Middleware Key Technologies Supporting Panoramic Display of Low-voltage Distribution Substation Area Operation

A panoramic display system for low-voltage distribution substation areas is built in this paper to meet the visual monitoring requirements of equipment operation status in low-voltage distribution substation areas. Aiming at the characteristics of complicated data structure, decentralized deployment, and equipment isomerization of power equipment, a middleware for panoramic display system of low-voltage distribution substation area operation is designed. The residual search method realizes the data identification to give a timely warning. Heterogeneous data is clustered by the K-means method. Simulation results verify the better performance of the proposed middleware for low-voltage substation operation.


Introduction
With the development of the new power system, the low-voltage distribution substation area needs to carry more operating equipment.Because of the construction of information systems and different applications of the relatively scattered, it is hard to form a centralized and unified information chain for making strategy, thereby being unable to support security control and auxiliary strategy-making.Hence, it is urgent to conduct panoramic and visual displays for low-voltage distribution substation areas.Among the existing technologies, the integration of cloud-edge-device collaboration and digital twin can provide a new idea for panoramic displays of low-voltage distribution substation area operation [1] [2] .However, due to the complex data structure, decentralized deployment, and equipment isomerization of the equipment in the low-voltage distribution substation area, it is difficult to cluster analyze the operation data.Meanwhile, due to the complex communication environment and interferences in the data acquisition environment, the equipment operation may generate bad data.It is necessary to develop a middleware to identify bad data of the power equipment operation data and cluster the operation data to support the panoramic display system's full situational awareness of the operation status of lowvoltage distribution substation areas.
Middleware is software for frequently encountered problems like heterogeneity, security, and resource-sharing.Middleware is integrated with the panoramic display system through the integration mechanism based on service-oriented architecture (SOA) and the defined open standard interface, which has the advantages of lightweight, loose coupling, flexibility, and so on [3]- [6] .Hua et al. [7] proposed a CORBA-based middleware for distributed power monitoring systems.However, the CORBA technology is huge and complex and lacks the standard configuration object application method, which is difficult to adapt to the development of low-voltage distribution substation areas of the current new power system.Zhou et al. [8] proposed a middleware with the integration of Oracle for the power information platform.However, Oracle installation operation is relatively complex and costly, which is difficult to be widely applied in low-voltage distribution substation areas.
Hence, we first propose the middleware application architecture supporting the panoramic display of low-voltage distribution substation area operation.Then, we study the specific functions of the middleware including the identification of bad data through the residual search method and the clustering processing of heterogeneous data through the K-means method.Finally, the proposed middleware is verified by the simulation to support the realization of panoramic displays for low-voltage distribution substation areas.The middleware integrates the bad data identification method and K-means clustering method for the heterogeneous complex data of power equipment operation [9].Middleware application architecture supporting panoramic display of low-voltage distribution substation area operation is shown in Figure 1.The power operation equipment of low-voltage distribution substation areas generates a large amount of operation data during operation.The data acquisition terminals are deployed to collect the operation data and transmit them to the data middle platform through 5G for storage [10].Then, the middleware detects the collected data and sends the bad data to the monitoring and warning center for warning in time [11] .Meanwhile, the middleware clusters the remaining data and delivers it to the data analysis center for analysis.Finally, the results are delivered to the panoramic display center to realize the visual analysis and display of equipment operations in low-voltage distribution substation areas.

Middleware Function Design of Panoramic Display System for Low-voltage Distribution Substation Area Operation
The middleware functions of the panoramic display system include bad data identification and K-means clustering, which are introduced as follows.

Bad data identification
In the process of data acquisition and processing, changes in the surrounding environment and the instability of communication signals will lead to certain errors in equipment operation data acquisition and affect the accuracy of data and data analysis results.Bad data is defined as measurement data with large deviation or error.At present, bad data identification is usually through data state estimation [12] .
The equipment operation state estimation can configure the real state of the equipment under the condition of measurement error, and provide the basis for ensuring the high quality of the data.Its measurement equation z is formulated as: ( ) where ( ) h x represents the measurement function and v is the measurement error subject to normal distribution.Because the measurement error often changes, the objective is redefined as: where i  is the measurement error weight.

Figure 2. Bad data identification process
Based on the idea of optimization, the weighted least squares method is adopted to express the error objective function as follows: where R is the variance matrix, its dimension is m, and the diagonal element is 2 i  .The residual search method is a widely used method to deal with residuals at present, and its workflow is shown in Figure 2.

K-means clustering
K-means clustering algorithm adopts Euclidean distance as the similarity evaluation index.This algorithm takes into account that clusters contain objects close to each other.Hence, the final target will obtain compact and independent clusters.K-means clustering is easy to execute, has efficient data mining ability and good scalability, and is adopted in image processing, fault diagnosis, and other fields widely.
The basic idea of K-means: there is a set of data points (i=1,2,...,n), determine K firstly, i.e., the number of clusters, and randomly select the data points i V (i=1,2,...,k) as the initial center of each cluster; Second, according to the principle of proximity, divide each data point into the cluster where the hithermost cluster center is located; further, each cluster center is updated The new cluster center is the data point mean of each cluster.Finally, calculate the sum of squares of the distance from each data point to the center of the cluster E through Formula ( 4).If the value of the objective function reaches the minimum, the clustering will be completed.
Otherwise, the clustering will be performed again according to the new cluster center.
where X is the data point in m R ; i V represents the mean value of the cluster i C .
Through the K-means clustering algorithm, the operation data of heterogeneous equipment is clustered and delivered to the panoramic display center to realize the visual analysis and display of equipment running status.

Simulation
We take the equipment operation data as the sample data to execute the simulation.The standard deviation of the measured value is 0.02, and the standard deviation of the phase angle is 0.005.The interaction assisted by the middleware between the equipment operation data and the panoramic display system is analyzed by using the simulation device of the low-voltage distribution substation area.Firstly, 100-500 sample data are randomly selected as bad data respectively to measure the data accuracy when the number of sample data is 2000.Then set the number of sample data as 1000-5000 respectively, and set the number of clusters k=6, to measure the analysis time of operation panoramic display systems.And compared with the C-B middleware and without middleware, to verify that the middleware proposed can effectively and reliably support the realization of panoramic display of low-voltage distribution substation area operation.3 shows the data accuracy result between the middleware proposed and the C-B middleware and without middleware in the interaction between the equipment operation data and the panoramic display system.The data accuracy of the proposed middleware outperforms that of the C-B middleware and without middleware by 11.2% and 14.2%, respectively.The reason is that the middleware proposed adds a bad data identification module based on the residual search method, which raises the data accuracy.
Figure 4.The data analysis time of the equipment operation data Figure 4 shows the comparison of data analysis time between the middleware proposed and the C-B middleware and without middleware.The analysis time of the equipment operation data with the participation of the proposed middleware is 17.3% and 44.8% less than that of the C-B middleware and without middleware, respectively.The reason is that the proposed middleware adds the clustering method based on K-means to reduce the pressure of the panoramic display system when analyzing the equipment operation data.

Conclusion
This paper proposed the middleware application architecture supporting the panoramic display of lowvoltage distribution substation area operation and studied the specific functions of the middleware facing the low-voltage distribution substation area operation panoramic display system.Bad data identification and the clustering processing of heterogeneous data are investigated through the residual search method and the K-means method, respectively.Finally, the proposed middleware is verified by the simulation to support the realization of the panoramic display for low-voltage distribution substation area operation.

Figure 1 .
Figure 1.Middleware application architecture supporting panoramic display of low-voltage distribution substation area operation

Figure
Figure3shows the data accuracy result between the middleware proposed and the C-B middleware and without middleware in the interaction between the equipment operation data and the panoramic display system.The data accuracy of the proposed middleware outperforms that of the C-B middleware and without middleware by 11.2% and 14.2%, respectively.The reason is that the middleware proposed adds a bad data identification module based on the residual search method, which raises the data accuracy.