摘要
针对传统BP神经网络在对继电器进行寿命预测时,由于初始权值、阈值具有随机性,常会遇见收敛速度过慢、陷入局部最优、稳定性差或者过学习、欠学习等问题,提出了通过果蝇优化算法与BP神经网络结合的预测方法来提高预测模型质量。利用果蝇算法寻求最优的初始权值、阈值,降低了模型误差。通过预测模型对电磁继电器贮存过程中接触电阻值的预测,间接预测了电磁继电器贮存寿命,并验证了模型的可行性。
When using traditional BP neural network to predict relay life,there often occur the problems of too slow convergence speed,local optimum,poor stability or over learning and less learning,because of the randomness of the initial weights and thresholds.To solve these problems,this paper proposed a prediction method combining drosophila optimization algorithm and BP neural network to improve the quality of prediction models.Using drosophila optimization algorithm to search for the optimal initial weights and thresholds to reduce model errors.By using the prediction model,the storage life of electromagnetic relay is predicted indirectly through the prediction of the contact resistance value in the storage process of electromagnetic relay,and the feasibility of the model was verified.
作者
王佳炜
王召斌
黄周霖
WANG Jiawei;WANG Zhaobin;HUANG Zhoulin(School of Electronics and Information,Jiangsu University of Science and Technology,Zhenjiang 212003,China)
出处
《电器与能效管理技术》
2019年第2期19-24,共6页
Electrical & Energy Management Technology
基金
国家自然科学基金资助项目(51507074)
江苏省高校自然科学研究面上资助项目(17KJB510014)
关键词
电磁继电器
贮存寿命预测
BP神经网络
果蝇优化算法
electromagnetic relay
prediction of storage life
BP neural network
drosophila optimization algorithm
