基于神经网络的反馈校正方法研究

在预测控制当中,往往需要取得精度尽可能高的预测模型,但由于实际被控对象中往往存在着非线性、参数时变、模型失配、外界干扰等因素,基于被控对象的模型预测与对象的实际输出必然存在着较大的误差.本文在提高模型预测精度的同时,在神经网络模型预测的基础上,采用基于神经网络的反馈校正来弥补神经网络模型预测的不足,并验证其较常规反馈校正有更好的校正质量.

The research of feedback correction methods based on neural network

In the prediction of control,we often need to get as much as possible the forecast model of high precision,but because the actual controlled object usually to exist in nonlinear time-varying,model and parameter mismatch,interference factors,based on the controlled object model predicted and actual output of the object there must be a bigger error.This paper improve the accuracy of the prediction model,at the same time,in the neural network model to predict,and on the basis of the neural network based feedback correction to supplement the neural network model for prediction of the shortage,and to verify the more routine feedback calibration have better alignment quality.