New methods for utilization of predictive information in neural network PID controller

  • Shukun Jia*
  • , Liang Wang
  • *Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In this paper, we analyze the utilization of predictive information in neural network PID controller (NN-PID). Based on the accuracy of predictive model, two novel methods are proposed to improve control performance. When predictive model is high-accuracy, two-step ahead predictive information is incorporated into loss function to adjust the weight of NN. When the predictive model is low-accuracy, only one-step ahead information is used and learning rate is adjusted based on the prediction error. Consequential simulation are conducted with each method.

Original languageEnglish
Title of host publicationProceedings of 2017 Chinese Intelligent Systems Conference
EditorsJunping Du, Weicun Zhang, Yingmin Jia
PublisherSpringer Verlag
Pages739-748
Number of pages10
ISBN (Print)9789811064951
DOIs
StatePublished - 2018
EventChinese Intelligent Systems Conference, CISC 2017 - Mudanjiang, China
Duration: 14 Oct 201715 Oct 2017

Publication series

NameLecture Notes in Electrical Engineering
Volume459
ISSN (Print)1876-1100
ISSN (Electronic)1876-1119

Conference

ConferenceChinese Intelligent Systems Conference, CISC 2017
Country/TerritoryChina
CityMudanjiang
Period14/10/1715/10/17

Keywords

  • Loss function
  • Neural network PID control
  • Predictive information
  • Variable learning rate

Fingerprint

Dive into the research topics of 'New methods for utilization of predictive information in neural network PID controller'. Together they form a unique fingerprint.

Cite this