Integrated method of transformer inrush current identification and early warning based on multi-sensor fusion

Abstract

The present invention discloses an integrated method for transformer inrush current identification and early warning based on multi-sensor fusion, which relates to the technical field of transformer equipment and includes the following steps: S1, distributed deployment of multiple sensors and construction of an adaptive sensing network; S2, multi-dimensional feature correction and time alignment; S3, extraction of fuzzy characteristics to capture the complex nonlinear relationship in the transformer operating state; S4, fusion of the data after fuzzy feature extraction; S5, construction of an inrush current identification model to accurately identify transformer inrush current; S6, real-time early warning. This integrated method for transformer inrush current identification and early warning based on multi-sensor fusion, by introducing distributed multi-sensor fusion technology, realizes dynamic extraction of multimodal features and deep learning modeling, and combines real-time dynamic threshold calculation, reinforcement learning optimization, and Bayesian fault probability inference to achieve integrated transformer inrush current identification and early warning.

Description

Technical Field

[0001] The present invention relates to the technical field of transformer equipment, and in particular to a transformer inrush current identification and early warning integrated method based on multi-sensor fusion. Background Art

[0002] Transformer inrush current is a common transient phenomenon in power system operation. Because it shares characteristics with fault current, it can easily lead to misjudgment and malfunction, impacting power supply stability and equipment safety. Currently, traditional inrush current identification methods primarily rely on the physical characteristics of a single sensor (such as the excitation current waveform) or rely on fixed empirical thresholds. These methods struggle to cope with environmental interference and changing operating conditions in complex power systems. Furthermore, some methods handle inrush current identification and early warning independently, failing to achieve real-time linkage and resulting in delayed fault prediction and early warning. Summary of the Invention

[0003] In view of the shortcomings of the existing technology, the present invention provides an integrated method for transformer inrush current identification and early warning based on multi-sensor fusion to solve the problems raised in the above background technology.

[0004] To achieve the above object, the present invention provides the following technical solutions:

[0005] In a first aspect, an embodiment of the present invention provides a method for integrating transformer inrush current identification and early warning based on multi-sensor fusion, comprising the following steps:

[0006] S1. Distributed deployment of multiple sensors and construction of adaptive perception networks, and collection of sensor data;

[0007] S2, performing multi-dimensional feature correction and time alignment on the collected sensor data;

[0008] S3, extracting fuzzy characteristics from the corrected data to capture the complex nonlinear relationship in the transformer operating state;

[0009] S4, fusing the data after extracting the fuzzy features and generating a multimodal fusion feature vector;

[0010] S5. Build an inrush current identification model, input the multimodal fusion feature vector, and accurately identify the transformer inrush current;

[0011] S6. Based on the surge flow identified by the surge flow identification model, analyze the precursory characteristics of the abnormal state and issue a real-time warning.

[0012] To further optimize the technical solution, in step S1, sensors including current, voltage, magnetic field, and temperature are arranged based on distributed deployment in the power equipment site where the transformer is located; the nodes in the adaptive sensing network are interconnected through a low-latency wireless protocol to ensure that the sensors work synchronously in real time under dynamic working conditions.

[0013] To further optimize this technical solution, in step S2, a dynamic multi-dimensional feature mapping model is constructed to perform feature alignment across time and dimensions on the sensor data;

[0014] In the dynamic multidimensional feature mapping model, it is assumed that there is sensors, and the recorded data matrix is ,in , Represents the feature dimension recorded by the sensor;

[0015] The features after mapping are defined as .

[0016] To further optimize this technical solution, the dynamic multi-dimensional feature mapping model includes:

[0017] Time correction:

[0018] Time alignment is based on interpolation and dynamic weight optimization. The corrected data is:

[0019] ;

[0020] in,

[0021] Indicates sensor Clock drift is dynamically updated based on the global synchronization mechanism;

[0022] is the correction factor, which is adjusted according to the real-time performance of the sensor;

[0023] Spatial bias correction:

[0024] Based on the linear and nonlinear combination model, the correction is:

[0025] ;

[0026] in,

[0027] It is a sensor The linear correction matrix is ​​obtained by least squares estimation;

[0028] is a nonlinear adjustment factor optimized by the local deviation model;

[0029] represents the nonlinear correction function;

[0030] Scale normalization:

[0031] For the scale difference of multidimensional features, the normalization formula is:

[0032] ;

[0033] in,

[0034] and Sensors The mean and standard deviation of the collected data are dynamically calculated based on the sliding window;

[0035] Final mapping:

[0036] Mapped features Calculated as:

[0037] ;

[0038] in,

[0039] is a weight matrix, which is generated based on mutual information constraint optimization to ensure that the correlation between features is maximized.

[0040] To further optimize this technical solution, in the dynamic multi-dimensional feature mapping model:

[0041] Time Correction :After acquiring sensor data, first correct the clock drift, and then dynamically adjust and , ensuring that the data of each sensor is synchronized in the time dimension;

[0042] Spatial bias correction :The time-corrected data is spatially corrected by the linear correction matrix and nonlinear functions , correct the deviation of the sensor caused by position or environmental factors;

[0043] Scale normalization :After spatial correction, the data is normalized and the sliding window is calculated and Dynamically adapt to changes in sensor performance to ensure data balance in a unified feature space;

[0044] Feature Map : Finally, through the weight matrix Feature mapping is performed on the normalized data, and the weight matrix is ​​generated by mutual information optimization to ensure high feature correlation and low redundancy, providing optimized multi-dimensional input for subsequent processing.

[0045] To further optimize the technical solution, in step S3, by performing feature decomposition on the corrected data and adopting a fuzzy logic analysis method, a rule-based fuzzy feature library is constructed. The specific steps include:

[0046] The eigenvalues ​​of current, voltage and magnetic field are divided into multiple fuzzy sets, including "normal", "critical" and "abnormal", using fuzzy clustering algorithm;

[0047] Design fuzzy rules between different fuzzy sets to quantify the state transition probability of each feature;

[0048] A fuzzy weight adjustment mechanism based on the entropy weight method is introduced to make the rule base adaptive to the importance of features.

[0049] Further optimizing the technical solution, in step S4, based on the data fusion of the graph neural network GNN, the sensor data is regarded as a graph structure for modeling;

[0050] The sensor nodes are used as the vertices of the graph, and the associations between different modal features are used as the edges of the graph. The global fusion features are extracted through the hierarchical propagation mechanism of GNN.

[0051] At the same time, the multi-head attention mechanism is used to enhance the weight of key features and give higher discrimination ability to important parameters;

[0052] Finally, a multimodal fusion feature vector is generated as the input of the subsequent surge recognition model.

[0053] Further optimizing the technical solution, in step S5, the input of the surge recognition model is the multimodal fusion feature vector output in step S4. ,dynamically analyze the complex relationships between features and generate the surge classification results;

[0054] The surge identification model includes local feature extraction, time-dependent modeling, dynamic optimization, and classification decision-making.

[0055] To further optimize this technical solution, in the surge identification model:

[0056] Local feature extraction:

[0057] The convolution operation extracts the local characteristics of the feature vector, and the feature representation is:

[0058] ;

[0059] in,

[0060] It is Feature matrix after layer convolution;

[0061] and are the convolution kernel and bias terms respectively;

[0062] is a nonlinear activation function;

[0063] Time-dependent modeling:

[0064] Temporal modeling of feature sequences extracted by convolution:

[0065] ;

[0066] in,

[0067] It's time The hidden state of

[0068] Represents the feature matrix extracted by convolution As input, the feature vector generated by convolution is used to model the temporal dynamics of the convolution feature sequence;

[0069] , , They are The mapping weight of , the mapping weight of the previous moment, and the overall bias term;

[0070] is the activation function;

[0071] Dynamic Optimization:

[0072] Introducing reinforcement learning to dynamically adjust classification thresholds , optimize the reward function:

[0073] ;

[0074] in,

[0075] is the reward value;

[0076] Indicates the model accuracy, which is the proportion of correctly classified samples to the total samples;

[0077] and are the false positive and false negative rates, respectively, Reflects the ratio of negative samples misclassified as positive samples, that is, the probability of false alarm. Reflects the proportion of positive samples that are misclassified as negative samples, that is, the probability of missing reports;

[0078] is the penalty coefficient;

[0079] Classification decision:

[0080] Classification is performed based on the output of local feature extraction and time-dependent modeling:

[0081] ;

[0082] in,

[0083] The classification results include “normal surge” and “abnormal surge”;

[0084] , are the output layer weights and biases.

[0085] To further optimize this technical solution, in step S6, the real-time warning includes:

[0086] Precursor feature extraction;

[0087] Dynamic threshold calculation;

[0088] Failure probability inference;

[0089] Real-time early warning decision-making.

[0090] In a second aspect, an embodiment of the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, the steps of an integrated method for transformer inrush current identification and early warning based on multi-sensor fusion as described in the first aspect of the present invention are implemented.

[0091] In a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, the steps of an integrated method for transformer inrush current identification and early warning based on multi-sensor fusion as described in the first aspect of the present invention are implemented.

[0092] Compared with the existing technology, the present invention provides an integrated method for transformer inrush current identification and early warning based on multi-sensor fusion, which has the following beneficial effects:

[0093] This integrated transformer inrush current identification and early warning method based on multi-sensor fusion utilizes distributed multi-sensor fusion technology to dynamically extract multimodal features and implement deep learning modeling. This method, combined with real-time dynamic threshold calculation, reinforcement learning optimization, and Bayesian fault probability inference, achieves integrated transformer inrush current identification and early warning. Compared with existing technologies, this method demonstrates higher inrush current identification accuracy and robustness under complex operating conditions. Furthermore, its early warning function can respond to system changes in real time, significantly improving intelligence and safety, effectively reducing false detection and missed detection rates, and ensuring the reliability and stability of power grid operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0094] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0095] Figure 1 This is a flow chart of an integrated method for transformer inrush current identification and early warning based on multi-sensor fusion proposed in the present invention. DETAILED DESCRIPTION

[0096] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0097] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0098] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it constitute a separate or selective embodiment that is mutually exclusive with other embodiments.

[0099] Example 1:

[0100] Reference Figure 1 , which is the first embodiment of the present invention, provides a transformer inrush current identification and early warning integrated method based on multi-sensor fusion, comprising the following steps:

[0101] S1. Conduct distributed deployment of multiple sensors and build an adaptive perception network, and collect sensor data.

[0102] In this embodiment, sensors for current, voltage, magnetic field, and temperature are deployed in a rationally distributed manner throughout the power equipment site where the transformer is located. Each sensor node can integrate an adaptive sensing module, using an embedded algorithm to dynamically adjust the sampling frequency and resolution to adapt to real-time environmental changes. For example, if a large current fluctuation is detected, the sensor will automatically switch to high-frequency sampling mode.

[0103] Each node in the adaptive sensing network is interconnected through a low-latency wireless protocol to achieve coordinated work between sensors while ensuring data quality, which is used to ensure that sensors can work synchronously in real time under dynamic working conditions.

[0104] S2. Perform multi-dimensional feature correction and time alignment on the collected sensor data.

[0105] In this embodiment, data acquired from a distributed sensor network may exhibit issues such as clock drift, spatial position deviation, and physical parameter scale differences. Multidimensional features include not only data from different time steps but also spatial and physical parameter differences between sensors. These features may reflect a variety of physical information related to surge phenomena at the same time point. Therefore, a dynamic multidimensional feature mapping model is constructed to align sensor data across time and dimensions, mapping data from different sensors into a unified feature space. This provides a standardized data foundation for subsequent surge identification.

[0106] In the dynamic multidimensional feature mapping model, it is assumed that there is sensors, and the recorded data matrix is ,in , Represents the feature dimensions recorded by the sensor. Multidimensional features are composed of multiple feature dimensions. For example, if a sensor provides both current and temperature data, then the sensor's multidimensional features include two primary dimensions: current and temperature, each of which may have multiple sub-dimensions.

[0107] During the multi-dimensional feature mapping process, we will calibrate, standardize, and align each feature dimension to ensure that the different dimensional data obtained by different sensors can be accurately integrated and reflect the complete system status.

[0108] The features after mapping are defined as .

[0109] The dynamic multi-dimensional feature mapping model includes:

[0110] Time correction:

[0111] Time alignment is based on interpolation and dynamic weight optimization. The corrected data is:

[0112] ;

[0113] in,

[0114] Indicates the current time step, used to identify the data collected by the sensor at this moment, and is used for interpolation calculations during the time correction process;

[0115] Indicates sensor Clock drift is dynamically updated based on the global synchronization mechanism;

[0116] is the correction factor, which is adjusted according to the real-time performance of the sensor.

[0117] Spatial bias correction:

[0118] Based on the linear and nonlinear combination model, the correction is:

[0119] ;

[0120] in,

[0121] It is a sensor The linear correction matrix is ​​obtained by least squares estimation;

[0122] is a nonlinear adjustment factor optimized by the local deviation model;

[0123] represents a nonlinear correction function (such as a polynomial fit or a kernel function).

[0124] Scale normalization:

[0125] For the scale difference of multidimensional features, the normalization formula is:

[0126] ;

[0127] in,

[0128] and Sensors The mean and standard deviation of the collected data are dynamically calculated based on a sliding window.

[0129] Final mapping:

[0130] Mapped features Calculated as:

[0131] ;

[0132] in,

[0133] is a weight matrix, which is generated based on mutual information constraint optimization to ensure that the correlation between features is maximized.

[0134] Furthermore, in the dynamic multidimensional feature mapping model:

[0135] Time Correction :After acquiring sensor data, first correct the clock drift, and then dynamically adjust and , ensuring that the data of each sensor is synchronized in the time dimension. For example, if the time deviation of a sensor is large, its , to match the global time reference. In practical applications, and The adjustment adopts an adaptive feedback mechanism: It will be dynamically updated based on the sensor's real-time signal-to-noise ratio and data fluctuation (reflecting its measurement stability). The real-time correction is based on the deviation between the global synchronization reference (such as GPS or master clock) and the sensor's local clock.

[0136] Spatial bias correction :The time-corrected data is spatially corrected by the linear correction matrix and nonlinear functions , correct the deviation of the sensor caused by position or environmental factors. For example, the subsequent correction matrix The sensor's location information (such as installation coordinates) and environmental parameters (such as temperature and electromagnetic interference level) are integrated into the spatial deviation correction to ensure effective correlation and unified compensation between the time-corrected data and the spatial correction.

[0137] Scale normalization :After spatial correction, the data is normalized and the sliding window is calculated and Dynamically adapt to changes in sensor performance and ensure data balance in a unified feature space.

[0138] Feature Map : Finally, through the weight matrix Feature mapping is performed on the normalized data, and the weight matrix is ​​generated by mutual information optimization to ensure high feature correlation and low redundancy, providing optimized multi-dimensional input for subsequent processing.

[0139] S3. Extract fuzzy characteristics from the corrected data to capture the complex nonlinear relationship in the transformer operating state.

[0140] In this embodiment, by performing feature decomposition on the corrected data and adopting a fuzzy logic analysis method, a rule-based fuzzy feature library is constructed. The specific steps include:

[0141] The fuzzy clustering algorithm is used to divide the characteristic values ​​of current, voltage and magnetic field into multiple fuzzy sets, including "normal", "critical" and "abnormal".

[0142] Design fuzzy rules between different fuzzy sets and quantify the state transition probability of each feature.

[0143] State transition probability describes the likelihood of transitioning from the current state to the next. It is calculated based on historical data and real-time observations. Its quantitative basis primarily includes feature membership, sample statistical distribution, and time series trends. A fuzzy membership function is used to measure the degree of transition between different states. The calculation involves counting the frequency of observations under different states and incorporating fuzzy rules to ensure the rationality and continuity of state transitions. Adjustments to state transition probabilities can be combined with real-time data updates to adapt to environmental changes and improve recognition accuracy.

[0144] Fuzzy rules include:

[0145] Definition of fuzzy sets;

[0146] According to the operating status of the transformer, the input characteristics (such as current, voltage, magnetic field strength, etc.) are divided into multiple fuzzy sets. Each fuzzy set is represented in the form of linguistic variables, for example:

[0147] Current: low (L), medium (M), high (H);

[0148] Voltage: normal (N), high (PH), low (PL);

[0149] Magnetic field strength: weak (W), medium (M), strong (S).

[0150] Design of fuzzy rules;

[0151] Fuzzy rules take the form of "if...then..." and determine the relationship between fuzzy sets through logical reasoning. For example:

[0152] Rule 1: If the current is high (H) and the voltage is low (PL), then the transformer status is abnormal (abnormal set).

[0153] Rule 2: If the current is medium (M) and the magnetic field strength is medium (M), then the transformer status is normal (normal set).

[0154] Rule 3: If the voltage is high (PH) and the magnetic field strength is strong (S), then the transformer state is critical (critical set).

[0155] Each rule implements state inference through fuzzy logic operations (such as maximum-minimum method or weighted average method).

[0156] In order to dynamically adjust the weights of different features in fuzzy rules, the entropy weight method is used to quantify the importance of features. The fuzzy weight adjustment mechanism based on the entropy weight method is introduced to make the rule base adaptive to the importance of features.

[0157] In practical applications, the fuzzy rule base will be dynamically adjusted according to the changes in weights. For example:

[0158] When the current entropy value is low (the characteristics are stable), the current characteristic weight is reduced and the rule is more dependent on the voltage or magnetic field strength.

[0159] When the voltage entropy value increases (feature fluctuations are drastic), the voltage feature weight increases, and the rule places more emphasis on the impact of voltage.

[0160] Through fuzzy characteristic decomposition, the complex state of transformer operation can be expressed in fuzzy space, providing nonlinear input features for the subsequent inrush current identification model.

[0161] S4. Fusing the data after fuzzy feature extraction and generating a multimodal fusion feature vector.

[0162] In this embodiment, in step S4, sensor data is modeled as a graph structure based on data fusion of a graph neural network (GNN).

[0163] Sensor nodes serve as graph vertices, and the relationships between different modal features serve as graph edges. Global fusion features are extracted through the hierarchical propagation mechanism of GNNs. Each sensor's raw data (such as current, voltage, and temperature) is normalized and dimensionally reduced before being embedded as the initial node features to form the input feature matrix. A graph neural network (GNN) is used to perform multi-layer feature propagation on the graph-structured data to obtain global fusion features.

[0164] At the same time, the multi-head attention mechanism is used to enhance the weight of key features and give higher discrimination ability to important parameters. The multi-head attention mechanism is introduced to enhance the weight of key modal features:

[0165] ;

[0166] Each attention head Focus on the importance of different modal features;

[0167] Representation node Enhanced features.

[0168] Finally, a multimodal fusion feature vector is generated , as the input of the subsequent surge identification model.

[0169] The spatiotemporal dynamic features of all nodes are integrated into a fusion feature vector through a fully connected layer and pooling operation :

[0170] ;

[0171] It can be global maximum pooling or average pooling to extract global features.

[0172] Represents the multimodal fusion feature vector, which includes the spatial correlation and temporal dynamic characteristics of sensors.

[0173] S5. Build an inrush current identification model, input the multimodal fusion feature vector, and accurately identify the transformer inrush current.

[0174] In this embodiment, the input of the surge recognition model is the multimodal fusion feature vector output in step S4. ,dynamically analyze the complex relationships between features and generate the surge classification results;

[0175] The surge identification model includes local feature extraction, time-dependent modeling, dynamic optimization, and classification decision-making.

[0176] Furthermore, in the surge identification model:

[0177] Local feature extraction:

[0178] The convolution operation extracts the local characteristics of the feature vector, and the feature representation is:

[0179] ;

[0180] in,

[0181] It is Feature matrix after layer convolution;

[0182] and are the convolution kernel and bias terms respectively;

[0183] is a non-linear activation function (such as ReLU).

[0184] Time-dependent modeling:

[0185] Temporal modeling of feature sequences extracted by convolution:

[0186] ;

[0187] in,

[0188] It's time The hidden state of

[0189] Represents the feature matrix extracted by convolution as input and the feature vector generated by convolution to achieve temporal dynamic modeling of the convolution feature sequence;

[0190] , , They are The mapping weight of , the mapping weight of the previous moment, and the overall bias term;

[0191] is the activation function;

[0192] Dynamic Optimization:

[0193] Introducing reinforcement learning to dynamically adjust classification thresholds , optimize the reward function:

[0194] ;

[0195] in,

[0196] is the reward value;

[0197] Indicates the model accuracy, which is the proportion of correctly classified samples to the total samples;

[0198] and are the false positive and false negative rates, respectively, Reflects the ratio of negative samples misclassified as positive samples, that is, the probability of false alarm. Reflects the proportion of positive samples that are misclassified as negative samples, that is, the probability of missing reports;

[0199] is the penalty coefficient.

[0200] Classification threshold The role of is to determine the decision boundary of the model when performing surge identification, affecting the false positive rate (FP) and false negative rate (FN). In the reinforcement learning module, the classification threshold It is not fixed, but dynamically adjusted by optimizing the reward function R to maximize the accuracy A while minimizing the false positives (FP) and false negatives (FN). When the false positive rate is too high, the threshold is appropriately increased. To reduce false alarms; when the missed alarm rate is too high, reduce In this way, the classification threshold can be adaptively adjusted under different operating environments to ensure the reliability and stability of recognition.

[0201] Classification decision:

[0202] Classification is performed based on the output of local feature extraction and time-dependent modeling:

[0203] ;

[0204] in,

[0205] The classification results include “normal surge” and “abnormal surge”;

[0206] is the hidden state vector at the final time step T, representing the model's feature extraction and memory results over the entire time series. It integrates information from all historical time steps and serves as the input for the final classification decision. After being transformed by the fully connected layer, it is used to determine the flow category.

[0207] , are the output layer weights and biases.

[0208] When used, the model includes:

[0209] Local feature extraction (CNN): multimodal fusion features Convolution kernel Extract local spatial patterns. Identification of surge characteristics (such as amplitude changes and specific frequency band responses) is particularly important. Nonlinear activation Emphasize key features and eliminate redundant information.

[0210] Time-dependent modeling (LSTM): The feature sequence output by the CNN is passed to the LSTM unit to capture long-term temporal dependencies, such as the dynamic changes in the surge waveform. Time series modeling can better distinguish between transient surges and steady-state surges.

[0211] Dynamic optimization: involving classification thresholds Dynamic optimization. Reinforcement learning is adjusted according to real-time classification results. , ensuring adaptability under different working conditions. For example, when the environmental disturbance is large, the reward function It will reduce the false positive rate weights to improve accuracy.

[0212] Classification decision: The combined output of CNN and LSTM passes through the fully connected layer to calculate the classification probability, and the flow state is determined based on the maximum probability value.

[0213] S6. Based on the surge flow identified by the surge flow identification model, analyze the precursory characteristics of the abnormal state and issue a real-time warning.

[0214] In this embodiment, real-time warning includes:

[0215] Precursor feature extraction;

[0216] Extract precursor features using a combination of time domain and frequency domain methods :

[0217] ;

[0218] in:

[0219] is the precursor eigenvector;

[0220] is the precursor weight matrix;

[0221] is the multimodal fusion feature of the sensor at the current moment;

[0222] represents the frequency domain features extracted by fast Fourier transform;

[0223] is the frequency domain feature adjustment factor.

[0224] pass Generate the multimodal features of the current sensor and combine them with FFT to obtain frequency domain features to form the precursor feature vector This combination can capture subtle characteristic changes before inrush current occurs, such as periodic fluctuations in current amplitude.

[0225] Multimodal features refer to multidimensional data collected from multiple different sensors. Through feature extraction and fusion methods, they form multimodal fusion features for inrush current identification. These features come from different types of sensors, such as current sensors, temperature sensors, and vibration sensors. Each sensor provides data in a different modality. After fusion, the accuracy and robustness of inrush current identification can be improved. They mainly include:

[0226] Time domain characteristics: such as the mean, variance, skewness, peak value, rise time, etc. of the current waveform;

[0227] Frequency domain characteristics: such as harmonic components, main frequency, spectrum energy distribution, etc.;

[0228] Spatial characteristics: signal correlation between different sensors, such as the current ratio of different measurement points at the same time;

[0229] Time series features: Use RNN / LSTM for time modeling to capture the pattern of signal changes over time.

[0230] Furthermore, frequency domain characteristics refer to the performance of the signal in the frequency domain, which reflects the periodicity and harmonic components of the signal and can reveal the changing trend of specific frequency components when the inrush current occurs. It mainly includes:

[0231] Main frequency component: the main frequency component of the signal;

[0232] Harmonic ratio: the energy ratio of fundamental wave and higher harmonics;

[0233] Spectral energy distribution: the energy proportion of different frequency bands, such as 0–50Hz, 50–150Hz, 150–500Hz, etc.

[0234] Spectral entropy: measures the complexity of the signal spectrum. The higher the entropy value, the more complex the signal frequency distribution.

[0235] Frequency domain features use Fast Fourier Transform (FFT) to convert time domain signals into frequency domain signals and calculate the above frequency domain features.

[0236] Dynamic threshold calculation;

[0237] Threshold Dynamic adjustment, the expression is:

[0238] ;

[0239] in:

[0240] is the initial threshold;

[0241] is the dynamic adjustment coefficient;

[0242] and are the mean and standard deviation of the precursor features (calculated using a dynamic sliding window).

[0243] During the early warning process, the system performs a standardized calculation of the volatility of the precursory characteristics (based on the mean and standard deviation of the sliding window). Dynamically adjusted thresholds Able to adapt to different operating conditions, such as load fluctuations or environmental influences.

[0244] Failure probability inference;

[0245] Calculate real-time failure probability based on Bayesian formula :

[0246] ;

[0247] in:

[0248] Precursor characteristics are observed under fault conditions probability;

[0249] is the prior failure probability;

[0250] is the total probability of the feature.

[0251] Use Bayesian inference methods combined with real-time data to estimate the probability of failure When the inrush precursor features are obvious, this value will increase rapidly, enhancing the sensitivity of the model.

[0252] Real-time early warning decision-making;

[0253] Early warning signals Output rules:

[0254] ;

[0255] in:

[0256] Indicates that an early warning is triggered;

[0257] Indicates normal status.

[0258] According to the probability of failure and threshold Compare and output real-time warning signals. , the inrush current warning mechanism is triggered and the signal is transmitted to perform subsequent protection actions.

[0259] Example 2:

[0260] This embodiment also provides a computer device, which is suitable for a method for integrating transformer inrush current identification and early warning based on multi-sensor fusion, and includes a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement a method for integrating transformer inrush current identification and early warning based on multi-sensor fusion as proposed in the above embodiment.

[0261] This embodiment also provides a storage medium having a computer program stored thereon. When the program is executed by a processor, the method for integrating transformer inrush current identification and early warning based on multi-sensor fusion as proposed in the above embodiment is implemented.

[0262] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0263] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the method of the present invention. The aforementioned storage medium includes various media that can store program code, such as USB flash drives, mobile hard drives, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0264] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0265] More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disc read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting, or processing it in another suitable manner as necessary, and then storing it in a computer memory.

[0266] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0267] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A transformer inrush current identification and early warning integrated method based on multi-sensor fusion, characterized in that: The following steps are involved: S1. Distributed deployment of multiple sensors and construction of adaptive perception networks, and collection of sensor data; S2. Perform multi-dimensional feature correction and time alignment on the collected sensor data; specifically, build a dynamic multi-dimensional feature mapping model to perform cross-time and cross-dimensional feature alignment on the sensor data; In the dynamic multidimensional feature mapping model, it is assumed that there is sensors, and the recorded data matrix is ,in , Represents the feature dimension recorded by the sensor; The features after mapping are defined as ; The dynamic multi-dimensional feature mapping model includes: Time correction: Time alignment is based on interpolation and dynamic weight optimization. The corrected data is: ； in, Indicates sensor Clock drift is dynamically updated based on the global synchronization mechanism; is the correction factor, which is adjusted according to the real-time performance of the sensor; Spatial bias correction: Based on the linear and nonlinear combination model, the correction is: ； in, It is a sensor The linear correction matrix is ​​obtained by least squares estimation; is a nonlinear adjustment factor optimized by the local deviation model; represents the nonlinear correction function; Scale normalization: For the scale difference of multidimensional features, the normalization formula is: ； in, and Sensors The mean and standard deviation of the collected data are dynamically calculated based on the sliding window; Final mapping: Mapped features Calculated as: ； in, is a weight matrix, generated based on mutual information constraint optimization to ensure that the correlation between features is maximized; S3, extracting fuzzy characteristics from the corrected data to capture the complex nonlinear relationship in the transformer operating state; S4, fusing the data after extracting the fuzzy features and generating a multimodal fusion feature vector; S5. Build an inrush current identification model, input the multimodal fusion feature vector, and accurately identify the transformer inrush current; S6. Based on the surge flow identified by the surge flow identification model, analyze the precursory characteristics of the abnormal state and issue a real-time warning.

2. The integrated method for transformer inrush current identification and early warning based on multi-sensor fusion according to claim 1 is characterized in that: In step S1, sensors including current, voltage, magnetic field, and temperature are arranged based on distributed deployment in the power equipment site where the transformer is located; the nodes in the adaptive sensing network are interconnected through a low-latency wireless protocol to ensure that the sensors work synchronously in real time under dynamic working conditions.

3. The integrated method for transformer inrush current identification and early warning based on multi-sensor fusion according to claim 2 is characterized in that: In the dynamic multi-dimensional feature mapping model: Time Correction :After acquiring sensor data, first correct the clock drift, and then dynamically adjust and , ensuring that the data of each sensor is synchronized in the time dimension; Spatial bias correction :The time-corrected data is spatially corrected by the linear correction matrix and nonlinear functions , correct the deviation of the sensor caused by position or environmental factors; Scale normalization :After spatial correction, the data is normalized and the sliding window is calculated and Dynamically adapt to changes in sensor performance to ensure data balance in a unified feature space; Feature Map : Finally, through the weight matrix Feature mapping is performed on the normalized data, and the weight matrix is ​​generated by mutual information optimization to ensure high feature correlation and low redundancy, providing optimized multi-dimensional input for subsequent processing.

4. The integrated method for transformer inrush current identification and early warning based on multi-sensor fusion according to claim 1 is characterized in that: In step S3, the corrected data is subjected to characteristic decomposition and a fuzzy logic analysis method is used to construct a rule-based fuzzy characteristic library. The specific steps include: The fuzzy clustering algorithm is used to divide the characteristic values ​​of current, voltage and magnetic field into multiple fuzzy sets, including "normal", "critical" and "abnormal"; Design fuzzy rules between different fuzzy sets to quantify the state transition probability of each feature; A fuzzy weight adjustment mechanism based on the entropy weight method is introduced to make the rule base adaptive to the importance of features.

5. The integrated method for transformer inrush current identification and early warning based on multi-sensor fusion according to claim 1 is characterized in that: In step S4, based on the data fusion of the graph neural network (GNN), the sensor data is regarded as a graph structure for modeling; The sensor nodes are used as the vertices of the graph, and the associations between different modal features are used as the edges of the graph. The global fusion features are extracted through the hierarchical propagation mechanism of GNN. At the same time, the multi-head attention mechanism is used to enhance the weight of key features and give higher discrimination ability to important parameters; Finally, a multimodal fusion feature vector is generated as the input of the subsequent surge recognition model.

6. The integrated method for transformer inrush current identification and early warning based on multi-sensor fusion according to claim 1 is characterized in that: In step S5, the input of the surge recognition model is the multimodal fusion feature vector output in step S4. ,dynamically analyze the complex relationships between features and generate the surge classification results; The surge identification model includes local feature extraction, time-dependent modeling, dynamic optimization, and classification decision-making.

7. The integrated method for transformer inrush current identification and early warning based on multi-sensor fusion according to claim 6 is characterized in that: In the surge identification model: Local feature extraction: The convolution operation extracts the local characteristics of the feature vector, and the feature representation is: ； in, It is Feature matrix after layer convolution; and are the convolution kernel and bias terms respectively; is a nonlinear activation function; Time-dependent modeling: Temporal modeling of feature sequences extracted by convolution: ； in, It's time The hidden state of Represents the feature matrix extracted by convolution As input, the feature vector generated by convolution is used to achieve temporal dynamic modeling of the convolution feature sequence; , , They are The mapping weight of , the mapping weight of the previous moment, and the overall bias term; is the activation function; Dynamic Optimization: Introducing reinforcement learning to dynamically adjust classification thresholds , optimize the reward function: ； in, is the reward value; Indicates the model accuracy, which is the proportion of correctly classified samples to the total samples; and are the false positive and false negative rates, respectively, Reflects the ratio of negative samples misclassified as positive samples, that is, the probability of false alarm. Reflects the proportion of positive samples that are misclassified as negative samples, that is, the probability of missing reports; is the penalty coefficient; Classification decision: Classification is performed based on the output of local feature extraction and time-dependent modeling: ； in, The classification results include "normal flow" and "abnormal flow"; , are the output layer weights and biases.

8. The integrated method for transformer inrush current identification and early warning based on multi-sensor fusion according to claim 1 is characterized in that: In step S6, the real-time warning includes: Precursor feature extraction; Dynamic threshold calculation; Failure probability inference; Real-time early warning decision-making.

Images