基于RVM联合SVR的低压开关电寿命预测方法  

A Low-voltage Switch Electrical Lifetime Prediction Method Based on RVM Joint SVR

作　　者：王伟光 许杰 王坤 WANG Weiguang;XU Jie;WANG Kun(Guangzhou Liwan Power Supply Bureau of Guangdong Power Grid Co.,Ltd.,Guangzhou 510000,China)

机构地区：[1]广东电网有限责任公司广州荔湾供电局,广东广州510000

出　　处：《微型电脑应用》2024年第11期112-115,123,共5页Microcomputer Applications

基　　金：2022年广东电网有限责任公司依托基建工程技术创新专题项目(030118DP22120065)。

摘　　要：继电器等低压开关的电寿命是其可靠性的重要指标,对电寿命进行准确预测对于整个电网系统的安全稳定运行至关重要。传统基于反向传播(BP)神经网络的电寿命预测方法精度低且泛化能力弱,限制了其在实际生活中的推广应用。针对该问题,提出一种基于相关向量机(RVM)联合支持向量回归(SVR)的组合模型实现继电器电寿命的高精度预测。分析接触电阻、线圈电感、吸合时间等10维特征参数与继电器电寿命之间的关系,建立RVM模型对10维特征参数进行特征选择,自动获得与电寿命相关性最高的3维特征参数构成最优特征集合,并将其作为SVR模型的输入从而建立电寿命预测模型,实现对继电器电寿命的高精度预测。针对SVR模型参数选择难题,提出改进的水循环优化算法(IWCA)对其全局寻优,提升预测性能。试验结果表明,相对于BP神经网络预测模型和单一SVR预测模型,所提组合模型预测精度分别提升11.5%和6.3%,实时性分别提升0.64 s和0.32 s,并且在小样本条件下表现出了更强的泛化能力,具有较好的应用前景。The electrical lifetime of low-voltage switches such as relays is an important indicator of their reliability,and accurate prediction of the electrical lifetime is crucial for the safe and stable operation of the entire power grid system.The traditional neural network-based method for predicting electrical lifetime has low accuracy and weak generalization ability,which limits its practical application.To address this issue,a combined model based on relevance vector machine(RVM)and support vector regression(SVR)is proposed to achieve high-precision prediction of relay electrical lifetime.The relationship between the 10 dimensional characteristic parameters such as contact resistance,coil inductance,and suction time and the electrical lifetime of the relay is analyzed.A RVM model is established to select the features of the 10 dimensional characteristic parameters,and automatically obtain the optimal feature set of the 3 dimensional characteristic parameters with the highest correlation with the electrical lifetime.These parameters are used as inputs to the SVR model to establish an electrical lifetime prediction model,achieving high-precision prediction of the electrical lifetime of the relay.Aimed at the problem of parameter selection in SVR model,an improved water cycle algorithm(IWCA)is proposed to globally optimize predictive performance.The experimental results show that compared to traditional BP neural network methods,the proposed combination model has higher prediction accuracy,better real-time performance,stronger generalization ability under small sample conditions and better application prospects.

关 键 词：低压开关 继电器 电寿命预测 反向传播神经网络 相关向量机 支持向量回归 

分 类 号：TM58[电气工程—电器]