摘要
In the actual using, for the power electron circuit fault data is time-varying, the prior knowledge mined in the data can't instruct the current fault prediction. This paper present an adaptive-time-window technology, based on the speed of data varying, adaptively adjust the length of time Then the feedforward neural network will mine the prediction knowledge on the data which are limited in the window, which contain the nearest data. The experiment demonstrates the effective of the presented algorithm.
- 出版日期2008
- 单位天津理工大学