Research On Photovoltaic Array Status Monitoring System For Photovoltaic Power Plants

With the development and promotion of photovoltaic power generation technology,various smal-scale distributed photovoltaic power plants are popularized in homes and factories.Due to quality and technology reasons,some unpredictable faults will occur during the use of the PV array.As a result,the output power of the entire array will decrease,the system efficiency will decrease,and the fire may be induced,resulting in serious economic losses.Existing PV array state detection methods have deficiencies in comprehensiveness and accuracy.Therefore,for the identification of array anomaly state,a PV array state detection method based on clustering technology is proposed.Aiming at the problem of historical fault sample data imbalance,a method based on improved SMOTE for PV array random forest fault diagnosis is proposed.The main research contents are as follows:1.Aiming at the situation that the existing PV array state detection method is greatly affected by environmental factors,combining the output characteristics of 3*3 PV arrays in four types of faults and normal conditions and the existing PV array condition monitoring technology,a clustering technology based PV array state detection method is proposed.The method clusters the photovoltaic array operation features at the same time point to form a plurality of cluster centers(normal operation states).The anomaly score is then calculated by the degree of deviation of each component feature from the normal state,and the anomaly state is identified by the anomaly detector.Then,the degree of deviation between each component feature and the normal state is calculated,and the abnormal state is identified by the abnormality detector.Among them,for the random selection problem of the initial cluster center of kmeans clustering,the outliers are excluded according to the closeness between the data,which limits the selection of the initial clustering center.2.Due to the imbalance of the PV module historical fault samples,there will be overfitting problems when using the SMOTE algorithm.This paper identifies sparse samples based on their density and uses them as seed samples.In the sampling process,the idea of the SMOTE algorithm is still used to synthesize a new sample between the seed sample and its neighbors.The fault diagnosis model is then implemented by a random forest formation fault diagnosis model.3.The requirements of the PV array condition monitoring system are analyzed,and its main functions and operation logic are briefly explained.According to the requirements of the PV array status monitoring system proposed in this paper,a PV array status monitoring platform is built to realize real-time monitoring,fault diagnosis and historical query.4.Some data of Fushan Photovoltaic Power Station were tested by the method proposed in this paper.The abnormal detection rate is slightly improved compared with LS-SVM algorithm and LM-BPNN algorithm.The fault diagnosis has also improved in terms of comprehensiveness,and the diagnostic accuracy rate for the four types of faults has reached more than 90%.