Photovoltaic Array Online Health Monitoring System Based On Extreme Learning Machine

With environmental pollution and energy shortage becoming more and more prominent,solar energy as a kind of sustainable and inexhaustible clean energy is getting more and more attention all over the world.Most photovoltaic power stations are located in remote area and tough environment,prone to various anomalies or faults,resulting in lower efficiency of power generation,safety risks,and other problems.Manual maintenance of the widely dispersed photovoltaic system in the region requires a lot of manpower and material resources,and this may be low efficiency and general maintenance effect.To solve these problems,on the basis of the existing studies on health monitoring and fault diagnosis of the photovoltaic array,this paper proposes an online health monitoring system of photovoltaic array,which based on extreme learning machine,it will improve the operation and maintenance efficiency of the photovoltaic array.The system consists of a data acquisition system,fault diagnosis,and remote monitoring server.The data acquisition system is composed of an open-source hardware,Raspberry Pi 3,and sensor acquisition circuit.The sensor acquisition circuit includes a power supply module,acquisition module,and analog-digital conversion module.During operation,the Hall voltage sensor and Hall current sensor are used in the acquisition module to obtain the voltage and current parameters of the photovoltaic array,and then output them to the digital-to-analog conversion module.The analog-digital conversion module adopts the ADC module,which is mainly realized by MCP3008 ADC chip with 8 input channels.Raspberry Pi 3 collects sensor data by controlling MCP3008,caches it locally,generates data files,and then sends the data files to remote monitoring server through File Zilla server.The fault diagnosis algorithm adopts the highly efficient mixed Machine Learning algorithm based on Bayesian classifier and Kernel-based Extreme Learning Machine,these will select each group of serial current,array voltage,reference plate current,and reference plate voltage together to form the fault feature vector.The Bayesian binary classifier is used to rapidly detect whether the photovoltaic array in faults,while the kernel limit learning machine can accurately classify and diagnose common fault,such as short circuit,open circuit,shadow,and aging.The remote monitoring server adopts B/S architecture,develops the Web based on the Python Flask framework,it stores analyzes and visualizes the parameters of photovoltaic modules collected by the data acquisition system.Independently designed a set of python-based information management platform on the server,it is able to facilitate the on-duty personnel to record fault information and timely troubleshooting.Based on the laboratory small photovoltaic grid-connected power generation platform and MATLAB/Simulink simulation model,the designed photovoltaic array health monitoring system was tested and verified.By building the same simulation model as the actual photovoltaic array,the possible faults of the photovoltaic array will be simulated and the simulation data are obtained.Through the fault simulation on the simulation and experimental platform,the fault data sets are obtained,and then the fault diagnosis model will be double-verified.Experimental results show that the fault diagnosis algorithm can diagnose common faults accurately and reliably,with an average accuracy of 98.05%.The remote monitoring system can collect the working parameters of photovoltaic modules and arrays and conduct fault diagnosis in real time,enabling the maintenance personnel to view the operating conditions of photovoltaic arrays in real time through the browser additionally,it can also make accurate operation and maintenance decisions in case of abnormal working conditions.The work of this paper has a certain reference value for the study of photovoltaic array on-line monitoring and fault diagnosis.