An early monitoring method for nonlinear power failure of a table area based on an internet of things table

Abstract

The application belongs to the technical field of power equipment, and particularly relates to an early monitoring method for nonlinear transformer area power failure based on an internet-of-things table and application thereof. The early monitoring method comprises the following steps: S1: obtaining power consumption data generated by each power user, and distinguishing nonlinear load users; S2: normalizing and manually marking the power consumption data; S3: constructing a multivariate data feature library; S4: using an improved SVM algorithm model of adaptive particle swarm optimization as a data classification network, and training the data classification network; S5: verifying the data classification network; S6: implanting the data classification network into the internet-of-things table to identify the power consumption data; S7: reporting abnormal data to an intelligent fusion terminal after the abnormal data are identified; S8: uploading historical data and abnormal states to a transformer area control center; and S9: judging the abnormal states of the transformer area by the transformer area control center, and notifying operation and maintenance personnel to investigate and dispose. The application can timely find power grid operation failure caused by the grid connection of nonlinear load users.

Description

Technical Field

[0001] This invention belongs to the field of power equipment technology, specifically relating to an early monitoring method for nonlinear transformer area power faults based on IoT meters, an abnormal data identification module, an early monitoring system for nonlinear transformer area power faults near the meter side, and an early monitoring system for nonlinear transformer area power faults near the main station side. Background Technology

[0002] With the development of smart grids, smart IoT meters have gradually become widely used. These meters possess the metering functions of traditional electricity meters, as well as data processing, monitoring, control, sensing, interaction, and routing capabilities. They are well-suited to meet the needs of the Internet of Things (IoT), laying the foundation for remote meter reading and monitoring in smart grids. As a cornerstone of smart grids and a key terminal product in the construction of new power systems, smart IoT meters play a crucial supporting role in achieving "informatization, automation, and interactivity" in smart grids. The research and development and implementation of IoT meters can greatly enhance the construction and regulation capabilities of smart grids.

[0003] Against the backdrop of the construction of new power systems and the energy internet, the operation of the power grid is facing two major new changes. On the supply side, the grid connection penetration rate of clean energy sources, primarily wind and solar power, is continuously increasing, meaning the power system needs to accommodate more fluctuating power sources. On the demand side, with the increasing penetration of electric vehicles into the automotive market and the widespread adoption of more diverse smart home appliances, electricity demand is becoming more unevenly distributed in time and space. These new types of electricity users, which significantly impact the supply and demand of the power grid, are called nonlinear load users. As more and more nonlinear load users are connected to the grid, they generate a large number of harmonics, which may cause voltage and current distortions, seriously affecting the power supply quality and the safety of grid operation. In this new context, how to effectively monitor the operating status of nonlinear load users, detect power anomalies in the grid earlier, and take timely measures before failures or losses occur has become a new challenge for those skilled in the art. Summary of the Invention

[0004] To reduce the impact of nonlinear load users connecting to the grid on the stability and security of the power grid, this invention provides an early monitoring method for nonlinear transformer area power faults based on IoT meters, an abnormal data identification module, an early monitoring system for nonlinear transformer area power faults near the meter side, and an early monitoring system for nonlinear transformer area power faults near the main station side.

[0005] This invention is achieved using the following technical solution:

[0006] An early monitoring method for nonlinear transformer substation power faults based on IoT meters is proposed. This method uses metering data collected from IoT meters installed at power user nodes to identify and address faults in power substations containing typical nonlinear load users. The early monitoring method for nonlinear transformer substation power faults comprises the following steps:

[0007] S1: Acquire electricity consumption data generated by each power user in the distribution area from the IoT meter's metering core, and distinguish the nonlinear load users included in it based on the proportion of electricity content.

[0008] S2: Normalize the electricity consumption data generated by nonlinear load users, and manually label normal and abnormal data to obtain the required feature sequence.

[0009] S3: Construct a multivariate data feature library containing normal and abnormal data, and divide it into training and test sets according to a preset ratio.

[0010] S4: An improved adaptive particle swarm optimization SVM algorithm model is used as the data classification network, and the data classification network is trained using the training set to solve for the objective function value of the globally optimal individual.

[0011] S5: Validate the trained data classification network using the test set until the data classification network identifies all anomalous data in the multivariate data feature library.

[0012] S6: The data classification network from the previous step is embedded into the IoT meter installed at the corresponding node of the nonlinear load user through an embedded module, for the purpose of identifying the electricity consumption data collected by the IoT meter's metering core.

[0013] S7: After any IoT meter monitors and identifies abnormal data, it generates an abnormal status flag and reports it to the smart fusion terminal.

[0014] S8: After receiving an abnormal status flag uploaded by any IoT meter in the distribution area, the intelligent fusion terminal extracts the historical data of the corresponding IoT meter and uploads the historical data and the abnormal status flag to the server of the distribution area control center simultaneously.

[0015] S9: The control center makes a preliminary judgment on the abnormal status of the area based on the status flags and historical data uploaded by each IoT meter in the area, and dispatches maintenance personnel to conduct on-site investigation and handling when necessary.

[0016] As a further improvement of the present invention, in step S1, typical nonlinear load users include: distributed photovoltaic power stations, electric vehicle charging stations, manufacturing plants, commercial buildings, and residential centralized charging stations. The types of electricity consumption data collected by the IoT metering chip include three-phase voltage value U, three-phase current value I, and daily load power W.

[0017] As a further improvement of the present invention, in step S2, the normalized values ​​of voltage, current, and daily load power are all within the range of [0,1]. The normalization expression used is:

[0018]

[0019] In the above formula, max|V|, max|I|, and max|w| are the absolute values ​​of the historical maximum values ​​of voltage, current, and daily load power collected by the IoT meter's metering core under normal conditions, respectively; V n I n W n These represent the voltage, current, and daily load power values ​​at the nth sampling moment of the IoT meter's metering core; v n i n w n These are the normalized voltage, current, and daily load power values ​​at the nth sampling time, respectively.

[0020] As a further improvement of the present invention, in step S3, the ratio of the training set to the test set divided by the multivariate data feature library is 6:4.

[0021] As a further improvement of the present invention, in step S4, the data classification network uses an adaptive particle swarm optimization algorithm to assist the support vector machine in parameter optimization during the training phase. The optimization parameters mainly include the penalty parameter C and the kernel parameter σ. 2 And the loss function. The initialization parameters of the data classification network include the inertia factor ω, learning factors c1 and c2, and initial particle positions, expressed by the following formula:

[0022]

[0023] In the above formula, k is the number of particle iterations, d = 1, 2, ..., D represents the spatial dimension; i = 1, 2, ..., N represents the number of particles in the population; R1 and R2 represent random values ​​in the range [0, 1]; c1 and c2 are learning factors, which adjust the step size of flight in the pb and gb directions, respectively; pb is the optimal position of the individual particle; gb is the global optimal position. Let i be the position of particle i in d-dimensional space during the k-th iteration; Let be the velocity of particle i in dimension d during the (k+1)th iteration.

[0024] As a further improvement of this invention, the parameter optimization process of the adaptive particle swarm optimization algorithm during the training phase of the data classification network is as follows:

[0025] S01: An elite back-learning strategy is introduced to initialize the individual quality of the population, thereby improving the global convergence speed and solution quality of the particle swarm optimization algorithm; the expression for the elite back-learning strategy is:

[0026] S2 = rand(ub-lb) - S1

[0027] In the above formula, rand is a random number, ub is the upper limit of the independent variable, lb is the lower limit of the independent variable; S1 represents the initial particle population; S2 represents the reverse particle population.

[0028] S02: Use the initialization function to generate an initial population, then generate its reverse population, and then merge the two populations.

[0029] S03: Solve for the fitness value of each individual particle in the population, sort the fitness values ​​from smallest to largest, and select the top 1 / 2 of the solutions as the initial population.

[0030] S04: Use the best fitness value among the particles as the global optimum.

[0031] S05: Based on the different convergence requirements at different stages, a polynomial mutation strategy is applied, using the generation distance obtained from the optimal solution set of two adjacent iterations as a reference index for the adaptive mutation scale adjustment mechanism, and mutation operations are carried out in stages.

[0032] S06: Compare the fitness value of each particle with the historical best; if it is better, replace it.

[0033] S07: Compare the fitness value of each particle with the historical global best; if it is better, replace it.

[0034] S08: Repeat steps S03-S07 until the preset termination condition is reached, and then apply the particle optimal solution obtained from the iteration to the SVM kernel parameters for optimization.

[0035] As a further improvement of this invention, in the training phase of the data classification network, in order to achieve better classification results, the support vector machine prediction model uses the mean squared error (MSE) as a performance evaluation function, the expression of which is:

[0036]

[0037] Where n represents the number of data samples, y i Represents the true value of the sample. This represents the predicted value of the sample. The smaller the mean square error, the better the prediction accuracy and reliability.

[0038] The present invention also includes an abnormal data identification module, which includes 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 constructs a data classification network as described in the aforementioned method for early monitoring of nonlinear transformer substation power faults based on IoT meters.

[0039] The present invention also includes a nonlinear transformer substation power fault early monitoring system located near the meter side, which is used to monitor the operating status of transformer substations containing nonlinear load users and promptly detect early faults during the operation of the substations. This type of nonlinear transformer substation power fault early monitoring system located near the meter side includes: an IoT meter, a data classification module, an intelligent fusion terminal, and a master station server.

[0040] The IoT meters are installed in each power node within the distribution area. Each IoT meter contains a metering chip for measuring the electricity consumption data of each power user. The IoT meters store and report the collected metering results.

[0041] The data classification module employs the aforementioned abnormal data identification module, which is embedded in the IoT meters located at the power nodes corresponding to each nonlinear load user within the distribution area. The data classification module identifies the metering data collected by the IoT meters and generates a corresponding abnormal status flag when abnormal data is identified.

[0042] The intelligent fusion terminal communicates with all IoT meters within the distribution area to acquire metering and status data collected by each IoT meter and report it to the main station server in the distribution area control center. The intelligent fusion terminal also extracts historical data from any IoT meter upon receiving an anomaly status flag uploaded by any IoT meter within the distribution area, and synchronously uploads the historical data and the anomaly status flag to the main station server.

[0043] The main station server is used to acquire metering and status data from various IoT meters uploaded by smart converged terminals in different distribution areas, and to perform statistical analysis on the data. The main station server also runs a fault analysis tool, which calculates the predicted probability of various fault states that may occur on-site in the distribution area based on historical data from each IoT meter that reports abnormal status flags. The main station server then issues maintenance instructions to designated maintenance personnel based on the prediction results from the fault analysis tool.

[0044] The present invention also includes an early monitoring system for nonlinear transformer substation power faults located near the master station. This system monitors the operating status of transformer substations containing nonlinear load users and promptly detects early faults during substation operation. The early monitoring system for nonlinear transformer substation power faults located near the master station includes: IoT meters, smart fusion terminals, and a master station server.

[0045] The IoT meters are installed in each power node within the distribution area. Each IoT meter contains a metering chip for measuring the electricity consumption data of each power user. The IoT meters store and report the collected metering results.

[0046] The intelligent fusion terminal communicates with all IoT meters in the distribution area to acquire metering data and status data collected by each IoT meter and report them to the main station server of a distribution area control center.

[0047] The main server is used to acquire metering and status data from various IoT meters uploaded by smart converged terminals in different distribution areas, and to perform statistical analysis on the data. The main server runs a data classification module and a fault analysis tool. The data classification module uses the aforementioned abnormal data classification network to identify metering data collected by the IoT meters. When abnormal data is identified, the data classification model generates a corresponding abnormal status flag. The fault analysis tool uses the acquired abnormal status flags to query historical data for each IoT meter, and then calculates the predicted probability of various fault states that may occur in the distribution area. Based on the prediction results of the fault analysis tool, the main server issues maintenance instructions to designated maintenance personnel.

[0048] The technical solution provided by this invention has the following beneficial effects:

[0049] The invention proposes an early monitoring method for nonlinear transformer area power faults based on IoT meters. This method can monitor and process multi-dimensional power data generated after nonlinear load users are connected to the grid in real time. It also uses a data classification network based on particle swarm optimization (SVM) strategy to classify the power data generated by nonlinear load users. Finally, it predicts and identifies potential power faults based on the abnormal states and data of different users in the transformer area.

[0050] The data classification network designed in this invention, based on a particle swarm optimization (SVM) strategy, can effectively utilize multi-data information collected from IoT tables. Furthermore, the designed network model architecture and training strategy improve the convergence speed of the network model, avoiding "jumping into" local optima.

[0051] This invention utilizes a designed data classification network to monitor operational data collected at the site in real time, either at the meter side or the master station side. This enables remote self-identification of the smart IoT meter's wiring status and greatly improves the efficiency of identifying abnormal data, laying the foundation for early identification and handling of power grid operation faults. Attached Figure Description

[0052] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0053] Figure 1 This is a flowchart illustrating the steps of an early monitoring method for nonlinear transformer area power faults based on IoT meters, as provided in Embodiment 1 of the present invention.

[0054] Figure 2 This is a time sequence diagram of the training strategy adopted by the data classification network designed in Embodiment 1 of the present invention during the training phase.

[0055] Figure 3 This is a system architecture diagram of a nonlinear transformer area power fault early monitoring system near the meter side provided in Embodiment 3 of the present invention.

[0056] Figure 4 This is a topology diagram of the nonlinear transformer area power fault early monitoring system near the main station side provided in Embodiment 4 of the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0058] Example 1

[0059] This embodiment provides an early monitoring method for nonlinear transformer substation power faults based on IoT meters. It is used to identify and handle faults in power substations containing typical nonlinear load users at an early stage, based on metering data collected by IoT meters installed in power user nodes. For example... Figure 1 As shown, the early monitoring method for power faults in this nonlinear distribution area follows these steps:

[0060] S1: Acquires electricity consumption data generated by each power user within the distribution area from the IoT meter's metering core, and distinguishes non-linear load users based on their proportion of electricity content. Typical non-linear load users include: distributed photovoltaic power stations, electric vehicle charging stations, manufacturing plants, commercial buildings, and residential centralized charging stations.

[0061] The data collected by the IoT metering core are diverse. Specifically, the types of electricity data required for early monitoring of nonlinear power faults in transformer substations include three-phase voltage values ​​U, three-phase current values ​​I, and daily load power W.

[0062] S2: Normalize the electricity consumption data generated by nonlinear load users, and manually label normal and abnormal data to obtain the required feature sequence.

[0063] The normalized values ​​of voltage, current, and daily load power are all within the range of [0,1]. The normalization expression used is as follows:

[0064]

[0065] In the above formula, max|V|, max|I|, and max|w| are the absolute values ​​of the historical maximum values ​​of voltage, current, and daily load power collected by the IoT meter's metering core under normal conditions, respectively; V n I n W n These represent the voltage, current, and daily load power values ​​at the nth sampling moment of the IoT meter's metering core; v n i n w n These are the normalized voltage, current, and daily load power values ​​at the nth sampling time, respectively.

[0066] S3: Construct a multivariate data feature library containing normal and abnormal data, and divide it into training and test sets according to a preset ratio. Specifically, in this embodiment, the training and test sets are divided in a 6:4 ratio.

[0067] S4: An improved adaptive particle swarm optimization SVM algorithm model is used as the data classification network, and the data classification network is trained using the training set to solve for the objective function value of the globally optimal individual.

[0068] like Figure 2 As shown, the data classification network designed in this embodiment uses an adaptive particle swarm optimization algorithm to assist the support vector machine in parameter optimization during the training phase. The optimization parameters mainly include the penalty parameter C and the kernel parameter σ. 2 And the loss function. The initialization parameters of the data classification network include the inertia factor ω, learning factors c1 and c2, and initial particle positions, expressed by the following formula:

[0069]

[0070] In the above formula, k is the number of particle iterations, d = 1, 2, ..., D represents the spatial dimension; i = 1, 2, ..., N represents the number of particles in the population; R1 and R2 represent random values ​​in the range [0, 1]; c1 and c2 are learning factors, which adjust the step size of flight in the pb and gb directions, respectively; pb is the optimal position of the individual particle; gb is the global optimal position. Let i be the position of particle i in d-dimensional space during the k-th iteration; Let be the velocity of particle i in dimension d during the (k+1)th iteration.

[0071] During the training phase of the data classification network, the parameter optimization process of the adaptive particle swarm optimization algorithm is as follows:

[0072] S01: An elite back-learning strategy is introduced to initialize the individual quality of the population, thereby improving the global convergence speed and solution quality of the particle swarm optimization algorithm; the expression for the elite back-learning strategy is:

[0073] S2 = rand(ub-lb) - S1

[0074] In the above formula, rand is a random number, ub is the upper limit of the independent variable, lb is the lower limit of the independent variable; S1 represents the initial particle population; S2 represents the reverse particle population.

[0075] S02: Use the initialization function to generate an initial population, then generate its reverse population, and then merge the two populations.

[0076] S03: Solve for the fitness value of each individual particle in the population, sort the fitness values ​​from smallest to largest, and select the top 1 / 2 of the solutions as the initial population.

[0077] S04: Use the best fitness value among the particles as the global optimum.

[0078] S05: Based on the different convergence requirements at different stages, a polynomial mutation strategy is applied, using the generation distance obtained from the optimal solution set of two adjacent iterations as a reference index for the adaptive mutation scale adjustment mechanism, and mutation operations are carried out in stages.

[0079] S06: Compare the fitness value of each particle with the historical best; if it is better, replace it.

[0080] S07: Compare the fitness value of each particle with the historical global best; if it is better, replace it.

[0081] S08: Repeat steps S03-S07 until the preset termination condition is reached, and then apply the particle optimal solution obtained from the iteration to the SVM kernel parameters for optimization.

[0082] During the training phase of the data classification network, in order to achieve better classification results, the Support Vector Machine prediction model uses the Mean Squared Error (MSE) as the performance evaluation function, the expression of which is:

[0083]

[0084] Where n represents the number of data samples, y i Represents the true value of the sample. This represents the predicted value of the sample. The smaller the mean square error, the better the prediction accuracy and reliability.

[0085] S5: Validate the trained data classification network using the test set until the data classification network identifies all anomalous data in the multivariate data feature library.

[0086] S6: The data classification network from the previous step is embedded into the IoT meter installed at the corresponding node of the nonlinear load user through an embedded module, for the purpose of identifying the electricity consumption data collected by the IoT meter's metering core.

[0087] S7: After any IoT meter monitors and identifies abnormal data, it generates an abnormal status flag and reports it to the smart fusion terminal.

[0088] S8: After receiving an abnormal status flag uploaded by any IoT meter in the distribution area, the intelligent fusion terminal extracts the historical data of the corresponding IoT meter and uploads the historical data and the abnormal status flag to the server of the distribution area control center simultaneously.

[0089] S9: The control center makes a preliminary judgment on the abnormal status of the area based on the status flags and historical data uploaded by each IoT meter in the area, and dispatches maintenance personnel to conduct on-site investigation and handling when necessary.

[0090] The technical solution provided in this embodiment mainly utilizes IoT metering chips to collect historical electricity consumption data from various electricity users, and simultaneously constructs a multivariate data feature library. The acquired data is then filtered and normalized into feature sequences. A network model capable of classifying data is then trained using this multivariate feature library. The network model designed in this embodiment primarily initializes the position and velocity parameters of individuals in the population by introducing an elite back-learning strategy into the particle swarm optimization algorithm. Then, a multinomial mutation strategy is applied at different iteration stages to perform mutation operations in stages, obtaining the globally optimal particle. Finally, the optimal example is input into an SVM classifier for iterative solution, focusing on optimizing the model kernel parameters and improving the global convergence speed of the algorithm.

[0091] In this embodiment, the algorithm model trained by the improved particle swarm optimization SVM strategy is used as the final data classification network. Then, the voltage and current information of each phase under the meter wiring status is input into the network model for identification. The network model compares the normalized real-time feature sequence with the feature vector in the multivariate data feature library, and selects the feature code that matches the feature sequence in the multivariate data feature library, thereby accurately predicting whether the real-time power consumption data of nonlinear load users is abnormal.

[0092] All abnormal electricity consumption data will eventually be aggregated at the distribution area control center. The distribution area control center will then identify potential early faults and their corresponding fault types based on the status data and historical metering data of different nodes.

[0093] Example 2

[0094] Building upon the existing embodiments, this embodiment further provides an abnormal data identification module, which includes 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 constructs a data classification network, similar to the early monitoring method for nonlinear transformer area power faults based on IoT meters in Embodiment 1. In this embodiment, the abnormal data identification module running a specific computer program is essentially a computer device capable of performing specific data processing tasks. This computer device can be an embedded computer device or a general-purpose computer device capable of executing programs.

[0095] The general-purpose computer equipment in this embodiment includes tablet computers, laptop computers, desktop computers, rack servers, blade servers, tower servers, or cabinet servers (including standalone servers or server clusters composed of multiple servers). The computer equipment in this embodiment includes, but is not limited to, memory and processors that can be interconnected via a system bus.

[0096] In this embodiment, the memory (i.e., the readable storage medium) includes flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory can be an internal storage unit of a computer device, such as the hard disk or RAM of the computer device. In other embodiments, the memory can also be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the computer device. Of course, the memory can also include both internal storage units and external storage devices of the computer device. In this embodiment, the memory is typically used to store the operating system and various application software installed on the computer device. In addition, the memory can also be used to temporarily store various types of data that have been output or will be output.

[0097] In some embodiments, the processor may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor is typically used to control the overall operation of a computer device. In this embodiment, the processor is used to run program code stored in memory or process data.

[0098] Example 3

[0099] Based on the foregoing embodiments, this embodiment further provides an early monitoring system for nonlinear transformer area power faults located near the meter side. This system is used to monitor the operating status of transformer areas containing nonlinear load users and promptly detect early faults during transformer area operation. For example... Figure 3 As shown, this type of nonlinear transformer area power fault early monitoring system near the meter includes: IoT meter, data classification module, intelligent fusion terminal, and main station server.

[0100] The IoT meters are installed in each power node within the distribution area. Each IoT meter contains a metering chip for measuring the electricity consumption data of each power user. The IoT meters store and report the collected metering results.

[0101] The data classification module adopts the abnormal data identification module as described in Example 2. The data classification module is embedded in the IoT meters located at the power nodes corresponding to each nonlinear load user within the distribution area. The data classification module is used to identify the metering data collected by the IoT meters and generate a corresponding abnormal status flag when abnormal data is identified.

[0102] The intelligent fusion terminal communicates with all IoT meters within the distribution area to acquire metering and status data collected by each IoT meter and report it to the main station server in the distribution area control center. The intelligent fusion terminal also extracts historical data from any IoT meter upon receiving an anomaly status flag uploaded by any IoT meter within the distribution area, and synchronously uploads the historical data and the anomaly status flag to the main station server.

[0103] The main station server is used to acquire metering and status data from various IoT meters uploaded by smart converged terminals in different distribution areas, and to perform statistical analysis on the data. The main station server also runs a fault analysis tool, which calculates the predicted probability of various fault states that may occur on-site in the distribution area based on historical data from each IoT meter that reports abnormal status flags. The main station server then issues maintenance instructions to designated maintenance personnel based on the prediction results from the fault analysis tool.

[0104] Example 4

[0105] Building upon Embodiments 1 and 2, this embodiment further provides an early monitoring system for nonlinear transformer substation power faults located near the master station. This system monitors the operational status of transformer substations containing nonlinear load users and promptly detects early faults during substation operation. Unlike Embodiment 3, this embodiment's early monitoring system for nonlinear transformer substation power faults located near the master station includes: IoT meters, smart fusion terminals, and a master station server.

[0106] The IoT meters are installed in each power node within the distribution area. Each IoT meter contains a metering chip for measuring the electricity consumption data of each power user. The IoT meters store and report the collected metering results.

[0107] The intelligent fusion terminal communicates with all IoT meters in the distribution area to acquire metering data and status data collected by each IoT meter and report them to the main station server of a distribution area control center.

[0108] The main server is used to acquire metering and status data from various IoT meters uploaded by smart converged terminals in different distribution areas, and to perform statistical analysis on the data. The main server runs a data classification module and a fault analysis tool. The data classification module uses the aforementioned abnormal data classification network to identify metering data collected by the IoT meters. When abnormal data is identified, the data classification model generates a corresponding abnormal status flag. The fault analysis tool uses the acquired abnormal status flags to query historical data for each IoT meter, and then calculates the predicted probability of various fault states that may occur in the distribution area. Based on the prediction results of the fault analysis tool, the main server issues maintenance instructions to designated maintenance personnel.

[0109] like Figure 4 As shown, in practical applications, the early monitoring system for nonlinear transformer area power faults near the main station consists of two parts: a data identification module software and a data interaction interface software. The data interaction interface software's specific functions include: periodically receiving information such as voltage, current, and daily load power from IoT meters; using a big data computing engine to perform data filtering, cleaning, integration, reduction, and transformation to eliminate noisy, conflicting, and missing data, ensuring data accuracy, authenticity, and completeness. Subsequently, it transmits the power data of nonlinear load users to the data classification module and receives the analysis results of each IoT meter from the data classification module's feedback. After organizing and summarizing these analysis results, they are displayed on the data interaction interface, including the meter number, quantity, and status flags of each IoT meter. Simultaneously, when the status flag of any IoT meter is abnormal, the status data and historical metering data of the corresponding transformer area's IoT meters are transmitted to the fault analysis tool. The fault analysis tool calculates the predicted probability of various transformer area faults and issues corresponding maintenance detailed orders when necessary.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for early monitoring of nonlinear transformer substation power faults based on IoT meters, characterized in that, It is used to identify and handle faults in power distribution areas containing typical nonlinear load users in the early stage based on the metering data collected by the IoT meters installed in the power user nodes; the early monitoring method for power faults in nonlinear distribution areas is as follows: S1: Acquire electricity consumption data generated by each power user in the distribution area from the IoT metering chip, and distinguish the nonlinear load users included in it according to the proportion of electricity content. S2: Normalize the electricity consumption data generated by nonlinear load users, and manually label normal and abnormal data to obtain the required feature sequence; S3: Construct a multivariate data feature library containing normal and abnormal data, and divide it into training and test sets according to a preset ratio; S4: An improved adaptive particle swarm optimization SVM algorithm model is used as the data classification network, and the data classification network is trained using the training set to solve for the objective function value of the globally optimal individual. S5: Validate the trained data classification network using the test set until the data classification network identifies all anomalous data in the multivariate data feature library; S6: The data classification network from the previous step is embedded into the IoT meter installed at the corresponding node of the nonlinear load user through an embedded module, which is used to identify the electricity consumption data collected by the IoT meter's metering core. S7: After any IoT meter monitors and identifies abnormal data, it generates an abnormal status flag and reports it to the smart fusion terminal; S8: After receiving an abnormal status flag uploaded by any IoT meter in the distribution area, the intelligent fusion terminal extracts the historical data of the corresponding IoT meter and uploads the historical data and the abnormal status flag to the server of the distribution area control center simultaneously. S9: The control center makes a preliminary judgment on the abnormal status of the area based on the status flags and historical data uploaded by each IoT meter in the area, and dispatches maintenance personnel to conduct on-site investigation and handling when necessary.

2. The early monitoring method for nonlinear transformer area power faults based on IoT meters as described in claim 1, characterized in that: In step S1, typical nonlinear load users include: distributed photovoltaic power stations, electric vehicle charging stations, manufacturing plants, commercial buildings, and residential centralized charging stations; The types of electricity data collected by the IoT metering chip include three-phase voltage value U, three-phase current value I, and daily load power W.

3. The early monitoring method for nonlinear transformer area power faults based on IoT meters as described in claim 2, characterized in that: In step S2, the normalized values ​​of voltage, current, and daily load power are all within the range of [0,1]. The normalization expression used is: In the above formula, max|V|, max|I|, and max|w| are the absolute values ​​of the historical maximum values ​​of voltage, current, and daily load power collected by the IoT meter's metering core under normal conditions, respectively; V n I n W n These represent the voltage, current, and daily load power values ​​at the nth sampling moment of the IoT meter's metering core; v n i n w n These are the normalized voltage, current, and daily load power values ​​at the nth sampling time, respectively.

4. The early monitoring method for nonlinear transformer area power faults based on IoT meters as described in claim 3, characterized in that: In step S3, the ratio of training set to test set data from the multivariate data feature library is 6:

4.

5. The early monitoring method for nonlinear transformer area power faults based on IoT meters as described in claim 1, characterized in that: In step S4, the data classification network uses an adaptive particle swarm optimization algorithm to assist the support vector machine in parameter optimization during the training phase. The optimization parameters include the penalty parameter C and the kernel parameter σ. 2 The initialization parameters of the data classification network include the inertia factor ω, learning factors c1 and c2, and initial particle positions, expressed as follows: In the formula, k is the number of particle iterations, d = 1, 2, ..., D represents the spatial dimension; i = 1, 2, ..., N represents the number of particles in the population; R1 and R2 represent random values ​​in the range [0, 1]; c1 and c2 are learning factors, which adjust the step size of flight in the pb and gb directions, respectively; pb is the optimal position of the individual particle; gb is the global optimal position. Let i be the position of particle i in d-dimensional space during the k-th iteration; Let be the velocity of particle i in dimension d during the (k+1)th iteration.

6. The early monitoring method for nonlinear transformer area power faults based on IoT meters as described in claim 5, characterized in that: During the training phase of the data classification network, the parameter optimization process of the adaptive particle swarm optimization algorithm is as follows: S01: An elite back-learning strategy is introduced to initialize the individual quality of the population, thereby improving the global convergence speed and solution quality of the particle swarm optimization algorithm; the expression for the elite back-learning strategy is: S2 = rand(ub-lb) - S1 In the above formula, rand is a random number, ub is the upper limit of the independent variable, lb is the lower limit of the independent variable; S1 represents the initial particle population; S2 represents the reverse particle population; S02: Use the initialization function to generate an initial population, then generate its reverse population, and then merge the two populations; S03: Solve for the fitness value of each individual particle in the population, sort the fitness values ​​from smallest to largest, and select the top 1 / 2 of the solutions as the initial population. S04: Use the best fitness value among the particles as the global optimum; S05: Based on the different convergence requirements at different stages, a polynomial mutation strategy is applied, using the generation distance obtained from the optimal solution sets of two adjacent iterations as a reference index for the adaptive mutation scale adjustment mechanism, and mutation operations are performed in stages: S06: Compare the fitness value of each particle with the historical best; if it is better, replace it. S07: Compare the fitness value of each particle with the historical global best; if it is better, replace it. S08: Repeat steps S03-S07 until the preset termination condition is reached, and then apply the particle optimal solution obtained from the iteration to the SVM kernel parameters for optimization.

7. The early monitoring method for nonlinear transformer area power faults based on IoT meters as described in claim 1, characterized in that: During the training phase of the data classification network, in order to achieve better classification results, the Support Vector Machine prediction model uses the Mean Squared Error (MSE) as the performance evaluation function, the expression of which is: Where n represents the number of data samples, y i Represents the true value of the sample. This represents the predicted value of the sample. The smaller the mean square error, the better the prediction accuracy and reliability.

8. An abnormal data identification module, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it constructs a data classification network in the early monitoring method for nonlinear transformer area power faults based on IoT tables as described in any one of claims 1-7.

9. A nonlinear transformer substation power fault early monitoring system located near the meter side, characterized in that: Used to monitor the operating status of transformer substations containing nonlinear load users and to detect early faults in the operation of the substations in a timely manner; The nonlinear transformer area power fault early monitoring system near the meter includes: The IoT meter is installed in each power node within the distribution area. Each IoT meter contains a metering chip for measuring the electricity consumption data of each power user. The IoT meter stores and reports the collected metering results. The data classification module adopts the abnormal data identification module as described in claim 8. The data classification module is embedded in the IoT meter in the power node corresponding to each nonlinear load user in the distribution area. The data classification module is used to identify the metering data collected by the IoT meter and generate a corresponding abnormal status flag when abnormal data is identified. The intelligent fusion terminal communicates with all IoT meters within the distribution area to acquire metering and status data collected by each IoT meter and report it to the main station server of the distribution area control center. The intelligent fusion terminal also extracts historical data from any IoT meter upon receiving an abnormal status flag uploaded by any IoT meter within the distribution area, and synchronously uploads the historical data and the abnormal status flag to the main station server. The main station server is used to acquire metering and status data of various IoT meters uploaded by smart converged terminals in different distribution areas, and to perform statistical analysis on the data. The main station server also runs a fault analysis tool, which is used to calculate the predicted probability of various fault states that may occur on-site in the distribution area based on the historical data of each IoT meter that reports abnormal status flags. The main station server issues operation and maintenance instructions to the designated operation and maintenance personnel based on the prediction results of the fault analysis tool.

10. A nonlinear transformer substation power fault early monitoring system near the main station side, which is used to monitor the operating status of transformer substations containing nonlinear load users and detect early faults in the operation of the transformer substation in a timely manner. The nonlinear transformer area power fault early monitoring system near the main station includes: The IoT meter is installed in each power node within the distribution area. Each IoT meter contains a metering chip for measuring the electricity consumption data of each power user. The IoT meter stores and reports the collected metering results. The intelligent fusion terminal communicates with all IoT meters in the distribution area to acquire metering data and status data collected by each IoT meter and report them to the main station server of the distribution area control center. The main station server is used to acquire metering and status data of various IoT meters uploaded by smart converged terminals in different distribution areas, and to perform statistical analysis on the data. The main station server runs a data classification module and a fault analysis tool. The data classification module uses the abnormal data classification network as described in claim 8 and is used to identify the metering data collected by the IoT meters. When abnormal data is identified, the data classification model generates a corresponding abnormal status flag. The fault analysis tool is used to query the historical data of each IoT meter based on the acquired abnormal status flags, and then calculate the predicted probability of various fault states that may occur on-site in the distribution area. The main station server issues maintenance instructions to designated maintenance personnel based on the prediction results of the fault analysis tool.

Images