IFA-PNN在电力变压器故障诊断中的应用
|
摘要
目前电力变压器故障分类模型中多是采用油中溶解气体分析(DGA)结合人工智能的方法,其中以神经网络为代表的智能方法应用最为广泛。提出一种基于改进的萤火虫算法(IFA)优化概率神经网络(PNN)的故障诊断方法。PNN结构简单,收敛速度快,但其中平滑因子对网络输出结果正确性影响较大,采用了IFA对平滑因子进行寻优。在标准萤火虫算法的基础上,将固定步长改为自适应步长,调整位置更新公式,不仅平衡局部搜索与全局搜索能力,也提高搜索的速度。此外,PNN对样本的代表性要求高,采用融合DGA算法实现输入数据的处理,有效降低了样本对分类结果的影响。通过实验数据验证,IFA-PNN算法可加快搜索的速度,提高诊断的精度,减小误差,分类效果明显,是一种有效的故障诊断方法。
At present, the dissolved gas analysis(DGA) combined with artificial intelligence is mainly used in the fault classification model of power transformers. The intelligent technology represented by neural network is the most widely used. A new fault diagnosis method for improving the probability neural network(PNN) is proposed. The structure of PNN is simple and the speed of convergence is fast, but the smoothing factor has a great influence on the correctness of the network output. Therefore, the improved firefly algorithm(IFA) is used to optimize the smoothing factor. Based on the standard firefly algorithm, the fixed step size is changed into the self-adaptive step length and the position update formula is adjusted, which not only balances the global search and local search capabilities, but also improves the search speed. In addition, PNN has high requirements on representative for samples, and the fusion DGA algorithm is adopted to realize the processing of input data,which effectively reduces the impact of samples on classification results. The IFA-PNN algorithm speeds up the search, improves the accuracy of the diagnosis, reduces the error, and has obvious classification effect. It is an effective fault diagnosis method.
At present, the dissolved gas analysis(DGA) combined with artificial intelligence is mainly used in the fault classification model of power transformers. The intelligent technology represented by neural network is the most widely used. A new fault diagnosis method for improving the probability neural network(PNN) is proposed. The structure of PNN is simple and the speed of convergence is fast, but the smoothing factor has a great influence on the correctness of the network output. Therefore, the improved firefly algorithm(IFA) is used to optimize the smoothing factor. Based on the standard firefly algorithm, the fixed step size is changed into the self-adaptive step length and the position update formula is adjusted, which not only balances the global search and local search capabilities, but also improves the search speed. In addition, PNN has high requirements on representative for samples, and the fusion DGA algorithm is adopted to realize the processing of input data,which effectively reduces the impact of samples on classification results. The IFA-PNN algorithm speeds up the search, improves the accuracy of the diagnosis, reduces the error, and has obvious classification effect. It is an effective fault diagnosis method.
引文
[1]何定,唐国庆,陈珩.电力变压器故障诊断专家系统TFDES[J].电力系统自动化,1993(7):32-35.
[2]王大忠,徐文,周泽存,等.模糊理论、专家系统及人工神经网络在电力变压器故障诊断中应用——基于油中溶解气体进行分析诊断[J].中国电机工程学报,1996(5):349-353.
[3]毕天姝,倪以信,吴复立,等.基于新型神经网络的电网故障诊断方法[J].中国电机工程学报,2002(2):74-79.
[4]李天云,程思勇,童建东,等.基于Elman神经网络的油浸式电力变压器故障诊断[J].中国电力,2006,39(11):55-57.
[5]张静.基于SOM的变压器故障诊断研究[J].电力学报,2014,(04):318-321.
[6]石鑫.基于深度学习的变压器故障诊断技术研究[D].北京:华北电力大学,2016.
[7]张国荣.基于SVM分类算法的电力变压器故障诊断[C]//第六届全国信息获取与处理学术会议论文集.2008:4.
[8]王永强,律方成,李和明.基于粗糙集理论和贝叶斯网络的电力变压器故障诊断方法[J].中国电机工程学报,2006,26(8):137-141.
[9]俞文燕.基于概率神经网络的电机故障诊断研究[D].扬州:扬州大学,2016.
[10]万怡骎.基于概率神经网络的变压器故障诊断[D].南昌:南昌大学,2007.
[11]范俊辉,彭道刚,黄义超,等.基于改进PSO优化PNN网络的变压器故障诊断方法[J].测控技术,2016,35(3):42-49.
[12]郑凯,袁丹,刘剑涛,等.基于SA-PSO优化自适应PNN网络的变压器故障诊断研究[J].计算机测量与控制,2014,22(4):1015-1017.
[13] Yang X S. Nature-Inspired Metaheuristic Algorithms[M].Luniver Press,2010(2):81-96.
[14] Specht D F. Probabilistic neural networks[J]. Neural Networks, 1990,3(1):109-118.
[15]李煜,马良.新型全局优化蝙蝠算法[J].计算机科学,2013,40(9):225-229.
[16]葛考,时贝贝,徐康耀,等.基于BAPSO-PNN神经网络算法的空气质量评价研究[J].环境工程,2015(S1):676-681.
[17]潘晓英,陈雪静,李昂儒,等.基于自适应步长的萤火虫划分聚类算法[J].计算机应用研究,2017(12):1-7.
[18]张明,张树群,雷兆宜.改进的萤火虫算法在神经网络中的应用[J].计算机工程与应用,2017,53(5):159-163.
[19]史峰,王辉,郁磊,等.MATLAB智能算法30个案例分析[M].北京:北京航空航天大学出版社,2011.
[20]赵玉新,Yang X S,刘利强.新兴元启发式优化方法[M].北京:科学出版社,2013.
[21]张珂斐,郭江,聂德鑫,等.基于化学反应优化神经网络与融合DGA算法的油浸式变压器模型研究[J].高电压技术,2016,42(4):1275-1281.
[22] Mineral Oil-Impregnated Electrical Equipment in Service Guide to the Interpretation of Dissolved and Free Gases Analysis:IEC-60599[S].
[2]王大忠,徐文,周泽存,等.模糊理论、专家系统及人工神经网络在电力变压器故障诊断中应用——基于油中溶解气体进行分析诊断[J].中国电机工程学报,1996(5):349-353.
[3]毕天姝,倪以信,吴复立,等.基于新型神经网络的电网故障诊断方法[J].中国电机工程学报,2002(2):74-79.
[4]李天云,程思勇,童建东,等.基于Elman神经网络的油浸式电力变压器故障诊断[J].中国电力,2006,39(11):55-57.
[5]张静.基于SOM的变压器故障诊断研究[J].电力学报,2014,(04):318-321.
[6]石鑫.基于深度学习的变压器故障诊断技术研究[D].北京:华北电力大学,2016.
[7]张国荣.基于SVM分类算法的电力变压器故障诊断[C]//第六届全国信息获取与处理学术会议论文集.2008:4.
[8]王永强,律方成,李和明.基于粗糙集理论和贝叶斯网络的电力变压器故障诊断方法[J].中国电机工程学报,2006,26(8):137-141.
[9]俞文燕.基于概率神经网络的电机故障诊断研究[D].扬州:扬州大学,2016.
[10]万怡骎.基于概率神经网络的变压器故障诊断[D].南昌:南昌大学,2007.
[11]范俊辉,彭道刚,黄义超,等.基于改进PSO优化PNN网络的变压器故障诊断方法[J].测控技术,2016,35(3):42-49.
[12]郑凯,袁丹,刘剑涛,等.基于SA-PSO优化自适应PNN网络的变压器故障诊断研究[J].计算机测量与控制,2014,22(4):1015-1017.
[13] Yang X S. Nature-Inspired Metaheuristic Algorithms[M].Luniver Press,2010(2):81-96.
[14] Specht D F. Probabilistic neural networks[J]. Neural Networks, 1990,3(1):109-118.
[15]李煜,马良.新型全局优化蝙蝠算法[J].计算机科学,2013,40(9):225-229.
[16]葛考,时贝贝,徐康耀,等.基于BAPSO-PNN神经网络算法的空气质量评价研究[J].环境工程,2015(S1):676-681.
[17]潘晓英,陈雪静,李昂儒,等.基于自适应步长的萤火虫划分聚类算法[J].计算机应用研究,2017(12):1-7.
[18]张明,张树群,雷兆宜.改进的萤火虫算法在神经网络中的应用[J].计算机工程与应用,2017,53(5):159-163.
[19]史峰,王辉,郁磊,等.MATLAB智能算法30个案例分析[M].北京:北京航空航天大学出版社,2011.
[20]赵玉新,Yang X S,刘利强.新兴元启发式优化方法[M].北京:科学出版社,2013.
[21]张珂斐,郭江,聂德鑫,等.基于化学反应优化神经网络与融合DGA算法的油浸式变压器模型研究[J].高电压技术,2016,42(4):1275-1281.
[22] Mineral Oil-Impregnated Electrical Equipment in Service Guide to the Interpretation of Dissolved and Free Gases Analysis:IEC-60599[S].