Box-type substation operation and maintenance management method and system based on digital twinning

Abstract

The invention relates to the technical field of substation operation and maintenance management, in particular to a box-type substation operation and maintenance management method and system based on digital twinning. The method comprises the following steps: initializing a particle swarm, wherein each particle corresponds to a group of parameter sets of the box-type substation; optimizing and iterating a particle swarm by using an improved particle swarm algorithm until a termination condition is met, and obtaining a global optimal parameter set; and applying the global optimal parameter set to the digital twinborn model to realize operation and maintenance management of the transformer substation. According to the scheme of the invention, the operation and maintenance management method which can predict the equipment state and meet the high-reliability operation and maintenance management requirement can be provided.

Description

Technical Field

[0001] This invention relates to the field of substation operation and maintenance management technology. More specifically, this invention relates to a method and system for operation and maintenance management of prefabricated substations based on digital twins. Background Technology

[0002] As a key node in the power distribution network, the prefabricated substation contains core equipment such as transformers and high and low voltage switchgear. During long-term operation, the equipment is affected by thermal stress, electrical stress and environmental factors, and its physical parameters (such as winding DC resistance, insulation thermal conductivity, contact resistance, etc.) will drift nonlinearly.

[0003] To maintain the high fidelity of digital twin models, existing technologies often employ particle swarm optimization (PSO) algorithms and use real-time monitoring data to perform reverse calibration of model parameters.

[0004] However, prefabricated substations have numerous parameters to be identified, forming a complex, high-dimensional, multi-peak solution space. Traditional PSO algorithms, during the optimization process, often suffer from premature convergence due to a lack of population diversity, causing particles to prematurely cluster near local extrema. While such local optima may exhibit small numerical fitting errors at the current moment, their corresponding physical parameter combinations often deviate from the true physical state (e.g., masking actual heat source anomalies with incorrect thermal conductivity). Once the substation's operating conditions change (such as sudden load changes or diurnal temperature variations), the digital twin model based on these local optima will quickly fail, unable to accurately predict equipment status and failing to meet the requirements of high-reliability operation and maintenance management.

[0005] Therefore, there is an urgent need for an operation and maintenance management method that can accurately predict equipment status and meet the requirements of high-reliability operation and maintenance management. Summary of the Invention

[0006] The purpose of this invention is to propose a digital twin-based operation and maintenance management method and system for prefabricated substations, in order to solve the problem that existing operation and maintenance management methods cannot accurately predict equipment status and meet the requirements of high-reliability operation and maintenance management; to this end, this invention provides solutions in the following two aspects.

[0007] In a first aspect, the present invention provides a digital twin-based method for the operation and maintenance management of prefabricated substations, comprising:

[0008] Initialize the particle swarm, with each particle corresponding to a set of parameters for the box-type substation; An improved particle swarm optimization algorithm is used to optimize and iterate the particle swarm until the termination condition is met, thus obtaining the globally optimal parameter set. The globally optimal parameter set is applied to the digital twin model to achieve substation operation and maintenance management; The improved particle swarm optimization algorithm includes adding a chaotic perturbation term based on the Lorentz distribution to the particle update velocity at each iteration, with the dynamic escape probability as the trigger condition. The dynamic escape probability is positively correlated with the steady-state trap risk index at the current iteration; the steady-state trap risk index is negatively correlated with both the aggregation index and the operating condition excitation intensity; the aggregation index represents the aggregation of all particle positions of the particle swarm to the global optimal position at the current iteration; the operating condition excitation intensity is positively correlated with the variance of the pre-obtained load current of the prefabricated substation within a set time window, negatively correlated with the mean of the load current, and positively correlated with the absolute value of the change index of the transformer top oil temperature.

[0009] The above scheme introduces a steady-state trap risk index and a dynamic escape probability, tightly coupling the algorithm's search strategy with the actual operating conditions of the physical system (such as load fluctuations and oil temperature changes). Compared to traditional algorithms that rely solely on mathematical convergence, this scheme can identify pseudo-convergence states caused by insufficient operating condition excitation (data scarcity) and forces particles to escape local extrema through a Lorentz chaotic perturbation mechanism. This ensures that the physical parameters identified by the digital twin model have high fidelity and accuracy under various complex operating environments, achieving precise substation operation and maintenance management.

[0010] Optionally, the method for calculating the clustering index includes: Calculate the root mean square distance between the position vector of each particle and the global optimal position vector at each iteration; use the root mean square distance as the aggregation index.

[0011] The aggregation index in the above scheme can more intuitively reflect whether the group is in an aggregation state, providing accurate data support for judging whether the algorithm has truly found the optimal solution or has merely fallen into stagnation, and is a prerequisite for subsequent judgment of the risk of steady-state trap.

[0012] Optionally, the excitation intensity under the operating condition The calculation method is as follows: ; in, and These represent the standard deviation and mean value of the load current within a set time window, respectively, where I is the load current. To set the change index of transformer top oil temperature within a time window, To prevent the default positive real number with a denominator of zero, To set the length of the time window.

[0013] By combining the load current variance, mean, and top oil temperature change indicators within a set time window, a condition excitation intensity is constructed to quantify the richness of the current data's dynamic characteristics. Compared to identification methods that do not distinguish data quality, this scheme can keenly sense whether external stimuli (such as sudden load changes or drastic temperature changes) are sufficient to excite the system's nonlinear characteristics. This strengthens identification when the data contains rich dynamic information and raises vigilance when the data is stable, avoiding parameter misidentification due to insufficient information.

[0014] Optionally, the steady-state trap risk index The calculation method is as follows: ; in, The normalized scaling coefficient. Let be the aggregation index at the t-th iteration. For the working condition excitation strength, The weighting factor is the influence of operating conditions.

[0015] The aforementioned steady-state trap risk index can accurately identify a specific dangerous state (i.e., a pseudo-steady-state trap) where the population is highly clustered but external operating conditions are extremely low. Compared to blind random perturbations, this steady-state trap risk index provides an intelligent decision-making basis for triggering the escape mechanism. It can maintain stability when the algorithm converges normally and provide timely warnings when it falls into a false local optimum.

[0016] Optionally, the dynamic escape probability The calculation method is as follows: ; in, Let be the steady-state trap risk index at the t-th iteration. This is a preset risk threshold constant.

[0017] Optionally, the method for generating the chaotic perturbation term includes: Perform a tangent function transformation on random numbers that are uniformly distributed in the interval (-0.5, 0.5).

[0018] Optionally, the step of using the dynamic escape probability as a trigger condition specifically means that the perturbation update is triggered when the generated random number is less than the dynamic escape probability.

[0019] Optionally, the parameter set includes at least one of the following: transformer equivalent circuit parameters, thermal resistance parameters of the thermal circuit model, and thermal capacity parameters; the transformer equivalent circuit parameters include winding DC resistance and leakage reactance; the thermal circuit model parameters include thermal resistance and thermal capacity.

[0020] Optionally, after applying the obtained globally optimal parameter set to the digital twin model, the method further includes: Obtain the operational dataset for a specified time period; each operational data point in the operational dataset must include at least the load voltage, load current, and transformer top oil temperature. The prediction accuracy of the calibrated digital twin model is verified using the running dataset; if the prediction accuracy meets the set requirements, the calibration is completed; if not, the improved particle swarm optimization algorithm is re-triggered.

[0021] In the second aspect, the digital twin-based operation and maintenance management system for prefabricated substations includes: processor; The memory stores computer instructions for operation and maintenance management of prefabricated substations based on digital twins. When the computer instructions are executed by the processor, the system performs the aforementioned operation and maintenance management method for prefabricated substations based on digital twins.

[0022] The beneficial effects of this invention are as follows: The present invention addresses the problem of local optima in the multi-peak solution space of prefabricated substations. By integrating particle swarm aggregation degree and operating condition excitation intensity (load fluctuation and oil temperature change), a steady-state trap risk index is constructed, which adaptively generates dynamic escape probability. Combined with the Lorentz chaotic perturbation mechanism with heavy-tailed characteristics, a large-span search is forcibly triggered when there is insufficient operating condition information due to swarm aggregation. This effectively distinguishes between true and false convergence, ensuring the accuracy and robustness of substation physical parameter identification. Attached Figure Description

[0023] Figure 1 This schematically illustrates the steps of the operation and maintenance management method for prefabricated substations based on digital twins in this embodiment. Figure 2 The diagram illustrates the dynamic escape probability curve. Figure 3 The diagram illustrates the structural block diagram of the digital twin-based prefabricated substation operation and maintenance management system in this embodiment. Detailed Implementation

[0024] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0025] This invention addresses the problem that existing particle swarm optimization (PSO) algorithms are prone to getting trapped in local optima (premature convergence) when identifying digital twin parameters in box-type substations. It introduces a dynamic escape probability and chaotic perturbation mechanism to improve the PSO algorithm. Specifically, by evaluating the aggregation degree of the swarm and the excitation intensity of external operating conditions in real time, a steady-state trap risk index is constructed, which then adaptively triggers the escape mechanism. This ensures that the true combination of physical parameters can be found in a complex and multi-peaked solution space, and can maintain high fidelity and accuracy in various complex operating environments, thereby achieving precise operation and maintenance management of substations.

[0026] Specifically, such as Figure 1 As shown, the operation and maintenance management method for prefabricated substations based on digital twins in this embodiment includes the following steps: Step S1: Construct the set of parameters to be identified for the box-type substation and initialize the particle swarm.

[0027] To build a high-fidelity digital twin model, it is necessary to randomly generate a parameter set, which is the solution vector in the particle swarm optimization algorithm.

[0028] The set of parameters to be identified includes at least one of the transformer equivalent circuit parameters and thermal circuit model parameters. The transformer equivalent circuit parameters include winding DC resistance and leakage reactance; the thermal circuit model parameters include thermal resistance and thermal capacitance.

[0029] In this embodiment, the population size is set to [value] when initializing the particle swarm. Each particle Corresponding to a set of parameters to be identified And randomly initialize the velocity vector of each particle. .

[0030] In the random generation process, the range of values ​​for each parameter in the parameter set is determined according to the upper and lower limits set on the equipment nameplate and design specifications.

[0031] Step S2: Optimize and iterate the particle swarm using the improved particle swarm algorithm until the termination condition is met, and obtain the globally optimal parameter set.

[0032] In this embodiment, the improved particle swarm optimization algorithm includes updating the particle state by superimposing a chaotic perturbation term based on the Lorentz distribution onto the particle update velocity at each iteration, with the dynamic escape probability as the trigger condition, while other steps remain unchanged.

[0033] Specifically, during the iterative process of the particle swarm optimization algorithm, a dynamic escape probability is constructed to determine whether the algorithm has fallen into a local optimum.

[0034] In this embodiment, the process of obtaining the dynamic escape probability is as follows: Step S21: Calculate the steady-state trap risk index.

[0035] When the particle swarm is highly concentrated (suspected convergence), but the external operating conditions are weak (insufficient dynamic information in the data, making it difficult to distinguish the response differences of different parameter combinations), the algorithm is highly likely to fall into a pseudo-steady-state trap caused by data scarcity. Therefore, to address the problem that traditional methods struggle to distinguish between population convergence and being trapped in local optima, this embodiment introduces a steady-state trap risk index.

[0036] Specifically, the process of obtaining the steady-state trap risk index includes the following steps: First, calculate the clustering index.

[0037] In this embodiment, the clustering index is used to characterize the density of the population distribution in the parameter space at each iteration.

[0038] The aggregation index is the calculated position vector of each particle in the t-th iteration. With the global optimal position vector The root mean square of the Euclidean distance.

[0039] Specifically, the clustering index The calculation formula is: ; in, The total number of particles, Represents the magnitude of a vector. For the first The particle in the first Position at the next iteration This is the global optimal position of the current particle swarm.

[0040] As can be seen from the above formula, when the aggregation index The value approaches This indicates that most particles in the swarm have gathered at the globally optimal position vector. Near the point, the particle swarm algorithm is either in a convergent or stagnant state.

[0041] Secondly, the excitation intensity under the working conditions is calculated.

[0042] In this embodiment, considering that the more severe the load fluctuations and the faster the temperature changes in the prefabricated substation, the more conducive it is to excite the nonlinear dynamic characteristics of the prefabricated substation, thereby improving the observability of parameter identification, it is necessary to construct the operating condition excitation intensity of the richness of dynamic features contained in the operating dataset of the prefabricated substation within a set time window before executing the particle swarm algorithm.

[0043] Specifically, the process of obtaining the operating condition excitation intensity is as follows: first, obtain the running dataset within the set time window, and then calculate the operating condition excitation intensity based on the running dataset.

[0044] In this embodiment, sensors installed at key nodes of the prefabricated substation are used to collect real-time operational data of the prefabricated substation at different times to form an operational dataset.

[0045] In one embodiment, the operating data includes electrical data and non-electrical data; the electrical data includes one or more combinations of load voltage, load current, and power factor; the non-electrical data includes one or more combinations of transformer top oil temperature, winding temperature, and ambient temperature and humidity.

[0046] In one embodiment, the operating condition excitation intensity The calculation formula is: ; in, The set time window length; The variance of the load current within a set time window is used to characterize the severity of load fluctuations. To set the average load current within a set time window; To prevent the use of pre-defined positive real numbers with a denominator of zero; To define an index for the change in transformer top-layer oil temperature within a set time window, thus characterizing the activity of the thermal dynamic process.

[0047] The value of the above-mentioned preset positive real number can be The aforementioned rate of change is the ratio of the rate of change of the transformer top-level temperature with respect to time to the nominal temperature rise rate. The nominal temperature rise rate is the historical maximum rate of change, obtained by retrieving the stored historical monitoring sequences of the device.

[0048] As can be derived from the above formula, when the load current fluctuates greatly (numerator increases) and the average load is low (denominator decreases, relative fluctuation is large), or when the oil temperature change index is large (logarithmic term increases), the operating condition excitation intensity... This will increase significantly, which means that the current data contains rich dynamic response information, which is beneficial for the accurate identification of parameters.

[0049] Then, a steady-state trap risk index is constructed.

[0050] In this embodiment, a steady-state trap risk index is constructed to identify the dangerous state of high aggregation under low excitation.

[0051] Specifically, the steady-state trap risk index The calculation formula is: ; in, To set the excitation intensity of the operating conditions within the time window, This is the normalization scaling factor, used to adjust the index to a reasonable order of magnitude; As the weighting factor for the influence of operating conditions, This represents the aggregation index at the t-th iteration.

[0052] In this embodiment, the preferred range of the weighting factor is: , specifically, The preferred setting is 2, because when When the data is too small, the influence of the excitation intensity is insufficiently weighted, which may lead to frequent misjudgments of traps during periods of data stability; when... When the value is too large, the operating conditions become too demanding, making it difficult to trigger the escape mechanism. Preferably, in this embodiment... Set as It can balance the effects of aggregation degree and working condition intensity.

[0053] The aforementioned normalized scaling coefficients are preset based on the search space magnitude of the parameter set to be identified, and are used to map the steady-state trap risk index to the sensitivity interval.

[0054] In one instance, the normalized scaling coefficient can also be the average of multiple historical normalized scaling coefficients when the initial steady-state trap risk index is 0.5, based on the aggregation index and operating condition excitation intensity at the time after the particle swarm initialization is completed in multiple historical operation and maintenance management processes of the prefabricated substation.

[0055] From the above formula, it can be seen that: when the aggregation index Very small (population aggregation) and operating condition excitation intensity When the value is very small (stable operating conditions, no excitation), the denominator approaches a minimum, leading to a steady-state trap risk index. The sharp increase suggests that the current high aggregation is very likely a spurious local optimum.

[0056] Step S22: Generate dynamic escape probability based on steady-state trap risk index.

[0057] In this embodiment, the steady-state trap risk index is converted into a specific probability value to control the triggering of disturbances.

[0058] Specifically, dynamic escape probability The calculation formula is: ; in, The preset risk threshold constant, Let be the steady-state trap risk index at the t-th iteration, and exp() is an exponential function with the natural constant as the base.

[0059] As can be seen from the above formula, with the steady-state trap risk index The increase in dynamic escape probability It exhibits a monotonically increasing trend and approaches the value. This means that the higher the risk of falling into a steady-state trap, the greater the probability of triggering the escape mechanism, thus giving the algorithm adaptive adjustment capabilities.

[0060] In the above embodiments, the steady-state trap risk index is transformed into a dynamic probability value between 0 and 1, thereby achieving adaptive adjustment of the algorithm's capabilities. As the risk of falling into a steady-state trap increases, the escape probability automatically and monotonically increases. Compared to mutation operations with fixed probabilities, this mechanism can dynamically balance the algorithm's development (fine-grained search) and exploration (escaping local traps) capabilities based on the current search state and operating environment, thus improving optimization efficiency.

[0061] For example, such as Figure 2 As shown, with the increase of the number of iterations, each iteration has a corresponding dynamic escape probability.

[0062] In this embodiment, the triggering condition is: when the generated random number is less than the dynamic escape probability, the perturbation update is triggered.

[0063] Specifically, triggering operator for: ; for Random numbers that are uniformly distributed between them. This represents the dynamic escape probability.

[0064] In the above formula, when hour, Values Otherwise .

[0065] The above method of triggering perturbation updates by comparing random numbers with dynamic escape probabilities introduces appropriate randomness into the algorithm, ensuring that the escape mechanism does not cause system oscillations due to deterministic rules. Instead, it adjusts the intervention frequency statistically based on the level of risk, ensuring that the algorithm maintains population diversity without disrupting the overall convergence trend.

[0066] In this embodiment, a chaotic perturbation term is introduced based on the traditional PSO velocity update formula. This chaotic perturbation term is introduced to provide a larger random kinetic energy when a particle gets trapped in a local optimum, enabling it to escape the current search region.

[0067] Specifically, update the velocity vector of the particle. The formula is: ; in, Inertial weight; For learning factors; for Uniformly distributed random numbers within a given range; For particles The historical optimal position of an individual; For the particle at the t-th iteration The current location; The globally optimal position; To trigger the operator, This is a disturbance term.

[0068] Among them, the chaotic perturbation term The generation method is as follows: For The tangent function is used to transform uniformly distributed random numbers within the interval (i.e., the generation method of Lorentz distribution).

[0069] Specifically: ,in for Random numbers between, This represents the disturbance intensity coefficient.

[0070] In the above embodiments, a chaotic perturbation term following a Lorentz distribution is generated using a tangent function transformation. Compared to the traditional Gaussian distribution perturbation, the Lorentz distribution has a more pronounced heavy-tailed characteristic, meaning that it provides small perturbations to maintain search accuracy in most cases, but can generate extremely large impulse perturbations with a low probability. This large-span jumping ability is crucial for helping particles escape deep and wide local extrema in the multi-peak solution space of the box-type substation, significantly improving the global optimization capability and effectively solving the problem of premature convergence in the complex parameter space of the box-type substation.

[0071] In this embodiment, the globally optimal parameter set is obtained by iterating the above process until a preset termination condition is met. .

[0072] The preset termination condition is reaching the maximum number of iterations or the variance of the fitness of the entire population being less than a preset threshold.

[0073] Step S3: Apply the globally optimal parameter set to the digital twin model to realize substation operation and maintenance management.

[0074] In this embodiment, the global optimal parameter set will be used to update the digital twin model of the box-type substation in order to realize the management of substation operation and maintenance.

[0075] Furthermore, after obtaining the identified parameter set, it is not used directly, but a self-calibration verification process is performed, specifically as follows: Obtain the operational dataset for a specified time period; each operational data point in the operational dataset must include at least the load voltage, load current, and transformer top oil temperature; The prediction accuracy of the calibrated digital twin model is verified using the running dataset; if the prediction accuracy meets the set requirements, the calibration is completed; if not, the self-calibration method is retried.

[0076] Specifically, the running dataset within the aforementioned set time period can be a running dataset acquired over a historical period. This historical period includes a set time window and the running dataset within that set time period. In other words, the running dataset within the set time period is a reserved segment of running data from the historical period used as a validation set. This data is then input into the updated digital twin model, and the error between the model output and the measured data is calculated. If the prediction accuracy meets the set requirements (e.g., the root mean square error RMSE is less than...), the model is considered valid. If the conditions are met, the calibration is complete. After calibration, the digital twin model is used for real-time condition monitoring, overload capacity assessment, and insulation life prediction. If the conditions are not met, the improved particle swarm optimization algorithm is re-triggered.

[0077] The output of the above model can be either the load current or the top oil temperature of the transformer.

[0078] In the above embodiments, by accurately identifying winding resistance, leakage reactance, and thermal resistance and capacitance, the digital twin model can not only reflect electrical characteristics but also accurately simulate the temperature rise process. This enables the operation and maintenance system to accurately assess the aging status, overload capacity, and insulation life of equipment based on its physical nature, rather than simply performing numerical fitting.

[0079] The present invention introduces a coupled evaluation mechanism of aggregation degree and operating condition excitation to construct a dynamic escape probability that can sense pseudo-steady state. Combined with chaotic perturbations with heavy-tailed characteristics, it significantly improves the robustness of the algorithm under varying operating conditions while ensuring the accuracy of parameter identification, thereby realizing the precision and intelligence of operation and maintenance management of box-type substations.

[0080] This invention also provides a digital twin-based operation and maintenance management system for prefabricated substations. For example... Figure 3 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the above-described digital twin-based operation and maintenance management method for prefabricated substations according to the present invention.

[0081] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and therefore will not be described in detail here.

[0082] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented by computer-readable / executable instructions stored or otherwise maintained on such a computer-readable medium.

[0083] In the description of this specification, "multiple" means at least two, such as two, three or more, etc., unless otherwise expressly and specifically defined.

[0084] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A digital twin-based operation and maintenance management method for prefabricated substations, characterized in that, include: Initialize the particle swarm, with each particle corresponding to a set of parameters for the box-type substation; An improved particle swarm optimization algorithm is used to optimize and iterate the particle swarm until the termination condition is met, thus obtaining the globally optimal parameter set. The globally optimal parameter set is applied to the digital twin model to achieve substation operation and maintenance management; The improved particle swarm optimization algorithm includes adding a chaotic perturbation term based on the Lorentz distribution to the particle update velocity at each iteration, with the dynamic escape probability as the trigger condition. The dynamic escape probability is positively correlated with the steady-state trap risk index at the current iteration; the steady-state trap risk index is negatively correlated with the aggregation index and the operating condition excitation intensity, respectively, and both are inversely proportional; the aggregation index represents the aggregation of all particle positions of the particle swarm to the global optimal position at the current iteration; the operating condition excitation intensity is positively correlated with the variance of the pre-obtained load current of the box-type substation within a set time window, negatively correlated with the mean of the load current, and positively correlated with the absolute value of the change index of the transformer top oil temperature.

2. The method for operation and maintenance management of prefabricated substations based on digital twins according to claim 1, characterized in that, The calculation method for the aggregation index includes: Calculate the root mean square distance between the position vector of each particle and the global optimal position vector at each iteration; use the root mean square distance as the aggregation index.

3. The method for operation and maintenance management of prefabricated substations based on digital twins according to claim 1, characterized in that, The excitation intensity under the operating conditions The calculation method is as follows: ； in, and These represent the standard deviation and mean value of the load current within a set time window, respectively, where I is the load current. To set the change index of transformer top oil temperature within a time window, To prevent the default positive real number with a denominator of zero, To set the length of the time window.

4. The method for operation and maintenance management of prefabricated substations based on digital twins according to claim 1, characterized in that, The steady-state trap risk index The calculation method is as follows: ； in, The normalized scaling coefficient. Let be the aggregation index at the t-th iteration. For the working condition excitation strength, The weighting factor is the influence of operating conditions.

5. The method for operation and maintenance management of prefabricated substations based on digital twins according to claim 1, characterized in that, The dynamic escape probability The calculation method is as follows: ； in, Let be the steady-state trap risk index at the t-th iteration. This is a preset risk threshold constant.

6. The method for operation and maintenance management of prefabricated substations based on digital twins according to claim 1, characterized in that, The method for generating the chaotic perturbation term includes: Perform a tangent function transformation on random numbers that are uniformly distributed in the interval (-0.5, 0.5).

7. The method for operation and maintenance management of prefabricated substations based on digital twins according to claim 1 or 6, characterized in that, The triggering condition based on the dynamic escape probability is specifically: when the generated random number is less than the dynamic escape probability, the perturbation update is triggered.

8. The method for operation and maintenance management of prefabricated substations based on digital twins according to claim 1, characterized in that, The parameter set includes at least one of the following: transformer equivalent circuit parameters, thermal resistance parameters of the thermal circuit model, and thermal capacity parameters; the transformer equivalent circuit parameters include winding DC resistance and leakage reactance; the thermal circuit model parameters include thermal resistance and thermal capacity.

9. The method for operation and maintenance management of prefabricated substations based on digital twins according to claim 1, characterized in that, After applying the obtained globally optimal parameter set to the digital twin model, the process further includes: Obtain the operational dataset for a specified time period; each operational data point in the operational dataset must include at least the load voltage, load current, and transformer top oil temperature. The prediction accuracy of the calibrated digital twin model is verified using the running dataset; if the prediction accuracy meets the set requirements, the calibration is completed; if not, the improved particle swarm optimization algorithm is re-triggered.

10. A digital twin-based operation and maintenance management system for prefabricated substations, characterized in that: include: processor; A memory storing computer instructions for operation and maintenance management of prefabricated substations based on digital twins, wherein when the computer instructions are executed by the processor, the system performs the operation and maintenance management method for prefabricated substations based on digital twins according to any one of claims 1-9.

Images