CN121055320A - Method and system for improving self-healing capacity of expressway micro-grid in extreme weather - Google Patents

Abstract

This invention proposes a method and system for enhancing the self-healing capability of highway microgrids under extreme weather conditions, relating to the field of power system automation technology. The method involves deploying sensors to collect real-time operational data, which is then fused with meteorological data to generate an integrated system state dataset. A digital twin engine generates system state prediction data through state estimation and dynamic simulation. A stochastic multi-objective optimization model solves this data to generate preventative scheduling commands, controlling the adjustment of operating points for generation, energy storage, and load units. Fault detection and identification are performed through residual analysis, and a sequence decision model is used to generate a sequence of self-healing control commands to control switching equipment to isolate faults and restore power supply. This achieves closed-loop management from prediction and scheduling to self-healing, improving the power supply reliability and rapid recovery capability of highway microgrids under extreme weather conditions.

Description

A method and system for enhancing the self-healing capability of highway microgrids under extreme weather conditions

Technical Field

This invention relates to the field of power system automation technology, and in particular to a method and system for enhancing the self-healing capability of highway microgrids under extreme weather conditions.

Background Technology

With the increasing frequency and intensity of extreme weather events, highway microgrids, as critical transportation energy infrastructure, face unprecedented operational risks. Traditional power grid architectures exhibit significant shortcomings in responding to extreme weather. For example, under the pressure of combined weather events such as heat waves or blizzards, system recovery times are excessively long, power supply reliability declines sharply, and large-scale power outages can even occur. While advanced technologies such as digital twins have made progress in some system components, they are often limited to isolated applications, lacking real-time synchronization and coordinated control of the overall operational status of highway microgrids. Specifically, traditional methods struggle to effectively integrate real-time operational data with meteorological forecasts, resulting in insufficient preventative scheduling and fault self-healing capabilities. This leads to an inability to respond and recover quickly during extreme events, causing severe energy outages and economic losses. Furthermore, existing systems suffer from a disconnect in the coordinated optimization between physical infrastructure and virtual models, failing to achieve closed-loop management from forecasting and scheduling to self-healing, thus hindering the comprehensive improvement of microgrid resilience.

Therefore, there is an urgent need for a method to enhance the self-healing capability of highway microgrids that can achieve full-process collaboration and dynamically adapt to extreme weather.

Summary of the Invention

To address the aforementioned problems in existing technologies, the first aspect of this invention proposes a method for enhancing the self-healing capability of highway microgrids under extreme weather conditions, comprising:

S1. Based on real-time operational data collected by sensors deployed in highway microgrids and weather forecast data input from external meteorological monitoring systems, the data fusion layer performs time alignment, redundancy removal, and consistency verification to generate an integrated system status dataset containing system electrical status, equipment operating status, and environmental status.

S2. Based on the integrated system state dataset, the unmeasurable state of the system is estimated by the state estimation algorithm in the digital twin engine, and dynamic simulation is performed by combining the built-in physical model and the input climate prediction data to generate system state prediction data including power balance, node voltage and frequency stability for a future preset period.

S3. Based on the system state prediction data, a stochastic multi-objective optimization model is used to solve the problem and generate preventive dispatch instructions for pre-adjusting the system operating point. The preventive dispatch instructions include the power setpoints of the power generation unit, energy storage unit and controllable load unit.

S4. Send preventive dispatch instructions to the physical execution unit of the highway microgrid, control the physical execution unit to adjust the active and reactive power output, and collect system status feedback data after the execution of preventive dispatch instructions based on sensors deployed near the physical execution unit;

S5. Based on system status feedback data, fault detection and identification are performed through residual analysis of the predicted data and real-time operating data by the digital twin engine. When a fault is confirmed, a sequence decision model is used to calculate and generate a self-healing control command sequence for isolating the fault and restoring power supply.

S6. The self-healing control command sequence is sent to the circuit breakers, contactors and power converters in the highway microgrid for execution, in order to isolate the faulty area and restore power supply to the non-faulty area.

Compared with the prior art, the beneficial effects of the present invention are as follows:

First, in S1, based on real-time operational data collected by sensors deployed in the highway microgrid and weather forecast data input from an external meteorological monitoring system, a data fusion layer performs time alignment, redundancy removal, and consistency verification to generate an integrated system state dataset. This ensures the accuracy and consistency of the data source, laying a reliable foundation for subsequent analysis. Next, in S2, based on the integrated system state dataset, the state estimation algorithm in the digital twin engine estimates the unmeasurable state of the system. Combined with the built-in physical model and input climate prediction data, dynamic simulation is performed to generate system state prediction data for a future preset time period, including power balance, node voltage, and frequency stability. This achieves advanced perception of the impact of extreme weather and prediction of system behavior. In S3, based on the system state prediction data, a stochastic multi-objective optimization model is used to generate preventative scheduling instructions for pre-adjusting system operating points. These instructions include power setpoints for generation units, energy storage units, and controllable load units, thereby optimizing resource allocation and reducing operational risks before climate events occur.

Then, in S4, preventative dispatch commands are issued to the physical execution units of the highway microgrid. These units adjust active and reactive power output and collect system status feedback data after the execution of the preventative dispatch commands based on sensors deployed near the physical execution units, forming a closed-loop control system to ensure accurate execution and real-time feedback of the dispatch commands. In S5, based on the system status feedback data, fault detection and identification are performed through residual analysis of the predicted data from the digital twin engine and the real-time operating data. Once a fault is confirmed, a sequence decision model is used to calculate and generate a self-healing control command sequence for fault isolation and power restoration, achieving rapid and accurate fault response. Finally, in S6, the self-healing control command sequence is issued to the circuit breakers, contactors, and power converters in the highway microgrid for execution, isolating the faulty area and restoring power to the non-faulty area, significantly shortening the power outage time.

The entire process, through the synergistic effect of data fusion, predictive optimization, instruction execution, and feedback control, achieves a full-chain resilience improvement from prevention to self-healing, effectively reducing fault recovery time, lowering operating costs, and maintaining a high system resilience index, thus solving the problems of slow response, poor coordination, and insufficient recovery capability of traditional methods.

Attached Figure Description

To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

Figure 1 is a flowchart illustrating a method for enhancing the self-healing capability of a highway microgrid under extreme weather conditions, according to an embodiment of the present invention.

Figure 2 shows a schematic diagram of a self-healing capability enhancement system for highway microgrids under extreme weather conditions, provided by an embodiment of the present invention.

Detailed Implementation

To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

The specific embodiments of the present invention will be described below.

Example 1

As shown in Figure 1, this invention proposes a method for improving the self-healing capability of highway microgrids under extreme weather conditions, including:

S1. Based on real-time operational data collected by sensors deployed in highway microgrids and weather forecast data input from external meteorological monitoring systems, the data fusion layer performs time alignment, redundancy removal, and consistency verification to generate an integrated system status dataset containing system electrical status, equipment operating status, and environmental status.

S2. Based on the integrated system state dataset, the unmeasurable state of the system is estimated by the state estimation algorithm in the digital twin engine, and dynamic simulation is performed by combining the built-in physical model and the input climate prediction data to generate system state prediction data including power balance, node voltage and frequency stability for a future preset period.

S3. Based on the system state prediction data, a stochastic multi-objective optimization model is used to solve the problem and generate preventive dispatch instructions for pre-adjusting the system operating point. The preventive dispatch instructions include the power setpoints of the power generation unit, energy storage unit and controllable load unit.

S4. Send preventive dispatch instructions to the physical execution unit of the highway microgrid, control the physical execution unit to adjust the active and reactive power output, and collect system status feedback data after the execution of preventive dispatch instructions based on sensors deployed near the physical execution unit;

S5. Based on system status feedback data, fault detection and identification are performed through residual analysis of the predicted data and real-time operating data by the digital twin engine. When a fault is confirmed, a sequence decision model is used to calculate and generate a self-healing control command sequence for isolating the fault and restoring power supply.

S6. The self-healing control command sequence is sent to the circuit breakers, contactors and power converters in the highway microgrid for execution, in order to isolate the faulty area and restore power supply to the non-faulty area.

In its implementation, the highway microgrid achieved enhanced resilience across the entire process under extreme weather events through the close integration of six key steps. First, in step S1, sensors deployed within the highway microgrid continuously collect real-time operational data, including electrical parameters such as voltage, current, and power. Simultaneously, an external meteorological monitoring system inputs weather forecast data, such as wind speed, irradiance, and temperature. The data fusion layer performs time alignment processing on this multi-source heterogeneous data to ensure that information from different sources is synchronized on the timeline. Redundancy is then eliminated to remove duplicate or contradictory data, and outliers are identified and corrected through consistency checks. Finally, an integrated system status dataset is generated. This dataset integrates the system's electrical status, equipment operating status, and environmental status, providing a unified and reliable data foundation for subsequent analysis. The technical effect of this step is to solve the decision delay problem caused by the scattered data sources and inconsistent formats in traditional methods. Through the processing of the data fusion layer, the accuracy and availability of the data are significantly improved, thus laying a solid foundation for the prediction and optimization of the digital twin engine. Time alignment avoids simulation errors caused by time sequence disorder, redundancy elimination reduces the computational burden, and consistency verification identifies data contradictions through rule bases or machine learning models to ensure the reliability of subsequent state estimation. The resulting high-quality dataset directly supports the accurate modeling of system behavior.

In step S2, based on the integrated system state dataset, the state estimation algorithm in the digital twin engine estimates the unmeasurable states of the system, such as inferring node voltage phase angles or equipment health status through filtering techniques. Simultaneously, the engine combines built-in physical models, such as grid topology and energy conservation equations, with input climate prediction data to perform dynamic simulation and generate system state prediction data for a preset future time period, including power balance, node voltage, and frequency stability. The technical advantage of this step lies in achieving advanced perception of the impact of extreme weather and multi-dimensional prediction of system behavior, avoiding the lag of traditional responsive control. The state estimation algorithm uses historical data and real-time measurements to compensate for variables that cannot be directly observed, while the dynamic simulation simulates system evolution under climate disturbances through physical models, thereby identifying potential risk points in advance, such as voltage exceedances or frequency fluctuations, providing forward-looking input for preventative scheduling. Step S3, based on system state prediction data, is solved using a stochastic multi-objective optimization model. This model comprehensively considers factors such as operating costs, risk exposure, and unmet energy needs to generate preventative scheduling instructions for pre-adjusting system operating points. The instructions specifically include power setpoints for power generation units, energy storage units, and controllable load units. The technical effect is to optimize resource allocation, balance economy and reliability before extreme events occur, and reduce system vulnerability. Stochastic optimization handles climate uncertainty through scenario analysis, and the multi-objective function achieves optimality through weight adjustment, thereby generating a scheduling strategy that balances safety and efficiency.

Step S4 issues preventative dispatch commands to the physical execution units of the highway microgrid, controlling them to adjust active and reactive power output, for example, by regulating generation and load through inverters or controllers. Simultaneously, based on system status feedback data collected by sensors deployed near the physical execution units after command execution, a closed-loop control loop is formed. The technical effect is to ensure accurate execution and real-time monitoring of dispatch commands, preventing command deviations from accumulating into system faults; the response of the physical execution units is verified through sensor feedback, and any deviations can be corrected promptly, thus maintaining system stability. Step S5, based on the system status feedback data, performs fault detection and identification through residual analysis of the predicted data from the digital twin engine and the real-time operating data, i.e., comparing the difference between expected and actual values; when a fault is confirmed, a sequence decision model is used to calculate and generate a self-healing control command sequence for fault isolation and power restoration. The technical effect is to achieve rapid and accurate fault diagnosis and recovery strategy generation, significantly shortening outage time; residual analysis identifies anomalies through threshold comparison, while the sequence decision model optimizes the recovery path based on system status and resource constraints, ensuring priority power supply to critical loads. Finally, step S6 sends the self-healing control command sequence to the circuit breakers, contactors, and power converters in the highway microgrid for execution, isolating faulty areas and restoring power to non-faulty areas, completing closed-loop management from prevention to self-healing. The technical benefits lie in reducing human intervention delays through automated execution, improving the overall system availability and resilience; the precise issuance of the command sequence ensures coordinated fault isolation and load restoration, preventing fault propagation, while seamless switching is achieved through the rapid response of power electronic equipment.

Overall, by leveraging the synergistic effects of data fusion, predictive simulation, optimized scheduling, command execution, and feedback control, the resilience of highway microgrids under extreme weather conditions was enhanced across the entire process. This method significantly improves the system's adaptability to climate disturbances, reduces operational risks through preventative scheduling, and rapidly restores power supply via self-healing control, thus addressing the slow response and poor coordination issues of traditional methods. Ultimately, this implementation ensures high system reliability during extreme events while optimizing economics, providing a sustainable operating paradigm for critical infrastructure.

In some implementations, S2 includes:

S21. Based on the node voltage, branch power and generator output data in the integrated system state dataset, a set of coupled differential-algebraic equations describing the power grid topology and the law of energy conservation are numerically solved to generate continuous dynamic estimation data of the system physical state.

S22. Based on the continuous dynamic estimation data of the system physical state and the direct sensor measurement values of the integrated system state data, the data is fused and error corrected by the Kalman filter algorithm with adaptive covariance matrix to generate a system state vector containing the voltage amplitude and phase angle of all nodes.

S23. Based on the system state vector, and incorporating the future wind speed, irradiance and temperature prediction data from the integrated system state dataset as external disturbance input, forward rolling time-domain simulation is performed through the linearized system model in the digital twin engine to generate system state prediction data including power balance, node voltage and frequency stability within a preset future time period.

In step S21, based on the node voltage, branch power, and generator output data in the integrated system state dataset, a set of coupled differential-algebraic equations describing the power grid topology and the law of energy conservation are numerically solved to generate continuous dynamic estimation data of the system's physical state. Specifically, the coupled differential-algebraic equations simulate energy flow and equipment dynamics in the power grid. For example, the power flow equations are solved iteratively using the Newton-Raphson method to obtain the continuous variation trajectory of parameters such as voltage and power. The technical effect is to provide high-precision physical state estimation, laying the foundation for subsequent predictions. Differential equations capture the dynamic behavior of the system, while algebraic equations constrain topological relationships. Numerical solutions avoid errors caused by simplified models, ensuring that the estimated data truly reflects the system's operating status. In step S22, based on the continuous dynamic estimation data of the system's physical state and the direct sensor measurements in the integrated system state dataset, data fusion and error correction are performed using a Kalman filter algorithm with an adaptive covariance matrix. The adaptive covariance matrix is dynamically adjusted according to the noise characteristics of real-time data, for example, through innovative sequence or covariance matching techniques, thereby generating a system state vector containing the voltage amplitude and phase angle of all nodes. The technical benefits lie in improving the robustness and accuracy of state estimation, especially when the data is noisy or the system undergoes sudden changes. Kalman filtering fuses estimation and measurement through a prediction-correction mechanism, while adaptive covariance optimizes the filter gain, reduces estimation bias, and ultimately forms a consistent and reliable state vector.

Step S23, based on the system state vector and incorporating future wind speed, irradiance, and temperature prediction data from the integrated system state dataset as external disturbance input, performs forward rolling time-domain simulation using a linearized system model in the digital twin engine. This generates system state prediction data for a preset future time period, including power balance, node voltage, and frequency stability. The linearized model, for example through small-signal analysis or state-space representation, updates the system response at each time step, simulating the evolution under climate disturbances. The technical effect is to achieve dynamic prediction of extreme weather events and identify system vulnerabilities in advance. The external disturbance input simulates the impact of climate, while the rolling simulation captures transient system behavior through time-domain progression, thereby predicting future states such as power deficits or voltage drops, providing reliable input for optimized scheduling.

Overall, the completeness and reliability of state estimation and prediction are improved through the synergy of coupled equation solving, adaptive filtering, and rolling simulation. This method ensures accurate modeling of the digital twin engine under complex climatic conditions and reduces prediction uncertainty through physical constraints and data fusion. Ultimately, this implementation enhances the system's predictability of climate disturbances, providing solid technical support for preventative scheduling and self-healing control.

In some implementations, S3 includes:

S31. Based on the probability and intensity data of extreme weather occurrence in the system state prediction data, a set of typical climate disturbance scenarios is constructed through scenario generation and reduction technology, and the expected value of system performance loss caused by each scenario is calculated to generate a quantitative climate disturbance risk value.

S32. Based on the climate disturbance risk value, power generation fuel cost and equipment operation and maintenance cost model, and the penalty cost model caused by power outage, construct a weighted summation objective function for a stochastic multi-objective optimization model, wherein the weight coefficients are dynamically adjusted according to the climate warning level;

S33. Under the conditions of satisfying the system power flow equation, generator output upper and lower limits, energy storage charging and discharging rate and line transmission capacity constraints, a stochastic programming algorithm is used to solve the weighted summation objective function to generate preventive scheduling instructions for pre-adjusting the system operating point.

In step S31, based on the probability and intensity data of extreme weather occurrence in the system state prediction data, a set of typical climate disturbance scenarios is constructed using scenario generation and reduction techniques. The expected value of system performance loss caused by each scenario is calculated, generating a quantified climate disturbance risk value. Scenario generation, for example, extracts weather parameters from the probability distribution through Monte Carlo sampling. Reduction techniques reduce the number of scenarios through clustering or importance sampling, retaining representative cases, thereby calculating the probability of load loss or expected equipment damage under each scenario. The technical effect is to transform climate uncertainty into quantifiable risk indicators, providing a basis for optimization. The scenario set covers a variety of possible climate events, and the system vulnerability is comprehensively assessed through expected value calculation, thus avoiding the limitations of a single scenario and ensuring the comprehensiveness of the risk value. Step S32, based on the climate disturbance risk value, the power generation fuel cost and equipment operation and maintenance cost model, and the penalty cost model caused by power outage, constructs a weighted summation objective function for a stochastic multi-objective optimization model. The weight coefficients are dynamically adjusted according to the climate warning level; for example, at a high warning level, the risk weight increases to prioritize safety, while the cost weight decreases accordingly. The technical effect lies in achieving a dynamic balance among multiple objectives, adapting to operational needs under different climatic conditions; weighted summation transforms multiple objectives into single-objective optimization, while dynamic weights are adjusted through rules or feedback mechanisms to ensure that the model focuses on minimizing risk in extreme events, while optimizing economic efficiency under normal conditions.

Step S33, under the constraints of system power flow equations, generator output limits, energy storage charging and discharging rates, and line transmission capacity, employs a stochastic programming algorithm to solve the weighted summation objective function, generating preventative scheduling instructions for pre-adjusting system operating points. Stochastic programming, for example, uses sample averaging or Benders decomposition to handle scenario uncertainties, seeking optimal generation, energy storage, and load setpoints under constraints. The technical effect is the generation of robust and economical scheduling strategies, enhancing the system's resilience to climate disturbances. The constraints ensure the feasibility of the scheduling instructions, while stochastic optimization avoids high-risk decisions through scenario analysis, thus generating instructions that both prevent climate impacts and consider cost.

Overall, climate-adaptive scheduling was achieved through the integration of scenario-based risk quantification, dynamic multi-objective optimization, and stochastic solution. This method significantly reduces operational risks and economic losses during extreme events, and avoids system overload or power outages through proactive optimization. Ultimately, this implementation ensures the stable operation of highway microgrids under variable climate conditions, providing a core optimization mechanism for enhancing resilience.

In some implementations, S31 includes:

S311. Based on long-term historical meteorological data and short-term numerical weather forecast data, a joint probability distribution function of key parameters of extreme weather is generated using the kernel density estimation method;

S312. Based on the network structure and load level in the system state prediction data, calculate the expected load loss probability and capacity of the system under different intensities of climate disturbances through a pre-established system vulnerability curve model;

S313. Based on the joint probability distribution function of key parameters of extreme weather and the expected probability and capacity of system load loss, the expected risk value is calculated through Monte Carlo simulation to generate a quantitative climate disturbance risk value.

In step S311, based on long-term historical meteorological data and short-term numerical weather forecast data, a joint probability distribution function of key parameters for extreme weather is generated using the kernel density estimation method. Kernel density estimation, as a non-parametric statistical method, uses kernel functions to smooth historical data points, thereby fitting the joint distribution of parameters such as wind speed and temperature, avoiding prior assumptions about the distribution form. The technical effect is to provide a more accurate characterization of climate uncertainty and reduce model bias; long-term data captures climate trends, short-term forecasts update real-time information, and kernel density estimation adapts to data sparsity through smoothing, ultimately generating a probability distribution that reflects the statistical characteristics of extreme events. In step S312, based on the network structure and load level in the system state prediction data, the expected load loss probability and capacity of the system under different intensities of climate disturbance are calculated using a pre-established system vulnerability curve model. Vulnerability curves, for example, map climate parameters to system performance losses through historical fault data or simulation calibration, such as high temperatures causing a decrease in photovoltaic efficiency or strong winds causing line faults. The technical effect lies in quantifying the system's sensitivity to climate disturbances, providing specific input for risk analysis; network structure and load level define the system's operating point, while vulnerability curves assess the impact of disturbances through functional relationships, thereby calculating the expected load loss index and identifying high-risk areas.

Step S313 calculates the expected risk value using Monte Carlo simulation based on the joint probability distribution function of key extreme weather parameters and the system's expected load loss probability and capacity, generating a quantified climate disturbance risk value. The Monte Carlo simulation generates numerous scenarios from the joint distribution through random sampling, applies vulnerability curves to each scenario to calculate losses, and finally averages the results to obtain the expected risk. The technical advantage lies in achieving a comprehensive and robust risk assessment, covering the full range of climate uncertainties. The Monte Carlo method reduces sampling errors through numerous repeated simulations, and by combining probability distributions and vulnerability models, it ultimately outputs a comprehensive risk value, providing reliable input for optimized scheduling.

Overall, the synergy of probability distribution modeling, vulnerability assessment, and Monte Carlo simulation enhances the scientific rigor and practicality of risk quantification. This method ensures accurate calculation of climate disturbance risk values and addresses the rarity and variability of extreme events through statistical methods. Ultimately, this implementation provides high-quality risk input for stochastic optimization, enhancing the reliability of the system's decision-making in climate emergencies.

In some implementations, S5 includes:

S51. Based on the voltage and current measurements in the system status feedback data, compare them point by point with the expected normal values generated by the digital twin engine at the corresponding time, and calculate the measurement residual vector.

S52. Based on current meteorological data and system average load rate, dynamically calculate the fault judgment threshold through a linear mapping function, and compare the measurement residual vector with the dynamic fault judgment threshold. When the residual continuously exceeds the threshold, generate a set of fault information containing fault component identification and type.

S53. Based on the fault information set and the current remaining energy storage capacity and adjustable generator standby capacity, a partially observable Markov decision process model is constructed. The immediate reward function of the Markov decision process model is positively correlated with the importance and quantity of the load restored in each decision step.

S54. The partially observable Markov decision process model is solved using a value iteration algorithm to generate a sequence of control actions, which constitutes a self-healing control command sequence for isolating faults and restoring power supply.

In step S51, based on the voltage and current measurements in the system status feedback data, a point-by-point comparison is made with the expected normal values generated by the digital twin engine at the corresponding time, and a measurement residual vector is calculated. The expected normal values come from the simulation output of the digital twin, and the residual vector quantifies the deviation between the actual and expected values through difference calculation. The technical effect is that it realizes the preliminary location of real-time fault detection, providing a data foundation for subsequent diagnosis; the point-by-point comparison ensures fine-grained detection, and the residual vector captures abnormal patterns, thereby identifying potential fault points, such as voltage drops or current overloads. In step S52, based on the current meteorological conditions and the system's average load rate, a fault judgment threshold is dynamically calculated through a linear mapping function, and the measurement residual vector is compared with the dynamic fault judgment threshold. When the residual continuously exceeds the threshold, a fault information set containing the fault component identification and type is generated; the linear mapping function maps parameters such as temperature and load rate to threshold adjustment amounts, and the dynamic threshold adapts to environmental changes. The technical benefits include improved accuracy and adaptability of fault detection, and reduced false alarms and missed alarms. Dynamic thresholds are adjusted by environmental factors to avoid the failure of fixed thresholds under extreme conditions. Continuous over-threshold judgment confirms the fault through time series analysis, and finally generates detailed fault information to support decision-making.

Step S53 constructs a partially observable Markov decision process model based on the fault information set, the current remaining energy storage capacity, and the adjustable generator reserve capacity. The immediate reward function of this model is positively correlated with the importance and quantity of the loads restored in each decision step. Partial observability refers to the fact that the system state is not fully known and needs to be inferred through observation. The Markov decision process models state transitions and rewards to optimize long-term recovery performance. The technical effect is the generation of an intelligent self-healing strategy that prioritizes critical loads. Fault information and resource constraints define the state space, and the reward function reflects load priority through weights, thereby guiding the model to select the optimal recovery action. Step S54 uses a value iteration algorithm to solve the partially observable Markov decision process model, generating a sequence of control actions that constitutes a self-healing control command sequence for fault isolation and power restoration. Value iteration updates the state value function iteratively, converging to the optimal strategy, and outputting sequences such as circuit breaker operation or power adjustment. The technical effect is the realization of efficient and reliable self-healing decision-making, shortening recovery time. Value iteration processes partial observability through belief states, optimizes reward accumulation, and ultimately generates an executable command sequence.

Overall, by integrating residual analysis, dynamic thresholding, decision modeling, and value iteration, the intelligence level of fault response and recovery is improved. This method ensures the speed and accuracy of self-healing control, adapting to varying operating conditions through an adaptive mechanism. Ultimately, this implementation significantly enhances the system's self-healing capability in extreme events, providing key technical support for continuous power supply to highway microgrids.

In some implementations, the dynamic calculation of the fault determination threshold in S52 includes:

S521. Based on the real-time temperature, humidity and wind speed data in the integrated system status dataset, the basic threshold adjustment amount is generated by looking up the table through the preset climate stress coefficient table.

S522. Based on the predicted regional load rate in the system status prediction data, generate an additional threshold adjustment amount related to the load level through a linear proportional relationship;

S523. Add the pre-set normal operating condition baseline threshold, basic threshold adjustment amount and additional threshold adjustment amount to obtain the dynamic fault judgment threshold.

In step S521, based on real-time temperature, humidity, and wind speed data in the integrated system status dataset, a basic threshold adjustment is generated by looking up a preset climate stress coefficient table. The climate stress coefficient table, established through historical data analysis or experimental calibration, maps different climate parameters to threshold adjustment coefficients; for example, high temperature or high humidity corresponds to a higher adjustment, reflecting the impact of environmental stress on system stability. The technical effect is that fault detection can adapt to changes in climate conditions, avoiding the failure of fixed thresholds under extreme weather conditions. The climate stress coefficient table quantifies the impact of environmental factors on equipment performance; for example, high temperature may increase line resistance or equipment failure rate, thereby dynamically adjusting the threshold to match the current risk level and ensuring that detection sensitivity is positively correlated with the severity of the climate. In step S522, based on the predicted regional load rate in the system status prediction data, an additional threshold adjustment related to the load level is generated through a linear proportional relationship. This linear proportional relationship, for example, maps the load rate to the adjustment amount through a function; the threshold is increased at high load rates to tolerate normal fluctuations, and decreased at low load rates to capture subtle anomalies. The technical effect lies in optimizing the detection threshold based on the system load status, reducing false alarms caused by load fluctuations. The load rate directly affects the system's operating stress; high load may mask fault signals, while low load makes the system more sensitive. By dynamically adjusting the threshold in a linear proportion, detection accuracy and stability can be balanced, avoiding excessive alarms during peak load periods or missed alarms during low load periods. Step S523 adds the pre-set normal operating condition baseline threshold, the basic threshold adjustment amount, and the additional threshold adjustment amount to obtain the dynamic fault judgment threshold. The normal operating condition baseline threshold is determined based on system design parameters or historical operating data, while the dynamic adjustment amount is calculated through the aforementioned steps. The final threshold integrates climate and load factors. The technical effect is that it realizes multi-factor collaborative threshold calculation, improving the robustness and reliability of fault detection. The baseline threshold provides a basic reference, while climate and load adjustments introduce real-time adaptability. By fusing multi-source information through additive operations, it ensures that the threshold can accurately distinguish between normal fluctuations and real faults under changing conditions, thereby supporting the accuracy of residual analysis.

Overall, by synergistically combining climate stress lookup, load ratio adjustment, and threshold fusion, the fault detection system can dynamically respond to changes in the environment and operating status. This method significantly reduces the probability of false alarms and false negatives, and ensures that the detection threshold remains optimal under extreme climate and load fluctuations through an adaptive mechanism. Ultimately, this implementation enhances the accuracy of fault identification, provides reliable input for self-healing control, and thus improves the overall resilience of the system under complex operating conditions.

In some implementations, the design of the immediate reward function in S53 includes:

S531. Based on the pre-classification level of loads in highway microgrids, the highest reward coefficient is assigned to key loads such as traffic lights, emergency lighting, and communication base stations;

S532. In the instant reward function, the weighted sum of the newly restored load power after each control action is executed is used as the positive reward item, where the weight is the reward coefficient corresponding to the load;

S533. In the immediate reward function, a penalty term is introduced as a negative reward term for the energy storage unit's discharge depth exceeding the safety limit and the distributed generator's overload operation;

S534. Add the positive reward items and the negative reward items to form an instant reward function that is positively correlated with the importance and quantity of the restored load, while constraining the safe operation of the equipment.

In step S531, based on the pre-classification level of loads in the highway microgrid, the highest reward coefficients are assigned to critical loads such as traffic lights, emergency lighting, and communication base stations. The load classification level is determined through system planning or operational strategies, for example, by classifying loads according to their importance to traffic safety and public safety, with critical loads assigned higher coefficients for priority protection. The technical effect is to ensure that the recovery strategy prioritizes critical infrastructure, enhancing the social and economic value of the system in emergency situations. Through pre-classification and coefficient allocation, the reward function naturally favors important loads in decision-making, thereby maximizing recovery benefits when resources are limited and avoiding the loss of key infrastructure due to balanced recovery. In step S532, in the immediate reward function, the weighted sum of the newly restored load power after each control action is used as a positive reward item, where the weight is the reward coefficient corresponding to the load. The weighted sum is calculated through mathematical operations, such as multiplying the restored power value by the coefficient and then summing the results, quantifying the positive contribution of each action. The technical effect lies in quantifying the recovery progress and guiding the decision-making model towards efficient recovery. Positive reward items are directly related to the quantity and importance of the restored loads, and through weighted summation, they reflect comprehensive value, thereby incentivizing the model to select actions that can quickly restore high-priority loads and accelerate system function reconstruction. Step S533 introduces penalty items as negative reward items into the immediate reward function for energy storage unit discharge depth exceeding safety limits and distributed generator overload operation. Safety limits are set based on equipment specifications or operating procedures, and penalty items are represented by negative values; for example, rewards are deducted when an action causes excessive energy storage discharge or generator overload. The technical effect is to constrain equipment operation within a safe range, preventing secondary failures caused by the self-healing process. Negative reward items, through risk avoidance mechanisms, suppress actions that may damage equipment or system stability, ensuring that the recovery strategy pursues speed without sacrificing long-term reliability. Step S534 adds the positive and negative reward items to form an immediate reward function that is positively correlated with the importance and quantity of the restored loads while constraining the safe operation of the equipment. The addition operation integrates positive incentives and negative constraints to form a comprehensive reward value. The technical effect lies in achieving a multi-objective balanced reward design, guiding the decision-making model to generate a safe and efficient self-healing strategy; through reward summation, the model considers both recovery benefits and equipment safety during the optimization process, thereby avoiding the one-sided pursuit of rapid recovery while ignoring system integrity, and ultimately generating a sustainable control sequence.

Overall, an intelligent reward mechanism is constructed by integrating load grading, weighted rewards, safety penalties, and function fusion. This method ensures that self-healing decisions prioritize the restoration of critical loads while strictly adhering to equipment operating constraints, thereby improving the safety and efficiency of the recovery process. Finally, this implementation guides the model to generate optimal strategies through a reward function, significantly enhancing the system's self-healing capability and overall resilience in the event of failures.

In some implementations, a model enhancement step is included after S6:

S7. Based on the actual system response data collected during the execution of the self-healing control command sequence, compare it with the expected response data obtained by the digital twin engine simulation under the same input, and calculate the model prediction error vector;

S8. Based on the model prediction error vector, the key parameter matrix of the state estimation algorithm in the digital twin engine is identified and updated online using the gradient descent algorithm;

S9. Load the updated key parameter matrix into the digital twin engine for subsequent state estimation and dynamic simulation.

In step S7, the actual system response data collected during the execution of the self-healing control command sequence is compared with the expected response data obtained by the digital twin engine simulation under the same input to calculate the model prediction error vector. The actual system response data comes from physical sensor measurements, while the expected response data comes from the digital twin simulation. The error vector quantifies the deviation between the actual and predicted values through interpolation, such as differences in voltage or power values. The technical effect is to provide direct feedback on model accuracy, providing a basis for parameter updates. By comparing the actual and expected data, the error vector captures the mismatch between the model and the physical system, such as model drift caused by equipment aging or environmental changes, thereby identifying the links that need correction. In step S8, based on the model prediction error vector, the key parameter matrix of the state estimation algorithm in the digital twin engine is identified and updated online using the gradient descent algorithm. The gradient descent algorithm iteratively adjusts parameters to minimize the error function, such as updating the covariance matrix or state transition matrix in the Kalman filter, thereby reducing prediction bias. The technical effect lies in achieving dynamic model calibration, improving the long-term accuracy of the digital twin; gradient descent optimizes parameters through error feedback, enabling the model to gradually adapt to system changes, such as the evolution of climate stress or load patterns, thereby maintaining the synchronization between simulation and reality and avoiding decision failures caused by accumulated errors. Step S9 loads the updated key parameter matrix into the digital twin engine for subsequent state estimation and dynamic simulation derivation; parameter loading is achieved through software interfaces or memory updates, ensuring that new parameters take effect immediately. The technical effect is the completion of a learning loop, enabling the digital twin to have continuous evolution capabilities; through periodic or event-driven parameter updates, the model can absorb the latest operational experience, thereby performing better in subsequent predictions and optimizations, ultimately improving the adaptability and reliability of the entire framework.

Overall, a self-improvement mechanism for the digital twin model was established through the coordinated efforts of error calculation, gradient optimization, and parameter loading. This method significantly reduces the uncertainty of model predictions and enhances the accuracy of preventative scheduling and self-healing control by adapting to dynamic changes in the system through online learning. Ultimately, this implementation ensures that the digital twin framework maintains high fidelity during long-term operation, providing sustainable technical support for the resilient management of highway microgrids.

In some implementations, the control process of the physical execution unit in S4 includes:

S41. A power generation unit, including a photovoltaic inverter and a wind turbine generator, receives the active power setpoint in the preventive dispatch command and adjusts the switching state of the power semiconductor devices through the internal control loop of the power generation unit to track the active power setpoint.

S42. Energy storage unit, including a bidirectional DC-AC converter of a battery energy storage system, receives charging and discharging power commands in preventive dispatch commands, and changes the direction and magnitude of power flow by adjusting the modulation wave signal of the energy storage unit;

S43. Controllable load unit, including the controller of electric vehicle charging pile, receives load power adjustment instructions in the preventive scheduling instructions, and adjusts the output current of the charging pile or temporarily interrupts the charging process through the communication interface.

In step S41, the power generation unit, including a photovoltaic inverter and a wind turbine generator, receives the active power setpoint from the preventive dispatch command and adjusts the switching state of the power semiconductor devices through the internal control loop of the power generation unit to track the active power setpoint. The internal control loop, for example, uses pulse width modulation or maximum power point tracking technology to change the output power by adjusting the switching frequency or duty cycle to match the setpoint. The technical effect is to achieve rapid and accurate adjustment of power generation, supporting system power balance. The power generation unit responds to the command through power electronic control, such as adjusting the DC-AC conversion of the inverter, to ensure that the output of renewable energy is consistent with the dispatch target, thereby maintaining stable power supply during climate events. In step S42, the energy storage unit, including a bidirectional DC-AC converter of a battery energy storage system, receives the charging and discharging power command from the preventive dispatch command and changes the direction and magnitude of power flow by adjusting the modulation wave signal of the energy storage unit. The modulation wave signal controls the switching on and off of the converter, for example, by adjusting the voltage or current reference value to achieve switching between charging and discharging modes. The technical benefits lie in the flexible management of energy storage and release, buffering the intermittency of renewable energy and load fluctuations. The energy storage unit responds to dispatch commands bidirectionally, such as charging during periods of excess power generation and discharging during peak demand, thereby smoothing the power curve and improving system reliability. In step S43, the controllable load unit, including the controller of the electric vehicle charging pile, receives load power adjustment commands from the preventative dispatch instructions and adjusts the output current of the charging pile or temporarily interrupts the charging process through a communication interface. The communication interface may employ a CAN bus or wireless protocol, and the controller reduces load demand by lowering the current or pausing operation. The technical benefits include achieving controllable adjustment on the load side, assisting the system in peak shaving and valley filling. The controllable load unit participates in dispatch through a demand response mechanism, such as suspending non-critical charging during extreme weather, thereby reducing system pressure and optimizing resource allocation.

Overall, by coordinating the control of power generation, energy storage, and load units, preventative dispatch commands are translated into concrete physical actions. This method ensures accurate execution of commands and real-time adjustment of system status, achieving rapid response through power electronics and communication technologies. Ultimately, this implementation forms a closed loop from virtual dispatch to physical execution, significantly improving the operational stability and resilience of highway microgrids under extreme weather conditions.

Example 2

As shown in Figure 2, in a second aspect, the present invention proposes a system for enhancing the self-healing capability of highway microgrids under extreme weather conditions. The system employs the method provided in any of the above embodiments and includes:

The data fusion and preprocessing module is used to generate an integrated system status dataset that includes system electrical status, equipment operating status and environmental status based on real-time operating data collected by sensors deployed in the highway microgrid and weather forecast data input from the external meteorological monitoring system, through time alignment, redundancy removal and consistency verification by the data fusion layer.

The digital twin engine module is used to estimate the unmeasurable state of the system based on the integrated system state dataset, through the state estimation algorithm in the digital twin engine, and to perform dynamic simulation and deduction by combining the built-in physical model and the input climate prediction data, to generate system state prediction data including power balance, node voltage and frequency stability for a future preset period.

The preventive dispatch optimization decision module is used to generate preventive dispatch instructions for pre-adjusting the system operating point by solving a stochastic multi-objective optimization model based on system state prediction data. The preventive dispatch instructions include power setpoints for power generation units, energy storage units, and controllable load units.

The instruction issuance and execution feedback module is used to issue preventive dispatch instructions to the physical execution units of the highway microgrid, control the physical execution units to adjust the active and reactive power output, and collect system status feedback data after the execution of preventive dispatch instructions based on sensors deployed near the physical execution units.

The fault detection and self-healing decision module is used to detect and identify faults based on system status feedback data and residual analysis of predicted data and real-time operating data by a digital twin engine. When a fault is confirmed, a sequence decision model is used to calculate and generate a sequence of self-healing control commands for isolating the fault and restoring power supply.

The self-healing control command execution module is used to send the self-healing control command sequence to the circuit breakers, contactors and power converters in the highway microgrid for execution, so as to isolate the faulty area and restore the power supply to the non-faulty area.

This system corresponds to the method provided in Embodiment 1 above, and will not be described in detail here.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the technical solutions of the embodiments of the present invention.

Claims (10)

1. The method for improving the self-healing capacity of the expressway micro-grid in extreme weather is characterized by comprising the following steps of:

S1, performing time alignment, redundancy elimination and consistency verification processing through a data fusion layer based on real-time operation data acquired by sensors deployed in a highway micro-grid and weather forecast data input by an external weather monitoring system, and generating an integrated system state data set containing a system electrical state, a device operation state and an environment state;

S2, estimating the undetectable state of the system through a state estimation algorithm in the digital twin engine based on an integrated system state data set, and carrying out dynamic simulation deduction by combining a built-in physical model and input climate prediction data to generate system state prediction data comprising power balance, node voltage and frequency stability in a future preset period;

S3, solving through a random multi-objective optimization model based on system state prediction data to generate a preventive scheduling instruction for pre-adjusting a system operation point, wherein the preventive scheduling instruction comprises power set values of a power generation unit, an energy storage unit and a controllable load unit;

S4, issuing the preventive dispatching instruction to a physical execution unit of the highway micro-grid, controlling the physical execution unit to adjust active power output and reactive power output, and acquiring system state feedback data after the preventive dispatching instruction is executed based on a sensor arranged near the physical execution unit;

S5, based on system state feedback data, performing fault detection and identification through residual analysis of pre-estimated data and real-time operation data of the digital twin engine, and after confirming that the fault occurs, performing calculation through a sequence decision model to generate a self-healing control instruction sequence for isolating the fault and recovering power supply;

s6, issuing a self-healing control instruction sequence to a breaker, a contactor and a power converter in the highway micro-grid to execute so as to isolate a fault area and restore power supply of a non-fault area.

2. The method for improving the self-healing capacity of the highway micro-grid under extreme weather according to claim 1, wherein S2 comprises:

S21, carrying out numerical solution through a set of coupled differential-algebraic equations describing the topology of a power grid and the conservation law of energy based on node voltage, branch power and generator output data in the integrated system state data set to generate continuous dynamic estimation data of the physical state of the system;

S22, based on continuous dynamic estimation data of the system physical state and direct measurement values of sensors in the integrated system state data set, carrying out data fusion and error correction through a Kalman filtering algorithm with a self-adaptive covariance matrix, and generating a system state vector comprising voltage amplitudes and phase angles of all nodes;

S23, based on the system state vector, future wind speed, irradiance and air temperature prediction data integrated into an integrated system state data set are used as external disturbance input, forward rolling time domain simulation is carried out through a linearization system model in the digital twin engine, and system state prediction data comprising power balance, node voltage and frequency stability in a future preset period are generated.

3. The method for improving the self-healing capacity of the highway micro-grid under extreme weather according to claim 1, wherein S3 comprises:

S31, constructing a typical climate disturbance scene set by a scene generation and reduction technology based on extreme weather occurrence probability and intensity data in system state prediction data, calculating a system performance loss expected value caused by the typical climate disturbance scene set for each scene, and generating a quantized climate disturbance risk value;

s32, constructing a weighted summation objective function of a random multi-objective optimization model based on a climate disturbance risk value, a power generation fuel cost and equipment operation and maintenance cost model and a punishment cost model caused by power interruption, wherein a weight coefficient is dynamically adjusted along with a climate early warning level;

s33, under the condition that the system tide equation, the upper and lower limits of the generator output, the energy storage charging and discharging rate and the line transmission capacity constraint are met, solving a weighted summation objective function by adopting a random programming algorithm, and generating a preventive scheduling instruction for pre-adjusting the system operating point.

4. The method for improving self-healing capacity of an expressway micro-grid in extreme weather according to claim 3, wherein S31 comprises:

s311, generating a joint probability distribution function of the extreme weather key parameters through a nuclear density estimation method based on long-term historical meteorological data and short-term numerical weather forecast data;

s312, calculating expected load loss probability and capacity of the system under different intensity climate disturbance by a pre-established system vulnerability curve model based on a network structure and load level in system state prediction data;

S313, calculating a risk expected value through Monte Carlo simulation based on a joint probability distribution function of extreme weather key parameters and expected load loss probability and capacity of the system, and generating a quantized climate disturbance risk value.

5. The method for improving the self-healing capacity of the highway micro-grid under extreme weather according to claim 1, wherein S5 comprises:

S51, comparing the voltage and current measured values in the feedback data of the system state with an expected normal value generated by the digital twin engine at the corresponding moment point by point, and calculating to obtain a measurement residual vector;

S52, dynamically calculating a fault judgment threshold value through a linear mapping function based on current meteorological condition data and system average load rate, comparing a measured residual vector with the dynamic fault judgment threshold value, and generating a fault information set containing fault element identifications and types when residual continuously exceeds the threshold;

S53, constructing a partly observable Markov decision process model based on the fault information set, the current energy storage residual capacity and the standby capacity of the adjustable generator, wherein an instant rewarding function of the Markov decision process model is positively related to the load importance degree and the load quantity recovered in each decision step;

S54, solving a part of observable Markov decision process model by adopting a value iterative algorithm, and generating a control action sequence to form a self-healing control instruction sequence for isolating faults and recovering power supply.

6. The method for improving the self-healing capacity of an expressway micro-grid in extreme weather according to claim 5, wherein dynamically calculating the fault determination threshold in S52 comprises:

s521, based on real-time air temperature, humidity and wind speed data in the integrated system state data set, performing table lookup calculation through a preset climate stress coefficient table to generate a basic threshold adjustment quantity;

S522, generating an additional threshold adjustment quantity related to a load level through a linear proportional relation based on a regional load rate predicted value in system state predicted data;

S523, adding the preset normal working condition reference threshold value, the preset basic threshold value adjustment quantity and the additional threshold value adjustment quantity to obtain a dynamic fault judgment threshold value.

7. The method for improving the self-healing capacity of an expressway micro-grid in extreme weather according to claim 5, wherein the designing of the instant prize function in S53 comprises:

S531, distributing the highest rewarding coefficient for key loads of traffic lights, emergency lighting and communication base stations based on the pre-classification level of the loads in the expressway micro-grid;

S532, taking a weighted sum of the load power newly restored after each control action is executed as a forward rewarding item in the instant rewarding function, wherein the weight is a rewarding coefficient corresponding to the load;

s533, introducing a penalty item for overload operation of the distributed generator, which exceeds a safety limit value on the discharge depth of the energy storage unit, as a negative-going reward item in the instant reward function;

s534, adding the positive rewarding item and the negative rewarding item to form an instant rewarding function positively related to the importance degree and the quantity of the restoring load and restraining the safe operation of the equipment.

8. The method for improving the self-healing capacity of an expressway micro-grid in extreme weather according to claim 1, further comprising the step of model enhancement after S6:

S7, comparing actual response data of the system acquired in the execution process of the self-healing control instruction sequence with expected response data obtained by simulation of the digital twin engine under the same input, and calculating to obtain a model prediction error vector;

S8, on-line identification and updating are carried out on key parameter matrixes of a state estimation algorithm in the digital twin engine through a gradient descent algorithm based on the model prediction error vector;

s9, loading the updated key parameter matrix into a digital twin engine for subsequent state estimation and dynamic simulation deduction processes.

9. The method for improving the self-healing capacity of the micro-grid on the expressway in extreme weather according to claim 1, wherein the control process of the physical execution unit in S4 comprises the following steps:

S41, a power generation unit comprises a photovoltaic inverter and a wind generating set, receives an active power set value in a preventive scheduling instruction, adjusts the switching state of a power semiconductor device through an inner control loop of the power generation unit, and tracks the active power set value;

S42, an energy storage unit, which comprises a bidirectional DC-AC converter of a battery energy storage system, receives a charge and discharge power instruction in a preventive scheduling instruction, and changes the direction and the magnitude of power flow by adjusting a modulation wave signal of the energy storage unit;

s43, a controllable load unit comprises a controller of the electric vehicle charging pile, receives a load power adjustment instruction in the preventive scheduling instruction, and adjusts the output current of the charging pile through a communication interface or temporarily interrupts the charging process.

10. A highway micro-grid self-healing capacity improving system in extreme weather, characterized in that the system adopts the method according to any one of claims 1 to 9, the system comprises:

The data fusion and preprocessing module is used for carrying out time alignment, redundancy elimination and consistency check processing through the data fusion layer based on real-time operation data acquired by the sensors deployed in the expressway micro-grid and weather forecast data input by the external weather monitoring system, and generating an integrated system state data set containing a system electrical state, an equipment operation state and an environment state;

the digital twin engine module is used for estimating the unmeasurable state of the system through a state estimation algorithm in the digital twin engine based on the integrated system state data set, and carrying out dynamic simulation deduction by combining a built-in physical model and input climate prediction data to generate system state prediction data comprising power balance, node voltage and frequency stability in a future preset period;

The preventive scheduling optimization decision module is used for solving through a random multi-objective optimization model based on system state prediction data to generate a preventive scheduling instruction for pre-adjusting a system operating point, wherein the preventive scheduling instruction comprises power set values of a power generation unit, an energy storage unit and a controllable load unit;

The instruction issuing and executing feedback module is used for issuing the preventive dispatching instruction to a physical execution unit of the highway micro-grid, controlling the physical execution unit to adjust active and reactive power output, and collecting system state feedback data after the preventive dispatching instruction is executed based on a sensor arranged near the physical execution unit;

The fault detection and self-healing decision module is used for carrying out fault detection and identification through residual analysis of the estimated data and the real-time operation data of the digital twin engine based on the system state feedback data, and calculating through the sequence decision model after confirming that the fault occurs, so as to generate a self-healing control instruction sequence for isolating the fault and recovering the power supply;

And the self-healing control instruction execution module is used for issuing a self-healing control instruction sequence to a breaker, a contactor and a power converter in the highway micro-grid for execution so as to isolate a fault area and restore power supply of a non-fault area.