PID index solving method of variation differential evolution-radial basis function neural network

Through the mutant differential evolution-radial basis neural network method, the PID parameters are optimized, and the problems of gain change sensitivity and overshoot in closed-loop dynamic adjustment of traditional PID controllers are solved, achieving rapid and accurate adjustment effects.

CN119168024AActive Publication Date: 2024-12-20NANCHANG HANGKONG UNIVERSITY
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202411233688.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-04
Publication Date
2024-12-20
Estimated Expiration
2044-09-04

AI Technical Summary

Technical Problem

Traditional PID controllers are sensitive to gain changes in closed-loop dynamic adjustment, resulting in large initial control force and frequent overshoots, making it difficult to achieve rapid adjustment.

Method used

Using the variant differential evolution-radial basis neural network (MDE-RBF) method, by introducing random variation and cross-operation in the differential evolution algorithm, combining radial basis neural network and conjugate gradient method, use adaptive control rules and exponential smoothing Prediction algorithm optimizes PID parameters to improve the system's adjustment accuracy and efficiency.

Benefits of technology

The online adjustment of PID parameters is realized, the local optimal trap is avoided, the system's global search ability and adjustment speed is improved, the sensitivity problem of traditional PID controllers in dynamic adjustment is solved, and the rapid and accurate adjustment effect is achieved. .

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119168024A_ABST
    Figure CN119168024A_ABST
Patent Text Reader

Abstract

The invention discloses a PID (Proportion Integration Differentiation) index solving method of a variation differential evolution-radial basis function neural network, which comprises the following steps of: introducing two times of value extraction and variation operators in a variation stage of a variation differential evolution (MDE) algorithm to obtain dominant individuals to participate in next-generation competition, so that local optimum is avoided, and the global search capability is improved; in a radial basis function (RBF) network, an exponential smoothing factor is combined with a conjugate gradient method, and adaptive step length is adopted to control data fluctuation in an iteration process. A dominant population is obtained through a variation differential evolution (MDE) algorithm, system loop information is identified through an optimized radial basis function (RBF) network, online adjustment of PID parameters is achieved, analysis is conducted through three industrial control models by comparing standard PID and RBF-PID control at the same time, and effect verification is conducted through the method disclosed by the invention.
Need to check novelty before this filing date? Find Prior Art

Citation Information

Patent Citations

  • PMA reliability optimization method based on hybrid reliability model and neural network response surface

    CN112307654A

  • Optimal selection method for control parameters of servo feeding system of numerical control machine tool

    CN113268040A

  • Adaptive superconductive magnetic energy storage (SMES) control method and system

    US8933572B1