网络控制系统中学习-预测环节的改进型BP算法

An Improved BP Algorithm in Learning-prediction Part of Networked Control System

在网络控制系统中,控制指令或状态信息经常因为传输延时或故障而不能及时到达目的地。该文尝试在受控方增加一个基于BP算法的学习-预测环节,对缺失的控制指令进行精确预测,并用于系统控制,提高了系统的鲁棒性。针对普通BP算法训练速度慢的缺点,引入了趋势矫正系数,有效地提高了预测速度和精度。

In networked control systems, control signals or state data often fail to reach their destinations on time due to transmission delays or breakdowns. A learning-prediction part based on BP algorithm is introduced into the controlled side to estimate precisely the missing control signals. The estimated values are then used to perform control function to improve the robustness of the control system. To accelerate the training process of traditional BP algorithm, a trend compensation coefficient is introduced which improves both the velocity and accuracy of prediction effectively.