy modeling requires a large number of fuzzy rules, relies too much on prior knowledge, has human subjective factors, and cannot accurately reflect the dynamic characteristics of nonlinear systems. When combined with predictive control, it is impossible to obtain the analytical solution of the optimal control law; Although network modeling has the ability to approximate any continuous function on compact sets in theory, there are still many problems to be solved in practical applications, such as: the local minimum problem of neural network and the determination of neural network topology. The training of the network is based on traditional statistics, requiring an infinite number of samples, but the samples obtained in actual production are often very limited, and the training goal of the neural network is to minimize the empirical risk, so over-fitting modeling often occurs As a result, due to the complex expression of the neural network model, when combined with predictive control, the approximate solution of the control law can only be obtained online through the method of numerical optimization; compared with the first two nonlinear modeling methods, based on statistical learning theory and The support vector machine modeling based on the principle of structural risk minimization has modeling advantages unmatched by other methods. Zhong et al. studied the nonlinear predictive control of trough reactors based on support vector machines. In order to facilitate the calculation of the control law, the quadratic Polynomial kernel function, the kernel function will obviously not bring satisfactory modeling effect to the severe nonlinear process, and at the same time, the calculation process of the multi-step predictive control law has a great approximation, which will have a great influence on the control accuracy influences

The modeling accuracy of support vector machines for nonlinear systems depends on the selection of kernel functions. In order to obtain better modeling results, it is generally necessary to select more complex kernel functions. When applied to predictive control, it is impossible to obtain an explicit maximum value. The optimal control law expression can only be obtained by nonlinear optimization or other approximate methods in each sampling period. Due to the online iterative nature of this type of algorithm, the online calculation time of the control law increases with the prediction time. Domain and control time domain increase and increase continuously, the amount of calculation is large, the real-time performance is poor, and it is impossible to control the fast sampling object (the sampling period of the tank reactor is usually 1 second), and it is impossible to give multi-step predictive control with the same form The analytical solution of the optimal control law for

Images