Dynamic Characteristics Models Of The Cloud Cluster For The Ultra-short-term Distributed Photovoltaic Power Prediction

With the deepening of economic globalization, energy and enviromental issues have become increasingly prominent. Solar photovoltaic generation has been developed rapidly with its superiotity. The statistical report indicates that distributed photovoltaic power generation will be the main direction of the photovoltaic generation in the near future. With a large number of distributed photovoltaic systems integrated into electricity grid, the demand of the stability of the output power of the photovoltaic power system has become increasingly high. Therefore, it is very significant to improve the accuracy of prediction of photovoltaic power generation.The wide fluctuations of output power of photovoltaic system caused by the instantaneous cover of cloud can not be captured by traditional short-term prediction of photovoltaic power generation. Thus, the accuracy of traditional short-term prediction is limited at time interval of15minutes. At present, in the study of ultra-short-term predition at home and abroad, the prediction accuracy is still limited becasue of the lack description of cloud characteristics impacting the solar radiation. This dissertation presents the dynamic characteristics models of the cloud cluster for the ultre-short-term prediction applicable to the distributed photovoltaic power generation system in urban communities, on the basis of studying the key characteristics of cloud impacting the solar radiation. The main work is as follows:(1) The distributed photovoltaic power ultra-short-term forecasting system based on the shielding effect of the cloud cluster is comprehensive described, including clouds monitoring device, the key dynamic chararcteristics models of cloud cluster and the indentification of the target cloud cluster, sparse cloud cluster shielding effect of solar radiation and prediction, offline model, the design of online compensator, and so on.(2)Through the study of the solar visual orbit, the position of the sun under a certain moment is described by the sun elevation angle and solar azimuth. The effective area of covering the photovoltaic panels in a certain moment is determined by the establishment of cloud shadows back projection model. Combined with a plane mirror projection model, the mapping relationship of the photovoltaic panels obscured by cloud cluster in the mirror vision is established.(3)The key characteristics of cloud cluster impacting the shielding effect of solar radiation is analysed in detail. The cloud cluster location is determined by extracting the common part. Through using the quadratic polynomial model to fit the timing cloud cluster location and using the least square method to determine the best curve fitting, the location prediction model is established. Through using the centroid angle function to describe the shape of cloud cluster and using the quadratic polynomial model to fit the corresponding point of the boundary of the timing cloud cluster, the shape prediction model is established. The cloud mass is divided into eight levels through using cluster analysis. The correspondence between the cloud mass and image gray is established. The different cloud mass regions are directly reflected through the pseudo-color processing. And the cloud mass distribution prediction model is established through extracting the corresponding cloud mass region of the timing cloud cluster and predict the cloud cluster mass distribution in the future like the shape prediction model. The index of solar radiation received by photovoltaic panels is estimated through combining the three kinds of prediction model above.(4)Through the design of the corresponding image processing algorithms and simulation analysis of three groups of timing cloud cluster, based on matlab software, the characteristics prediction of cloud cluster for the future is basically realized.The thesis has built the framework of the distributed photovoltaic power ultra-short-term forecasting system based on the shielding effect of the sparse cloud cluster. The description and prediction models of the key dynamic characteristics including cloud position, velocity, shape and mass distribution are established. The characteristics prediction of cloud cluster for the future is basically realized. Future research will improve the existing characteristics of cloud model, build more key characteristics of cloud cluster model and combine with wind speed, temperature and other factors to improve the prediction accuracy.