Digital-twin-driven substation equipment whole life cycle intelligent management method

Abstract

The application provides a digital twin driven substation equipment whole life cycle intelligent management method, relates to the substation management technical field, and comprises the following steps: feeding the output parameter of the physical layer model to the data layer model as input to generate an equipment health state prediction value; feeding the equipment health state prediction value output by the data layer model to the economic layer model as input to generate a current period maintenance cost estimation value; if the current period maintenance cost estimation value exceeds a cost threshold value, a first control signal is generated, the prediction step length for generating the equipment health state prediction value in the data layer model is adjusted, so that the prediction step length is shortened with the increase of the maintenance cost; the adjusted prediction step length is fed back to the data layer model, and the adjusted equipment health state prediction value is regenerated; if the regenerated equipment health state prediction value exceeds a health threshold value, a second control signal is generated; and a maintenance instruction is generated and output according to the second control signal.

Description

Technical Field

[0001] This invention belongs to the field of substation management technology, specifically relating to a digital twin-driven intelligent management method for the entire lifecycle of substation equipment. Background Technology

[0002] Life cycle cost (LCC) management of substation equipment is a key means for the power industry to achieve lean asset operation. Existing LCC management methods usually adopt static models, which perform ex-post accounting based on historical data of equipment and fixed cost parameters. There is a time lag between the cost assessment results and the real-time operating status of the equipment, and it cannot reflect the impact of changes in the current health status of the equipment on maintenance costs.

[0003] When equipment operating conditions change (such as accelerated performance degradation), existing cost assessments cannot reflect these changes in a timely manner. This leads to maintenance decisions still being based on lagging cost data, creating a disconnect between condition awareness and cost optimization. This disconnect causes maintenance strategy adjustments to lag behind changes in equipment condition, resulting in either cost waste due to over-maintenance or increased equipment failure losses due to untimely maintenance, making it difficult to achieve optimized control of equipment lifecycle costs. Summary of the Invention

[0004] In view of the above-mentioned defects or deficiencies in the existing technology, a digital twin-driven intelligent management method for the entire life cycle of substation equipment is provided, including the following steps: Establish a three-layer digital twin model for the entire life cycle of equipment. The three-layer digital twin model includes a physical layer model, a data layer model, and an economic layer model. The output parameters of the physical layer model are fed into the data layer model as input, and the data layer model generates a predicted value of the device health status. The predicted health status of the equipment output by the data layer model is fed into the economic layer model as input, and the economic layer model generates the estimated cost of the cost cycle maintenance. The estimated maintenance cost for the current cycle output by the economic layer model is compared with a cost threshold. If the estimated maintenance cost for the current cycle exceeds the cost threshold, a first control signal is generated. According to the first control signal, the prediction step size used to generate the predicted value of the device health status in the data layer model is adjusted so that the prediction step size is shortened as the maintenance cost increases; The adjusted prediction step size is fed back to the data layer model to regenerate the adjusted device health status prediction value. The adjusted predicted device health status value is compared with a preset health threshold. If the regenerated predicted device health status value exceeds the health threshold, a second control signal is generated. Based on the second control signal, a maintenance command is generated and output.

[0005] According to the technical solution provided in this application, if the estimated maintenance cost for this period exceeds the cost threshold, the method further includes the following steps: When the estimated maintenance cost for this period exceeds the cost threshold, the cost contribution values ​​of each cost item in the economic layer model are sorted. The cost items include equipment aging cost, fault repair cost, planned maintenance cost, and operating loss cost. Identify cost items that exceed a preset contribution threshold in the cost contribution value, or identify at least one cost item that ranks at the top in the cost contribution value ranking, and obtain the equipment component identifier corresponding to each identified cost item; Based on the device component identifier, locate the health status prediction submodule corresponding to the device component identifier in the data layer model; Adjusting the prediction step size in the data layer model used to generate the predicted value of the device health status includes the following steps: The prediction step size is adjusted only for the identified health status prediction submodule, while the prediction step size of other health status prediction submodules in the data layer model remains unchanged.

[0006] According to the technical solution provided in this application, adjusting the prediction step size of only the located health status prediction submodule includes the following steps: Obtain the cost items corresponding to each health status prediction sub-module located, and extract the cost contribution value of each cost item in the estimated maintenance cost of the current period; Calculate the ratio of each cost contribution value to the cost threshold to obtain the excess contribution of each cost item; Based on the contribution of each cost item to exceeding the standard, the prediction step size adjustment coefficient of each health status prediction sub-module is determined, and the contribution of exceeding the standard is positively correlated with the prediction step size adjustment coefficient. Multiply the current prediction step size of each health status prediction submodule by the corresponding prediction step size adjustment coefficient to obtain the adjusted prediction step size of each health status prediction submodule. Among them, the greater the contribution of the cost item exceeding the standard, the greater the reduction in the prediction step size of the corresponding health status prediction submodule.

[0007] According to the technical solution provided in this application, after feeding back the adjusted prediction step size to the data layer model and regenerating the adjusted device health status prediction value, the method further includes the following steps: Obtain the adjusted prediction step size of each health status prediction submodule, calculate the ratio of each adjusted prediction step size to the original prediction step size, and obtain the step size compression ratio of each health status prediction submodule. Obtain the health status prediction value sequence of each health status prediction submodule within multiple consecutive historical prediction periods, and calculate the fluctuation amplitude of the health status prediction value sequence. The step size compression rate and fluctuation amplitude of each health status prediction submodule are compared. If the step size compression rate is greater than the first preset threshold and the fluctuation amplitude is less than the second preset threshold, the change in the current prediction value of the health status prediction submodule is determined to be a pseudo fluctuation caused by the shortening of the step size, and a pseudo fluctuation identifier is generated. If the step size compression rate is greater than the first preset threshold and the fluctuation amplitude is greater than or equal to the second preset threshold, then the change in the current predicted value of the health status prediction submodule is determined to be a real fluctuation in the device's status, and a real fluctuation identifier is generated.

[0008] According to the technical solution provided in this application, after generating the pseudo-fluctuation identifier or the real fluctuation identifier, the following steps are also included: When the determination result is a pseudo fluctuation identifier, the preset health threshold corresponding to the health status prediction submodule is obtained, the threshold adjustment range is calculated according to the step size compression rate, and the preset health threshold is adjusted by the threshold adjustment range to obtain the target health threshold. When the judgment result is a true fluctuation indicator, the preset health threshold corresponding to the health status prediction submodule is kept unchanged as the target health threshold. Obtain the sequence of health status prediction values ​​of the health status prediction submodule in multiple consecutive historical prediction periods, calculate the slope of the trend of the health status prediction value sequence, and if the slope of the trend is positive and greater than a third preset threshold, then lower the preset health threshold corresponding to the health status prediction submodule to obtain the target health threshold. The step of comparing the adjusted predicted device health status with a preset health threshold, and generating a second control signal if the regenerated predicted device health status exceeds the health threshold, includes the following steps: The adjusted device health status prediction value of each health status prediction submodule is compared with the target health threshold corresponding to each health status prediction submodule. If the adjusted device health status prediction value of any health status prediction submodule exceeds its corresponding target health threshold, a second control signal is generated.

[0009] According to the technical solution provided in this application, after multiplying the current prediction step size of each health status prediction submodule by the corresponding prediction step size adjustment coefficient to obtain the adjusted prediction step size of each health status prediction submodule, the method further includes the following steps: Obtain the device component identifiers corresponding to each health status prediction submodule with adjusted prediction step size; Based on the preset substation operation logic rule base, query the operation logic relationship between the identifiers of each equipment component; Based on the operational logic relationship, the adjusted prediction step size of logically related device components is synchronously corrected so that the prediction step size between logically related device components meets the constraints defined by the operational logic relationship.

[0010] According to the technical solution provided in this application, the step of synchronously correcting the prediction step size of logically related device components based on the operational logic relationship includes the following steps: When the operating logic relationship is a primary / backup relationship, the primary device and the backup device are identified, and the prediction step size of the backup device is adjusted to be the same as the prediction step size of the primary device. When the operating logic relationship is an interlocking relationship, identify all devices in the interlocking group and adjust the prediction step size of all devices in the interlocking group to the prediction step size of the device with the shortest prediction step size in the interlocking group. When the running logic relationship is an operation timing relationship, identify the preceding and following devices in the timing chain, and adjust the prediction step size of the following device to be no less than the prediction step size of the preceding device. When the operational logic relationship is a capacity constraint relationship, all devices in the capacity constraint group are identified. When the prediction step size of any device in the capacity constraint group is shortened, the prediction step size of all devices in the capacity constraint group is shortened synchronously.

[0011] According to the technical solution provided in this application, after generating the second control signal, the method further includes the following steps: Obtain the device component identifier corresponding to the second control signal, and query the available maintenance window for that device component in the preset maintenance window cycle table; The step of generating and outputting maintenance instructions based on the second control signal includes the following steps: If an available maintenance window is found, the second control signal is bound to the available maintenance window to generate a maintenance instruction with execution time constraints.

[0012] According to the technical solution provided in this application, after querying the available maintenance windows of the equipment component in the preset maintenance window cycle table, the method further includes the following steps: If no available maintenance window is found, the maintenance priority level of the device component is obtained, and the second control signal is inserted into the maintenance waiting queue according to the maintenance priority level. When there are multiple second control signals in the maintenance waiting queue, the execution order in the maintenance waiting queue is reordered according to the operational logic relationship between the equipment components corresponding to each second control signal, so that the maintenance instructions of equipment components with interlocking relationships or operation timing relationships are executed in the order defined by the operational logic relationship.

[0013] According to the technical solution provided in this application, after obtaining the target health threshold, the method further includes the following steps: Obtain the historical health status prediction value sequence of this health status prediction submodule before the generation of pseudo-fluctuation identifier, and fit it to obtain the historical degradation trend of health status. Based on the historical degradation trend, estimate the estimated time from the current moment when the predicted health status value will reach the target health threshold; Obtain historical fault data of the equipment component corresponding to the health status prediction submodule, and calculate the historical average interval between when the equipment component reaches the health threshold and when a fault actually occurs. Calculate the total risk exposure duration of the device component based on the expected arrival time and the historical average interval. Obtain the preset maintenance cycle for the equipment component, and compare the total risk exposure time with the preset maintenance cycle: If the total duration of risk exposure is less than or equal to the preset maintenance cycle, then the target health threshold remains unchanged. If the total duration of risk exposure exceeds the preset maintenance cycle, the target health threshold is lowered until the total duration of risk exposure is less than or equal to the preset maintenance cycle.

[0014] Compared with the prior art, the beneficial effects of this application are as follows: I. Real-time linkage between cost assessment and status awareness: By establishing a three-layer digital twin model that couples the physical layer model, data layer model and economic layer model, the real-time status parameters of the equipment (physical layer) are transformed into health status prediction values ​​(data layer), and then further transformed into maintenance cost estimates (economic layer). This enables the cost assessment results to be updated in real time as the equipment status changes, overcoming the defect of cost accounting lagging behind status changes in the existing technology.

[0015] II. Establishing a Cost-Driven Predictive Maintenance Closed Loop: When the estimated maintenance cost exceeds a cost threshold, the prediction step size is automatically shortened, the monitoring frequency of equipment health status is increased, and the adjusted predicted value is compared with the health threshold to trigger a maintenance command. This mechanism transforms cost indicators from passive results of ex-post accounting into proactive adjustment factors that participate in operation and maintenance decisions, forming a closed-loop control chain of cost exceeding limits → increased monitoring → status assessment → maintenance decision-making, thus achieving predictive maintenance under cost constraints.

[0016] 3. Optimize maintenance timing and reduce total life cycle cost: By triggering predictive step size adjustment when the cost exceeds the threshold, it is possible to increase monitoring in advance before the equipment is about to enter the high-cost maintenance stage. When the health status reaches the threshold, maintenance instructions are output in a timely manner to avoid the expansion of failure losses caused by the equipment operating with defects, and at the same time to prevent the waste of resources caused by over-maintenance, thereby achieving optimal control of total life cycle cost. Attached Figure Description

[0017] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 A flowchart illustrating the steps of the digital twin-driven intelligent management method for the entire lifecycle of substation equipment provided in this application. Detailed Implementation

[0018] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0019] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] As mentioned in the background section, this application proposes a digital twin-driven intelligent management method for the entire lifecycle of substation equipment, such as... Figure 1 As shown, it includes the following steps: S1. Establish a three-layer digital twin model for the entire life cycle of the equipment. The three-layer digital twin model includes a physical layer model, a data layer model, and an economic layer model. S2. The output parameters of the physical layer model are fed into the data layer model as input, and the data layer model generates a predicted value of the device health status. S3. The predicted health status of the equipment output by the data layer model is fed into the economic layer model as input, and the economic layer model generates the estimated cost of the cost cycle maintenance. S4. Compare the estimated maintenance cost for this period output by the economic layer model with the cost threshold. If the estimated maintenance cost for this period exceeds the cost threshold, generate a first control signal. S5. According to the first control signal, adjust the prediction step size in the data layer model used to generate the predicted value of the device health status so that the prediction step size is shortened as the maintenance cost increases. S6. Feed the adjusted prediction step size back to the data layer model and regenerate the adjusted device health status prediction value. S7. Compare the adjusted predicted device health status value with the preset health threshold. If the regenerated predicted device health status value exceeds the health threshold, generate a second control signal. S8. Generate and output maintenance instructions based on the second control signal.

[0021] Specifically, this embodiment is executed in a computer system, which includes a data acquisition server, a digital twin modeling server, and a decision output terminal. The data acquisition server is connected to intelligent electronic devices, online monitoring devices, SCADA systems, and the enterprise's ERP system and production management system in the substation via industrial Ethernet, and collects in real time the design parameters of the equipment during the design phase, the process data during the manufacturing phase, the operating condition data and environmental data during the operation phase, the maintenance records during the maintenance phase, and the disposal cost data during the decommissioning phase.

[0022] This method first constructs a three-layer digital twin model. The physical layer model is built using finite element analysis (FEM). CAD models of major equipment such as transformers, circuit breakers, and disconnectors are imported into FEM software like ANSYS or COMSOL. Material property parameters, including elastic modulus, Poisson's ratio, density, thermal conductivity, and coefficient of thermal expansion, are set. Mesh elements are generated, and boundary conditions are set to form a computational model capable of simulating the physical behavior of the equipment under the coupled effects of electromagnetic, temperature, and stress fields. For transformers, the physical layer model focuses on the magnetic flux density distribution of the core, the current density distribution of the windings, the temperature field distribution of the oil flow, and the degree of polymerization of the insulating paper. For circuit breakers, the physical layer model focuses on the pressure changes in the arc-extinguishing chamber, the degree of contact erosion, and the mechanical stress distribution of the operating mechanism. The output parameters of the physical layer model include, but are not limited to: transformer top-layer oil temperature, winding hot spot temperature, core grounding current, vibration amplitude spectrum, and partial discharge quantity; circuit breaker opening and closing coil current waveforms, contact stroke curves, arc-extinguishing chamber gas pressure, and cumulative mechanical operation counts.

[0023] The data layer model is constructed using machine learning methods. The parameters output by the physical layer model are used as input features, and the device health status is used as the output label to construct the training dataset. The training data comes from the device's historical operation records and a fault case library. Health status labels are continuous values ​​between 0 and 1, where 1 indicates the device is in a brand-new state, and 0 indicates the device is completely failed and cannot operate. The labeling method combines expert experience with fault data: for normal operation, the health status value is determined comprehensively based on the device's commissioning time, statistical lifespan curves of similar devices, and periodic test data; for the moment of failure, the health status value is labeled as the critical value before the failure. The data layer model uses a Long Short-Term Memory (LSTM) network structure, which includes an input layer, two hidden layers, and an output layer. The number of nodes in the input layer has the same dimension as the output parameters of the physical layer model. Each hidden layer has 128 memory units, and the output layer outputs a single node's predicted health status value. The network training uses mean squared error as the loss function, and the Adam optimizer is used for parameter optimization. The training epochs are set to 200, and the batch size is 32. After training is complete, the real-time output parameters of the physical layer model are input into the data layer model, and the data layer model outputs a sequence of predicted health status values ​​for multiple future time steps.

[0024] The economic layer model employs cost accounting methods. The life-cycle cost of substation equipment is divided into five major categories: initial investment cost, operating cost, maintenance cost, failure loss cost, and decommissioning cost. Each cost category is further subdivided into specific cost items. Initial investment cost includes equipment procurement costs, installation and commissioning costs, and land occupation costs. Operating costs include energy consumption costs, daily inspection costs, and environmental control costs. Maintenance costs include planned maintenance labor costs, spare parts costs, and tool costs, as well as emergency repair costs, spare parts costs, and downtime losses for fault repairs. Failure loss costs include power outage losses due to equipment failure, direct economic losses from equipment damage, and losses due to the impact on downstream equipment. Decommissioning costs include equipment dismantling costs, transportation costs, and environmental treatment costs. The economic layer model achieves cost estimation by establishing a mapping relationship between maintenance actions and cost items. Specifically, each maintenance action corresponds to a set of cost item weight vectors, where each element represents the cost consumption coefficient of that maintenance action in that cost item. When the predicted health status of the equipment is input, the economic layer model maps the health status value to the corresponding maintenance action type and then calculates the estimated maintenance cost for the current cycle using the weight vectors. The mapping relationship uses a piecewise linear function. For example, when the health status value is greater than 0.8, it corresponds to planned maintenance actions with a low estimated cost; when the health status value is between 0.5 and 0.8, it corresponds to condition-based maintenance actions with a medium estimated cost; and when the health status value is less than 0.5, it corresponds to emergency repair actions with a high estimated cost.

[0025] This method presets a cost threshold of 1.2 times the average annual maintenance budget for the equipment, with the specific value set according to the equipment type and importance. For example, for a 220kV main transformer, with an average annual maintenance budget of 500,000 yuan, the cost threshold is set to 600,000 yuan. The estimated maintenance cost for the current cycle output by the economic layer model is compared with this cost threshold. The comparison process is implemented through a comparator, which receives two inputs: the estimated cost and the cost threshold, and outputs a Boolean value. If the estimated cost is less than or equal to the cost threshold, the comparator outputs 0, indicating that the current cost is controllable and no adjustment is triggered. If the estimated cost is greater than the cost threshold, the comparator outputs 1, indicating that the cost exceeds the threshold, and a first control signal is generated.

[0026] The first control signal triggers the adjustment of the prediction step size. The prediction step size refers to the time interval in days during which the data layer model performs a health status prediction. The default prediction step size is set to 30 days, meaning that the data layer model reads real-time data from the physical layer model and generates a health status prediction value every 30 days. The method for adjusting the prediction step size uses a piecewise function, defining the cost exceeding the threshold ratio r = estimated cost / cost threshold. When r is between 1 and 1.2, the prediction step size is adjusted to 20 days; when r is between 1.2 and 1.5, the prediction step size is adjusted to 15 days; and when r is greater than 1.5, the prediction step size is adjusted to 7 days. The parameters of this piecewise function can be configured according to equipment characteristics and maintenance requirements. The adjusted prediction step size is transmitted to the timer module of the data layer model through a feedback loop, and the timer module resets the execution time of the next prediction based on the new prediction step size.

[0027] The data layer model regenerates the adjusted equipment health status predictions using the adjusted prediction step size. The regeneration process is the same as the initial generation process, but the input data is updated with the latest physical layer model output parameters. Due to the shortened prediction step size, the temporal resolution of the prediction value sequence is improved, enabling a more refined reflection of the changing trajectory of the equipment status.

[0028] This method presets a health threshold, which is set based on the importance level and safety requirements of the equipment. For critical equipment such as main transformers, the health threshold is set to 0.7; for general equipment such as disconnect switches, the health threshold is set to 0.5. The health threshold is stored in the system's parameter configuration library and can be modified through the operation and maintenance management interface. The adjusted predicted health status value of the equipment is compared with the preset health threshold, using the same comparison method as the cost threshold. If the predicted health status value is less than or equal to the health threshold, maintenance is not triggered. If the predicted health status value is greater than the health threshold, a second control signal is generated.

[0029] The second control signal triggers the generation of maintenance instructions. The generation process includes: determining the maintenance type based on the degree to which the predicted health status value exceeds a threshold; if the exceedance is less than 0.1, a check maintenance instruction is generated; if the exceedance is between 0.1 and 0.3, a repair maintenance instruction is generated; if the exceedance is greater than 0.3, an emergency repair maintenance instruction is generated. The maintenance instruction includes equipment identification, maintenance type, suggested maintenance time, and estimated maintenance cost, and is pushed to the operation and maintenance management platform via a message queue. Operation and maintenance personnel then arrange specific maintenance work according to the instruction.

[0030] The technical principle of this embodiment lies in constructing a closed-loop feedback mechanism between cost indicators and monitoring frequency. In existing technologies, cost assessment and condition monitoring are independent of each other; cost data is used for post-event statistics, while condition monitoring is used for real-time early warning, and the two lack linkage. This embodiment, through data feed and feedback control between three-layer models, makes the cost estimate an active adjustment factor for the monitoring frequency. When the cost estimate exceeds a threshold, the system automatically shortens the prediction step size and intensifies monitoring, capturing abnormal signals in the early stages of equipment condition deterioration and promptly outputting maintenance instructions when the health status reaches a threshold. This mechanism brings cost constraints forward to the monitoring stage, achieving synergistic optimization of economy and reliability.

[0031] The technical effects achieved by this embodiment include: First, the cost assessment results affect the monitoring frequency in real time, avoiding the time lag between cost overruns and condition deterioration. Second, the prediction step size adaptively shortens as costs increase, realizing on-demand allocation of monitoring resources. Third, the triggering timing of maintenance commands is jointly determined by cost thresholds and health thresholds, forming optimal decision-making under dual constraints, preventing over-maintenance due to excessive costs and avoiding missed fault detection due to insufficient monitoring.

[0032] In a preferred embodiment, if the estimated maintenance cost for the current period exceeds the cost threshold, the method further includes the following steps: When the estimated maintenance cost for this period exceeds the cost threshold, the cost contribution values ​​of each cost item in the economic layer model are sorted. The cost items include equipment aging cost, fault repair cost, planned maintenance cost, and operating loss cost. Identify cost items that exceed a preset contribution threshold in the cost contribution value, or identify at least one cost item that ranks at the top in the cost contribution value ranking, and obtain the equipment component identifier corresponding to each identified cost item; Based on the device component identifier, locate the health status prediction submodule corresponding to the device component identifier in the data layer model; Adjusting the prediction step size in the data layer model used to generate the predicted value of the device health status includes the following steps: The prediction step size is adjusted only for the identified health status prediction submodule, while the prediction step size of other health status prediction submodules in the data layer model remains unchanged.

[0033] Specifically, after the estimated cost exceeds the cost threshold and a first control signal is generated, the system executes a cost contribution analysis. The cost contribution analysis module reads detailed data from each cost item in the economic layer model. Cost items are categorized according to their relative positions, including equipment aging cost, fault repair cost, planned maintenance cost, and operating loss cost. The cost contribution value of each cost item is calculated as follows: Equipment aging cost is obtained by mapping state parameters such as the polymerization degree of insulating materials, the ablation depth of contacts, and the wear of mechanical parts in the physical layer model. The mapping function is that aging cost equals the degree to which the state parameters deviate from the standard value multiplied by the unit aging cost coefficient. Fault repair cost is obtained by multiplying the predicted fault probability value in the data layer model by the average cost of a single fault repair. Planned maintenance cost is obtained by multiplying the predetermined maintenance items and frequency in the maintenance plan by the standard cost of each item. Operating loss cost is obtained by multiplying the decrease in equipment operating efficiency by the unit loss cost.

[0034] After calculating the cost contribution value of each cost item, the system performs a sorting operation. The sorting uses a quicksort algorithm to arrange the four cost items in descending order of cost contribution value. For example, in a certain execution, the cost contribution value of equipment aging is 450,000 yuan, the cost contribution value of fault repair is 280,000 yuan, the cost contribution value of planned maintenance is 120,000 yuan, and the cost contribution value of operating loss is 80,000 yuan. The cost threshold is 600,000 yuan, and the estimated total cost is 930,000 yuan. The sorting result is: equipment aging cost first, fault repair cost second, planned maintenance cost third, and operating loss cost fourth.

[0035] The system has a preset contribution threshold, set at 30% of the cost threshold, i.e., 180,000 yuan. The system identifies cost items with a contribution value exceeding 180,000 yuan. In this example, equipment aging costs of 450,000 yuan and fault repair costs of 280,000 yuan both exceed 180,000 yuan, therefore these two cost items are identified. The system simultaneously or alternatively employs a ranking identification method, identifying at least one cost item that ranks highly in the list, such as identifying the top three items. The two identification methods can be selected according to the application scenario. This embodiment uses a combination of contribution threshold identification and ranking identification, taking the union of the two to ensure that no key cost sources are missed.

[0036] After identifying key cost items, the system queries a pre-established mapping table between cost items and equipment component identifiers. In the mapping table, equipment aging costs correspond to three component identifiers: transformer core, transformer windings, and circuit breaker contacts; fault repair costs correspond to three component identifiers: circuit breaker operating mechanism, protection device plug-in, and disconnector switch contacts; planned maintenance costs correspond to all major components; and operating loss costs correspond to two component identifiers: transformer cooling system and circuit breaker arc-extinguishing chamber. Based on the identified equipment aging costs and fault repair costs, the system obtains the corresponding component identifier sets, including transformer core, transformer windings, circuit breaker contacts, circuit breaker operating mechanism, protection device plug-in, and disconnector switch contacts.

[0037] The system locates the health status prediction sub-modules corresponding to these component identifiers in the data layer model. The data layer model adopts a modular design, with each device component corresponding to an independent health status prediction sub-module. The naming convention for sub-modules uses the component identifier plus the suffix "Predictor," for example, TransformerCorePredictor corresponds to the transformer core, TransformerWindingPredictor to the transformer winding, BreakerContactPredictor to the circuit breaker contact, BreakerMechanismPredictor to the circuit breaker operating mechanism, ProtectionDevicePredictor to the protection device plug-in, and DisconnectorContactPredictor to the disconnector switch contact. Each sub-module operates independently, using the component's unique feature data for health status prediction. The location process is implemented through string matching, matching the component identifier with the sub-module name and returning the instance of the successfully matched sub-module.

[0038] After location is completed, the system only adjusts the prediction step size of the located health status prediction submodules, while the prediction step size of other submodules remains unchanged. The adjustment method uses piecewise functions that apply only to the located submodules. For example, the original prediction step size for the transformer core submodule was 30 days, but it may be adjusted to 12 days based on the subsequently calculated contribution of exceeding the standard; the original prediction step size for the circuit breaker operating mechanism submodule was 30 days, but it may be adjusted to 18 days; and although the disconnector contact finger submodule is located, its corresponding cost contribution value is relatively small, so the adjustment range is also correspondingly small.

[0039] The technical principle of this embodiment lies in decomposing the total cost exceeding the threshold problem into a cost item contribution problem, and further decomposing it into a device component level problem. Existing technologies typically employ a global adjustment strategy when costs exceed the limit, uniformly increasing the monitoring frequency for all components, resulting in wasted computing resources on non-critical components. This embodiment, through cost contribution ranking and component location, identifies the root cause components that truly lead to cost exceeding the limit, adjusting the monitoring frequency only for these components, while maintaining the original frequency for normal components.

[0040] The technical effects achieved by this embodiment include: First, it enables refined tracing of cost overruns, pinpointing macro-level cost exceeding thresholds to specific equipment components, providing clear direction for subsequent maintenance decisions. Second, it avoids the scattered waste of monitoring resources, concentrating computational resources on components that truly require attention. Third, by mapping cost items to equipment components, it establishes a correlation between economic and technical indicators, enabling cost analysis results to directly guide adjustments to condition monitoring strategies. Fourth, the location results can be further used to guide spare parts procurement plans, allowing for advance stockpiling of spare parts for critical components and shortening fault response time.

[0041] In a preferred embodiment, adjusting the prediction step size only for the located health status prediction submodule includes the following steps: Obtain the cost items corresponding to each health status prediction sub-module located, and extract the cost contribution value of each cost item in the estimated maintenance cost of the current period; Calculate the ratio of each cost contribution value to the cost threshold to obtain the excess contribution of each cost item; Based on the contribution of each cost item to exceeding the standard, the prediction step size adjustment coefficient of each health status prediction sub-module is determined, and the contribution of exceeding the standard is positively correlated with the prediction step size adjustment coefficient. Multiply the current prediction step size of each health status prediction submodule by the corresponding prediction step size adjustment coefficient to obtain the adjusted prediction step size of each health status prediction submodule. Among them, the greater the contribution of the cost item exceeding the standard, the greater the reduction in the prediction step size of the corresponding health status prediction submodule.

[0042] Specifically, the system first obtains the cost items corresponding to each identified health status prediction sub-module. Based on the location results, the transformer core and transformer winding sub-modules correspond to equipment aging costs, while the circuit breaker contacts, circuit breaker operating mechanisms, protection device plug-ins, and disconnector switch contact sub-modules correspond to fault repair costs. The system extracts the cost contribution value of each cost item in the estimated maintenance cost for this period from the economic layer model, with equipment aging costs at 450,000 yuan and fault repair costs at 280,000 yuan.

[0043] The system calculates the ratio of each cost contribution value to a preset cost threshold to obtain the excess contribution rate of each cost item. The preset cost threshold is 600,000 yuan. The excess contribution rate of equipment aging costs is 450,000 divided by 600,000, which equals 0.75. The excess contribution rate of failure repair costs is 280,000 divided by 600,000, which is approximately 0.4667. The excess contribution rate indicates the relative contribution of that cost item to the cost excess; the larger the value, the greater the contribution of that item to the total cost excess.

[0044] The system determines the prediction step size adjustment coefficient for each health status prediction submodule based on the excess contribution of each cost item. The adjustment coefficient is determined using a linear function, equal to 1 minus the excess contribution multiplied by a scaling factor. The scaling factor is set according to the importance of the equipment: 0.8 for critical equipment and 0.6 for general equipment. In this embodiment, the transformer is a critical equipment, so the scaling factor is 0.8; the circuit breaker is also a critical equipment, so the scaling factor is 0.8. The adjustment coefficient for equipment aging costs is 1 minus 0.75 multiplied by 0.8, which equals 0.4. The adjustment coefficient for fault repair costs is 1 minus 0.4667 multiplied by 0.8, which equals 0.6267, approximately 0.63.

[0045] After determining the adjustment coefficients, the system multiplies the current prediction step size of each health status prediction submodule by the corresponding adjustment coefficient to obtain the adjusted prediction step size. The current prediction step size for the transformer core submodule is 30 days, multiplied by 0.4 to get 12 days. The same applies to the transformer winding submodule. The current prediction step size for the circuit breaker contact submodule is 30 days, multiplied by 0.63 to get 18.9 days, rounded to 19 days. The same applies to the circuit breaker operating mechanism submodule. The protection device plug-in submodule and the disconnector switch contact submodule are also adjusted to 19 days with a coefficient of 0.63.

[0046] For the sub-modules corresponding to planned maintenance costs and operating loss costs, since they are not located, their prediction step size remains unchanged at 30 days. However, this embodiment further considers that although the cost items corresponding to these sub-modules are not exceeded, their cost contribution values ​​may be appropriately adjusted if they are close to the threshold. The system sets a secondary adjustment threshold. For cost items whose cost contribution value exceeds the cost threshold by 20% but does not reach 30%, the corresponding sub-module prediction step size adjustment coefficient is set to 0.9, that is, the step size is slightly shortened. In this embodiment, the planned maintenance cost of 120,000 yuan does not exceed 180,000 yuan, so no secondary adjustment is triggered; the operating loss cost of 80,000 yuan also does not exceed the threshold, so no adjustment is triggered.

[0047] After the system completes the adjustment, it writes the adjusted prediction step size into the configuration file of the data layer model. At the beginning of the next prediction cycle, the data layer model reads the new prediction step size and performs health status prediction according to the new step size.

[0048] The technical effects achieved in this embodiment include: First, the step size adjustment range precisely corresponds to cost risk, with components of higher risk receiving higher monitoring frequencies, conforming to the risk-first principle. Second, the calculation of the adjustment coefficient has a clear mathematical formula, is repeatable and verifiable, avoiding the subjectivity of human experience judgment. Third, by setting a scaling factor, the sensitivity of the adjustment can be flexibly adjusted to adapt to the importance of different equipment and the operation and maintenance requirements of different scenarios. Fourth, by using secondary adjustment thresholds, cost items close to the critical value are taken into account, avoiding marginal omissions caused by judging a single threshold. Fifth, it achieves a precise quantitative conversion of cost indicators into technical indicators, providing a unified quantitative basis for subsequent threshold optimization and maintenance decisions.

[0049] In a preferred embodiment, after feeding the adjusted prediction step size back to the data layer model and regenerating the adjusted device health status prediction value, the method further includes the following steps: Obtain the adjusted prediction step size of each health status prediction submodule, calculate the ratio of each adjusted prediction step size to the original prediction step size, and obtain the step size compression ratio of each health status prediction submodule. Obtain the health status prediction value sequence of each health status prediction submodule within multiple consecutive historical prediction periods, and calculate the fluctuation amplitude of the health status prediction value sequence. The step size compression rate and fluctuation amplitude of each health status prediction submodule are compared. If the step size compression rate is greater than the first preset threshold and the fluctuation amplitude is less than the second preset threshold, the change in the current prediction value of the health status prediction submodule is determined to be a pseudo fluctuation caused by the shortening of the step size, and a pseudo fluctuation identifier is generated. If the step size compression rate is greater than the first preset threshold and the fluctuation amplitude is greater than or equal to the second preset threshold, then the change in the current predicted value of the health status prediction submodule is determined to be a real fluctuation in the device's status, and a real fluctuation identifier is generated.

[0050] Specifically, after adjusting the prediction step size, the system writes the adjusted prediction step size into the timer configuration of each submodule of the data layer model. The data layer model then executes the next round of health status prediction according to the new prediction step size, generating the adjusted device health status prediction value. This prediction value is a sequence that changes over time, with each time point corresponding to a health status value between 0 and 1.

[0051] The system then performs a step size compression ratio calculation. The step size compression ratio is defined as the ratio of the adjusted predicted step size to the original predicted step size. The system reads the current predicted step size of each health state prediction submodule from the configuration file of the data layer model, i.e., the adjusted predicted step size, denoted as T_new. Simultaneously, it reads the predicted step size of the submodule before executing the previous implementation method, i.e., the current predicted step size, denoted as T_old. The formula for calculating the step size compression ratio is r = T_new / T_old. Since the application scenario of this invention is that the step size is shortened due to cost exceeding a threshold, T_new is less than T_old, and the step size compression ratio r is between 0 and 1. For example, the transformer core submodule had a step size of 30 days before adjustment and a step size of 12 days after adjustment, with a step size compression ratio of 0.4. The circuit breaker operating mechanism submodule had a step size of 30 days before adjustment and a step size of 19 days after adjustment, with a step size compression ratio of approximately 0.63.

[0052] The system simultaneously calculates the fluctuation range of the health status prediction value sequence. The fluctuation range is calculated using a sliding window method, with the window size set to 10 prediction periods. The system reads the health status prediction value sequence of this health status prediction submodule within the most recent 10 prediction periods, denoted as H1, H2, ..., H10. The standard deviation of this sequence is calculated as a quantitative indicator of the fluctuation range. The formula for calculating the standard deviation is: first, calculate the mean of the sequence μ = (H1+H2+...+H10) / 10; then calculate the sum of squared deviations of each value from the mean, divide by 10, and take the square root to obtain the standard deviation σ. The range of the standard deviation σ is between 0 and 0.5. The closer σ is to 0, the more stable the sequence and the smaller the fluctuation; the larger σ is, the more drastic the fluctuation. For the transformer core submodule, if its predicted health status sequence is 0.82, 0.81, 0.83, 0.80, 0.82, 0.81, 0.79, 0.80, 0.81, 0.82, the calculated standard deviation is approximately 0.011, indicating very small fluctuations. For the circuit breaker operating mechanism submodule, if its sequence is 0.75, 0.78, 0.72, 0.80, 0.71, 0.79, 0.74, 0.82, 0.70, 0.77, the calculated standard deviation is approximately 0.038, indicating larger fluctuations.

[0053] The system has two preset thresholds. The first preset threshold is used to determine whether the step size compression rate reaches a significant level that requires attention; in this embodiment, it is set to 0.7. That is, only when the step size compression rate is less than 0.7 does it indicate that the step size reduction is large, and the fluctuation nature needs to be analyzed. If the step size compression rate is greater than or equal to 0.7, it indicates that the step size adjustment is small, and the fluctuation nature judgment is not triggered; the original health threshold is used directly. The second preset threshold is used to determine whether the fluctuation amplitude belongs to the stationary range; in this embodiment, it is set to 0.02. That is, when the fluctuation amplitude is less than 0.02, the predicted value sequence is considered to be in a stationary state.

[0054] The system compares the step size compression rate of each health status prediction submodule with a first preset threshold and the fluctuation amplitude with a second preset threshold. For the transformer core submodule, the step size compression rate is 0.4, which is less than the first preset threshold of 0.7, meeting the judgment condition; the fluctuation amplitude is 0.011, which is less than the second preset threshold of 0.02, meeting the stationarity condition. The system determines that the current predicted value change of this submodule is a pseudo-fluctuation caused by the shortened step size and generates a pseudo-fluctuation identifier. The pseudo-fluctuation identifier means that the change in the predicted value is mainly due to the increased time resolution caused by the shortened prediction step size, rather than a significant deterioration of the actual equipment status. For the circuit breaker operating mechanism submodule, the step size compression rate is 0.63, which is less than 0.7, meeting the judgment condition; the fluctuation amplitude is 0.038, which is greater than 0.02, not meeting the stationarity condition. The system determines that the current predicted value change of this submodule is a true fluctuation of the equipment status and generates a true fluctuation identifier. The true fluctuation identifier means that the change in the predicted value reflects a true deterioration of the equipment status, which requires attention.

[0055] For submodules with a step size compression ratio greater than or equal to the first preset threshold, the system does not generate any identifier and directly uses the existing health threshold for subsequent comparisons. The system stores the pseudo-fluctuation identifier and the true fluctuation identifier in the identifier record table of the database for subsequent steps to read.

[0056] The technical principle of this embodiment lies in distinguishing between two different sources of change in predicted values ​​through a dual comparison of step size compression ratio and fluctuation amplitude. After shortening the step size, the temporal resolution of the predicted value increases, and subtle fluctuations that were originally smoothed out by low-frequency sampling may become apparent. These fluctuations are not true deterioration of the equipment status, but rather pseudo-fluctuations caused by the increased sampling frequency. If these pseudo-fluctuations are misjudged as status deterioration, it will lead to the generation of a large number of invalid maintenance commands. This embodiment introduces a fluctuation amplitude threshold to classify situations with a large step size compression ratio and small fluctuation amplitude as pseudo-fluctuations, and situations with a large step size compression ratio and large fluctuation amplitude as true fluctuations, thus achieving accurate identification of the nature of the fluctuations.

[0057] The technical effects achieved by this embodiment include: First, it accurately distinguishes between pseudo-fluctuations and true fluctuations, avoiding misjudgment of predicted value fluctuations caused by shortened step sizes as equipment failures. Second, the generation of pseudo-fluctuation identifiers can guide subsequent steps to raise the health threshold, thereby suppressing the output of invalid maintenance commands. Third, the generation of true fluctuation identifiers can guide subsequent steps to maintain or lower the health threshold, ensuring that true equipment failures are not missed. Fourth, by setting a first preset threshold, unnecessary fluctuation nature analysis is avoided when the step size adjustment range is small, saving computational resources. Fifth, by calculating the fluctuation amplitude through a sliding window, the judgment result is based on recent historical data, enabling timely response to changes in equipment status.

[0058] In a preferred embodiment, after generating the pseudo-fluctuation identifier or the real fluctuation identifier, the method further includes the following steps: When the determination result is a pseudo fluctuation identifier, the preset health threshold corresponding to the health status prediction submodule is obtained, the threshold adjustment range is calculated according to the step size compression rate, and the preset health threshold is adjusted by the threshold adjustment range to obtain the target health threshold. When the judgment result is a true fluctuation indicator, the preset health threshold corresponding to the health status prediction submodule is kept unchanged as the target health threshold. Obtain the sequence of health status prediction values ​​of the health status prediction submodule in multiple consecutive historical prediction periods, calculate the slope of the trend of the health status prediction value sequence, and if the slope of the trend is positive and greater than a third preset threshold, then lower the preset health threshold corresponding to the health status prediction submodule to obtain the target health threshold. The step of comparing the adjusted predicted device health status with a preset health threshold, and generating a second control signal if the regenerated predicted device health status exceeds the health threshold, includes the following steps: The adjusted device health status prediction value of each health status prediction submodule is compared with the target health threshold corresponding to each health status prediction submodule. If the adjusted device health status prediction value of any health status prediction submodule exceeds its corresponding target health threshold, a second control signal is generated.

[0059] Specifically, the system reads the identifier type of each health status prediction submodule from the identifier record table in the database. For each submodule, different threshold adjustment strategies are executed based on its identifier type.

[0060] When a submodule is marked as a pseudo-fluctuation indicator, the system obtains the corresponding preset health threshold. The preset health threshold is stored in the configuration parameters of each submodule; for example, the preset health threshold for the transformer core submodule is 0.70, and the preset health threshold for the circuit breaker operating mechanism submodule is 0.65. The system calculates the threshold increase based on the step compression rate. The calculation of the threshold increase uses a linear mapping function, mapping the step compression rate range from 0 to the first preset threshold of 0.7 to an increase range of 0 to 0.1. The specific mapping formula is: Increase = (First preset threshold - Step compression rate) / First preset threshold × Maximum increase. The maximum increase is set to 0.1, meaning the health threshold can be increased by a maximum of 0.1. For the transformer core submodule, with a step compression rate of 0.4, the calculated increase is (0.7 - 0.4) / 0.7 × 0.1 = 0.0429, approximately 0.043. The system adds 0.043 to the preset health threshold of 0.70 to obtain the target health threshold of 0.743.

[0061] When a submodule is marked as a true fluctuation indicator, the system keeps the corresponding preset health threshold unchanged and directly uses the preset health threshold as the target health threshold. For the circuit breaker operating mechanism submodule, the preset health threshold is 0.65, and this value remains unchanged; the target health threshold is 0.65.

[0062] Regardless of whether a submodule is marked, the system performs trend slope calculation. The trend slope calculation uses a linear regression method to obtain the sequence of predicted health status values ​​for the health status prediction submodule over multiple consecutive historical prediction periods. In this embodiment, data from the most recent 15 prediction periods are used. Using the prediction period number as the independent variable t and the predicted health status value as the dependent variable H, a linear regression model H = a × t + b is constructed, where a is the trend slope. A positive trend slope indicates a downward trend in equipment health status; the larger the absolute value of the slope, the faster the deterioration. A negative trend slope indicates improved equipment health status, which typically occurs after maintenance or when the load is reduced, and does not trigger a threshold reduction. The system presets a third threshold, which is set to 0.01 in this embodiment. That is, when the trend slope is greater than 0.01, the equipment is considered to be in an accelerated deterioration state. For the circuit breaker operating mechanism submodule, if the calculated trend slope is 0.015, which is greater than 0.01, the system lowers the preset health threshold for this submodule. The downward adjustment range is determined based on the trend slope. A linear mapping function is used to map the trend slope from the third preset threshold to 0.05 to a downward adjustment range of 0 to 0.1. The downward adjustment range is calculated as: (Trend slope - Third preset threshold) / (0.05 - Third preset threshold) × Maximum downward adjustment range. The maximum downward adjustment range is set to 0.1. Therefore, the calculated downward adjustment range is: (0.015 - 0.01) / (0.05 - 0.01) × 0.1 = 0.0125. The system subtracts 0.0125 from the preset health threshold of 0.65 to obtain the target health threshold of 0.6375.

[0063] After performing the above processing on each submodule, the system obtains the target health threshold for each submodule. For submodules that simultaneously meet multiple conditions, the system executes according to a preset priority order: first, it processes pseudo-fluctuation or real fluctuation indicators, and then adjusts the trend slope on top of that. The priority order is set to avoid conflicts in the adjustment logic. In this embodiment, the adjustment of the trend slope has the highest priority because the risk of accelerated equipment degradation needs to be addressed most promptly. If the trend slope is positive and greater than the third preset threshold, the threshold is further lowered on top of the pseudo-fluctuation or real fluctuation adjustment.

[0064] The system stores the target health thresholds for each submodule in the configuration file of the data layer model, replacing the original preset health thresholds. Subsequent comparisons between predicted health values ​​and the health thresholds will use these updated target health thresholds.

[0065] The technical principle of this embodiment lies in differentiating the health threshold based on the different sources of fluctuations and the differences in degradation trends. For devices with spurious fluctuations, the predicted value fluctuations are caused by shortening step sizes rather than actual equipment deterioration; therefore, the threshold is raised to avoid triggering invalid maintenance commands due to spurious fluctuations. For devices with genuine fluctuations, the predicted value fluctuations reflect changes in the actual equipment condition; therefore, the threshold is kept unchanged to ensure that genuine fault signals are not filtered out. For devices with accelerated degradation, the threshold is lowered to reduce the threshold for triggering maintenance and achieve early warning. This differentiated adjustment strategy enables the health threshold to adaptively match the current state characteristics and risk level of the equipment.

[0066] The technical effects achieved in this embodiment include: First, the increased threshold triggered by the pseudo-fluctuation indicator effectively suppresses false alarms caused by the shortened step size, avoids the generation of invalid maintenance commands, and saves operation and maintenance resources. Second, the threshold triggered by the true fluctuation indicator remains unchanged, ensuring that real equipment faults can be detected in a timely manner and avoiding missed risk assessments. Third, the decreased threshold triggered by the trend slope provides early warning of accelerated equipment degradation, transforming the maintenance strategy from passive response to proactive prevention. Fourth, by setting a priority order, the logical conflict problem when multiple conditions are met simultaneously is resolved, ensuring the determinism of threshold adjustment. Fifth, the adjustment range of the threshold is quantitatively correlated with the step size compression rate or trend slope, achieving precise control of the adjustment intensity and avoiding the arbitrariness of subjective experience judgment.

[0067] In a preferred embodiment, after multiplying the current prediction step size of each health status prediction submodule by the corresponding prediction step size adjustment coefficient to obtain the adjusted prediction step size of each health status prediction submodule, the method further includes the following steps: Obtain the device component identifiers corresponding to each health status prediction submodule with adjusted prediction step size; Based on the preset substation operation logic rule base, query the operation logic relationship between the identifiers of each equipment component; Based on the operational logic relationship, the adjusted prediction step size of logically related device components is synchronously corrected so that the prediction step size between logically related device components meets the constraints defined by the operational logic relationship.

[0068] Specifically, after performing the previous step size adjustment, the system obtains the device component identifiers corresponding to each health status prediction submodule with the adjusted prediction step size. The device component identifiers adopt a unified coding rule. For example, the identifier for the transformer core is TRF_CORE_001, the identifier for the transformer winding is TRF_WND_001, the identifier for the circuit breaker contact is CBR_CT_001, the identifier for the circuit breaker operating mechanism is CBR_MECH_001, the identifier for the protection device plug-in is PROT_PLG_001, and the identifier for the disconnecting switch contact finger is DIS_CT_001.

[0069] The system calls the preset substation operation logic rule base. This rule base is a relational database that stores the operational logic relationships between all equipment components within the substation. Operational logic relationships include four types: primary / standby relationships, interlocking relationships, operation sequence relationships, and capacity constraint relationships. Primary / standby relationships describe redundant configurations between equipment, such as a main transformer and a standby transformer; when the main transformer is running, the standby transformer is in hot standby mode. Interlocking relationships describe safety interlocks between equipment; for example, a disconnecting switch can only be operated after the circuit breaker is open, otherwise a safety accident may occur. Operation sequence relationships describe the sequential requirements of equipment operations; for example, during power-on operations, the disconnecting switch must be closed before the circuit breaker is closed, and during power-off operations, the circuit breaker must be opened before the disconnecting switch is opened. Capacity constraint relationships describe capacity matching requirements between equipment; for example, the transformer capacity must match the rated capacity of the incoming circuit breaker, and when the transformer load rate exceeds a certain threshold, the incoming circuit breaker is also at risk of overload.

[0070] The system queries the operational logic relationships between the identifiers of various equipment components. For transformer cores and transformer windings, they belong to the same transformer body component and have a strong correlation, but the rule base does not define a direct operational logic relationship because they do not involve operational constraints between different devices. For circuit breaker contacts and circuit breaker operating mechanisms, they belong to the same circuit breaker equipment and similarly do not involve cross-device constraints. For circuit breaker contacts and disconnector switch contacts, the rule base defines an interlocking relationship because the disconnector switch can only be operated after the circuit breaker is opened. For main transformers and standby transformers, the rule base defines a main / standby relationship.

[0071] Based on the retrieved operational logic relationships, the system synchronously corrects the adjusted prediction step sizes of logically related device components. The purpose of this synchronous correction is to ensure that the prediction step sizes of logically related device components meet the constraints defined by the operational logic relationships.

[0072] For primary / standby relationships, the system identifies primary and standby equipment. Primary equipment refers to the equipment currently in operation, while standby equipment refers to redundant equipment in hot or cold standby status. The system adjusts the prediction step size of the standby equipment to be the same as that of the primary equipment. For example, if the prediction step size of the primary transformer is adjusted to 12 days, and the prediction step size of the standby transformer was originally 30 days, the system will synchronously correct the prediction step size of the standby transformer to 12 days. This is because when the primary equipment fails and needs to be switched over, the status of the standby equipment must be monitored synchronously with that of the primary equipment; otherwise, it might be discovered that the standby equipment is also in a faulty state only during the switchover.

[0073] For interlocking relationships, the system identifies all equipment within the interlocking group. An interlocking group refers to a set of equipment with interlocking relationships; for example, a circuit breaker and a disconnector form an interlocking group. The system adjusts the prediction step size of all equipment in the interlocking group to the prediction step size of the equipment with the shortest prediction step size in the group. For example, if the prediction step size of the circuit breaker is 19 days, and the prediction step size of the disconnector was originally 30 days, the system will synchronously correct the prediction step size of the disconnector to 19 days. This is because the status of both devices must be reliable simultaneously during interlocking operations. If the monitoring frequency of the disconnector is lower than that of the circuit breaker, the status of the disconnector may not be updated in time even when the status of the circuit breaker is known, leading to safety risks during interlocking operations.

[0074] For operational timing relationships, the system identifies preceding and following devices in the timing chain. A preceding device is the one operated on first in the sequence, and a following device is the one operated on last. The system adjusts the prediction step size of the following device to be no less than that of the preceding device. For example, in a power outage operation, the circuit breaker is the preceding device, and the disconnector is the following device. If the prediction step size of the circuit breaker is 19 days, and the original prediction step size of the disconnector is 30 days, 30 days is no less than 19 days, satisfying the constraint, and no adjustment is needed. If the original prediction step size of the disconnector is 15 days, which is less than 19 days, the system adjusts the prediction step size of the disconnector to 19 days. This is because the status monitoring frequency of the following device should not be lower than that of the preceding device; otherwise, changes in the status of the preceding device may not be promptly transmitted to the operational decisions of the following device.

[0075] For capacity constraints, the system identifies all devices within the capacity constraint group. A capacity constraint group refers to a group of devices with capacity matching requirements, such as a transformer and its incoming circuit breaker. When the prediction step size of any device in the capacity constraint group is shortened, the system synchronously shortens the prediction step size of all devices in the group. For example, if the prediction step size of the transformer core submodule is adjusted to 12 days, the system will also synchronously adjust the prediction step size of the incoming circuit breaker submodule to 12 days, even if the cost contribution of the circuit breaker itself does not exceed the limit. This is because when the monitoring frequency of critical equipment increases, the monitoring frequency of equipment with matching capacity should also increase to avoid safety risks caused by capacity mismatch.

[0076] The system will write the revised prediction step size to the configuration file of the data layer model, overwriting the previous adjustment results. The data layer model will then perform health status predictions according to the revised prediction step size.

[0077] The technical principle of this embodiment lies in incorporating the substation's operational logic rules into the prediction step size adjustment process. Existing technologies, when adjusting step sizes, only consider the cost-effectiveness of individual devices, neglecting the operational logic constraints between devices. This can lead to the adjusted step size configuration potentially violating the safety requirements of device operation. This embodiment, by establishing an operational logic rule base and performing synchronous correction after step size adjustment, ensures that the adjustment result satisfies both economic and safety requirements, achieving a balance between economy and safety.

[0078] The technical effects achieved by this embodiment include: First, the synchronous correction of the primary / standby relationship ensures that the standby equipment maintains the same monitoring frequency as the primary equipment, avoiding the risk of unknown status of the standby equipment during primary / standby switching. Second, the synchronous correction of the interlocking relationship ensures that the monitoring frequency of all equipment within the interlocking group is consistent, guaranteeing the synchronization of status during interlocking operations and improving operational safety. Third, the synchronous correction of the operation sequence relationship ensures that the monitoring frequency of the downstream equipment is not lower than that of the upstream equipment, maintaining the continuity of status transmission in the operation sequence and avoiding operational errors caused by monitoring blind spots. Fourth, the synchronous correction of the capacity constraint relationship ensures that the monitoring frequencies of equipment groups with matching capacities are synchronized, avoiding the omission of capacity risks due to monitoring asymmetry. Fifth, as a post-processing step of step size adjustment, the synchronous correction does not change the original cost-driven adjustment logic, but rather adds safety constraints on top of it, achieving a balance between economy and safety.

[0079] In a preferred embodiment, the step of synchronously correcting the prediction step size of logically related device components according to the operational logic relationship includes the following steps: When the operating logic relationship is a primary / backup relationship, the primary device and the backup device are identified, and the prediction step size of the backup device is adjusted to be the same as the prediction step size of the primary device. When the operating logic relationship is an interlocking relationship, identify all devices in the interlocking group and adjust the prediction step size of all devices in the interlocking group to the prediction step size of the device with the shortest prediction step size in the interlocking group. When the running logic relationship is an operation timing relationship, identify the preceding and following devices in the timing chain, and adjust the prediction step size of the following device to be no less than the prediction step size of the preceding device. When the operational logic relationship is a capacity constraint relationship, all devices in the capacity constraint group are identified. When the prediction step size of any device in the capacity constraint group is shortened, the prediction step size of all devices in the capacity constraint group is shortened synchronously.

[0080] Specifically, when performing synchronization correction, the system first queries the operational logic rule base for the type of operational logic relationship. The rule base uses a relational database table structure for storage, with each record containing three fields: relationship type, equipment component identifier group, and relationship parameters. The relationship type field can take values ​​such as master-slave relationship, interlocking relationship, operation sequence relationship, or capacity constraint relationship. The equipment component identifier group field stores a list of equipment component identifiers participating in the logical relationship. The relationship parameters field stores the corresponding additional parameters for the relationship: for master-slave relationships, it stores the primary equipment identifier; for operation sequence relationships, it stores the lists of preceding and following equipment identifiers; and for capacity constraint relationships, it stores the capacity matching coefficient.

[0081] When the retrieved operational logic relationship is a primary / standby relationship, the system identifies the primary and standby devices from the relationship parameter fields. The primary device is the device currently in operation, while the standby device is a redundant device in hot or cold standby mode. The primary and standby devices are functionally equivalent, and the standby device should be able to seamlessly switch to operation when the primary device fails or requires maintenance. The system obtains the adjusted prediction step size of the health status prediction submodule corresponding to the primary device and uses this step size as the target step size. Then, the system locates the health status prediction submodule corresponding to the standby device and reads its current prediction step size. If the current prediction step size of the standby device is the same as that of the primary device, no adjustment is needed. If they are different, the system modifies the prediction step size of the standby device to that of the primary device and writes the modified value into the data layer model configuration file. For example, if the prediction step size of the primary transformer core submodule is adjusted to 12 days, and the original prediction step size of the standby transformer core submodule was 30 days, the system will synchronously correct the prediction step size of the standby transformer core submodule to 12 days. After synchronization correction, the health status prediction frequency of the primary and backup devices remains consistent. When the status of the primary device changes, the status of the backup device can also be updated synchronously, ensuring that the status information of the backup device is up-to-date during primary-backup switchover.

[0082] When the retrieved operational logic relationship is an interlocking relationship, the system identifies all devices in the interlocking group. An interlocking relationship refers to the interlocking of operations between devices; for example, a disconnector switch can only be operated after the circuit breaker has been opened, otherwise it may cause a safety accident involving the opening and closing of the disconnector switch under load. Devices in the interlocking group must maintain synchronized states during operation; a change in the state of any device may affect the operating permissions of other devices in the interlocking group. The system iterates through each device component in the interlocking group, obtaining the prediction step size of the corresponding health status prediction submodule for each device. The system finds the minimum prediction step size and uses it as the target step size. Then, the system modifies the prediction step size of all devices in the interlocking group to this minimum value. For example, if the prediction step size of the circuit breaker operating mechanism submodule is 19 days, and the prediction step size of the disconnector switch contact submodule was originally 30 days, the system will synchronously correct the prediction step size of the disconnector switch contact submodule to 19 days. After this correction, the monitoring frequency of all equipment in the interlocking group is unified to the highest frequency in the group, ensuring that when any interlocking operation occurs, the status of all relevant equipment has been updated with the latest predicted value, avoiding the lag in status information caused by inconsistent monitoring frequencies.

[0083] When the retrieved operational logic relationship is an operation sequence relationship, the system identifies the preceding and following devices in the sequence chain. The operation sequence relationship refers to the order in which device operations must be performed. For example, in a power-on operation, the isolating switch must be closed before the circuit breaker, and in a power-off operation, the circuit breaker must be opened before the isolating switch. In the sequence chain, the preceding device is the device that is operated on first, and the following device is the device that is operated on last. The system obtains the prediction step size of the health status prediction submodule corresponding to the preceding device, denoted as T_front. It also obtains the prediction step size of the health status prediction submodule corresponding to the following device, denoted as T_rear. The system compares T_rear with T_front. If T_rear is greater than or equal to T_front, the constraint condition is met, and no adjustment is needed. If T_rear is less than T_front, the system modifies the prediction step size of the following device to T_front. For example, in a power-on operation, the isolating switch is the preceding device, and the circuit breaker is the following device. If the prediction step size of the disconnector contact finger submodule is 19 days, and the original prediction step size of the circuit breaker operating mechanism submodule is 15 days, 15 days is less than 19 days, which does not meet the constraint that the step size of the downstream device must not be less than that of the upstream device. The system corrects the prediction step size of the circuit breaker operating mechanism submodule to 19 days. After this correction, the monitoring frequency of the downstream device is not lower than that of the upstream device. When the state of the upstream device changes, the downstream device can capture the state change at the same or higher frequency, ensuring the continuous and complete state transmission in the operation sequence.

[0084] When the retrieved operational logic relationship is a capacity constraint relationship, the system identifies all devices in the capacity constraint group. A capacity constraint relationship refers to a capacity matching requirement between devices. For example, the rated capacity of a transformer must match the rated capacity of an incoming circuit breaker; when the transformer load rate increases, the load on the incoming circuit breaker also increases accordingly. Devices in the capacity constraint group are electrically connected in series or parallel; a change in the load of any device will affect the load status of other devices in the group. The system detects whether the predicted step size of any device in the capacity constraint group has been shortened. This shortening may originate from cost overrun adjustments or other synchronous corrections. If a shortened predicted step size is detected for any device, the system synchronously shortens the predicted step size of all devices in the capacity constraint group, with the shortening magnitude consistent with the triggering device. For example, if the predicted step size of the transformer core submodule is shortened to 12 days, the system will also synchronously shorten the predicted step size of the incoming circuit breaker submodule to 12 days, even if the cost contribution of the circuit breaker itself does not exceed the limit. If the predicted step sizes of multiple devices in the capacity constraint group are shortened, the system takes the minimum value among all shortened step sizes as the target step size and applies it synchronously to all devices in the group.

[0085] The system writes the synchronized and corrected prediction step size into the configuration file of the data layer model. In subsequent prediction cycles, the data layer model performs health status predictions according to the corrected step size.

[0086] The technical principle of this embodiment lies in transforming the operational logic relationships between substation equipment into synchronous correction rules for the predicted step size. The primary / standby relationship requires that standby equipment maintain the same monitoring frequency as the primary equipment, ensuring that the status of the standby equipment is known during primary / standby switching. The interlocking relationship requires that all equipment within an interlocking group maintain the same monitoring frequency, ensuring state synchronization during interlocking operations. The operation sequence relationship requires that the monitoring frequency of downstream equipment is not lower than that of upstream equipment, ensuring continuous state transmission in the operation sequence. The capacity constraint relationship requires that the monitoring frequencies of equipment groups with matching capacities be synchronized, ensuring that capacity risks are detected in a timely manner. These four correction rules cover the main logical association types between substation equipment, ensuring that the adjustment results of the predicted step size meet both economic and operational safety requirements.

[0087] In a preferred embodiment, after generating the second control signal, the method further includes the following step: Obtain the device component identifier corresponding to the second control signal, and query the available maintenance window for that device component in the preset maintenance window cycle table; The step of generating and outputting maintenance instructions based on the second control signal includes the following steps: If an available maintenance window is found, the second control signal is bound to the available maintenance window to generate a maintenance instruction with execution time constraints.

[0088] Specifically, after performing the health threshold comparison, if the adjusted device health status prediction value of any health status prediction submodule exceeds its corresponding target health threshold, a second control signal is generated. The second control signal includes information such as the device component identifier that triggered the signal, the predicted health status value, the target health threshold, and the extent of the exceedance. The system then transmits this signal to the maintenance instruction generation module.

[0089] The maintenance instruction generation module first parses the second control signal and extracts the equipment component identifier. The system calls the maintenance window periodic table query interface, passing the equipment component identifier as the query parameter. The maintenance window periodic table is a pre-built database table that stores information on all planned power outage windows of the substation within the year. Each record includes fields such as window number, window start time, window end time, window type, list of maintainable equipment, and window status. Window types are divided into annual maintenance windows, temporary power outage windows, and holiday maintenance windows. The maintainable equipment list field stores a list of equipment component identifiers that can be scheduled for maintenance during the window period. The window status field indicates whether the window is occupied, partially occupied, or idle.

[0090] The system searches the maintenance window cycle table for maintenance windows in the maintainable equipment list that contain the corresponding equipment component identifier and whose window status is idle or partially occupied, based on the equipment component identifier. The query results may return multiple available windows, which are sorted by start time from earliest to latest, and the window with the earliest start time is selected as the preferred window. If the query result is empty, it means that there are currently no available maintenance windows, and the system executes the processing procedure described later.

[0091] If an available maintenance window is found, the system binds the second control signal to that window. The binding process includes: obtaining the window number and start / end time; filling the window's maintenance task list with information such as the equipment component identifier and maintenance type from the second control signal; and updating the window status to partially occupied or occupied. The system generates a maintenance instruction with execution time constraints. The instruction includes the equipment component identifier, maintenance type, suggested maintenance time, maintenance window number, estimated maintenance duration, and estimated maintenance cost. The suggested maintenance time is taken from the window's start time, and the estimated maintenance duration is obtained from the standard working hours database based on the maintenance type. For example, the estimated maintenance duration for transformer overhaul is 8 hours, and for circuit breaker maintenance, it is 4 hours.

[0092] The generated maintenance instructions are pushed to the operation and maintenance management platform via a message queue. Operation and maintenance personnel can view all pending maintenance instructions on the platform. Maintenance instructions are displayed in order of suggested maintenance time, with reminders sent in advance for instructions nearing their execution time. After the operation and maintenance personnel confirm the instruction, the system updates the window status to "occupied" and synchronizes the maintenance plan to the production management system and dispatch system, ensuring that maintenance work is coordinated with the power grid operation plan.

[0093] The technical principle of this embodiment lies in associating the generation of maintenance instructions with available maintenance windows, thereby coordinating the maintenance plan with the power grid outage plan. Most maintenance work on substation equipment needs to be performed under power outage conditions, and outage windows are a limited and scarce resource. Existing technologies often ignore the constraints of outage windows when generating maintenance instructions, leading to instructions being unable to be executed due to the inability to schedule a power outage, or temporary power outage requests causing risks to power grid operation. This embodiment binds maintenance instructions to specific windows through a pre-established maintenance window periodic table, synchronizing the maintenance plan with the outage plan and ensuring the executability of the instructions.

[0094] In a preferred embodiment, after querying the available maintenance windows of the device component within a preset maintenance window cycle table, the method further includes the following steps: If no available maintenance window is found, the maintenance priority level of the device component is obtained, and the second control signal is inserted into the maintenance waiting queue according to the maintenance priority level. When there are multiple second control signals in the maintenance waiting queue, the execution order in the maintenance waiting queue is reordered according to the operational logic relationship between the equipment components corresponding to each second control signal, so that the maintenance instructions of equipment components with interlocking relationships or operation timing relationships are executed in the order defined by the operational logic relationship.

[0095] Specifically, when the maintenance instruction generation module performs a window query, if the query result is empty, that is, there is no available maintenance window, the system enters the waiting queue processing flow.

[0096] The system first obtains the maintenance priority level of the equipment component corresponding to the second control signal. The maintenance priority level is a pre-defined level based on factors such as the importance of the equipment, the consequences of a failure, and its health status, and is divided into three levels: Level 1, Level 2, and Level 3. Level 1 priority corresponds to critical equipment, such as main transformers and bus circuit breakers; failures in these devices could lead to large-scale power outages and require priority maintenance. Level 2 priority corresponds to important equipment, such as feeder circuit breakers and protection devices; failures in these devices might lead to localized power outages. Level 3 priority corresponds to general equipment, such as disconnecting switches and grounding switches; failures in these devices have a smaller impact. The maintenance priority level is stored in the equipment ledger database, and the system retrieves it by querying the equipment component identifier. For example, the maintenance priority of the main transformer is Level 1, the feeder circuit breaker is Level 2, and the disconnecting switch is Level 3.

[0097] The system inserts the second control signal into the maintenance waiting queue based on its maintenance priority level. The maintenance waiting queue is a priority-based data structure, where each element contains complete information about the second control signal and an insertion timestamp. The insertion rule is as follows: the priority queue is sorted by maintenance priority level, with level 1 priority at the front, level 2 next, and level 3 at the back. Within the same priority level, elements are sorted according to the order of their insertion timestamps, with earlier inserted elements at the front and later inserted elements appended to the back. This insertion rule ensures that high-priority maintenance requests are processed first, and within the same priority level, a first-come, first-served principle is applied.

[0098] Multiple second control signals may exist simultaneously in the maintenance waiting queue, triggered by health thresholds of different device components. When a new maintenance window is released, the system retrieves a signal from the head of the queue for window matching. However, before retrieving the signal, the system reorders the execution order in the queue. The reordering is based on the operational logic relationships between the device components corresponding to each second control signal. These operational logic relationships include primary / standby relationships, interlocking relationships, operation timing relationships, and capacity constraint relationships.

[0099] The system iterates through each second control signal in the maintenance waiting queue and extracts the corresponding device component identifier. The system then calls the runtime logic rule base to query the runtime logic relationships between these device component identifiers. If interlocking or timing relationships are found, the system reorders these signals.

[0100] For interlocking relationships, the system identifies all equipment components within the interlocking group. Since the equipment within an interlocking group must operate synchronously or sequentially, centralizing the execution of maintenance commands for these devices reduces the number of power outages. The system removes the second control signals corresponding to all equipment within the interlocking group from the queue and re-inserts them as a consecutive group. This group of signals is placed at the front of the queue as a whole, and the order of signals within the group follows the interlocking operation sequence. For example, if a circuit breaker and a disconnector are interlocked, with the circuit breaker having a maintenance priority of level two and the disconnector a maintenance priority of level three, the circuit breaker would normally be prioritized before the disconnector. However, the interlocking relationship requires simultaneous maintenance of both. The system merges them into one group, placing it at the front of the queue, with the order within the group being disconnector first, then circuit breaker, conforming to the power outage operation sequence.

[0101] Regarding the timing of operations, the system identifies the preceding and following devices in the timing chain. The timing requires that maintenance of the preceding device be performed before maintenance of the following device. The system adjusts the queue order so that the signal corresponding to the preceding device precedes the signal of the following device. For example, in a power outage operation sequence, disconnector maintenance precedes circuit breaker maintenance. If the circuit breaker signal precedes the disconnector signal in the queue, the system will swap their positions to ensure that the disconnector signal is processed first.

[0102] After reordering is complete, the system stores the updated maintenance wait queue in the database. When a new maintenance window is released, the system retrieves a signal from the head of the queue and executes the window matching and instruction generation process.

[0103] The technical principle of this embodiment lies in combining the maintenance waiting queue with operational logic relationships to achieve optimized scheduling of maintenance execution order. Existing technologies, when handling multiple maintenance requests, typically execute them simply in chronological or priority order, ignoring the operational logic constraints between devices. This may lead to maintenance sequences violating operational requirements or requiring interlocked devices to be maintained separately, increasing the number of power outages. This embodiment introduces operational logic relationships to reorder the waiting queue, ensuring the maintenance order meets the safety requirements of equipment operation, while simultaneously centralizing the processing of interlocked devices to reduce the number of power outages.

[0104] In a preferred embodiment, after obtaining the target health threshold, the method further includes the following steps: Obtain the historical health status prediction value sequence of this health status prediction submodule before the generation of pseudo-fluctuation identifier, and fit it to obtain the historical degradation trend of health status. Based on the historical degradation trend, estimate the estimated time from the current moment when the predicted health status value will reach the target health threshold; Obtain historical fault data of the equipment component corresponding to the health status prediction submodule, and calculate the historical average interval between when the equipment component reaches the health threshold and when a fault actually occurs. Calculate the total risk exposure duration of the device component based on the expected arrival time and the historical average interval. Obtain the preset maintenance cycle for the equipment component, and compare the total risk exposure time with the preset maintenance cycle: If the total duration of risk exposure is less than or equal to the preset maintenance cycle, then the target health threshold remains unchanged. If the total duration of risk exposure exceeds the preset maintenance cycle, the target health threshold is lowered until the total duration of risk exposure is less than or equal to the preset maintenance cycle.

[0105] Specifically, this embodiment further refines the processing steps after obtaining the target health threshold to ensure that the adjusted health threshold will not cause equipment failure before the next planned maintenance. After performing the threshold adjustment, the system obtains the target health threshold for each health status prediction submodule. Taking the transformer core submodule as an example, the threshold is adjusted upward due to a pseudo-fluctuation flag, and the target health threshold is 0.743. The system performs risk verification on this threshold.

[0106] The system first obtains the historical health status prediction value sequence of the health status prediction submodule before the generation of the pseudo-fluctuation identifier, such as the monthly prediction values ​​for the past 12 months. Based on this historical data, a linear regression method is used to fit a historical degradation trend function of the health status, which reflects the changing pattern of the equipment's health status over time.

[0107] Based on the historical degradation trend function and the current predicted health status value, the system estimates the estimated time required for the predicted health status value to reach the target health threshold from the current moment. For example, if the current health status is 0.78 and the historical degradation rate is a decrease of 0.0125 per month, the estimated time to reach the target health threshold is approximately 89 days.

[0108] The system acquires historical fault data for the device component and calculates the average interval between when the component reaches a healthy threshold and when a fault actually occurs. For example, based on multiple fault records, the average interval is calculated to be 48 days.

[0109] The system adds the expected arrival time to the historical average interval to obtain the total duration of risk exposure. Taking the above value as an example, the total duration of risk exposure is 137 days.

[0110] The system obtains the preset maintenance cycle for the equipment component. For example, a 220kV transformer is typically maintained every 12 months, with a preset maintenance cycle of 365 days. The total risk exposure duration is compared to the preset maintenance cycle. If the total risk exposure duration is less than or equal to the preset maintenance cycle, it indicates that the risk of equipment failure before the next planned maintenance is controllable, and the target health threshold remains unchanged. If the total risk exposure duration is greater than the preset maintenance cycle, it indicates that the target health threshold is set too high, and the equipment may fail before the next maintenance. The system then lowers the target health threshold.

[0111] The threshold is lowered using a gradual approximation method, decreasing the threshold by a small step each time and recalculating the total risk exposure time until the total risk exposure time meets the condition of being less than or equal to the preset maintenance cycle. The lowered health threshold is used as the updated target health threshold for subsequent health status comparisons. The technical principle of this embodiment lies in combining the adjustment of the health threshold with the historical degradation pattern of the equipment and the maintenance cycle constraint. Although threshold increases triggered by pseudo-fluctuations can suppress false alarms, they may make the threshold too high, causing the equipment to reach a critical failure state before the next planned maintenance. By estimating the time to reach the threshold, statistically analyzing the average interval from the threshold to the failure, and comparing it with the maintenance cycle, the rationality of the threshold can be verified, and the threshold can be adjusted back if necessary to ensure safety.

[0112] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A digital twin-driven intelligent management method for the entire lifecycle of substation equipment, characterized in that, Includes the following steps: Establish a three-layer digital twin model for the entire life cycle of equipment. The three-layer digital twin model includes a physical layer model, a data layer model, and an economic layer model. The output parameters of the physical layer model are fed into the data layer model as input, and the data layer model generates a predicted value of the device health status. The predicted health status of the equipment output by the data layer model is fed into the economic layer model as input, and the economic layer model generates the estimated cost of the cost cycle maintenance. The estimated maintenance cost for the current cycle output by the economic layer model is compared with a cost threshold. If the estimated maintenance cost for the current cycle exceeds the cost threshold, a first control signal is generated. Based on the first control signal, the prediction step size used to generate the predicted value of the device health status in the data layer model is adjusted so that the prediction step size is shortened as maintenance costs increase. The adjusted prediction step size is fed back to the data layer model to regenerate the adjusted device health status prediction value. The adjusted predicted device health status value is compared with a preset health threshold. If the regenerated predicted device health status value exceeds the health threshold, a second control signal is generated. Based on the second control signal, a maintenance command is generated and output.

2. The digital twin driven substation equipment life cycle intelligent management method of claim 1, wherein, If the estimated maintenance cost for this period exceeds the cost threshold, the process further includes the following steps: When the estimated maintenance cost for this period exceeds the cost threshold, the cost contribution values ​​of each cost item in the economic layer model are sorted. The cost items include equipment aging cost, fault repair cost, planned maintenance cost, and operating loss cost. Identify cost items that exceed a preset contribution threshold in the cost contribution value, or identify at least one cost item that ranks at the top in the cost contribution value ranking, and obtain the equipment component identifier corresponding to each identified cost item; Based on the device component identifier, locate the health status prediction submodule corresponding to the device component identifier in the data layer model; Adjusting the prediction step size in the data layer model used to generate the predicted value of the device health status includes the following steps: The prediction step size is adjusted only for the identified health status prediction submodule, while the prediction step size of other health status prediction submodules in the data layer model remains unchanged.

3. The digital twin driven substation equipment life cycle intelligent management method of claim 2, wherein, The adjustment of the prediction step size only for the located health status prediction submodule includes the following steps: Obtain the cost items corresponding to each health status prediction sub-module located, and extract the cost contribution value of each cost item in the estimated maintenance cost of the current period; Calculate the ratio of each cost contribution value to the cost threshold to obtain the excess contribution of each cost item; Based on the contribution of each cost item to exceeding the standard, the prediction step size adjustment coefficient of each health status prediction sub-module is determined, and the contribution of exceeding the standard is positively correlated with the prediction step size adjustment coefficient. Multiply the current prediction step size of each health status prediction submodule by the corresponding prediction step size adjustment coefficient to obtain the adjusted prediction step size of each health status prediction submodule. Among them, the greater the contribution of the cost item exceeding the standard, the greater the reduction in the prediction step size of the corresponding health status prediction submodule.

4. The digital twin driven substation equipment lifecycle intelligent management method of claim 3, wherein, After feeding the adjusted prediction step size back to the data layer model and regenerating the adjusted device health status prediction value, the method further includes the following steps: Obtain the adjusted prediction step size of each health status prediction submodule, calculate the ratio of each adjusted prediction step size to the original prediction step size, and obtain the step size compression ratio of each health status prediction submodule. Obtain the health status prediction value sequence of each health status prediction submodule within multiple consecutive historical prediction periods, and calculate the fluctuation amplitude of the health status prediction value sequence. The step size compression rate and fluctuation amplitude of each health status prediction submodule are compared. If the step size compression rate is greater than the first preset threshold and the fluctuation amplitude is less than the second preset threshold, the change in the current prediction value of the health status prediction submodule is determined to be a pseudo fluctuation caused by the shortening of the step size, and a pseudo fluctuation identifier is generated. If the step size compression rate is greater than the first preset threshold and the fluctuation amplitude is greater than or equal to the second preset threshold, then the change in the current predicted value of the health status prediction submodule is determined to be a real fluctuation in the device's status, and a real fluctuation identifier is generated.

5. The digital twin driven substation equipment life cycle intelligent management method of claim 4, wherein, After generating pseudo-fluctuation indicators or real fluctuation indicators, the following steps are also included: When the determination result is a pseudo fluctuation identifier, the preset health threshold corresponding to the health status prediction submodule is obtained, the threshold adjustment range is calculated according to the step size compression rate, and the preset health threshold is adjusted by the threshold adjustment range to obtain the target health threshold. When the judgment result is a true fluctuation indicator, the preset health threshold corresponding to the health status prediction submodule is kept unchanged as the target health threshold. Obtain the sequence of health status prediction values ​​of the health status prediction submodule in multiple consecutive historical prediction periods, calculate the slope of the trend of the health status prediction value sequence, and if the slope of the trend is positive and greater than a third preset threshold, then lower the preset health threshold corresponding to the health status prediction submodule to obtain the target health threshold. The step of comparing the adjusted predicted device health status with a preset health threshold, and generating a second control signal if the regenerated predicted device health status exceeds the health threshold, includes the following steps: The adjusted device health status prediction value of each health status prediction submodule is compared with the target health threshold corresponding to each health status prediction submodule. If the adjusted device health status prediction value of any health status prediction submodule exceeds its corresponding target health threshold, a second control signal is generated.

6. The digital twin driven substation equipment lifecycle intelligent management method of claim 3, wherein, After multiplying the current prediction step size of each health status prediction submodule by the corresponding prediction step size adjustment coefficient to obtain the adjusted prediction step size of each health status prediction submodule, the method further includes the following steps: Obtain the device component identifiers corresponding to each health status prediction submodule with adjusted prediction step size; Based on the preset substation operation logic rule base, query the operation logic relationship between the identifiers of each equipment component; Based on the operational logic relationship, the adjusted prediction step size of logically related device components is synchronously corrected so that the prediction step size between logically related device components meets the constraints defined by the operational logic relationship.

7. The digital twin driven substation equipment lifecycle intelligent management method of claim 6, wherein, The step of synchronously correcting the prediction step size of logically related device components according to the operational logic relationship includes the following steps: When the operating logic relationship is a primary / backup relationship, the primary device and the backup device are identified, and the prediction step size of the backup device is adjusted to be the same as the prediction step size of the primary device. When the operating logic relationship is an interlocking relationship, identify all devices in the interlocking group and adjust the prediction step size of all devices in the interlocking group to the prediction step size of the device with the shortest prediction step size in the interlocking group. When the running logic relationship is an operation timing relationship, identify the preceding and following devices in the timing chain, and adjust the prediction step size of the following device to be no less than the prediction step size of the preceding device. When the operational logic relationship is a capacity constraint relationship, all devices in the capacity constraint group are identified. When the prediction step size of any device in the capacity constraint group is shortened, the prediction step size of all devices in the capacity constraint group is shortened synchronously.

8. The digital twin driven substation equipment lifecycle intelligent management method of claim 5, wherein, After generating the second control signal, the method further includes the following steps: Obtain the device component identifier corresponding to the second control signal, and query the available maintenance window for that device component in the preset maintenance window cycle table; The step of generating and outputting maintenance instructions based on the second control signal includes the following steps: If an available maintenance window is found, the second control signal is bound to the available maintenance window to generate a maintenance instruction with execution time constraints.

9. The digital twin driven substation equipment lifecycle intelligent management method of claim 8, wherein, After querying the available maintenance windows for the device component within the preset maintenance window cycle table, the process further includes the following steps: If no available maintenance window is found, the maintenance priority level of the device component is obtained, and the second control signal is inserted into the maintenance waiting queue according to the maintenance priority level. When there are multiple second control signals in the maintenance waiting queue, the execution order in the maintenance waiting queue is reordered according to the operational logic relationship between the equipment components corresponding to each second control signal, so that the maintenance instructions of equipment components with interlocking relationships or operation timing relationships are executed in the order defined by the operational logic relationship.

10. The digital twin driven substation equipment lifecycle intelligent management method of claim 5, wherein, After obtaining the target health threshold, the following steps are also included: Obtain the historical health status prediction value sequence of this health status prediction submodule before the generation of pseudo-fluctuation identifier, and fit it to obtain the historical degradation trend of health status. Based on the historical degradation trend, estimate the estimated time from the current moment when the predicted health status value will reach the target health threshold; Obtain historical fault data of the equipment component corresponding to the health status prediction submodule, and calculate the historical average interval between when the equipment component reaches the health threshold and when a fault actually occurs. Calculate the total risk exposure duration of the device component based on the expected arrival time and the historical average interval. Obtain the preset maintenance cycle for the equipment component, and compare the total risk exposure time with the preset maintenance cycle: If the total risk exposure duration is less than or equal to the preset maintenance period, the target health threshold is maintained unchanged; If the total risk exposure duration is greater than the preset maintenance period, the target health threshold is lowered until the total risk exposure duration meets the condition of being less than or equal to the preset maintenance period.

Images