Research On State Detection Of Photovoltaic Arrays Based On Enhanced Kernel Limit Learning Machine

Photovoltaic array is an important part of the photovoltaic power plant to achieve s olar power generation,its working condition is of vital importance to the secure and reli able operation of the entire photovoltaic power plan.Therefore,it is of great significance to carry out research on the condition detection of PV arrays and accurately identify the fault state of PV arrays to ensure the safe operation of PV power plants and avoid the oc currence of major accidents.The current common PV array condition detection method i s mainly the detection of PV modules,but the traditional detection methods for PV mod ules will inevitably incur high costs and low detection accuracy.In this paper,we propos e a comprehensive learning PSO with linearly-decreasing inertia weight(L-CLPSO)alg orithm to optimize the kernel extreme learning machine(KELM),this method is propose d to achieve effective detection of PV array states.Firstly,based on the basic principles of photovoltaic power generation,this paper re lies on the MATLAB platform to build a simulation model of PV arrays to realize the si mulation output of the characteristic curves of PV arrays under different states and obtai n the original feature quantity data set based on the analysis of this characteristic curve;secondly,by assessing the data distribution of the original feature quantity,this paper int roduces the fault factor(K)and fill factor(FF)to improve the feature volume,enhances the variability between different faults,and obtains a new feature volume dataset;then,t his paper adopts the adaptive semi-unsupervised weighted oversampling(A-SUWO)alg orithm to reduce the adverse effects of data imbalance on the state diagnosis of PV array s,and uses a comprehensive evaluation criterion for validation analysis;finally,in this p aper,the KELM algorithm is selected through comparative analysis and the L-CLPSO al gorithm is used to optimize the kernel parameters and penalty factors of KELM to form a L-CLPSO-KELM algorithm with high detection accuracy as an identification algorith m for PV array state detection.In this paper,the improved characteristic quantity datasets of K and FF factors and the improved characteristic quantity datasets of A-SUWO algorithm are input into the K ELM diagnostic model optimized by particle swarm optimization(PSO)algorithm and c ompared with the original characteristic quantity datasets,respectively,and the comparis on results show that the K factor with FF factor and A-SUWO method can effectively im prove the detection accuracy of each fault of PV arrays;By comparing the diagnostic ac curacy of the L-CLPSO-KELM algorithm and the kernel extreme learning machine opti mized by other algorithms,the results show that the average test set accuracy of the opti mized algorithm proposed in this paper can reach 98.30%,which verifies the effectivene ss of the L-CLPSO-KELM identification algorithm in PV array condition detection.Figure [54] Table [12] Reference [102]...