This paper taking one-hidden layer combined BP network model used in the fault diagnosis of power transformer as a exapmle and on the based of training flow process chart, points out that the number of hidden nodes, initial weight values, training error, the max training times and the order of training samples have different influence on the BP network generalization.

This theory shows that genetic algorithm,which uses minimum marginal as fitness function and where error is less than 50%,can have convergence influence on training error regarding two kinds of problems and have boundary on the general-ization error of testing integration.

This problem only causes a waste of neurons,but also results in the training error too great to meet the request of training precision and hardly completes the task of classification or clustering.

Voltage flicker measurement and its error correction based on DSP platform;

The simulation results show that the system precision and robust is greatly improved through generalized predictive control and output error compensation instead of error correction.

A dynamic error correction ele ment is supplemented.

This paper proposes an error calibration strategy,using the two-step iterative parameter estimation algorithm to solve this problem,which is combining the non-linear parameters into linear parameter,estimating the states using Kalman filter and then getting the original parameter using Gauss-.

Combining with correlative mathematics principle,Least Square Regression(LSR) is imported to set up error calibration scheme.