Research On Fast Visualization Analysis Model Of Power Big Data Based On Three-dimensional Scene

Visualization methods has always been an important means in data presentation. Research shows that human obtained over 80% information from the outside world through visual system.When data is presented in a perpetual graphical form, users can dig out the information conveyed by data immediately and comprehend and transform it into knowledge quickly. Due to the exploded increase in volume and great extension in dimension of data, traditional visualization methods are no longer well applicable in the era of big data. In view of the traditional two-dimensional visualization methods are actually two-dimensional projection of the N-dimension data, they perform high density and excessive compressed dimension in one graph with spatial information missed, which is not able to reflect the characteristics of the data efficiently. Power big data is distinguished from other categories of data for it has a structure of typical spatial topology, where discrete data can generally be corresponded to the electrical equipment in three-dimensional space. This study proposes a new conception which takes the three-dimensional scene as the fundamental element of visualization analysis of power big data,where data analysis and visualization transactions take place, thus implements the demonstration of spatial characteristic of data and of relations between data, which is useful in fast human-computer interaction.Based on the conception of integrating VR scene and power big data analysis, firstly a fast visualization analysis model of power big data based on three-dimensional VR scene is proposed,in which levels are designed, targeting at different abstract modules like transaction, engine,computation, control and store. The regularly departed modules of power data storing, data mining and analysis, data visualization are integrated into one platform by this model. Secondly,when considering the issue of large scene rendering in this model, a highly efficient hidden surface removal algorithm and a fast multi-resolution rendering algorithm are designed in order to accelerate the constructing and rendering of large scene of power system. The former method implements a scene graph by octree which enables efficient occlusion culling, thus reduces the graphic data that is delivered to pipeline. The latter method implements an LOD weight function aiming at the issue of reducing mesh complexity, which quantizes the contribution of all elements in a mesh and optimizes them progressively and dynamically in a non-destructive strategy. Finally, an example of a visualization analysis of transformer oil chromatography in a substation is demonstrated to show the model and methods proposed by this study are feasible and effective.The experiments shows the visualization analysis model of power big data based on three-dimensional VR scene meets the requirement of data analysis in power system applications,and meets the requirement of enhancing the comprehension when human-computer interactivity is in progress. The model is able to improve the efficiency of visualization analysis, in which method of hidden surface removal performs high efficiency in dealing with large scene, and method of multi-resolution rendering performs effective shape retaining in electrical equipment models.