Research and Performance Optimization of Visualization of Panoramic Monitoring in Responsive Distribution Network

As the scale of the power grid continues to expand, the number of power grid devices in the national power system is increasing, the structure of the power grid is becoming more and more complex, and the information construction of the distribution network is improving, but the visualization is relatively low in real time. In terms of informationization, the current information presentation and interaction mode cannot adapt to the rapid expansion of power supply scale and daily distribution network operation and maintenance management. In this paper, we study GIS-based grid panoramic visualization display technology and responsive visualization solutions for multiple terminals. These solutions not only improve the monitoring efficiency of the distribution network, but also reduce the time of finding and overhauling when problems occur. It also enables visualization of topological data and timely access to the required information. The experiments in this paper use clustering methods to optimize the visualization elements as well as to optimize the GIS rendering, which effectively improves the visualization efficiency.


Introduction
The distribution grid is an important part of the smart grid, and this part is also directly connected between the power system and each customer. The intelligent distribution operation and maintenance management system aims to collect and summarize distribution grid data and user data, grid structure and geo-graphic information. This system is used to monitor the operational status of the distribution system to provide early warning in case of accidents, or faults in the distribution network. The current distribution network information display is simple, the data source is relatively single, the visualization is not high in real time, and it still relies on manual regular inspection to monitor and manage the operation of the traditional distribution network. When a fault occurs in the power equipment, the corresponding fault point cannot be quickly identified for maintenance, which directly reduces the safety, reliability and efficiency of the distribution cluster.
This paper analyzes the current distribution network based on GIS and topology data for responsive distribution network panoramic monitoring visualization. It realizes the panoramic monitoring of distribution network operation status and the comprehensive display of equipment ledger data, topology data and operation data related to distribution network. It realizes layer customization and multi-layer overlay, and diversified display methods such as micro-charts, hover boxes, and special elements. Realize interactive distribution network automation based on logical and business relationships of distribution network automation panoramic data.

Topological model construction
Topological data structure is a way of organizing spatial data according to the principles of topological geometry. For distribution network graphics, topological data structure only understands the interrelationship among the distribution network diagram elements (points, lines and surfaces) from the abstract concept, without considering the coordinate positions of nodes and line segments, but only paying attention to their adjacency and linkage relationships. Each link of the distribution network generates a large amount of data, from power plant generation, to line transmission, substation transformation, distribution of electricity in distribution stations, customer level consumption to dispatching in operation and maintenance classes. The power plant generates data on power production, operation monitoring and equipment overhaul during the power generation process; telematics, telemetry, remote control, telecontrol, logs and other data generated during the operation of power equipment. Data visualization graphically conveys data information to the grid data as a carrier.
In this paper, we analyze the physical topological connections of typical distribution networks, construct the main topological connections and attributes as diagrams, and save them in the form of diagram databases.
The relationship of substation, transformer, pole tower, switch and line, construct topological relationship, save to the diagram database, and other data save to the relational database. According to the relationship between distribution transformer, pole tower, switch and line, define the nodes of the single-data parent-child relationship and sibling relationship for subsequent display and calculation, and the topological model is shown in figure 1. The node carries attributes such as ID, name, voltage level, latitude and longitude coordinates of the object. The connection relationship represents a section of line between nodes with line ID, line name and other attributes. A connection relationship with the same line ID constitutes a line in the actual service, and various distribution substations, towers, switches, and other node objects are mounted under the line.
In this paper, we consider the commercial graph database and open source graph database, and make a comprehensive comparison in terms of performance, openness and security, and choose the open source software neo4j as the graph database to store topological connection relations.

Application of GIS
GIS is a comprehensive high-tech integrating geography, geometry, computer science and various application models [4]. Because of its powerful geographic data management and analysis capabilities, GIS provides strong support for the management, analysis and maintenance of complex power grid data in the power system. it is of great help to power application decision-making. The complex structure of the power system and the wide geographical distribution are the best development platform and information visualization environment of GIS, especially suitable for monitoring of power systems, such as transmission systems, distribution systems and power-using systems [5].

Visualization Platform Construction
The visualization is divided into three layers: map visualization, topology visualization and data visualization, as shown in figure 2. The map visualization is loaded and displayed by reducing the map by column, allowing you to zoom in and out, switch between different graphics.It can switch between different layers to quickly and precisely locate the corresponding location and view on-site information according to your needs [6].
Based on topology data, topology visualization adds topology layers on the basis of maps to graphically represent the topology of equipment in the power system such as substations, towers, distribution substations, cable wells, branch boxes, lines and other objects. It generates the distribution network topology according to the geographical location latitude and longitude where they are located and their relationship with each other. Different distinctions are made for main lines, branch lines and feeders, and the equipment to be queried is marked and hidden [7].Provide topology editing of towers Data visualization uses topology elements bound to ledger, power, fault and outage data, allowing direct interaction with the elements on the topology diagram to view or maintain related equipment ledger, power, fault and outage data [8].
First, the construction of the geographic information engine is completed to form a map display of the area. On the basis of the map, the distribution network topology data is accessed and we superimpose the topology related to substations, towers, lines on the geographic map. [9]. At the same time, objects such as towers, substations and lines are data-bound with the model to provide humancomputer interaction operation, and the object's ledger data can be viewed. Meanwhile, corresponding pages are provided to display business indicators and business data.

Optimisation
After completing the construction of the topological model and visualization platform, this paper optimizes the visualization image. The element points are superimposed on the map using latitude and longitude information, and the value of the composite index is represented by the size and color of the points. This value is derived by comprehensive calculation and processing. [10]. During map zooming, when the map is scaled down to smaller layers, the amount of data for points within the visible range of the map will be larger, and there may be tens of thousands or hundreds of thousands of points to be rendered. The current general browsers are struggling to handle the rendering of points of this scale, so it is of greater significance to optimize the display of large data volumes.
In this paper, by means of clustering, those that are geo-graphically close to each other in a certain range on the map are fused into a single point. The values of the points are calculated by subjective and objective weighting of multiple indicators to arrive at a combined value. The location of the points is calculated by the k-means clustering algorithm. This reduces the data level of points by one level, which can greatly improve the performance of the system.

Graph element fusion optimization
For a certain zoom level S of the map, each element that needs to be rendered is P0 ,P1 ,P2 ,...Pn ,The latitude and longitude coordinates of each point are (x0 ,y0 ),(x1 ,y1 ),(x2 ,y2 )...(xn ,yn ), clustering according to the distance of each point by k-means algorithm, the specific steps are as follows: (1) Determining the number of clusters K, according to the scaled hierarchy.
(2) Randomly select K points from all the objects to be displayed as the initial center of mass (3) Assigning the remaining objects to the nearest center of mass.
(4) Update the central value of the class using methods such as the mean. (5) Repeat S1, S2, S3, S4 until the K center points no longer change.
For each cluster point after clustering, Pa, Pb, Pc ..., Pn. According to the distance as the weight value, calculate the coordinates of the center point (Xc, Xc ) is Vc , The calculation steps are as follows (table 1).

Determine the sample sequence
According to the clustering results of the above steps, calculate the distance between each sample point and the centroid in each category, and arrange them according to the distance from large to small,denoted as: 1 2 [ , , , ] (1) Among them, i refers to the i-th class after clustering, k is the number of samples included in the i-th class, is the distance between each sample in the i-th class and the centre point, and 1  Determine the sample correction factor The weight of each sample point is determined according to the principle that the larger the distance is, the smaller the contribution of the sample point to the class is. First, the reciprocal of the distance is calculated, that is, the correction coefficient in the above table 1.
 Determine sample weights The correction coefficient obtained by S2 is normalized, and the result is recorded as the weight . Determine the position of each category index value Vi * in all index values, calculate the above According to the method in this paper, in practical applications, the number of primitives that can be displayed in the same window is increased from the original 10,000 to 100,000, and the rendering time for displaying 10,000 primitive points is 0.91s to 0.13s (figure3, 4).

4.2.1.
3D GIS rendering optimization. In order to more realistically and effectively demonstrate the operation of actual lines, substations, etc., support for 3D simulation models is added to GIS. the representation of models in GIS is mainly expressed through the Levels of Detail (LOD) technology, which expresses the process of GIS processing models from simple to complex. [11]. The LOD technique refers to the multi-scale representation of the scene model, thus achieving a balance between data volume reduction and model distortion reduction. It can be expressed by equation Where obj is the true object, S is the approximate formula, and d is the distance from the viewpoint to the object. There is a difference between the real object and the approximate object f(x) represents the true object, y is the object to the viewpoint distance, the view range 0-d, obviously the smaller the difference, the better the approximation effect. When the viewpoint is close to a device, only the LOD4 level of the model is loaded.
The experiments were conducted to test the rendering time and frame rate of the model in the scene, using performance in Chrome developer tools to check the performance of the page. The performance tests were performed on the original scene without data hierarchy and on the mixed model scene with LOD hierarchy, and then the frame rate was read out by writing a program to render the scene. This test experiment took 100 seconds of frame rate variation for each of the original model and LOD1 to LOD4 models respectively, and thus the data obtained was plotted as a corresponding frame rate variation graph, as shown in figure 5. The frame rate indicates the smoothness of the rendering process, the higher the frame rate, the smoother the picture. From the graphs, we can see that LOD1-LOD2 ends up at 60fps, LOD3 at 40fps, and LOD4 at 35fps. All cases of direct rendering of the original model end up at 20pfs, and the final stable values of LOD1-LOD4 frame rate curves are higher than the original model. According to the analysis of the graph, if the user is not interested in model details, the scene browsing will be smoother if the view is at LOD1-LOD2 level when the large volume of model interiors and model details are not rendered. Even if the user is interested in a model and goes to the LOD4 level to explore its details, only a single facility is rendered. This greatly reduces the work pressure on the system and avoids the resource-wasting disadvantage of the system rendering all complete models no matter where the viewpoint is. It improves the orderly management and scheduling of spatial data, and can effectively improve the efficiency of visualization.

4.2.2.
Experiment analysis. According to the distribution network topology model and visual monitoring architecture proposed in this paper, and the optimization method based on graph element fusion and LOD hierarchical rendering. The panoramic visualization display of large-scale graphical elements and 2-3D fusion is realized. Firstly, the topological model is constructed, as in figure 6 and secondly, the panoramic monitoring visualization platform is constructed based on GIS, and finally the experiments are applied to optimize the graph fusion and visualize the model based on multi-layer LOD rendering, as in figure 6. These optimizations greatly reduce the space occupied by the graph element data and the graph element rendering time in the system, and the LOD layering process of the visualization system greatly reduces the workload of the system and greatly improves the visualization efficiency.

5.
Conclusion The panoramic monitoring system of distribution network based on GIS technology studied in this paper integrates dis-tribution network topology information, network frame equipment status parameters and geospatial information. It realizes a panoramic three-dimensional multi-dimensional view of line and geographic information from the perspectives of time, space, topological dimension, geography, and business dimension, revealing the hidden laws and values behind the power data. Through the large screen visualization display, it can intuitively view the operation status of the distribution network and realize the rapid access to the regular information of the distribution network in the safety and quality, energy efficiency and other dimensions. Mastering the GIS-based distribution grid panoramic monitoring method proposed in this paper can be widely used in the design and development work of future distribution grid panoramic monitoring visualization.