Voltage transformer error prediction model construction method, error evaluation method and device

Abstract

The present application relates to the technical field of electric power metering, and provides a voltage transformer error prediction model construction method, an error evaluation method and a device, the construction method comprising: collecting voltage and current phasors of a monitoring node and obtaining three-phase admittance parameters; constructing a node feature matrix based on the voltage and current phasors, establishing a graph adjacency matrix based on the three-phase admittance parameters, and using the node feature matrix and the graph adjacency matrix to perform error prediction by using a graph neural network to obtain voltage and current prediction errors; obtaining voltage estimated true values based on the voltage phasors and the voltage prediction errors, obtaining current estimated true values based on the current phasors and the current prediction errors, and determining voltage drop and current balance constraint losses of a three-phase transmission line based on the voltage estimated true values and the current estimated true values; and performing parameter iteration on the graph neural network based on the voltage drop and the current balance constraint losses to obtain an error prediction model. The method and device provided by the present application can maintain stable performance under different operating conditions and network topologies.

Description

Technical Field

[0001] This invention relates to the field of power metering technology, and in particular to a method for constructing a voltage transformer error prediction model, an error assessment method, and a device. Background Technology

[0002] With the increasing proportion of new energy sources, the operating environment of the power system is undergoing profound changes. The intermittency and volatility of new energy sources, as well as the integration of a large number of power electronic devices, have led to complex and variable power flow in the power grid, an expanded range of voltage fluctuations, and increased harmonic pollution, posing a severe challenge to the accuracy of system measurements.

[0003] As a critical measurement device, the accuracy of voltage transformers directly affects the system's operation and control level. In traditional power grids, their operating environment is relatively stable, and error characteristics can be maintained through periodic calibration. However, with the integration of a high proportion of renewable energy sources, factors such as frequent voltage fluctuations, high-frequency harmonic interference, and dynamic changes in secondary loads collectively cause voltage transformer errors to exhibit significant time-varying nonlinear characteristics, making it difficult for traditional static error models to accurately describe their true performance.

[0004] Current error detection mainly relies on offline calibration and online monitoring. The former offers high accuracy but fails to reflect actual operating conditions, while the latter, although capable of real-time monitoring, is prone to misjudgment under complex conditions. Existing methods struggle to effectively distinguish between systematic errors and random disturbances, thus failing to provide support for precise operation and maintenance. Therefore, there is an urgent need to develop new methods for voltage transformer error assessment adapted to the characteristics of modern power systems. Summary of the Invention

[0005] This invention provides a method for constructing a voltage transformer error prediction model, an error assessment method, and an apparatus to address the shortcomings of existing voltage transformer error detection methods in terms of low error estimation accuracy and insufficient stability under complex operating conditions.

[0006] This invention provides a method for constructing a voltage transformer error prediction model, comprising:

[0007] The three-phase voltage phasors and three-phase current phasors of each monitoring node are collected synchronously, and the three-phase admittance parameters between each monitoring node are obtained.

[0008] Based on the three-phase voltage phasors and three-phase current phasors, a node feature matrix is ​​constructed. Based on the three-phase admittance parameters between each monitoring node, a graph adjacency matrix is ​​established. The node feature matrix and the graph adjacency matrix are used to perform error prediction using a graph neural network to obtain the voltage prediction error and the current prediction error.

[0009] The true voltage estimate is obtained based on the three-phase voltage phasors and voltage prediction error, the true current estimate is obtained based on the three-phase current phasors and current prediction error, and the voltage drop constraint loss and current balance constraint loss of the three-phase transmission line are determined based on the true voltage estimate and the true current estimate.

[0010] Based on the voltage drop constraint loss and current balance constraint loss, the graph neural network is iterated to obtain the voltage transformer error prediction model.

[0011] According to the voltage transformer error prediction model construction method provided by the present invention, the step of determining the voltage drop constraint loss and current balance constraint loss of the three-phase transmission line based on the true voltage estimate and the true current estimate includes:

[0012] Based on the true values ​​of voltage estimates at the first and last nodes of a three-phase transmission line, and the true values ​​of current estimates flowing from the first node to the last node, the line voltage residual is calculated, and the voltage drop constraint loss of all three-phase transmission lines within a preset time period is determined based on the line voltage residual.

[0013] Based on the true value of the estimated current at the head end of the monitoring node and the true value of the estimated voltage of the monitoring node, the node current residual is calculated, and the current balance constraint loss of all monitoring nodes within a preset time period is determined based on the node current residual.

[0014] According to the voltage transformer error prediction model construction method provided by the present invention, the calculation of the line voltage residual based on the true voltage estimates of the first and last nodes of a three-phase transmission line, and the true current estimates flowing from the first node to the last node, includes:

[0015] Based on the series impedance matrix of the three-phase transmission line, the true value of the current estimated from the first node to the last node, the parallel admittance matrix of the line, and the true value of the voltage estimated from the first node, the line voltage drop of the three-phase transmission line is determined.

[0016] Based on the true voltage estimates of the first and last nodes of the three-phase transmission line and the voltage drop of the line, the line voltage residual of the three-phase transmission line is calculated.

[0017] According to the voltage transformer error prediction model construction method provided by the present invention, the step of calculating the node current residual based on the true value of the head-end current estimation of the monitoring node and the true value of the voltage estimation of the monitoring node includes:

[0018] Based on the true values ​​of the three-phase parallel admittance and voltage estimation of the monitoring node, the parallel current of the monitoring node is calculated;

[0019] Based on the true value of the estimated current at the head end of the monitoring node, the load current of the monitoring node, and the parallel current, the node current residual of the monitoring node is calculated.

[0020] According to the voltage transformer error prediction model construction method provided by the present invention, the step of iterating the parameters of the graph neural network based on the voltage drop constraint loss and the current balance constraint loss includes:

[0021] Error analysis is performed on the three-phase voltage phasors of each monitoring node at each time point, and the error analysis results are used as the error labels corresponding to the three-phase voltage phasors.

[0022] Based on the difference between the error label and the voltage prediction error, the error prediction loss is determined;

[0023] The graph neural network is iterated based on the voltage drop constraint loss, the current balance constraint loss, and the error prediction loss.

[0024] The present invention also provides an error assessment method, comprising:

[0025] Real-time acquisition of new measurement data, including three-phase voltage phasors and three-phase current phasors;

[0026] Error prediction is performed based on the measurement data and the voltage transformer error prediction model to obtain the real-time error. The voltage transformer error prediction model is obtained based on the voltage transformer error prediction model construction method.

[0027] Calculate the confidence level based on the voltage drop constraint loss and current balance constraint loss obtained from the real-time error calculation.

[0028] Online error assessment is performed based on the real-time error and the confidence level.

[0029] According to the error assessment method provided by the present invention, the online error assessment based on the real-time error and the confidence level includes:

[0030] Based on the confidence level and the preset confidence threshold, adjust the grading error threshold;

[0031] Online error assessment is performed based on the real-time error and the adjusted graded error threshold.

[0032] The present invention also provides a device for constructing a voltage transformer error prediction model, comprising:

[0033] The data acquisition unit is used to synchronously acquire the three-phase voltage phasors and three-phase current phasors of each monitoring node, and to obtain the three-phase admittance parameters between each monitoring node.

[0034] The error prediction unit is used to construct a node feature matrix based on the three-phase voltage phasors and the three-phase current phasors, establish a graph adjacency matrix based on the three-phase admittance parameters between each monitoring node, and use the node feature matrix and the graph adjacency matrix to perform error prediction using a graph neural network to obtain the voltage prediction error and the current prediction error.

[0035] The loss determination unit is used to obtain the true value of voltage estimation based on the three-phase voltage phasors and voltage prediction error, obtain the true value of current estimation based on the three-phase current phasors and current prediction error, and determine the voltage drop constraint loss and current balance constraint loss of the three-phase transmission line based on the true value of voltage estimation and the true value of current estimation.

[0036] The parameter iteration unit is used to perform parameter iteration on the graph neural network based on the voltage drop constraint loss and current balance constraint loss to obtain the voltage transformer error prediction model.

[0037] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the voltage transformer error prediction model construction method or the error assessment method described above.

[0038] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the voltage transformer error prediction model construction method or error assessment method as described above.

[0039] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the voltage transformer error prediction model construction method or error assessment method as described above.

[0040] The voltage transformer error prediction model construction method, error assessment method, and apparatus provided by this invention significantly improve estimation accuracy and stability by establishing an error estimation framework that considers three-phase line models and network topology constraints, and by fully utilizing redundant information from multiple measurement points for cross-validation. Combining graph neural networks and physical constraints, it exhibits good generalization ability and robustness, maintaining stable performance under different operating conditions and network topologies. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0042] Figure 1 This is one of the flowcharts illustrating the method for constructing a voltage transformer error prediction model provided by the present invention.

[0043] Figure 2 This is one of the flowcharts of the error assessment method provided by the present invention.

[0044] Figure 3 This is the second flowchart of the error assessment method provided by the present invention.

[0045] Figure 4 This is a schematic diagram of the structure of the voltage transformer error prediction model construction device provided by the present invention.

[0046] Figure 5 This is a schematic diagram of the error assessment device provided by the present invention.

[0047] Figure 6 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0049] As a critical measuring device in power systems, the accuracy of voltage transformers directly impacts the accuracy of energy metering, the reliability of relay protection operations, and the effectiveness of grid dispatch control. In traditional power systems, the operating environment of voltage transformers is relatively stable, with slow load changes, and error parameters such as ratio error and phase error can be maintained within acceptable ranges through periodic verification. However, in grid environments with a high proportion of renewable energy integration, voltage transformers face unprecedented operational challenges. The power output of renewable energy plants fluctuates drastically with weather conditions, leading to frequent changes in grid voltage; high-frequency harmonic components generated by power electronic devices such as inverters alter the magnetization characteristics of voltage transformers; and the random grid connection and disconnection operations of distributed generation sources make secondary loads highly dynamic. These factors combined result in voltage transformer error characteristics exhibiting strong time-varying and nonlinear features, making traditional static error models insufficient to accurately describe their true performance.

[0050] Existing voltage transformer error detection and evaluation technologies mainly rely on offline verification and online monitoring. Offline verification uses dedicated testing equipment to accurately measure voltage transformers, achieving high measurement accuracy, but it suffers from drawbacks such as long testing cycles, high costs, and inability to reflect actual operating conditions. While online monitoring technology can acquire voltage transformer operating data in real time, existing methods are mostly based on single-point measurement information, lacking consideration of the overall constraints of the power grid, and are prone to misjudgment under complex operating conditions. More importantly, traditional methods generally use simple criteria based on threshold comparisons, which cannot effectively distinguish between systematic and random errors, making it difficult to accurately identify the root causes of errors and providing limited support for operation and maintenance decisions.

[0051] From the physical perspective of power grid operation, there are strict network constraints between voltage and current measurements at various measuring points. According to Kirchhoff's laws, the sum of voltage drops in any closed loop is zero, and the algebraic sum of currents at any node is zero. These fundamental laws form the theoretical basis for power grid flow calculations and state estimation. In three-phase unbalanced systems, the self-impedance and mutual impedance of the lines must also be considered. Furthermore, for lines equipped with regulating equipment such as on-load tap-changing transformers and static var compensators, changes in the operating state of these devices will have a deterministic impact on electrical measurements at both ends of the line. These network constraints provide important redundant information for identifying and estimating voltage transformer errors, but current technology has not fully utilized this advantage. If a mathematical relationship between measurement errors and network constraints can be established, it may be possible to significantly improve the accuracy and reliability of error estimation without increasing additional hardware investment.

[0052] In recent years, artificial intelligence (AI) technology has been increasingly applied in power systems, providing new approaches to solving complex nonlinear problems. In particular, graph neural network (GNN) technology can effectively handle complex systems like power grids, which exhibit significant graph structure characteristics, and has shown promising application prospects in areas such as load forecasting, fault diagnosis, and optimal dispatching. A power system is essentially a large-scale graph network, with nodes representing buses and edges representing lines or transformers. This inherent graph structure gives GNN technology a broad application scope. However, in the field of voltage transformer error estimation, how to organically combine data-driven learning methods with the physical constraints of the power grid to ensure both the accuracy of the estimation results and the preservation of the physical meaning of the interpretation remains a challenging research topic.

[0053] A comprehensive analysis of current technological status and practical engineering needs necessitates the development of a new voltage transformer error estimation technology that can adapt to the complex operating environment of new energy power grids, fully utilize multi-source information, and possess good interpretability and generalization capabilities. This technology should overcome the limitations of existing technologies, ensuring the reliability and accuracy of the power system measurement system under large-scale new energy integration, and providing a solid technical guarantee for the construction of a new power system.

[0054] The applicant's long-term research has revealed that existing voltage transformer error detection methods primarily rely on offline calibration and single-point threshold monitoring, which are ill-suited to the complex operating environment of new energy power grids, characterized by frequent voltage fluctuations, severe harmonic pollution, and dynamic load changes. Traditional methods neglect grid network constraints and lack effective utilization of the physical correlations between measurement data from multiple points, resulting in low accuracy and instability in error estimation under complex operating conditions. Furthermore, existing technologies struggle to handle complex situations involving lines with on-load tap-changing equipment, failing to simultaneously consider the coupled effects of line parameter deviations, transformer system errors, and tap ratio deviations, and lacking the ability to accurately track dynamic changes in voltage transformer errors.

[0055] To address the aforementioned problems, this invention proposes a method for constructing a voltage transformer error prediction model. This method first synchronously acquires the three-phase voltage phasors and three-phase current phasors of each monitoring node, and obtains the three-phase admittance parameters between each monitoring node. Then, a node feature matrix is ​​constructed based on the three-phase voltage phasors and three-phase current phasors, and a graph adjacency matrix is ​​established based on the three-phase admittance parameters between each monitoring node. A graph neural network is then used to perform error prediction using the node feature matrix and the graph adjacency matrix to obtain the voltage prediction error and current prediction error. The true voltage estimate is obtained based on the three-phase voltage phasors and voltage prediction error, and the true current estimate is obtained based on the three-phase current phasors and current prediction error. The voltage drop constraint loss and current balance constraint loss of the three-phase transmission line are then determined based on the true voltage estimate and current estimate. Finally, the graph neural network is iterated based on the voltage drop constraint loss and current balance constraint loss to obtain the voltage transformer error prediction model.

[0056] The method provided in this invention combines the power grid topology with multi-node synchronous measurement data and utilizes graph neural networks for modeling. The voltage transformer error prediction model constructed in this way can fully learn and utilize the inherent physical correlations and network constraints between measurement data from multiple measurement points in the power grid, rather than performing isolated single-point analysis. This effectively addresses complex operating conditions such as rapid voltage fluctuations and harmonic interference caused by the integration of new energy sources, significantly improving the accuracy of error prediction and enhancing the model's anti-interference stability.

[0057] Furthermore, by constructing a joint loss function that integrates voltage drop constraints and current balance constraints, and iteratively training the graph neural network model, the prediction model acquires a powerful ability to learn physical laws. The model not only learns data features but is also strictly constrained to satisfy the fundamental physical laws of circuits. The model trained in this way can more profoundly reveal the intrinsic relationship between errors and operating states, thereby achieving dynamic and accurate tracking of the time-varying and nonlinear characteristics of voltage transformer errors, overcoming the poor adaptability of traditional static models.

[0058] This invention can be applied to scenarios requiring voltage transformer error assessment, such as metering points at the outlet of the collection line in new energy power plants, harmonic voltage monitoring points at grid connection points, and key power flow observation points in active distribution networks. The execution entity of this method can be an electronic device such as a terminal device, computer, server, server cluster, or a specially designed voltage transformer error assessment device, or an error assessment device installed within such an electronic device. This device can be implemented through software, hardware, or a combination of both.

[0059] Figure 1 This is one of the flowcharts illustrating the voltage transformer error prediction model construction method provided by the present invention, such as... Figure 1 As shown, the method includes the following steps 110-140.

[0060] Step 110: Synchronously collect the three-phase voltage phasors and three-phase current phasors of each monitoring node, and obtain the three-phase admittance parameters between each monitoring node.

[0061] Specifically, this step is the data preparation phase, which prepares high-quality, spatiotemporally synchronized input data for the graph neural network model. Monitoring nodes are key locations within the power network where synchronized phasor measurement devices are selected and installed. These nodes are typically located at important grid connections, such as substation busbars, new energy plant grid connection points, and important load access points. Synchronous acquisition ensures that the voltage and current phasor data collected by all nodes are strictly synchronized at the same time.

[0062] In some possible implementations, three-phase voltage and current synchronization phasor data of each node are collected synchronously within the new energy power station. The total collection period is set to [missing information]. The set of monitoring nodes is denoted as ,in This indicates the total number of monitoring nodes. At time [time]... Data collection nodes Three-phase voltage phasors and three-phase current phasors Construct a synchronous measurement data matrix:

[0063]

[0064] Simultaneously obtain the line electrical parameter matrix , representing a node With nodes Three-phase admittance parameters between:

[0065]

[0066] in For series admittance matrix, This is a parallel admittance matrix. A complete time-segment dataset is established. .

[0067] Step 120: Construct a node feature matrix based on the three-phase voltage phasors and the three-phase current phasors, establish a graph adjacency matrix based on the three-phase admittance parameters between each monitoring node, and use a graph neural network to perform error prediction using the node feature matrix and the graph adjacency matrix to obtain the voltage prediction error and the current prediction error.

[0068] Specifically, the purpose of this step is to build an intelligent model that can integrate the physical structure of the power grid with real-time operational data, and to use this model to infer systematic errors hidden in the raw measurement data.

[0069] The node characteristic matrix is ​​a matrix used to describe the characteristics of all nodes in a graph. The node characteristic matrix consists of the real and imaginary parts (or magnitudes, phase angles) of the three-phase voltage phasors and three-phase current phasors measured at that node.

[0070] For example, node feature matrix Including a sliding window of size The real and imaginary parts of the time-series voltage phasors and the real and imaginary parts of the current phasors are represented as follows:

[0071]

[0072] A graph adjacency matrix is ​​a mathematical matrix used to describe the connection relationships between nodes in a graph. If there is a direct electrical connection between two nodes (such as a power line or transformer), a non-zero value is used in the corresponding position of the matrix to represent the connection strength. In this embodiment, the connection strength is determined by the three-phase admittance parameter; the larger the admittance value, the tighter the electrical connection.

[0073] A graph neural network model based on the topological relationships of monitoring nodes is constructed. Monitoring nodes are treated as graph nodes, and electrical connections as graph edges; a graph adjacency matrix is ​​then established. :

[0074]

[0075] in This represents the average value of the three-phase admittance.

[0076] Multi-layer graph neural networks, optionally graph convolutional neural networks, are used for feature extraction and error prediction.

[0077]

[0078]

[0079] in To standardize the adjacency matrix, For degree matrix, For the first Layer weight parameters.

[0080] Graph neural networks use a message passing mechanism to allow information from one node to influence the judgment of its neighboring nodes, ultimately outputting the voltage and current prediction errors for each monitoring node.

[0081] Step 130: Obtain the true voltage estimate based on the three-phase voltage phasor and voltage prediction error; obtain the true current estimate based on the three-phase current phasor and current prediction error; and determine the voltage drop constraint loss and current balance constraint loss of the three-phase transmission line based on the true voltage estimate and the true current estimate.

[0082] Specifically, the estimated true values ​​refer to the voltage and current values ​​that are closer to the physical reality after model correction. The estimated true values ​​are the optimal estimates of real physical quantities that cannot be directly measured by the model.

[0083] Under the small-error approximation condition, the relationship between the measured value and the true value is as follows:

[0084]

[0085]

[0086] in This represents the magnitude error vector in the voltage prediction error. This represents the phase error vector in the voltage prediction error. This represents the magnitude error vector in the current prediction error. This is the phase error vector in the current prediction error.

[0087] The estimated true value is obtained by subtracting the error from the model output from the measured value, which is the three-phase voltage phasor. Subtracting the voltage prediction error yields the true voltage estimate. Transform the three-phase current phasors Subtracting the current prediction error yields the true value of the current estimate. .

[0088] Voltage drop constraint loss is a loss function based on Kirchhoff's voltage law. KVL states that in any closed loop, the algebraic sum of the voltage drops across all components is zero. For a transmission line, its voltage drop should equal the line impedance multiplied by the current flowing through it. Voltage drop constraint loss calculates the difference between the line voltage drop estimated from the true value and the voltage drop calculated according to KVL theory. The smaller the loss value, the more the estimated true value satisfies KVL.

[0089] The current balance constraint loss is a loss function based on Kirchhoff's Current Law (KCL). KCL states that the sum of the currents flowing into any node is zero. The current balance constraint loss calculates the algebraic sum of the estimated true values ​​of all line currents flowing into a node (i.e., the net current), which ideally should be zero. The smaller the loss value, the more the estimated true value satisfies KCL.

[0090] Substituting the estimated true values ​​into the voltage drop equation and the current balance equation respectively, we obtain the voltage drop constraint loss and the current balance constraint loss of the three-phase transmission line.

[0091] This embodiment links data-driven error prediction with the physical model by estimating the true value, providing a clear objective for subsequent optimization—finding the error value that makes the estimated true value best conform to the physical laws.

[0092] Step 140: Based on the voltage drop constraint loss and current balance constraint loss, perform parameter iteration on the graph neural network to obtain the voltage transformer error prediction model.

[0093] Specifically, the voltage drop constraint loss and current balance constraint loss functions do not rely on manually labeled real error data (which is extremely difficult to obtain). Instead, they are based on eternal and precise physical laws, providing powerful and free supervision signals for model training. The level of the loss value directly reflects the physical rationality of the model's prediction results. By minimizing the loss function defined by physical laws, the graph neural network is guided to adjust its internal parameters so that its error predictions maximize the satisfaction of the final true estimate with KVL and KCL.

[0094] In each iteration, the algorithm fine-tunes the weight parameters in the graph neural network based on the gradient calculated from the loss function, with the goal of reducing the value of the loss function.

[0095] Node feature matrix Standardized adjacency matrix The input is fed into the constructed model, and training is performed using a composite loss function consisting of voltage drop constraint loss and current balance constraint loss. A learning rate decay strategy is set, with an initial learning rate... Every time The learning rate decays to its original value after each training epoch. Times:

[0096]

[0097] in This indicates the current training round. An early stopping mechanism is set up so that the validation set loss continues... Training stops when the change in rounds is less than a set threshold.

[0098] The trained, parameter-fixed, and practically usable model is the voltage transformer error prediction model. This model has learned how to accurately predict the errors of each voltage transformer from the input grid synchronous measurement data. When new measurement data is input, the model can directly output the current error estimate.

[0099] In some embodiments, step 130, which determines the voltage drop constraint loss and current balance constraint loss of the three-phase transmission line based on the true voltage estimate and the true current estimate, includes:

[0100] Step 131: Based on the true values ​​of voltage estimates of the first and last nodes on the three-phase transmission line and the true values ​​of current estimates flowing from the first node to the last node, calculate the line voltage residual and determine the voltage drop constraint loss of all three-phase transmission lines within a preset time period based on the line voltage residual.

[0101] Step 132: Based on the true value of the estimated current at the head end of the monitoring node and the true value of the estimated voltage of the monitoring node, calculate the node current residual, and determine the current balance constraint loss of all monitoring nodes within the preset time period based on the node current residual.

[0102] Specifically, the preset time period refers to the length of the time window used for loss calculation. That is, loss calculation is no longer a single point in time, but a series of continuous time steps, which allows the model to learn the dynamic characteristics of the error over time and utilize more data to improve the stability of the estimation.

[0103] The voltage residual of a single line at a single time point is calculated using the true voltage estimates of the first and last nodes of a three-phase transmission line, as well as the true current estimates flowing from the first to the last node. The squared L2 norm of the voltage residual vector represents the line voltage drop loss. The voltage drop loss of all lines within a preset time period is summed to obtain the voltage drop constraint loss.

[0104] For any node, the current residual at that moment is calculated using the true values ​​of the estimated current at the node's head and the true values ​​of the estimated voltage at the node. The squared L2 norm of the current residual vector is the node current loss. The current balance constraint loss is obtained by summing the node current losses of all nodes over a preset time period.

[0105] In some embodiments, step 131 specifically includes:

[0106] Step 131-1: Based on the line series impedance matrix of the three-phase transmission line, the true value of the current estimated from the first node to the last node, the line parallel admittance matrix, and the true value of the voltage estimated from the first node, determine the line voltage drop of the three-phase transmission line.

[0107] Step 131-2: Based on the true voltage estimates of the first and last nodes of the three-phase transmission line and the line voltage drop, calculate the line voltage residual of the three-phase transmission line.

[0108] Specifically, line voltage drop refers to the voltage drop that occurs when current flows through the line itself. It is a theoretical voltage drop caused solely by the impedance of a three-phase transmission line.

[0109] The line voltage drop can be expressed by the formula: .in, This is the series impedance matrix of the line. as the first node Flow to the end node The true value of the current estimate, , is the parallel admittance matrix of the line. Half of as the first node The true value of the voltage estimate.

[0110] The residual voltage of a three-phase transmission line is expressed as:

[0111] = - -

[0112] in, End node The true value of the voltage estimate. Ideally... .

[0113] Furthermore, the voltage drop constraint loss of all three-phase transmission lines within the preset time period is expressed as:

[0114]

[0115] In some embodiments, step 132 specifically includes:

[0116] Step 132-1: Calculate the parallel current of the monitoring node based on the true values ​​of the three-phase parallel admittance and voltage estimation of the monitoring node;

[0117] Step 132-2: Based on the true value of the estimated current at the head end of the monitoring node, the load current of the monitoring node, and the parallel current, calculate the node current residual of the monitoring node.

[0118] Specifically, the parallel current of the monitoring node is expressed by the formula: ,in , is the parallel admittance matrix of the line. Half of it.

[0119] Node current residuals at monitoring nodes Expressed as a formula:

[0120]

[0121] In the formula, To estimate the true value of the current at the starting end of the monitoring node m, This is the load current.

[0122] Furthermore, the current balance constraint loss of all monitoring nodes within the preset time period is expressed as:

[0123]

[0124] In some embodiments, step 140 involves iterating the parameters of the graph neural network based on voltage drop constraint loss and current balance constraint loss, including:

[0125] Step 141: Perform error analysis on the three-phase voltage phasors of each monitoring node at each time point, and use the error analysis results as the error labels corresponding to the three-phase voltage phasors.

[0126] Step 142: Determine the error prediction loss based on the difference between the error label and the voltage prediction error;

[0127] Step 143: Iterate the parameters of the graph neural network based on voltage drop constraint loss, current balance constraint loss and error prediction loss.

[0128] Specifically, existing error analysis algorithms are used to process the acquired three-phase voltage phasors, generating a reference error value for each monitoring node. An error calculation function is defined. , for time node Error analysis was performed on the three-phase voltage measurements:

[0129]

[0130] in express Compared to the error, express Phase angle error. Constructing an error label matrix. To form a labeled dataset .

[0131] Then, based on the difference between the error label and the voltage prediction error, the error prediction loss is determined. The error prediction loss is expressed as:

[0132]

[0133] Based on this, the total model loss is determined using voltage drop constraint loss, current balance constraint loss, and error prediction loss as the overall optimization objective function. This can be expressed as:

[0134]

[0135] in and These are weighting coefficients, which control the importance of voltage drop constraints and current balance constraints, respectively.

[0136] Figure 2 This is one of the flowcharts illustrating the error assessment method provided by the present invention, such as... Figure 2 The present invention provides an error assessment method, including steps 210-240.

[0137] Step 210: Acquire new measurement data in real time, including three-phase voltage phasors and three-phase current phasors;

[0138] Step 220: Based on the measurement data and the voltage transformer error prediction model, perform error prediction to obtain the real-time error. The voltage transformer error prediction model is obtained based on the voltage transformer error prediction model construction method.

[0139] Step 230: Calculate the confidence level based on the voltage drop constraint loss and current balance constraint loss obtained from the real-time error calculation;

[0140] Step 240: Perform online error assessment based on real-time error and confidence level.

[0141] Specifically, the trained model is deployed for real-time error evaluation. New measurement data. The data is newly acquired from the synchronous phasor measurement devices deployed at various monitoring nodes in the power grid, providing the latest, real-time input data for the online error assessment system.

[0142] New measurement data Input to voltage transformer error prediction model, output real-time error The degree to which physical constraints are satisfied is verified. Voltage drop constraint loss and current balance constraint loss are calculated based on real-time error, and then confidence level is calculated based on voltage drop constraint loss and current balance constraint loss.

[0143] Confidence The calculation method is as follows:

[0144]

[0145] in, For voltage drop constraint loss, This is due to current balance constraint losses.

[0146] Online error assessment refers to combining the error value predicted by the model with the confidence level of that prediction to arrive at a comprehensive and guiding final conclusion for operation and maintenance. Online error assessment based on real-time error and confidence level can generate an actionable assessment result that reflects both the potential error of the equipment and the uncertainty of prediction, providing decision support for operation and maintenance personnel and avoiding false alarms.

[0147] This embodiment not only utilizes the predictive power of the model, but also introduces confidence level, using the degree of compliance with physical constraints as an intrinsic measure of the reliability of the prediction results, thereby greatly improving the practical value and reliability of the system in industrial field applications.

[0148] In some embodiments, step 240 specifically includes:

[0149] Adjust the grading error threshold based on the confidence level and the preset confidence threshold;

[0150] Online error assessment is performed based on the magnitude of the real-time error and the adjusted graded error threshold.

[0151] Specifically, the preset confidence threshold is a pre-defined critical value (e.g., 0.9) used to distinguish between high and low confidence levels. When the model predicts high confidence, a strict error threshold is used to ensure high sensitivity; when the confidence level is low, the threshold is appropriately relaxed to avoid false alarms.

[0152] For example, if the calculated confidence level is greater than 0.9, the error threshold remains unchanged; if the confidence level is less than 0.9, all error thresholds are adjusted to 120% of their original values.

[0153] Then, based on the real-time error and the adjusted graded error threshold, an online error assessment is performed. Optionally:

[0154]

[0155] in, For ratio error, This represents the phase angle error.

[0156] Based on the above embodiments, Figure 3 This is the second flowchart of the error assessment method provided by the present invention, as shown below. Figure 3As shown, a method for evaluating the error of a voltage transformer is provided, including:

[0157] S1: Collect the three-phase voltage and three-phase current phasors of each monitoring node within the new energy power station, and simultaneously import the system's cable admittance / parameters to form a complete time-period dataset. Based on existing error analysis algorithms, generate evaluation data including voltage measurement ratio difference and angle difference for each time point and node, forming a labeled dataset.

[0158] S2 uses monitoring nodes as graph nodes and primary electrical connections as graph edges. Adjacency relationships are constructed and normalized based on three-phase admittance information to obtain a graph consistent with the station's topology. The node feature matrix is ​​obtained by splicing time-series voltage and current phasor data within a sliding window, and the graph's adjacency matrix is ​​determined by the three-phase admittances between nodes.

[0159] S3 introduces both a supervision term and a physical consistency constraint into the training objective. The physical constraint includes the voltage drop consistency residual and the node injection current residual based on the line equivalent model. Finally, a composite loss function is constructed by weighting the supervision error, voltage constraint loss and current constraint loss.

[0160] S4 inputs the node feature matrices and normalized graph structure at all time points into the model, trains it using the composite loss, and sets an early stopping mechanism to avoid overfitting.

[0161] S5 deploys the trained model, evaluates newly arrived measurement data in real time, outputs the ratio / angle difference estimate of each node and the confidence level calculated based on the voltage and current physical residuals, and performs graded alarms according to preset thresholds.

[0162] The method provided in this invention significantly improves estimation accuracy and stability by establishing an error estimation framework that considers three-phase line models and network topology constraints, and by fully utilizing redundant information from multiple measurement points for cross-validation. Combining graph neural networks and physical constraints, it exhibits good generalization ability and robustness, maintaining stable performance under different operating conditions and network topologies.

[0163] The voltage transformer error prediction model construction device provided by the present invention is described below. The voltage transformer error prediction model construction device described below and the voltage transformer error prediction model construction method described above can be referred to in correspondence.

[0164] Figure 4 This is a schematic diagram of the voltage transformer error prediction model construction device provided by the present invention, as shown below. Figure 4 As shown, the device includes:

[0165] The data acquisition unit 410 is used to synchronously acquire the three-phase voltage phasors and three-phase current phasors of each monitoring node, and to obtain the three-phase admittance parameters between each monitoring node.

[0166] Error prediction unit 420 is used to construct a node feature matrix based on the three-phase voltage phasors and three-phase current phasors, establish a graph adjacency matrix based on the three-phase admittance parameters between each monitoring node, and use the node feature matrix and the graph adjacency matrix to perform error prediction using a graph neural network to obtain voltage prediction error and current prediction error.

[0167] The loss determination unit 430 is used to obtain the true value of voltage estimation based on the three-phase voltage phasors and voltage prediction error, obtain the true value of current estimation based on the three-phase current phasors and current prediction error, and determine the voltage drop constraint loss and current balance constraint loss of the three-phase transmission line based on the true value of voltage estimation and the true value of current estimation.

[0168] The parameter iteration unit 440 is used to perform parameter iteration on the graph neural network based on the voltage drop constraint loss and current balance constraint loss to obtain the voltage transformer error prediction model.

[0169] Based on the above embodiments, the loss determination unit is specifically used for:

[0170] Based on the true values ​​of voltage estimates at the first and last nodes of a three-phase transmission line, and the true values ​​of current estimates flowing from the first node to the last node, the line voltage residual is calculated, and the voltage drop constraint loss of all three-phase transmission lines within a preset time period is determined based on the line voltage residual.

[0171] Based on the true value of the estimated current at the head end of the monitoring node and the true value of the estimated voltage of the monitoring node, the node current residual is calculated, and the current balance constraint loss of all monitoring nodes within a preset time period is determined based on the node current residual.

[0172] Based on the above embodiments, the loss determination unit is specifically used for:

[0173] Based on the series impedance matrix of the three-phase transmission line, the true value of the current estimated from the first node to the last node, the parallel admittance matrix of the line, and the true value of the voltage estimated from the first node, the line voltage drop of the three-phase transmission line is determined.

[0174] Based on the true voltage estimates of the first and last nodes of the three-phase transmission line and the voltage drop of the line, the line voltage residual of the three-phase transmission line is calculated.

[0175] Based on the above embodiments, the loss determination unit is specifically used for:

[0176] Based on the true values ​​of the three-phase parallel admittance and voltage estimation of the monitoring node, the parallel current of the monitoring node is calculated;

[0177] Based on the true value of the estimated current at the head end of the monitoring node, the load current of the monitoring node, and the parallel current, the node current residual of the monitoring node is calculated.

[0178] Based on the above embodiments, the parameter iteration unit is specifically used for:

[0179] Error analysis is performed on the three-phase voltage phasors of each monitoring node at each time point, and the error analysis results are used as the error labels corresponding to the three-phase voltage phasors.

[0180] Based on the difference between the error label and the voltage prediction error, the error prediction loss is determined;

[0181] The graph neural network is iterated based on the voltage drop constraint loss, the current balance constraint loss, and the error prediction loss.

[0182] Figure 5 This is a schematic diagram of the error assessment device provided by the present invention, as shown below. Figure 5 As shown, the device includes:

[0183] The data acquisition unit 510 acquires new measurement data in real time, including three-phase voltage phasors and three-phase current phasors.

[0184] The error acquisition unit 520 performs error prediction based on the measurement data and the voltage transformer error prediction model to obtain the real-time error. The voltage transformer error prediction model is obtained based on the voltage transformer error prediction model construction method.

[0185] The confidence calculation unit 530 is used to calculate the confidence level based on the voltage drop constraint loss and current balance constraint loss obtained by the real-time error calculation.

[0186] Error assessment unit 540 is used to perform online error assessment based on the real-time error and the confidence level.

[0187] Based on the above embodiments, the error evaluation unit is specifically used for:

[0188] Based on the confidence level and the preset confidence threshold, adjust the grading error threshold;

[0189] Online error assessment is performed based on the real-time error and the adjusted graded error threshold.

[0190] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6As shown, the electronic device may include: a processor 610, a communications interface 620, a memory 630, and a communications bus 640, wherein the processor 610, the communications interface 620, and the memory 630 communicate with each other through the communications bus 640. The processor 610 can call logic instructions in the memory 630 to execute a voltage transformer error prediction model construction method or an error evaluation method. The model construction method includes: synchronously acquiring the three-phase voltage phasors and three-phase current phasors of each monitoring node, and obtaining the three-phase admittance parameters between each monitoring node; constructing a node feature matrix based on the three-phase voltage phasors and three-phase current phasors, establishing a graph adjacency matrix based on the three-phase admittance parameters between each monitoring node, applying the node feature matrix and the graph adjacency matrix to perform error prediction using a graph neural network to obtain voltage prediction error and current prediction error; obtaining the true voltage estimate based on the three-phase voltage phasors and voltage prediction error, obtaining the true current estimate based on the three-phase current phasors and current prediction error, and determining the voltage drop constraint loss and current balance constraint loss of the three-phase transmission line based on the true voltage estimate and current estimate; performing parameter iteration on the graph neural network based on the voltage drop constraint loss and current balance constraint loss to obtain the voltage transformer error prediction model.

[0191] The error assessment method includes: acquiring new measurement data in real time, the measurement data including three-phase voltage phasors and three-phase current phasors; performing error prediction based on the measurement data and a voltage transformer error prediction model to obtain a real-time error, the voltage transformer error prediction model being obtained based on a voltage transformer error prediction model construction method; calculating a confidence level based on the voltage drop constraint loss and current balance constraint loss calculated from the real-time error; and performing online error assessment based on the real-time error and the confidence level.

[0192] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0193] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the voltage transformer error prediction model construction method or error evaluation method provided by the above methods. The model construction method includes: synchronously acquiring the three-phase voltage phasors and three-phase current phasors of each monitoring node, and obtaining the three-phase admittance parameters between each monitoring node; constructing a node feature matrix based on the three-phase voltage phasors and three-phase current phasors, establishing a graph adjacency matrix based on the three-phase admittance parameters between each monitoring node, applying the node feature matrix and the graph adjacency matrix to perform error prediction using a graph neural network to obtain voltage prediction error and current prediction error; obtaining the true voltage estimate based on the three-phase voltage phasors and voltage prediction error, obtaining the true current estimate based on the three-phase current phasors and current prediction error, and determining the voltage drop constraint loss and current balance constraint loss of the three-phase transmission line based on the voltage drop constraint loss and current balance constraint loss; performing parameter iteration on the graph neural network based on the voltage drop constraint loss and current balance constraint loss to obtain the voltage transformer error prediction model.

[0194] The error assessment method includes: acquiring new measurement data in real time, the measurement data including three-phase voltage phasors and three-phase current phasors; performing error prediction based on the measurement data and a voltage transformer error prediction model to obtain a real-time error, the voltage transformer error prediction model being obtained based on a voltage transformer error prediction model construction method; calculating a confidence level based on the voltage drop constraint loss and current balance constraint loss calculated from the real-time error; and performing online error assessment based on the real-time error and the confidence level.

[0195] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements a voltage transformer error prediction model construction method or error evaluation method provided by the above methods. The model construction method includes: synchronously acquiring three-phase voltage phasors and three-phase current phasors of each monitoring node, and obtaining three-phase admittance parameters between each monitoring node; constructing a node feature matrix based on the three-phase voltage phasors and three-phase current phasors, establishing a graph adjacency matrix based on the three-phase admittance parameters between each monitoring node, applying the node feature matrix and the graph adjacency matrix to perform error prediction using a graph neural network to obtain voltage prediction error and current prediction error; obtaining the true voltage estimate based on the three-phase voltage phasors and voltage prediction error, obtaining the true current estimate based on the three-phase current phasors and current prediction error, and determining the voltage drop constraint loss and current balance constraint loss of the three-phase transmission line based on the voltage drop constraint loss and current balance constraint loss; performing parameter iteration on the graph neural network based on the voltage drop constraint loss and current balance constraint loss to obtain the voltage transformer error prediction model.

[0196] The error assessment method includes: acquiring new measurement data in real time, the measurement data including three-phase voltage phasors and three-phase current phasors; performing error prediction based on the measurement data and a voltage transformer error prediction model to obtain a real-time error, the voltage transformer error prediction model being obtained based on a voltage transformer error prediction model construction method; calculating a confidence level based on the voltage drop constraint loss and current balance constraint loss calculated from the real-time error; and performing online error assessment based on the real-time error and the confidence level.

[0197] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0198] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0199] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing a voltage transformer error prediction model, characterized in that, include: The three-phase voltage phasors and three-phase current phasors of each monitoring node are collected synchronously, and the three-phase admittance parameters between each monitoring node are obtained. Based on the three-phase voltage phasors and three-phase current phasors, a node feature matrix is ​​constructed. Based on the three-phase admittance parameters between each monitoring node, a graph adjacency matrix is ​​established. The node feature matrix and the graph adjacency matrix are used to perform error prediction using a graph neural network to obtain the voltage prediction error and the current prediction error. The true voltage estimate is obtained based on the three-phase voltage phasors and voltage prediction error, the true current estimate is obtained based on the three-phase current phasors and current prediction error, and the voltage drop constraint loss and current balance constraint loss of the three-phase transmission line are determined based on the true voltage estimate and the true current estimate. Based on the voltage drop constraint loss and current balance constraint loss, the graph neural network is iterated to obtain the voltage transformer error prediction model. The determination of voltage drop constraint loss and current balance constraint loss of three-phase transmission lines based on the true values ​​of voltage estimation and current estimation includes: Based on the true values ​​of voltage estimates at the first and last nodes of a three-phase transmission line, and the true values ​​of current estimates flowing from the first node to the last node, the line voltage residual is calculated, and the voltage drop constraint loss of all three-phase transmission lines within a preset time period is determined based on the line voltage residual. Based on the true value of the estimated current at the head end of the monitoring node and the true value of the estimated voltage of the monitoring node, the node current residual is calculated, and the current balance constraint loss of all monitoring nodes within a preset time period is determined based on the node current residual.

2. The method for constructing a voltage transformer error prediction model according to claim 1, characterized in that, The calculation of the line voltage residual based on the true voltage estimates of the first and last nodes of a three-phase transmission line, and the true current estimates flowing from the first node to the last node, includes: Based on the series impedance matrix of the three-phase transmission line, the true value of the current estimated from the first node to the last node, the parallel admittance matrix of the line, and the true value of the voltage estimated from the first node, the line voltage drop of the three-phase transmission line is determined. Based on the true voltage estimates of the first and last nodes of the three-phase transmission line and the voltage drop of the line, the line voltage residual of the three-phase transmission line is calculated.

3. The method for constructing a voltage transformer error prediction model according to claim 1, characterized in that, The calculation of the node current residual based on the true value of the head-end current estimation of the monitoring node and the true value of the voltage estimation of the monitoring node includes: Based on the true values ​​of the three-phase parallel admittance and voltage estimation of the monitoring node, the parallel current of the monitoring node is calculated; Based on the true value of the estimated current at the head end of the monitoring node, the load current of the monitoring node, and the parallel current, the node current residual of the monitoring node is calculated.

4. The method for constructing a voltage transformer error prediction model according to claim 1, characterized in that, The parameter iteration of the graph neural network based on the voltage drop constraint loss and current balance constraint loss includes: Error analysis is performed on the three-phase voltage phasors of each monitoring node at each time point, and the error analysis results are used as the error labels corresponding to the three-phase voltage phasors. Based on the difference between the error label and the voltage prediction error, the error prediction loss is determined; The graph neural network is iterated based on the voltage drop constraint loss, the current balance constraint loss, and the error prediction loss.

5. An error assessment method, characterized in that, include: Real-time acquisition of new measurement data, including three-phase voltage phasors and three-phase current phasors; Error prediction is performed based on the measurement data and the voltage transformer error prediction model to obtain the real-time error. The voltage transformer error prediction model is obtained based on the voltage transformer error prediction model construction method according to any one of claims 1 to 4. Calculate the confidence level based on the voltage drop constraint loss and current balance constraint loss obtained from the real-time error calculation. Online error assessment is performed based on the real-time error and the confidence level.

6. The error assessment method according to claim 5, characterized in that, The online error assessment based on the real-time error and the confidence level includes: Based on the confidence level and the preset confidence threshold, adjust the grading error threshold; Online error assessment is performed based on the real-time error and the adjusted graded error threshold.

7. A device for constructing a voltage transformer error prediction model, characterized in that, include: The data acquisition unit is used to synchronously acquire the three-phase voltage phasors and three-phase current phasors of each monitoring node, and to obtain the three-phase admittance parameters between each monitoring node. The error prediction unit is used to construct a node feature matrix based on the three-phase voltage phasors and the three-phase current phasors, establish a graph adjacency matrix based on the three-phase admittance parameters between each monitoring node, and use the node feature matrix and the graph adjacency matrix to perform error prediction using a graph neural network to obtain the voltage prediction error and the current prediction error. The loss determination unit is used to obtain the true value of voltage estimation based on the three-phase voltage phasors and voltage prediction error, obtain the true value of current estimation based on the three-phase current phasors and current prediction error, and determine the voltage drop constraint loss and current balance constraint loss of the three-phase transmission line based on the true value of voltage estimation and the true value of current estimation. The parameter iteration unit is used to perform parameter iteration on the graph neural network based on the voltage drop constraint loss and current balance constraint loss to obtain the voltage transformer error prediction model. The loss determination unit is specifically used for: Based on the true values ​​of voltage estimates at the first and last nodes of a three-phase transmission line, and the true values ​​of current estimates flowing from the first node to the last node, the line voltage residual is calculated, and the voltage drop constraint loss of all three-phase transmission lines within a preset time period is determined based on the line voltage residual. Based on the true value of the estimated current at the head end of the monitoring node and the true value of the estimated voltage of the monitoring node, the node current residual is calculated, and the current balance constraint loss of all monitoring nodes within a preset time period is determined based on the node current residual.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the voltage transformer error prediction model construction method as described in any one of claims 1 to 4, or the error assessment method as described in any one of claims 5 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the voltage transformer error prediction model construction method as described in any one of claims 1 to 4, or the error assessment method as described in any one of claims 5 to 6.

Images