A circuit breaker state evolution prediction method based on a digital twin model

Abstract

The application discloses a circuit breaker state evolution prediction method based on a digital twin model, relates to the technical field of digital twinning, and comprises the following steps: collecting multi-source operation data, processing and generating operation condition state data sets; constructing a circuit breaker digital twin model, performing state calibration and generating a state vector; establishing a state domain and an energy domain bidirectional mapping, performing forward mapping to generate an energy state vector; constructing a condition evolution scheduler, extracting condition semantics, determining a channel activation sequence and activation intensity; constructing a multiphase flow type transition channel and jointly driving, and outputting an energy state prediction result; reflecting and mapping to generate a future state evolution result, and outputting a state trend and a health prediction value. The application realizes accurate modeling and forward-looking prediction of the state evolution process of the circuit breaker under complex conditions by constructing the circuit breaker digital twin model and introducing the state energy bidirectional mapping and the multiphase flow type state transition mechanism.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and in particular to a method for predicting the state evolution of circuit breakers based on a digital twin model. Background Technology

[0002] As a critical control and protection device in power systems, the operating status of circuit breakers directly affects the safety and reliability of the power grid. With the continuous expansion of power system capacity and the increasing complexity of operating conditions, circuit breakers are subjected to various influences during long-term operation, including current surges, mechanical wear, thermal aging, and environmental factors, resulting in a continuous evolution of their internal state over time. Therefore, effectively modeling and predicting the operating status of circuit breakers has become an important research topic in the field of power equipment condition monitoring and intelligent operation and maintenance.

[0003] In existing technologies, circuit breaker condition assessment and prediction methods are mostly based on single operating indicators or simple time series analysis models. They estimate the health status of the equipment by statistically or trend-fitting parameters such as contact wear, number of operations, or temperature rise. Some methods introduce physical models or data-driven models to simulate and analyze the circuit breaker condition. However, the above methods usually treat the circuit breaker condition change as a single degradation path, making it difficult to simultaneously depict the real state evolution process under the superposition of multiple behaviors such as long-term operational degradation, sudden shocks such as short-circuit breaking, and condition recovery caused by maintenance.

[0004] Existing technologies generally lack systematic modeling of operating condition sequences and their evolution order, making it difficult to reflect the impact of different combinations of operating conditions, operating durations, and historical state paths on the state evolution results. State variables are mostly processed through direct mapping or linear relationships, failing to effectively reflect the characteristics of energy accumulation, dissipation, and irreversible evolution. This results in prediction results that are insensitive to changes in complex operating conditions, with insufficient prediction stability and foresight, making it difficult to provide reliable support for the refined operation and maintenance decisions of circuit breakers.

[0005] Therefore, how to provide a method for predicting the state evolution of circuit breakers based on digital twin models is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a circuit breaker state evolution prediction method based on a digital twin model. This invention constructs a digital twin model of the circuit breaker by performing computational processing and state modeling on multi-source operating data of the circuit breaker. It introduces a bidirectional mapping mechanism between the state domain and the energy domain, operating condition evolution scheduling, and multiphase flow-type state transition processing to achieve predictive analysis of the circuit breaker's state evolution process under complex operating conditions. This invention fully utilizes the computer's modeling and evolutionary computation capabilities for equipment operating data, enabling it to simultaneously characterize long-term operational degradation, sudden operating condition shocks, and state changes caused by maintenance and repair. It possesses advantages such as sensitivity to complex operating condition sequences, stable prediction results, detailed state evolution characterization, and strong support for operation and maintenance decision-making.

[0007] A circuit breaker state evolution prediction method based on a digital twin model according to an embodiment of the present invention includes:

[0008] Collect multi-source operating data during the operation of the circuit breaker, process the multi-source operating data, and form an operating condition status dataset;

[0009] A digital twin model of a circuit breaker is constructed. The state of the digital twin model of the circuit breaker is calibrated based on the operating condition state dataset. A state vector is defined in the digital twin model of the circuit breaker to represent the health state of the circuit breaker.

[0010] Based on the current state vector and the historical state change information in the operating condition state dataset, a bidirectional mapping relationship between the circuit breaker state domain and energy domain is established. The current state vector is then subjected to state-energy domain mapping processing to generate an energy state vector.

[0011] A condition evolution scheduler is constructed. Based on the current operating condition and historical operating condition sequence in the operating condition state dataset, the condition semantic information and condition influence trajectory are generated by the condition evolution scheduler, and the channel activation order and channel activation intensity used to control state transitions are determined.

[0012] Based on the energy state vector, a multiphase flow state transition channel is constructed, which includes an asymptotic phase channel, abrupt phase channel, and a recovery phase channel. The multiphase flow state transition channel is jointly driven according to the channel activation order and channel activation intensity to generate energy state prediction results.

[0013] Based on the bidirectional mapping relationship, the energy state prediction results are processed by inverse mapping from the energy domain to the state domain to obtain the state evolution prediction results of the circuit breaker in the future multiple time steps, and output the state change trend and comprehensive health state prediction results of the circuit breaker.

[0014] Optionally, the multi-source operating data includes circuit breaker opening and closing time data, opening and closing stroke data, opening and closing operation speed data, contact rebound characteristic data, arc duration data, breaking current data, contact resistance data, contact temperature rise data, circuit breaker housing temperature data, ambient temperature data, load current level data, operating frequency data, and maintenance record data.

[0015] Optionally, the processing of multi-source operational data includes time synchronization, anomaly removal, and standardization of the multi-source operational data.

[0016] Optionally, defining a state vector in the circuit breaker digital twin model to characterize the health state of the circuit breaker includes:

[0017] In the operating condition status dataset, the multi-source operating data is divided into fields. The breaking current data, arc duration data, load current level data, ambient temperature data, and operating frequency data are set as operating condition input fields. The opening and closing time data, opening and closing stroke data, opening and closing operation speed data, contact rebound characteristic data, contact resistance data, contact temperature rise data, and circuit breaker shell temperature data are set as observation fields. The maintenance record data is set as maintenance event fields. All fields are sorted by time according to a unified sampling period.

[0018] A digital twin model of a circuit breaker is constructed. The digital twin model of the circuit breaker consists of a state layer, an operating condition input layer, an observation layer, and a parameter set. The state layer contains a state vector composed of contact wear state quantity, operating mechanism fatigue state quantity, and insulation and thermal aging state quantity. The operating condition input layer receives the operating condition input quantity field for each sampling period. The observation layer receives the observation quantity field for the same sampling period. The parameter set records the state update coefficient and the observation mapping coefficient.

[0019] Within each sampling period, the working condition input layer and the state layer are called. According to the state update rules in the parameter set, the state vector of the previous sampling period is mapped to the working condition input field of the current sampling period to generate the state vector of the current sampling period. The observation layer is called to map the state vector of the current sampling period to generate the observation layer output according to the observation mapping rules.

[0020] Using the observation fields of the same sampling period in the operating condition status dataset as a reference, the error of the observation layer output is calculated. The initial value of the corresponding component in the status layer is reset in the sampling period triggered by the maintenance event field. The status update coefficient and observation mapping coefficient in the parameter set are iteratively adjusted until the error meets the preset threshold.

[0021] After the error meets the preset threshold, the parameter set is fixed, the state vector of the current sampling period is output and the state vector sequence is saved.

[0022] Optionally, the generation of the energy state vector includes:

[0023] Extract the state vector corresponding to the current sampling period and the historical state change information aligned with time from the operating condition status dataset. Generate a historical state change sequence based on the historical state change information. Extract the breaking current data, arc duration data, load current level data, ambient temperature data and operating frequency data within the same time range as the operating condition driving sequence.

[0024] A bidirectional mapping relationship is established between the circuit breaker's state domain and energy domain. This bidirectional mapping relationship includes a forward mapping from the state domain to the energy domain and a reverse mapping from the energy domain to the state domain, wherein:

[0025] The positive mapping relationship is determined by the state vector and the operating condition driving sequence, which together determine the composition, order, and weight of the components of the energy state vector.

[0026] The reverse mapping relationship reconstructs the corresponding state vector and corrects the state evolution hysteresis characteristics based on the component composition, component order and component weight, combined with the corresponding working condition driving sequence and historical state change sequence.

[0027] When establishing a positive mapping relationship, the state vector is processed by path identification based on the historical state change sequence to generate path identifiers to distinguish between degenerate paths and recovery paths. The path identifiers are used as input items for the positive mapping relationship.

[0028] Based on the positive mapping relationship, the state vector of the current sampling period is processed to map from state to energy domain. The energy state vector is output according to the composition and order of the components of the energy state vector, and the energy state vector is bound and stored with the path identifier.

[0029] The energy state vector is converted into the inverse mapping result of the state vector based on the inverse mapping relationship. The consistency of the inverse mapping result with the state vector of the current sampling period is checked. After the consistency check is passed, the bidirectional mapping relationship is fixed and the energy state vector is output.

[0030] Optionally, determining the channel activation order and channel activation intensity used to control state transitions includes:

[0031] Extract the current operating condition and historical operating condition sequences from the operating condition status dataset, complete time alignment and unified sampling, and form the operating condition input set corresponding to the current sampling period;

[0032] A working condition evolution scheduler is constructed, which consists of three layers: a semantic deconstruction layer, an influence trajectory planning layer, and an orchestration execution layer. The semantic deconstruction layer receives the working condition input set and outputs the working condition semantic information and path identifier. The influence trajectory planning layer outputs the working condition influence trajectory and the orchestration time window set based on the historical operating working condition sequence. The orchestration execution layer outputs the channel activation order and channel activation intensity based on the working condition semantic information and the working condition influence trajectory.

[0033] The semantic deconstruction layer performs semantic parsing on the current operating condition, generating operating condition semantic information composed of driving semantic information, modulation semantic information, and disturbance semantic information. Based on the maintenance records and short circuit interruption records, recovery path identifiers and sudden change path identifiers are generated for the corresponding operating condition segments.

[0034] Through the influence trajectory planning layer, short-cycle impact trajectories, long-cycle evolution trajectories, and recovery sensitive intervals are generated based on historical operating condition sequences and path identifiers, and corresponding time window sets and priority sequences are generated.

[0035] Through the orchestration execution layer, based on the semantic information of the working condition, the trajectory of the working condition influence, the set of orchestration time windows and the priority sequence, the channel activation order, channel activation intensity and activation duration for controlling state transitions are determined, and conflict suppression and mutual exclusion constraints are applied within the time window of channel overlap.

[0036] Optionally, the generation of energy state prediction results includes:

[0037] A multiphase flow state transition channel is constructed based on the energy state vector. The multiphase flow state transition channel includes an asymptotic phase channel, abrupt phase channel, and a recovery phase channel. A channel energy ledger, a sliding formation window, and a channel memory unit are established to record the channel processing order, channel processing intensity, and channel historical residual.

[0038] According to the channel activation order and channel activation intensity, the progressive phase channel is driven to perform piecewise accumulation and temperature fatigue coupled window processing, and a monotonic envelope and rate upper limit constraint are applied to the energy state vector to generate the first intermediate energy state vector.

[0039] Identify the impact segments triggered by short-circuit interruption and arc duration in the operating condition state dataset, select the mutation processing template according to the path identifier, drive the mutation phase channel to avoid repeated inclusion within the refractory period window and perform dynamic saturation clipping, and generate a second intermediate energy state vector.

[0040] Based on the maintenance records in the operating condition dataset, two types of recovery units are distinguished: lubrication recovery and component replacement recovery. The maximum pullback limit and gradual entry time window are set in the drive recovery phase channel to perform recovery processing on the second intermediate energy state vector and generate the third intermediate energy state vector.

[0041] Within the sliding formation window, conflict suppression and mutual exclusion constraints are applied to the overlapping time windows of the three types of channels. The channel residual buffer of the channel memory unit is enabled to perform sequential processing on the undigested residuals. The channels are merged according to the channel activation order and the order of completion of channel activation intensity to obtain candidate energy state prediction results.

[0042] The energy consistency of the channel energy ledger is checked, and the energy state prediction result is confirmed after the preset threshold is met.

[0043] Optionally, the state change trend and comprehensive health status prediction results of the output circuit breaker include:

[0044] Obtain the energy state prediction result, and call the mapping relationship from energy domain to state domain in the bidirectional mapping relationship between the circuit breaker state domain and energy domain as the basis for the inverse mapping processing from energy domain to state domain;

[0045] Based on the mapping relationship from the energy domain to the state domain, the energy state prediction results are processed by component-by-component mapping according to the component composition, component order and component weight of the energy state vector, and each energy component is converted into the corresponding state change quantity to generate the predicted state vector.

[0046] According to the definition structure of the state vector in the circuit breaker digital twin model, the predicted state vector is decomposed to obtain the contact wear state quantity, the operating mechanism fatigue state quantity, and the insulation and thermal aging state quantity. The value range constraint and the rate of change constraint are applied to each state quantity to form the constrained predicted state vector.

[0047] The constrained predicted state vector is re-input into the mapping relationship from the circuit breaker state domain to the energy domain to obtain the corresponding energy remapping result. The consistency of the energy remapping result and the energy state prediction result is compared. When the consistency meets the preset condition, the predicted state vector is confirmed as a valid result.

[0048] The confirmed and valid predicted state vector is used as the current state vector for the next time step. The bidirectional mapping process between the state domain and the energy domain, the operating condition evolution scheduling process, and the multiphase flow state transition process are sequentially input to obtain the state evolution prediction results of the circuit breaker for the next multiple time steps. The state change trend and comprehensive health state prediction results of the circuit breaker are then output.

[0049] The beneficial effects of this invention are:

[0050] This invention constructs a digital twin model of a circuit breaker and performs unified modeling and processing of multi-source operational data during circuit breaker operation. This enables continuous representation and dynamic updating of the circuit breaker's health status, allowing the circuit breaker's state to no longer be limited to a single indicator or static evaluation result, but rather to reflect its evolutionary characteristics during operation in the form of a state vector. This invention can continuously track the state changes of key internal components of the circuit breaker while maintaining consistency in physical meaning, providing a stable and reliable foundation for subsequent state evolution analysis and prediction.

[0051] This invention introduces a bidirectional mapping mechanism between the state domain and the energy domain, transforming the circuit breaker's state change process into an energy evolution process. This allows state changes caused by long-term operational degradation, short-circuit breaking impacts, and maintenance to be characterized within a unified energy expression space. Through the combination of bidirectional mapping and consistency verification, this invention effectively improves the continuity and stability of state evolution modeling, avoiding prediction biases caused by discrete state variables and missing path dependencies in existing technologies. This makes the state evolution prediction results more sensitive to historical state paths and operating condition sequences.

[0052] This invention constructs an operating condition evolution scheduler and combines it with multiphase flow-type state transition processing to uniformly regulate different operating conditions and their evolution rhythms. This enables the circuit breaker's state evolution process to dynamically switch between gradual degradation, sudden changes, and state recovery according to changes in operating conditions. This invention can more realistically reflect the actual state change patterns of circuit breakers in complex operating environments, improve the foresight and practicality of state evolution prediction, and provide effective support for the refined operation and maintenance management and life assessment of circuit breakers. Attached Figure Description

[0053] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0054] Figure 1 This is a flowchart of a circuit breaker state evolution prediction method based on a digital twin model proposed in this invention;

[0055] Figure 2 This is a schematic diagram of the operating condition evolution scheduler for a circuit breaker state evolution prediction method based on a digital twin model proposed in this invention.

[0056] Figure 3 This is a schematic diagram of the processing of a multiphase flow state transition channel in a circuit breaker state evolution prediction method based on a digital twin model proposed in this invention. Detailed Implementation

[0057] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0058] refer to Figure 1 , Figure 2 and Figure 3 A method for predicting the state evolution of circuit breakers based on a digital twin model, comprising:

[0059] Collect multi-source operating data during the operation of the circuit breaker, process the multi-source operating data, and form an operating condition status dataset;

[0060] A digital twin model of a circuit breaker is constructed. The state of the digital twin model of the circuit breaker is calibrated based on the operating condition state dataset. A state vector is defined in the digital twin model of the circuit breaker to represent the health state of the circuit breaker.

[0061] Based on the current state vector and the historical state change information in the operating condition state dataset, a bidirectional mapping relationship between the circuit breaker state domain and energy domain is established. The current state vector is then subjected to state-energy domain mapping processing to generate an energy state vector.

[0062] A condition evolution scheduler is constructed. Based on the current operating condition and historical operating condition sequence in the operating condition state dataset, the condition semantic information and condition influence trajectory are generated by the condition evolution scheduler, and the channel activation order and channel activation intensity used to control state transitions are determined.

[0063] Based on the energy state vector, a multiphase flow state transition channel is constructed, which includes an asymptotic phase channel, abrupt phase channel, and a recovery phase channel. The multiphase flow state transition channel is jointly driven according to the channel activation order and channel activation intensity to generate energy state prediction results.

[0064] Based on the bidirectional mapping relationship, the energy state prediction results are processed by inverse mapping from the energy domain to the state domain to obtain the state evolution prediction results of the circuit breaker in the future multiple time steps, and output the state change trend and comprehensive health state prediction results of the circuit breaker.

[0065] In this embodiment, the multi-source operating data includes circuit breaker opening and closing time data, opening and closing stroke data, opening and closing operation speed data, contact rebound characteristic data, arc duration data, breaking current data, contact resistance data, contact temperature rise data, circuit breaker housing temperature data, ambient temperature data, load current level data, operating frequency data, and maintenance record data.

[0066] In this embodiment, the processing of multi-source operational data includes time synchronization, anomaly removal, and standardization of the multi-source operational data.

[0067] In this embodiment, defining a state vector in the circuit breaker digital twin model to characterize the health state of the circuit breaker includes:

[0068] In the operating condition status dataset, the multi-source operating data is divided into fields. The breaking current data, arc duration data, load current level data, ambient temperature data, and operating frequency data are set as operating condition input fields. The opening and closing time data, opening and closing stroke data, opening and closing operation speed data, contact rebound characteristic data, contact resistance data, contact temperature rise data, and circuit breaker shell temperature data are set as observation fields. The maintenance record data is set as maintenance event fields. All fields are sorted by time according to a unified sampling period.

[0069] A digital twin model of a circuit breaker is constructed. The digital twin model of the circuit breaker consists of a state layer, an operating condition input layer, an observation layer, and a parameter set. The state layer contains a state vector composed of contact wear state quantity, operating mechanism fatigue state quantity, and insulation and thermal aging state quantity. The operating condition input layer receives the operating condition input quantity field for each sampling period. The observation layer receives the observation quantity field for the same sampling period. The parameter set records the state update coefficient and the observation mapping coefficient.

[0070] Within each sampling period, the operating condition input layer and the state layer are invoked. Based on the state update rules in the parameter set, the state vector of the previous sampling period is mapped to the operating condition input field of the current sampling period to generate the state vector of the current sampling period. The observation layer is then invoked to map the state vector of the current sampling period to generate the observation layer output based on the observation mapping rules. The state update rules are as follows:

[0071] Based on the state vector of the previous sampling period as the initial state value, the operating condition input field corresponding to the current sampling period is introduced, and the state increment superposition processing is performed on the contact wear state quantity, the operating mechanism fatigue state quantity, and the insulation and thermal aging state quantity respectively.

[0072] Based on the load current level, breaking current, arc duration and operation frequency in the operating condition input field, different weights are applied to the state increments corresponding to each state quantity, and the state update amplitude is constrained by the state change amplitude of the previous sampling period.

[0073] When the operating condition input field contains maintenance records, limited rollback processing is performed on the corresponding state quantity, and the state vector of the current sampling period is generated after the rollback processing is completed.

[0074] The observation mapping rule is:

[0075] Based on the state vector of the current sampling period, the corresponding observation mapping relationship is selected according to the composition order of the contact wear state quantity, the operating mechanism fatigue state quantity, and the insulation and thermal aging state quantity in the state vector;

[0076] Based on the aforementioned observation mapping relationship, the contact wear state quantity is mapped to the observation of opening and closing time and contact resistance, the operating mechanism fatigue state quantity is mapped to the observation of opening and closing operation speed and stroke deviation, and the insulation and thermal aging state quantity is mapped to the observation of contact temperature rise and shell temperature.

[0077] The time alignment and dimension consistency processing of the observations obtained from the mapping are performed to form the observation layer output of the current sampling period;

[0078] Using the observation fields of the same sampling period in the operating condition status dataset as a reference, the error of the observation layer output is calculated. The initial value of the corresponding component in the status layer is reset in the sampling period triggered by the maintenance event field. The status update coefficient and observation mapping coefficient in the parameter set are iteratively adjusted until the error meets the preset threshold.

[0079] After the error meets the preset threshold, the parameter set is fixed, the state vector of the current sampling period is output and the state vector sequence is saved. The preset threshold is the relative error threshold that the comprehensive error between the observation layer output and the corresponding measured running data of the current sampling period does not exceed 5%.

[0080] In this embodiment, generating the energy state vector includes:

[0081] Extract the state vector corresponding to the current sampling period and the historical state change information aligned with time from the operating condition status dataset. Generate a historical state change sequence based on the historical state change information. Extract the breaking current data, arc duration data, load current level data, ambient temperature data and operating frequency data within the same time range as the operating condition driving sequence.

[0082] A bidirectional mapping relationship is established between the circuit breaker's state domain and energy domain. This bidirectional mapping relationship includes a forward mapping from the state domain to the energy domain and a reverse mapping from the energy domain to the state domain, wherein:

[0083] The positive mapping relationship is determined by the state vector and the operating condition driving sequence, which together determine the component composition, component order, and component weights of the energy state vector. Specifically, the determination of the component composition, component order, and component weights is as follows:

[0084] Based on the composition of the contact wear state quantity, the operating mechanism fatigue state quantity, and the insulation and thermal aging state quantity in the state vector, the basic components of the energy state vector are determined, and each basic component is assigned to a different type of state change source.

[0085] Based on the time sequence, duration, and type of the operating conditions in the operating condition drive sequence, the basic components are sorted to determine the order of each component in the energy state vector, so that the component order is consistent with the time path of state evolution.

[0086] Based on the changes in load current level, breaking current, arc duration and operation frequency in the operating condition drive sequence, corresponding component weights are set for each component, and the component weights are adjusted in combination with historical state change information to complete the construction of the energy state vector.

[0087] The current operating parameters are: load current level of 690A, breaking current of 8200A, arc duration of 7.5ms, and operation frequency of 4 times / hour, all obtained during the current sampling period. The baseline parameters are: load current level of 610A, breaking current of 0A, arc duration of 0ms, and operation frequency of 1 time / hour, corresponding to the previous stable phase. The resulting variation amplitudes are 80A, 8200A, 7.5ms, and 3 times / hour, respectively. Normalization is performed on these variation amplitudes according to preset normalization limits of 120A, 10000A, 12ms, and 6 times / hour, yielding normalization results of 0.667, 0.820, 0.625, and 0.500, respectively. The four basic contribution coefficients for the contact wear component are set to 0.15, 0.45, 0.30 and 0.10, respectively; the four basic contribution coefficients for the operating mechanism fatigue component are set to 0.10, 0.20, 0.10 and 0.60, respectively; and the four basic contribution coefficients for the insulation and thermal aging component are set to 0.35, 0.15, 0.30 and 0.20, respectively. The initial component values ​​of the three basic components are calculated to be 0.707, 0.593 and 0.644, respectively. Combining the state change sequences of the previous three consecutive sampling periods, where the contact wear state quantity increased from 0.41 to 0.43 and then to 0.49, the operating mechanism fatigue state quantity increased from 0.46 to 0.48 and then to 0.54, and the insulation and thermal aging state quantity increased from 0.44 to 0.45 and then to 0.47, historical correction coefficients of 1.08, 1.05, and 1.03 were set to correct the three basic components, resulting in corrected component values ​​of 0.763, 0.623, and 0.663, respectively. Normalization of the corrected component values ​​yielded a weight of 0.372 for the contact wear component, 0.304 for the operating mechanism fatigue component, and 0.324 for the insulation and thermal aging component in the current sampling period, thus completing the determination of the weights of each component of the energy state vector.

[0088] The reverse mapping relationship reconstructs the corresponding state vector and corrects the state evolution hysteresis characteristics based on the component composition, component order and component weight, combined with the corresponding working condition driving sequence and historical state change sequence.

[0089] When establishing a positive mapping relationship, the state vector is processed by path identification based on the historical state change sequence to generate path identifiers to distinguish between degenerate paths and recovery paths. The path identifiers are used as input items for the positive mapping relationship.

[0090] Based on the forward mapping relationship, the state vector of the current sampling period is processed to perform state-to-energy domain mapping. The energy state vector is output according to its component composition and component order, and the energy state vector is bound and stored with the path identifier. Specifically, the process of performing state-to-energy domain mapping based on the forward mapping relationship on the state vector of the current sampling period is as follows:

[0091] According to the pre-determined component composition in the positive mapping relationship, the contact wear state quantity, the operating mechanism fatigue state quantity, and the insulation and thermal aging state quantity are extracted from the state vector of the current sampling period, and respectively mapped to the basic energy component in the energy state vector.

[0092] Based on the component order determined in the forward mapping relationship, each basic energy component is arranged in the time sequence corresponding to the working condition drive sequence to form an ordered energy component sequence.

[0093] Based on the component weights determined in the forward mapping relationship, each energy component is weighted and the weighted energy components are combined and output as the energy state vector of the current sampling period.

[0094] The energy state vector is converted into the inverse mapping result of the state vector based on the inverse mapping relationship. The consistency of the inverse mapping result with the state vector of the current sampling period is checked. After the consistency check is passed, the bidirectional mapping relationship is fixed and the energy state vector is output.

[0095] In this embodiment, determining the channel activation order and channel activation intensity used to control state transitions includes:

[0096] Extract the current operating condition and historical operating condition sequences from the operating condition status dataset, complete time alignment and unified sampling, and form the operating condition input set corresponding to the current sampling period;

[0097] A working condition evolution scheduler is constructed, which consists of three layers: a semantic deconstruction layer, an influence trajectory planning layer, and an orchestration execution layer. The semantic deconstruction layer receives the working condition input set and outputs the working condition semantic information and path identifier. The influence trajectory planning layer outputs the working condition influence trajectory and the orchestration time window set based on the historical operating working condition sequence. The orchestration execution layer outputs the channel activation order and channel activation intensity based on the working condition semantic information and the working condition influence trajectory.

[0098] The semantic deconstruction layer performs semantic parsing on the current operating condition to generate operating condition semantic information composed of driving semantic information, modulation semantic information, and disturbance semantic information. Based on maintenance records and short-circuit interruption records, recovery path identifiers and sudden change path identifiers are generated for the corresponding operating condition segments. Specifically, the semantic parsing of the current operating condition through the semantic deconstruction layer involves:

[0099] Based on the load current level, operation frequency and ambient temperature data corresponding to the current sampling period in the operating condition status dataset, the basic operating condition is identified, and the operating condition information reflecting the continuous operating characteristics is extracted as driving semantic information.

[0100] Based on the arc duration, opening and closing time variation, and contact temperature rise variation in the centralized data of operating conditions, the regulatory influencing factors in the current operating conditions are identified, and the corresponding operating condition information is extracted as modulation semantic information.

[0101] Based on the short-circuit interruption records and maintenance records in the operating condition status dataset, sudden events in the current operating condition are marked. When a short-circuit interruption record is detected, a sudden change path identifier is generated, and when a maintenance record is detected, a recovery path identifier is generated.

[0102] Through the aforementioned impact trajectory planning layer, short-cycle impact trajectories, long-cycle evolution trajectories, and recovery sensitive intervals are generated based on historical operating condition sequences and path identifiers. Corresponding time window sets and priority sequences are also generated. Specifically, the generation of short-cycle impact trajectories, long-cycle evolution trajectories, and recovery sensitive intervals based on historical operating condition sequences and path identifiers involves:

[0103] Based on the operating condition segments marked as abrupt change path in the historical operating condition sequence, extract the information on changes in breaking current, arc duration and opening and closing time within the corresponding time range, and determine the operating condition change trajectory within the time range as a short-cycle impact trajectory.

[0104] Based on the operating condition segments in the historical operating condition sequence that have not been marked with a sudden change path and whose duration exceeds the preset duration, the corresponding load current level, operating frequency and ambient temperature change information are extracted, and the operating condition change trajectory is determined as a long-cycle evolution trajectory.

[0105] Based on the operating condition segments marked as recovery path identifiers in the historical operating condition sequence, and combined with the occurrence time and duration interval of the corresponding maintenance records, the time interval that has a significant impact on the state evolution is determined, and the time interval is determined as the recovery sensitive interval.

[0106] Through the orchestration execution layer, based on the working condition semantic information, the working condition influence trajectory, the orchestration time window set, and the priority sequence, the channel activation order, channel activation intensity, and activation duration for controlling state transitions are determined. Conflict suppression and mutual exclusion constraints are applied within the time windows of channel overlap. Specifically, determining the channel activation order, channel activation intensity, and activation duration for controlling state transitions involves:

[0107] Based on the temporal relationship between the short-cycle impact trajectory, the long-cycle evolution trajectory, and the recovery sensitive interval in the working condition influence trajectory, and combined with the priority sequence, the gradual phase channel, the abrupt phase channel, and the recovery phase channel are sorted to determine the channel activation order, so that the channel activation order is consistent with the working condition evolution time order.

[0108] Based on the proportions of driving semantic information, modulation semantic information and disturbance semantic information in the working condition semantic information, and combined with the change amplitude of the corresponding trajectory segment in the working condition influence trajectory, the corresponding channel activation intensity is assigned to each channel so that the activation intensity of different channels matches the degree of influence of their corresponding working conditions.

[0109] Based on the start and end times and duration of each time window in the time window set, and combined with the corresponding channel's effective range in the historical operating condition sequence, the activation duration of each channel is determined. When multiple channel activation time windows overlap, the channel activation duration is adjusted according to the priority sequence.

[0110] In this embodiment, the generation of energy state prediction results includes:

[0111] A multiphase flow state transition channel is constructed based on the energy state vector. The multiphase flow state transition channel includes an asymptotic phase channel, abrupt phase channel, and a recovery phase channel. A channel energy ledger, a sliding formation window, and a channel memory unit are established to record the channel processing order, channel processing intensity, and channel historical residual.

[0112] According to the channel activation order and channel activation intensity, the asymptotic phase channel is driven to perform piecewise accumulation and temperature fatigue coupling window processing. A monotonic envelope and rate upper limit constraint are applied to the energy state vector to generate the first intermediate energy state vector. Specifically, the piecewise accumulation and temperature fatigue coupling window processing for driving the asymptotic phase channel is as follows:

[0113] Based on the channel activation order and the set of time windows, the energy state vectors corresponding to the current sampling period and its adjacent sampling periods are divided into continuous time segments, and the energy state vectors in each segment are accumulated in time order.

[0114] Within each time segment, contact temperature rise data and ambient temperature data corresponding to the segment are introduced to establish the correspondence between temperature change and energy accumulation, and temperature fatigue coupling processing is performed on the energy state vector after segmented accumulation.

[0115] The energy state vectors of each time segment that have completed the temperature fatigue coupling treatment are merged in chronological order and used as the processing output of the asymptotic phase channel.

[0116] The monotonic envelope constraint refers to restricting the direction of change of each energy component in the energy state vector output by the asymptotic phase channel within a continuous sampling period, so that each energy component is only allowed to change in a predetermined degradation direction, and is not allowed to have a regression change opposite to the historical change trend in adjacent sampling periods.

[0117] The rate upper limit constraint refers to limiting the change amplitude of each energy component in the energy state vector output by the asymptotic phase channel within a continuous sampling period, so that the change of each energy component within a single sampling period does not exceed the maximum allowable change range obtained from the statistics of historical operating conditions.

[0118] The system identifies impact segments triggered by short-circuit interruption and arc duration in the operating condition dataset, selects a mutation processing template based on path identifiers, drives the mutation phase channel to avoid repeated inclusion within the refractory period window, and performs dynamic saturation trimming to generate a second intermediate energy state vector, where:

[0119] The mutation processing template refers to a predefined set of mutation processing rules for the operating condition segments marked as mutation path identifiers in the operating condition state dataset. The set of rules is used to limit the processing method of the energy state vector by the mutation phase channel. The mutation processing template includes the energy increment mapping relationship corresponding to the short circuit breaking current amplitude, arc duration length and opening and closing time change amplitude, and constrains the update order and update amplitude of the corresponding components in the energy state vector according to the combination characteristics of the operating parameters.

[0120] Dynamic saturation pruning is performed as follows: During the processing of the abrupt phase channel, the energy state vector after the abrupt processing template is applied is checked component by component. When the change of energy component in a continuous sampling period exceeds the preset upper limit, the change value of energy component is truncated. Within the refractory period window, the energy increment corresponding to the same short-circuit breaking event is only allowed to be counted once, so as to avoid repeated accumulation of the same impact event in adjacent sampling periods.

[0121] Based on the maintenance records in the operating condition dataset, two types of recovery units are distinguished: lubrication recovery and component replacement recovery. A maximum pullback limit and a gradual entry time window are set in the drive recovery phase channel. The second intermediate energy state vector is processed to generate a third intermediate energy state vector, where:

[0122] The maximum pullback limit refers to the maximum allowable pullback amplitude set for each energy component in the second intermediate energy state vector based on the maintenance record data in the operating condition state dataset during the recovery phase channel processing. It is used to limit the upper limit of the pullback of each energy component relative to the state before maintenance during the recovery process. The upper limit of the pullback is taken in different ranges according to the two types of recovery units: lubrication recovery and component replacement recovery.

[0123] The gradual entry time window refers to a continuous time interval set in the recovery phase channel processing, corresponding to the maintenance record data. Within the time interval, the second intermediate energy state vector is subjected to a phased withdrawal process, so that each energy component gradually approaches the maximum withdrawal limit in chronological order, rather than completing the full withdrawal within a single sampling period.

[0124] Within the sliding formation window, conflict suppression and mutual exclusion constraints are applied to the overlapping time windows of the three types of channels. The channel residual buffer of the channel memory unit is enabled to perform sequential processing on the undigested residuals. The channels are merged according to the channel activation order and the order of completion of channel activation intensity to obtain candidate energy state prediction results.

[0125] The channel energy ledger undergoes an energy consistency check. Once a preset threshold is met, the energy state prediction result is confirmed. Specifically, the energy consistency check of the channel energy ledger involves:

[0126] According to the processing order of the progressive phase channel, abrupt phase channel and recovery phase channel recorded in the channel energy ledger, the energy change records of each channel in the current sampling period are summarized to form the channel energy change summary value;

[0127] The summary value of the channel energy change is compared with the actual energy change amplitude between the second intermediate energy state vector and the third intermediate energy state vector, and the deviation between the ledger record energy change and the state vector energy change is calculated.

[0128] When the deviation does not exceed a preset threshold, it is determined that the channel energy ledger record is consistent with the energy state evolution result, and the energy state prediction result of the current sampling period is confirmed.

[0129] In this embodiment, the state change trend and comprehensive health status prediction results of the output circuit breaker include:

[0130] Obtain the energy state prediction result, and call the mapping relationship from energy domain to state domain in the bidirectional mapping relationship between the circuit breaker state domain and energy domain as the basis for the inverse mapping processing from energy domain to state domain;

[0131] Based on the mapping relationship from the energy domain to the state domain, the energy state prediction results are processed by component-by-component mapping according to the component composition, component order and component weight of the energy state vector, and each energy component is converted into the corresponding state change quantity to generate the predicted state vector.

[0132] Based on the definition and structure of the state vector in the digital twin model of a circuit breaker, the predicted state vector is decomposed to obtain the contact wear state quantity, the operating mechanism fatigue state quantity, and the insulation and thermal aging state quantity. Value range constraints and rate of change constraints are applied to each state quantity to form the constrained predicted state vector, where:

[0133] The value range constraint is based on the definition structure of the state vector in the circuit breaker digital twin model. The minimum and maximum value boundaries are set for the contact wear state quantity, the operating mechanism fatigue state quantity, and the insulation and thermal aging state quantity, respectively, so that each state quantity is always limited to an effective range consistent with historical operating data and equipment design parameters during the prediction process.

[0134] The rate of change constraint is based on the statistical characteristics of the changes of each state quantity in the historical operating condition state dataset within adjacent sampling periods. The maximum allowable change range of the contact wear state quantity, the operating mechanism fatigue state quantity, and the insulation and thermal aging state quantity is limited within a unit sampling period, so that each state quantity in the predicted state vector keeps changing continuously between consecutive sampling periods.

[0135] The constrained predicted state vector is re-input into the mapping relationship from the circuit breaker state domain to the energy domain to obtain the corresponding energy remapping result. The consistency of the energy remapping result and the energy state prediction result is compared. When the consistency meets the preset condition, the predicted state vector is confirmed as a valid result.

[0136] The confirmed and valid predicted state vector is used as the current state vector for the next time step. The bidirectional mapping process between the state domain and the energy domain, the operating condition evolution scheduling process, and the multiphase flow state transition process are sequentially input to obtain the state evolution prediction results of the circuit breaker for the next multiple time steps. The state change trend and comprehensive health state prediction results of the circuit breaker are then output.

[0137] Example 1:

[0138] To verify the feasibility of this invention in practice, it was applied to an outdoor high-voltage vacuum circuit breaker in a 220kV substation. The circuit breaker has a rated current of 1250A and a rated short-circuit breaking current of 31.5kA and operates in an environment with frequent load fluctuations. Because the substation is responsible for regulating urban loads and supplying power to industrial users, the circuit breaker experiences both long-term stable operation and periodic alternations of various conditions, including sudden load increases, fault clearing, and planned maintenance. In actual management, the maintenance unit found that simply relying on the number of operations or experience thresholds to judge the circuit breaker's condition is insufficient to accurately reflect its true health level. In particular, changes in condition after short-circuit breaking or maintenance are difficult to quantify and assess, posing a risk of premature or delayed maintenance.

[0139] In this operational scenario, the circuit breaker state evolution prediction method based on a digital twin model proposed in this invention is used to continuously monitor and predict the circuit breaker's operating status. During circuit breaker operation, multi-source operational data, including opening and closing time, opening and closing stroke, opening and closing operation speed, contact rebound characteristics, arc duration, breaking current, contact resistance, contact temperature rise, circuit breaker casing temperature, ambient temperature, load current level, operation frequency, and maintenance records, are collected through the integrated online monitoring system within the station. This data is recorded with a 5-minute sampling period and undergoes time alignment, outlier removal, and standardization to form an operational status dataset for state modeling and analysis.

[0140] Based on this operational condition dataset, a digital twin model of the circuit breaker is constructed, and a state vector is defined in the model to characterize the health status of the circuit breaker. The state vector consists of contact wear state variables, operating mechanism fatigue state variables, and insulation and thermal aging state variables, reflecting the comprehensive state of the circuit breaker in terms of electrical, mechanical, and thermal aging aspects, respectively. By calibrating historical operating data, the state change trend output by the digital twin model is made consistent with the measured opening and closing times, contact temperature rise, and other observations, thus achieving continuous characterization of the circuit breaker's state.

[0141] Based on state modeling, a bidirectional mapping relationship between the state domain and the energy domain is introduced, mapping the current state vector and its historical change paths to an energy state vector. This energy state vector is used to uniformly represent the energy accumulation, sudden impact energy input, and energy withdrawal caused by maintenance during long-term operation of the circuit breaker, enabling different types of state changes to be processed within the same evolution space. Through this mapping method, the state transitions of the circuit breaker under sudden operating conditions such as short-circuit breaking can be accurately captured, while avoiding the problems of state variable discrepancies and path information loss in traditional methods.

[0142] During operation, a condition evolution scheduler is constructed based on the current operating conditions and historical operating condition sequences in the operating condition status dataset. This scheduler comprehensively analyzes the operating conditions at different operating stages, generating semantic information and influence trajectories of the operating conditions, and determines the channel activation sequence and intensity for controlling state transitions accordingly. When the circuit breaker is in a long-term stable operation phase, the scheduler primarily activates the progressive phase channel to reflect the slow degradation process; when a short-circuit breaking condition is detected, the scheduler preferentially activates the abrupt phase channel to characterize the state abrupt change; when maintenance records are found in the operating conditions, the scheduler activates the recovery phase channel to model the state rollback process.

[0143] In the multiphase flow state transition processing, the asymptotic phase channel, the abrupt phase channel, and the recovery phase channel jointly drive the energy state vector according to the activation order and activation intensity given by the scheduler, generating energy state prediction results. Subsequently, based on the bidirectional mapping relationship between the state domain and the energy domain, the energy state prediction results are back-mapped into the future multi-time-step state evolution results of the circuit breaker, and the predicted state vector is recursively used as the input for the next time step, forming a continuous state evolution prediction sequence.

[0144] Table 1. Examples of Circuit Breaker Operating Conditions and State Evolution Data

[0145] Load current (A) Breaking current (A) Opening and closing time (ms) Contact temperature rise (°C) Ambient temperature (°C) Contact wear condition measurement Fatigue status of operating mechanism Insulation and thermal aging status 610 0 47.8 31.6 18.4 0.41 0.46 0.44 645 0 48.3 32.9 24.7 0.43 0.48 0.45 690 8200 50.6 38.5 29.2 0.49 0.54 0.47 620 0 49.2 34.1 22.5 0.47 0.52 0.46 630 0 49.8 35.0 20.1 0.51 0.56 0.48

[0146] As can be seen from the data in Table 1, during the stable operation phase without any interruption operations, the circuit breaker load current remained in the range of 610A to 645A, the opening and closing time remained in the range of 47.8ms to 48.3ms, and the contact temperature rise was in the range of 31.6℃ to 32.9℃. The corresponding contact wear state, operating mechanism fatigue state, and insulation and thermal aging state all showed a slow increasing trend, indicating that under normal load conditions, the circuit breaker state evolved in a gradual manner, which is consistent with the actual law of long-term operation degradation.

[0147] When a short-circuit breaking condition occurs during operation, the load current rises to 690A, and the breaking current reaches 8200A. The opening and closing time increases significantly to 50.6ms, and the contact temperature rises to 38.5℃. At the same time, the contact wear status and the operating mechanism fatigue status show a significant jump, reflecting the transient impact of the sudden condition on the internal state of the circuit breaker. This state change characteristic exhibits typical abrupt behavior during the evolution process.

[0148] After the short circuit was interrupted and normal operation was restored, the load current dropped to 620A to 630A, and the opening and closing time and contact temperature rise decreased accordingly. However, the various state quantities were still higher than the level before the short circuit and continued to rise slowly. This indicates that maintenance or state mitigation has a certain restorative effect on the circuit breaker state, but it cannot completely eliminate the cumulative degradation effect. This further reflects the objective characteristics of the superposition of long-term degradation and sudden impact during the evolution of the circuit breaker state.

[0149] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for predicting the state evolution of circuit breakers based on a digital twin model, characterized in that, include: Collect multi-source operating data during the operation of the circuit breaker, process the multi-source operating data, and form an operating condition status dataset; A digital twin model of a circuit breaker is constructed. The state of the digital twin model of the circuit breaker is calibrated based on the operating condition state dataset. A state vector is defined in the digital twin model of the circuit breaker to represent the health state of the circuit breaker. Based on the current state vector and the historical state change information in the operating condition state dataset, a bidirectional mapping relationship between the circuit breaker state domain and energy domain is established. The current state vector is then subjected to state-energy domain mapping processing to generate an energy state vector. A condition evolution scheduler is constructed. Based on the current operating condition and historical operating condition sequence in the operating condition state dataset, the condition semantic information and condition influence trajectory are generated by the condition evolution scheduler, and the channel activation order and channel activation intensity used to control state transitions are determined. Based on the energy state vector, a multiphase flow state transition channel is constructed, which includes an asymptotic phase channel, abrupt phase channel, and a recovery phase channel. The multiphase flow state transition channel is jointly driven according to the channel activation order and channel activation intensity to generate energy state prediction results. Based on the bidirectional mapping relationship, the energy state prediction results are processed by inverse mapping from the energy domain to the state domain to obtain the state evolution prediction results of the circuit breaker in the future multiple time steps, and output the state change trend and comprehensive health state prediction results of the circuit breaker.

2. The method for predicting the state evolution of a circuit breaker based on a digital twin model according to claim 1, characterized in that, The multi-source operating data includes circuit breaker opening and closing time data, opening and closing stroke data, opening and closing operation speed data, contact rebound characteristic data, arc duration data, breaking current data, contact resistance data, contact temperature rise data, circuit breaker housing temperature data, ambient temperature data, load current level data, operating frequency data, and maintenance record data.

3. The method for predicting the state evolution of a circuit breaker based on a digital twin model according to claim 1, characterized in that, The processing of multi-source operational data includes time synchronization, anomaly removal, and standardization of the multi-source operational data.

4. The method for predicting the state evolution of a circuit breaker based on a digital twin model according to claim 1, characterized in that, The definition of a state vector in the circuit breaker digital twin model to characterize the health state of the circuit breaker includes: In the operating condition status dataset, the multi-source operating data is divided into fields. The breaking current data, arc duration data, load current level data, ambient temperature data, and operating frequency data are set as operating condition input fields. The opening and closing time data, opening and closing stroke data, opening and closing operation speed data, contact rebound characteristic data, contact resistance data, contact temperature rise data, and circuit breaker shell temperature data are set as observation fields. The maintenance record data is set as maintenance event fields. All fields are sorted by time according to a unified sampling period. A digital twin model of a circuit breaker is constructed. The digital twin model of the circuit breaker consists of a state layer, an operating condition input layer, an observation layer, and a parameter set. The state layer contains a state vector composed of contact wear state quantity, operating mechanism fatigue state quantity, and insulation and thermal aging state quantity. The operating condition input layer receives the operating condition input quantity field for each sampling period. The observation layer receives the observation quantity field for the same sampling period. The parameter set records the state update coefficient and the observation mapping coefficient. Within each sampling period, the working condition input layer and the state layer are called. According to the state update rules in the parameter set, the state vector of the previous sampling period is mapped to the working condition input field of the current sampling period to generate the state vector of the current sampling period. The observation layer is called to map the state vector of the current sampling period to generate the observation layer output according to the observation mapping rules. Using the observation fields of the same sampling period in the operating condition status dataset as a reference, the error of the observation layer output is calculated. The initial value of the corresponding component in the status layer is reset in the sampling period triggered by the maintenance event field. The status update coefficient and observation mapping coefficient in the parameter set are iteratively adjusted until the error meets the preset threshold. After the error meets the preset threshold, the parameter set is fixed, the state vector of the current sampling period is output and the state vector sequence is saved.

5. The method for predicting the state evolution of a circuit breaker based on a digital twin model according to claim 1, characterized in that, The generated energy state vector includes: Extract the state vector corresponding to the current sampling period and the historical state change information aligned with time from the operating condition status dataset. Generate a historical state change sequence based on the historical state change information. Extract the breaking current data, arc duration data, load current level data, ambient temperature data and operating frequency data within the same time range as the operating condition driving sequence. A bidirectional mapping relationship is established between the circuit breaker's state domain and energy domain. This bidirectional mapping relationship includes a forward mapping from the state domain to the energy domain and a reverse mapping from the energy domain to the state domain, wherein: The positive mapping relationship is determined by the state vector and the operating condition driving sequence, which together determine the composition, order, and weight of the components of the energy state vector. The reverse mapping relationship reconstructs the corresponding state vector and corrects the state evolution hysteresis characteristics based on the component composition, component order and component weight, combined with the corresponding working condition driving sequence and historical state change sequence. When establishing a positive mapping relationship, the state vector is processed by path identification based on the historical state change sequence to generate path identifiers to distinguish between degenerate paths and recovery paths. The path identifiers are used as input items for the positive mapping relationship. Based on the positive mapping relationship, the state vector of the current sampling period is processed to map from state to energy domain. The energy state vector is output according to the composition and order of the components of the energy state vector, and the energy state vector is bound and stored with the path identifier. The energy state vector is converted into the inverse mapping result of the state vector based on the inverse mapping relationship. The consistency of the inverse mapping result with the state vector of the current sampling period is checked. After the consistency check is passed, the bidirectional mapping relationship is fixed and the energy state vector is output.

6. The method for predicting the state evolution of a circuit breaker based on a digital twin model according to claim 1, characterized in that, The determination of the channel activation order and channel activation intensity used to control state transitions includes: Extract the current operating condition and historical operating condition sequences from the operating condition status dataset, complete time alignment and unified sampling, and form the operating condition input set corresponding to the current sampling period; A working condition evolution scheduler is constructed, which consists of three layers: a semantic deconstruction layer, an influence trajectory planning layer, and an orchestration execution layer. The semantic deconstruction layer receives the working condition input set and outputs the working condition semantic information and path identifier. The influence trajectory planning layer outputs the working condition influence trajectory and the orchestration time window set based on the historical operating working condition sequence. The orchestration execution layer outputs the channel activation order and channel activation intensity based on the working condition semantic information and the working condition influence trajectory. The semantic deconstruction layer performs semantic parsing on the current operating condition, generating operating condition semantic information composed of driving semantic information, modulation semantic information, and disturbance semantic information. Based on the maintenance records and short circuit interruption records, recovery path identifiers and sudden change path identifiers are generated for the corresponding operating condition segments. Through the influence trajectory planning layer, short-cycle impact trajectories, long-cycle evolution trajectories, and recovery sensitive intervals are generated based on historical operating condition sequences and path identifiers, and corresponding time window sets and priority sequences are generated. Through the orchestration execution layer, based on the semantic information of the working condition, the trajectory of the working condition influence, the set of orchestration time windows and the priority sequence, the channel activation order, channel activation intensity and activation duration for controlling state transitions are determined, and conflict suppression and mutual exclusion constraints are applied within the time window of channel overlap.

7. The method for predicting the state evolution of a circuit breaker based on a digital twin model according to claim 1, characterized in that, The generated energy state prediction results include: A multiphase flow state transition channel is constructed based on the energy state vector. The multiphase flow state transition channel includes an asymptotic phase channel, abrupt phase channel, and a recovery phase channel. A channel energy ledger, a sliding formation window, and a channel memory unit are established to record the channel processing order, channel processing intensity, and channel historical residual. According to the channel activation order and channel activation intensity, the progressive phase channel is driven to perform piecewise accumulation and temperature fatigue coupled window processing, and a monotonic envelope and rate upper limit constraint are applied to the energy state vector to generate the first intermediate energy state vector. Identify the impact segments triggered by short-circuit interruption and arc duration in the operating condition state dataset, select the mutation processing template according to the path identifier, drive the mutation phase channel to avoid repeated inclusion within the refractory period window and perform dynamic saturation clipping, and generate a second intermediate energy state vector. Based on the maintenance records in the operating condition dataset, two types of recovery units are distinguished: lubrication recovery and component replacement recovery. The maximum pullback limit and gradual entry time window are set in the drive recovery phase channel to perform recovery processing on the second intermediate energy state vector and generate the third intermediate energy state vector. Within the sliding formation window, conflict suppression and mutual exclusion constraints are applied to the overlapping time windows of the three types of channels. The channel residual buffer of the channel memory unit is enabled to perform sequential processing on the undigested residuals. The channels are merged according to the channel activation order and the order of completion of channel activation intensity to obtain candidate energy state prediction results. The energy consistency of the channel energy ledger is checked, and the energy state prediction result is confirmed after the preset threshold is met.

8. The method for predicting the state evolution of a circuit breaker based on a digital twin model according to claim 1, characterized in that, The state change trend and comprehensive health status prediction results of the output circuit breaker include: Obtain the energy state prediction result, and call the mapping relationship from energy domain to state domain in the bidirectional mapping relationship between the circuit breaker state domain and energy domain as the basis for the inverse mapping processing from energy domain to state domain; Based on the mapping relationship from the energy domain to the state domain, the energy state prediction results are processed by component-by-component mapping according to the component composition, component order and component weight of the energy state vector, and each energy component is converted into the corresponding state change quantity to generate the predicted state vector. According to the definition structure of the state vector in the circuit breaker digital twin model, the predicted state vector is decomposed to obtain the contact wear state quantity, the operating mechanism fatigue state quantity, and the insulation and thermal aging state quantity. The value range constraint and the rate of change constraint are applied to each state quantity to form the constrained predicted state vector. The constrained predicted state vector is re-input into the mapping relationship from the circuit breaker state domain to the energy domain to obtain the corresponding energy remapping result. The consistency of the energy remapping result and the energy state prediction result is compared. When the consistency meets the preset condition, the predicted state vector is confirmed as a valid result. The confirmed and valid predicted state vector is used as the current state vector for the next time step. The bidirectional mapping process between the state domain and the energy domain, the operating condition evolution scheduling process, and the multiphase flow state transition process are sequentially input to obtain the state evolution prediction results of the circuit breaker for the next multiple time steps. The state change trend and comprehensive health state prediction results of the circuit breaker are then output.

Images