Power distribution network transformer operation state monitoring and energy efficiency collaborative optimization method

Abstract

The embodiment of the invention provides a power distribution network transformer operation state monitoring and energy efficiency collaborative optimization method, which is applied to the technical field of power grid remote control and telemetering and network control systems, and is characterized in that the system synchronously acquires transformer state data through acoustics, vibration, load and other sensors, and performs feature fusion extraction by adopting deep learning; and a health reference library of the sub-load intervals is established. Through multi-dimensional difference calculation of real-time monitoring data and a health template, a health degree score is generated and state early warning is realized. An energy efficiency adaptive optimization module is innovatively introduced, an optimal load set value is solved based on a multi-objective optimization algorithm, and an optimization strategy is executed through closed-loop control. The system is suitable for various scenes such as an urban power grid, a new energy power station, an industrial environment and a remote area, realizes the conversion from fixed cycle maintenance to state-driven operation and maintenance, improves the operation reliability, the service life and the operation and maintenance efficiency of the transformer, and provides a solution for the full-life-cycle management of power equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power grid remote control telemetry and network control systems, and in particular to a power distribution network transformer operating state monitoring and energy efficiency collaborative optimization method. BACKGROUND

[0002] As the core energy conversion device in the power grid, the operating state of the transformer is directly related to the safety and reliability of the entire power system. With the development of social economy and the increasing demand for electricity, the operating environment faced by the transformer is becoming increasingly complex.

[0003] Traditional transformer operation and maintenance generally adopts periodic maintenance and monitoring mode based on a single parameter threshold. This mode has obvious drawbacks: fixed cycle maintenance lacks pertinence, which may lead to over-maintenance or insufficient maintenance, wasting resources and causing safety hazards; traditional monitoring methods, such as electrical quantities and oil chromatographic analysis, cannot sensitively and accurately capture the progressive changes of internal mechanical states of the transformer, such as winding deformation, core loosening and insulation deterioration, and cannot achieve early warning of faults.

[0004] In modern power grids, transformers usually work in groups. However, the current management mode mainly focuses on independent monitoring and optimization of single devices, lacks collaborative optimization strategies from the global perspective of the system, and cannot fully tap the potential of the device group.

[0005] In summary, the field of power distribution network transformer state monitoring and energy efficiency management needs a complete technical system of multi-modal perception, working condition adaptive benchmark, health quantification evaluation, predictive optimization and group collaborative control to solve the problem that device reliability, system economy and operation and maintenance efficiency are difficult to balance under complex operating environment. SUMMARY

[0006] Embodiments of the present application provide a power distribution network transformer operating state monitoring and energy efficiency collaborative optimization method, which realizes an intelligent operation and maintenance technology system that integrates multi-dimensional information, adapts to different operating conditions, has edge intelligence and supports group collaborative decision-making, to realize the transition of the transformer from passive maintenance to active early warning. To achieve the above purpose, the present application adopts the following technical solutions: A power distribution network transformer operating state monitoring and energy efficiency collaborative optimization method, the method comprising the following steps: Collecting sound signals, vibration acceleration signals and load rate time series data during transformer operation, extracting key features through feature fusion processing and generating a multi-dimensional health state feature vector; Based on the multi-dimensional health state feature vector and transformer operating history data, a comprehensive feature model including acoustic features, vibration features and their correlation with load rate is constructed, and the load response sensitivity coefficient under different loads is calculated; Based on the comprehensive feature model, standard health state feature templates are generated and stored for different load level intervals, thereby constructing a hierarchical load feature template library; In the online monitoring stage, the corresponding standard health state feature template is retrieved from the feature template library according to the current transformer load rate; the multi-dimensional difference index between the multi-dimensional health state feature vector currently collected and generated and the retrieved corresponding standard health state feature template is calculated; The multi-dimensional difference index is fused and calculated to generate a health degree score reflecting the deviation degree of the transformer operating state, and the change trend of the health degree score is calculated based on the time sequence of the health degree score; The health degree score is compared with the multi-level diagnosis threshold, and the multi-level diagnosis threshold includes a fault threshold, a warning threshold and a high-efficiency operation threshold; when the health degree score is lower than the fault threshold, an early warning work order is generated and uploaded to the operation and maintenance management system; When the health degree score is between the warning threshold and the high-efficiency operation threshold, energy efficiency adaptive optimization is performed, and based on the health degree score change trend and the load response sensitivity coefficient, the load rate set value of the transformer is determined through a multi-objective optimization algorithm; According to the load rate set value, closed-loop feedback adjustment is performed to realize transformer load adjustment; When system-level optimization is needed, the boundary parameters of the economic operation interval are determined based on the load rate set value, and the load distribution strategy is adjusted according to the boundary parameters to realize the coordinated adjustment of operation stability and energy efficiency balance.

[0007] Among them, the execution of energy efficiency adaptive optimization, based on the health degree score change trend and the load response sensitivity coefficient, determines the load rate set value of the transformer through a multi-objective optimization algorithm, including: Based on the health degree score change trend and the load response sensitivity coefficient, a multi-objective optimization function is established, and the multi-objective optimization function takes improving the health degree score and suppressing the increase of the load response sensitivity coefficient as the optimization objective; Taking the multi-objective optimization function as a fitness function, a health operation boundary condition including transformer load rate upper and lower limit constraints and load rate change rate constraints is constructed; A non-dominated sorting genetic algorithm with elitist strategy is adopted to perform multi-objective search based on the fitness function in the feasible solution space defined by the health operation boundary condition to obtain a Pareto front solution set; The relative closeness of each solution on the Pareto front solution set to the ideal solution is calculated, and a selected solution is determined from the Pareto front solution set according to the size order of the relative closeness; The load rate value corresponding to the selected solution is output as the load rate set value.

[0008] The calculating the relative closeness of each solution on the Pareto frontier solution set to the ideal solution and determining the selected solution from the Pareto frontier solution set according to the size order of the relative closeness comprises: The relative closeness of each Pareto frontier solution is calculated by using an entropy weight method-based technique for order preference by similarity to ideal solution (TOPSIS); The solutions in the Pareto frontier solution set are arranged in descending order according to the corresponding relative closeness values, and the solution arranged in the first place is selected as the selected solution; The relative closeness is determined according to the Euclidean distance between the solution set and the ideal solution and the information entropy distribution characteristics of the solution set.

[0009] The determining the boundary parameters of the economic operation interval based on the load rate set value comprises: The change rate of the health degree score corresponding to the load rate set value is calculated; The boundary function of the economic operation interval is constructed based on the change rate and the energy efficiency index; The upper and lower limit values output by the boundary function are used as the boundary parameters of the economic operation interval.

[0010] The adjusting the load distribution strategy according to the boundary parameters comprises: The target load rate of each transformer unit is determined according to the boundary parameters; The load distribution balancing algorithm is executed under the constraint of the target load rate to obtain a load adjustment result, and the transformer operation state is adjusted according to the load adjustment result to realize load redistribution.

[0011] The generating and storing the standard health state feature template for different load level intervals comprises: For each load level interval, the principal component analysis method is used to reduce the dimension of the features output by the comprehensive feature model, and the main feature components are extracted; The time evolution constraint is introduced, and the sliding window mechanism is used to analyze the change trend of the features on the time axis to ensure the feature stability information; The standard health state feature template is generated by using the maintenance mechanism corresponding to the feature stability information, and the maintenance mechanism corresponding to the feature stability information includes feature smoothing and outlier removal; The standard health state feature template of each load level interval is stored in the hierarchical load feature template library, and an index is established.

[0012] The calculating the multi-dimensional difference index between the multi-dimensional health state feature vector currently collected and generated and the corresponding standard health state feature template comprises: calculating a cosine distance between the current multi-dimensional health state feature vector and the retrieved standard health state feature template, the cosine distance being used to represent the spatial distribution difference between the two; extracting the trend of the current feature vector over time series, and comparing it with the expected evolution trend of the standard health state feature template, calculating a weighted time series difference coefficient, wherein the weight is assigned based on feature importance; linearly combining the cosine distance and the weighted time series difference coefficient as input parameters to generate a multi-dimensional difference indicator, which is used to comprehensively reflect the deviation degree of the current state from the standard health state feature template.

[0013] wherein the fusion calculation of the multi-dimensional difference indicator to generate a health degree score reflecting the deviation degree of the transformer operating state includes: calculating feature fluctuation stability by calculating the variance and standard deviation of the current feature vector within a sliding window to evaluate the degree of feature fluctuation; calculating feature entropy change rate based on information entropy theory to analyze the uncertainty change of feature distribution; calculating time series correlation by evaluating the correlation between the current feature sequence and the standard health state feature template sequence through Pearson correlation coefficient; assigning weights to feature fluctuation stability, feature entropy change rate and time series correlation, and performing weighted summation to generate a health degree score, wherein the weights are adaptively adjusted according to the transformer type and operating environment.

[0014] wherein the comparison of the health degree score with the multi-level diagnosis threshold includes: presetting a fault threshold, a warning threshold and an efficient operation threshold, wherein the fault threshold is set based on historical fault data, the warning threshold is set based on a device aging model, and the efficient operation threshold is set based on the optimal energy efficiency value; when the health degree score is lower than the fault threshold, determining that the transformer is in a fault state; when the health degree score is between the fault threshold and the warning threshold, determining that the transformer is in a warning state; when the health degree score is between the warning threshold and the efficient operation threshold, determining that the transformer is in a normal operating state; when the health degree score is higher than the efficient operation threshold, determining that the transformer is in an efficient operation state.

[0015] wherein the generation of a warning work order and uploading to the operation and maintenance management system includes: when the health degree score is lower than the fault threshold, automatically generating a warning work order, the work order including the transformer identifier, the detection time, the health degree score, the abnormal feature description and the recommended repair measures; Packaging the early warning work order into a standardized data format conforming to the interface protocol of the operation and maintenance management system; Uploading the early warning work order to the operation and maintenance management system through an encrypted communication protocol and triggering an alarm notification.

[0016] From the above technical solutions, the present application has the following beneficial effects: 1. The method realizes a fundamental mode change of transformer operation and maintenance management from "passive response, experience-driven" to "active early warning, intelligent-driven", overcomes the limitations of traditional single parameter monitoring through a multi-modal information fusion perception system, can sensitively capture early progressive degradation characteristics of the internal mechanical state of the transformer, realizes early warning and accurate diagnosis of faults. At the same time, the system links the health state evaluation result with the operation control strategy, forms a closed-loop intelligent operation and maintenance system of "monitoring-evaluation-early warning-optimization-control", realizes the transition from fixed period maintenance to state-driven operation and maintenance, improves the operation and maintenance efficiency and reduces the cost, realizes the coordinated improvement of safety and energy efficiency under the premise of ensuring the safety of equipment through health state-based energy efficiency adaptive optimization, and provides complete technical support for the whole life cycle management of power equipment.

[0017] 2. For the urban central business district, the system ensures reliable operation and optimal control under high-density and periodic load through load-adaptive feature fusion and refined health benchmark library. In the face of strong volatility of new energy power generation, the system successfully balances power generation benefit and equipment life through working condition identification, predictive health management and flexible adjustment strategy. In complex industrial environments, the fault-oriented feature extraction and mechanism data fusion diagnosis method improves the fault identification accuracy under harmonic pollution and impact load. For remote areas, highly intelligent edge computing and neighbor coordination mechanism effectively overcome the constraints of communication and environment, ensuring reliable monitoring under unattended operation. By building a transformer group collaborative optimization system, the limitations of single device independent optimization are broken through, realizing the leap from local optimum to system overall efficiency maximization, and improving the overall resilience, economy and intelligence level of the power grid. BRIEF DESCRIPTION OF DRAWINGS

[0018] The present application will be further described below with reference to the accompanying drawings.

[0019] Figure 1 The first flowchart of the embodiment of the present application; Figure 2 The second flowchart of the embodiment of the present application; Figure 3 The third flowchart of the embodiment of the present application; Figure 4 The fourth flowchart of the embodiment of the present application; Figure 5 The fifth flowchart of the embodiment of the present application; Detailed Implementation

[0020] The terms "first," "second," and "third," etc., used in this application specification, claims, and drawings are for distinguishing different objects, not for specifying a particular order.

[0021] In the embodiments of this application, the words "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the words "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0022] Research has revealed that traditional transformer operation and maintenance generally employs a periodic inspection and monitoring model based on a single parameter threshold. This model has significant drawbacks: fixed-cycle inspections lack specificity, potentially leading to over-maintenance or under-maintenance, wasting resources and creating safety hazards; traditional monitoring methods, such as electrical quantity and oil chromatography analysis, are insufficient to sensitively and accurately capture the internal mechanical state of the transformer, such as the gradual changes in winding deformation, core loosening, and insulation deterioration, failing to provide early warning of faults.

[0023] To address the aforementioned issues, this application provides a method for synergistic optimization of the operating status and energy efficiency of distribution network transformers: Example

[0024] To solve the above problems, such as Figure 1 As shown, this embodiment focuses on the intelligent operation and maintenance management of the main transformer in a key hub substation in the central business district of a large city. The power supply load in this area exhibits significant temporal periodicity, with highly concentrated commercial load during the day and residential power consumption dominating at night, showing a clear difference in load characteristics between weekdays and holidays. As a critical piece of equipment in the power grid, the transformer gradually enters a period of performance change after long-term operation. Traditional monitoring methods are insufficient to accurately assess the gradual changes in its internal mechanical state, let alone optimize and adjust operating parameters based on its health status.

[0025] At the data perception level, the system constructs a multi-dimensional information synchronous acquisition system. Acoustic sensors are responsible for capturing the characteristic sounds generated by the magnetostriction of the transformer core, the specific frequency sound waves formed by the vibration of the windings, and the transient acoustic signals caused by partial discharge; vibration sensors monitor the mechanical vibration state of the transformer body and auxiliary support structure, covering a wide frequency band of vibration characteristics from low frequency to high frequency; the load monitoring unit records the transformer load change process, forming a complete load time sequence characteristic.

[0026] The feature fusion processing link adopts a deep neural network architecture to realize intelligent feature extraction. The acoustic signal is processed by time-frequency transformation and converted into a time-frequency spectrum to input a convolutional neural network, which automatically learns the harmonic distribution characteristics and transient acoustic patterns. The vibration signal is decomposed and the features are mined to extract the energy distribution and vibration modal characteristics of each frequency band. The load data is analyzed for time series features to capture the load change pattern. The system innovatively introduces a weighted fusion mechanism to automatically adjust the fusion weights of acoustic and vibration features according to the current load level of the transformer: in the low load state, the acoustic feature analysis is emphasized, and in the high load state, the vibration feature analysis is emphasized, realizing adaptive matching of the sensing focus and operating conditions.

[0027] In the health benchmark construction phase, the system establishes standard health templates for each load interval based on long-term historical operation data. First, the load range of the transformer is divided into multiple continuous intervals. For each load interval, the set of acoustic and vibration features under normal state in the interval is extracted, and through feature dimension reduction and time series smoothing processing, a feature template representing the health state of the load interval is formed. All templates together form a hierarchical health feature library of the transformer, which serves as a reference system for subsequent health state assessment.

[0028] In the online monitoring phase, the system tracks the load level of the transformer and automatically selects the standard template corresponding to the load interval from the health feature library. By calculating the multi-dimensional differences between the features and the standard template, including spatial distribution differences and time series evolution trend differences, the system can sensitively capture state abnormalities. These difference indicators are converted into intuitive health scores after weighted fusion, and their change trends are calculated to provide decision basis for state prediction.

[0029] When the health score enters the warning interval, the system starts the energy efficiency adaptive optimization module. This module takes improving the health change trend and reducing the load response sensitivity as the optimization objective, and solves the optimal load setting value through a multi-objective optimization algorithm under the consideration of operation safety constraints. This setting value takes into account the improvement needs of the device health state and also considers the requirements of operation economy.

[0030] The system executes the optimization results through a closed-loop control mechanism. According to the actual operation requirements of the power grid, two control modes can be selected: the direct control mode accurately adjusts the load to the set value; the interval control mode constructs a safe operation interval centered on the set value and flexibly adjusts the load distribution within the interval. This differentiated control strategy ensures the feasibility of the optimization scheme in complex power grid environments.

[0031] This scheme realizes the technical leap from single parameter threshold judgment to multi-modal fusion intelligent evaluation in transformer state monitoring, the mode change from fixed period maintenance to state driven operation and maintenance, and the system upgrade from independent safety monitoring to safety and energy efficiency collaborative optimization. By building a complete intelligent operation and maintenance technology system, an innovative solution is provided for the whole life cycle management of important power equipment. Embodiments

[0032] As shown in Figure 2 , specifically: the transformer operating conditions under new energy power generation environment have significant particularity. The load characteristics of wind farm booster transformers are directly affected by natural wind speed changes, showing irregular and severe fluctuation characteristics; photovoltaic power station transformers need to cope with the power step changes caused by day and night alternation and cloud cover. This strong fluctuation operation mode makes the transformer bear the challenge of mechanical stress and thermal stress that traditional power systems have never encountered, and puts forward new requirements for equipment state monitoring.

[0033] The perception level is specially optimized for the characteristics of new energy scenarios. The acoustic monitoring module uses wideband acquisition technology to focus on capturing the transient acoustic characteristics generated inside the transformer when the power changes rapidly. When the power change rate exceeds a certain threshold, the system automatically starts the high-frequency sampling mode to record the transient process characteristics completely. The vibration monitoring module intensifies the monitoring ability of the frequency band directly related to electromagnetic force, especially the tracking analysis of winding and core vibration state during load mutation.

[0034] The feature extraction link innovatively introduces an operating condition recognition mechanism. The system analyzes the load change trend and subdivides the operating state into multiple operating condition modes such as steady-state operation, slow fluctuation, rapid fluctuation, and impact load. For different operating conditions, the system automatically calls the corresponding feature analysis strategy. For example, during stable operation, the focus is on analyzing the long-term evolution trend of the feature; during rapid fluctuation, the focus is on the instantaneous response characteristics of the feature; and in the impact load operating condition, the focus is on the peak value characteristics and recovery process of the feature.

[0035] The health benchmark library is constructed taking into full consideration the characteristics of new energy. In addition to the traditional load rate interval division, a health benchmark based on power change rate is established. The system records the normal response range of transformer acoustic and vibration characteristics under different power change rates, forming a multi-dimensional health reference system. This refined benchmark setting enables the system to accurately distinguish between normal power fluctuation response and potential equipment abnormalities.

[0036] The state evaluation module adopts a dual diagnosis mechanism. First, a rapid comparison of features and corresponding health benchmarks is made to generate a basic health score; second, cross-condition trend correlation analysis is performed to identify possible progressive degradation characteristics. When a feature is found to be abnormal in multiple operating conditions, the system raises the warning level and prompts the possible existence of a substantial defect.

[0037] The energy efficiency optimization module realizes the innovation of predictive health management. The system accesses meteorological forecast data, combines the current health status of the equipment, and predicts the operating environment that the transformer will face in the future period. When it is predicted that an operating mode that is not conducive to the health of the equipment may occur, the system generates an operating optimization suggestion in advance. For example, when it is predicted that there will be a continuous windy weather, the system suggests to appropriately smooth the power output, although a small amount of power generation income is lost, but the thermal cycle stress of the equipment is effectively reduced, and the insulation life is significantly prolonged.

[0038] The control execution layer adopts a flexible adjustment strategy. The system realizes the optimization adjustment of the operating state of the transformer by coordinating the output distribution of multiple power generation units in the new energy power station. Under the premise of ensuring the total output power, the load fluctuation of the transformer is smoothed through an optimization control algorithm, and the transient impact is controlled within the bearing range of the equipment. This preventive control based on the health status effectively balances the dialectical relationship between power generation benefit and equipment life.

[0039] The present scheme breaks through the application limitations of traditional monitoring systems in new energy scenarios, and establishes a special monitoring and evaluation system suitable for fluctuating loads. Through working condition self-adaptive perception, predictive health management and flexible operation control, a technical leap from passive bearing to active adaptation of new energy transformers is realized, providing key technical support for the reliable operation of power equipment under the background of high proportion of new energy access. Embodiments

[0040] As shown in Figure 3 , specifically: the transformer in the industrial environment faces multiple technical challenges: harmonic pollution brought by power electronic equipment, impact load caused by starting and stopping of large motors, cyclic load changes in continuous production processes, etc. These special working conditions not only accelerate the aging process of the equipment, but also make the extraction and identification of fault features extremely complex, and the traditional monitoring method cannot meet the precise diagnosis demand.

[0041] The perception layer adopts a multi-sensor cooperative working mode. The acoustic monitoring system is equipped with a high-range acquisition unit, which can simultaneously capture weak discharge signals and strong mechanical noise. Through advanced sound separation technology, the system effectively distinguishes the acoustic characteristics of the transformer body from the background environmental noise. The vibration monitoring network is distributed at key parts of the transformer, and a complete mechanical vibration panoramic view is constructed, providing sufficient data support for state evaluation.

[0042] For the industrial harmonic environment, the system develops a special electrical-mechanical correlation analysis algorithm. By synchronously collecting transformer electrical quantity signals and acoustic vibration signals, a quantitative relationship model between harmonic current and specific frequency mechanical vibration is established. When abnormal harmonic content is detected, the system automatically starts correlation analysis to determine whether harmful mechanical resonance has been triggered, providing an important basis for early fault diagnosis.

[0043] The feature extraction module adopts a fault-oriented design. Based on the analysis of a large number of industrial transformer fault cases, the system focuses on extracting feature parameters that are strongly related to typical fault modes. For example, for winding deformation faults, the system focuses on the asymmetric characteristics of axial and radial vibrations; for core faults, the system focuses on monitoring acoustic emission characteristics in specific frequency bands; for insulation degradation, the system focuses on analyzing the evolution rules of partial discharge acoustic characteristics. This targeted feature extraction strategy significantly improves the accuracy of fault identification.

[0044] The health assessment system adopts a hierarchical early warning mechanism. The system establishes a multi-level evaluation system from normal state, attention state, abnormal state to dangerous state. Each level corresponds to clear feature criteria and disposal suggestions. When the system detects a state change, it not only gives the current level assessment, but also predicts the time window of level change based on trend analysis, providing sufficient basis for operation and maintenance decision-making.

[0045] The diagnostic reasoning module combines mechanism models and data-driven methods. The system has a built-in transformer multi-physical field coupling model that can simulate the acoustic and vibration response characteristics under different fault types. When abnormal features are detected, the system starts a multi-hypothesis verification process, analyzes the matching degree between actual monitoring data and various fault simulation results, and gives the most likely fault type and its probability evaluation, realizing intelligent diagnosis from phenomenon to essence.

[0046] The system has been deployed and applied in multiple industrial scenarios, effectively addressing the monitoring challenges in complex electromagnetic environments. In the application of a large rectifier transformer in a certain steel enterprise, the system successfully identified early signs of winding looseness, issued an early warning, and avoided a major equipment accident. In the monitoring of an impact load transformer in a chemical plant, the system accurately diagnosed the insulation degradation problem of the core clamp, guided maintenance personnel to accurately handle it, and improved the reliability of the equipment. Embodiments

[0047] As shown in Figure 4 , specifically: the operation and maintenance of transformers in remote areas face three major technical challenges: communication conditions restrict data transmission, harsh environments affect equipment reliability, and limited operation and maintenance resources cause response delays. These factors collectively require the monitoring system to have a high degree of intelligence and autonomy, enabling it to complete equipment state monitoring and operation and maintenance decisions with limited external support.

[0048] The hardware platform adopts a fully sealed and robust design, with wide temperature range working capability, suitable for extreme environments such as high temperature, high humidity, and high altitude. The power supply system supports multiple energy complementary modes to ensure continuous and stable operation under power grid fluctuations. The core computing unit adopts a low-power architecture to ensure processing performance under limited cooling conditions, meeting the long-term unattended operation requirements.

[0049] Edge intelligence computing is the core innovation of the system. The on-site device is built-in with complete signal processing and feature extraction algorithms, which can independently complete the whole process of data acquisition to health state evaluation. This architecture greatly reduces data transmission requirements, enabling the system to maintain effective monitoring under extremely low bandwidth conditions. The edge intelligence node has self-learning ability and can optimize the evaluation model according to the characteristics of the local equipment to improve the accuracy of state identification.

[0050] The communication module adopts intelligent adaptive strategy. The system adjusts the communication frequency and data content according to the device health state, power grid operating conditions and communication link quality. Under normal conditions, it periodically sends a simplified heartbeat signal, automatically increases the communication priority in abnormal conditions, and starts multi-channel redundant transmission in emergency conditions to ensure that critical information reaches the main station system reliably.

[0051] The state evaluation algorithm is specially optimized for remote areas. The system has strong noise suppression capability and can effectively identify and eliminate false features caused by temporary interference. The adaptive threshold mechanism established through long-term data accumulation enables the system to adapt to the normal state fluctuations of equipment with seasonal changes, avoiding false alarms caused by environmental factors and improving the reliability of the monitoring system.

[0052] The system innovatively introduces a neighbor device cooperative monitoring mechanism. Multiple monitoring devices deployed within a certain area automatically form a network, sharing state information and environmental data. When communication of a certain site is interrupted, neighboring sites will automatically strengthen monitoring coverage of the area, forming a redundant monitoring network to ensure no dead angles in monitoring. This distributed intelligent cooperation greatly improves the applicability of the system in remote areas.

[0053] The operation and maintenance decision support system fully considers the characteristics of remote areas. The operation and maintenance recommendations generated by the system not only include technical analysis, but also take into account traffic conditions, climate factors, resource allocation and other actual situations to provide the most economical and effective disposal scheme. For example, it suggests strengthening the inspection of key equipment before the rainy season, extending the monitoring period in areas with poor transportation conditions, and prioritizing the handling of high-risk equipment when resources are scarce, etc., to optimize the allocation of operation and maintenance resources.

[0054] In the large-scale application of the system in multiple remote areas, the system successfully achieved effective monitoring of numerous transformers. Since the system was put into operation, it has accurately warned of equipment anomalies multiple times, with an average early warning time significantly better than traditional methods. Based on the differential operation and maintenance of the system's accurate state evaluation, the regional transformer failure rate has decreased significantly, and the operation and maintenance cost has been effectively reduced, providing a successful example of intelligent management of power equipment in remote areas and having important promotional value. EMBODIMENT

[0055] As Figure 5As shown, specifically: In modern power grids, transformers often work in groups to jointly undertake regional power supply tasks. The traditional single-device independent optimization mode is difficult to achieve system-wide optimization, and there is an urgent need to establish a new optimization decision system from the perspective of group coordination, fully tap the potential of group coordination, and improve the overall operational efficiency of the power grid.

[0056] The group state perception network builds a unified data collection system to ensure time synchronization and standardization of transformer monitoring data. The system establishes a digital twin model for each transformer, mapping its health status, load characteristics, energy efficiency level, and operational constraints, forming a complete device state panoramic view. Through data fusion technology, dispersed device information is integrated into system-level cognition, laying the foundation for collaborative optimization.

[0057] The health status assessment uses a combination of relative and absolute evaluation methods. The system focuses on both the absolute health level of individual devices and their relative state position in the group. Through clustering analysis and pattern recognition, the system automatically identifies weak links and strong resources in the group, discovers complementary characteristics between devices, and provides multi-dimensional state information support for optimization decisions.

[0058] The collaborative optimization model uses a multi-layer optimization architecture. The bottom layer optimization aims to improve the health and energy efficiency of individual devices, determining the ideal operating range for each transformer. The middle layer optimization coordinates the load distribution of multiple transformers within a substation, achieving optimal operation at the station level. The upper layer optimization considers the overall power grid, coordinating the operation of multiple substations to maximize system-wide efficiency. This hierarchical optimization architecture ensures both local optimization and global coordination.

[0059] The optimization algorithm fully considers actual operational constraints. The system incorporates rich power grid operation rules and safety criteria to ensure that the optimization scheme meets key technical requirements such as safety verification, voltage quality requirements, protection coordination conditions, etc. Through multi-scenario simulation verification, the system can identify potential risks and avoid possible operational conflicts, ensuring the safety and feasibility of the optimization scheme.

[0060] The decision support system provides multi-dimensional visual analysis. Through health status heat maps, load distribution topology maps, and optimization effect comparison maps, the system helps operators intuitively understand system status and optimization recommendations. The system also supports hypothesis analysis, allowing operators to predict the consequences of different decision-making schemes, aiding in making scientific decisions and improving power grid operation and management.

[0061] The operation boundary management is a characteristic function of the system. The system adjusts the operation limits of each transformer according to the health status of the group, the load prediction results, and the weather environment factors. When the system pressure is large, the safety margin is appropriately tightened, and when the conditions allow, the potential of the equipment is fully tapped to achieve the balance between safety and economy. This adaptive operation strategy enables the power grid to maintain the optimal operating state under complex environments.

[0062] The successful application of the system in multiple regional power grids shows that transformer group collaborative optimization can produce significant comprehensive benefits. Through accurate state assessment and scientific load distribution, the average service life of equipment is effectively extended, the system operation energy efficiency is significantly improved, and the operation and maintenance resource allocation is more reasonable. This intelligent management method at the system level provides important technical support and practical experience for the digital transformation of the power grid, and promotes the power grid operation into a new stage of intelligent collaboration.

[0063] The above shows and describes the basic principles of the present application. Main features and advantages. Those skilled in the art should understand that the present application is not limited to the above examples, and the above examples and descriptions in the specification are only to illustrate the principles of the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection claimed by the present application is defined by the appended claims and their equivalents.

Claims

1. A power distribution network transformer operating state monitoring and energy efficiency collaborative optimization method, characterized in that, The method comprises the following steps: Collecting sound signals, vibration acceleration signals and load rate time series data during transformer operation, extracting key features through feature fusion processing and generating a multi-dimensional health state feature vector; Based on the multi-dimensional health state feature vector and the transformer operation history data, a comprehensive feature model including acoustic features, vibration features and their correlation with load rate is constructed, and the load response sensitivity coefficient under different loads is calculated; Based on the comprehensive feature model, standard health state feature templates are generated and stored for different load level intervals, thereby constructing a hierarchical load feature template library; In the online monitoring stage, the corresponding standard health state feature template is retrieved from the feature template library according to the current transformer load rate; the multi-dimensional difference index between the multi-dimensional health state feature vector currently collected and generated and the retrieved corresponding standard health state feature template is calculated; The multi-dimensional difference index is fused and calculated to generate a health degree score reflecting the deviation degree of the transformer operating state, and the change trend of the health degree score is calculated based on the time series of the health degree score; The health degree score is compared with the multi-level diagnosis threshold, and the multi-level diagnosis threshold includes a fault threshold, a warning threshold and a high-efficiency operation threshold; when the health degree score is lower than the fault threshold, an early warning work order is generated and uploaded to the operation and maintenance management system; When the health degree score is between the warning threshold and the high-efficiency operation threshold, energy efficiency adaptive optimization is performed, and based on the health degree score change trend and the load response sensitivity coefficient, the load rate set value of the transformer is determined through a multi-objective optimization algorithm; According to the load rate set value, closed-loop feedback adjustment is performed to realize transformer load adjustment; When system-level optimization is needed, the boundary parameters of the economic operation interval are determined based on the load rate set value, the load distribution strategy is adjusted according to the boundary parameters, and the collaborative adjustment of operation stability and energy efficiency balance is realized.

2. The method of claim 1, wherein, The execution of energy efficiency adaptive optimization, based on the health degree score change trend and the load response sensitivity coefficient, through a multi-objective optimization algorithm to determine the load rate set value of the transformer, comprises: Based on the health degree score change trend and the load response sensitivity coefficient, a multi-objective optimization function is established, which takes improving the health degree score and suppressing the increase of the load response sensitivity coefficient as the optimization objectives; Taking the multi-objective optimization function as the fitness function, the health operation boundary conditions including the upper and lower limits of the transformer load rate and the load rate change rate constraints are constructed; Using a non-dominated sorting genetic algorithm with an elitist strategy, a multi-objective search based on the fitness function is performed in the feasible solution space defined by the health operation boundary conditions to obtain a Pareto front solution set; The relative closeness of each solution on the Pareto front solution set to the ideal solution is calculated, and a selected solution is determined from the Pareto front solution set according to the size order of the relative closeness; The load rate value corresponding to the selected solution is output as the load rate set value.

3. The method of claim 2, wherein, The calculation of the relative closeness of each solution on the Pareto front solution set to the ideal solution, and the determination of a selected solution from the Pareto front solution set according to the size order of the relative closeness, comprises: The relative closeness degrees of the Pareto front solutions are calculated by using an entropy weight-based technique for order preference by similarity to ideal solution (TOPSIS) method; The solutions in the Pareto front solution set are arranged in descending order according to the corresponding relative closeness degree values, and a solution ranked at the first position is selected as the selected solution; The relative closeness degree is determined according to the Euclidean distance between the solution set and an ideal solution and the information entropy distribution characteristics of the solution set.

4. The method of claim 1, wherein, The boundary parameters of the economic operation interval are determined based on the load rate set value, including: A health degree score change rate corresponding to the load rate set value is calculated; An economic operation interval boundary function is constructed based on the change rate and an energy efficiency index; Upper and lower limit values output by the boundary function are used as the boundary parameters of the economic operation interval.

5. The method of claim 4, wherein, The load distribution strategy is adjusted according to the boundary parameters, including: A target load rate of each transformer unit is determined according to the boundary parameters; A load distribution balancing algorithm is executed under the constraint of the target load rate to obtain a load adjustment result, and the transformer operation state is adjusted according to the load adjustment result to realize load redistribution.

6. The method of claim 1, wherein, The standard health state feature templates are generated and stored for different load level intervals, including: For each load level interval, the features output by the comprehensive feature model are reduced in dimension by using a principal component analysis method to extract main feature components; A time sequence evolution constraint is introduced, and a sliding window mechanism is used to analyze the change trend of the features on the time axis to ensure the feature stability information; A standard health state feature template is generated by applying a maintenance mechanism corresponding to the feature stability information, and the maintenance mechanism corresponding to the feature stability information includes feature smoothing and outlier removal; The standard health state feature templates of each load level interval are stored in a hierarchical load feature template library, and an index is established.

7. The method of claim 1, wherein, The multi-dimensional difference index between the multi-dimensional health state feature vector currently collected and generated and the corresponding standard health state feature template is calculated, including: A cosine distance between the current multi-dimensional health state feature vector and the retrieved standard health state feature template is calculated, and the cosine distance is used to represent the spatial distribution difference between the two; A change trend of the current feature vector on the time sequence is extracted and compared with an expected evolution trend of the standard health state feature template to calculate a weighted time sequence difference coefficient, wherein the weight is allocated based on feature importance; The cosine distance and the weighted time sequence difference coefficient are linearly combined as input parameters to generate a multi-dimensional difference index, and the multi-dimensional difference index is used to comprehensively reflect the deviation degree of the current state from the standard health state feature template.

8. The method of claim 1, wherein, The multi-dimensional difference index is fused and calculated to generate a health degree score reflecting the deviation degree of the transformer operation state, including: A feature fluctuation stability is calculated by statistically analyzing the variance and standard deviation of the current feature vector within a sliding window to evaluate the feature fluctuation degree; A feature entropy change rate is calculated based on information entropy theory to analyze the uncertainty change of the feature distribution; A time sequence correlation degree is calculated by using a Pearson correlation coefficient to evaluate the correlation between the current feature sequence and the standard health state feature template sequence. The feature fluctuation stability, feature entropy change rate and time sequence correlation degree are assigned weights, and a health degree score is generated by weighted summation, wherein the weights are adaptively adjusted according to the transformer type and operation environment.

9. The method of claim 1, wherein, The comparison of the health degree score with the multi-level diagnosis threshold value comprises: The preset failure threshold value, early warning threshold value and efficient operation threshold value, wherein the failure threshold value is set based on historical failure data, the early warning threshold value is set based on a device aging model, and the efficient operation threshold value is set based on an energy efficiency optimal value; When the health degree score is lower than the failure threshold value, it is determined that the transformer is in a failure state; When the health degree score is between the failure threshold value and the early warning threshold value, it is determined that the transformer is in an early warning state; When the health degree score is between the early warning threshold value and the efficient operation threshold value, it is determined that the transformer is in a normal operation state; When the health degree score is higher than the efficient operation threshold value, it is determined that the transformer is in an efficient operation state.

10. The method of claim 9, wherein, The generation of the early warning work order and uploading to the operation and maintenance management system comprises: When the health degree score is lower than the failure threshold value, an early warning work order is automatically generated, and the work order comprises a transformer identifier, detection time, health degree score, abnormal feature description and recommended repair measures; The early warning work order is packaged into a standardized data format conforming to an interface protocol of the operation and maintenance management system; The early warning work order is uploaded to the operation and maintenance management system through an encrypted communication protocol, and an alarm notification is triggered.

Images