A life prediction method of a modular electric energy metering box
By combining a digital twin constraint engine and a transfer learning engine, the coupling aging effect between modules in a modular power metering box is quantified, which solves the problem of rapid adaptation of the prediction model after module replacement and improves the accuracy of life prediction and operation and maintenance efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CETSDEC CO LTD
- Filing Date
- 2026-03-03
- Publication Date
- 2026-06-05
Smart Images
Figure CN122154452A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of power equipment condition monitoring and predictive maintenance technology, and more specifically, to a method for predicting the lifespan of a modular power metering box. Background Technology
[0002] Modular energy metering boxes, as key terminal equipment in smart grids, directly impact the quality of electricity data collection due to their reliability. Currently, the industry generally adopts a functional modular design architecture, decoupling subsystems such as metering, communication, and power conversion into independently replaceable modular units. However, while this design improves maintenance convenience, it also presents technical challenges in lifespan prediction.
[0003] Traditional lifespan prediction methods have significant limitations. First, existing technologies typically evaluate each functional module in isolation, relying solely on the module's own aging parameters, such as the resistance drift of the sampling resistor in the metering module or the signal packet loss rate of the communication module, to estimate lifespan, completely ignoring the physical coupling effects between modules. In reality, capacitor aging in the power conversion module can cause output voltage fluctuations, and this electrical interference accelerates the performance degradation of precision components in the metering module. Similarly, the temperature gradient changes within the enclosure caused by the decreased efficiency of the heat dissipation module will simultaneously exacerbate the thermal aging rate of the RF chip in the communication module and the ADC converter in the metering module. This coupled aging effect becomes a "blind spot" in existing technologies, leading to a systematic deviation between the predicted results and actual failure conditions.
[0004] Secondly, the dynamic maintenance characteristics brought about by modular design have not been effectively addressed. When a faulty module is replaced, the new module forms a new aging interaction system with the original module group, but traditional static prediction models cannot quickly adapt to this architectural change. For example, matching the newly installed power module with the original heat dissipation module requires re-establishing an aging association model, and existing technologies lack effective knowledge transfer mechanisms, resulting in inaccurate assessments of the overall system health after replacement.
[0005] These problems directly impact the efficiency of power grid operation and maintenance: on the one hand, sudden cascading failures due to unidentified coupling risks occur frequently; on the other hand, after module replacement, a long period of time is required to re-accumulate operational data to obtain reliable predictions, severely restricting the effectiveness of predictive maintenance. Currently, there is an urgent need to establish a lifespan prediction system capable of quantifying the aging interactions between modules and possessing dynamic adaptability. Existing technologies urgently need improvement to address these issues. Summary of the Invention
[0006] (a) Technical problems to be solved The purpose of this application is to provide a method for predicting the lifespan of a modular power metering box, which has the technical advantages of accurately quantifying the coupling aging effect between modules and realizing dynamic adaptive maintenance.
[0007] (II) Technical Solution This application provides a method for predicting the lifespan of a modular energy metering box. The technical solution is as follows: 1. Obtain module ontology parameters and cross-module coupling parameters for multiple functional modules, where module ontology parameters characterize the aging state of the module itself, and cross-module coupling parameters characterize the mutual influence between modules; 2. Based on the module ontology parameters and cross-module coupling parameters, calculate the coupling aging factor between multiple functional modules using a digital twin constraint engine; 3. Based on the module ontology parameters and coupling aging factors, predict the remaining lifespan of multiple functional modules and the overall health of the energy metering box using a transfer learning engine; 4. Based on the predicted remaining lifespan and overall health, generate and output operation and maintenance decision information.
[0008] Furthermore, this application also proposes that obtaining the module body parameters and cross-module coupling parameters of multiple functional modules includes: collecting the module body parameters and cross-module coupling parameters through sensing units configured on each functional module, wherein the sensing units are connected to the power metering box system through a standardized interface.
[0009] Furthermore, this application proposes that the functional modules include at least two of the following: an energy metering module, a power conversion module, a communication module, and a heat dissipation module; wherein, the module body parameters include at least one of the following: sampling resistor value drift, MCU operating temperature, output ripple coefficient, capacitor ESR value, RF chip power consumption, signal packet loss rate, fan speed, and heat sink temperature difference; and the cross-module coupling parameters include at least one of the following: input voltage fluctuation amplitude, output signal noise value, output voltage deviation, load regulation rate, transmit power fluctuation, heat sink temperature, temperature gradient of each area inside the box, and heat conduction efficiency.
[0010] Furthermore, this application proposes that calculating the coupling aging factor through a digital twin constraint engine includes: constructing a digital twin for each functional module, the digital twin mapping the aging-sensitive parameters of the components in the module; establishing a coupling model between modules, the coupling model describing the electrical or thermal conduction relationships between modules based on physical equations; and calculating the coupling aging factor through the coupling model based on the module's ontological parameters and cross-module coupling parameters, wherein the coupling aging factor is a value between 0 and 1, and the larger the value, the stronger the coupling aging effect.
[0011] Furthermore, this application proposes a method for predicting remaining lifespan and overall health using a transfer learning engine, including: using an LSTM neural network as the base model to train a module aging baseline; when a module in the energy metering box is replaced, employing a transfer learning strategy to transfer the aging data experience of the old module to the new module, achieving rapid adaptation by fine-tuning the LSTM neural network; and predicting the remaining lifespan of each functional module and the overall health of the energy metering box based on module ontology parameters and coupling aging factors.
[0012] Furthermore, this application also proposes a transfer learning strategy that includes: when a new module is connected, the coupling aging factor and lifetime decay data of the old module are used as source domain knowledge, and adaptation is achieved by fine-tuning the fully connected layer of the LSTM neural network, wherein the adaptation time is less than 15 minutes.
[0013] Furthermore, this application also proposes that the output operation and maintenance decision information includes at least one of the following: lifespan visualization information, used to display the remaining lifespan and overall health of each module, and to mark high-coupling-risk module pairs; dynamic maintenance suggestions, adjusting the maintenance cycle based on the prediction results after module replacement; and fault warning, triggering a linkage warning when the predicted remaining lifespan is lower than a first threshold and the coupling aging factor is higher than a second threshold.
[0014] Furthermore, this application also proposes that the method further includes: achieving data-driven prediction under physical constraints through the collaborative interaction of a digital twin constraint engine and a transfer learning engine, wherein the digital twin constraint engine outputs physical constraint conditions to the transfer learning engine, and the transfer learning engine feeds back data deviation signals to the digital twin constraint engine to trigger model correction.
[0015] Furthermore, this application also proposes a computing device comprising: at least one processor; and a memory storing instructions that, when executed by the at least one processor, cause the at least one processor to perform the aforementioned lifespan prediction method for a modular power metering box.
[0016] Furthermore, this application also proposes a non-transitory machine-readable storage medium storing executable instructions that, when executed, cause the machine to perform the aforementioned lifespan prediction method for modular energy metering boxes.
[0017] (III) Beneficial Effects Compared with the prior art, the beneficial effects of the present invention are as follows: This invention quantifies the physical coupling effect between modules through a digital twin constraint engine and achieves dynamic adaptation by combining it with a transfer learning engine. This effectively solves the problems of neglecting coupling aging and model inaccuracy after maintenance in traditional methods, and has the technical advantages of improving prediction accuracy and maintenance efficiency.
[0018] This application can accurately assess the aging effects of coupling between modules, avoiding misjudgments of overall lifespan due to isolated predictions. In module replacement scenarios, transfer learning is used to quickly adapt to the characteristics of new modules, maintaining the effectiveness of the prediction model. The generated operation and maintenance decision information can identify combinations of modules with high coupling risks, guiding maintenance personnel to prioritize the handling of related aging issues, thereby improving the operation and maintenance efficiency and reliability of metering boxes. Attached Figure Description
[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0020] Figure 1 This is a schematic diagram of the logical structure of a life prediction method for modular energy metering boxes. Detailed Implementation
[0021] The technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application. It should be noted that similar reference numerals and letters in the following drawings indicate similar items; therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Furthermore, in the description of this application, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0022] In existing technologies, modular energy metering boxes achieve flexible maintenance by combining different functional modules, but traditional lifespan prediction methods have significant drawbacks. Existing technologies typically assess the aging status of each functional module independently, without considering the interactions between modules. For example, capacitor aging in the power conversion module may cause output voltage fluctuations, thereby accelerating the wear and tear of the MCU chip in the metering module; temperature rise caused by heat dissipation module failure may simultaneously exacerbate the aging of both the communication and metering modules. Furthermore, after module replacement, the original prediction model cannot quickly adapt to the mixed architecture of old and new modules, leading to increased prediction bias. These limitations result in insufficient overall lifespan prediction accuracy and high operation and maintenance costs.
[0023] To address the aforementioned issues, the inventors discovered that the aging effect of inter-module coupling is the core reason for insufficient prediction accuracy and recognized the need to establish a quantitative evaluation mechanism. Simultaneously, the requirement for dynamic adaptation after module replacement spurred the exploration of rapid model adjustment methods. By analyzing the relationship between physical correlation and data-driven approaches, they proposed combining digital twin models with transfer learning to construct a dual-engine collaborative mechanism. The digital twin model is used to quantify the inter-module coupling effect, while transfer learning enables rapid adaptation in mixed scenarios of old and new modules, forming a closed-loop optimization based on physical rules and data-driven approaches.
[0024] Example 1
[0025] Therefore, this application proposes a method for predicting the lifespan of a modular energy metering box, which includes multiple functional modules. The method includes the following steps: S100. Obtain the module body parameters and cross-module coupling parameters of multiple functional modules, where the module body parameters represent the aging state of the module itself, and the cross-module coupling parameters represent the mutual influence between modules. Among them, module-level parameters refer to indicators reflecting the aging state of the functional module itself. Specifically, these can be achieved by sampling resistor drift, MCU operating temperature, or capacitor ESR value, and are used to directly characterize the aging degree of the components inside the module. Cross-module coupling parameters refer to indicators reflecting the mutual influence between different functional modules. Specifically, these can be achieved by input voltage fluctuation amplitude, internal temperature gradient, or thermal conductivity, and are used to quantify the electrical or thermal correlation effects between modules.
[0026] In this step, parameter acquisition is accomplished through sensing units configured on each functional module. The selection and deployment location of the sensing units must match the functional characteristics of the module: the output ripple coefficient of the power module is achieved using a ripple sensor, which is deployed at the output end of the power module near the connection line of the metering module; the input voltage fluctuation amplitude of the metering module is achieved using a Hall voltage sensor, which is deployed at the sampling resistor input end of the metering module; and the thermal conductivity of the heat dissipation module is achieved using a temperature sensor, which is deployed at the air inlet and outlet of the heat sink and on the surface of the RF chip of the communication module.
[0027] Specifically, all sensing units are connected to the system through the RS485 Modbus RTU standardized interface. The interface pins are defined as pin 1 VCC (5V), pin 2 GND, pin 3 A, and pin 4 B. The data transmission rate is set to 9600bps, the parity is even parity, and the stop bit is 1 bit. It is compatible with a maximum of 32 sensing nodes networking simultaneously.
[0028] The parameter acquisition frequency is differentiated by type: for module body parameters, the output ripple coefficient is measured once every 0.5 seconds, the capacitor ESR value is measured once every 5 minutes, and the fan speed is measured once every 1 second; for cross-module coupling parameters, the input voltage fluctuation amplitude is measured once every 0.5 seconds, and the internal temperature gradient is measured once every 1 second; the measurement accuracy threshold for all parameters is ±0.1%. When the measurement error exceeds this threshold, the sensing unit automatically triggers re-acquisition.
[0029] S200: Based on module ontology parameters and cross-module coupling parameters, the coupling aging factor between multiple functional modules is calculated through a digital twin constraint engine. A digital twin constraint engine is a computational model that constructs the coupling relationship between modules based on physical equations. Specifically, it can be implemented using electrical network equations or heat conduction equations, and is used to transform cross-module parameters into coupling aging factors in the range of 0-1.
[0030] When calculating the coupling aging factor, this engine needs to establish specific physical equations and specify parameter values for different coupling types: the thermal coupling model uses the differential form of Fourier's law. ,in The thermal conductivity coefficient of the heat sink material is used; specifically, for aluminum alloy, it is taken as 205 W / (m・K). The contact area between the heat sink and the communication module is 0.02 m² in this embodiment, measured according to the module design drawings. The temperature gradient between the heat sink and the communication module is calculated by dividing the temperature difference collected by the sensor by the contact distance, which is 0.01m.
[0031] The electrical coupling model uses the AC circuit ripple effect equation. ,in Let be the ripple sensitivity coefficient of the sampling resistor, and take the metal film resistor as 0.001Ω / (V・h). The output ripple coefficient of the power module. For duration. The results of the physical equation calculations are transformed into a 0-1 coupled aging factor through linear normalization, the normalization formula being: ,in Calculated values for physical equations, such as thermal conductivity. Change in resistance , This is the minimum value when the equipment is operating normally, for example... , , For example, the critical value for equipment failure , ), and This was determined through statistical analysis of failures in a certain number of similar measuring boxes.
[0032] S300, based on module ontology parameters and coupling aging factors, uses a transfer learning engine to predict the remaining lifespan of multiple functional modules and the overall health of the power metering box. A transfer learning engine is a machine learning framework that supports rapid model adaptation. Specifically, it can be implemented by fine-tuning the fully connected layers of a pre-trained LSTM neural network, which is used to inherit historical aging patterns and adapt to the characteristics of new modules after module replacement.
[0033] Specifically, the basic LSTM model structure of this engine needs to be clearly defined: the number of input layer neurons = the number of module ontology parameters + the number of cross-module coupling parameters. For example, the number of module ontology parameters is 4, including output ripple coefficient, capacitor ESR value, fan speed, and heat sink temperature difference; the number of cross-module coupling parameters is 2, including input voltage fluctuation amplitude and heat conduction efficiency. In this case, the number of input layer neurons is 6, the number of hidden layers is 2, the number of hidden layers is 64 neurons per layer, and the activation function is tanh; the number of fully connected layers is 1, the number of neurons is 2, and the outputs are remaining lifetime and coupling aging factor, respectively, and the activation function is Linear; the loss function is mean squared error (MSE).
[0034] When a module in the electricity metering box is replaced, the transfer learning strategy must be executed according to the following rules: Source domain data filtering: Only filter historical data whose function type is consistent with the new module and whose coupled object is consistent with the current system. For example, if the new module is a 12V power supply module, then filter all old 12V power supply modules. The amount of data after filtering is no less than 1,000 records. The old power supply module was once operated in combination with the same model of metering module and heat dissipation module.
[0035] Model fine-tuning: Freeze the LSTM hidden layer parameters and adjust only the weights of the fully connected layers. In this embodiment, the fine-tuning parameters are set to a learning rate of 0.001, 50 iterations, Adam optimizer, and batch size of 32 to ensure that the adaptation time is less than 15 minutes.
[0036] Finally, based on the module's intrinsic parameters and coupling aging factor, the remaining lifespan of each functional module and the overall health of the power metering box are predicted through a fine-tuned model.
[0037] S400 generates and outputs operation and maintenance decision information based on predicted remaining lifespan and overall health.
[0038] Specifically, this method first collects module-specific parameters and cross-module coupling parameters through a standardized interface. For example, the output ripple coefficient of the power supply module is used as a module-specific parameter, while the resulting input voltage fluctuation amplitude of the metering module is used as a cross-module parameter.
[0039] The digital twin constraint engine establishes a coupling model between modules based on physical equations. For example, it calculates the impact of heat dissipation module failure on the communication module temperature using the heat conduction equation, and outputs a quantified coupling aging factor. For instance, when a heat dissipation module fan failure leads to a decrease in heat conduction efficiency, the sensor detects a 15K temperature difference between the heatsink outlet temperature and the communication module's RF chip temperature, with a contact distance of 0.01m. Substituting this into Fourier's law... Then, through the normalization formula , here The critical thermal conduction power for the fault. Based on the power at the extreme failure state, the coupling aging factor between the heat dissipation module and the communication module is determined to be 0.95, indicating an extremely high coupling risk.
[0040] The transfer learning engine combines coupling factors with ontology parameters, uses a pre-trained neural network to predict the remaining lifespan of each module, and achieves rapid adaptation by fine-tuning the network layers when modules are replaced. For example, after replacing a new 12V power module, 1200 data points of old 12V power modules that have historically operated in combination with the same model of metering and heat dissipation modules are selected. The LSTM hidden layers are frozen, and after 50 iterations of fine-tuning the fully connected layers, the model adaptation is completed in 12 minutes. The real-time output ripple coefficient of the new power module is 0.3%, the capacitor ESR value is 0.5Ω, and the coupling aging factor is 0.8. Its remaining lifespan is predicted to be 1.7 years, and the overall health of the power metering box is 82 points (out of 100).
[0041] The final generated operation and maintenance decision information includes annotations of highly coupled risk modules and dynamic maintenance suggestions, forming a closed loop from prediction to execution.
[0042] Compared to existing technologies, traditional methods assess the aging status of individual modules in isolation, failing to quantify the accelerated wear effect of power module aging on metering modules. This solution transforms implicit cross-module effects into explicit coupling factors through a digital twin model, for example, incorporating the correlation between decreased heat dissipation efficiency and increased power consumption of the communication module into the calculation. Furthermore, while existing technologies require model retraining after module replacement, this solution inherits historical data experience through transfer learning, completing adaptation within 15 minutes and avoiding prediction interruptions.
[0043] Through the above technical solution, this application can accurately assess the coupling aging effect between modules, avoiding misjudgments of overall lifespan due to isolated predictions. In module replacement scenarios, transfer learning is used to quickly adapt to the characteristics of new modules, maintaining the effectiveness of the prediction model. The generated operation and maintenance decision information can identify combinations of modules with high coupling risks, guiding maintenance personnel to prioritize the handling of related aging issues, thereby improving the operation and maintenance efficiency and reliability of the metering box.
[0044] This application further proposes a system that collects module body parameters and cross-module coupling parameters by means of sensing units configured on each functional module, wherein the sensing units are connected to the power metering box through a standardized interface.
[0045] Among them, the sensing unit refers to the miniaturized and modular design of the sensor component, which can be implemented by using a surface-mount temperature sensor or a Hall current sensor. Its size is adapted to the plug-in structure of the functional module, and it is enabled or disabled synchronously with the installation or removal of the module, in order to solve the problem of flexibility in sensor deployment and maintenance in modular scenarios.
[0046] A standardized interface refers to a data access port with unified physical specifications and communication protocols. Specifically, it can be implemented using RS485 bus or Modbus protocol to eliminate the differences in heterogeneous sensor data formats between different functional modules and ensure the compatibility and continuity of cross-module parameter acquisition.
[0047] Specifically, the sensing unit of each functional module is configured to simultaneously collect two types of parameters: module-specific parameters to characterize its own aging state, such as the resistance drift of the sampling resistor in the power metering module; and cross-module coupling parameters to characterize the impact on associated modules, such as the sensing unit of the power conversion module directionally collecting the input voltage fluctuation amplitude of the communication module.
[0048] The sensing unit transmits the collected data to the system in real time through a standardized interface. For example, it uses the RS485 bus protocol to synchronously upload the fan speed of the heat dissipation module and the heat sink temperature of the communication module. When a functional module is replaced, the sensing unit of the new module automatically connects to the system through the same standardized interface, without the need to reconfigure the sensor parameters, ensuring that the acquisition of coupling parameters between modules is not affected by module plugging and unplugging operations.
[0049] Compared to existing technologies, traditional solutions use fixed, universal sensors to collect device parameters, which cannot adapt to the pluggable characteristics of modular structures and do not specifically collect coupling parameters for inter-module interactions. This solution, through a modular sensing unit design, allows sensors to be replaced synchronously with functional modules, enabling targeted collection of relevant parameters from target modules. For example, the sensing unit of a heat dissipation module only collects the temperature of the heat sink of adjacent modules, rather than the overall temperature data of the entire enclosure, thus avoiding redundant data transmission and improving the effectiveness of coupling parameters.
[0050] Through the above technical solution, this application eliminates the data fragmentation problem caused by the fixed deployment of sensors in modular power metering boxes, dynamically captures the real-time interaction effects between specific modules, such as the voltage disturbance effect of power module aging on metering module being accurately collected; the standardized interface design enables seamless access to the system of sensor data from replacement modules produced by different manufacturers, and there is no need to recalibrate the sensors after module replacement, significantly reducing the complexity of operation and maintenance.
[0051] This application further proposes functional modules including at least two of the following: an energy metering module, a power conversion module, a communication module, and a heat dissipation module; module body parameters including at least one of the following: sampling resistor value drift, MCU operating temperature, output ripple coefficient, capacitor ESR value, RF chip power consumption, signal packet loss rate, fan speed, and heat sink temperature difference; and cross-module coupling parameters including at least one of the following: input voltage fluctuation amplitude, output signal noise value, output voltage deviation, load regulation rate, transmit power fluctuation, heat sink temperature, temperature gradient of each area inside the box, and heat conduction efficiency.
[0052] Among them, the power metering module refers to the core unit used to measure power consumption. Specifically, it can be implemented using a circuit board with a sampling resistor and an MCU. The drift of the sampling resistor value in its main parameters is directly related to the attenuation of metering accuracy.
[0053] A power conversion module is a functional unit that converts input voltage into a stable output. Specifically, it can be implemented using a circuit with electrolytic capacitors and voltage regulator chips. The ESR value of the capacitor in its main parameters reflects the degree of electrolyte drying.
[0054] The communication module is a functional unit that enables wireless data transmission. It can be implemented using an RF chip and an antenna module, with the RF chip power consumption representing signal transmission efficiency. The heat dissipation module is a functional unit that maintains temperature balance within the enclosure. It can be implemented using a combination of a fan and heat sink, with the fan speed reflecting active cooling capability.
[0055] The input voltage fluctuation amplitude in the cross-module coupling parameters refers to the voltage change amplitude from the power supply module output to the metering module. This can be acquired using a voltage sensor to quantify the impact of power supply module aging on the metering module. Thermal conductivity efficiency refers to the heat transfer efficiency between the heat dissipation module and the communication module. This can be measured using a temperature gradient sensor to quantify the impact of heat dissipation failure on the communication module.
[0056] Specifically, this solution constructs an aging monitoring network covering the core functional chain by defining the types of functional modules and their parameter systems. The combination of the power metering module, power conversion module, communication module, and heat dissipation module covers the four major functional units of the metering box: electrical metering, energy conversion, data transmission, and thermal management, ensuring that parameter acquisition covers typical interaction scenarios. The module parameters are designed for the failure paths of the core functions of each module. For example, the resistance drift of the sampling resistor in the power metering module directly reflects the degradation of metering accuracy; the ESR value of the capacitor in the power conversion module reveals the aging degree of the electrolytic capacitor; the signal packet loss rate of the communication module characterizes the reliability of data transmission; and the temperature difference of the heat sink in the heat dissipation module quantifies the heat dissipation efficiency.
[0057] Cross-module coupling parameters map the interactions between modules. For example, the output voltage deviation of the power conversion module is transmitted to the energy metering module through the input voltage fluctuation amplitude parameter, and the heat conduction efficiency of the heat dissipation module affects the communication module through the heat sink temperature parameter. These parameters are collected in real time by standardized sensing units, providing directional correlation data for subsequent calculation of coupling aging factors, avoiding modeling distortion caused by ambiguous parameter correlations in traditional methods.
[0058] Compared to existing technologies, traditional methods typically use single-type parameters or fail to differentiate module functions for data acquisition, such as monitoring only general parameters like voltage and current, thus failing to capture the differences in aging characteristics among different functional units within a modular metering box. Existing base station monitoring schemes do not define parameters according to module type, making it impossible to distinguish the impact paths of power module aging from communication module aging. This solution, through module classification and parameter-oriented correlation design, achieves for the first time the systematic acquisition of multi-dimensional aging data, including electrical and thermodynamic performance, from modular metering boxes, providing a data foundation for quantifying coupling effects.
[0059] Through the above technical solutions, this application solves the problems of incomplete parameter acquisition and difficulty in quantifying cross-module effects in modular energy metering boxes. By limiting the types of functional modules and defining their core sensitive parameters, it ensures coverage of the critical aging paths of each functional unit in the metering box; by designing cross-module coupling parameters in a targeted manner, it transforms the implicit inter-module effects into measurable physical quantities, providing data support for accurately assessing the coupled aging effects.
[0060] This application further proposes a lifespan prediction method for modular power metering boxes, wherein the calculation of the coupling aging factor through a digital twin constraint engine includes: constructing a digital twin for each functional module, the digital twin mapping the aging-sensitive parameters of the components in the module; establishing a coupling model between modules, the coupling model describing the electrical or thermal conduction relationships between modules based on physical equations; and calculating the coupling aging factor through the coupling model based on the module body parameters and cross-module coupling parameters, wherein the coupling aging factor is a value between 0 and 1, and the larger the value, the stronger the coupling aging effect.
[0061] The digital twin refers to a dynamic virtual mapping of the aging state of components within a functional module. This can be constructed using parametric modeling tools, such as importing initial values of aging-sensitive parameters from component specifications and dynamically updating them in real-time by receiving module-level parameters collected by sensing units. This feature reflects the microscopic aging state of components in real time, providing accurate input for the coupling model. The coupling model is a mathematical model of inter-module interactions based on physical laws, which can be implemented using circuit equations or heat conduction equations. For example, Ohm's law can be used to describe the impact of power module output voltage fluctuations on the resistance aging of the metering module. This feature transforms the implicit correlations between modules into calculable mathematical relationships, ensuring the physical rationality of the coupling effect analysis. The coupling aging factor is an index that quantifies the aging interactions between modules. This can be derived from physical equations, such as normalizing the correlation between voltage fluctuation amplitude and resistance aging rate to a value between 0 and 1. This feature provides quantifiable parameters of coupling effect strength for subsequent lifetime prediction.
[0062] Specifically, when calculating the coupling aging factor, a digital twin is first generated for each functional module. This twin continuously receives real-time monitoring data from the sensing unit, such as hourly updates to the equivalent series resistance of the capacitor or the temperature parameters of the heat sink. Subsequently, based on the physical correlation characteristics between modules, a coupling model including electrical or thermodynamic equations is established. For example, for the electrical coupling between the power supply module and the metering module, Kirchhoff's voltage law is used to construct an equation illustrating the impact of voltage fluctuations on the aging of the sampling resistor.
[0063] To address the thermal conduction correlation between the heat dissipation module and the communication module, Fourier's law is used to establish the relationship between thermal conduction efficiency and chip temperature rise. By inputting real-time acquired module body parameters and cross-module coupling parameters into the coupling model, physical equations are calculated, and a normalized coupling aging factor is finally output. In this process, the dynamic parameter updates of the digital twin and the quantitative derivation of the physical equations form a closed loop, ensuring the accuracy and real-time performance of the coupling effect analysis.
[0064] Compared to existing technologies, traditional methods typically rely on empirical formulas to estimate the aging correlation between modules, such as statistically analyzing the relationship between temperature rise and failure rate based solely on historical data, without establishing a quantitative model based on physical laws. While existing technology CN120358531A applies digital twin technology, it is only used for equipment status recording and does not involve physical modeling of the coupling effects between modules. This solution, by combining a dynamic digital twin with a physical coupling model, achieves, for the first time, a quantitative analysis of the aging interactions between modules, resolving the subjective bias problem of traditional empirical assessments.
[0065] Through the above technical solution, this application can accurately quantify the intensity of the coupling aging effect between modules. For example, it can accurately identify the impact of power module aging on the lifespan of the metering module, avoiding prediction errors caused by isolated evaluation. At the same time, the coupling model based on physical equations ensures that the aging factor calculation conforms to actual physical laws, preventing unreasonable prediction results that violate thermodynamics or circuit theory, and significantly improving the reliability of the overall lifespan prediction of modular energy metering boxes.
[0066] This application further proposes a method for predicting remaining lifespan and overall health using a transfer learning engine, including using an LSTM neural network as the base model to train the module aging baseline; when a module in the energy metering box is replaced, a transfer learning strategy is used to transfer the aging data experience of the old module to the new module, and rapid adaptation is achieved by fine-tuning the LSTM neural network; the remaining lifespan of each functional module and the overall health of the energy metering box are predicted based on the module ontology parameters and coupling aging factors.
[0067] Among them, LSTM neural network refers to recurrent neural network with long short-term memory capability. Specifically, it can be implemented using a network structure that includes an input layer, hidden layers and fully connected layers. It captures the nonlinear degradation characteristics during the aging process of modules through time-series data processing.
[0068] Transfer learning strategies refer to using the coupling aging factors and lifetime decay data of old modules as source domain knowledge for model fine-tuning. Specifically, this can be achieved by freezing the hidden layer parameters of the LSTM and adjusting only the weights of the fully connected layers, allowing the new module to quickly inherit the coupling aging association experience between the old module and other modules. Module ontology parameters refer to physical quantities characterizing the aging state of the module itself. These can be implemented using sensor data such as sampling resistor drift, capacitor ESR values, or fan speed, reflecting the aging degree of the internal components of the module.
[0069] The coupling aging factor is a parameter that quantifies the degree of mutual influence between modules. Specifically, it can be implemented using a value between 0 and 1 calculated based on a digital twin constraint engine, which is used to describe the intensity of cross-module aging effects.
[0070] Specifically, this method first uses an LSTM neural network to train on historical aging data to establish a baseline model of the degradation patterns of each module. When a module replacement event is detected, the transfer learning engine extracts the coupling aging factors and corresponding lifetime decay trajectories between the old module and other modules from the old module's operational data, and inputs this as source domain knowledge into the LSTM network. By preserving the temporal feature extraction capability of the network's hidden layers, only the parameters of the fully connected layers are fine-tuned, enabling the new module to quickly adapt to the coupling environment of the existing system. During the prediction process, the real-time ontology parameters of the new module and the coupling aging factors of the current system are input into the fine-tuned model, and the remaining lifetime of the module and the overall health indicators affected by it are calculated simultaneously.
[0071] Compared to existing technologies, traditional methods require the re-collection of complete lifecycle data to train new models after module replacement, failing to utilize historical coupling and correlation information. This solution, however, employs a transfer learning strategy to directionally transfer the coupling experience of old modules to new modules, achieving rapid adaptation while preserving the original system-level aging patterns. Existing transfer learning methods only perform data transfer on independent modules, neglecting the coupling factors between modules, leading to prediction results deviating from actual operating conditions.
[0072] Through the above technical solution, this application can quickly establish a coupling aging correlation model between the new module and other parts of the system after module replacement, effectively reducing prediction bias caused by the mixed architecture of old and new modules. This method avoids the large amount of new data collection required for traditional full model reconstruction by targeted migration of coupled aging experience data, while ensuring that the prediction model continuously reflects the dynamic interaction between modules, improving the timeliness and accuracy of operation and maintenance decisions.
[0073] This application further proposes that when a new module is connected, the coupling aging factor and lifetime decay data of the old module are used as source domain knowledge, and adaptation is achieved by fine-tuning the fully connected layer of the LSTM neural network, wherein the adaptation time is less than 15 minutes.
[0074] Among them, the coupling aging factor is a quantitative indicator that characterizes the degree of aging caused by the mutual influence between modules. Specifically, it can be calculated using the correlation model between cross-module coupling parameters and module body parameters, and is used to reflect the transmission strength of aging effects between modules.
[0075] Lifetime degradation data refers to the performance degradation record of a module over time during operation. Specifically, it can be constructed by collecting historical change trends of the module's parameters through sensors, and is used to describe the aging pattern of the module.
[0076] Source domain knowledge refers to the historical data set of old modules that have the same functional type and coupled objects as the new module. Specifically, data filtering algorithms can be used to match the coupling aging factor and lifetime decay data of modules of the same type to retain prior information on coupling effects related to the new module.
[0077] Fine-tuning the fully connected layer refers to adjusting only the parameters of the terminal network layer responsible for feature mapping in the LSTM neural network. Specifically, this can be achieved by freezing the hidden layer parameters and updating the weights of the fully connected layer. This is used to inherit the general temporal aging rules while adapting to the coupling characteristics of the new module.
[0078] The adaptation time refers to the upper limit of the time required to adjust the model parameters. Specifically, it can be achieved by limiting the size of the trainable parameters of the fully connected layer and the number of iterations of the optimization algorithm, so as to ensure that the model update meets the real-time requirements of the operation and maintenance scenario.
[0079] Specifically, when a functional module in the electricity metering box is replaced, the data of the old module with the same functional type and coupled objects as the new module is first selected, and its coupling aging factor and lifespan decay data are extracted as source domain knowledge. Then, keeping the parameters of the hidden layers in the LSTM neural network unchanged, only the weights of the fully connected layers are updated with gradients, and the model is jointly trained using the initial running data of the new module and the source domain knowledge. This process, by limiting the parameter adjustment range and optimization step size of the fully connected layers, ensures that the model completes adaptation within 15 minutes, quickly establishing a coupling aging relationship model between the new module and other modules.
[0080] Compared to existing technologies, traditional methods require retraining the entire prediction model after module replacement, consuming significant computational resources and leading to decreased accuracy due to the lack of source domain data screening. This solution avoids interference from irrelevant data by selectively screening coupled data for similar modules; and by locally fine-tuning the fully connected layers, it accurately adapts to the characteristics of the new module while preserving general aging patterns, effectively balancing model update speed and prediction accuracy.
[0081] Through the above technical solution, this application can quickly establish a life prediction model under the hybrid architecture of old and new modules after module replacement, shorten the model adaptation time and reduce prediction deviation, and ensure the timeliness and accuracy of operation and maintenance decisions.
[0082] This application further proposes that the output operation and maintenance decision information include at least one of the following: lifespan visualization information, used to display the remaining lifespan and overall health of each module, and to mark modules with high coupling risk; dynamic maintenance suggestions, adjusting the maintenance cycle based on the prediction results after module replacement; and fault warning, triggering a linkage warning when the predicted remaining lifespan is lower than a first threshold and the coupling aging factor is higher than a second threshold.
[0083] Among them, lifespan visualization information refers to the visualization data of the remaining lifespan and overall health status of each module displayed through a graphical interface. Specifically, it can be implemented in the form of heat map or topology map. The remaining lifespan range of different modules is distinguished by color gradient, and the coupling aging factor value is marked on the module connection line to intuitively reflect the aging coupling strength between modules.
[0084] Dynamic maintenance recommendations refer to a strategy that dynamically adjusts maintenance cycles based on the predicted remaining lifespan after module replacement. This can be implemented using a weighted calculation model based on the coupling aging factor and remaining lifespan, determining the adjustment range of the maintenance cycle by quantifying the degree of aging impact between modules. Fault warnings are alert signals triggered when a module's remaining lifespan falls below a preset threshold and its coupling aging factor exceeds a critical value. This can be implemented using dual-threshold comparison logic, triggering the warning mechanism by simultaneously monitoring real-time data of remaining lifespan and coupling aging factor.
[0085] Specifically, the lifespan visualization information displays the remaining lifespan and overall health of each module through a graphical interface. For example, the remaining lifespan range is marked with different colors in the topology diagram, and the coupling aging factor value is displayed on the connection line between modules. When the value exceeds the set threshold, it is automatically marked as a high-coupling-risk module pair. For example, when the coupling aging factor of the power module and the metering module reaches 0.7, a red warning sign is displayed.
[0086] Dynamic maintenance recommendations adjust maintenance cycles based on predicted results after module replacement. For example, if the remaining lifespan of a new power module is predicted to be 2 years, but its coupling aging factor with the metering module is 0.8, the maintenance cycle is shortened from the original 2 years to 1.5 years. Fault warnings employ a dual-threshold triggering mechanism. For instance, when the remaining lifespan of a module is less than 30 days and its coupling aging factor exceeds 0.6, a linkage warning signal is triggered and pushed to the operation and maintenance system.
[0087] Compared to existing technologies, current operational decision-making information is generated based solely on the lifespan of a single module or its independent fault state, failing to reflect the coupling aging risks between modules. For example, traditional methods only display the remaining lifespan of each module while ignoring the aging interactions between modules, resulting in fixed maintenance cycles that cannot adapt to the coupling effects after module replacement. This solution addresses the shortcomings of traditional methods in identifying coupling risks and generating dynamic maintenance strategies by introducing coupling aging factor labeling, dynamic maintenance cycle adjustment, and a dual-threshold early warning mechanism.
[0088] Through the above technical solutions, this application achieves multi-dimensional and accurate generation of operation and maintenance decision information: by marking high-coupling risk module pairs, operation and maintenance personnel can quickly locate module combinations with collaborative degradation risks; by dynamically adjusting the maintenance cycle, the problem of maintenance strategy failure after module replacement is avoided; and by using dual-threshold linkage early warning, early warning is triggered in a timely manner when the lifespan of a single module is insufficient and the coupling effect is significant, effectively preventing the occurrence of chain failures.
[0089] This application further proposes to achieve data-driven prediction under physical constraints through the collaborative interaction of a digital twin constraint engine and a transfer learning engine. The digital twin constraint engine outputs physical constraints to the transfer learning engine, and the transfer learning engine feeds back data deviation signals to the digital twin constraint engine to trigger model correction.
[0090] The digital twin constraint engine refers to a simulation model built based on the physical characteristics of a modular energy metering box. Specifically, it can be implemented using multiphysics coupling modeling technology to calculate dynamic constraint parameters between modules in real time. The transfer learning engine refers to a prediction model with cross-module knowledge transfer capabilities. Specifically, it can be implemented using a transfer learning framework based on LSTM neural networks to adjust the prediction boundary according to physical constraints.
[0091] Physical constraints refer to the coupling aging thresholds reflecting the current module combination. These can be implemented using dynamically generated electrical or thermodynamic equation parameters, such as the upper limit of output voltage fluctuation when the power supply module and metering module are coupled. Data deviation signals refer to the difference between the predicted results and the values calculated by the physical model. This can be implemented using a residual calculation module, triggering a correction command when the difference exceeds a preset threshold.
[0092] Specifically, during the operation of the modular energy metering box, the digital twin constraint engine generates dynamic constraint parameters based on the physical relationships between the current module combinations. For example, when there is a heat conduction relationship between the heat dissipation module and the communication module, it outputs a coupling threshold between the heat sink temperature rise rate and the power consumption of the RF chip. The transfer learning engine receives this threshold as a prediction boundary condition to restrict the physical compliance of the neural network output values during training.
[0093] When the difference between the predicted remaining lifetime and the physical model's calculated value exceeds a threshold, the transfer learning engine generates a data deviation signal. This signal carries the module identifier and parameter type of the deviation source. The digital twin constraint engine locates the coupled model parameters that need correction based on the deviation signal. For example, when the communication module's lifetime prediction deviation stems from an inaccurate thermal conductivity coefficient, it only adjusts the coefficient values in the thermal conductivity equation, rather than globally modifying the model parameters. This forms a closed-loop optimization mechanism of "constraint generation - prediction calibration - deviation correction".
[0094] Compared to existing technologies, traditional methods operate the physical model and the data-driven model independently. For example, CN120358531A only uses digital twins for state monitoring without interacting with the prediction model; CN120528095A uses fixed physical equations to constrain the prediction model, making it unable to adapt to parameter changes after module replacement. This solution defines the prediction boundary through dynamic constraint parameters, preventing the data-driven model from producing conclusions contrary to physical laws; it uses a threshold-triggered bias feedback mechanism to initiate corrections only when prediction deviations are significant, reducing invalid computations; and it avoids stability issues caused by global model reconstruction through targeted parameter adjustments.
[0095] Through the above technical solution, this application solves the problem of model bias accumulation caused by the disconnect between physical constraints and data-driven prediction, ensuring that the prediction results always conform to the actual physical laws, and quickly correcting model parameters when modules are replaced or the environment changes, thus maintaining the stability of prediction accuracy.
[0096] Example 2
[0097] This application further proposes a computing device, including at least one processor and a memory, the memory storing instructions that, when executed by at least one processor, cause at least one processor to perform a life prediction method for a modular power metering box.
[0098] The processor, or processor, refers to the core that executes program instructions. It can be implemented using a multi-core CPU or an embedded microcontroller. It parses algorithm instructions from memory and performs data calculations, enabling real-time processing of module parameters, aging factor calculations, and remaining lifetime predictions. The memory, or memory, refers to the physical medium that stores program instructions and data. It can be implemented using a solid-state drive or flash memory chips. It stores the physical equations of the digital twin constraint engine, the neural network model of the transfer learning engine, and historical aging data, providing a knowledge base to support dynamic model adaptation after module replacement.
[0099] Specifically, the processor executes instructions from memory to drive the sensing unit to collect module-specific parameters and cross-module coupling parameters of each functional module in the energy metering box. The digital twin constraint engine constructs electrical and thermal conduction correlation models between modules based on physical equations, and calculates coupling aging factors in real time to quantify the mutual influence between modules. The transfer learning engine trains the module aging baseline based on an LSTM neural network. When a module replacement event is detected, it fine-tunes the fully connected layer to transfer the aging data of the old module to the new module, achieving rapid adaptation of the life prediction model under the hybrid architecture. The memory continuously records real-time data generated during operation and corrected model parameters, forming a closed-loop feedback mechanism to optimize prediction accuracy.
[0100] Compared to existing technologies, traditional computing devices only support independent lifetime prediction algorithms for single modules, failing to address the error accumulation problem caused by inter-module coupling effects. Furthermore, the entire model needs to be retrained after module replacement, leading to response delays. This solution, through a collaborative architecture of processor and memory, integrates physical constraint models with data-driven algorithms, enabling dynamic calculation of coupling aging factors and real-time updates of model parameters at the hardware level. This overcomes the technical bottleneck of prediction system lag after module replacement.
[0101] Through the above technical solution, this application can quantify the coupling aging effect between modules in real time, eliminate the prediction bias caused by isolated evaluation, and complete the prediction model adaptation within 15 minutes after module replacement, ensuring the accuracy of the overall life prediction results under the new and old hybrid architecture, and reducing the risk of unplanned downtime and maintenance costs.
[0102] Example 3
[0103] This application further proposes a non-transitory machine-readable storage medium storing executable instructions that, when executed, cause the machine to perform a life prediction method for a modular energy metering box.
[0104] Non-transitory machine-readable storage media refers to physical carriers with persistent storage capabilities, specifically solid-state drives, flash memory chips, or EEPROMs, used to solidify the execution logic of the lifetime prediction method. Executable instructions refer to machine-readable instruction sets composed of computer program code, specifically compiled binary files or scripting languages, used to drive computing devices to perform parameter acquisition, coupling factor calculation, and lifetime prediction processes. Data-driven prediction under physical constraints refers to physical rule constraints established based on digital twins and coupled models, specifically implemented using finite element equations or thermodynamic equations, to ensure that the data prediction process conforms to actual physical laws. Dynamic adaptation refers to the ability to quickly adjust the prediction model after module replacement, specifically implemented using parameter fine-tuning algorithms in transfer learning, to shorten the model adaptation time after the integration of new modules.
[0105] Specifically, when the instructions stored in the storage medium are executed by the machine, the module-specific parameters and cross-module coupling parameters of each functional module in the energy metering box are first acquired. The module-specific parameters are collected in real time by sensors connected via standardized interfaces. Subsequently, a digital twin of each module is constructed based on a digital twin constraint engine, and the coupling aging factor between modules is calculated through a coupling model. This factor reflects the cumulative aging effect caused by electrical or thermal conduction correlations. The transfer learning engine predicts the remaining lifespan and overall health of each module based on historical aging data and the current coupling factor. When a module is replaced, the coupling aging factor of the old module is used as source domain knowledge input into the transfer learning model, enabling rapid adaptation to the new module by fine-tuning the parameters of the fully connected layer. The digital twin constraint engine and the transfer learning engine continuously optimize prediction accuracy through bidirectional data interaction. Physical constraints ensure that data predictions do not deviate from actual physical laws, and data deviation signals trigger model corrections to maintain prediction reliability.
[0106] Compared to existing technologies, traditional storage media only store static prediction models, failing to handle dynamic adaptation needs arising from module replacements, and lacking physical constraint mechanisms that cause prediction results to deviate from actual operating conditions. In existing technologies using general-purpose storage devices, model updates require redeploying the entire prediction system, increasing operational downtime. In contrast, this solution, by embedding physical constraints and dynamic adaptation logic, enables the storage media to automatically trigger the transfer learning process after module replacement, while continuously verifying the physical rationality of the prediction results through a digital twin.
[0107] Through the above technical solutions, this application solves the problem of lifespan prediction errors caused by coupling aging effects in modular equipment, and achieves rapid adaptation in scenarios where new and old modules operate in combination. The standardized deployment capability of storage media allows the prediction method to be implemented across different hardware platforms, avoiding the repetitive development costs caused by equipment differences in traditional solutions. The synergistic mechanism of physical constraints and data-driven approaches effectively suppresses non-physical prediction results that may arise from purely data-driven methods, ensuring the reliability of operation and maintenance decision-making information. The automatic adaptation function after module replacement reduces the need for manual intervention, shortening the system maintenance cycle.
[0108] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.
Claims
1. A method for predicting the lifespan of a modular energy metering box, wherein the energy metering box comprises multiple functional modules, characterized in that, The method includes: Obtain the module ontology parameters and cross-module coupling parameters of the multiple functional modules, wherein the module ontology parameters characterize the aging state of the module itself, and the cross-module coupling parameters characterize the mutual influence between modules; Based on the module ontology parameters and cross-module coupling parameters, the coupling aging factor between the multiple functional modules is calculated using a digital twin constraint engine; Based on the module ontology parameters and the coupling aging factor, the remaining lifespan of the multiple functional modules and the overall health of the power metering box are predicted using a transfer learning engine. Based on the predicted remaining lifespan and overall health, operation and maintenance decision information is generated and output.
2. The lifespan prediction method for the modular energy metering box according to claim 1, characterized in that, The process of obtaining the module body parameters and cross-module coupling parameters of the multiple functional modules includes: The module body parameters and cross-module coupling parameters are collected by the sensing units configured on each functional module, wherein the sensing units are connected to the system of the power metering box through a standardized interface.
3. The lifespan prediction method for the modular energy metering box according to claim 2, characterized in that, The functional modules include at least two of the following: an energy metering module, a power conversion module, a communication module, and a heat dissipation module. The module body parameters include at least one of the following: sampling resistor value drift, MCU operating temperature, output ripple coefficient, capacitor ESR value, RF chip power consumption, signal packet loss rate, fan speed, and heat sink temperature difference. The cross-module coupling parameters include at least one of the following: input voltage fluctuation amplitude, output signal noise value, output voltage deviation, load regulation rate, transmit power fluctuation, heat sink temperature, temperature gradient of each area inside the box, and heat conduction efficiency.
4. The lifespan prediction method for the modular energy metering box according to claim 1, characterized in that, The calculation of the coupling aging factor through the digital twin constraint engine includes: A digital twin is constructed for each functional module, and the digital twin maps the aging-sensitive parameters of the components in the module; Establish a coupling model between modules, which describes the electrical or thermal conduction relationships between modules based on physical equations; Based on the module ontology parameters and cross-module coupling parameters, the coupling aging factor is calculated through the coupling model, wherein the coupling aging factor is a value between 0 and 1, and the larger the value, the stronger the coupling aging effect.
5. The lifespan prediction method for the modular energy metering box according to claim 1, characterized in that, The prediction of remaining lifespan and overall health using a transfer learning engine includes: Use an LSTM neural network as the base model and train the module to age the baseline; When a module in the power metering box is replaced, a transfer learning strategy is adopted to transfer the aging data experience of the old module to the new module, and the LSTM neural network is fine-tuned to achieve rapid adaptation. Based on the module's intrinsic parameters and coupling aging factor, the remaining lifespan of each functional module and the overall health of the power metering box are predicted.
6. The lifespan prediction method for the modular energy metering box according to claim 5, characterized in that, The transfer learning strategy includes: when a new module is connected, the coupling aging factor and lifetime decay data of the old module are used as source domain knowledge, and adaptation is achieved by fine-tuning the fully connected layer of the LSTM neural network, wherein the adaptation time is less than 15 minutes.
7. The lifespan prediction method for the modular energy metering box according to claim 1, characterized in that, The output operation and maintenance decision information includes at least one of the following: Lifetime visualization information is used to display the remaining lifetime and overall health of each module, and to mark high-coupling-risk module pairs; Dynamic maintenance recommendations adjust maintenance cycles based on predicted results after module replacement; Fault warning: When the predicted remaining lifespan is lower than the first threshold and the coupling aging factor is higher than the second threshold, a linkage warning is triggered.
8. The lifespan prediction method for the modular energy metering box according to claim 1, characterized in that, The method further includes: Through the collaborative interaction of the digital twin constraint engine and the transfer learning engine, data-driven prediction under physical constraints is achieved. The digital twin constraint engine outputs physical constraints to the transfer learning engine, and the transfer learning engine feeds back data deviation signals to the digital twin constraint engine to trigger model correction.
9. A computing device, characterized in that, include: At least one processor; as well as A memory storing instructions that, when executed by the at least one processor, cause the at least one processor to perform the life prediction method for a modular energy metering box as described in any one of claims 1-8.
10. A non-transitory machine-readable storage medium, characterized in that, It stores executable instructions that, when executed, cause the machine to perform the life prediction method for the modular energy metering box as described in any one of claims 1-8.