| 基于云计算关联分析的电力设备故障识别模型 |
| 点此下载全文 |
| 引用本文：翁东雷1,王露民1,莫建国1,邱云1,杨东东2.基于云计算关联分析的电力设备故障识别模型[J].电网与清洁能源,2023,39(10):38~44 |
| 摘要点击次数: 382 |
| 全文下载次数: 401 |
|
| 基金项目:国家电网有限公司科技项目（2018YJ26632）；宁波供电公司科技项目（KJCX015） |
|
| 中文摘要:为解决目前电力设备故障识别系统识别敏感度低的问题，提出基于云计算关联分析的电力设备故障识别模型。利用关联分析法、Model-1故障特征提取法、Copula函数的故障特征分类法，对电力设备故障特征进行提取和分类，将分类后的特征数据随机组成训练集X，并在此基础上获得故障特征优化的二维数据，将Copula函数的输出结果导入优化ID3的井漏类型分类算法中以完成对故障特征的优化，得到电力设备故障特征分类矩阵；利用非对称性卷积层的CNN模型，实现对电力设备多种故障类型的快速识别。实验结果表明：在进行故障准确性检测时，所提方法的故障识别率平均高达87.2%、识别精准率平均高达71.06%；在不同负荷对系统灵敏性影响的测试中，所提方法在任意负荷状态下的故障识别数据计数不低于40次，优于对比方法；在对电力设备匝间短路故障位点的识别性能测试中，所提方法在任意匝间短路故障位点的故障识别数据计数均高于140次，优于对比方法。所提方法的故障识别精确度高、故障位置识别敏感性高，可促进电网安全运行和发展。 |
| 中文关键词:故障识别 关联性分析 Model-1特征提取 Copula函数 非对称性卷积层 |
| |
| A Power Equipment Fault Identification Model Based on Cloud Computing Association Analysis |
|
|
| Abstract:To address low identification sensitivity of the current power equipment fault identification system， a power equipment fault identification model based on cloud computing correlation analysis is proposed. The fault features of the power equipment are extracted and classified by using the correlation analysis method， the Model-1 fault feature extraction method and the copula function fault feature classification method. The classified feature data are randomly composed of training set X， and on this basis， the two-dimensional data of fault feature optimization is obtained. The output result of Copula function is introduced into the lost circulation type classification algorithm of optimized ID3 to optimize the fault characteristics， and the fault characteristic classification matrix of the power equipment is obtained. The CNN model of the asymmetric convolution layer is used to realize the rapid identification of multiple fault types of power equipment. The experimental results show that the average fault recognition rate of the proposed method is as high as 87.2% and the average recognition accuracy is as high as 71.06%； In the test of the influence of different loads on the sensitivity of the system， the fault identification data count of the proposed method under any load state is not less than 40 times， which is better than the compared method； In the identification performance test of the turn to turn short circuit fault location of the power equipment， the fault identification data count of the proposed method at any turn to turn short circuit fault location is higher than 140 times， which is better than the compared method. The proposed method has high accuracy and high sensitivity of fault location identification， which can promote the development of power grid safe operation. |
| keywords:fault identification correlation analysis Model-1 feature extraction Copula function asymmetric convolution |
| 查看全文 查看/发表评论 下载PDF阅读器 |
