Distribution transformer monitoring system and monitoring method

Abstract

The invention discloses a distribution transformer monitoring system and method, and relates to the technical field of electric power intelligent monitoring, and the method comprises the steps: obtaining the three-dimensional temperature gradient field distribution and mechanical stress distribution of a distribution transformer based on a standardized multi-source data set, and outputting a comprehensive field matrix through a field coupling analysis method; performing spatial analysis on the comprehensive field matrix by adopting spatial hot spot analysis, identifying and marking temperature and mechanical stress abnormal areas, performing spatial aggregation and noise filtering through an OPTICS density clustering algorithm, and outputting abnormal feature vectors; inputting the abnormal feature vector into a pre-trained intelligent diagnosis model, outputting a comprehensive fault risk score and performing preliminary diagnosis; searching fault data in a historical case library based on the preliminary diagnosis result for comparison verification, and optimizing the preliminary diagnosis result through a machine learning algorithm to generate a final diagnosis result; according to the method, a field coupling analysis method is adopted to establish a heat-force bidirectional action model, and the spatial distribution characteristics of the temperature gradient and the stress tensor are accurately reflected.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent power monitoring, and in particular to a distribution transformer monitoring system and a monitoring method. Background Art

[0002] With the rapid development of smart grids, distribution transformer condition monitoring technology continues to innovate, leveraging sensor networks and machine learning to achieve intelligent data collection and fault diagnosis. Current technologies primarily utilize infrared temperature measurement, vibration detection, and oil chromatography analysis to acquire real-time, multi-dimensional data on transformer operating conditions, including temperature, vibration, and electrical parameters. Using threshold determination, statistical analysis, and signal processing algorithms, the system can initially identify abnormal conditions.

[0003] In the field of intelligent power monitoring technology, traditional methods typically employ a step-by-step process, first independently calculating the temperature gradient field and mechanical stress distribution, then performing a simple correlation through empirical formulas or linear superposition. This approach struggles to accurately reflect the nonlinear interactions of multiple physical fields under complex transformer operating conditions. This can easily lead to misjudgments and missed detections of abnormal areas, especially in scenarios such as transient overloads or localized insulation degradation. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a distribution transformer monitoring method to solve the problem of low fault diagnosis accuracy caused by insufficient collaborative analysis of multi-physical fields of distribution transformers.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a distribution transformer monitoring method, which comprises: The operating status data of distribution transformers is collected in real time and preprocessed to obtain a standardized multi-source dataset. Based on the standardized multi-source dataset, the three-dimensional temperature gradient field distribution and mechanical stress distribution of the distribution transformer are obtained, and a comprehensive field matrix is ​​output through field coupling analysis. Spatial hotspot analysis is used to perform spatial analysis on the comprehensive field matrix, identifying and marking abnormal temperature and mechanical stress areas. Spatial aggregation and noise filtering are performed using the OPTICS density clustering algorithm to output abnormal feature vectors. The abnormal feature vectors are input into a pre-trained intelligent diagnosis model, which outputs a comprehensive fault risk score and performs a preliminary diagnosis. Based on the preliminary diagnosis results, fault data in the historical case library is retrieved for comparison and verification. The preliminary diagnosis results are optimized using a machine learning algorithm to generate a final diagnosis result. Dynamic warnings are issued based on the final diagnosis results, and a fault detection report is generated.

[0007] As a preferred solution of the distribution transformer monitoring method of the present invention, wherein: the distribution transformer operating status data includes temperature data, core vibration data, shell vibration data, load current data, operating voltage data, and insulating oil and gas data; The preprocessing includes noise suppression through sliding window filtering, data normalization using the Z-score standardization method, and missing value compensation using linear interpolation.

[0008] As a preferred solution of the distribution transformer monitoring method of the present invention, wherein: based on the standardized multi-source data set, the three-dimensional temperature gradient field distribution and mechanical stress distribution of the distribution transformer are obtained, and the comprehensive field matrix is ​​output through the field coupling analysis method. The specific steps are as follows: Based on standardized multi-source data sets, the finite element analysis method is used to calculate the three-dimensional temperature gradient field distribution, and the mechanical stress distribution is obtained by generalized Hooke's law; The three-dimensional temperature gradient field distribution and mechanical stress distribution are input into the field coupling analysis method to establish a thermal-mechanical coupling relationship, and a comprehensive field matrix containing multidimensional data of spatial coordinates, three-dimensional temperature gradient field distribution and mechanical stress distribution is output.

[0009] As a preferred solution of the distribution transformer monitoring method of the present invention, the method uses spatial hotspot analysis to perform spatial analysis on the comprehensive field matrix, identifies and marks abnormal temperature and mechanical stress areas, performs spatial aggregation and noise filtering through the OPTICS density clustering algorithm, and outputs abnormal feature vectors. The specific steps are as follows: The comprehensive field matrix is ​​input into the spatial hot spot analysis to identify abnormal areas of temperature gradient and mechanical stress and output the abnormal points; The outliers are spatially clustered using the OPTICS density clustering algorithm, which merges adjacent outliers and filters out isolated noise points, outputting the cluster core area. Based on the cluster core area, the spatial coordinates, average temperature gradient and maximum mechanical stress of the abnormal points are extracted through the spatial statistical feature extraction method to generate the abnormal feature vector.

[0010] As a preferred solution of the distribution transformer monitoring method of the present invention, wherein: the abnormal feature vector is input into the pre-trained intelligent diagnosis model, the comprehensive fault risk score is output and the preliminary diagnosis is performed, the specific steps are as follows: Based on historical abnormal feature vectors, the DNN deep neural network is trained through the back-propagation algorithm to obtain an intelligent diagnosis model that can output a comprehensive fault risk score; The abnormal feature vector is input into the intelligent diagnosis model to analyze the comprehensive fault risk score of each abnormal area; Based on historical fault data, low-risk thresholds, high-risk thresholds and fault classification rules are defined, and the comprehensive fault risk score is judged to obtain preliminary diagnosis results.

[0011] As a preferred solution of the distribution transformer monitoring method of the present invention, wherein: based on the preliminary diagnosis results, the fault data in the historical case library is retrieved for comparison and verification, the preliminary diagnosis results are optimized by a machine learning algorithm to generate the final diagnosis results. The specific steps are as follows: Based on the preliminary diagnosis results, the historical case library is searched to match the fault data records under similar working conditions and output the Top-K similar cases; The preliminary diagnosis results and the treatment effect data of the Top-K similar cases are input into the machine learning algorithm, and the diagnosis conclusions are optimized through probability weighting to generate the final diagnosis results.

[0012] As a preferred solution of the distribution transformer monitoring method of the present invention, wherein: the dynamic early warning is performed based on the final diagnosis result and a fault detection report is generated. The specific steps are as follows: Based on the final diagnosis results, the rule engine automatically matches the warning rule library to trigger multi-level warnings; The final diagnosis results, multi-level warnings and distribution transformer operating status data are aligned in time and space through feature fusion algorithms to construct structured diagnostic data; Based on structured diagnostic data, intelligent analysis is performed through natural language generation methods to obtain a fault detection report.

[0013] In a second aspect, the present invention provides a distribution transformer monitoring system, comprising: Data acquisition module, coupling analysis module, anomaly extraction module, diagnosis module, optimization module and report generation module; data acquisition module, used to collect distribution transformer operating status data in real time, and preprocess it to obtain a standardized multi-source data set; coupling analysis module, used to obtain the three-dimensional temperature gradient field distribution and mechanical stress distribution of the distribution transformer based on the standardized multi-source data set, and output the comprehensive field matrix through the field coupling analysis method; anomaly extraction module, used to perform spatial analysis on the comprehensive field matrix using spatial hotspot analysis, identify and mark temperature and mechanical stress abnormal areas, perform spatial aggregation and noise filtering through the OPTICS density clustering algorithm, and output anomaly feature vectors; diagnosis module, used to input the anomaly feature vectors into the pre-trained intelligent diagnosis model, output a comprehensive fault risk score and perform preliminary diagnosis; optimization module, used to retrieve fault data in the historical case library based on the preliminary diagnosis results for comparison and verification, optimize the preliminary diagnosis results through machine learning algorithms, and generate final diagnosis results; report generation module, used to perform dynamic early warning based on the final diagnosis results and generate a fault detection report.

[0014] In a third aspect, 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 is executed by the processor, any step of the distribution transformer monitoring method as described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the distribution transformer monitoring method as described in the first aspect of the present invention is implemented.

[0016] The beneficial effects of the present invention are as follows: the three-dimensional temperature gradient field distribution and mechanical stress distribution are calculated through finite element analysis and generalized Hooke's law, and a thermal-mechanical bidirectional action model is established by using a field coupling analysis method, thereby realizing the collaborative calculation and closed-loop feedback of the temperature field and the stress field, and accurately reflecting the spatial distribution characteristics of the temperature gradient and the stress tensor; at the same time, the OPTICS density clustering algorithm is used to intelligently aggregate the spatial hotspot analysis results, extract the spatial coordinates, average temperature gradient and maximum mechanical stress characteristics of the cluster core area, realize the automatic identification and quantitative characterization of the distributed fault area, and achieve the purpose of improving the accuracy and reliability of fault diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] 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.

[0018] Figure 1 Flowchart of the distribution transformer monitoring method.

[0019] Figure 2 Flowchart for preprocessing of distribution transformer operating status data.

[0020] Figure 3 Flowchart for field coupling analysis.

[0021] Figure 4 Schematic diagram of the distribution transformer monitoring system. DETAILED DESCRIPTION

[0022] 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.

[0023] 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.

[0024] 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 refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0025] Reference Figures 1 to 4 , this embodiment provides a distribution transformer monitoring method, comprising the following steps: S1. Collect the operating status data of distribution transformers in real time and preprocess them to obtain a standardized multi-source data set.

[0026] S1.1. The operating status data of the distribution transformer includes temperature data, core vibration data, casing vibration data, load current data, operating voltage data, and insulating oil and gas data.

[0027] It should be noted that the operating status data of distribution transformers is collected through the collaborative efforts of various sensors deployed on the distribution transformers. Temperature data is collected by fiber optic temperature sensors embedded in the windings and core, as well as by infrared temperature measurement modules attached to the outer casing. These sensors provide information on both the overall and local heating status of the equipment. Vibration data on the core and outer casing are acquired using high-precision accelerometers, monitoring internal mechanical vibration and external structural resonance caused by magnetostriction, respectively. Hall effect sensors and voltage transformers installed on the high and low voltage sides measure load current and operating voltage in real time for analyzing electrical load conditions. Insulating oil gas data is collected using an online oil chromatograph, detecting dissolved gas components and their concentration changes to assess insulation aging.

[0028] S1.2. Preprocessing includes noise suppression through sliding window filtering, data normalization using the Z-score standardization method, and missing value compensation using linear interpolation.

[0029] It should be noted that sliding window filtering is used to suppress high-frequency noise. For vibration signals, a window width adjustment mechanism based on signal energy analysis is employed. When random pulse interference is detected, the window width is automatically reduced to preserve detailed features. For temperature data, the window width is dynamically expanded based on the temperature change rate, effectively smoothing out abnormal jumps caused by sensor transient responses. A Z-score normalization method is used to process heterogeneous physical quantities, eliminating numerical scale differences between temperature data, core vibration data, housing vibration data, load current data, operating voltage data, and insulating oil and gas data, ensuring comparability. Missing values ​​caused by network delays or equipment failures during transmission are compensated using a linear interpolation algorithm based on time series correlation to ensure data continuity. After layered processing, the multi-source data is synchronized and aligned in the time domain, forming a standardized multi-source dataset suitable for advanced analysis.

[0030] S2. Based on the standardized multi-source data set, the three-dimensional temperature gradient field distribution and mechanical stress distribution of the distribution transformer are obtained, and the comprehensive field matrix is ​​output through the field coupling analysis method.

[0031] S2.1. Based on the standardized multi-source data set, the finite element analysis method is used to calculate the three-dimensional temperature gradient field distribution, and the mechanical stress distribution is obtained by the generalized Hooke's law.

[0032] It should be noted that after importing the standardized multi-source data set into the finite element analysis environment, a three-dimensional geometric model of the distribution transformer was established based on its actual physical structure, including key components such as the core, windings, and insulation layer. The spatial discretization was performed using tetrahedral elements. Temperature data was loaded as boundary conditions at the corresponding node locations, and the heat conduction partial differential equation was solved to obtain the continuous three-dimensional temperature gradient field distribution. Solving the partial differential equation for heat conduction to obtain the continuously distributed temperature gradient field expression is: ; in, It is the three-dimensional temperature gradient field distribution, which means the temperature in space The rate of change of position, is the differential change of the temperature field, It is along The differential displacement of the coordinate axis, It is along The differential displacement of the coordinate axis, It is along The differential displacement of the coordinate axis, is the total number of nodes, is the index variable for the total number of nodes, It is The temperature value of each node, It is The shape function of each node is related to the spatial coordinates The partial derivative of The core and shell vibration data are loaded as displacement boundary conditions onto the corresponding node positions of the finite element model, and the stiffness matrix is ​​constructed in combination with parameters such as the material elastic modulus and Poisson's ratio. The stress-strain relationship σ=C:ε is established based on the generalized Hooke's law, where σ is the stress tensor, C is the elastic stiffness tensor, and ε is the strain tensor. The strain component is calculated using the displacement field gradient as ε=(∇u+∇uᵀ) / 2, where ∇u is the gradient of the displacement field u, ∇uᵀ is the transposed matrix of the displacement gradient, and u is the vibration displacement vector. The Newton-Raphson iterative method is used for nonlinear solution: the displacement field distribution is initially assumed, the residual force vector is calculated, the linearized equilibrium equation is solved to update the displacement field, and the residual force is recalculated until the convergence criterion ‖R‖<ε is met. The Jacobian matrix is ​​updated at each iteration, and the preconditioned conjugate gradient method is used to solve the linear equations. After convergence, the stress components of the Gaussian integral points of each unit are calculated using the constitutive relationship, the nodal stresses are interpolated, and the mechanical stress distribution is finally output.

[0033] S2.2. Input the three-dimensional temperature gradient field distribution and mechanical stress distribution into the field coupling analysis method to establish a thermal-mechanical coupling relationship, and output a comprehensive field matrix containing multidimensional data of spatial coordinates, three-dimensional temperature gradient field distribution and mechanical stress distribution.

[0034] It should be noted that the three-dimensional temperature gradient field distribution and mechanical stress distribution are imported into the coupled analysis platform, and field quantity matching and alignment operations are performed under a unified grid system. The temperature gradient field distribution data is embedded in the mechanical equilibrium equation as a thermal load term, while the mechanical stress distribution data is fed back to correct the thermophysical property parameters. A bidirectional coupling control equation set is constructed, in which the temperature gradient field distribution is converted into an equivalent thermal stress term through the thermal expansion coefficient, and the mechanical stress distribution affects the temperature field distribution through deformation work. A step-by-step iterative solution process is performed. The temperature field solver updates the 3D temperature gradient field distribution data, while the mechanical solver updates the mechanical stress distribution data. Boundary conditions are transferred alternately until the global residual reaches the convergence threshold. After the coupling converges, the complete physical quantities of all mesh nodes are extracted, including the node spatial coordinates, three-dimensional temperature gradient components, and six independent stress tensor components. These quantities are organized and stored according to spatial topology, and the output is a comprehensive field matrix containing the spatial coordinates, 3D temperature gradient field distribution, and mechanical stress distribution.

[0035] S3. Use spatial hotspot analysis to perform spatial analysis on the comprehensive field matrix, identify and mark abnormal temperature and mechanical stress areas, perform spatial aggregation and noise filtering through the OPTICS density clustering algorithm, and output abnormal feature vectors.

[0036] S3.1. Input the comprehensive field matrix into the spatial hotspot analysis to identify abnormal areas of temperature gradient and mechanical stress and output the abnormal points.

[0037] It should be noted that the comprehensive field matrix is ​​parsed into spatial coordinates, temperature gradient components, and mechanical stress component data of the grid nodes. The temperature gradient field distribution data identifies abnormally high temperature gradient areas exceeding the set threshold by calculating the standard deviation of the temperature gradient values ​​of each node and the adjacent nodes. The mechanical stress distribution data uses the von Mises stress criterion to evaluate the equivalent stress of each node and mark areas that exceed the material yield strength limit. Spatial association rules are established to cross-validate nodes where both the temperature gradient and mechanical stress exceed the limit, and isolated noise points are excluded. Verified abnormal nodes are recorded as abnormal points, containing complete location coordinates, the degree of temperature gradient abnormality, and the mechanical stress exceeding the limit value, and the abnormal points are finally output.

[0038] S3.2. Use the OPTICS density clustering algorithm to spatially cluster outliers, merge adjacent outliers, filter out isolated noise points, and output the cluster core area.

[0039] It should be noted that outliers are loaded into the computational environment of the OPTICS density clustering algorithm, where a point set topology is constructed based on spatial coordinate information. For each outlier, the Euclidean distance to the kth nearest neighbor is calculated as the core distance, and a minimum sample number parameter is set to determine the neighborhood range. An ordered sequence of reachable distances is generated by traversing all outliers, and potential clusters are identified by analyzing significant depressions in the sequence curve. Density expansion is performed on adjacent outliers that meet the core distance criteria in space to form connected clusters of outliers. A statistically based noise filtering strategy is used to remove isolated points that consistently exceed a dynamic threshold. A convex hull is calculated for each connected cluster to determine the spatial boundary range, retaining the dense region containing the vast majority of outliers as a valid cluster. The cluster core region is then output.

[0040] S3.3. Based on the cluster core area, the spatial statistical feature extraction method is used to extract the spatial coordinates, average temperature gradient and maximum mechanical stress of the abnormal point to generate the abnormal feature vector.

[0041] It should be noted that the cluster core regions are loaded into the feature extraction environment for spatial statistical analysis. For each identified cluster core region, all outlier data records within the region are traversed. The geometric center coordinates of the cluster core region are calculated by taking the average spatial coordinates of all outliers within the region. For the temperature gradient field distribution data, the temperature gradient components in three orthogonal directions are extracted for each outlier within the cluster core region. The average gradient values ​​in the X, Y, and Z directions are calculated, and these three average values ​​together constitute the representative temperature gradient signature of the region. For the mechanical stress distribution data, the six independent stress components of each outlier within the cluster core region are examined. The absolute values ​​of all stress components are compared, and the stress component with the largest value is recorded as the maximum mechanical stress signature of the region. The obtained geometric center coordinates, the average temperature gradient values ​​in the three directions, and the maximum mechanical stress values ​​are arranged and combined in a predetermined order to form a fixed-dimensional feature description vector. A complete outlier feature vector is generated for each cluster core region. This vector contains key parameters: spatial location information, temperature gradient characteristics, and mechanical stress characteristics. The final output is the outlier feature vector corresponding to all cluster core regions.

[0042] S4. Input the abnormal feature vector into the pre-trained intelligent diagnosis model, output the comprehensive fault risk score and perform preliminary diagnosis.

[0043] S4.1. Based on historical abnormal feature vectors, the DNN deep neural network is trained through the back-propagation algorithm to obtain an intelligent diagnosis model that can output a comprehensive fault risk score.

[0044] It should be noted that the historical anomaly feature vector collection is loaded into the DNN deep neural network training environment to construct a DNN deep neural network architecture consisting of an input layer, multiple hidden layers, and an output layer. The input layer nodes strictly correspond to the spatial coordinates, average temperature gradient, and maximum mechanical stress characteristic dimensions of the anomaly feature vectors. Each hidden layer is configured with an activation function to achieve nonlinear transformations, and the output layer uses a specific activation function to map the results into probability values. The training process initializes the weight parameters, sets the learning rate to control the magnitude of parameter updates, and defines the batch size to determine the sample size for each parameter update. A backpropagation algorithm is used for iterative optimization. During the forward propagation phase, the abnormal feature vector is passed layer by layer and the predicted output is calculated. This is then compared with the actual fault type annotated in historical fault records to obtain the loss function value. During the backpropagation phase, the network weight parameters are adjusted layer by layer based on the gradient of the loss function. A separate validation set is used to monitor the training process, triggering an early stopping mechanism when the validation loss fails to decrease after multiple consecutive iterations. The trained DNN deep neural network solidifies the parameters of each layer and establishes a stable mapping relationship from abnormal features to fault risks. After the new abnormal feature vector is input, the network calculation can obtain a standardized comprehensive fault risk score. The score value reflects the probability of fault occurrence, and finally an intelligent diagnosis model that can output a comprehensive fault risk score is obtained.

[0045] S4.2. Input the abnormal feature vector into the intelligent diagnosis model to analyze the comprehensive fault risk score of each abnormal area; It should be noted that the abnormal feature vector is loaded into the input interface of the trained intelligent diagnosis model. After receiving the abnormal feature vector, the intelligent diagnosis model activates the input layer node to pass the spatial coordinates, average temperature gradient, and maximum mechanical stress eigenvalues ​​to the first hidden layer. The hidden layer performs a weighted sum operation and applies an activation function for nonlinear transformation. The processing results are passed to the subsequent hidden layers in sequence. Each hidden layer repeats the feature extraction and transformation operations to gradually abstract high-order feature representations. The output layer receives the output of the last hidden layer and calculates a scalar value in the range of zero to one as a comprehensive fault risk score through a specific activation function. A mapping relationship is established between the score result and the abnormal feature vector. For each input abnormal feature vector, the intelligent diagnosis model outputs the corresponding comprehensive fault risk score.

[0046] S4.3. Based on historical fault data, define low-risk thresholds, high-risk thresholds, and fault classification rules, and determine the comprehensive fault risk score to obtain preliminary diagnostic results.

[0047] It should be noted that the correspondence between the comprehensive fault risk score and the actual fault occurrence in the past cases was analyzed. According to the statistical distribution of historical fault data, the threshold boundary for distinguishing low-risk and high-risk states is determined. The low-risk threshold is defined as the upper limit of the score interval with a low probability of fault occurrence, and the high-risk threshold is defined as the lower limit of the score interval with a significantly increased probability of fault occurrence. A multi-level fault classification rule is established. When the comprehensive fault risk score is between zero and the low-risk threshold, it is determined to be a "normal state", indicating that the current abnormal area is in a safe operating range and only requires routine monitoring; when the comprehensive fault risk score exceeds the low-risk threshold but does not reach the high-risk threshold, it is determined to be a "warning state", indicating that potential fault signs appear in the abnormal area, and it is necessary to strengthen monitoring and prepare maintenance plans; when the comprehensive fault risk score exceeds the high-risk threshold, it is determined to be a "fault state"; The comprehensive fault risk score output by the intelligent diagnostic model is compared with the defined threshold range, and the risk level of each abnormal area is labeled according to the preset classification rules. The spatial coordinates of the abnormal area, the comprehensive fault risk score, and the risk level label are combined to form a structured diagnostic record. The results of all abnormal areas are summarized to generate a preliminary diagnostic report.

[0048] S5. Based on the preliminary diagnosis results, the fault data in the historical case library is retrieved for comparison and verification, and the preliminary diagnosis results are optimized through machine learning algorithms to generate the final diagnosis results.

[0049] S5.1. Based on the preliminary diagnosis results, search the historical case library, match the fault data records under similar working conditions, and output the Top-K similar cases.

[0050] It should be noted that the preliminary diagnosis results are input into the case retrieval engine, and the spatial coordinates of the abnormal area, the comprehensive fault risk score and the risk level in the diagnosis results are extracted as retrieval features. A multi-dimensional index structure is established in the historical case library, and the index fields contain the abnormal feature vectors, fault types and disposal records of historical cases. A similarity calculation algorithm is used to compare the characteristics of the current abnormal area with the records in the historical case library one by one, and the spatial position similarity, risk score deviation and fault type matching are calculated. A composite weight coefficient is set, with spatial position similarity having the highest weight, risk score deviation having the second highest weight, and fault type matching having the lowest weight. Historical cases are sorted in descending order according to the comprehensive similarity score, and the top K case records are intercepted. The Top-K similar cases are validated, and the output Top-K similar case set contains complete information such as case number, similarity score, historical fault details and disposal plan.

[0051] S5.2. Input the preliminary diagnosis results and the treatment effect data of the Top-K similar cases into the machine learning algorithm, optimize the diagnosis conclusion through probability weighting, and generate the final diagnosis result.

[0052] It should be noted that the preliminary diagnosis results and the Top-K similar case set are loaded into the machine learning optimization environment, and the risk level classification in the preliminary diagnosis results and the treatment effect records in the Top-K similar cases are extracted as the optimization basis; The treatment effect record contains complete maintenance data such as the maintenance measures actually taken in historical cases, changes in the operating status of the equipment after maintenance, and the fault recurrence cycle. After quantitative coding, it forms a computable treatment effect indicator. The treatment effect indicator specifically includes the evaluation values ​​of three dimensions: the matching degree between the maintenance measures and the fault type, the degree of equipment performance recovery after maintenance, and the interval time of fault recurrence. Each dimension is converted into a standardized value according to the predefined scoring standard. The matching degree between the maintenance measures and the fault type reflects the pertinence of the treatment plan, the degree of equipment performance recovery after maintenance measures the maintenance effect, and the interval time of fault recurrence evaluates the durability of the maintenance plan. The evaluation values ​​of the three dimensions are weighted and fused to form a comprehensive treatment effect score. The higher the score, the more successful the treatment plan of the historical case. The comprehensive treatment effect score and the case similarity score together constitute the dual basis for optimizing the diagnosis conclusion; A probability-weighted matrix is ​​constructed, where the rows correspond to the possible fault types of the preliminary diagnosis results, the columns correspond to the treatment effects of the top-K similar cases, and the matrix elements are the normalized values ​​of the case similarity scores. Using the Bayesian inference framework, the treatment effects of the top-K similar cases are used as observational evidence, and the preliminary diagnosis results are used as prior probabilities to calculate the posterior probability distribution. For each fault type hypothesis, the weighted treatment effect data of the relevant cases are accumulated and normalized to obtain the revised probability. A probability confidence threshold is set based on the accuracy of historical diagnosis cases. When the revised probability exceeds the probability confidence threshold (0.7-0.9), the diagnosis conclusion is confirmed; otherwise, the preliminary diagnosis result is retained. The fault type judgment after probability weighted optimization is integrated with the original risk level classification to form a final diagnosis result that includes the fault type, risk score, confidence level, and recommended treatment plan.

[0053] S6. Provide dynamic warning based on the final diagnosis results and generate a fault detection report.

[0054] S6.1. Based on the final diagnosis results, the rule engine automatically matches the warning rule library to trigger multi-level warnings.

[0055] It should be noted that the final diagnosis result is input into the rule engine execution environment, and the fault type, risk score and confidence level fields in the diagnosis result are parsed. The early warning rule library establishes a multi-level index structure according to the fault type and risk level. Each early warning rule clearly defines the triggering conditions, warning level and response measures. The rule engine matches the final diagnosis result with the conditional expression in the early warning rule library one by one. When the fault type is completely matched and the risk score reaches the rule threshold, the corresponding early warning rule is activated. For the early warning rules that are successfully matched, the preset warning level and response measures are extracted to generate a standardized early warning notification. The early warning notification contains structured fields such as the coordinates of the fault location, the warning level code, the recommended disposal measures and the timeliness requirements. The multi-level warnings are sorted according to the degree of urgency. High-risk warnings are pushed immediately, and medium and low-risk warnings enter the pending queue. All triggered warning records are stored in the early warning log database.

[0056] S6.2. The final diagnosis results, multi-level warnings and distribution transformer operating status data are aligned in time and space through the feature fusion algorithm to construct structured diagnostic data.

[0057] It should be noted that the final diagnostic results, multi-level warning records, and distribution transformer operating status data are loaded into the feature fusion processing environment. The three data sources are time-series aligned according to a unified timestamp to ensure temporal consistency of the data records. Based on the physical structure topology of the distribution transformer, a mapping relationship is established between the coordinates of the abnormal area in the final diagnostic result and the spatial location of the operating status monitoring point. The warning level field of the multi-level warning record is correlated with the changing trends of the operating status parameters. A feature-level fusion strategy is used to combine the fault type code of the final diagnostic result, the warning level identifier of the multi-level warning, and the key parameters of the distribution transformer operating status. A five-tuple data structure is constructed, comprising time, space, diagnosis, warning, and operation dimensions. Each data unit records the complete time stamp, spatial location, diagnostic conclusion, warning level, and operating parameters, forming structured diagnostic data with temporal and spatial correlation.

[0058] S6.3. Based on the structured diagnostic data, intelligent analysis is performed through natural language generation methods to obtain a fault detection report.

[0059] It should be noted that the structured diagnostic data is input into the natural language generation engine, which parses the time stamp, spatial location, diagnostic conclusion, warning level, and operating parameters in the five-tuple data structure. A report template library is established, defining chapter structures such as fault overview, anomaly location, risk analysis, and disposal recommendations. Key features in the structured diagnostic data are extracted, time dimension data is converted into a readable time description, spatial location data is mapped into a description of the physical location of the equipment, diagnostic conclusions and warning levels are converted into risk level statements, and quantitative analysis statements are generated for the changing trends of operating parameters. Using a rule-based text generation method, structured fields are populated into corresponding locations within the report template according to semantic relationships. The diagnostic conclusion section combines the fault type and risk score to generate a severity description. The anomaly location section integrates spatial coordinates and the distribution transformer structure diagram to generate a location description. The risk analysis section links historical cases and current parameters to generate trend forecasts. The action recommendations section matches the early warning rule base to generate operational instructions. The generated text fragments undergo syntax verification and logical coherence checks before being assembled into a complete fault detection report that complies with technical document specifications. The fault detection report includes standardized sections such as title, equipment information, detection time, fault description, analysis conclusions, and recommended measures.

[0060] This embodiment also provides a distribution transformer monitoring system, including: Data acquisition module, coupling analysis module, anomaly extraction module, diagnosis module, optimization module and report generation module; Data acquisition module, used to collect distribution transformer operating status data in real time and preprocess it to obtain a standardized multi-source data set; A coupling analysis module is used to obtain the three-dimensional temperature gradient field distribution and mechanical stress distribution of the distribution transformer based on a standardized multi-source data set, and output a comprehensive field matrix through a field coupling analysis method; The anomaly extraction module is used to perform spatial analysis on the comprehensive field matrix using spatial hotspot analysis, identify and mark abnormal temperature and mechanical stress areas, perform spatial aggregation and noise filtering using the OPTICS density clustering algorithm, and output anomaly feature vectors; The diagnostic module is used to input abnormal feature vectors into the pre-trained intelligent diagnostic model, output a comprehensive fault risk score and perform preliminary diagnosis; The optimization module is used to retrieve fault data from the historical case library based on the preliminary diagnosis results for comparison and verification, and optimize the preliminary diagnosis results through machine learning algorithms to generate the final diagnosis results; The report generation module is used to provide dynamic early warning based on the final diagnosis results and generate a fault detection report.

[0061] This embodiment also provides a computer device, which is applicable to the distribution transformer monitoring method, including: 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 the distribution transformer monitoring method proposed in the above embodiment.

[0062] 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.

[0063] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the distribution transformer monitoring method proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0064] In summary, the present invention calculates the three-dimensional temperature gradient field distribution and mechanical stress distribution through finite element analysis and generalized Hooke's law, and adopts the field coupling analysis method to establish a thermal-mechanical bidirectional action model, thereby realizing the collaborative calculation and closed-loop feedback of the temperature field and stress field, and accurately reflecting the spatial distribution characteristics of the temperature gradient and stress tensor; at the same time, the OPTICS density clustering algorithm is used to intelligently aggregate the spatial hotspot analysis results, extract the spatial coordinates, average temperature gradient and maximum mechanical stress characteristics of the cluster core area, realize the automatic identification and quantitative characterization of distributed fault areas, and achieve improved fault diagnosis accuracy and reliability.

[0065] 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 distribution transformer monitoring method, characterized in that: include, Collect distribution transformer operating status data in real time and preprocess it to obtain a standardized multi-source data set; Based on standardized multi-source data sets, the three-dimensional temperature gradient field distribution and mechanical stress distribution of the distribution transformer are obtained, and the comprehensive field matrix is ​​output through the field coupling analysis method; Spatial hotspot analysis is used to perform spatial analysis on the comprehensive field matrix to identify and mark abnormal temperature and mechanical stress areas. The OPTICS density clustering algorithm is used to perform spatial aggregation and noise filtering, and output abnormal feature vectors. Input the abnormal feature vector into the pre-trained intelligent diagnosis model, output the comprehensive fault risk score and perform preliminary diagnosis; Based on the preliminary diagnosis results, the fault data in the historical case library is retrieved for comparison and verification. The preliminary diagnosis results are optimized through machine learning algorithms to generate the final diagnosis results. Dynamic early warning is provided through the final diagnostic results, and a fault detection report is generated.

2. The distribution transformer monitoring method according to claim 1, wherein: The distribution transformer operating status data includes temperature data, core vibration data, shell vibration data, load current data, operating voltage data, and insulating oil and gas data; The preprocessing includes noise suppression through sliding window filtering, data normalization using the Z-score standardization method, and missing value compensation using linear interpolation.

3. The distribution transformer monitoring method according to claim 2, wherein: The method is based on a standardized multi-source data set to obtain the three-dimensional temperature gradient field distribution and mechanical stress distribution of the distribution transformer, and output a comprehensive field matrix through a field coupling analysis method. The specific steps are as follows: Based on standardized multi-source data sets, the finite element analysis method is used to calculate the three-dimensional temperature gradient field distribution, and the mechanical stress distribution is obtained by generalized Hooke's law; The three-dimensional temperature gradient field distribution and mechanical stress distribution are input into the field coupling analysis method to establish a thermal-mechanical coupling relationship, and a comprehensive field matrix containing multidimensional data of spatial coordinates, three-dimensional temperature gradient field distribution and mechanical stress distribution is output.

4. The distribution transformer monitoring method according to claim 3, wherein: The spatial hotspot analysis is used to perform spatial analysis on the comprehensive field matrix, identify and mark abnormal temperature and mechanical stress areas, perform spatial aggregation and noise filtering using the OPTICS density clustering algorithm, and output abnormal feature vectors. The specific steps are as follows: The comprehensive field matrix is ​​input into the spatial hot spot analysis to identify abnormal areas of temperature gradient and mechanical stress and output the abnormal points; The outliers are spatially clustered using the OPTICS density clustering algorithm, which merges adjacent outliers and filters out isolated noise points, outputting the cluster core area. Based on the cluster core area, the spatial coordinates, average temperature gradient and maximum mechanical stress of the abnormal points are extracted through the spatial statistical feature extraction method to generate the abnormal feature vector.

5. The distribution transformer monitoring method according to claim 4, wherein: The abnormal feature vector is input into the pre-trained intelligent diagnosis model, a comprehensive fault risk score is output, and a preliminary diagnosis is performed. The specific steps are as follows: Based on historical abnormal feature vectors, the DNN deep neural network is trained through the back-propagation algorithm to obtain an intelligent diagnosis model that can output a comprehensive fault risk score; The abnormal feature vector is input into the intelligent diagnosis model to analyze the comprehensive fault risk score of each abnormal area; Based on historical fault data, low-risk thresholds, high-risk thresholds and fault classification rules are defined, and the comprehensive fault risk score is judged to obtain preliminary diagnosis results.

6. The distribution transformer monitoring method according to claim 5, wherein: The method of retrieving fault data from the historical case library based on the preliminary diagnosis results for comparison and verification, optimizing the preliminary diagnosis results through a machine learning algorithm, and generating a final diagnosis result is as follows: Based on the preliminary diagnosis results, the historical case library is searched to match the fault data records under similar working conditions and output the Top-K similar cases; The preliminary diagnosis results and the treatment effect data of the Top-K similar cases are input into the machine learning algorithm, and the diagnosis conclusions are optimized through probability weighting to generate the final diagnosis results.

7. The distribution transformer monitoring method according to claim 6, wherein: The final diagnosis result is used to provide a dynamic early warning and generate a fault detection report. The specific steps are as follows: Based on the final diagnosis results, the rule engine automatically matches the warning rule library to trigger multi-level warnings; The final diagnosis results, multi-level warnings and distribution transformer operating status data are aligned in time and space through feature fusion algorithms to construct structured diagnostic data; Based on structured diagnostic data, intelligent analysis is performed through natural language generation methods to obtain a fault detection report.

8. A distribution transformer monitoring system, based on the distribution transformer monitoring method according to any one of claims 1 to 7, characterized in that: Including data acquisition module, coupling analysis module, anomaly extraction module, diagnosis module, optimization module and report generation module; Data acquisition module, used to collect distribution transformer operating status data in real time and preprocess it to obtain a standardized multi-source data set; A coupling analysis module is used to obtain the three-dimensional temperature gradient field distribution and mechanical stress distribution of the distribution transformer based on a standardized multi-source data set, and output a comprehensive field matrix through a field coupling analysis method; The anomaly extraction module is used to perform spatial analysis on the comprehensive field matrix using spatial hotspot analysis, identify and mark abnormal temperature and mechanical stress areas, perform spatial aggregation and noise filtering using the OPTICS density clustering algorithm, and output anomaly feature vectors; The diagnostic module is used to input abnormal feature vectors into the pre-trained intelligent diagnostic model, output a comprehensive fault risk score and perform preliminary diagnosis; The optimization module is used to retrieve fault data from the historical case library based on the preliminary diagnosis results for comparison and verification, and optimize the preliminary diagnosis results through machine learning algorithms to generate the final diagnosis results; The report generation module is used to provide dynamic early warning based on the final diagnosis results and generate a fault detection report.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the distribution transformer monitoring method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the distribution transformer monitoring method according to any one of claims 1 to 7 are implemented.

Images