基于神经网络的电参数反演载荷算法

提出一种基于神经网络的抽油机载荷反演算法,从电机输入电功率和载荷输出非线性逼近的角度建立高维输入输出系统模型,通过对网络误差的负梯度反馈,自动调整权值参数,求解电参数到载荷的映射网络. 通过实测数据软件仿真,算法能实现电参数到载荷之间的非线性拟合,误差小于10-3,相关系数达到0.95以上. 

Algorithm for Electrical Parameters and Payload Based on Neural Network

A BP neural network based algorithm for payload calculation was proposed. A high dimensional input-output system was formed to describe the non-linear relation between electrical parameters and payload. By the use of negative gradient of the network error feedback, network parameters were automatically adjusted, therefore the mapping connection between electrical parameters and payload can be solved. Experimental results and simulations show that the algorithm can solve the problem of the non-linear approximation, and achieve an error probability of e-3 with correlation coefficient up to 0.95.