Automatic update method of GIS platform drawing model based on machine learning

Abstract

Aiming at the problem that the number of data bytes in the traditional automatic update technology of GIS platform is small, a method of automatic update of GIS platform graph model based on machine learning is studied. Firstly, the data of the GIS platform model is convolved by the iso-linear feature detection operator in the automatic updating technology of the GIS platform model, and the calculated data of the GIS platform model is expressed as spatial data. A reasonable updating criterion is established, the spatial relationship of GSI data is reconstructed by the measure of updating criterion, the data vector of GIS platform model updated within the updating time range is calculated, and the regional data elements in the space are constantly changed to complete the data updating of GIS platform model. The experimental results show that compared with the automatic updating method of GIS platform model, the proposed method can update more data bytes with the same number of data bytes.

Keywords

Machine learning GIS platform automatic updating of model spatial relations

1. Introduction

Automatic data update technology is the process of automatic data update in the system. Geographic information system is a geographic information system developed on the basis of geographic science, computer technology, remote sensing technology and information science. Its basic function is to collect data, convert data and ensure the integrity and logical consistency of stored data. With the continuous development of urban informatization, most cities in China have begun to prepare digital city construction planning based on GIS. The resource generation, transmission, conversion, distribution and utilization of smart grid are distributed in a broad space field. Spatial data is one of the core data for the survival of power grid enterprises. Geographic information system (GIS) is widely used in power grid automation, production, marketing, power grid planning, power communication, emergency command and other core businesses. GIS technology is of special significance to improve the technical level and application value of relevant application systems in power grid enterprises [1, 2]. Distribution automation system is the combination of modern electronic technology, computer and network technology, communication technology and power equipment, which integrates user information, real-time information, grid structure parameters, offline information and geographic information of distribution network. The monitoring, protection, control, measurement and work management of power supply department under normal and accident conditions are organically integrated to form a complete automatic management system [3]. It is a huge, complex and highly comprehensive system, which is mainly used to ensure the quality of power supply, improve service level and reduce operating costs [4, 5]. With the continuous development of social economy and power supply and demand market, users’ requirements for the reliability and quality of GIS platform are constantly improving. The traditional operation mode and management method of GIS platform have been difficult to meet the requirements of safe, high-quality and economic operation [6, 7]. Improving the equipment and operation of the whole power system and realizing the dispatching automation and distribution automation of local power grid are the only way to ensure the safe and economic operation of power grid and improve power quality.

The application system of power grid public information maintenance is a kind of system in which the GIS platform can further supplement the power grid related public information based on the power grid model [8, 9]. In such applications, there are practical business functions involving the change of power grid related public information, such as power marketing management system, which can provide power user information. Based on the research background, machine learning is applied to the automatic updating of GIS platform to improve the updating ability of GIS platform. In the conventional tightly coupled architecture, GIS is embedded in each graphics related application system. GIS data and various application data are mixed together and cannot be disassembled. All kinds of application functions of the system are directly carried on GIS. The public power grid GIS platform in this architecture is not a commercial GIS platform in the conventional sense, but an enterprise level shared GIS platform with advanced encapsulation and built-in power grid model perception ability [10, 11]. It is very important to decouple the GIS platform from the application system and upgrade it to the enterprise level. The innovation of the research content is not only to provide highly consistent shared geographic information for various application systems of power grid enterprises, but also to establish enterprise level shared power grid basic model through reasonable division of labor and collaborative synchronization, so as to realize the continuous update of power grid model through normal power grid production and business activities.

2. Design of automatic updating method for gis platform graph module

2.1 Using automatic data updating technology to process GIS platform change data

Before processing GIS change data, we use the supporting distributed data mechanism of automatic data update technology to extract GIS platform model data [12]. According to a certain strategy and cycle, a small part of GIS platform map model data is edited offline. When editing, check out/check in is used to copy the structure, as shown in Fig. 1.

Figure 1.

Check out/check in replication structure.

Using the replication structure shown in Fig. 1, the data is replicated by using replica, and finally the total GIS platform model database is formed. The architecture positions the public grid GIS platform as an application transparent platform combining location awareness and grid awareness. On this basis, the detailed integration scheme of data sharing, information complementation and function interaction between grid geographic graphics application and public grid GIS platform is given. Then, using the operator of equal linear feature detection, machine learning processing is performed on the formed GIS platform map model data [13]. The calculation formula is as follows:

$\displaystyle\textit{Score}=\frac{G\left({\left|{a_{1}-a_{2}}\right|}\right)% \times G\left({\left|{a_{2}-a_{3}}\right|}\right)}{\sum\limits_{i=1}^{3}{\left% ({F\left({a_{i}-b_{i}}\right)+F\left({a_{i}-c_{i}}\right)}\right)}}$ (1)

In the formula, $G\left(\right)$ is a function related to the uniformity of linear features, $F\left(\right)$ depends on the contrast between linear features and their surrounding background, $a$ , $b$ , $c$ represent the existing GIS platform model data set. The distribution of data sets in each part of the space calculated by Eq. (1) is shown in Fig. 2.

Figure 2.

Distribution of change data sets.

As shown in Fig. 2, the data marked by letters are the existing data sets, and the rest are the data sets of GIS platform model data that have changed [14]. In order to prevent false detection of linear features in the change set of GIS platform model data, the detection factor in Eq. (1) is modified. The calculation expression is as follows:

$\displaystyle S^{\prime}_{\textit{core}}=\frac{G\left({\left|{a_{1}-a_{2}}% \right|}\right)\times G\left({\left|{a_{2}-a_{3}}\right|}\right)}{\max\left({% \sum\limits_{i=1}^{3}F\left({a_{i}-b_{i}}\right),\sum\limits_{i=1}^{3}F\left({% a_{i}-c_{i}}\right)}\right)}$ (2)

The original GIS platform model data with rough linear characteristics are cleaned by the linear detection factor modified by Eq. (2). The processing width of the model data of GIS platform is set to 1–3 pixels [15]. At this time, the cleaned GIS change data set can be expressed as:

$\displaystyle S_{\textit{core}}^{\prime\prime}=\frac{G\left({\left|{a_{1}-a_{2% }}\right|}\right)\times G\left({\left|{a_{2}-a_{3}}\right|}\right)\times G% \left({\left|{a_{3}-a_{4}}\right|}\right)\times G\left({\left|{a_{4}-a_{5}}% \right|}\right)}{\max\left({\sum\limits_{i=1}^{5}F\left({a_{i}-b_{i}}\right),% \sum\limits_{i=1}^{5}F\left({a_{i}-c_{i}}\right)}\right)}$ (3)

In order to prevent the accidental loss of the change data set calculated by Eq. (3), the GIS platform model data set calculated by Eq. (3) is stored in the form of “snapshot” according to different time periods, and the pixels of the processing data are used as the stored data, which are sorted according to their own time values to obtain GIS platform model data sets in different time orders. After using Eq. (3), the detection factor in the formula will trace the GIS platform model data in the specified operation time sequence according to the needs, so as to ensure the integrity of the GIS platform model data. After the processing, the GIS platform model data will form a new spatial relationship database. In order to realize the automatic updating of GIS platform model data, reconstruct a spatial relationship of GIS platform model data, and complete the research on the automatic updating technology of GIS platform model data.

2.2 Reconstructing spatial relationship of GIS platform data

When using machine learning to reconstruct the spatial relationship of GIS platform data, the calculated GIS data is described as spatial data, and the final results are shown in Table 1 [16].

Table 1
Spatial data after description

SDO_GID	SDO_X1	SDO_Y1	SDO_X2	SDO_Y2
1011	P1(X)	P1(Y)	P2(X)	P2(Y)
1012	P2(X)	P2(Y)	P3(X)	P3(Y)
1015	P3(X)	P3(Y)	P4(X)	P4(Y)
1010	P4(X)	P4(Y)	P5(X)	P5(Y)
1013	P5(X)	P5(Y)	P6(X)	P6(Y)
1012	P6(X)	P6(Y)	P7(X)	P7(Y)
1013	P7(X)	P7(Y)	P8(X)	P8(Y)
1015	P8(X)	P8(Y)	P1(X)	P1(Y)
1018	G1(X)	G1(Y)	G1(X)	G1(Y)
1012	G2(X)	G2(Y)	G2(X)	G2(Y)
1019	G3(X)	G3(Y)	G3(X)	G3(Y)
1010	G4(X)	G4(Y)	G4(X)	G4(Y)

From the spatial data shown in Table 1, when reconstructing the GIS platform data space, the raster data structure and vector data structure are used to reconstruct the spatial relationship. Firstly, machine learning is used to represent GIS platform data. In order to eliminate the restriction of attribute word length of geographical entities, the coordinates of sampling points are directly recorded based on geometric spatial coordinates [17]. Then, the grid structure is used to divide the coordinates of the sampling points into grid arrays with uniform size. Each grid is recorded as a pixel and marked as a row and column number. Fuzzy clustering is used to divide the row and column numbers at this time. Suppose that the row and column number obtained at this time is $X$ , and the pixel $x_{k}$ belongs to the length of class $i$ , which is expressed by ambiguity $u_{ik}$ . The calculation formula is as follows:

$\displaystyle U=\left[{{\begin{array}[]{ccccc}{u_{11}}&{u_{12}}&{u_{13}}&% \cdots&{u_{1n}}\\ {u_{21}}&{u_{22}}&{u_{23}}&\cdots&{u_{2n}}\\ \vdots&\vdots&\vdots&\ddots&\vdots\\ {u_{i1}}&{u_{i2}}&{u_{i3}}&\cdots&{u_{ik}}\\ \end{array}}}\right]$ (4)

When the fuzzy matrix of Eq. (4) is used to deal with row and column numbers, the following constraints are satisfied:

$\displaystyle\left\{{{\begin{array}[]{l}{u_{ik}\in[0,2]\quad\forall i,k}\\ {\sum\limits_{i=1}^{i}{u_{ik}}=1\quad\forall k}\\ {0<\sum\limits_{k=0}^{n}{u_{ik}}<n}\\ \end{array}}}\right.$ (5)

According to the constraint relationship of Eq. (5), the GIS data is converted to the unique row and column number containing the code. In order to avoid the problem that column numbers can not update data codes due to the poor current situation of data source and the change of update scale, a reasonable update criterion is established. The measurement function of updating criteria is defined as:

$\displaystyle J_{\text{FCM}}(U,V)=\sum\limits_{k=0}^{n}{\sum\limits_{i=1}^{c}{% u_{ik}^{m}}}\left\|{x_{k}-v_{i}}\right\|^{2}$ (6)

In the formula, $m$ is the weighted index and $J_{\text{FCM}}(U,V)$ is the weighted sum of squares of all kinds of data. With the gradual increase of $m$ value, the fuzziness of the updating criterion becomes greater. When $m=$ 1, the measurement of the updating criterion function is the most accurate, so the spatial relationship of GIS can be reconstructed as shown in Fig. 3.

Figure 3.

Reconstructed GIS spatial relationship.

The spatial relationship of the final reconstructed GIS data is shown in Fig. 3. P1, P2, P3, P4, P5 and P6 respectively represent 6 points marked on the map. In the spatial region in the figure, in order to prevent the inconsistency between the data edge processing and the common boundary line of the map spot, which leads to the problem that the spatial relationship of GIS data cannot be superimposed, the update criterion metric function is split into pure relational mode to express [18, 19]. The GIS data transmitted by the server is used to form a tolerance structure on the WAN. The tolerance structure is shown in Fig. 4.

Figure 4.

Tolerance structure.

Using the tolerance structure shown in Fig. 4, the spatial relationship of GIS data is superimposed to form data elements in different regions, and the regional elements are constantly changed to update the map model data of GIS platform. For the realization of GIS platform graphics module automatic update to prepare.

2.3 Realization of automatic updating of graph module on GIS platform

On the basis of the research, under the new idea of integrated architecture, the integrated maintenance function of graph and model is completed by the division and cooperation of power grid model maintenance application system and GIS platform. Generally, the business entry point is the power grid equipment change business, and the power grid model maintenance application system interacting with GIS platform is the power grid production management system [20]. In this integrated mode, we need to use the version control technology of graphics and equipment account in the GIS platform and production management system respectively, and finally form the power grid topology and equipment public information in the GIS platform, as shown in Fig. 5.

Table 2
Attribute and number of bytes of map module data boundary point of GIS platform

Name	Boundary point properties	Number of bytes
Coordinate 1954_x	JZZX54X	13
Coordinate 1954_y	JZZX54Y	12
Coordinate 1954_B	JZZX54B	13
Coordinate 1954_L	JZZX54L	12
Coordinate 1954_H	JZZX54H	13
Coordinate 1980_x	JZZXS0X	13
Coordinate 1980_x	JZZXB0Y	12
Coordinate 1980_y	JZZX80B	13
Coordinate 1980_B	JZZX80L	12
Coordinate 1980_L	JZZX80H	13
Coordinate 1980_H	JZZX00X	12
Coordinate 2000_x	JZZX00Y	13
Coordinate 2000_y	JZZX00B	13
Coordinate 2000_B	JZZX00L	12
Coordinate 2000_L	JZZX00H	13
Coordinate 2000_H	JZZX84X	13
Coordinate WGS84_x	JZZX84Y	13
Coordinate WGS84_y	JZZX84B	12
Coordinate WGS84_B	JZZX84L	13

Figure 5.

New maintenance mode of integration of power grid diagram and model.

The production management system starts the process of equipment change and transfers the initialization information of equipment changing to the GIS platform. According to the received equipment change initialization information, the GIS platform locates the graphics that need to be changed and modifies the related graphics, which may involve creating a new equipment account or modifying or deleting the existing equipment account, but only within the GIS platform and only maintaining the public data of the equipment account. GIS platform establishes and maintains the grid topology information synchronously during the integrated maintenance of grid graph and model to meet the needs of subsequent topology applications [21]. When the GIS platform saves the graphics and equipment account, it calls the interface service of the production management system to create or delete the corresponding equipment account in the production management system, and then it can maintain the non-public equipment account data through the equipment account maintenance page embedded in the production management system.

After the maintenance of graphics and equipment account is completed, the GIS platform sends a message to the production management system to inform it to complete the audit of graphics and equipment account [22]. During the process of auditing the graphics and equipment account, the production management system can call GIS platform to obtain the latest graphics and their change information needed for auditing, and can synchronously return the graphics and equipment account to the specified version. If the final audit fails, the production management system will inform the GIS platform to draw and modify the graphics or equipment account again. If it is approved, the production management system will call the graphics publishing interface service of GIS platform to publish the latest version of graphics (including the public data of equipment account), and publish the latest equipment account data after confirming that the graphics are published successfully. In the meantime, if the graphics are not published successfully, the equipment account will not be published.

After the synchronous release of graphics and equipment accounts, the GIS platform can also provide a special verification tool to accurately verify the corresponding consistency of graphics and equipment accounts between the GIS platform and the production management system, so as to deal with the unexpected situation during or after the interaction between the two systems.

In the integrated interaction process, the production management system is responsible for assigning unique identification to the new equipment, and the GIS platform localizes and saves the assigned unique identification, through which the association between grid elements and grid equipment is realized [23].

The production management system can also regularly synchronize the complete equipment information to the data center according to the unified data model for the use of GIS platform and other application systems.

The integrated design of integrated maintenance of power grid diagram and model is the most typical application in the distribution network information management, which is the basic guarantee function for the normal operation of all kinds of distribution network information applications. Through the integrated maintenance of the distribution network model, it can effectively avoid the repeated maintenance of the frequently changed basic equipment account by the grass-roots users, and ensure the consistency of the basic data in each system.

To sum up, the automatic updating technology is used to process the GIS platform change data, reconstruct the spatial relationship of GIS platform data, and realize the automatic updating of GIS platform model.

3. Experimental comparative analysis

3.1 Experimental preparation

The hardware environment is prepared for the experiment, including a LT30 handset, a DELLPC compatible machine (with relevant hardware configuration), and the operating system is Windows 2007 or Windows 2010. The software adopts virtual PDA environment and Microsoft ActiveSync for data transmission between PC and handset. The development tool adopts Microsoft Embedded Visual C++ 3.0 and Microsoft Visual Studio 2008. Based on the hardware and software preparation, the attribute and byte number of boundary point of GIS platform map model data collected are shown in Table 2.

Based on the experimental preparation, two kinds of traditional GIS platform map model data automatic updating technology and GIS platform map model data automatic updating technology based on data automatic updating technology are used to carry out the experiment, and the number of bytes that can be updated by three kinds of automatic updating technology is compared.

3.2 Experimental results

For the same attribute and number of GIS platform map model data boundary points, the experimental results of the number of updated bytes are shown in Fig. 6.

Figure 6.

Experimental results of automatic data update technology.

As shown in Fig. 6, for the GIS platform model data collected in the preparation stage of the experiment, the average number of bytes finally updated by traditional update technology 1 is 8 for different boundary point attributes. When traditional update technology 2 updates the GIS platform model data in Table 2, the number of bytes finally updated is 4. The average number of update bytes is 12, which can update all the bytes of GIS platform model data, and the update effect of GIS platform model data is the best.

4. Conclusion

A method based on machine learning for automatic updating of GIS platform model data is proposed. The updating of GIS platform model data can be used to maintain the real-time of GIS platform model data. Updating the unrealistic data in the database with real-time status data and change data can update the data as a whole and locally, enrich the ways of data collection, and improve the timeliness of data update. It can realize the correct connection between the updated data and the original data while keeping the original data unchanged.

For future research, it can be discussed from the following aspects:

Some “broken” polygons may be generated in the process of updating. Further research work should explore how to automatically process the “broken” polygons of GIS platform model.

In further research, the storage, query and recovery of historical data should be considered.

We can study how to establish a whole of a complete object and realize the association update relationship between the whole and parts of a complete object. That is, if a part of a complete object is modified, the whole of the complete object will also be updated automatically, including automatic update of attributes of GIS platform model.

Footnotes

Acknowledgments

Guizhou Power Grid Co., Ltd. technology projects for No. GZKJXM20190715.

References

Lan

. Debris flow analyst (DA): A debris flow model considering kinematic uncertainties and using a GIS platform. Engineering Geology. 2020; 279: 105877.

Alhamwi

Medjroubi

Vogt

Agert

. Development of a GIS-based platform for the allocation and optimisation of distributed storage in urban energy systems. Applied Energy. 2019; 251(1): 113-120.

Samal

Maurer

Blitterswijk

Truckenmüller

Giselbrecht

. A new microengineered platform for 4D tracking of single cells in a stem-cell-based in vitro morphogenesis model. Advanced Materials. 2020; 32(24): 207-212.

Nawrocki

Grzywacz

Sniezynski

. Adaptive resource planning for cloud-based services using machine learning. Journal of Parallel and Distributed Computing. 2021; 3(6): 59-62.

Najafi

Pourghasemi

Ghanbarian

Fallah Shamsi

. Land-subsidence susceptibility zonation using remote sensing, GIS, and probability models in a Google Earth Engine platform. Environmental Earth Sciences. 2020; 79(21): 491-500.

Bhattacharjee

Mohanty

. GIS-based approach for the measurement of variability in tectonomorphic signatures using DEM’s data: A case study from the Habo Dome in the Kachchh area, India. Environmental Earth Sciences. 2020; 79(18): 56-63.

Liu

. Development and application of meteorological disaster monitoring and early warning platform for characteristic agriculture in Huzhou city based on GIS. Asian Agricultural Research. 2017; (1): 50-52.

Thomas

Michael

Brown

Shaghayegh

Ulrich

Ward

KCH

Alexander

Maximilian

Tanja

. Genetic interaction mapping and exon-resolution functional genomics with a hybrid Cas9-Cas12a platform. Nature Biotechnology. 2020; 38(5): 638-648.

Sanchez-Aparicio

Martin-Jimenez

Pozo

Gonzalez-Gonzalez

Laguela

. Ener3DMap-SolarWeb roofs: A geospatial web-based platform to compute photovoltaic potential. Renewable and Sustainable Energy Reviews. 2021; 135(12): 59-63.

10.

Teschendorff

. Avoiding common pitfalls in machine learning omic data science. Nature Materials. 2019; 26(18): 422-427.

11.

Dubitzky

Lopes

Davis

Berrar

. The open international soccer database for machine learning. Machine Learning. 2019; 108(8): 9-28.

12.

Hulbert

Rouet-Leduc

Johnson

Ren

Riviere

Bolton

Marone

. Similarity of fast and slow earthquakes illuminated by machine learning. Nature Geoscience. 2019; 37(12): 69-74.

13.

Mojtaba

Milad

Meysam

Hamed

Mohammad-Hassan

. DC/DC power converter control-based deep machine learning techniques: Real-time implementation. IEEE Transactions on Power Electronics. 2020; 12(99): 1-10.

14.

Du-Harpur

Arthurs

Ganier

Woolf

Lynch

. Clinically-relevant vulnerabilities of deep machine learning systems for skin cancer diagnosis. Journal of Investigative Dermatology. 2020; 14(56): 85-92.

15.

Miller

Gallidabino

MacRae

Owen

Bury

Barrona

. Prediction of bioconcentration factors in fish and invertebrates using machine learning. Science of The Total Environment. 2019; 648(15): 80-89.

16.

Kelley

Ziatdinov

Collins

Susner

Vasudevan

Balke

Kalinin

Jesse

. Fast scanning probe microscopy via machine learning: Non-rectangular scans with compressed sensing and Gaussian process optimization. Small. 2020; 16(37): 65-72.

17.

Fouedjio

Klump

. Exploring prediction uncertainty of spatial data in geostatistical and machine learning approaches. Environmental Earth Sciences. 2019; 78(38): 55-63.

18.

Lam

Nguyen

Bui

Chung

Chang

Nehmetallah

Raub

. Quantitative scoring of epithelial and mesenchymal qualities of cancer cells using machine learning and quantitative phase imaging. Journal of Biomedical Optics. 2020; 25(2): 1-11.

19.

Wang

Rui

. Research and application of machine learning method based on swarm intelligence optimization. Journal of Computational Methods in Sciences and Engineering. 2019; 19(1): 179-187.

20.

Liu

Ding

Shaham

Rahayu

Lin

. When machine learning meets privacy a survey and outlook. ACM Computing Surveys. 2021; 54(2): 69-72.

21.

Libera

Pedretti

Tateo

Mason

Piccinini

Fabbri

. Conceptual model of arsenic mobility in the shallow alluvial aquifers near Venice (Italy) elucidated through machine learning and geochemical modeling. Water Resources Research. 2020; 56(9): 65-72.

22.

Tan

Wang

Gong

. Discriminating soybean seed varieties using hyperspectral imaging and machine learning. Journal of Computational Methods in Sciences and Engineering. 2019; 19(4): 1001-1015.

23.

Thrampoulidis

Mavromatidis

Lucchi

Orehounig

. A machine learning-based surrogate model to approximate optimal building retrofit solutions. Applied Energy. 2021; 281(12): 52-60.