Power transmission and transformation grid monitoring method and system based on Internet of Things

Abstract

The invention relates to the technical field of intelligent power grids, in particular to a power transmission and transformation grid monitoring method and system based on the Internet of Things, and the method comprises the steps: representing the noise complexity of an initial data matrix through the reconstruction difference evaluation of the initial data matrix formed by monitoring data vectors, thereby determining the number of denoising iterations, and improving the denoising precision. According to the method and the system, the target data matrix is subjected to fault identification, iteration denoising is carried out through the denoising model, the expected denoising effect can be achieved with the number of denoising iterations as small as possible, and therefore on the premise that the reliability of the target data matrix is guaranteed, that is, on the premise that the accuracy of power grid monitoring is guaranteed, denoising time consumption is reduced, and in addition, fault identification is carried out in a matrix mode. The power grid early warning information is determined according to the obtained fault identification probability vector, and compared with a time sequence prediction model iterative prediction mode in the prior art, time consumed by iterative prediction does not need to be introduced, and the timeliness of power grid monitoring is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of smart grids, and particularly to a monitoring method and system for transmission and transformation power grids based on the Internet of Things. Background Art

[0002] In the operation of modern transmission and transformation power grid substations, the stability and security of grid equipment are crucial for the reliable operation of the power system. With the continuous expansion of the grid scale and the improvement of equipment complexity, the real-time monitoring and fault warning of equipment have become increasingly important. With the rapid development of the Internet of Things technology, new possibilities have been provided for the health monitoring of grid equipment. By using a variety of sensors to monitor the operating status and environmental conditions of equipment in real time, comprehensive data support has been provided for the safe operation of the power grid.

[0003] Traditional monitoring methods for transmission and transformation power grid substations usually rely on a single data source or simple rule judgment to achieve equipment status monitoring, and it is difficult to effectively integrate various data types. When facing complex equipment failures and changing environmental conditions, traditional monitoring methods cannot accurately judge the health status of equipment, resulting in delays or misjudgments in fault identification.

[0004] According to the above situation, the prior art has proposed a data monitoring method based on an artificial intelligence model, which usually realizes the prediction of equipment data based on a time series prediction model, and then realizes the early prediction of faults. However, the time series prediction model usually adopts an iterative prediction method. In the scenario of multiple grid equipment, the prediction time will increase due to the increase in data volume, making it difficult to meet the timeliness requirements of grid monitoring.

[0005] In addition, the data collected by sensors may have various types of noise. Without knowing the type of noise, it is difficult to perform denoising processing on the data through an effective denoising algorithm, which will reduce the reliability of the data collected by sensors, and further reduce the accuracy of grid monitoring.

[0006] Therefore, how to improve the accuracy and timeliness of grid monitoring has become an urgent problem to be solved. Summary of the Invention

[0007] In view of the above technical problems, the technical solution adopted by the present invention is a monitoring method for transmission and transformation power grids based on the Internet of Things. The monitoring method for transmission and transformation power grids based on the Internet of Things includes the following steps: S101, forming an initial data matrix from the monitoring data vectors respectively obtained at M preset time points, where M is a positive integer.

[0008] S102, performing masking processing on the initial data matrix to obtain a masked data matrix.

[0009] S103. Input the masked data matrix into the trained reconstruction model to obtain a reconstructed data matrix.

[0010] S104. Determine a reconstruction difference evaluation value based on the initial data matrix and the reconstructed data matrix.

[0011] S105. Map the reconstruction difference evaluation value to the number of denoising iterations according to a preset first mapping function.

[0012] S106. Perform iterative denoising on the initial data matrix through the trained denoising model according to the number of denoising iterations to obtain a target data matrix.

[0013] S107. Input the target data matrix into the trained fault identification model to obtain a fault identification probability vector.

[0014] S108. Generate a power grid warning message when the fault identification probability vector meets the preset conditions.

[0015] The present invention also provides a transmission and transformation power grid monitoring system based on the Internet of Things. The transmission and transformation power grid monitoring system based on the Internet of Things includes: A matrix formation module, configured to form an initial data matrix from monitoring data vectors respectively obtained at M preset time points, where M is a positive integer.

[0016] A matrix masking module, configured to perform masking processing on the initial data matrix to obtain a masked data matrix.

[0017] A matrix reconstruction module, configured to input the masked data matrix into the trained reconstruction model to obtain a reconstructed data matrix.

[0018] A difference evaluation module, configured to determine a reconstruction difference evaluation value based on the initial data matrix and the reconstructed data matrix.

[0019] An iteration number mapping module, configured to map the reconstruction difference evaluation value to the number of denoising iterations according to a preset first mapping function.

[0020] A matrix denoising module, configured to perform iterative denoising on the initial data matrix through the trained denoising model according to the number of denoising iterations to obtain a target data matrix.

[0021] A fault identification module, configured to input the target data matrix into the trained fault identification model to obtain a fault identification probability vector.

[0022] A fault warning module, configured to generate a power grid warning message when the fault identification probability vector meets the preset conditions.

[0023] The present invention has at least the following beneficial effects: By evaluating the reconstruction difference of the initial data matrix formed by the monitoring data vectors, the noise complexity of the initial data matrix is characterized, so as to determine the number of denoising iterations. Then, iterative denoising is performed through the denoising model, and the expected denoising effect can be achieved with as few denoising iterations as possible. Thus, on the premise of ensuring the reliability of the target data matrix, that is, on the premise of ensuring the accuracy of power grid monitoring, the time consumption of the denoising process is reduced. In addition, fault identification is performed in the form of a matrix, and the power grid warning information is determined according to the obtained fault identification probability vector. Compared with the existing technology that uses the time series prediction model for iterative prediction, the time consumption of iterative prediction is not introduced, effectively improving the timeliness of power grid monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0025] Figure 1 It is a flowchart of a method for monitoring a transmission and transformation power grid based on the Internet of Things provided in Embodiment 1 of the present invention; Figure 2 It is a schematic diagram of a system for monitoring a transmission and transformation power grid based on the Internet of Things provided in Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0026] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above drawings are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It can be understood that, under appropriate circumstances, the above terms used to distinguish similar objects can be interchanged, so that the present invention can also implement other embodiments other than the above-described illustrated embodiments or described embodiments. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or server including a series of steps or units does not necessarily have to be limited to those clearly listed steps or units, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0028] Example 1 Example 1 provides a monitoring method for transmission and distribution power grids based on the Internet of Things. As Figure 1 shown, it is a flowchart of a monitoring method for transmission and distribution power grids based on the Internet of Things provided by Embodiment 1 of the present invention. The monitoring method for transmission and distribution power grids based on the Internet of Things includes the following steps: S101, forming an initial data matrix from the monitoring data vectors respectively obtained at M preset time points, where M is a positive integer; S102, performing mask processing on the initial data matrix to obtain a masked data matrix; S103, inputting the masked data matrix into a trained reconstruction model to obtain a reconstructed data matrix; S104, determining a reconstruction difference evaluation value according to the initial data matrix and the reconstructed data matrix; S105, mapping the reconstruction difference evaluation value to the number of denoising iterations according to a preset first mapping function; S106, performing iterative denoising on the initial data matrix through a trained denoising model according to the number of denoising iterations to obtain a target data matrix; S107, inputting the target data matrix into a trained fault identification model to obtain a fault identification probability vector; S108, generating a power grid warning message when the fault identification probability vector meets a preset condition.

[0029] Among them, the preset time point may refer to the acquisition time point of the power grid equipment status. The time interval between adjacent preset time points may be a fixed value, which can be set by the implementer himself, but it should be ensured that the fixed value is greater than or equal to the time taken from obtaining the monitoring data vector at the Mth preset time point to determining whether the fault identification probability vector meets the preset condition.

[0030] The monitoring data vector can be obtained by splicing the monitoring data collected by each sensor, and the initial data matrix can be obtained by splicing M monitoring data vectors.

[0031] The reconstruction difference evaluation value can be used to characterize the noise complexity of the initial data matrix. The prior condition of this embodiment is that the higher the noise complexity, the higher the difficulty of accurately reconstructing the masked data matrix after mask processing of the initial data matrix through the reconstruction model.

[0032] Specifically, the denoising model can be implemented using a diffusion generative model. The architecture of the diffusion generative model can only include a denoising neural network during the usage phase. The architecture of this denoising neural network can adopt a U-Net architecture. The training process of the diffusion generative model will not be elaborated here. It should be noted that in the prior art, the diffusion generative model usually requires multiple iterations to recover the denoised data from an initial data matrix that may contain various types of noise. Obviously, multiple iterations will introduce a high denoising time consumption. In this embodiment, the denoising iteration times are mapped according to the reconstruction difference evaluation value, aiming to limit the excessive denoising iteration times, so as to minimize the denoising iteration times as much as possible on the premise of meeting the denoising effect, thereby improving the timeliness of the overall power grid monitoring.

[0033] If the obtained denoising iteration times still cannot meet the time consumption requirements, the implementer can perform knowledge distillation based on the trained diffusion generative model to reduce the denoising iteration times. If in the prior art, when inputting an initial sample matrix into the trained diffusion generative model for denoising processing, usually P times of iterative denoising are adopted, then use the result of the ((R - 1)×P / Q)-th iterative denoising as the sample, and the result of the (R×P / Q)-th iterative denoising as the label to train a new diffusion generative model. P, Q, and R are all positive integers. Update the diffusion generative model with the new diffusion generative model. Specifically, when R = 1, use the initial sample matrix as the sample. By this means, the denoising iteration times can be effectively reduced, but correspondingly, a certain denoising accuracy will be lost. Therefore, only when the denoising iteration times obtained by mapping according to the reconstruction difference evaluation value cannot meet the time consumption requirements, will the diffusion generative model be updated through knowledge distillation processing.

[0034] In a specific implementation manner, the monitoring data vector includes the monitoring values corresponding to N monitoring data types respectively. The monitoring data types belong to a monitoring data type set, and the monitoring data type set at least includes current, voltage, frequency, vibration, and temperature.

[0035] Among them, the monitoring data type set can also include power belonging to power grid equipment monitoring, as well as humidity, wind speed, air pressure, air quality, etc. belonging to environmental monitoring.

[0036] In a specific implementation manner, the monitoring data vector is a column vector; The masking the initial data matrix to obtain a masked data matrix includes: According to the preset sampling probabilities corresponding to M columns in the initial data matrix, perform weighted random selection on the M columns to obtain reference columns; Set the column vectors corresponding to the reference columns in the initial data matrix to a first preset value to obtain the masked data matrix.

[0037] Among them, the monitoring data vector is a column vector of size N×1, and each column of the monitoring data vector corresponds to a different monitoring data type respectively.

[0038] The preset sampling probability can refer to the weight when randomly selecting the corresponding column, and the first preset value can be 0.

[0039] In a specific implementation manner, the steps for obtaining the preset sampling probabilities corresponding to M columns in the initial data matrix include: For any column in the initial data matrix, according to a preset second mapping function, map the column number corresponding to this column to the reference value corresponding to this column; Perform normalization processing on the reference values corresponding to M columns respectively to obtain the preset sampling probabilities corresponding to M columns respectively.

[0040] Among them, the second mapping function can be w m =m / M, where m is the column number and w m is the reference value of the m-th column.

[0041] Specifically, the normalization processing result of the reference value of the m-th column, that is, the preset sampling probability, can be W m =w m / (∑Me=1(w e ))

[0042] In a specific implementation manner, the trained reconstruction model includes a trained first encoder and a trained first decoder; The step of inputting the masked data matrix into the trained reconstruction model to obtain a reconstructed data matrix includes: Input the masked data matrix into the trained first encoder for feature extraction to obtain a first feature vector; Input the first feature vector into the trained first decoder for feature reconstruction to obtain the reconstructed data matrix.

[0043] Among them, the reconstruction model can also adopt a U-Net architecture. The training samples of the reconstruction model can be a reference data matrix formed by reference monitoring data vectors in the case of theoretical noiselessness, and the training label of the reconstruction model is the reference data matrix itself. The training loss of the reconstruction model can be calculated using a mean squared error loss function.

[0044] It should be noted that if it is difficult to obtain the reference monitoring data vector in the case of theoretical noiselessness, the implementer can also use the result obtained by inputting the initial sample matrix into a trained diffusion generation model for iterative denoising as the reference data matrix.

[0045] In a specific implementation manner, determining the reconstruction difference evaluation value according to the initial data matrix and the reconstructed data matrix includes: Calculating the Euclidean distance between the initial data matrix and the reconstructed data matrix to obtain a difference distance; Performing normalization processing on the difference distance according to a second preset value to obtain the reconstruction difference evaluation value.

[0046] Wherein, the second preset value can be obtained by calculating the Euclidean distance between the initial data matrix and a zero matrix, and the normalization processing may refer to taking the ratio of the difference distance to the second preset value.

[0047] Specifically, when calculating the Euclidean distance between the initial data matrix and the reconstructed data matrix, the implementer can only calculate the Euclidean distance between the column vectors corresponding to the reference column in the initial data matrix and the reconstructed data matrix respectively. Correspondingly, the second preset value can be obtained by calculating the Euclidean distance between the column vector corresponding to the reference column in the initial data matrix and a zero vector.

[0048] In a specific implementation manner, mapping the reconstruction difference evaluation value to the denoising iteration times according to a preset first mapping function includes: Mapping the reconstruction difference evaluation value to an intermediate value according to the first mapping function; Performing ceiling processing on the intermediate value to obtain the denoising iteration times.

[0049] Wherein, the first mapping function can be y = tx, where t can be the maximum denoising iteration times, x is the reconstruction difference evaluation value, and y is the intermediate value.

[0050] In a specific implementation manner, iteratively denoising the initial data matrix through a trained denoising model according to the denoising iteration times to obtain a target data matrix includes: Initializing the temporary iteration times K = 1, and using the initial data matrix as the temporary data matrix; Inputting the temporary data matrix into the trained denoising model to obtain a denoised data matrix; If K is less than the denoising iteration times, using the denoised data matrix as the temporary data matrix, updating K = K + 1, and returning to execute the step of inputting the temporary data matrix into the trained denoising model; If K is equal to the denoising iteration times, using the denoised data matrix finally output by the trained denoising model as the target data matrix.

[0051] In a specific implementation manner, the fault recognition probability vector corresponds to the Mth preset time point; When the fault recognition probability vector meets a preset condition, generating a power grid early warning message, including: Determining I probability change vectors according to the fault recognition probability vector and the historical probability vectors corresponding to the (M - I)-th to (M - 1)-th preset time points respectively, where the probability change vector includes probability change values corresponding to J fault type dimensions respectively, and both I and J are positive integers; If the probability change values corresponding to any fault type dimension in the I probability change vectors are all greater than a third preset value, it is determined that the fault recognition probability vector meets the preset condition, and the power grid early warning message is generated according to this fault type dimension.

[0052] Among them, subtracting the historical probability vectors corresponding to the (M - i + 1)-th and (M - i)-th preset time points respectively to obtain the probability change vector corresponding to the (M - i + 1)-th preset time point, where i is an integer within the range of [2, I], and the third preset value can be set to 0, and the implementer can adjust the third preset value according to the actual situation.

[0053] Specifically, the fault recognition model may include a second encoder and a fully connected layer. The initial parameters of the second encoder may be the same as those of the trained first encoder. Since the trained first encoder is trained based on the reconstruction task of the data matrix, it can be considered that the trained first encoder can learn part of the correlation between different monitoring data types. Making the initial parameters of the second encoder the same as those of the trained first encoder can improve the feature extraction ability of the second encoder, so that compared with training based on randomly initialized parameters of the second encoder, the recognition accuracy of the trained fault recognition model can be improved. The training of the fault recognition model can adopt the training method of a conventional multi-classification model.

[0054] After the monitoring data vector is collected at each preset time point, the fault recognition probability vector corresponding to the preset time point is processed. If the fault recognition probability vector at a certain preset time point does not meet the preset condition, the fault recognition probability vector at this preset time point is stored as the historical probability vector.

[0055] It can be known that compared with the prior art that uses a time series prediction model for fault prediction, this embodiment avoids the time series iterative prediction process, thereby reducing the processing time of power grid monitoring data and improving the timeliness of power grid monitoring. Moreover, during fault prediction, the fault recognition probability vector is obtained based on the actually collected monitoring data vector, and then the probability change trend analysis is performed according to the fault recognition probability vector and the historical probability vector for fault prediction, avoiding the introduction of time series prediction errors and improving the accuracy of fault prediction. And the historical probability vector can be stored in advance, further improving the timeliness of power grid monitoring.

[0056] In the first embodiment, by evaluating the reconstruction difference of the initial data matrix formed by the monitoring data vectors, the noise complexity of the initial data matrix is characterized, so as to determine the number of denoising iterations. Then, through the denoising model for iterative denoising, the expected denoising effect can be achieved with as few denoising iterations as possible. Thus, on the premise of ensuring the reliability of the target data matrix, that is, on the premise of ensuring the accuracy of power grid monitoring, the time consumption of denoising processing is reduced. In addition, for fault identification in matrix form, the power grid warning information is determined according to the obtained fault identification probability vector. Compared with the existing technology that uses the time series prediction model for iterative prediction, there is no need to introduce the time consumption of iterative prediction, effectively improving the timeliness of power grid monitoring.

[0057] Embodiment Two Embodiment Two of the present invention provides a transmission and transformation power grid monitoring system based on the Internet of Things. As Figure 2 shown, it is a schematic diagram of a transmission and transformation power grid monitoring system based on the Internet of Things provided by the second embodiment of the present invention. The transmission and transformation power grid monitoring system based on the Internet of Things includes: A matrix formation module 201, configured to form an initial data matrix from the monitoring data vectors respectively obtained at M preset time points, where M is a positive integer; A matrix masking module 202, configured to perform masking processing on the initial data matrix to obtain a masked data matrix; A matrix reconstruction module 203, configured to input the masked data matrix into a trained reconstruction model to obtain a reconstructed data matrix; A difference evaluation module 204, configured to determine a reconstruction difference evaluation value according to the initial data matrix and the reconstructed data matrix; A number mapping module 205, configured to map the reconstruction difference evaluation value to the number of denoising iterations according to a preset first mapping function; A matrix denoising module 206, configured to perform iterative denoising on the initial data matrix through a trained denoising model according to the number of denoising iterations to obtain a target data matrix; A fault identification module 207, configured to input the target data matrix into a trained fault identification model to obtain a fault identification probability vector; A fault warning module 208, configured to generate power grid warning information when the fault identification probability vector meets a preset condition.

[0058] It should be noted that for the specific limitations of the transmission and transformation power grid monitoring system based on the Internet of Things, reference can be made to the limitations on the transmission and transformation power grid monitoring method based on the Internet of Things in the above text, which will not be elaborated here. For the information interaction, execution process, etc. between the above modules, since they are based on the same concept as the method embodiment of the present invention, their specific functions and the technical effects brought can be specifically referred to the method embodiment part, which will not be elaborated here.

[0059] The above are only the preferred embodiments of the present invention, and do not impose any form of limitation on the present invention. Although the present invention has been disclosed above with the preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art, without departing from the scope of the technical solution of the present invention, can make some changes or modifications to the equivalent embodiments of equivalent changes by using the above-disclosed technical content. However, as long as it does not depart from the content of the technical solution of the present invention, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A monitoring method for transmission and transformation power grids based on the Internet of Things, characterized in that, The method for monitoring the power transmission and transformation network based on the Internet of Things includes the following steps: S101, forming an initial data matrix from the monitoring data vectors respectively obtained at M preset time points, where M is a positive integer; S102, performing masking processing on the initial data matrix to obtain a masked data matrix; S103, inputting the masked data matrix into a trained reconstruction model to obtain a reconstructed data matrix; S104, determining a reconstruction difference evaluation value according to the initial data matrix and the reconstructed data matrix; S105, mapping the reconstruction difference evaluation value to the number of denoising iterations according to a preset first mapping function; S106, performing iterative denoising on the initial data matrix through a trained denoising model according to the number of denoising iterations to obtain a target data matrix; S107, inputting the target data matrix into a trained fault identification model to obtain a fault identification probability vector; S108, generating a power grid warning message when the fault identification probability vector meets a preset condition.

2. The method for monitoring a transmission and distribution power grid based on the Internet of Things according to claim 1, characterized in that The monitoring data vector includes monitoring values corresponding to N monitoring data types respectively, the monitoring data types belong to a monitoring data type set, and the monitoring data type set includes at least current, voltage, frequency, vibration and temperature.

3. The method for monitoring a power transmission and transformation network based on the Internet of Things according to claim 1, wherein The monitoring data vector is a column vector; The performing masking processing on the initial data matrix to obtain a masked data matrix includes: Performing weighted random selection on M columns according to the preset sampling probabilities corresponding to the M columns in the initial data matrix to obtain reference columns; Setting the column vectors corresponding to the reference columns in the initial data matrix to a first preset value to obtain the masked data matrix.

4. The method for monitoring the transmission and transformation power grid based on the Internet of Things according to claim 3, wherein, The steps for obtaining the preset sampling probabilities corresponding to the M columns in the initial data matrix include: For any column in the initial data matrix, mapping the column number corresponding to the column to a reference value corresponding to the column according to a preset second mapping function; Performing normalization processing on the reference values corresponding to the M columns to obtain the preset sampling probabilities corresponding to the M columns.

5. The method for monitoring a power transmission and transformation network based on the Internet of Things according to claim 1, wherein The trained reconstruction model includes a trained first encoder and a trained first decoder; The inputting the masked data matrix into a trained reconstruction model to obtain a reconstructed data matrix includes: Inputting the masked data matrix into the trained first encoder for feature extraction to obtain a first feature vector; Inputting the first feature vector into the trained first decoder for feature reconstruction to obtain the reconstructed data matrix.

6. The method for monitoring the transmission and transformation power grid based on the Internet of Things according to claim 1, characterized in that The determining a reconstruction difference evaluation value according to the initial data matrix and the reconstructed data matrix includes: Calculating the Euclidean distance between the initial data matrix and the reconstructed data matrix to obtain a difference distance; Performing standardization processing on the difference distance according to a second preset value to obtain the reconstruction difference evaluation value.

7. The monitoring method for the transmission and transformation power grid based on the Internet of Things according to claim 1, characterized in that The mapping the reconstruction difference evaluation value to the number of denoising iterations according to a preset first mapping function includes: Mapping the reconstruction difference evaluation value to an intermediate value according to the first mapping function; Performing ceiling processing on the intermediate value to obtain the number of denoising iterations.

8. The method for monitoring a power transmission and transformation network based on the Internet of Things according to claim 1, characterized in that, Performing iterative denoising on the initial data matrix through the trained denoising model according to the denoising iteration times, to obtain a target data matrix, includes: Initializing a temporary iteration times K = 1, and using the initial data matrix as a temporary data matrix; Inputting the temporary data matrix into the trained denoising model to obtain a denoised data matrix; If K is less than the denoising iteration times, using the denoised data matrix as the temporary data matrix, updating K = K + 1, and returning to execute the step of inputting the temporary data matrix into the trained denoising model; If K is equal to the denoising iteration times, using the denoised data matrix finally output by the trained denoising model as the target data matrix.

9. The method for monitoring the transmission and transformation power grid based on the Internet of Things according to claim 1, wherein The fault recognition probability vector corresponds to the Mth preset time point; When the fault recognition probability vector meets the preset conditions, generating a power grid warning message, includes: Determining I probability change vectors according to the fault recognition probability vector and the historical probability vectors respectively corresponding to the (M - I)th preset time point to the (M - 1)th preset time point, where the probability change vector includes probability change values respectively corresponding to J fault type dimensions, and both I and J are positive integers; If the probability change values respectively corresponding to any fault type dimension in the I probability change vectors are all greater than the third preset value, determining that the fault recognition probability vector meets the preset conditions, and generating the power grid warning message according to this fault type dimension.

10. An Internet of Things-based transmission and distribution network monitoring system, characterized in that, The Internet of Things-based transmission and transformation power grid monitoring system includes: A matrix formation module, configured to form an initial data matrix from the monitoring data vectors respectively obtained at M preset time points, where M is a positive integer; A matrix masking module, configured to perform masking processing on the initial data matrix to obtain a masked data matrix; A matrix reconstruction module, configured to input the masked data matrix into the trained reconstruction model to obtain a reconstructed data matrix; A difference evaluation module, configured to determine a reconstruction difference evaluation value according to the initial data matrix and the reconstructed data matrix; A times mapping module, configured to map the reconstruction difference evaluation value to denoising iteration times according to a preset first mapping function; A matrix denoising module, configured to perform iterative denoising on the initial data matrix through the trained denoising model according to the denoising iteration times to obtain a target data matrix; A fault recognition module, configured to input the target data matrix into the trained fault recognition model to obtain a fault recognition probability vector; A fault warning module, configured to generate a power grid warning message when the fault recognition probability vector meets the preset conditions.

Images