Predictive maintenance strategy optimization method for transformer area equipment maintenance

Abstract

The invention discloses a predictive maintenance strategy optimization method for transformer area equipment maintenance, relates to the technical field of transformer area equipment maintenance, and is used for solving the problem of poor maintenance effect of transformer equipment. Sensors are installed in a transformer area to collect current, voltage, temperature and other key data, preprocessing, frequency alignment and fusion are carried out to generate the same equipment state data, and when data fluctuation exceeds a threshold value, re-collection and calibration are triggered, a health state model is constructed, and a health index is calculated to evaluate the equipment state. The method comprises the following steps: predicting a future health state through an LSTM model, triggering early warning when the future health state is lower than a baseline, dividing risk levels according to health indexes and risk factors, setting maintenance priorities based on health assessment and fault prediction, applying a multi-objective optimization algorithm in combination with an emergency degree and equipment importance to assess resource requirements, and dynamically monitoring resource use conditions. The prediction precision is improved, and the equipment resource allocation efficiency and the maintenance response speed are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of transformer area equipment maintenance, and more specifically, to a predictive maintenance strategy optimization method for transformer area equipment maintenance. Background Art

[0002] Transformer substation equipment refers to all related equipment installed in a transformer substation in the power system, which is mainly used to reduce high-voltage electricity to low-voltage electricity suitable for residential, commercial and industrial users. Substation is a division or unit of the power supply network, usually serving a specific geographical area (such as a community, village or industrial area), responsible for converting high-voltage transmission into usable low-voltage distribution; With the rapid development of the power system and the acceleration of urbanization, the load faced by transformer substation equipment is increasing, the equipment operating environment is more complex, and traditional maintenance strategies (such as fixed periodic maintenance and post-fault maintenance) have gradually exposed many shortcomings. In this context, predictive maintenance has gradually become a key strategy for transformer substation equipment management. Predictive maintenance relies on real-time monitoring, data collection and intelligent analysis to collect equipment operation data and predict faulty equipment, so as to identify potential fault risks in advance, optimize maintenance timing and maintenance resource allocation, and ensure stable operation of substation equipment.

[0003] Deficiencies in existing technologies: When maintenance needs arise for multiple devices or substations at the same time, there is a lack of effective assessment of the urgency of the equipment's health status, which makes it difficult to reasonably allocate maintenance resources when resources are limited. It is impossible to prioritize the maintenance of high-risk equipment, and low-priority equipment occupies too many resources. Equipment that really needs emergency treatment cannot receive timely maintenance, thereby increasing the risk of equipment failure and maintenance costs. Summary of the invention

[0004] In order to overcome the above defects of the prior art, there is a solution as follows to solve the problem of poor maintenance effect of transformer equipment in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: A predictive maintenance strategy optimization method for transformer area equipment maintenance includes the following steps: Sensors are installed in the transformer area to collect data, and the data is preprocessed and aligned to the same frequency. The aligned data is weighted fused using the Dempster-Shafer evidence theory to generate equipment status data, and data re-collection and calibration are triggered when the fluctuation exceeds the threshold. Build a health status model based on fused data, calculate the health index to evaluate the equipment status, use exponential decay filtering to smooth short-term fluctuations, predict future health status through the LSTM model, trigger an alert when it is lower than the baseline, divide the risk level according to the health index and risk factors, and generate a fault alert when the risk is high; Set maintenance priorities based on health assessment and fault prediction, calculate urgency based on health index and fault risk, determine equipment importance through analytic hierarchy process, apply multi-objective optimization algorithm to evaluate resource requirements, and dynamically monitor resource usage; The particle swarm optimization is used to dynamically optimize the adjustment strategy.

[0006] In a preferred embodiment, sensors are installed in the transformer area to collect data, and the data is preprocessed and aligned to the same frequency. The aligned data is weighted fused by Dempster-Shafer evidence theory to generate equipment status data, and the data is triggered to be re-collected and calibrated when the fluctuation exceeds the threshold. The specific steps are as follows: The data collected by the sensors include temperature, current, voltage, and load parameters; Normalize the collected data to make the data range of different physical quantities consistent; Use the isolation forest algorithm to detect anomalies on each normalized data, remove noise data, and perform spatiotemporal alignment on the collected data to align different collected data to the same frequency; After completing the time-space alignment, the collected data are fused using a dynamic weighted fusion algorithm based on the Dempster-Shafer evidence theory; The fused data is used as the equipment status input of the station area at the current time point, and a feedback mechanism is set. When the difference between the fused value and the value at the previous time point exceeds the set threshold, the feedback mechanism restarts the equipment data collection and calibrates the sensor.

[0007] In a preferred embodiment, a health status model is constructed based on the fused data, a health index is calculated to evaluate the device status, and an exponential decay filter is used to smooth short-term fluctuations. The specific steps include: The fused data is used as the input data for health status assessment. A health status model is established for the fused data. The health index of the equipment is calculated by principal component extraction and feature vector generation. The health status of the equipment is determined based on the health index. Use exponential decay filtering to dynamically smooth the health index of the device to deal with short-term fluctuations; For equipment whose health status is unstable, the deviation degree between the characteristic value of the fused data and the reference value is extracted to determine the risk factor, which is used to analyze the failure risk.

[0008] In a preferred embodiment, the future health status is predicted by the LSTM model, and an early warning is triggered when it is lower than the baseline. The specific steps are as follows: Build a fault risk prediction model based on historical data and apply the LSTM model to predict the health index at future moments; The inputs to the failure risk prediction model are current and historical health indices and risk factors; If the predicted health index continues to be lower than the baseline health status baseline, an early warning signal is triggered, indicating the risk of equipment failure.

[0009] In a preferred embodiment, the risk level is divided according to the health index and the risk factor, and a fault warning is generated when the risk is high, including the following steps: Conduct risk assessment on the predicted health index, set risk levels and generate early warning signals. The risk levels are determined based on the degree of deviation of the health index and the weight of the risk factors. When the risk level is less than or equal to the low risk level threshold, it means that the risk is low, the equipment status is stable, and no immediate maintenance is required; When the risk level is greater than the low risk level threshold and less than or equal to the high risk level threshold, the device is marked as medium risk and an observation plan is formulated; When the risk level is greater than the high risk level threshold, the equipment risk is judged to have reached a high level, and a fault warning signal is generated to remind maintenance personnel to perform inspections and preventive maintenance.

[0010] In a preferred embodiment, maintenance priorities are set based on health assessment and failure prediction, urgency is calculated by combining health index and failure risk, and equipment importance is determined by analytic hierarchy process, including the following steps: Based on the health status assessment and failure risk prediction results, the urgency of the maintenance task for each device is quantified. The urgency index is used to indicate the urgent maintenance needs of the device at the current moment. Determine the relative importance of equipment through the hierarchical analysis method, build an equipment importance judgment matrix, and assign values ​​according to the role of the equipment; The eigenvalue method is used to calculate the eigenvector of the judgment matrix, and the eigenvector is standardized to obtain the weight of each device as the device importance index; After the calculation is completed, the importance of the equipment is determined according to the obtained equipment importance index value; After obtaining the task urgency and equipment importance, the comprehensive priority index is calculated, and the expression formula is as follows: ,in, , and To adjust the parameters, is the urgency index of device i, represents the sum of risk factors of device i, and is a constant parameter used to adjust the impact of risk factors on priority. is the importance index of equipment i; The maintenance tasks of each device are sorted according to the comprehensive priority index to generate a maintenance task list. The tasks are sorted according to the principle of high to low priority, and the first task in the list is maintained first.

[0011] In a preferred embodiment, a multi-objective optimization algorithm is applied to evaluate resource requirements and dynamically monitor resource usage, including the following steps: Evaluate the human resources required based on the priority and complexity of each maintenance task; Determine the list of tools and equipment required for each task based on the type of maintenance task and the characteristics of the equipment required, and calculate the tool requirements necessary to meet all task requirements in combination with tool quantity limitations and equipment availability; For each maintenance task, analyze material requirements, including parts and materials, and calculate material requirements; Based on the task priority index and human resource requirements, tool requirements, and material requirements, the particle swarm optimization algorithm is applied to dynamically allocate resources and determine the resource allocation cost.

[0012] In a preferred embodiment, the particle swarm optimization is used to dynamically optimize the adjustment strategy, comprising the following steps: Construct a multi-objective optimization model based on resource requirements and task priorities. The objective function of the model includes resource allocation cost, task completion time and resource utilization rate. The allocated resources include manpower, tools and materials. The particle swarm optimization algorithm is selected to optimize the resource allocation scheme.

[0013] Technical effects and advantages of a predictive maintenance strategy optimization method for transformer area equipment maintenance of the present invention: The present invention collects key data such as current, voltage, and temperature by installing sensors in the transformer area, performs preprocessing and frequency alignment, and uses Dempster-Shafer evidence theory to generate the same device status data. When the data fluctuation exceeds the threshold, re-collection and calibration are triggered to ensure the accuracy of the data; The health status model is constructed by integrating data, and the health index is calculated to evaluate the equipment status. The exponential decay filter is used to smooth short-term fluctuations. The future health status is predicted through the LSTM model, and an early warning is triggered when it is lower than the baseline. According to the health index and risk factor, the risk is divided into "low, medium, and high" levels. High risk generates fault warnings. Based on health assessment and fault prediction, maintenance priorities are set. In combination with the urgency and importance of the equipment, a multi-objective optimization algorithm is applied to evaluate resource requirements, and resource usage is dynamically monitored. Finally, particle swarm optimization is used to achieve intelligent adjustment of strategies, give priority to the maintenance needs of high-risk equipment, increase prediction accuracy, improve equipment resource allocation efficiency and maintenance response speed, and enhance the stability and reliability of equipment operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention is a flowchart of a method for optimizing a predictive maintenance strategy for transformer area equipment maintenance. DETAILED DESCRIPTION

[0015] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0016] In order to achieve the above objectives, Figure 1 A structural schematic diagram of a predictive maintenance strategy optimization method for transformer area equipment maintenance according to the present invention is given, which specifically includes the following steps: Sensors are installed in the transformer area to collect data, and the data is preprocessed and aligned to the same frequency. The aligned data is weighted fused using the Dempster-Shafer evidence theory to generate equipment status data, and data re-collection and calibration are triggered when the fluctuation exceeds the threshold. Build a health status model based on fused data, calculate the health index to evaluate the equipment status, use exponential decay filtering to smooth short-term fluctuations, predict future health status through the LSTM model, trigger an alert when it is lower than the baseline, divide the risk level according to the health index and risk factors, and generate a fault alert when the risk is high; Set maintenance priorities based on health assessment and fault prediction, calculate urgency based on health index and fault risk, determine equipment importance through analytic hierarchy process, apply multi-objective optimization algorithm to evaluate resource requirements, and dynamically monitor resource usage; The particle swarm optimization is used to dynamically optimize the adjustment strategy.

[0017] Step 1: deploy smart sensors in the transformer area to collect key equipment parameters including temperature, current, voltage, load, etc. in real time, and transmit the data to the central data platform through the Internet of Things. The specific steps are as follows: Install a variety of sensors in the transformer area to collect key operating parameters, including current, voltage, temperature, humidity, load and other data. Assume that the number of sensors is N, and each sensor (where i=1, 2, ..., N) according to different frequencies Collect data and generate real-time data streams , where t is the time point; Data Flow The initialization includes basic data formatting and preprocessing to meet the needs of subsequent data fusion. The initialization step includes normalization processing to make the data range of different physical quantities consistent. The formula is: ,in, and Respectively represent sensors The minimum and maximum values ​​recorded in the historical data to ensure that the relative proportions of each data are consistent; For each normalized data Perform anomaly detection, remove noise data, use the Isolation Forest algorithm to build an anomaly detection model, and define anomaly factors To represent the abnormality of each data point, the calculation of the abnormal factor is as follows: ,in, For sensor Normal working value benchmark, represents the data volatility parameter, is a minimum value to prevent the denominator from being zero. If (Threshold setting), then the data Treat it as abnormal data and remove it, otherwise proceed to the next step; It should be noted that the degree of deviation between the data and the benchmark value is used to determine whether it is abnormal, avoiding the mean and simple statistical methods; parameters It is obtained through historical fluctuations and is used to characterize the normal data fluctuation range of the sensor. It is used to ensure computing stability; Multi-source data collection frequency Therefore, it is necessary to perform spatiotemporal alignment so that the data can be Fusion is performed at the kth time point, assuming that the target frequency of target alignment is , use interpolation to calculate the uncollected data points , the formula is: ,in, , and The aligned data obtained by interpolation is two adjacent sampling moments Go to the next step; The interpolation formula of time-space alignment can fill in the missing data according to the numerical change trend between adjacent sampling points, ensuring that the data of each sensor is aligned at the same time point, which is helpful for subsequent data fusion. The target frequency is set according to the system's demand for data update rate; After completing the time-space alignment, the dynamic weighted fusion algorithm based on the Dempster-Shafer evidence theory is used to fuse the multi-source data. Assume that the fused data is represented as , the specific fusion formula is as follows: , where the weight Dynamic calculation, taking into account the credibility of each sensor and real-time deviation factor , the specific formula is: ,in Indicates the previous time point The fused data value of is the smoothing coefficient, which is used to control the influence of deviation. The final fused data Ability to balance data contributions from different sensors and achieve accurate data integration; The fused data is used as the equipment status input of the current time point and passed to the next module for health status assessment. A feedback mechanism is set up. When the difference between the fused value and the value at the previous time point exceeds the set threshold, the equipment data collection is restarted and the sensor is calibrated.

[0018] Step 2: Perform health status assessment and fault risk prediction. Use the fused real-time data to quantitatively assess the health status of the transformer equipment and predict future fault risks based on the assessment results. The specific steps are as follows: Fusion Data As the input data for health status assessment, a health status model based on multidimensional features is first established for the fused data, and the health index of the equipment is calculated by principal component extraction and feature vector generation. , assuming that the fused data contains M key features, the health index calculation formula is: ,in, is the weight of the jth feature, Indicates that the jth feature is The value of the moment, It is the health status benchmark obtained through feature vector analysis, which represents the degree of influence of the feature on the health of the device; The health index reflects the overall health status of the device. The weight is determined based on the importance of the features, so that important features contribute more to the health index. The health status benchmark is obtained based on the principal component analysis of historical data to ensure that the relative influence of each feature is consistent with the actual operating status of the device. In equipment analysis, the health index of the equipment has short-term fluctuations. Exponential Decay Filtering (EDF) is used to dynamically smooth the health index. The formula of the smoothed health index is as follows: ,in, is the smoothing coefficient, which takes a value in the range of [0, 1] to control the impact of short-term fluctuations. Baseline health status For comparison, if , then the health status of the equipment is determined to be unstable and its failure risk needs to be further analyzed; Exponential decay filtering eliminates short-term fluctuations, ensuring that the health index can accurately reflect the long-term health change trend of the device. The health status baseline represents the health level of the device under normal conditions, which can be obtained through the analysis of historical health data. The smoothing coefficient controls the filtering intensity of short-term fluctuations, making the evaluation more robust. For devices whose health status is unstable, further analyze their possible failure risks. By extracting the risk factor of the fused data, analyze the contribution of each feature to the failure risk. The risk factor formula is: ,in, is the reference value of the ith feature (standard value under normal operating conditions), represents the risk sensitivity weight of the ith feature. By calculating each risk factor, the feature with the greatest impact on the failure risk can be identified; The calculation of the risk factor is based on the deviation between the current feature value and the reference value. The risk sensitivity weight is used to adjust the impact of the feature on the failure risk, ensuring that highly sensitive features can play a greater role in risk calculation. The goal of this step is to find out the potential causes of the failure and facilitate targeted maintenance. After completing the risk factor extraction, a fault risk prediction model is built based on historical data, and the long short-term memory network (LSTM) is used to predict the health index at future moments. , where n is the prediction time step and the model input is the current and historical health index and risk factors, the prediction formula of the LSTM model is: ; like Persistently below baseline , then the relevant system triggers an early warning signal to indicate potential failure risks; The long short-term memory network (LSTM) model is good at capturing long-term dependencies in time series data and is suitable for predicting future trends in the health index. The model input includes smoothed historical health indexes and risk factors to ensure that the prediction results can comprehensively reflect the health status and risk status of the current equipment. If the predicted health index is lower than the baseline for a long time, the possibility of equipment failure is high and maintenance should be arranged in time; Based on the failure risk prediction, the predicted health index Conduct risk assessment and set risk levels And generate early warning signals. The risk level is divided according to the degree of deviation of the health index and the weight of the risk factor. The specific formula is: ,in, represents the weight coefficient of the risk factor, The control level is divided into three levels: "low, medium, and high". When the risk level reaches "high", the system generates a fault warning signal; According to the calculated risk level value, it is divided into three levels; Low risk: When When the risk is low, the equipment is in a stable state and no immediate maintenance is required; Medium risk: When the system marks the device as medium risk, an observation plan can be made according to the situation; High risk: When When the equipment risk reaches a high level, a fault warning signal is generated to remind maintenance personnel to conduct inspections and preventive maintenance. Risk level threshold For low risk level threshold, The high risk level threshold is determined based on historical equipment data and expert experience, reflecting the risk level of the equipment in different health states; When the risk level Exceeding the high risk threshold When a fault occurs, a fault warning signal is generated, which automatically sends an alarm message to the maintenance management system and records the health status, risk factors and other relevant data of the equipment for maintenance personnel to conduct detailed analysis and preventive maintenance decisions.

[0019] Step 3: Maintenance priority setting and task sorting. Based on the equipment health status assessment and fault risk prediction, set the priority of the maintenance task. According to the health index and the predicted probability of fault occurrence, calculate the priority score through the multi-factor weighted method, and sort the tasks based on this score. Further optimize the task sorting, and apply the priority scheduling algorithm (such as the analytic hierarchy process AHP or multi-criteria decision-making based on TOPSIS) to subdivide the tasks to ensure that high-risk and critical equipment receive the highest priority. Task sorting should not only consider the urgency of the equipment status, but also analyze the impact of the equipment on the power supply system. The specific steps are as follows: Quantify the urgency of maintenance tasks for each device based on health status assessment and failure risk prediction results, urgency index It is used to express the urgent maintenance demand of equipment i at the current moment. The formula is as follows: ,in, Indicates the degree of deviation of the health status of device i, is the failure risk level of device i, and To adjust the weights, used to control the impact of health status and risk level, and obtain the urgency index After that, it is used for priority calculation; It should be noted that the urgency index combines the health status deviation and the fault risk level to ensure that high-risk and poorly healthy devices receive higher priority and adjust the weight. and It can be adjusted according to system requirements, so that the emergency maintenance needs of different equipment can be effectively differentiated; In order to reasonably allocate maintenance resources, it is also necessary to consider the importance of the equipment in the area and define the equipment importance index , the relative importance of equipment is determined by the Analytic Hierarchy Process (AHP). The main steps of applying the AHP are as follows: Construct an equipment importance judgment matrix and assign values ​​based on the role of the equipment in the system (such as load support capacity, power supply area importance, etc.); The eigenvalue method is used to calculate the eigenvector of the judgment matrix, and the eigenvector is standardized to obtain the weight of each device as the device importance index; After the calculation is completed, the equipment with a higher equipment importance index value is the key equipment, and its priority will be increased in the subsequent steps; After obtaining the task urgency and equipment importance, the comprehensive priority index is calculated, and the expression formula is as follows: ,in, , and To adjust the parameters, represents the sum of the risk factors of the equipment, and is a constant parameter used to adjust the impact of risk factors on priority, the comprehensive priority index The larger the value, the higher the maintenance priority of the equipment; is the importance index of equipment i; According to the comprehensive priority index, the maintenance tasks of each device are sorted and a maintenance task list is generated. Task sorting follows the principle of high to low priority, and the first task in the list is maintained first. Set thresholds Used to determine whether to perform maintenance tasks immediately. When the device task is marked as "urgent task", other tasks are marked as "normal tasks"; During the task execution, as the health status of the equipment changes in real time, the comprehensive priority index The priority index of the device may change, so a dynamic adjustment mechanism is set up to recalculate the priority index of the device at each time interval and update the task list. When the new priority ranking appears at the top of the task list and its priority is significantly higher than the current task, the task execution order is dynamically adjusted.

[0020] Step 4: Conduct resource demand assessment and dynamic resource matching. According to the priority and specific content of each maintenance task, evaluate the required human resources, tools and materials to ensure accurate resource allocation. Use linear programming algorithms to optimize the allocation of human and material requirements to ensure that high-priority tasks are allocated to high-quality resources. At the same time, adapt to real-time changes in maintenance needs and use greedy algorithms to match maintenance tasks with available resources in real time. In an emergency, personnel and equipment can be quickly adjusted to cope with sudden tasks. Through the dynamic resource matching mechanism, maintenance tasks can be carried out in an orderly manner under resource constraints to reduce the risk of resource waste. The specific steps are as follows: Based on the priority and complexity of each maintenance task, evaluate the required human resources, match the appropriate skill types, and set the number of technical workers required for task i. and the number of people required for each type of work ,in , then the total manpower requirement formula for task i is: ,in, Indicates the skill requirement satisfaction rate of job type j (value ranges from [0, 1]), which is used to ensure that the skills of the assigned personnel meet the task requirements; In the human resource demand assessment, skill matching ensures that each task is assigned to the right skilled personnel. Indicates the degree of match between current personnel skills and task requirements, reduces the task failure rate, and calculates , obtain the total manpower requirements for each task and guide subsequent resource allocation; Determine the list of tools and equipment required for each task based on the type of maintenance task and the characteristics of the equipment required (where k represents the equipment type number), combined with the tool quantity limit and equipment availability, calculate the total number of tools required to meet all task requirements: ,in, represents the number of equipment types required for task i, is the availability coefficient of tool equipment k, which is used to indicate whether the equipment is available (value is 0 or 1); Quantification of tool requirements ensures the availability of equipment in task implementation, availability factor Check the current status of your tools and reduce the risk of delays due to resource shortages. Calculate results Indicates the number of tools required for a task and forms the basis for tool resource allocation; For each maintenance task, analyze the material requirements, including parts, materials, etc., to ensure the smooth completion of the task. The material requirements calculation formula is as follows: ,in, is the number of materials required for task i, represents the demand quantity of the lth material, is the inventory availability factor (value 0 or 1), if , indicating that the inventory is insufficient and materials need to be replenished. The material demand assessment determines the availability of the current inventory by calculating the number of parts and materials required for the task. The inventory availability coefficient is used for inventory inspection to ensure that the materials are fully prepared and trigger the replenishment plan when the inventory is insufficient to avoid task delays due to material shortages; Task-based priority index and resource requirements , , , a multi-objective optimization algorithm (such as particle swarm optimization algorithm or genetic algorithm) is applied to perform dynamic resource allocation. The objective function of the resource allocation model is: ,in, , , The weight factors for manpower, tools, and materials are used to balance the allocation priorities of different resources by adjusting the weights to ensure that resources are tilted toward high-priority tasks; The goal of resource allocation optimization is to minimize the resource allocation cost of high-priority tasks through the resource allocation model, improve the efficiency of resource allocation through multi-objective optimization algorithms, and set relevant weight factors so that resources tend to be biased towards urgent and important tasks, ensuring that limited resources are reasonably allocated to the most needed tasks; During the resource allocation and task execution process, the resource usage of each task is monitored in real time, and a feedback mechanism is set. When the resource usage exceeds the threshold of the predicted value, the resource adjustment mechanism is triggered to reallocate manpower, tools or material resources to ensure the smooth progress of the maintenance task. After the resource requirements are completed, the remaining manpower, tools and materials in the resource pool are regularly checked. For resources below the safety stock, an automatic replenishment plan is generated, and scarce materials are purchased first to ensure sufficient resources for subsequent tasks.

[0021] Step 5: Optimize resource allocation based on intelligent scheduling. To improve resource utilization efficiency, genetic algorithm or particle swarm optimization algorithm is used to take the priority of each maintenance task, personnel skill requirements and task time as the input of the scheduling model, and dynamically generate resource allocation scheme. The scheduling model can switch between global optimality and local optimality to achieve centralized use of resources in high-priority tasks. In the resource allocation process, the equipment status data is dynamically updated, and the scheduling strategy is adjusted through real-time feedback to ensure that high-risk equipment is maintained first. At the same time, the optimization algorithm improves allocation efficiency through continuous iteration, so that resource allocation is closer to maintenance needs. The specific steps are as follows: Construct a multi-objective optimization model based on resource requirements and task priorities. The objective function of the model contains three main parts: resource allocation cost, task completion time, and resource utilization rate. Suppose the priority index of task i is , the allocated resources include manpower, tools and materials, and the optimization objective function is defined as follows: ,in, is the time factor weight, Indicates the total time to complete the task. is the weight of resource utilization, Indicates the effective utilization rate of resources; To solve the multi-objective optimization problem, the particle swarm optimization algorithm (PSO) is selected. First, the particle swarm is randomly initialized, where each particle represents a resource allocation scheme. and speed In turn, they represent the resource allocation status and optimization direction of the scheme. The initialization of the particle swarm includes: Randomly generate particle positions It represents the initial resource allocation, that is, the initial matching of task priority and resource demand. The steps of position initialization are as follows: for each task's resource demand (manpower, tools, materials, etc.), an initial allocation plan that meets the constraints is randomly generated. The allocation plan must meet basic constraints, such as manpower cannot exceed the available number of people, the number of tools cannot exceed the inventory limit, etc. In actual resource allocation scenarios, maintenance tasks in transformer substations may have special resource requirements. For example, high-priority tasks may require more professional technicians and key equipment. Therefore, the generation of initial positions needs to ensure that these key tasks can obtain appropriate resources. Set particle speed is set to zero so that the particle swarm algorithm explores the solution space at a slower speed in the initial iteration. The particle speed determines the moving direction and stride of the particles in the solution space, which will be updated in subsequent iterations. In actual scheduling, the speed vector can control the range of changes in the resource allocation plan. For example, when resources are scarce, a smaller initial speed can be set to ensure that resource allocation adjustments are more precise; when resources are more abundant, a larger initial speed can be set to speed up the search progress; Set the number of iterations and control parameters such as inertia weight; Each particle is based on its current position Calculate the fitness function , the fitness function is to optimize the value of the target Z and record the optimal position of each particle so far and the global optimal position , when the fitness function of the particle's current position is better than the local or global optimum, update and ; Based on the current optimal position, update the speed and position of each particle to optimize the resource allocation plan. The speed update formula is: ; The position update formula is: ,in, is the inertia weight, which controls the exploration ability of particles; and is the acceleration factor, and is a random number between [0, 1], used to enhance the randomness of the search; In the resource allocation of transformer substations, random factors and It can help particle swarms make more attempts in resource allocation schemes. Especially when resource allocation is close to the global optimum, random factors enable particles to continue to explore other possible optimization schemes and avoid falling into suboptimal solutions. After each iteration, the particle speed and position are updated, and the fitness value of the current solution is checked. When the fitness change is less than the convergence threshold or the number of iterations reaches the maximum, the algorithm stops iterating and obtains the final resource allocation plan. The optimal resource allocation plan is recorded as , represents the resource scheduling scheme that satisfies task priorities and optimal resource configuration; Fitness calculation is the basic step of optimization. The rationality of the resource allocation scheme is evaluated by calculating the fitness value of each particle scheme. For resource scheduling in transformer substations, the value of the fitness function can reflect whether the current resource allocation meets the needs of high-priority tasks and provide a reference for subsequent iterations; The final resource allocation plan It is provided to the maintenance management system to execute resource scheduling, and a feedback mechanism is set up to monitor resource usage in real time. If the deviation between actual resource usage and the allocation plan exceeds the set threshold, the re-optimization process is triggered.

[0022] Resource allocation optimization of intelligent scheduling can accurately allocate limited resources (manpower, tools, and materials) to maintenance tasks with higher priority, avoiding excessive use of resources on low-priority tasks. Allocation optimization not only reduces idleness and waste of resources, but also ensures that the maintenance needs of high-risk and high-priority equipment are met first. Through more reasonable resource allocation, the overall response speed of predictive maintenance is accelerated, and the operating efficiency of equipment in the substation is improved. Especially in the case of resource shortage or intensive maintenance tasks, intelligent scheduling optimization ensures the optimal use of each key resource, reduces unnecessary resource scheduling, enhances the overall maintenance support capability of the substation, and makes predictive maintenance more efficient and accurate. Through intelligent scheduling algorithms, high-priority tasks can be quickly identified and more resources can be allocated to them, which significantly improves the response speed to potential faults. Especially in the case of multiple tasks in parallel, the system can dynamically adjust resource allocation according to task priority, so that high-priority tasks can be adequately supported in a short time, reducing the risk of key equipment failure. The improvement of priority response capabilities ensures that high-risk equipment can receive timely intervention when abnormal health status occurs, effectively preventing the further spread or deterioration of faults. This optimization strategy plays an important role in improving system stability and power supply reliability, making predictive maintenance more adaptable and agile when facing sudden problems. Intelligent scheduling resource allocation optimization brings higher resource utilization efficiency, task response speed, cost control effect and maintenance accuracy to the predictive maintenance strategy of transformer substation equipment maintenance, making it more intelligent and efficient when facing multiple task requirements and resource constraints. Through the optimized intelligent maintenance strategy, substation equipment management is not only more stable at present, but also has good adaptability and scalability, ensuring that it remains efficient and reliable in future development, providing a solid guarantee for the safe and stable operation of the power system.

[0023] It should be noted that the threshold information in this embodiment is pre-set by professionals and will not be explained in detail here. Some parameter English letters in the embodiments have the same situation, but different meanings are explained when used, which will not be explained one by one here.

[0024] The present invention collects key data such as current, voltage, and temperature by installing sensors in the transformer area, performs preprocessing and frequency alignment, and uses the Dempster-Shafer evidence theory to generate the same device status data. When the data fluctuation exceeds the threshold, re-collection and calibration are triggered to ensure the accuracy of the data. The health status model is constructed by integrating data, and the health index is calculated to evaluate the equipment status. The exponential decay filter is used to smooth short-term fluctuations. The future health status is predicted through the LSTM model. An early warning is triggered when it is lower than the baseline. According to the health index and risk factor, the risk is divided into "low, medium, and high" levels. A high-risk fault warning is generated. Based on health assessment and fault prediction, maintenance priorities are set. In combination with the urgency and importance of the equipment, a multi-objective optimization algorithm is applied to evaluate resource requirements, and resource usage is dynamically monitored. Finally, particle swarm optimization is used to achieve intelligent adjustment of strategies, give priority to the maintenance needs of high-risk equipment, increase prediction accuracy, improve equipment resource allocation efficiency and maintenance response speed, and enhance the stability and reliability of equipment operation.

[0025] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0026] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0027] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0028] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0029] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0030] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A predictive maintenance strategy optimization method for transformer area equipment maintenance, characterized by: The steps include: Sensors are installed in the transformer area to collect data, and the data is preprocessed and aligned to the same frequency. The aligned data is weighted fused using the Dempster-Shafer evidence theory to generate equipment status data, and data re-collection and calibration are triggered when the fluctuation exceeds the threshold. Build a health status model based on fused data, calculate the health index to evaluate the equipment status, use exponential decay filtering to smooth short-term fluctuations, predict future health status through the LSTM model, trigger an alert when it is lower than the baseline, divide the risk level according to the health index and risk factors, and generate a fault alert when the risk is high; Set maintenance priorities based on health assessment and fault prediction, calculate urgency based on health index and fault risk, determine equipment importance through analytic hierarchy process, apply multi-objective optimization algorithm to evaluate resource requirements, and dynamically monitor resource usage; The particle swarm optimization is used to dynamically optimize the adjustment strategy.

2. The predictive maintenance strategy optimization method for transformer area equipment maintenance according to claim 1 is characterized by: Sensors are installed in the transformer area to collect data, and the data is preprocessed and aligned to the same frequency. The aligned data is weighted fused using the Dempster-Shafer evidence theory to generate equipment status data, and data re-collection and calibration are triggered when the fluctuation exceeds the threshold. The specific steps are as follows: The data collected by the sensors include temperature, current, voltage, and load parameters; Normalize the collected data to make the data range of different physical quantities consistent; Use the isolation forest algorithm to detect anomalies on each normalized data, remove noise data, and perform spatiotemporal alignment on the collected data to align different collected data to the same frequency; After completing the time-space alignment, the collected data are fused using a dynamic weighted fusion algorithm based on the Dempster-Shafer evidence theory; The fused data is used as the equipment status input of the station area at the current time point, and a feedback mechanism is set. When the difference between the fused value and the value at the previous time point exceeds the set threshold, the feedback mechanism restarts the equipment data collection and calibrates the sensor.

3. The predictive maintenance strategy optimization method for transformer area equipment maintenance according to claim 2 is characterized by: The health status model is constructed based on the fused data, the health index is calculated to evaluate the device status, and the exponential decay filter is used to smooth short-term fluctuations. The specific steps include: The fused data is used as the input data for health status assessment. A health status model is established for the fused data. The health index of the equipment is calculated by principal component extraction and feature vector generation. The health status of the equipment is determined based on the health index. Use exponential decay filtering to dynamically smooth the health index of the device to deal with short-term fluctuations; For equipment whose health status is unstable, the deviation degree between the characteristic value of the fused data and the reference value is extracted to determine the risk factor, which is used to analyze the failure risk.

4. The predictive maintenance strategy optimization method for transformer area equipment maintenance according to claim 3 is characterized by: The LSTM model is used to predict future health status and trigger an alert when it is lower than the baseline. The specific steps are as follows: Build a fault risk prediction model based on historical data and apply the LSTM model to predict the health index at future moments; The inputs to the failure risk prediction model are current and historical health indices and risk factors; If the predicted health index continues to be lower than the baseline health status baseline, an early warning signal is triggered, indicating the risk of equipment failure.

5. The predictive maintenance strategy optimization method for transformer area equipment maintenance according to claim 4 is characterized by: The risk level is divided according to the health index and risk factors, and a fault warning is generated when the risk is high, including the following steps: Conduct risk assessment on the predicted health index, set risk levels and generate early warning signals. The risk levels are determined based on the degree of deviation of the health index and the weight of the risk factors. When the risk level is less than or equal to the low risk level threshold, it means that the risk is low, the equipment status is stable, and no immediate maintenance is required; When the risk level is greater than the low risk level threshold and less than or equal to the high risk level threshold, the device is marked as medium risk and an observation plan is formulated; When the risk level is greater than the high risk level threshold, the equipment risk is judged to have reached a high level, and a fault warning signal is generated to remind maintenance personnel to perform inspections and preventive maintenance.

6. The predictive maintenance strategy optimization method for transformer area equipment maintenance according to claim 5 is characterized by: Maintenance priorities are set based on health assessment and failure prediction, the urgency is calculated by combining the health index and failure risk, and the importance of equipment is determined through the hierarchical analysis method, including the following steps: Based on the health status assessment and failure risk prediction results, the urgency of the maintenance task for each device is quantified. The urgency index is used to indicate the urgent maintenance needs of the device at the current moment. Determine the relative importance of equipment through the hierarchical analysis method, build an equipment importance judgment matrix, and assign values ​​according to the role of the equipment; The eigenvalue method is used to calculate the eigenvector of the judgment matrix, and the eigenvector is standardized to obtain the weight of each device as the device importance index; After the calculation is completed, the importance of the equipment is determined according to the obtained equipment importance index value; After obtaining the task urgency and equipment importance, the comprehensive priority index is calculated, and the expression formula is: ,in, , and To adjust the parameters, is the urgency index of device i, represents the sum of risk factors of device i, and is a constant parameter used to adjust the impact of risk factors on priority. is the importance index of equipment i; The maintenance tasks of each device are sorted according to the comprehensive priority index to generate a maintenance task list. The tasks are sorted according to the principle of high to low priority, and the first task in the list is maintained first.

7. The predictive maintenance strategy optimization method for transformer area equipment maintenance according to claim 6 is characterized by: Apply multi-objective optimization algorithms to evaluate resource requirements and dynamically monitor resource usage, including the following steps: Evaluate the human resources required based on the priority and complexity of each maintenance task; Determine the list of tools and equipment required for each task based on the type of maintenance task and the characteristics of the equipment required, and calculate the tool requirements necessary to meet all task requirements in combination with tool quantity limitations and equipment availability; For each maintenance task, analyze material requirements, including parts and materials, and calculate material requirements; Based on the task priority index and human resource requirements, tool requirements, and material requirements, the particle swarm optimization algorithm is applied to dynamically allocate resources and determine the resource allocation cost.

8. The predictive maintenance strategy optimization method for transformer area equipment maintenance according to claim 7 is characterized by: The particle swarm optimization is used to dynamically optimize the adjustment strategy, including the following steps: Construct a multi-objective optimization model based on resource requirements and task priorities. The objective function of the model includes resource allocation cost, task completion time and resource utilization rate. The allocated resources include manpower, tools and materials. The particle swarm optimization algorithm is selected to optimize the resource allocation plan.

Images