| 摘要: |
| 为了提高继电保护装置遥信插件的出厂质量，解决产品检验自动化的问题，提出了一种基于PSO优化多层感知器(Multi-Layer Perceptron, MLP)神经网络的遥信插件质量识别方法。首先，建立了继电保护装置遥信插件自动化硬件测试平台。然后，改进了PSO优化算法，调整了惯性权重 的滑动特性，使其根据粒子间距实时调整步进。最后，在SPSS中使用k-s检验对原始起动电压数据进行正态性检验，得到了具有正态性样本的频率分布及其拟合曲线，提取特征训练集，并对神经网络进行训练和测试。实验结果表明，该方法能够有效且准确地对遥信插件进行质量识别，实现了产品检验和质量识别的自动化及智能化。PSO-MLP神经网络训练时间短，收敛速度快，质量识别准确度高，约为97%，且泛化能力强。 |
| 关键词: 质量识别 MLP神经网络 继电保护 遥信 智能化 |
| DOI：10.19783/j.cnki.pspc.190345 |
| 投稿时间：2019-03-28修订日期：2019-06-23 |
| 基金项目:天津市轨道交通重大专项资助(18ZXGDGX00010);天津市“一带一路”科技创新合作项目资助(18YDYGHZ00030) |
|
| Investigation on method of quality identification for telesignalization plug-in based on PSO-MLP neural network |
| CHEN Dongyang |
| (Tianjin Keyvia Electric Co., Ltd., Tianjin 300392, China) |
| Abstract: |
| In order to improve the outgoing quality of the telesignalization plug-in of the relay protection device and solve the problem of product inspection automation, this paper involves in a quality identification method of telesignalization plug-in based on PSO-MLP neural network. Firstly, the automation hardware test platform for telesignalization plug-in of relay protection device is established. Secondly, PSO algorithm is improved, and then the inertia weigh sliding characteristics is adjusted, which makes its real-time adjustment step by step according to particle spacing. Finally, the original start-up voltage data is examined by normality test using the k-s test in SPSS, then the frequency distribution of samples with normality and its fitting curve is obtained, to extract the characteristics of the training set and then to train and test the neural network. The experiment results show that the method can effectively and accurately identify the quality for telesignalization plug-in and realize the product inspection and quality identification automation and intelligent. The training time of PSO-MLP neural network is short, convergence rate is fast, and the identification accuracy is high, about 97%, and the generalization ability is strong. This work is supported by Tianjin Rail Transit Major Special Project (No. 18ZXGDGX00010) and Tianjin “Belt and Road” Initiative Science and Technology Innovation Cooperation Project (No. 18YDYGHZ00030). |
| Key words: quality identification MLP neural network relay protection telesignalization intelligence |
