Simulation difference region calibration method based on power flow of power system and electronic equipment

Abstract

The invention discloses a simulation difference region calibration method based on power flow of a power system and electronic equipment, and belongs to the technical field of simulation error identification of the power system. According to the method, a directed weighting matrix is constructed on the basis of constructing an adjacent matrix on the basis of a power grid topological structure and node power flow information, and power grid region division is realized through an MCL algorithm in combination with physical connection constraints. For isolated PQ or PV nodes in a division result, a coupling strength-based re-attribution strategy is introduced. On the basis of region division, dynamic simulation tracks and actual measurement data of nodes in all regions are collected, a root-mean-square error and a Pearson's correlation coefficient are constructed to serve as error evaluation indexes, and a difference region with remarkable errors is recognized according to the error indexes. The method provides an accurate basis for power grid parameter correction and simulation model optimization, has physical rationality, direction sensitivity and quantitative evaluation capability, and is suitable for simulation error diagnosis and regional accurate treatment of a complex power system.

Description

Technical Field

[0001] The present invention belongs to the technical field of power system simulation error identification, and in particular relates to a simulation error area calibration method based on power system flow and electronic equipment. Background Art

[0002] In the field of power system simulation error identification, existing technologies mainly focus on regional division methods based on topological structures and node relationships. For example, the minimum spanning tree (MST) method uses grid disturbance information to construct a tree diagram with the lowest connection cost, thereby revealing the most physically connected parts of the grid. This method achieves the grouping of grid nodes by dividing the connection strength, thereby locating possible error propagation paths. However, the MST method is highly dependent on disturbance information, and the disturbance data in the actual power grid is often insufficient, which makes the method limited in describing the complex coupling relationship of data transmission in the power grid. In addition, since the type and amplitude of the disturbance information have a great influence on the results of the method, the grid connection strength it describes may not be sufficient to provide accurate error region division.

[0003] Another approach divides the grid into multiple subregions based on its geometric topology and identifies the source of errors through step-by-step simulation verification. These methods primarily partition regions based on the distance between nodes and the number of adjacent nodes, but often overlook the importance of dynamic grid information. Dividing the grid solely from a static geometric perspective can overlook the dynamic changes in the grid's real-time operating state, thereby affecting the accuracy of error identification. Furthermore, these methods often rely on manually set parameters, limiting their adaptability and generalization capabilities, and can lead to computational inefficiencies when applied to large-scale power systems.

[0004] Therefore, although existing technologies can complete the preliminary identification of error areas to a certain extent, they fail to fully utilize the actual operation data of the power grid, especially the dynamic characteristics of the power grid, and have limitations in the accuracy of regional division and computational efficiency.

[0005] A new technical solution is urgently needed in the existing technology to solve this problem. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a simulation error area calibration method and electronic equipment based on power system flow, which is used to solve the problems in the existing technology that error identification relies too much on disturbance information, or mainly relies on the static topology structure of the network, ignoring the dynamic coupling and directionality of flow transmission, and these methods rely on manually set parameters during implementation, resulting in insufficient error positioning accuracy and low computational efficiency.

[0007] The technical solution adopted by the present invention is to provide a method for calibrating the simulation difference area based on the power system flow, comprising the following steps:

[0008] (1) Based on the physical connection information and operation data of the power grid, the entire network is divided into several sub-areas. The power flow data of each node is collected from the power system and normalized. The positive and negative information are retained during normalization, and the global maximum absolute value is used as the normalization factor.

[0009] (2) A directed weighted matrix is ​​constructed based on the physical connection information and flow data of the power grid. The coupling relationship between nodes is reflected by the flow direction and intensity. The symmetrical physical connection matrix formed by the actual connection is combined with the actual connection to establish edges only between the nodes that are actually connected.

[0010] (3) The MCL algorithm applicable to directed graphs is used to automatically partition the directed weighted matrix of the power grid into communities, identifying sub-regions with strong internal coupling and weak external connections, which are used to reflect the actual structure of power flow transmission and error propagation in the power grid;

[0011] (4) When the isolated node is a PQ or PV node after community division, the isolated node is merged into the adjacent community based on the weight information in the directed weighted matrix. The adjacent community satisfies the maximum sum of the connection weights between the isolated node and all nodes in the adjacent community.

[0012] (5) In each region, the difference between the simulation data and the measured data is quantitatively evaluated by calculating the root mean square error and the correlation coefficient. The error indicators of each node are combined with the weighted average method to identify the area with larger error, that is, the difference area.

[0013] The power flow data of each node in step (1) includes the transmission direction and intensity of active and reactive power. The normalization formula of active data is:

[0014]

[0015] Among them, P ij is the active power from node i to node j, P norm,ij is the normalized active power value, P ref is the reference value of line active power;

[0016] The normalization formula for reactive data is:

[0017]

[0018] Among them, Q ij is the reactive power from node i to node j, Q norm,ij is the normalized reactive power value, Q ref It is the reference value of line reactive power.

[0019] The directed weighted matrix in step (2) is constructed based on Euclidean distance by fusing normalized active and reactive power flow data. The flow direction and intensity are represented by directed weights. For the direction from node i to node j, when there is a positive active flow from node i to node j, that is, P norm,ij > 0, the weight of this direction is expressed as:

[0020]

[0021] When there is no positive active power flow from node i to node j, that is, P ij <0, then check the reverse active power flow P ji , if there is a reverse active power flow, the weight of this direction is expressed as:

[0022]

[0023] The directed weighted matrix W is an asymmetric matrix, which is constructed as follows:

[0024]

[0025] Among them, w ij It reflects the power flow transmission intensity from node i to node j, i∈V, j∈V, where V={1,2,3,...,n}, and n represents the total number of nodes in the system.

[0026] The optimization objectives of the MCL algorithm in step (3) include:

[0027] 1) Maximize the sum of edge weights between nodes within the community;

[0028] 2) Minimize the sum of edge weights across communities.

[0029] The MCL algorithm takes the constructed directed weighted matrix as input, and through the "expansion" and "dilation" operations of random walks, it classifies the strongly coupled nodes in the graph into the same "attractor" to achieve region partitioning.

[0030] The directed weighted matrix is ​​converted into a row-normalized transfer matrix, and self-loops are added on the diagonal to ensure that each node is at least connected to itself, which is used to simulate the random walk process of the MCL algorithm. The addition of the self-loop and the normalization process are expressed as:

[0031] M=normalize(W+I)

[0032] Where I is the identity matrix, normalize(·) means dividing each row by the row sum so that the sum of each row is 1;

[0033] The "expand" operation is expressed as:

[0034]

[0035] Where e is the expansion index, usually e = 2, r is the expansion index, r is greater than 1;

[0036] After multiple iterations, the final matrix has a block structure, and the non-zero values ​​within the block correspond to the same community. That is, while maintaining the physical connection constraint, the random walk stability is extracted through the directed flow data, thereby automatically identifying areas with tight internal coupling and weak cross-region connectivity.

[0037] The isolated node merging criterion in step (4) is:

[0038]

[0039] Among them, C * For nearby communities, C * Satisfy the node g and its * The sum of the connection weights between all nodes in the network is the largest.

[0040] In step (5), in each region, for each node n, the root mean square error between the hybrid dynamic simulation data and the measured data is expressed as:

[0041]

[0042] Where N is the number of sampling points, and represent the simulated and measured values ​​of node n at time t respectively;

[0043] In order to integrate the errors of all nodes in the region, the node weight w is given n , according to the node importance or data quality setting, the regional root mean square error is defined as the weighted average:

[0044]

[0045] The Pearson correlation coefficient of each node n is defined in step (5) as:

[0046]

[0047] in, and are the means of the simulation and measured data of node n, respectively;

[0048] The regional comprehensive correlation coefficient is obtained by weighted average:

[0049]

[0050] Select all variable trajectories in the same area and calculate two error indicators respectively to quantitatively characterize the size of the error. Then, based on the two indicators, calculate the average value of the error values ​​of all trajectories in the area as the overall error evaluation of the area.

[0051] By calculating the above two error indicators for each region and setting a predefined threshold, when RMSE region Exceeds a predefined threshold or r region When the value is lower than the predefined threshold, the regional simulation accuracy can be quantitatively evaluated and poor areas with large errors can be identified.

[0052] An electronic device includes a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to execute any one of the methods described above.

[0053] A computer-readable storage medium stores a computer program, wherein the computer program implements any of the above methods when executed by a computer.

[0054] Through the above design scheme, the present invention can bring the following beneficial effects:

[0055] 1. Preservation of flow direction and intensity information: A directed weighted matrix is ​​constructed using normalized active and reactive flow data, and the Euclidean modulus of the flow vector is directly calculated. This ensures that the edge weight not only reflects the transmission intensity but also preserves the flow direction information, thus better fitting the actual power transmission process.

[0056] 2. Automatic community detection: The MCL algorithm is used to divide the directed graph into communities, and the expansion and dilation operations of random walks are used to automatically identify areas with strong internal connectivity and weak external connectivity. This avoids the need to manually set the number of clusters or thresholds, and improves the adaptability and robustness of the division.

[0057] 3. Clear quantitative error metrics: By calculating the root mean square error (RMSE) and correlation coefficient for each region, we quantitatively evaluated the differences between the simulated and measured data. In the experiment, the simulation error in Region 3 was significantly higher than in other regions. The RMSE and correlation coefficient for active power in this region were 0.39 and 0.91, respectively, and the RMSE and correlation coefficient for reactive power were 0.27 and 0.96, respectively. The error locations were included in the identified communities, demonstrating the effectiveness of the method.

[0058] 4. Real-time and automated method: The community detection algorithm does not require a preset number of clusters and a large number of manual parameters. It can automatically identify regional divisions under the global graph structure and achieve rapid dynamic calibration in combination with error indicators. Compared with traditional methods, it has higher computational efficiency and real-time response capabilities.

[0059] 5. This method fully retains the direction and intensity information of the tide in data utilization, and at the same time uses physical connection constraints to ensure the practical significance of regional division. Combined with automatic community detection and quantitative error indicators, it achieves efficient and accurate identification of error-causing areas, providing advanced and reliable technical support for the subsequent location of error sources and parameter correction. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 This is a schematic diagram of a 5-machine 14-node system of the present invention's simulation-based differential area calibration method and electronic equipment;

[0061] Figure 2 This is a schematic diagram of the simulation-induced difference area calibration method based on power system flow and the partitioning result of 5 machines and 14 nodes of the electronic equipment according to the present invention;

[0062] Figure 3 This is a schematic diagram of the overall process of the simulation error area calibration method based on power system flow and the electronic equipment of the present invention;

[0063] Figure 4 Schematic diagram of the device of the simulation difference area calibration method based on power system flow and electronic equipment of the present invention. DETAILED DESCRIPTION

[0064] The present invention will be further described below with reference to the accompanying drawings and specific embodiments:

[0065] 1. Construction of adjacency matrix based on node information

[0066] First, a one-way adjacency matrix A is constructed based on the physical connection relationship of the power grid, where for any nodes i and j:

[0067]

[0068] Matrix A is a half-width matrix, and its form is:

[0069]

[0070] Since the physical connection of the power grid is non-directional, we further construct a symmetric physical connection matrix, namely the adjacency matrix:

[0071] A sym =sgn(A+A T ) (3)

[0072] Among them, sgn(x) means that if x>0, the output is 1, otherwise the output is 0. sym Indicates whether there is a physical connection between any two nodes in the power grid.

[0073] After collecting the power flow data (active power P(t) and reactive power Q(t)) from each node, the power flow data is first normalized to capture the grid's operating status. During normalization, positive and negative information is retained, and the global maximum absolute value is used as the normalization factor. The normalization formula for active power data is:

[0074]

[0075] Reactive power data, its normalization formula is:

[0076]

[0077] 2. Construction of directed weighted matrix W

[0078] Next, in order to construct a directed weight that can reflect the direction and intensity of power transmission, we add a sym The combined strength of power flows in both directions is calculated when (i, j) = 1. Specifically, by fusing normalized active and reactive power flow data, a directed weighted matrix is ​​constructed based on Euclidean distance. Euclidean distance combines both active and reactive power information, reflecting the combined effects of power transmission and voltage support. Reactive power flow information is implicitly incorporated into the weight calculation using the normalized Euclidean distance, providing a more comprehensive reflection of the overall power flow status.

[0079] For the direction from node i to node j, if there is a positive active power flow from node i to node j, that is, P norm,ij > 0 (indicates that there is actual forward active transmission), the weight of this direction is defined as:

[0080]

[0081] If there is no positive active power flow, that is, P ij <0, then check the reverse active power flow P ji If a reverse power flow exists, the weight of that direction is defined as:

[0082]

[0083] This matrix W is directed and only retains weights on node pairs where there is actual power flow. Edges in the opposite direction where there is no power flow are ignored (i.e., weights are 0). This ensures that the direction of power flow is preserved, while the weight of each edge reflects the actual strength of the power flow.

[0084] The constructed directed weighted matrix W is an asymmetric matrix, as shown below:

[0085]

[0086] Among them, wij It reflects the power flow transmission intensity from node i to node j, i∈V, j∈V, where V={1,2,3,...,n}, and n represents the total number of nodes in the system.

[0087] 3. Region Partitioning Based on Directed Weighted Matrix

[0088] Using the constructed directed weighted matrix W and physical connectivity constraints, we obtain a directed graph that truly reflects the operational state of the power grid. The goal of regionalization is to divide the graph into subgraphs (communities) with strong internal connectivity and weak external connections. Therefore, the MCL method can be used for calculation. This goal implies the following optimization objectives:

[0089] 1) Maximize the sum of edge weights between nodes within the community;

[0090] 2) Minimize the sum of edge weights across communities.

[0091] The basic idea of ​​this method is to take the constructed directed weighted matrix W as input, and use the "expansion" and "dilation" operations of random walks to classify the strongly coupled nodes in the graph into the same "attractor", thereby achieving region division.

[0092] In order for the MCL algorithm to simulate the random walk process, W needs to be converted into a row-normalized transfer matrix. Self-loops are usually added to the diagonal to ensure that each node is connected to at least itself. The process of adding self-loops and normalization can be expressed as:

[0093] M=normalize(W+I) (9)

[0094] Where I is the identity matrix, and normalize(·) means dividing each row by the row sum so that the sum of each row is 1. Next, the expansion operation is performed:

[0095]

[0096] The parameter e is the expansion exponent (typically e = 2), and r is the dilation exponent (greater than 1). After multiple iterations, the final matrix has a block-like structure, with non-zero values ​​within a block corresponding to the same community. This means that while maintaining physical connectivity, random walk stability is extracted from directed flow data, automatically identifying regions with tight internal coupling and weak cross-regional connectivity.

[0097] The system is divided into multiple sub-areas through community division, and the division results are further refined to ensure that each sub-area has practical significance in terms of physics and operation. Specifically, after community detection, the system obtains the community set C = {C1, C2, ..., C m}, where C k For a sub-region.

[0098] For a sub-region C g If |C g | = 1, the node is considered an isolated node g. If node g is a PQ or PV node (i.e., a key location for power injection or consumption), it is considered that the node cannot independently reflect the coupling characteristics and grid operating conditions of the entire region. It is necessary to further analyze the set of nodes directly connected to the physical neighborhood N(g) in the physical wiring diagram, so as to merge (re-attribute) the node with the neighborhood.

[0099] From the directed weighted matrix, we can know the power flow coupling weight w between nodes ij , define the merging criteria of node g:

[0100]

[0101] That is, merge the isolated node g into the community C * In, C * Satisfy the node g and its * The sum of the connection weights between all nodes in the grid is maximized. This ensures that the isolated PQ or PV node, as a power source or consumer, can reflect the energy flow trend in the actual grid and form a whole with the areas with strong coupling in its physical neighborhood.

[0102] 4. Construction of error indicators and regional error assessment

[0103] In each region, for each node n, the root mean square error between the hybrid dynamic simulation data and the measured data is defined as:

[0104]

[0105] This indicator can refine the details of the error from the perspective of time domain. Where N is the number of sampling points, and Represent the simulated and measured values ​​of node n at time t respectively. In order to integrate the errors of all nodes in the region, the node weight w is given n (can be set according to node importance or data quality), the regional root mean square error can be defined as a weighted average:

[0106]

[0107] For the overall similarity of trajectories, the correlation coefficient method is used, and the Pearson correlation coefficient of each node n is defined as:

[0108]

[0109] in and are the means of the simulated and measured data of node n, respectively. Similarly, the regional comprehensive correlation coefficient can be obtained by weighted averaging:

[0110]

[0111] All variable trajectories in the same area are selected to calculate two error indicators respectively to quantitatively characterize the size of the error, and the average value of the error values ​​of all trajectories in the area is calculated based on the two indicators as the overall error evaluation of the area.

[0112] By calculating the above two error indicators for each region, the smaller the RMSE, the smaller the error, and the closer the correlation coefficient is to 1, the smaller the error. And set a predefined threshold, when RMSE region Exceeds a predefined threshold or r region When the value is lower than a predefined threshold, the regional simulation accuracy can be quantitatively evaluated, and the “poor area” with large errors can be identified.

[0113] 5. Case Verification

[0114] According to the IEEE standard 5-machine 14-node system, the case analysis and verification are carried out. Figure 1 In this system, due to the difference between the parameters set in the simulation and the actual parameter data of the system, there is an error between the variable trajectory obtained by simulation and the actual trajectory.

[0115] The data from the standard system is used as real data. A generator in the system is randomly selected and its parameters are randomly perturbed. After long-term operation, the actual parameters of the simulated system change, which can lead to a loss of simulation accuracy. For the simulation, a fourth-order model is used for the generator and a second-order model for the excitation system. A fault occurs at node 02, lasts 0.1 seconds, and has a sampling time of 5 seconds.

[0116] The original one-way adjacency matrix A:

[0117]

[0118] Directed weight matrix W:

[0119]

[0120] The community classification results detected based on the MCL algorithm are as follows Figure 2 shown.

[0121] The regional error evaluation based on the measurement trajectory of active power P and reactive power Q is shown in Table 1. The threshold of RMSE is set to 0.3, and the threshold of correlation coefficient is set to 0.95.

[0122] Table 1 Regional error evaluation based on measurement trajectory

[0123]

[0124] The results show that the error in community 3 is significantly higher than that in communities 1 and 2. Based on the defined thresholds, the RMSE of active power for nodes in community 3 is above the threshold of 0.3, and the correlation coefficient is below the threshold of 0.95. While the RMSE and correlation coefficient of reactive power do not exceed the threshold, the error is significantly greater than that in communities 1 and 2. The error in community 3 is the most significant, making it identifiable as a poorly performing area. The system randomly selected a generator on node 06, which was assigned to community 3, further validating the effectiveness of the method.

[0125] This method can identify the areas that need to be corrected, which will greatly reduce the cost required for the subsequent parameter calibration process.

[0126] like Figure 4 As shown, the present invention also provides an electronic device, including a memory and a processor, the memory is used to store computer programs, and the processor is used to run the computer programs so that the electronic device executes the above-mentioned simulation difference area calibration method based on power system flow.

[0127] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which implements the above-mentioned simulation error area calibration method based on power system flow when executed by a computer.

[0128] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0129] The implementation methods of the present invention are not limited to the above-mentioned embodiments. Any other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principles of the present invention should be considered as equivalent replacement methods and are included in the scope of protection of the present invention.

Claims

1. A method for calibrating the simulation-induced difference region based on power system flow, characterized by: The following steps are involved: (1) Based on the physical connection information and operation data of the power grid, the entire network is divided into several sub-areas. The power flow data of each node is collected from the power system and normalized. The positive and negative information are retained during normalization, and the global maximum absolute value is used as the normalization factor. (2) A directed weighted matrix is ​​constructed based on the physical connection information and flow data of the power grid. The coupling relationship between nodes is reflected by the flow direction and intensity. The symmetrical physical connection matrix formed by the actual connection is combined with the actual connection to establish edges only between the nodes that are actually connected. (3) The MCL algorithm applicable to directed graphs is used to automatically partition the directed weighted matrix of the power grid into communities, identifying sub-regions with strong internal coupling and weak external connections, which are used to reflect the actual structure of power flow transmission and error propagation in the power grid; (4) When the isolated node is a PQ or PV node after community division, the isolated node is merged into the adjacent community based on the weight information in the directed weighted matrix. The adjacent community satisfies the maximum sum of the connection weights between the isolated node and all nodes in the adjacent community. (5) In each region, the difference between the simulation data and the measured data is quantitatively evaluated by calculating the root mean square error and the correlation coefficient. The error indicators of each node are combined with the weighted average method to identify the area with larger error, that is, the difference area.

2. The method for calibrating the simulation-induced difference region based on power system power flow according to claim 1 is characterized in that: The power flow data of each node in step (1) includes the transmission direction and intensity of active and reactive power. The normalization formula of active data is: Among them, P ij is the active power from node i to node j, P norm,ij is the normalized active power value, P ref is the reference value of line active power; The normalization formula for reactive data is: Among them, Q ij is the reactive power from node i to node j, Q norm,ij is the normalized reactive power value, Q ref It is the reference value of line reactive power.

3. The method for calibrating the simulation-induced difference region based on power system power flow according to claim 2, characterized in that: The directed weighted matrix in step (2) is constructed based on Euclidean distance by fusing normalized active and reactive power flow data. The flow direction and intensity are represented by directed weights. For the direction from node i to node j, when there is a positive active flow from node i to node j, that is, P norm,ij > 0, the weight of this direction is expressed as: When there is no positive active power flow from node i to node j, that is, P ij <0, then check the reverse active power flow P ji , if there is a reverse active power flow, the weight of this direction is expressed as: The directed weighted matrix W is an asymmetric matrix, which is constructed as follows: Among them, w ij It reflects the power flow transmission intensity from node i to node j, i∈V, j∈V, where V={1,2,3,...,n}, and n represents the total number of nodes in the system.

4. The method for calibrating the simulation-induced difference region based on power system power flow according to claim 1, characterized in that: The optimization objectives of the MCL algorithm in step (3) include: 1) Maximize the sum of edge weights between nodes within the community; 2) Minimize the sum of edge weights across communities.

5. The method for calibrating the simulation-induced difference region based on power system power flow according to claim 4 is characterized in that: The MCL algorithm takes the constructed directed weighted matrix as input, and through the expansion and dilation operations of random walks, it classifies the strongly coupled nodes in the graph into the same attractor to achieve region partitioning.

6. The method for calibrating the simulation-induced difference region based on power system power flow according to claim 5, characterized in that: The directed weighted matrix is ​​converted into a row-normalized transfer matrix, and self-loops are added on the diagonal to ensure that each node is at least connected to itself, which is used to simulate the random walk process of the MCL algorithm. The addition of the self-loop and the normalization process are expressed as: M=normalize(W+I) Where I is the identity matrix, normalize(·) means dividing each row by the row sum so that the sum of each row is 1; The expansion operation is expressed as: Where e is the expansion index, usually e = 2, r is the expansion index, r is greater than 1; After multiple iterations, the final matrix has a block structure, and the non-zero values ​​within the block correspond to the same community. That is, while maintaining the physical connection constraint, the random walk stability is extracted through the directed flow data, thereby automatically identifying areas with tight internal coupling and weak cross-region connectivity.

7. The method for calibrating the simulation-induced difference region based on power system power flow according to claim 1, characterized in that: The isolated node merging criterion in step (4) is: Among them, C * For nearby communities, C * Satisfy the node g and its * The sum of the connection weights between all nodes in the network is the largest.

8. The method for calibrating the simulation-induced difference region based on power system power flow according to claim 1, characterized in that: In step (5), in each region, for each node n, the root mean square error between the hybrid dynamic simulation data and the measured data is expressed as: Where N is the number of sampling points, and represent the simulated and measured values ​​of node n at time t respectively; In order to integrate the errors of all nodes in the region, the node weight w is given n , according to the node importance or data quality setting, the regional root mean square error is defined as the weighted average:

9. The method for calibrating the simulation-induced difference region based on power system power flow according to claim 1, characterized in that: The Pearson correlation coefficient of each node n is defined in step (5) as: in, and are the means of the simulation and measured data of node n, respectively; The regional comprehensive correlation coefficient is obtained by weighted average: Select all variable trajectories in the same area and calculate two error indicators respectively to quantitatively characterize the size of the error. Then, based on the two indicators, calculate the average value of the error values ​​of all trajectories in the area as the overall error evaluation of the area. By calculating the above two error indicators for each region and setting a predefined threshold, when RMSE region Exceeds a predefined threshold or r region When the value is lower than the predefined threshold, the regional simulation accuracy can be quantitatively evaluated and poor areas with large errors can be identified. 10 . An electronic device comprising a memory and a processor, wherein the memory is used to store a computer program, and the processor runs the computer program to enable the electronic device to perform the method according to claim 1 .

11. A computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the method according to any one of claims 1 to 9 when executed by a computer.

Images