利用贝叶斯优化多层感知机的光伏阵列在线故障诊断方法

Online Fault Diagnosis Method for Photovoltaic Arrays Using Bayesian Optimized Multilayer Perceptron

光伏阵列在非正常状态下运行,将严重降低光伏电站的发电效率,并缩短光伏组件的使用寿命。及时准确地识别故障类型变得尤为重要,提出了一种贝叶斯优化算法(Bayesian optimization algorithm,BOA)与多层感知机(multilayer perceptron,MLP)相结合的光伏阵列实时远程监控在线故障诊断方法。首先,利用MATLAB/Simulink建立光伏阵列的仿真模型,模拟4种典型故障情况,并设置了7种故障状态,分析不同故障下的电流-电压(I-V)、功率-电压(P-V)曲线。考虑光伏阵列运行过程中能实时测量的物理量,提取故障特征并构造了一个新特征电流占比。随后,对特征数据进行了标准化处理。接着,利用BOA对MLP的超参数进行优化,训练得到了具有较高准确性的光伏阵列故障诊断模型,测试准确率达到99.67%,诊断效果很好。最后,通过与随机森林和支持向量机等算法进行对比,并进行了仿真实验验证了该方法的准确性和稳定性。

Photovoltaic arrays running under abnormal conditions will significantly reduce the power generation efficiency of photovoltaic power stations and shorten the service life of photovoltaic components.Timely and accurate identification of fault types becomes particularly important.A real-time remote monitoring and online fault diagnosis method for photovoltaic arrays was proposed,combining Bayesian optimization algorithm(BOA)with multilayer perceptron(MLP).Firstly,a simulation model of the photovoltaic array was established using Matlab/Simulink,simulating four typical fault scenarios and setting seven fault states.I-V and P-V curves were analyzed under different fault conditions.Considering the real-time measurable physical quantities during the operation of the photovoltaic array,fault features were extracted,and a new feature“current ratio”was constructed.Subsequently,the feature data was standardized.Then,BOA was used to optimize the hyper parameters of MLP,resulting in a highly accurate fault diagnosis model for the PV array with a test accuracy of 99.67%,demonstrating excellent diagnostic performance.Finally,a comparative study with algorithms such as random forest and support vector machine was conducted,and simulation experiments verify the accuracy and stability of the proposed method.