Data Object Modeling for Industrial Internet of Things and Its Efficient Implementation in a Cloud Test System for Relay Protection

The IEC61850 standard contains a hierarchically structured data model, which is abstracted into Server, Logical Device, Logical Node, Data Object, Data Attribute, etc. by virtualizing the actual physical device IED [29].There are 10 parts in the IEC61850 standard, and the first 5 parts of the IEC61850 mainly focuses on the general requirements, terminology, system and engineering management, communication requirements, and other basic descriptions about the system. Part 6 deals with the configuration description language in terms of functionality, modeling, and communication. Part 7 describes the information exchange model, which describes the basic model of the standard, abstract communication services, and basic data types. Parts 8-9 are about mapping abstract services to communication protocols and transmitting communications according to the protocols, and Part 10 is about conformance testing of the system.

The IEC61850 standard proposes the concept of station information layering, which utilizes the method of information layering to achieve the purpose of station communication data division. The advantages of adopting a layered distributed structure are as follows: 1)

It is conducive to the independent installation of electrical equipment and the corresponding protection and measurement functions, and the system operation is not affected by the failure of a single element.

From a hierarchical point of view, the substation automation system is divided into station control layer, interval layer and process layer, which defines the communication interfaces between and within the layers, effectively realizing the exchange of various types of data between the two sides of the communication. The process layer collects analog quantities of system operation, monitors the operating status of equipment, sends control commands related to the function of primary equipment, etc., and interacts with the interval layer. The interval layer provides a summary of the data processing for an interval and operates its primary equipment according to communication information. The station control layer summarizes the data information of the whole station, executes the tripping and blocking signals of the whole station, and realizes the communication functions such as human-machine interface and remote control.

The IEC 61850 standard uses an object-oriented model that decomposes substation functions into logical nodes representing the smallest functional units. It also defines a communication service model and an information model, including a server (Server), logical devices (LD), logical nodes (LN) and data objects (DO) and data attributes (DA).

The Server encapsulates logical nodes and data and offers interfaces for external access. A logical device consists of logical nodes and additional services representing a set of functions, each representing a logical node. A logical node is the smallest unit that implements a data function. Data objects provide a means of illustrating the representation of information, and data attributes define the name, format, range, or allowed values, and the notation of the sampled values to be transmitted.

Overall, to build the IEC61850 information model, it is first necessary to define the specific logical nodes and data objects required for the goal, and to construct logical devices based on the selected objects, which are then constructed into servers to publish/subscribe the communication data to realize the transfer of information between different substation systems.

The IEC61850 standard is an object-oriented modeling approach to model the relay protection cloud test system, which not only allows for modification of the system’s functionality at any time, but also works well with other units and has good scalability. Therefore, information modeling is the core key point of the IEC61850 standard.

The primary task of modeling a relay protection device using the IEC 61850 standard protocol is to describe and define the various functions of relay protection as follows: 1)

Protection functions. Multi-segment grounding and inter-phase distance element protection, multi-segment zero sequence and inter-phase overcurrent element protection, longitudinal protection, circuit breaker failure protection, and oscillation blocking protection.

The relay protection configuration generally includes fast main protection in the direction of the longitudinal fault component and the zero sequence direction, fast I section protection consisting of distance elements, and backup protection consisting of three-section phase-to-phase and grounding distance and a number of sections of zero sequence direction overcurrent.

It is known from the IEC 61850 standard that the various functions in a line protection device can be defined in terms of logical nodes, with each function of the actual device defined as an instance of a functional logical node. The geometry of all logical nodes plus ancillary services constitute a logical device. A logical device can be defined as a protected line, and each line is an instance of a logical device. All functions and data of the line protection device can be provided and accessed via a server, which in the actual device is the communication structure of the physical protection device with the corresponding communication address. In the logical node there should also be defined an object model for the Generalized Data Variables (GGIO), which is the common data attribute geometry in the protection device. It defines data objects for controlling and accessing the configuration parameters of the protective device.The GGIOs are used for amplitude, saving and activating the configuration parameters of the protective device as well as for determining the operating paradigm of the protection. The values of phase voltages, phase-to-phase voltages, phase currents, voltages, and currents of each sequence obtained by the metrological value measurement unit can be defined in this object model for retrieval and use by the user.

1)

Database data modeling design

Based on the configuration of relay protection devices, it can be seen that the various types of grid data converge in the logical node layer, and the basis of database modeling is to determine the entity or object class of the database and its attributes. The database data of the relay protection cloud test system is divided into basic data, equipment parameter data, and operation paradigm data. The basic data mainly corresponds to grid area information and provincial grid information. The entities corresponding to the equipment parameter data are mainly plant stations, transformers, bus bars, lines, generators, series compensation capacitor reactors, and other grid components and equipment. The entities corresponding to the operation mode data mainly include the system operation mode and plant operation mode, and the plant operation mode includes the operation status (whether it is put into operation or not) of each grid component. Each of the above types of data needs to be indexed in a suitable way to help users understand the changes in relay protection data and provide data support for the realization of relay protection.

The fixed value inspection module mainly realizes the summoning and checking of the fixed values of the protection devices in the field of the sub-station, including the summoning and checking of the fixed values of the protection devices in the operation area and the summoning of the fixed values in the non-operation area. At the same time, in special mode, some protection devices need to execute temporary fixed values, and thus two or more sets of fixed values will be stored in the protection devices in different zones. If the fixed value of different fixed value area is mixed, it will cause the protection device to misoperate or refuse to operate and other serious consequences, so the summoning of the current fixed value area number of the protection device is also a basic function that the fixed value inspection module should be able to realize. Fixed-value inspection module work, the first need to detect the fixed value of the protection device and the system stored in the standard library of the fixed value of the same, when the detection of the fixed value of the content of the difference, or with a set of protection device information transmission is not smooth, then issued an alarm to remind the staff to check.

Fixed-value inspection module work that finds on-site fixed-value anomalies should be issued in a timely manner to remind the staff to deal with the alarm in a timely manner. The realization of the alarm function requires the use of a fixed-value inspection alarm module. When the alarm information is generated, the fixed value inspection alarm module will transmit the alarm information to the corresponding workstation to remind the relevant staff to check and deal with it promptly. Reminders generally use sound and light alarm modes. Alarm information can be used in the form of a light sign, so that the alarm is more direct and eye-catching. The first level of information is displayed as the specific name of the substation, the second level of information is displayed as the specific interval of the substation’s protective devices, and the third level of information is displayed as the specific list of entries. Through the three-level information display, maintenance personnel can locate specific protective devices, which helps them deal with them timely.

Relay protection sensitivity requirements, that is, in the protection range can be sensitive and rapid response to faults or abnormal operating conditions, and make the corresponding action, the sensitivity of the advantages and disadvantages of the sensitivity coefficient is measured by the size of the sensitivity coefficient. 1)

Zero sequence current protection I paragraph

Consider the maximum zero sequence current 3I_{0, max} to avoid the end of the protected line grounded short-circuit to verify the requirements: (1)Iset13I0,max≥1.3$$\frac{{I_{set}^1}}{{3{I_{0,\max }}}} \geq 1.3$$

Where IIset$${I^I}_{set}$$ is the calibration current to be calibrated. 2)

Zero sequence current protection II

Consider the minimum zero sequence current 3I0,min$$3{I_{0,\min }}$$ to be calibrated when escaping the ground fault at the end of the protected line, i.e.: (2)3II0,minIIIset≥1.3~1.5$$\frac{{3I{I_{0,\min }}}}{{{I^{II}}_{set}}} \geq 1.3\sim 1.5$$

When used as a remote backup protection, it is calibrated to the minimum zero sequence current at the time of a ground fault at the end of an adjacent line, i.e: (4)3Inext,−,minIIIIIset≥1.2$$\frac{{3{I_{next, - ,\min }}}}{{{I^{III}}{I_{set}}}} \geq 1.2$$

The selectivity of relay protection requires that the fault be removed in the first instance by the circuit breaker closest to the power side of the point of fault. If the protection for this action refuses to act, then the higher level of protection on the power side should act.

Line protection selectivity of the calibration mainly depends on when the fault occurs in the protection line outside the measurement, the protection can reliably not action. When checking, it is mainly to see whether there is selectivity between the protection delay section and the neighboring sections of protection. Zero sequence current protection selective checking of the basic rules are: to be checked according to the time of the protection action, to find out the neighboring and the protection of the action time is most similar to the two segments, set as M-1 paragraph and M paragraph, and its action time is between the two. If the full length of the range to be protected by the check is less than the range of the protection M-1 segment, the system selectivity requirements are met. Otherwise, it is not selective with the neighboring protection M segment.

Figure 2 shows a schematic diagram of protection coordination, A protection II segment action range is smaller than the B protection M-1 segment action range, when the fault occurs in the range of 0% ~ a%, B protection M-1 segment has a quick-acting first to excise the fault, and the protection A will not occur in the false action. If the fault occurs in the range of a%~100%, which is beyond the range of A protection, A does not act, i.e. the selectivity requirement is satisfied.

Figure 2.

A schematic of protection

Figure 3 shows another schematic diagram of protection coordination, A protection II section action range exceeds the B protection M-1 section action range, if the fault occurs in the a%~b% section, due to the A protection II section action time is smaller than the B protection M section action time, so the A protection may be the first to take action to remove the fault first or the two are removed at the same time. At this time, the A protection will be faulty, which does not meet the system selectivity requirements.

Figure 3.

Another schematic of protection

1)

Grounding distance I checking

Grounding distance I protection in the protection line range should be reliable when the end of the fault does not act. Grounding distance I protection is quick-acting protection, no delay instantaneous action. In order to ensure that the end of the scope of protection (adjacent line outlet) fault occurs, the protective device of the grounding distance I section does not operate, must be set in the action impedance setting, so that the scope of protection does not reach the full length of the line. The calibration formula is: (5)K=Z0IZI$$K = \frac{{{Z_{0I}}}}{{{Z_I}}}$$

Where, Z_0I is the grounding distance protection I-section setting value, Z₁ is the positive sequence impedance of the entire length of the line protected by this section. K should be a number less than 1. 2)

Grounding distance II section calibration

Grounding distance II protection belongs to the fixed-time protection, and grounding distance I protection together constitute the main protection of the distance protection, when the line end of the ground fault occurs when it must have enough sensitivity to cut off the fault. That is, it should satisfy: (6)K=Z0IIZI$$K = \frac{{{Z_{0II}}}}{{{Z_I}}}$$

In the formula, Z_0II is the setting value of grounding distance protection II, and Z₁ is the positive sequence impedance of the whole length of the line protected by this section. 3)

Grounding distance III section calibration

Sensitivity, phase distance III protection can not only be used as the line I, II section of the protection of the near backup protection, but also as a neighboring lower equipment protection of the far backup, which must meet a certain degree of sensitivity. The far backup protection calibration formula is: (7)K=Zd23Zt+Kb,maxZnext$$K = \frac{{{Z_{d23}}}}{{{Z_t} + {K_{b,\max }}{Z_{next}}}}$$

Where Z₁ is the positive sequence impedance of the line, Z_next is the impedance of the neighboring equipment, and K_{b, max} is the maximum value of the branching factor. K is greater than or equal to 1.2.

Near backup protection calibration formula is: (8)K=Zd23Zl$$K = \frac{{{Z_{d23}}}}{{{Z_l}}}$$

Where Z_l is the positive sequence impedance of this line. K is greater than or equal to 1.5.

Selectivity does not exceed the range of the next level of protection that cooperates in time.

In order to realize the fast warning of smart grid relay protection fixed value for relay protection cloud testing system, this paper proposes the fast warning method of relay protection fixed value based on radial basis network (RBF) [31].

The relay protection fixed value that does not satisfy the sensitivity as well as selectivity requirements is taken as the input to the radial basis neuron network, which is represented by the input vector Xm=(x1,x2,…,xm)$${X_m} = \left( {{x_1},{x_2}, \ldots ,{x_m}} \right)$$, the implicit layer response vector is represented by Gj=(g1,g2,…,gj)$${G_j} = \left( {{g_1},{g_2}, \ldots ,{g_j}} \right)$$, and the smart grid relay protection fixed value judgment result is represented by a.

Assuming that the neuron network weight is W, the weight is weighted, and M and the threshold d are added to the network, and the result obtained after processing with the transfer function is the neuron network output a, which is calculated by the following formula: (9)a=f(MW+d)$$a = f\left( {MW + d} \right)$$

Radial basis neuron network learning consists of Gaussian function center value c_j and output layer weights obtained by least squares algorithm w_jt. For training samples, the radial basis function center is explicitly expected to output y_t after known, and the output weights are obtained by least squares method w˜jt$${\tilde w_{jt}}$$. 1)

Adjust the hidden layer neuron center and width radial basis neuron network error function formula is as follows: (10)E=12∑n=1NEn$$E = \frac{1}{2}\sum\limits_{n = 1}^N {{E^n}}$$

where N and Eⁿ denote the number of samples and the error of entering the nrd sample, respectively, and Eⁿ is calculated as follows: (11)En=∑k=1s(tkn−ykn)2$${E^n} = \sum\limits_{k = 1}^s {{{\left( {t_k^n - y_k^n} \right)}^2}}$$

where tkn$$t_k^n$$ and ykn$$y_k^n$$ denote the actual output and the desired output, respectively, where n = 1, 2, ⋯, N.

The parameter correction is proportional to the negative gradient to minimize the error function, then the parameter correction ΔA_j and the negative gradient Δδ_j are calculated as: (12)ΔAj=−μ1∂E∂Aj$$\Delta {A_j} = - {\mu _1}\frac{{\partial E}}{{\partial {A_j}}}$$ (13)Δδj=−μ2∂E∂σj$$\Delta {\delta _j} = - {\mu _2}\frac{{\partial E}}{{\partial {\sigma _j}}}$$

Where, μ₁ and μ₂ denote the parameter correction amount and the correction coefficient of negative gradient, respectively.

After inputting all the samples, the parameter equations are adjusted by iterative method as follows: (14)zj(m+1)=zj(m)+ΔAj$${z_j}(m + 1) = {z_j}(m) + \Delta {A_j}$$ (15)kj(m+1)=kj(m)+Δδj$${k_j}(m + 1) = {k_j}(m) + \Delta {\delta _j}$$

Where z_j and k_j denote the learning rate of the neuron center of the hidden layer and the width change rate, respectively; m denotes the number of iterations.

When the width change rate is higher than the learning rate of the neuron center of the hidden layer, the radial basis neural network generalization performance is better. 2)

Determine the radial basis neural network weights

Let Q ∈ Z^{r × N}, P ∈ Z^{u × N} and Y ∈ Z^{s × N} denote the radial basis neural network input matrix, the hidden layer input matrix and the hidden layer output matrix, respectively, and W ∈ Z^{s × u} denote the radial basis neuron network output layer weights, which can be obtained as follows: (16)Y=Q×PT$$Y = \frac{{Q \times P}}{T}$$

In the formula, T=(t1,t2,⋯,ts)T∈Zs×N$$T = {\left( {{t_1},{t_2}, \cdots ,{t_s}} \right)^T} \in {Z^{s \times N}}$$ represents the target output of the radial-based neuron network, and the linear least squares method is used to reduce the error between the actual output and the target output, and when the error value is the lowest, the output of the smart grid relay protection fixed-value judgment result is output, and when the value has not exceeded the critical fixed-value, the early warning instruction is not output to avoid the occurrence of false alarms. When the value exceeds the critical value, the warning command is output.

The RBF-based relay protection value calibration process proposed in this paper is shown in Fig. 4, which mainly contains two phases of offline training and online calibration, aiming at realizing the intelligent and fast warning of relay protection value calibration, and fully ensuring the stable operation and scheduling of the power network.

Figure 4.

Based on RBF’s relay protection value check process

Offline training consists of four parts: 1)

Extreme operation mode data generation. Generally, through offline simulation, a large number of extreme operation mode data is generated for a specific network to constitute a sample set. It is also possible to collect the extreme modes of operation that have occurred in the past period of time, but the extreme modes of operation collected in this way are often not comprehensive enough, and need to be supplemented by offline simulation calculations.

The online calibration process is broken down into two parts: input data construction and fixed-value calibration. When a new extreme operation mode occurs, the mode fixed value matrix is first generated according to the current operation mode information and the protection fixed values information. Since only part of the lines are put into and taken out of service under extreme operation mode, the mode fixed value matrix can be obtained by modifying the mode fixed value matrix under the original operation mode. Then the matrix is inputted into the trained RBF model to obtain the results of relay protection value checking.

In the data transmission process of relay protection cloud testing system, this paper designs a semantic indexing method based on quadtree algorithm, which aims to better realize the efficient transmission of relay protection related data. In order to verify the effectiveness of the application of this method in the relay protection cloud testing system, this paper quantitatively analyzes the four indexes, namely, query response time, query comparison times, the amount of data transferred by the server and the number of server connections. The indexing model without cache, semantic cache model (SCM) and quad-tree indexing methods are selected for comparison, and the data object query performance of different methods is obtained as shown in Fig. 5, where Fig. 5(a)~(b) shows the comparison results of query response time and query comparison times, respectively.

Figure 5.

Data object query performance

Generally speaking, the query and the processing process with larger data volume takes more time. In Fig. 5(a), it can be seen that the slope of the response time of the no-cache model is larger than that of the semantic cache model, and the response time of the quadtree-based indexing is less than that of the traditional semantic cache model, and the average response time of the indexing query based on the quadtree algorithm is only 3.17 s in the process of 30 queries.This is due to the fact that both the query time and the volume of the data returned from the server are more important in the design of the indexing based on the quadtree algorithm are less than the traditional semantic caching model. Moreover, in Fig. 5(b), the number of query comparisons is proportional to the number of queries for the uncached model, while the slope of the curve for the cached model is very small. More importantly, the number of query comparisons for the indexing method based on the quadtree algorithm is less than that of the traditional semantic caching model. This is due to the fact that location-dependent queries have a high level of similarity, and therefore unnecessary comparisons can be eliminated by using the quadtree algorithm. The results above show that the indexing method based on the quadtree algorithm can better utilize the data cache of the relay protection cloud testing system and improve the efficiency of data queries.

In addition, this paper further compares the amount of data transferred to the server and the number of connections between the no-cache and the methods in this paper, and Fig. 6 shows the comparison results of the different methods, where Fig. 6(a)~(b) shows the comparison results of the amount of data transferred to the server and the number of connections, respectively. From the figure, we can see that in the absence of caching model, the amount of data transmitted over the network is very large, almost proportional to the number of queries. With the same amount of data transferred in the semantic caching model, the slope is significantly lower. This is due to the fact that all or part of the desired query results are contained in the semantic cache. Additionally, the number of server connections is inversely proportional to the number of queries in the no-cache model. This is due to the fact that in the no-cache scenario, the server is accessed for each query request. Whereas, in the quadtree-based index design, the number of service connections is less, this is because the required query results are contained in the data cache in whole or in part. In summary, the quadtree index design for query processing can fully utilize cache resources, reduce network load, and also support frequent disconnections occurring during the query.

Figure 6.

Comparison results of different methods

In the data storage experiments, this paper selects 5 hours of monitoring data from a substation for testing and conducts parallel writing experiments on the above 5 hours of data respectively. In order to verify the effectiveness of this paper’s indexing algorithm based on the quadtree algorithm, the five hours of grid monitoring data were constructed in parallel, respectively, and the traditional parallel writing algorithm as well as the parallel resampling algorithm were selected as comparisons, and Fig. 7 shows the results of the comparison of different methods.

Figure 7.

Comparison results of different methods

As can be seen from the figure, the indexing method based on the quadtree algorithm in this paper takes only 93.54ms on average for parallel writing of grid monitoring data into the database, which is 65.41% and 33.97% lower than the traditional algorithm and parallel resampling algorithm, respectively, and the time required is much smaller than the two comparative algorithms. When the amount of data gradually increased, the traditional algorithm to write the database time increase is larger and this paper based on the quadtree indexing method of the time required although increased but the overall rise is smaller, which can be seen in this paper proposed method effectively reduces the reduction of the time required to write the grid monitoring data into the database, can be higher storage efficiency will be related to the logical nodes of the data written into the database, for the user’s timely application to provide efficient and fast data service.

In addition, this paper also analyzes the acceleration ratio for different grid monitoring data, from the analysis results, when the experimental grid data is less the acceleration ratio effect of the parallel algorithm is not very ideal, the data import time is almost unchanged, but as the data increases the acceleration ratio gradually increases, the time of data writing decreases, and the indexing efficiency based on the quadtree algorithm increases. This is mainly because, when the amount of data is small, the number of regions to be divided when storing on HBase is small, and the data exists only in a fixed number of nodes, which leads to little change in the data import time. However, with the increase in the amount of data HBase need to increase the number of region, then different logical nodes can be written to HBase, the algorithm accelerates the effect is more obvious.

Taking a regional localized power grid as an experimental object, the system of this paper is applied within the regional localized power grid to verify the automatic alarm effect of the system of this paper for electrical relay protection equipment. The topology of the regional localized power grid is shown in Figure 8. The topology of the regional local power grid contains 7 substations, 10 relay protection devices, 1 power plant, 21 lines, and the three protection values of each relay protection device are shown in Table 1.

Figure 8.

Local grid topology diagram

Using the system in this paper in the EMS and offline calibration calculation system, according to the relay protection equipment operating state information, generate relay protection equipment operating state information samples, to R1 to R5 of the first five relay protection equipment as an example, using the system in this paper to generate the five relay protection equipment operating state information samples distribution is shown in Figure 9. As can be seen from the figure, the system can effectively generate relay protection equipment operating state information according to the operating state information of relay protection equipment to generate relay protection equipment operating state information samples, and generated between the five relay protection equipment operating state information, there is no confusion, can be clearly distinguished between the five relay protection equipment operating state information for the subsequent automatic warning of relay protection equipment to provide more convenient data support. Experiments have proved that the system in this paper can effectively generate operational state information samples for relay protection equipment.

Figure 9.

Operation status information sample distribution

Using the system in this paper, the local relay protection equipment in the region is calibrated and automatically alarms according to the results of the calibration. The results of the calibration of the relay protection equipment are shown in Table 2, and Figure 10 shows the results of the automatic alarms, with the red markers in the figure as the alarm relay protection equipment modules.

Figure 10.

Automatic alarm result of the fixed value check

According to the intelligent calibration results, it can be seen that the relay protection cloud testing system designed in this paper can effectively complete the calibration of the fixed value of the relay protection equipment and carry out automatic alarms according to the calibration results. In the local power grid in the region, the relay protection equipment R5, R6, and R7 have faults, which are prone to mis-actuation or refusal to act, affecting the stability of power grid operation. The relay protection cloud testing system designed in this paper relies on the intelligent warning model of relay protection fixed value calibration designed by RBF network, which can send the alarm information to maintenance personnel in time. The above results show that the relay protection cloud test system designed in this paper can effectively verify the fixed value of relay protection equipment, realize automatic alarm of relay protection equipment, and provide support for ensuring stable operation of the power grid and rapid location and identification of abnormal points.

The application scope of the relay protection cloud testing system designed in this paper is oriented to the industrial Internet of Things, and in order to further verify the effectiveness of the system, this paper aims to apply the system to the provincial relay protection integration. The system is spliced with power grid data from different regions to verify cross-region relay protection values. Based on this, this paper selects the grid data of a province’s A and B areas, and splices them with the relay protection cloud test system designed in this paper, and the fixed-value calibration model before and after splicing is shown in Fig. 11, of which Fig. 11(a)~(c) are the model of the local regulator and the provincial regulator before splicing, as well as the complete model after splicing, respectively.

Figure 11.

The fixed value check model before and after the splicing

In order to objectively analyze the difference between the results of relay protection value verification before and after data splicing, the calculation results of single-phase and three-phase short-circuit faults occurring in the system at three typical locations in the large and small ways are selected for comparison. Take the 220kV bus short-circuit faults of BGS, JJL, and DBJ as an example, in which BGS and JJL become the boundaries of provincial and local data splicing, while DBJ becomes the boundary of non-splicing. The calculation results are shown in Table 3, where SPF and TPF are single-phase faults and three-phase faults respectively.

The closer the nodes are to the data splicing boundary, the more obvious the difference between the fault calculation results before and after splicing is, with the maximum difference reaching 5.31% (primary current of about 1274A), but overall within the allowable range of engineering calibration. This confirms the rationale for ignoring the 110kV system in the current calculation of the provincial regulator’s relay protection calibration. Under the situation of rapid development of extra-high voltage grid, integration of province, prefecture and county, and even integration of state grid and province, the relay protection cloud testing system based on IEC61850 standard proposed in this paper has many advantages, such as simple principle, high reliability, low requirements for grid topology modeling, and convenient engineering realization, etc., and has a good prospect for popularization and application. The method proposed in this paper is based on the general principle of multilevel dispatching and hierarchical management. Regardless of the type of division method used for the interface between the upper and lower levels of the dispatching units, the boundaries of the grid in a certain region are always solidified and predictable. Therefore, the method proposed in this paper is effective regardless of the number of levels of vertical dispatching used or the number of levels of dispatching used to carry out integrated calibration calculations.