Comprehensive measurement and control method, device and medium for box transformer

Abstract

The application belongs to the field of wind power generation, and discloses a kind of box transformer comprehensive measurement and control method, device and medium, through the collection of box transformer electrical, environment, mechanical vibration and insulation state parameter of multi-source heterogeneous sensor, the weighted clustering algorithm of attention mechanism optimization is introduced to realize data classification, the threshold is dynamically adjusted in combination with equipment health degree evaluation model, the improved Bi-LSTM life prediction model is used to predict the remaining life of key components, and the communication scheduling strategy based on priority and network bandwidth adaptive matching is designed.

Description

Technical Field

[0001] This manual relates to the field of wind power generation technology, and in particular to a method, device and medium for integrated measurement and control of transformer substations. Background Technology

[0002] With the rapid development of the new energy industry, the scale of photovoltaic power stations and wind power stations is constantly expanding. As the core equipment of the power station, the prefabricated substation undertakes the important task of power conversion and transmission. Since new energy power stations are usually located in harsh outdoor environments, the prefabricated substations are scattered and far from the central control room, which brings many challenges to the measurement and control work.

[0003] Traditional methods for monitoring and controlling prefabricated substations often rely on distributed equipment for data collection, resulting in low data processing efficiency and delayed fault warnings. While integrated monitoring and control devices exist in current technologies, they often fail to adequately consider the varying importance of different parameters during data classification and processing, leading to the overloading of critical data. Threshold settings are frequently fixed, unable to adapt to dynamic changes in equipment operating conditions. Furthermore, inefficient allocation of communication resources can cause transmission delays that impact processing efficiency for critical fault information. In addition, existing technologies lack the ability to predict the lifespan of critical components in prefabricated substations, hindering early fault warnings and preventative maintenance, thus affecting the reliability and economic efficiency of substation operation.

[0004] Therefore, there is an urgent need for a comprehensive measurement and control solution for prefabricated substations that can achieve accurate multi-dimensional data processing, dynamic threshold adjustment, lifespan prediction and early warning, and adaptive communication scheduling to solve the above-mentioned technical problems. Summary of the Invention

[0005] This invention addresses the problems existing in the prior art by proposing a comprehensive measurement and control method, device, and medium for prefabricated substations. It collects electrical, environmental, mechanical vibration, and insulation status parameters of the prefabricated substations through multi-source heterogeneous sensors, introduces an attention mechanism-optimized weighted clustering algorithm to achieve data classification, dynamically adjusts thresholds in conjunction with an equipment health assessment model, uses an improved Bi-LSTM lifetime prediction model to predict the remaining lifetime of key components, and designs a communication scheduling strategy based on priority and network bandwidth adaptive matching.

[0006] To achieve the above objectives, this application provides the following technical solution:

[0007] In a first aspect, a comprehensive measurement and control method for a prefabricated substation is characterized by comprising the following steps:

[0008] S1, real-time acquisition of multi-dimensional operational data of the transformer substation using a distributed sensor array, including electrical parameters, environmental parameters, equipment status parameters, and insulation status parameters; S2, preprocessing and feature extraction of the multi-dimensional operational data; S3, classification of the preprocessed multi-dimensional operational data based on an improved weighted K-means clustering algorithm, assigning different weights according to the importance of each data category to obtain several data categories; Step S3 includes: S3.1, constructing an improved weighted K-means clustering algorithm, introducing an attention mechanism to dynamically adjust parameter weights; S3.2, calculating the contribution of each parameter to the impact of transformer substation faults through an attention network, and correcting the weight values ​​in real time, as shown in the following formula: (2), where, The corrected weights for the i-th type of parameters; Basic weights; The sum of attention scores for all parameters; For attention scores, training is conducted based on the correlation between historical fault data and parameters; S3.3, the number of cluster categories is set, the clustering effectiveness is verified by the silhouette coefficient, and various data categories are output; S4, for each data category, a dynamic threshold model is constructed by combining the historical operating data and fault records of the transformer substation. The dynamic threshold model updates the threshold range in real time using a sliding window algorithm; S5, based on the improved Bi-LSTM life prediction model of transfer learning-digital twin fusion, the remaining life of key components of the transformer substation is calculated according to the multidimensional operating data and the output results of the dynamic threshold model, and fault warning signals are generated; S6, based on the urgency of the fault warning signals and the importance of data categories, an adaptive communication scheduling strategy is adopted to allocate communication resources and prioritize the transmission of high-priority data.

[0009] Optionally, step S2 includes:

[0010] S2.1, adopts an outlier detection algorithm based on isolated forest to remove invalid data, and combines Kalman filtering to smooth high-frequency fluctuation data such as vibration and current;

[0011] S2.2, the time-domain and frequency-domain features of the mechanical vibration signal are extracted by wavelet packet transform, and the harmonic features of the electrical parameters are extracted by Fourier transform to form a high-dimensional feature vector;

[0012] S2.3, The high-dimensional feature vector is standardized by unifying its dimensions based on the improved Z-Score normalization method, and parameter importance weight correction normalization is introduced. The specific formula is as follows:

[0013] (1),

[0014] in, These are the original parameter values; The weights of the i-th type of parameters; The mean of the parameters; The standard deviation of the parameter; These are the standardized parameter values.

[0015] Optionally, step S4 includes the following steps:

[0016] S4.1 Construct a health assessment model for the transformer substation equipment. Input the feature vector of the multidimensional operating data, historical fault repair records, and equipment operating years into the health assessment model for the transformer substation equipment for calculation. Use the analytic hierarchy process (AHP) combined with the entropy weight method to calculate the combined weights and output the health index.

[0017] S4.2 introduces an environmental stress factor to quantify the dynamic influence of temperature, humidity, atmospheric pressure, and electromagnetic interference intensity on the threshold parameters of the transformer substation.

[0018] S4.3, Establish a fault mode correlation matrix to quantify the mapping relationship between different fault types and parameter thresholds;

[0019] S4.4, combining the health index, comprehensive environmental stress factor and failure mode correlation matrix, constructs a three-dimensional coupled dynamic threshold model;

[0020] S4.5 introduces a threshold robustness verification mechanism, using Monte Carlo simulation to verify the false alarm rate of the threshold under different combinations of environmental stress and health status.

[0021] Optionally, in step S4.2, the temperature stress factor is:

[0022] (3)

[0023] Where T is the real-time ambient temperature; Standard reference temperature; This is the temperature influence coefficient;

[0024] Humidity stress factor:

[0025] (4)

[0026] Where H represents the real-time relative humidity; Standard reference humidity; Humidity influence coefficient;

[0027] Comprehensive environmental stress factor:

[0028] (5)

[0029] Wherein, EMI stands for Electromagnetic Interference Intensity; is a coefficient.

[0030] Optionally, step S5 includes:

[0031] S5.1, based on the three-dimensional model of the transformer substation, design parameters, material properties, and real-time acquired data, a digital twin is constructed;

[0032] S5.2, construct an improved Bi-LSTM lifetime prediction model. In the input layer, the physical acquisition features and digital twin virtual features are fused, and a multi-scale attention fusion mechanism is introduced to strengthen the weight of key features. In the hidden layer, a double hidden layer structure is adopted, and the output parameters of the improved Bi-LSTM lifetime prediction model are optimized based on the transfer learning training model. The output is the remaining lifetime and failure mode probability distribution of the key components of the transformer.

[0033] S5.3 establishes a four-level early warning mechanism based on the deviation between remaining lifetime and dynamic threshold, combined with digital twin simulation.

[0034] Optionally, in step S6, the communication scheduling strategy further includes a bandwidth prediction mechanism. The bandwidth prediction mechanism uses the ARIMA model to predict network bandwidth changes in the future, adjusts data transmission priority and compression rate in advance, and avoids parameter delays caused by bandwidth fluctuations.

[0035] Secondly, the present invention provides a comprehensive monitoring and control device for transformer substations, comprising:

[0036] The data acquisition module is used to collect multi-dimensional operating data of the transformer in real time. The data acquisition module includes a current sensor, a voltage sensor, a temperature and humidity sensor, an oil level sensor, an oil temperature sensor, and a switch status detector.

[0037] The data preprocessing module is used to perform outlier removal and standardization on the multidimensional operational data;

[0038] The data classification module is used to classify preprocessed multidimensional running data based on an improved weighted K-means clustering algorithm;

[0039] The threshold modeling module is used to build dynamic threshold models and update the threshold range in real time;

[0040] The life prediction and early warning module is used to predict the remaining life of key components of the transformer substation using equipment operation life prediction algorithms and generate fault early warning signals.

[0041] The communication scheduling module is used to allocate communication resources and transmit data based on an adaptive communication scheduling strategy;

[0042] The processor is used to execute the functions of the above modules;

[0043] The memory is used to store multi-dimensional operational data, historical fault records, model parameters, and program instructions.

[0044] Optional, also includes:

[0045] The digital twin modeling module uses the Unity3D engine to build a digital twin of the transformer substation, supports real-time data-driven state mapping, and has virtual fault simulation and intervention strategy simulation functions.

[0046] The transfer learning training module, equipped with an edge computing chip, supports adaptive training with an adaptive number of source and target domains.

[0047] The environmental stress monitoring unit integrates an electromagnetic interference sensor and a high-precision barometric pressure sensor to collect environmental stress parameters in real time and provide input for the three-dimensional coupled dynamic threshold model.

[0048] The early warning and self-healing execution module includes a relay output interface, a fan cooling control interface, and a load adjustment interface, which can execute the optimal intervention strategy generated by the digital twin.

[0049] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the integrated measurement and control method for transformer substations as described in the first aspect.

[0050] The beneficial effects of this invention are as follows:

[0051] 1. By collecting electrical, environmental, mechanical vibration and insulation status parameters of the transformer substation through multi-source heterogeneous sensors, a weighted clustering algorithm with attention mechanism optimization is introduced to classify the data. The threshold is dynamically adjusted in combination with the equipment health assessment model. An improved Bi-LSTM lifetime prediction model is used to predict the remaining lifetime of key components. A communication scheduling strategy based on priority and network bandwidth adaptive matching is designed.

[0052] 2. This invention solves the problems of single data dimension, fixed threshold, low life prediction accuracy and waste of communication resources in traditional measurement and control. It enables early warning of transformer substation failures up to 72 hours in advance and reduces the latency of key data transmission to within 50ms. It is suitable for complex operating conditions of new energy power plants such as high altitude and strong electromagnetic interference, and improves the reliability and operation and maintenance efficiency of transformer substations. Attached Figure Description

[0053] This specification will be further described by way of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. The same numbers in the drawings denote the same structures or steps.

[0054] Figure 1 This is a schematic diagram of a comprehensive measurement and control method for a transformer substation according to one embodiment of this application.

[0055] Figure 2 For the purposes of this application Figure 1The diagram shows a three-dimensional coupled dynamic threshold modeling in one embodiment.

[0056] Figure 3 For the purposes of this application Figure 1 The diagram illustrates an improved Bi-LSTM lifetime prediction and hierarchical early warning system based on transfer learning-digital twin fusion in one embodiment.

[0057] Figure 4 For the purposes of this application Figure 1 The diagram shows a comparison of dynamic threshold adjustment in one embodiment. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description of this application is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely one preferred embodiment of this application and are only used to explain this application. They do not limit the scope of protection of this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0059] Example 1:

[0060] like Figure 1-4 As shown, a comprehensive measurement and control method for a transformer substation includes the following steps:

[0061] S1. A distributed sensor array is used to collect multi-dimensional operational data of the transformer substation in real time. This multi-dimensional operational data includes electrical parameters, environmental parameters, equipment status parameters, and insulation status parameters. The electrical parameters include three-phase current, three-phase voltage, power factor, active power, reactive power, harmonic content, and frequency. The environmental parameters include the internal temperature of the transformer substation, external ambient temperature, relative humidity, and atmospheric pressure. The equipment status parameters include the transformer core vibration amplitude and frequency, winding vibration signal, and mechanical displacement of the switch operating mechanism. The insulation status parameters include transformer oil dielectric loss, dissolved gas content in the oil, and bushing insulation resistance. The sensor array's acquisition frequency is dynamically adjusted according to the parameter type: the electrical parameter acquisition frequency is 50Hz / cycle with 64 sampling points, the mechanical status parameter acquisition frequency is 1kHz, and the insulation status parameter acquisition frequency is once / hour.

[0062] S2, preprocess and extract features from the multidimensional operational data;

[0063] S3, based on the improved weighted K-means clustering algorithm, classifies the preprocessed multidimensional running data, assigns different weights according to the importance of each data category, and obtains several data categories;

[0064] S4. For each data category, a dynamic threshold model is constructed by combining the historical operating data and fault records of the transformer substation. The dynamic threshold model updates the threshold range in real time through a sliding window algorithm.

[0065] S5, based on the improved Bi-LSTM life prediction model of transfer learning-digital twin fusion, calculates the remaining life of key components of the transformer substation and generates fault warning signals according to the output results of multi-dimensional operating data and dynamic threshold model.

[0066] S6. Based on the urgency of the fault warning signal and the importance of the data category, an adaptive communication scheduling strategy is adopted to allocate communication resources and prioritize the transmission of high-priority data.

[0067] This embodiment collects electrical, environmental, mechanical vibration, and insulation status parameters of the transformer substation using multi-source heterogeneous sensors. An attention-based optimized weighted clustering algorithm is introduced for data classification. Thresholds are dynamically adjusted using an equipment health assessment model. An improved Bi-LSTM life prediction model is employed to predict the remaining lifespan of critical components. A communication scheduling strategy based on priority and network bandwidth adaptive matching is designed. This embodiment addresses the problems of single data dimension, fixed thresholds, low lifespan prediction accuracy, and wasted communication resources in traditional measurement and control systems. It achieves 72-hour early warning of transformer substation failures, reduces critical data transmission latency to within 50ms, and is suitable for complex operating conditions in new energy power plants such as high-altitude areas and areas with strong electromagnetic interference, improving the reliability and maintenance efficiency of transformer substation operations.

[0068] In one specific embodiment, in step S1, the distributed sensor array adopts an anti-electromagnetic interference design, the current sensor adopts the Rogowski coil principle, the vibration sensor adopts a piezoelectric accelerometer, and the dissolved gas sensor in oil adopts photoacoustic spectroscopy technology.

[0069] In one specific embodiment, step S2 includes:

[0070] S2.1, adopts an outlier detection algorithm based on isolated forest to remove invalid data, and combines Kalman filtering to smooth high-frequency fluctuation data such as vibration and current;

[0071] S2.2, the time-domain and frequency-domain features of the mechanical vibration signal are extracted by wavelet packet transform, and the harmonic features of the electrical parameters are extracted by Fourier transform to form a high-dimensional feature vector;

[0072] S2.3, The high-dimensional feature vector is standardized by unifying its dimensions based on the improved Z-Score normalization method, and parameter importance weight correction normalization is introduced. The specific formula is as follows:

[0073] (1),

[0074] in, These are the original parameter values; The weight of the i-th type of parameter ranges from 0.1 to 0.8. The mean of the parameters; The standard deviation of the parameter; These are the standardized parameter values.

[0075] In one specific embodiment, step S3 includes:

[0076] S3.1, Construct an improved weighted K-means clustering algorithm and introduce an attention mechanism to dynamically adjust parameter weights; wherein, the basic weight of electrical parameters is 0.5, the basic weight of mechanical state parameters is 0.3, the basic weight of insulation state parameters is 0.15, and the basic weight of environmental parameters is 0.05.

[0077] S3.2, the contribution of each parameter to the impact of transformer substation faults is calculated using an attention network, and the weight values ​​are adjusted in real time. The specific formula is shown below:

[0078] (2),

[0079] in, The corrected weights for the i-th type of parameters; Basic weights; The sum of attention scores for all parameters; n is the total number of parameter categories; Attention scores are obtained through training based on the correlation between historical fault data and parameters;

[0080] S3.3 sets the number of clusters to 6-8, verifies the effectiveness of clustering through silhouette score, and outputs various data categories, such as current and voltage, vibration characteristics, insulation status, power, environment, and switching status.

[0081] In one specific embodiment, step S3, the training process of the attention network includes:

[0082] 1. Construct a training dataset, collect parameter data of several sets of transformers under normal and fault conditions, and label the fault type and related parameters;

[0083] 2. The attention network is trained using the Adam optimizer, and the loss function is cross-entropy loss, so that the attention score is positively correlated with the contribution of the parameters to the fault.

[0084] 3. Optimize network parameters through multi-fold cross-validation, and stop training when the validation set accuracy is ≥92%.

[0085] In one specific embodiment, step S4 includes:

[0086] S4.1 Construct a health assessment model for the transformer substation equipment. Input the feature vector of the multidimensional operating data, historical fault repair records, and equipment operating years into the health assessment model for the transformer substation equipment for calculation. Use the analytic hierarchy process (AHP) combined with the entropy weight method to calculate the combined weights and output the health index.

[0087] S4.2 introduces an environmental stress factor to quantify the dynamic influence of temperature, humidity, atmospheric pressure, and electromagnetic interference intensity on the threshold parameters of the transformer substation.

[0088] S4.3, Establish a fault mode correlation matrix to quantify the mapping relationship between different fault types and parameter thresholds, matrix elements The correlation strength between the i-th type of fault and the j-th type of parameter is represented by the parameter, which is obtained based on multiple sets of fault data and Bayesian network training.

[0089] S4.4, combining the health index, comprehensive environmental stress factor and failure mode correlation matrix, constructs a three-dimensional coupled dynamic threshold model;

[0090] S4.5 introduces a threshold robustness verification mechanism, using Monte Carlo simulation to verify the false alarm rate of the threshold under different combinations of environmental stress and health status. When the false alarm rate is >3%, the window length and threshold coefficient are automatically adjusted.

[0091] The three-dimensional coupled dynamic threshold model can adapt to different environmental stresses and fault modes. Under extreme conditions (-35℃, humidity 85%, electromagnetic interference intensity 0.8), the threshold adjustment response time is ≤1 second, the false alarm rate is reduced from the original 8%-12% to ≤3%, and the fault false alarm rate is ≤0.5%, which is 80% lower than the original model.

[0092] Further, in step S4.2, the temperature stress factor is:

[0093] (3)

[0094] Where T is the real-time ambient temperature; Standard reference temperature; This is the temperature influence coefficient;

[0095] Humidity stress factor:

[0096] (4)

[0097] Where H represents the real-time relative humidity; Standard reference humidity; Humidity influence coefficient;

[0098] Comprehensive environmental stress factor:

[0099] (5)

[0100] Wherein, EMI stands for Electromagnetic Interference Intensity; is a coefficient.

[0101] Furthermore, in step S4.4, the length of the sliding window is dynamically adjusted, as shown in the following formula:

[0102] (6)

[0103] in, The baseline window length is 30 cycles of electrical parameters; HI is the health index; SF is the comprehensive environmental stress factor.

[0104] The threshold coefficient is adaptive, and the calculation formula is shown below:

[0105] (7)

[0106] (8)

[0107] in, The baseline threshold coefficient is denoted by m, which represents the total number of fault types. This is the fault mode correlation matrix;

[0108] The threshold range is calculated using the following formula:

[0109] (9)

[0110] (10)

[0111] in, , These are the upper and lower limits of the dynamic threshold, respectively; The exponentially weighted mean of the data within the window; The standard deviation is the index-weighted average. These are all threshold coefficients, dynamically adjusted based on the health index. The higher the HI, the better. The larger the value, the better.

[0112] In step S4, the hierarchical analysis method (AHP) judgment matrix construction rules for the equipment health assessment model are as follows: the target layer is equipment health; the criterion layer consists of electrical health sub-indices, mechanical health sub-indices, insulation health sub-indices, and environmental adaptability sub-indices; the indicator layer contains 3-5 specific parameters for each sub-indice, and the indicator weights are determined by expert scoring. Specific parameters, such as current harmonic distortion rate, voltage deviation, and power factor, are included in the electrical health sub-index.

[0113] In one specific embodiment, step S5 includes:

[0114] S5.1, based on the three-dimensional model of the transformer substation, design parameters, material properties, and real-time acquired data, a digital twin is constructed;

[0115] S5.2, construct an improved Bi-LSTM lifetime prediction model. In the input layer, the physical acquisition features and digital twin virtual features are fused, and a multi-scale attention fusion mechanism is introduced to strengthen the weight of key features. In the hidden layer, a double hidden layer structure is adopted, and the output parameters of the improved Bi-LSTM lifetime prediction model are optimized based on the transfer learning training model. The output is the remaining lifetime and failure mode probability distribution of the key components of the transformer.

[0116] S5.3, based on the remaining lifespan and the deviation of the dynamic threshold, combined with digital twin simulation, sets up a four-level early warning mechanism. For early warnings of level three and below, the optimal intervention strategy is automatically generated by simulating the effects of different intervention measures (such as load adjustment and heat dissipation optimization) through digital twin simulation, and then sent to the transformer substation control system for execution to delay the evolution of the fault.

[0117] By setting up step S5, the life prediction accuracy MAPE is improved by 37.5% compared to the original model (MAPE≤8%), the fault mode recognition accuracy is ≥95%, the fault location time for maintenance personnel is shortened from 2 hours to 10 minutes, the early warning time is extended from 72 hours to 96 hours, the new self-healing mechanism reduces the potential fault delay rate by ≥40%, and the model has strong adaptability to small sample scenarios. Only 2 years of real data are needed for the target transformer to achieve stable accuracy (traditional models require more than 5 years of data).

[0118] Further, in step S5.2, the input layer integrates physical acquisition features (trend change rate of high-dimensional feature vectors, dynamic threshold deviation) and digital twin virtual features (virtual fault evolution features, simulated lifetime decay features), totaling 128 dimensions of input features; a feature enhancement layer is added, introducing a multi-scale attention fusion mechanism to extract attention weights for short-term (within 1 hour), medium-term (within 24 hours), and long-term (within 7 days) features, as shown in the following formula:

[0119] (11)

[0120] in, Features at different scales; For training parameters; For feature dimensions.

[0121] Further, in step S5.2, the hidden layer adopts a double hidden layer structure, introducing a hybrid architecture of gated recurrent unit (GRU) and bidirectional long short-term memory (Bi-LSTM) to reduce gradient vanishing and enhance the ability to capture long sequence dependencies. The transfer learning training model uses the real operation data of the same type of box-type substation as the source domain data, and the real data + digital twin virtual data of the target box-type substation as the target domain data. The model parameters are optimized through the domain adaptation loss function to improve the generalization ability in the small sample scenario. The domain adaptation loss function is specifically shown in the following formula:

[0122] (12)

[0123] Where,[[]]END]] is the remaining useful life prediction loss; is the domain difference loss; By = 0.3 to balance the two parts of the loss, enabling the model to quickly adapt to the target box-type substation.

[0124] Further, in step S5.2, the output layer adopts a multi-task learning design, and simultaneously outputs the remaining useful life (RUL) of key components (windings, switch mechanisms, insulating bushings) and the probability distribution of failure modes (such as the probability of arc discharge, the probability of core failure, etc.).

[0125] In a specific embodiment, in step S5.3, the four-level early warning mechanism includes:

[0126] a) Level 1 early warning (urgent): The remaining useful life RUL ≤ 24 hours or any parameter exceeds 10% / is lower than 10%, triggering a trip signal and an audible and visual alarm;

[0127] b) Level 2 early warning (important): 24 hours < remaining useful life RUL ≤ 72 hours or the parameter exceeds 5% - 10% / is lower than 5% - 10%, sending an operation and maintenance work order to the centralized control room (including fault location, repair plan, spare parts list);

[0128] c) Level 3 early warning (prompt): 72 hours < remaining useful life RUL ≤ 96 hours or the parameter is close to the threshold range ( -5% ~ or ~ +5%), recording the data change trend and generating suggestions for improving the health status;

[0129] d) Level 4 early warning (potential risk): 96 hours < RUL ≤ 168 hours and the digital twin simulation shows a deteriorating trend of the parameter, triggering regular special monitoring (the acquisition frequency is doubled).

[0130] In one specific embodiment, in step S6, the data is divided into three priorities: highest priority (Level 1 warning signals, tripping commands, remaining lifespan), allocated 30%-50% bandwidth, using UDP protocol with redundant transmission; medium priority (Level 2 warning signals, electrical parameters, mechanical status parameters), allocated 20%-30% bandwidth, using TCP protocol; and low priority (Level 3 warning signals, environmental parameters, insulation status parameters (non-real-time), allocated 10%-20% bandwidth, using batch transmission with compression algorithm. Specifically, network bandwidth changes are monitored in real time, and when the bandwidth is below a threshold, the sampling frequency of low-priority data is automatically reduced to prioritize the transmission of high-priority data.

[0131] In one specific embodiment, in step S6, the communication scheduling strategy further includes a bandwidth prediction mechanism. The bandwidth prediction mechanism uses the ARIMA model to predict network bandwidth changes in the future, adjusts data transmission priority and compression rate in advance, and avoids parameter delays caused by bandwidth fluctuations.

[0132] Communication scheduling phase: The specific formula for the transmission priority factor is as follows:

[0133] (13)

[0134] in, Let x be the transmission priority factor for the x-th data packet; This is the normalization function; Let be the cosine similarity between the x-th and y-th data packets; Y is the total number of data packets. The criticality of information in the x-th data packet; This is the key acquisition time sequence for the x / y-th data packet.

[0135] Example 2:

[0136] A comprehensive monitoring and control device for a transformer substation includes:

[0137] The data acquisition module is used to collect multi-dimensional operating data of the transformer in real time. The data acquisition module includes a current sensor, a voltage sensor, a temperature and humidity sensor, an oil level sensor, an oil temperature sensor, and a switch status detector.

[0138] The data preprocessing module is used to perform outlier removal and standardization on the multidimensional operational data;

[0139] The data classification module is used to classify preprocessed multidimensional running data based on an improved weighted K-means clustering algorithm;

[0140] The threshold modeling module is used to build dynamic threshold models and update the threshold range in real time;

[0141] The life prediction and early warning module is used to predict the remaining life of key components of the transformer substation using equipment operation life prediction algorithms and generate fault early warning signals.

[0142] The communication scheduling module is used to allocate communication resources and transmit data based on an adaptive communication scheduling strategy;

[0143] The processor is used to execute the functions of the above modules;

[0144] The memory is used to store multi-dimensional operational data, historical fault records, model parameters, and program instructions.

[0145] Specifically, the sensor interfaces of the data acquisition module adopt a standardized design: the current / voltage signal interface is a BNC connector (anti-interference), the vibration signal interface is an M12 aviation plug, the gas sensor interface is RS485, and the dielectric loss test interface is a dedicated high-voltage interface (withstand voltage ≥2000V). All interfaces are designed to prevent mis-insertion.

[0146] Specifically, the adaptive communication scheduling module also supports edge computing and cloud collaboration; it completes high-priority data processing and early warning locally, and uploads low-priority data to the cloud platform (supporting the MQTT protocol) to achieve historical data tracing and big data analysis.

[0147] Example 3:

[0148] This embodiment provides a computer-readable storage medium storing instructions that, when executed on a computer or processor, cause the computer or processor to perform the above-described instructions. Figure 1 One or more steps in the illustrated embodiment. If the constituent modules of the above-described electronic device are implemented as software functional units and sold or used as independent products, they can be stored in the computer-readable storage medium.

[0149] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this specification are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in or transmitted through a computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Versatile Discs (DVDs)), or semiconductor media (e.g., Solid State Disks (SSDs)).

[0150] Those skilled in the art will understand that all or part of the processes in the method of Embodiment 1 described above can be implemented by a computer program instructing related hardware. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks. Unless otherwise specified, the technical features of this embodiment and the implementation scheme can be combined arbitrarily.

[0151] The above-described specific embodiments are preferred embodiments of the integrated measurement and control method, device and medium for transformer substations of this application, and are not intended to limit the specific scope of this application. The scope of this application includes but is not limited to the specific embodiments described above. All equivalent changes made in accordance with the shape and structure of this application are within the protection scope of this application.

Claims

1. A comprehensive measurement and control method for transformer substations, characterized in that, Includes the following steps: S1, real-time acquisition of multi-dimensional operating data of the transformer substation through a distributed sensor array, the multi-dimensional operating data including electrical parameters, environmental parameters, equipment status parameters and insulation status parameters; S2, preprocess and extract features from the multidimensional operational data; S3, classifying the preprocessed multidimensional running data based on the improved weighted K-means clustering algorithm, assigning different weights according to the importance of each data category, and obtaining several data categories; step S3 includes: S3.1, Construct an improved weighted K-means clustering algorithm and introduce an attention mechanism to dynamically adjust parameter weights; S3.2, the contribution of each parameter to the impact of transformer substation faults is calculated using an attention network, and the weight values ​​are adjusted in real time. The specific formula is shown below: （2）， in, The corrected weights for the i-th type of parameters; Basic weights; The sum of attention scores for all parameters; Attention scores are obtained through training based on the correlation between historical fault data and parameters; S3.3 sets the number of cluster categories, verifies the clustering effectiveness using the silhouette coefficient, and outputs the data categories for each type. S4. For each data category, a dynamic threshold model is constructed by combining the historical operating data and fault records of the transformer substation. The dynamic threshold model updates the threshold range in real time through a sliding window algorithm. S5. Based on the improved Bi-LSTM lifetime prediction model of transfer learning-digital twin fusion, the remaining lifetime of key components of the transformer substation is calculated according to the multi-dimensional operating data and the output results of the dynamic threshold model, and a fault early warning signal is generated. S6. Based on the urgency of the fault warning signal and the importance of the data category, an adaptive communication scheduling strategy is adopted to allocate communication resources and prioritize the transmission of high-priority data.

2. The integrated measurement and control method for transformer substations according to claim 1, characterized in that, Step S2 includes: S2.1, adopts an outlier detection algorithm based on isolated forest to remove invalid data, and combines Kalman filtering to smooth high-frequency fluctuation data such as vibration and current; S2.2, the time-domain and frequency-domain features of the mechanical vibration signal are extracted by wavelet packet transform, and the harmonic features of the electrical parameters are extracted by Fourier transform to form a high-dimensional feature vector; S2.3, The high-dimensional feature vector is standardized by unifying its dimensions based on the improved Z-Score normalization method, and parameter importance weight correction normalization is introduced. The specific formula is as follows: （1）， in, These are the original parameter values; The weights of the i-th type of parameters; The mean of the parameters; The standard deviation of the parameter; These are the standardized parameter values.

3. The integrated measurement and control method for transformer substations according to claim 1, characterized in that, Step S4 includes: S4.1 Construct a health assessment model for the transformer substation equipment. Input the feature vector of the multidimensional operating data, historical fault repair records, and equipment operating years into the health assessment model for the transformer substation equipment for calculation. Use the analytic hierarchy process (AHP) combined with the entropy weight method to calculate the combined weights and output the health index. S4.2 introduces an environmental stress factor to quantify the dynamic influence of temperature, humidity, atmospheric pressure, and electromagnetic interference intensity on the threshold parameters of the transformer substation. S4.3, Establish a fault mode correlation matrix to quantify the mapping relationship between different fault types and parameter thresholds; S4.4, combining the health index, comprehensive environmental stress factor and failure mode correlation matrix, constructs a three-dimensional coupled dynamic threshold model; S4.5 introduces a threshold robustness verification mechanism, using Monte Carlo simulation to verify the false alarm rate of the threshold under different combinations of environmental stress and health status.

4. The integrated measurement and control method for prefabricated substations according to claim 3, characterized in that, In step S4.2, the temperature stress factor is: （3） Where T is the real-time ambient temperature; Standard reference temperature; This is the temperature influence coefficient; Humidity stress factor: （4） Where H represents the real-time relative humidity; Standard reference humidity; Humidity influence coefficient; Comprehensive environmental stress factor: （5） Wherein, EMI stands for Electromagnetic Interference Intensity; is a coefficient.

5. The integrated measurement and control method for transformer substations according to claim 1, characterized in that, Step S5 includes: S5.1, based on the three-dimensional model of the transformer substation, design parameters, material properties, and real-time acquired data, a digital twin is constructed; S5.2, construct an improved Bi-LSTM lifetime prediction model. In the input layer, the physical acquisition features and digital twin virtual features are fused, and a multi-scale attention fusion mechanism is introduced to strengthen the weight of key features. In the hidden layer, a double hidden layer structure is adopted, and the output parameters of the improved Bi-LSTM lifetime prediction model are optimized based on the transfer learning training model. The output is the remaining lifetime and failure mode probability distribution of the key components of the transformer. S5.3 establishes a four-level early warning mechanism based on the deviation between remaining lifetime and dynamic threshold, combined with digital twin simulation.

6. The integrated measurement and control method for prefabricated substations according to claim 1, characterized in that, In step S6, the communication scheduling strategy also includes a bandwidth prediction mechanism. The bandwidth prediction mechanism uses the ARIMA model to predict network bandwidth changes in the future, adjusts data transmission priority and compression rate in advance, and avoids parameter delays caused by bandwidth fluctuations.

7. A comprehensive monitoring and control device for a prefabricated substation, characterized in that, include: The data acquisition module is used to collect multi-dimensional operating data of the transformer in real time. The data acquisition module includes a current sensor, a voltage sensor, a temperature and humidity sensor, an oil level sensor, an oil temperature sensor, and a switch status detector. The data preprocessing module is used to perform outlier removal and standardization on the multidimensional operational data; The data classification module is used to classify preprocessed multidimensional running data based on an improved weighted K-means clustering algorithm; The threshold modeling module is used to build dynamic threshold models and update the threshold range in real time. The life prediction and early warning module is used to predict the remaining life of key components of the transformer substation using equipment operation life prediction algorithms and generate fault early warning signals. The communication scheduling module is used to allocate communication resources and transmit data based on an adaptive communication scheduling strategy; The processor is used to execute the functions of the above modules; The memory is used to store multi-dimensional operational data, historical fault records, model parameters, and program instructions.

8. The integrated monitoring and control device for transformer substations according to claim 7, characterized in that, Also includes: The digital twin modeling module uses the Unity3D engine to build a digital twin of the transformer substation, supports real-time data-driven state mapping, and has virtual fault simulation and intervention strategy simulation functions. The transfer learning training module, equipped with an edge computing chip, supports adaptive training with an adaptive number of source and target domains. The environmental stress monitoring unit integrates an electromagnetic interference sensor and a high-precision barometric pressure sensor to collect environmental stress parameters in real time and provide input for the three-dimensional coupled dynamic threshold model. The early warning and self-healing execution module includes a relay output interface, a fan cooling control interface, and a load adjustment interface, which can execute the optimal intervention strategy generated by the digital twin.

9. A 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 integrated measurement and control method for transformer substations as described in any one of claims 1-6.

Images