摘要
针对传统灰色模型对变压器油中气体进行分析时预测时间跨度较长、序列精度会大幅降低的问题,提出了灰色模型残差修正融合算法。在建立灰色多变量模型的基础上确定关联度大于0.5的气体变量,然后融合了自适应回归算法和马尔可夫修正模型对残差进行修正,避免了残差的持续累积,提高了预测精度。通过仿真,将传统灰色模型、仅通过自适应回归修正以及进一步马尔可夫修正后的误差进行对比,结果表明,融合了两种模型的算法准确度最高,测试集误差减小到最初的34%。该灰色模型残差修正融合算法可有效提高传统灰色预测模型的准确性。
Aiming at the problem that the accuracy of the sequence with long time span is greatly reduced when the traditional grey model analyzes gas in the transformer oil, a residual-modified fusion algorithm based on grey model is proposed. On the basis of a multi-variable grey model, the gases with a correlation degree larger than 0.5 were selected. Then, the adaptive regression algorithm and Markov modified model were used to correct the residual, which avoided the continuous accumulation of residual errors and improves the prediction accuracy. The comparison of the traditional grey model, or with the error only through adaptive regression and further Markov correction through simulation shows that the accuracy of the fusion using the two models is the highest, which is reduced by 34% compared to the original error. The facts show that the residual-modified fusion algorithm based on grey model can effectively improve the accuracy of the traditional grey prediction model.
出处
《电子测量与仪器学报》
CSCD
北大核心
2018年第10期87-94,共8页
Journal of Electronic Measurement and Instrumentation
基金
国家自然科学基金(51577046)
国家自然科学基金重点项目(51637004)
国家重点研发计划“重大科学仪器设备开发”项目(2016YFF0102200)
装备预先研究重点项目(41402040301)资助