Power system source-grid-load integrated high-order uncertainty cooperative control method and system based on distributed robust optimization

Abstract

The invention provides an electric power system source network load integrated high-order uncertainty cooperative control method and system based on distributed robust optimization, and relates to the technical field of electric power system optimization control. The method comprises the following steps: forming a multi-layer collaborative risk allocation structure by constructing a hybrid uncertainty envelope with high-order statistical constraints and combining cross-layer risk budget and sensitive mapping; based on the real-time operation data and the distribution sudden change identification indexes, self-calibration and safety order reduction control are implemented, and a recoverable control strategy is obtained; and dynamic updating of uncertainty envelope and risk budget is realized through rolling feedback and incremental learning, so that a distributed robust scheduling system with self-repairing and continuous robustness is constructed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system optimal control, in particular to a power system source-grid-load integrated high-order uncertainty collaborative control method and system based on distribution robust optimization. BACKGROUND

[0002] With the continuous expansion of new energy installed capacity, the operation of the power system presents the characteristics of multi-source heterogeneity, strong uncertainty and high coupling. The current source-grid-load integrated control is mostly based on the deterministic optimization or single-layer robust optimization framework, and only uses low-order statistical indicators such as mean value and variance to describe the renewable energy output and load fluctuation. In actual operation, the asymmetric distribution, peak characteristics of wind and light output and the dependence between source and load make it difficult for traditional prediction and scheduling models to accurately reflect the real operation risk. In addition, the source, grid, load and storage are optimized with independent objectives and constraints, lacking risk linkage mechanism, resulting in fragmented response and insufficient coordination of the system in the event of sudden disturbances.

[0003] In order to balance the economy and safety of the power system, distribution robust optimization has gradually become a new direction for uncertainty modeling of the power system. By constructing a fuzzy distribution set around the historical data, the optimal decision can be sought in the most adverse situation. However, there are still three bottlenecks in existing research: first, the fuzzy set only focuses on low-order moment characteristics, and it is difficult to describe the complex non-Gaussian dependence structure between multiple variables; second, the robust boundary is generally fixed, and cannot be automatically calibrated according to the data drift; third, the cross-layer decision information flow is scattered, and there is a lack of real-time risk coordination channel between source, grid, load and storage. In the event of extreme weather or sudden disturbances, the system often appears over-control or response delay, forming new potential safety hazards.

[0004] Although the existing distribution robust scheduling method enhances the robustness of the system to prediction errors, it still cannot solve the following key problems:

[0005] Uncertainty description distortion, lack of high-order description method for skewness, kurtosis and multi-source dependence characteristics; Robust boundary rigidity, unable to automatically calibrate according to real-time deviation or distribution drift; Risk linkage fragmentation, risk budget and constraints between source, grid, load and storage are not synchronized, making it difficult to form overall coordination; Mutation response lag, in the event of wind and light mutation, there is no reliable mutation detection and safety reduction control mechanism. SUMMARY

[0006] In order to overcome the shortcomings of the prior art, the purpose of the present application is to provide a power system source-grid-load integrated high-order uncertainty collaborative control method and system based on distribution robust optimization, which solves the problems of high-order uncertainty description, lack of adaptive robust boundary, insufficient cross-layer risk coordination and lack of distribution mutation response in the prior art.

[0007] To achieve the above object, the present application provides the following scheme:

[0008] A power system source-network-load integrated high-order uncertainty collaborative control method based on distribution robust optimization, comprising:

[0009] Collecting multi-layer operation data of the power source side, the network side, the load side and the energy storage side, extracting key features reflecting skewness, kurtosis and correlation dependence, and obtaining a high-order dependent feature data set;

[0010] Based on the high-order dependent feature data set, a mixed uncertainty envelope with high-order statistical constraints is constructed to define the credible boundary of the disturbance variable;

[0011] The mixed uncertainty envelope is self-calibrated by using real-time operation deviation, so that the envelope range can be automatically adjusted according to data drift, and a dynamically adjusted uncertainty robust boundary is obtained;

[0012] Under the dynamically adjusted uncertainty robust boundary, a risk budget allocation relationship of the power source side, the network side, the load side and the energy storage side is established, and the risk is linked and balanced between layers through inter-layer sensitivity mapping, and a coordinated cross-layer risk budget result is obtained;

[0013] According to the cross-layer risk budget result, a real-time correction strategy meeting monotonicity and smoothness constraints is generated, and a verified recoverable control strategy is obtained;

[0014] Combined with the verified recoverable control strategy and the cross-layer risk budget result, the collaborative optimization problem is equivalently reconstructed, the structure containing high-order constraints is converted into a solvable form of a single layer, and an operable distribution robust optimization model is constructed;

[0015] According to the operation result of the operable distribution robust optimization model, a distribution mutation identification index is set, when the observation breaks through the uncertainty envelope, a safety reduction response is triggered, and an operation result with sudden disturbance safety reduction capability is obtained;

[0016] The operation result is fed back to the high-order dependent feature data set, and the uncertainty envelope and the risk budget are updated, forming a rolling learning and self-repairing scheduling cycle.

[0017] The present application discloses the following technical effects:

[0018] This invention provides a method and system for integrated high-order uncertainty collaborative control of power system sources, grid, and load based on sub-Brooker robust optimization. The method includes: collecting multi-layer operational data from the power source, grid, load, and energy storage sides; extracting key features reflecting skewness, sharpness, and dependencies to obtain a high-order dependency feature dataset; constructing a hybrid uncertainty envelope with high-order statistical constraints based on the high-order dependency feature dataset to define the credible boundaries of disturbance variables; performing self-calibration of the hybrid uncertainty envelope using real-time operational deviations, enabling the envelope range to automatically expand and contract with data drift, resulting in a dynamically adjusted uncertainty robust boundary; and establishing a risk budget allocation relationship among the power source, grid, load, and energy storage sides under the dynamically adjusted uncertainty robust boundary, and realizing risk allocation between layers through inter-layer sensitive mapping. By linking and balancing, a coordinated cross-layer risk budget result is obtained. Based on the cross-layer risk budget result, a real-time correction strategy satisfying monotonicity and smoothness constraints is generated, resulting in a verified recoverable control strategy. Combining the verified recoverable control strategy with the cross-layer risk budget result, the collaborative optimization problem is equivalently reconstructed, transforming the structure containing high-order constraints into a single-layer solvable form to construct an operable sub-Bruker optimization model. Based on the operating results of the operable sub-Bruker optimization model, a distribution mutation identification index is set. When the observation breaks through the uncertainty envelope, a safety reduction response is triggered, resulting in an operating result with the ability to reduce the order of sudden disturbances. The operating result is fed back to the high-order dependency feature dataset to update the uncertainty envelope and risk budget, forming a rolling learning and self-healing scheduling loop. This invention addresses the problems in existing power system source-grid-load-storage collaborative optimization, such as insufficient characterization of high-order features of disturbance uncertainty, lack of cross-layer coordination in risk budget allocation, and insufficient adaptability of control strategies to distribution drift. It proposes a high-order uncertainty collaborative control method based on sub-Bruker optimization. This method constructs a hybrid uncertainty envelope by introducing skew and spiking features, achieving high-precision modeling of complex disturbance distributions. It enhances the risk coupling coordination capability among power sources, grids, loads, and storage by dynamically adjusting the risk budget through a cross-layer sensitive mapping mechanism. At the same time, based on rolling learning and self-repair mechanisms, the system can automatically identify and adaptively adjust to data drift and sudden disturbances, thereby significantly improving the robustness, safety resilience, and optimization efficiency of the power system in time-varying environments. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 A flowchart of a high-order uncertainty collaborative control method for power system source-grid-load integration based on distributed bar optimization is provided for an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] like Figure 1 As shown, this invention provides a high-order uncertainty collaborative control method for power system source-grid-load integration based on distributed bar optimization, comprising:

[0024] Step 100: Collect multi-layer operation data from the power supply side, network side, load side, and energy storage side, extract key features reflecting skewness, sharpness, and related dependencies, and obtain a high-order dependency feature dataset;

[0025] Step 200: Based on the aforementioned high-order dependency feature dataset, construct a hybrid uncertainty envelope with high-order statistical constraints to define the credible boundaries of the perturbation variables;

[0026] Step 300: Self-calibrate the mixed uncertainty envelope using real-time running deviation, so that the envelope range can automatically expand and contract with data drift, and obtain a dynamically adjusted robust uncertainty boundary;

[0027] Step 400: Under the dynamically adjusted uncertainty robust boundary, establish the risk budget allocation relationship between the power supply side, network side, load side and energy storage side, and realize the linkage and balance of risks between layers through inter-layer sensitive mapping to obtain a coordinated cross-layer risk budget result.

[0028] Step 500: Based on the cross-layer risk budget results, generate a real-time correction strategy that satisfies the monotonicity and smoothness constraints to obtain a verified recoverable control strategy;

[0029] Step 600: Combining the verified recoverable control strategy with the cross-layer risk budget results, the collaborative optimization problem is equivalently reconstructed, transforming the structure containing high-order constraints into a single-layer solvable form, in order to construct an operable sub-Bruker optimization model;

[0030] Step 700: Based on the running results of the operable sub-Bruker optimization model, set the distribution mutation identification index. When the observation breaks through the uncertainty envelope, trigger the safe order reduction response to obtain the running results with the ability to safely reduce order during sudden disturbances.

[0031] Step 800: Feed the running results back to the high-order dependency feature dataset, update the uncertainty envelope and risk budget, and form a rolling learning and self-healing scheduling loop.

[0032] Furthermore, the specific implementation process of step 100 is as follows:

[0033] In this embodiment, the acquisition of high-order dependency feature datasets first involves unified sampling and time synchronization of key operating parameters from the power supply side, network side, load side, and energy storage side. To ensure data quality, a multi-scale wavelet denoising method is used to filter out transient random disturbances, and abrupt changes are eliminated using an anomaly detection algorithm based on Mahalanobis distance. Furthermore, Z-standardization is used to achieve dimensional uniformity, resulting in a multi-layered operational data base with consistent comparability under the same time benchmark. This provides reliable data support for subsequent high-order statistical feature extraction.

[0034] Building upon this foundation, this embodiment utilizes skewness coefficients and kurtosis factors to quantitatively describe asymmetric fluctuations and extreme value clustering behaviors. It also combines empirical distribution functions to estimate the tail sensitivity of variables, thereby obtaining a set of statistical features characterizing the higher-order distribution patterns of the variables. Subsequently, a Copula function is introduced to construct a multidimensional dependency structure model. By estimating the conditional correlations and tail dependency strengths among variables, the linkage mechanisms between power sources, networks, loads, and energy storage are identified. Joint analysis of these higher-order statistical features and dependency parameters yields structural correlation information on multi-layered disturbances, reflecting potential covariant relationships during system operation.

[0035] This embodiment further performs weighted integration and temporal rearrangement of the obtained dependency structure features. Based on the weighted fusion of each feature's contribution to system fluctuations, a dynamic feature sequence is constructed using a time-sliding window approach to achieve a high-order characterization of the evolution of the operating state. The integrated dependency feature combination forms a high-order dependency feature dataset with time-varying and inter-layer coupling characteristics. This dataset can serve as the core input for subsequent uncertainty envelope construction and robust optimization modeling, enabling a deep description and quantifiable expression of the multi-layered disturbance mechanism of the power system's source-grid-load-storage system.

[0036] Furthermore, the specific implementation process of step 200 is as follows:

[0037] In constructing a hybrid uncertainty envelope with higher-order statistical constraints, this embodiment first analyzes the statistical characteristics of each disturbance source item by item, relying on the aforementioned higher-order dependency feature dataset. By calculating the mean trend, fluctuation range, and higher-order skewness and kurtosis indices for each disturbance variable, statistical feature parameters that can quantitatively describe the distribution pattern are extracted, forming an uncertainty feature set containing higher-order statistical indices of multiple levels of disturbances. Among them, the skewness index is used to measure the asymmetry of the disturbance distribution, and the kurtosis index is used to characterize the peak concentration of fluctuations; both reflect the degree to which the disturbance distribution deviates from the normal state.

[0038] After obtaining the uncertainty feature set, this embodiment estimates the degree of distributional differences among multiple variables and employs a dependency modeling method based on correlation entropy or Copula functions to identify the dependency strength, direction, and cooperative change patterns among disturbances at different levels, thereby obtaining the dependency structure feature results of multidimensional disturbance correlation. This dependency structure is used to reveal the implicit coupling relationship between disturbance variables. For example, when the load fluctuation amplitude increases, the power output on the power source side and the energy storage switching strategy will have a common response trend. After combining the dependency structure feature results with higher-order statistical features, a higher-order constraint set under multiple disturbance conditions is formed. By setting confidence interval thresholds and system steady-state boundaries, the acceptable boundary range of uncertainty in the disturbance space can be defined, resulting in a preliminary constraint boundary model.

[0039] This embodiment integrates higher-order statistical constraints and dependency structure constraints when constructing the hybrid uncertainty envelope, thereby forming a unified uncertainty description framework that considers both the single-dimensional perturbation distribution pattern and multi-dimensional correlation characteristics. The parameters used to calculate the constraint boundaries can be derived from the higher-order dependency feature dataset and the running samples: the multi-perturbation variable vector is extracted from the aforementioned multi-layer running data base set; the expectation vector is calculated from the mean of the perturbation samples; the covariance matrix is ​​deduced based on the variance and correlation between the perturbation samples; the skew statistical function and the ridge statistical function are calculated using the third and fourth order central moments; the skew amplification factor and the ridge amplification factor can be empirically set based on the distribution characteristics of historical perturbation samples; and the confidence envelope threshold is set to a percentile range according to the system safety tolerance requirements. This fusion model can dynamically characterize the nonlinear coupling structure between perturbation variables, making the obtained hybrid uncertainty envelope both statistically interpretable and providing an accurate and stable constraint basis for subsequent sub-Bruker optimization models.

[0040] Furthermore, the specific implementation process of step 300 is as follows:

[0041] In this embodiment, during the envelope self-calibration process, real-time operational data from the power supply side, network side, load side, and energy storage side are first compared and analyzed with planned operational data within corresponding time windows. By identifying the deviation direction, offset magnitude, and duration of each disturbance variable relative to the planned state, deviation information characterizing data drift is extracted. To address random measurement errors and sudden anomalies in the real-time data, this embodiment employs a robust regression-based noise suppression method to locally weight and correct abnormal offset segments, thereby obtaining a cleaned real-time operational deviation dataset. This dataset is used to reflect the degree to which the current system operation deviates from the envelope center in real time.

[0042] After acquiring the real-time operational deviation dataset, this embodiment utilizes a sensitivity mapping model to calculate the displacement between the uncertainty envelope center and the boundaries of each layer. Specifically, by analyzing the impact of perturbation variable deviations on envelope morphological parameters, a set of mapping factors describing changes in envelope scale and directionality is obtained. The essence of the sensitivity mapping model is to establish a response mapping relationship between perturbation variable offsets and envelope morphological changes, thereby quantifying the stretching or contraction effects caused by different perturbations in the envelope space. For example, when the power output continuously shifts positively, the boundary radius of the envelope in that direction will expand accordingly, while a decrease in load fluctuations or an increase in energy storage output compensation will cause boundary contraction. The result of this step is the formation of a set of mapping factors that can directly drive envelope boundary adjustments to achieve adaptive morphological correction in a multidimensional distribution space.

[0043] Subsequently, this embodiment uses the mapping factor set as adjustment parameters input into the envelope self-calibration process, and determines the adjusted uncertainty boundary parameter set through multi-stage iterations. Specifically, this includes: using the current inner points of the envelope as a reference, gradually correcting the Euclidean distance between the outer points and the boundary, and gradually correcting the envelope radius and shape curve according to the adjustment factors. When the iteration step size is lower than a set threshold and the envelope coverage rate meets the confidence requirement, the update stops and the self-calibrated uncertainty boundary parameter set is output. Finally, the shape of the envelope after expansion and contraction is smoothed to eliminate local inflections and abrupt changes introduced by the self-calibration iteration, so that the envelope maintains continuity and geometric stability under dynamic conditions, thereby forming a dynamically adjusted and smoothed uncertainty robust boundary, providing a real-time updated perturbation confidence constraint basis for the subsequent sub-bullscrew optimization model.

[0044] Furthermore, the specific implementation process of step 400 is as follows:

[0045] In constructing the cross-layer risk budget allocation relationship, this embodiment first analyzes the operational data of the power generation side, network side, load side, and energy storage side in real time based on dynamically adjusted uncertainty robust boundaries to identify risk-sensitive variables at each layer, including generation power margin, grid cross-sectional power flow margin, load response deviation, and available energy storage capacity. For these variables, this embodiment extracts risk assessment indicators reflecting safety margin, power deviation, and resource availability, and sets initial risk constraint ranges based on the operational characteristics of different layers, forming a basic risk indicator set containing multiple layers of risk variables and constraint intervals. This indicator set describes the potential risk level of each layer's operational status and serves as the basic input for subsequent risk budget calculations.

[0046] After obtaining the basic risk indicator set, this embodiment establishes a multi-layered nested risk budget model based on the operational characteristics and collaborative relationships of the source, grid, load, and storage layers. This model aims at risk equalization and determines the risk budget proportions for different layers by constructing hierarchical objective functions and constraint equations. Budget allocation comprehensively considers the importance and sensitivity weights of each layer, thereby deriving preliminary risk budget results for each layer from an independent operational perspective. For example, when the transmission margin risk on the network side is high, the system will tilt towards increasing the risk budget proportion on the network side to enhance the defense capability of that layer, thereby reducing the overall risk transmission and accumulation. The result of this step is the formation of independent risk budget matrices between layers, used for subsequent sensitivity mapping analysis.

[0047] Subsequently, this embodiment inputs the layered risk budget results into a sensitive mapping model to analyze the interaction relationships between risks at different layers. The sensitive mapping model is a mathematical framework for characterizing the transmission and response patterns of risks across different levels. It constructs a cross-layer risk sensitive mapping matrix by calculating the rate of change and acceleration of change of the risk response function at each layer relative to the budget parameters. The elements in the resulting matrix represent the sensitivity of the risk allocation ratio to the risk responses of other layers, and can be used to identify risk transmission paths and nonlinear coupling strength. The first-order risk rate of change reflects the direct transmission effect of risk between layers, while the second-order risk acceleration reflects the nonlinear strengthening trend of risk propagation. Correction coefficients are used to adjust the impact of the second-order terms on the overall mapping output. For example, when changes in the energy storage-side risk budget have a nonlinear amplification effect on the load-side risk response, the correction coefficients can be used to suppress excessive coupling effects and maintain a steady-state balance in cross-layer risk allocation. The results of this sensitive mapping model are used to make linked corrections to the hierarchical risk budget, so that the risks can be synchronously coordinated and balanced across the four levels of source, grid, load and storage. Finally, the cross-level risk budget results are obtained after being consistent, which provides a stable risk tolerance basis for subsequent robust response and dynamic control decisions.

[0048] Furthermore, the specific implementation process of step 500 is as follows:

[0049] In generating a real-time correction strategy that satisfies monotonicity and smoothness constraints, this embodiment first identifies key control variables on the power generation side, network side, load side, and energy storage side based on the aforementioned cross-layer risk budget results, and constructs a control constraint input set according to their corresponding risk tolerance ranges. These key control variables include power generation output, node voltage, grid power flow distribution, load response amplitude, and energy storage charging and discharging power. For these variables, this embodiment extracts boundary conditions such as power output regulation, voltage constraints, load adjustable ratio, and energy storage capacity constraints, and determines initial control adjustment parameters based on the cross-layer risk budget results to ensure the executability and stability of the control inputs within the risk tolerance range, thereby forming a control constraint input set containing multi-layer control variables, constraint boundaries, and initial parameters.

[0050] After obtaining the control constraint input set, this embodiment establishes a real-time correction strategy model. This model focuses on achieving convergence of the target indicator and balancing risk constraints. By introducing monotonic constraints, it ensures that the direction of change of the control variables aligns with the improvement trend of the target indicator, avoiding reverse oscillations or undesirable deviations. For example, during power adjustment, when the system cost function decreases, the monotonic constraints in the model restrict the update direction of the control variables to ensure that the cost function continues to decrease without fluctuations or rebounds. To address different operating scenarios, this embodiment sets a control correction step size parameter to adjust the magnitude of each control quantity update; the sensitivity matrix describes the responsiveness of each control variable to the constraints; the operating cost vector and the target vector are derived from system performance indicator evaluation and risk budget balance calculation, respectively. The control correction step size can be automatically adjusted according to the real-time iterative convergence rate, and its initial value can be set based on historical control effect experience. During iterative execution, the model continuously corrects the control variables, ensuring that the results always meet the monotonic change requirement and achieve optimal approximation within the feasible region.

[0051] After obtaining a real-time correction strategy that satisfies monotonic constraints, this embodiment further introduces a smoothing constraint model to smooth the control process over time. This constraint model limits the variation amplitude of control decision variables between adjacent time steps, achieving continuous and flexible transitions in control updates and preventing amplification of system disturbances caused by instantaneous jumps. The smoothing constraint model consists of a control coefficient matrix, a constraint threshold vector, and a smoothing tolerance parameter. The control coefficient matrix identifies the types of constraints affecting boundary conditions, the constraint threshold vector defines the allowable safety upper limit for each variable, and the smoothing tolerance parameter limits the rate of control change between adjacent time steps. This embodiment solves the control problem under smoothing constraints by driving a predictive control algorithm or a real-time optimizer, and performs recoverability verification and operational feasibility checks on the optimization results to ensure that the strategy remains recoverable and continuously schedulable under sudden disturbances and boundary drift conditions. Finally, a recoverable control strategy, verified mathematically and logically, is formed, providing implementable dynamic adjustment commands and risk response support for subsequent sub-Bruker optimization models.

[0052] Furthermore, the specific implementation process of step 600 is as follows:

[0053] In constructing an operable sub-optimal model, this embodiment first integrates and analyzes validated recoverable control strategies with cross-layer risk budget results to clarify the collaborative optimization logic structure of the source, grid, load, and storage layers. By analyzing the interrelationships between the objective functions and constraints at each layer, coupled control variables and their channels of action are identified, thereby constructing a structural mapping set describing the multi-layer decision-making logic of the system. This structural mapping set reflects the cross-dependencies between power supply-side output regulation, network-side power flow allocation, load-side response characteristics, and energy storage scheduling, serving as the fundamental input for the equivalent reconstruction of the collaborative optimization model. At this stage, this embodiment establishes mapping links between risk tolerance, robust constraints, and dependent characteristic parameters in higher-order constraints and control variables to achieve a unified characterization of multi-layer objective functions and constraints.

[0054] Subsequently, this embodiment utilizes the aforementioned collaborative optimization structure mapping set to perform equivalent replacements on the higher-order dependencies, risk constraint boundaries, and uncertain disturbance terms involved in the original model. Specifically, by setting a disturbance confidence set and introducing a dual transformation method, the original high-order constraint expression, which was difficult to solve directly, is transformed into a constraint expression that can be linearized or optimized by convex means. This process reconstructs the multi-layer collaborative optimization model, which originally contained high-order nonlinear risk terms and nested dependency structures, into a robust optimization form that can be directly solved in a single-layer structure. In this process, the risk tolerance is provided by the cross-layer risk budget results as an adjustable parameter on the constraint side; the control variables and their constraints are provided by the recoverable control strategy model. Through variable internalization and inter-layer constraint folding methods, the operational constraints of each layer are uniformly mapped to the feasible decision domain, ensuring the overall connectivity and solvability of the model.

[0055] Based on this equivalent reconstruction model framework, this embodiment introduces a confidence set boundary for the uncertain disturbance distribution during the solution process, using the Wasserstein distance as a metric to characterize the deviation between the actual disturbance distribution and the reference sample distribution. For the control variables in the model, their value range is determined by the optimized output of the aforementioned recoverable control strategy, ensuring both operational operability and dynamic adaptability of robust constraints. For the expectation calculation of the cost function, an approximate solution of the statistical expectation is performed using the true confidence set of the disturbance distribution, thereby reflecting the expected performance under distribution drift. The confidence radius parameter is used to limit the acceptable range of disturbance deviation; its value can be taken as the percentile ratio of the root mean square deviation of historical disturbance samples to achieve a balance between stability and sensitivity. The resulting sub-Bruker robust optimization model, while maintaining model solvability, can dynamically reflect the impact of uncertainty, achieving robust solution and distribution adaptive control for the integrated collaborative optimization problem of power sources, grids, loads, and storage. This provides a directly implementable optimization decision-making basis for the safe, stable, and efficient operation of power systems under high uncertainty scenarios.

[0056] Furthermore, the specific implementation process of step 700 is as follows:

[0057] In implementing an operational strategy with the capability to safely reduce the impact of sudden disturbances, this embodiment first continuously monitors key operational status parameters on the power supply side, network side, load side, and energy storage side based on the real-time operational results of an operable sub-Bruker optimization model. It focuses on tracking characteristic variables such as power output change rate, voltage offset, load response rate, and energy storage charging and discharging fluctuations. By calculating the local statistical characteristics of the operational data within a time sliding window, this embodiment identifies abnormal signals that may lead to disturbance instability, including abrupt slope changes, amplitude spikes, or periodic abnormal behaviors. For the extracted potential anomalies and abrupt change trends, a robust filtering and clustering weighting method is used for screening and feature extraction, forming a dataset of abrupt change symptom patterns that reflects the rapid changes in multi-layered operational states. This dataset accurately describes the multi-dimensional dynamic characteristic changes of the system in the disturbance precursor stage.

[0058] After obtaining the dataset of mutation symptoms, this embodiment performs a dynamic evaluation of key disturbance variables at the distribution level. Specifically, the distribution deviation rate is obtained by calculating the deviation of each variable from its historical statistical distribution; the change rate of the disturbance sample sequence over a continuous period is calculated to obtain the time-varying gradient; and an abnormal growth coefficient is constructed by combining the relationship between the variable growth rate and the fluctuation amplitude. These three types of indicators together constitute a distribution mutation identification index set used to characterize the speed of disturbance propagation and the risk of envelope breakthrough. Based on this, this embodiment establishes the statistical threshold range of the multidimensional identification indicators through backtracking training of historical operating data, and determines the triggering critical values ​​corresponding to different levels to distinguish between normal disturbances, recoverable deviations, and high-risk mutations. This index system can quantitatively characterize the degree of breakthrough of the disturbance distribution relative to the uncertain envelope, providing a model-based criterion for triggering a safety response.

[0059] Subsequently, this embodiment performs real-time comparisons based on the aforementioned distribution mutation identification indicators and uncertainty robustness boundaries. If any indicator exceeds the set hierarchical trigger threshold, a safety degradation response mechanism is immediately triggered. The safety degradation response mechanism is an operational emergency strategy based on risk grading and dynamic decoupling, used to rapidly reduce system complexity and energy flow pressure under disturbance exceeding limits. After triggering the response, this embodiment invokes a verified recoverable control strategy to perform output reduction adjustment on the power source side, impedance path reconstruction and power flow redistribution on the network side, zoned load reduction or peak shifting response on the load side, and rapid charging and discharging on the energy storage side to balance instantaneous energy deviations. By establishing a hierarchically linked safety degradation response scheme, coordinated degradation and risk release among the source, grid, load, and storage are achieved. Finally, after the execution of the above strategies, the system will recover to a stable operating state within the envelope, thus obtaining an operational result with the ability to safely reduce degradation under sudden disturbances, effectively improving the robustness and self-recovery of the overall power system under extreme disturbances.

[0060] Furthermore, the specific implementation process of step 800 is as follows:

[0061] In this embodiment, when forming a rolling learning and self-healing scheduling loop, the operational results with the ability to safely reduce the order of sudden disturbances are first used as feedback. The operational status, control response, and risk deviation characteristics of the power supply side, network side, load side, and energy storage side are extracted and classified in detail. Specifically, by monitoring actual output, power flow distribution, voltage and current changes, load response signals, and energy storage status changes, dynamic time-series characteristics in the operational results are captured. Furthermore, spatial topology information is combined to encode and analyze the synergistic effects between different regions or nodes. Through dimensionality reduction encoding and feature combination processing, an operational feedback dataset characterizing the dynamic behavior and coupling relationships of the system is generated. This dataset reflects the multidimensional response patterns of the system under the influence of real-time disturbances, providing fundamental data support for subsequent feature updates and envelope correction.

[0062] After obtaining the runtime feedback dataset, this embodiment inputs it into a higher-order dependency feature dataset to achieve rolling updates and dynamic calibration of the system feature distribution. Specifically, firstly, by comparing the newly sampled data with historical feature distributions, abnormal drift and model mismatch behaviors caused by changes in runtime conditions are identified; then, an incremental learning method is used to correct parameters and update the structure of the deviating features. The core mechanism of the incremental learning method is to gradually absorb new sample features while preserving the existing statistical structure, enabling the model to maintain stable convergence in changing environments. The updated higher-order dependency feature set can dynamically reflect the evolution trend of perturbation correlation structure and feature parameters, avoiding envelope calculation distortion due to model lag, and achieving adaptive continuous correction of higher-order dependency features.

[0063] Based on the dynamically updated higher-order dependency feature set, this embodiment re-estimates the distribution of disturbance variables and their boundary fluctuation amplitudes, performs adaptive adjustment of envelope expansion and contraction, and obtains a self-corrected uncertainty envelope update result. This uncertainty envelope update result is used to redefine the credibility range and higher-order constraint weights of each disturbance variable, and recalculates and corrects the cross-layer risk budget in conjunction with changes in risk distribution, determining the latest risk allocation ratio and inter-layer coupling coefficient. After the update is completed, this embodiment inputs the corrected uncertainty envelope and risk budget results into the distributed robust optimization model, iteratively optimizing the strategy parameters and scheduling instructions, so that the model forms a closed-loop self-learning scheduling structure of "analysis-execution-feedback-update". Through rolling learning and self-repair mechanism, the scheduling strategy can be continuously evolved and achieve long-term adaptive stability, thereby ensuring the robustness, agility and intelligent response capability of the system in uncertain operating environments.

[0064] This embodiment also provides a high-order uncertainty collaborative control system for power system source-grid-load integration based on sub-blob bar optimization, including:

[0065] The acquisition module is used to collect multi-layer operation data from the power supply side, network side, load side and energy storage side, extract key features reflecting skewness, sharpness and related dependencies, and obtain a high-order dependency feature dataset.

[0066] An uncertainty envelope construction module is used to construct a hybrid uncertainty envelope with high-order statistical constraints based on the high-order dependency feature dataset, so as to define the credible boundary of the perturbation variable;

[0067] A boundary construction module is used to self-calibrate the mixed uncertainty envelope using real-time running deviations, so that the envelope range can automatically expand and contract with data drift, resulting in a dynamically adjusted robust uncertainty boundary.

[0068] The risk budget result calculation module is used to establish the risk budget allocation relationship between the power supply side, network side, load side and energy storage side under the dynamically adjusted uncertainty robust boundary, and realize the linkage and balance of risks between layers through inter-layer sensitive mapping to obtain a coordinated cross-layer risk budget result.

[0069] The recoverable control strategy setting module is used to generate a real-time correction strategy that satisfies monotonicity and smoothness constraints based on the cross-layer risk budget results, thereby obtaining a verified recoverable control strategy.

[0070] The sub-Brubar optimization model construction module is used to combine the verified recoverable control strategy with the cross-layer risk budget results to equivalently reconstruct the collaborative optimization problem, transforming the structure containing high-order constraints into a single-layer solvable form, so as to construct an operable sub-Brubar optimization model.

[0071] The operation result calculation module is used to set the distribution mutation identification index based on the operation results of the operable sub-Bruker optimization model. When the observation breaks through the uncertainty envelope, it triggers the safe order reduction response and obtains the operation results with the ability to safely reduce order during sudden disturbances.

[0072] The update module is used to feed back the running results to the high-order dependency feature dataset, update the uncertainty envelope and risk budget, and form a rolling learning and self-healing scheduling loop.

[0073] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0074] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A high-order uncertainty collaborative control method for power system source-grid-load integration based on distributed bar optimization, characterized in that, include: Multi-layer operation data from the power supply side, network side, load side, and energy storage side are collected, and key features reflecting skewness, sharpness, and related dependencies are extracted to obtain a high-order dependency feature dataset. Based on the aforementioned high-order dependency feature dataset, a hybrid uncertainty envelope with high-order statistical constraints is constructed to define the credible boundary of the perturbation variable. The mixed uncertainty envelope is self-calibrated using real-time operational deviations, so that the envelope range can automatically expand and contract with data drift, resulting in a dynamically adjusted robust uncertainty boundary. Under the dynamically adjusted uncertainty robust boundary, the risk budget allocation relationship between the power supply side, network side, load side and energy storage side is established, and the linkage and balance of risks between layers are realized through inter-layer sensitive mapping, so as to obtain a coordinated cross-layer risk budget result. Based on the cross-layer risk budget results, a real-time correction strategy that satisfies the monotonicity and smoothness constraints is generated, resulting in a verified recoverable control strategy. By combining the verified recoverable control strategy with the cross-layer risk budget results, the collaborative optimization problem is equivalently reconstructed, transforming the structure containing high-order constraints into a single-layer solvable form, in order to construct an operable sub-Bruker optimization model. Based on the operational results of the aforementioned operable sub-Bruker optimization model, a distribution mutation identification index is set. When the observation breaks through the uncertainty envelope, a safe order reduction response is triggered, resulting in operational results with the ability to safely reduce order during sudden disturbances. The running results are fed back to the higher-order dependency feature dataset to update the uncertainty envelope and risk budget, forming a rolling learning and self-healing scheduling loop.

2. The method for integrated high-order uncertainty collaborative control of power system source-grid-load based on distributed bar optimization according to claim 1, characterized in that, The system collects multi-layered operational data from the power supply side, network side, load side, and energy storage side, extracts key features reflecting skewness, sharpness, and dependencies, and obtains a high-order dependency feature dataset, including: Key operational data from the power supply side, network side, load side, and energy storage side were uniformly sampled and synchronized in time, and noise reduction, anomaly screening, and standardization were performed to obtain a unified multi-layer operational data base set. Based on the aforementioned multi-layered operational data base, statistical features describing asymmetric fluctuations and peak concentrations were extracted, resulting in a statistical feature set characterizing the higher-order distribution of variables. Based on the statistical feature set, the correlation and linkage patterns of variables at different levels are analyzed, the dependency structure among power source, network, load and energy storage is identified, and the statistical features and dependency features are fused to obtain a combination of dependency features that jointly reflect the correlation of multi-level variables. The dependent features are combined, weighted, and arranged chronologically to obtain a high-order dependent feature dataset.

3. The method for integrated high-order uncertainty collaborative control of power system source-grid-load based on distributed bar optimization according to claim 1, characterized in that, The construction of a hybrid uncertainty envelope with higher-order statistical constraints based on the higher-order dependency feature dataset includes: Feature analysis was performed on the high-order dependency feature dataset to identify the mean trend, fluctuation range, and high-order skewness and sharp constraints of each perturbation variable, and statistical indicators that can reflect the distribution pattern were extracted to obtain an uncertainty feature set containing high-order statistical indicators of multi-level perturbations. Using the uncertainty feature set, the degree of distributional difference among multiple variables is calculated, and the dependence strength, direction and cooperative change pattern among variables are identified, thus obtaining the dependency structure feature results describing the multidimensional perturbation correlation. By combining the uncertainty feature set with the dependency structure feature results, and based on the acceptable confidence range and the steady-state boundary of operation, a high-order constraint set under multiple perturbation conditions is established, and the boundary range of the perturbation space is described by the constraint set, thus obtaining a constraint boundary model that preliminarily defines the confidence interval of multiple perturbations. Based on the constrained boundary model, higher-order statistical constraints and dependencies from different sources of disturbance are fused to obtain a hybrid uncertainty envelope with higher-order statistical constraints and dependency structure constraints. The expression for the constraint boundary model that initially defines the multi-perturbation confidence interval is: ； in, A vector of multiple perturbation variables; Let be the expected vector of the perturbation variable; Let be the covariance matrix of the disturbance variables; This is a skewed statistical function for the perturbation distribution, used to characterize the asymmetric distribution. is a sharp statistical function of the perturbation distribution, used to represent the peak concentration. This is the statistical amplification factor for the skewness feature, used to measure the degree of influence of skewness on the envelope shape; is the statistical amplification factor for the kurtosis feature, used to balance the strength of the effect of the distribution kurtosis on the envelope boundary; The confidence envelope threshold is used to limit the maximum tolerance range of the perturbation confidence interval.

4. The high-order uncertainty collaborative control method for power system source-grid-load integration based on distributed bar optimization according to claim 1, characterized in that, The method of self-calibrating the mixed uncertainty envelope using real-time operational deviations, enabling the envelope range to automatically expand and contract with data drift, and obtaining a dynamically adjusted robust uncertainty boundary, includes: The real-time operation data from the power supply side, network side, load side and energy storage side are compared with the planned operation data to identify the deviation direction and offset magnitude of each disturbance variable. Noise suppression and anomaly removal are then performed to obtain a real-time operation deviation dataset that characterizes the disturbance deviation features. Based on the real-time operational deviation dataset, the displacement of the envelope center and boundary under the current disturbance conditions is analyzed. The adjustment factors of the envelope shape and scale are calculated using a sensitivity mapping model, resulting in a set of mapping factors that reflect the dynamic adjustment requirements of the envelope. The set of mapping factors is used to perform multi-stage self-calibration on the mixed uncertainty envelope. The envelope radius and shape are adjusted by iterative comparison between the inner and outer points of the boundary, and the set of uncertainty boundary parameters adjusted by self-calibration iteration is obtained. Based on the set of uncertainty boundary parameters, the envelope after expansion and contraction is smoothed and corrected to obtain a robust uncertainty boundary that has been dynamically adjusted and smoothed.

5. The method for integrated high-order uncertainty collaborative control of power system source-grid-load based on distributed bar optimization according to claim 1, characterized in that, Under the dynamically adjusted uncertainty robust boundary, a risk budget allocation relationship is established among the power supply side, network side, load side, and energy storage side. Through inter-layer sensitive mapping, the linkage and balance of risks across layers are achieved, resulting in a coordinated cross-layer risk budget outcome, including: Based on the dynamically adjusted uncertainty robust boundary, risk-sensitive variables under the operating conditions of the power supply side, network side, load side, and energy storage side are identified. Risk assessment indicators reflecting safety margin, power deviation, and resource availability are extracted, and the initial risk constraint range is determined, resulting in a set of basic risk indicators containing multiple layers of risk variables and constraint intervals. Based on the aforementioned risk baseline indicator set, objective functions and constraints for risk budgets at each level are established. A multi-layered nested budget allocation mechanism is adopted, and the hierarchical risk budget results reflecting the independent risk allocation at each level are determined according to the importance and sensitivity weights between levels. The hierarchical risk budget results are input into the sensitive mapping model to calculate the mutual influence coefficient of risk changes between different levels, identify the coupling transmission path of risk between source, network, load and storage, and quantify the synchronous change trend of cross-layer risk, thus obtaining a cross-layer sensitive mapping matrix that can describe the degree of risk coupling. Based on the aforementioned cross-layer sensitive mapping matrix, the hierarchical risk budget results are adjusted in a coordinated manner, and the risk budget ratio and constraint boundary are adjusted to obtain a balanced and adjusted set of cross-layer risk budget relationships. The cross-layer risk budget relationship set was standardized to obtain cross-layer risk budget results that can be used to determine the risk tolerance of control variables and the range of robust response. The expression for the sensitive mapping model is: ； Where S is the cross-layer risk sensitivity mapping matrix; Indicates risk budget parameters hierarchical Risk response volume The sensitivity coefficient; For the first The risk response function corresponding to the layer; For the first Layered risk budget ratio parameters; This is the first-order rate of change of risk, used to characterize the direct coupling effect of risk transmission; This is a second-order risk acceleration, used to reflect the nonlinear strength and stability trend of risk transmission; This is the correction coefficient for the second-order sensitivity term, used to balance the impact of linear and nonlinear risk coupling on the overall mapping structure.

6. The method for integrated high-order uncertainty collaborative control of power system source-grid-load based on distributed bar optimization according to claim 1, characterized in that, Based on the cross-layer risk budget results, a real-time correction strategy satisfying monotonicity and smoothness constraints is generated to obtain a validated recoverable control strategy, including: Based on the cross-layer risk budget results, key control variables and their risk tolerance ranges on the power supply side, network side, load side, and energy storage side are identified. Constraints involving output regulation, voltage control, load response, and energy storage scheduling are extracted, resulting in a control constraint input set containing multi-layer control variables, constraint boundaries, and initial parameters. Based on the control constraint input set, a real-time correction strategy model is established, and a monotonic constraint condition is introduced to obtain a real-time correction strategy model that satisfies the monotonic constraint condition. By combining the real-time correction strategy model with the cross-layer risk budget results and performing time-series smoothing processing to avoid abrupt changes in control actions, a feasible domain constraint model with smooth transition characteristics was obtained. Under the feasible domain constraint model, the real-time adjustment command is solved by iterative optimization or predictive control algorithm, and the recoverability of the result is verified to obtain a recoverable control strategy that has been mathematically verified and tested for operational feasibility. The expression for the real-time correction strategy model is: ； The expression for the feasible region constraint model with smooth transition characteristics is: ； in, For the first Control decision vector of the time-matter system; This is the updated control vector for the next time step; To control the correction step factor; This is the sensitivity matrix of the system constraints to the control variables; This is a vector representing the current operating cost or performance metric. This serves as a reference target or expected indicator vector. The set of feasible domains for the control strategy; This represents the coefficient matrix of the control boundary and safety constraints; This is the corresponding constraint threshold vector; This is a smoothing tolerance allowed for changes in control values ​​at adjacent time points, used to ensure policy continuity and recoverability.

7. The method for integrated high-order uncertainty collaborative control of power system source-grid-load based on distributed bar optimization according to claim 1, characterized in that, The method combines the verified recoverable control strategy with the cross-layer risk budget results to perform an equivalent reconstruction of the collaborative optimization problem, transforming the structure containing higher-order constraints into a single-layer solvable form, in order to construct an operable sub-Bruker optimization model, including: Based on the verified recoverable control strategy and the cross-layer risk budget results, the correlation between the optimization objectives and coupled decision variables of each layer (source, network, load, and storage) is analyzed, and a collaborative optimization structure mapping set describing the multi-layer decision logic and constraint relationships of the system is obtained. Using the aforementioned collaborative optimization structure mapping set, the higher-order dependencies, risk tolerances, and robust constraints are equivalently represented by uncertainty set delimitation and dual transformation, resulting in the constraint transformation result of equivalent replacement of higher-order structures. The control variables and constraints in the verified recoverable control strategy are embedded into the constraint transformation result. Through variable internalization and inter-layer constraint folding, the original multi-layer collaborative optimization problem is transformed into a single-layer robust solvable optimization model structure, resulting in a robust model framework with single-layer solvability. Based on the robust model framework, and combining the confidence set boundary and risk response mechanism of the uncertain perturbation distribution, the target robustness weight and constraint confidence level are defined, resulting in an operable sub-robust optimization model. The expression for the operational sub-Blule bar optimization model is: ； in, The system's collaborative decision-making variable vector; This is the feasible decision domain; For perturbation random variables; To perturb the true distribution; Let be the set of uncertain distributions, representing all possible distribution regions that satisfy the confidence range; The empirical distribution of the sample; For about distribution Expectation calculation; This is the system operating cost function under actual disturbances; These are the system operation constraints. Wasserstein distance is a metric for inter-distribution distance, used to measure the degree of deviation between a perturbation distribution and a reference distribution. The Wasserstein confidence radius is used to define the acceptable range of uncertainty.

8. The high-order uncertainty collaborative control method for power system source-grid-load integration based on distributed bar optimization according to claim 1, characterized in that, Based on the operational results of the described sub-Bruker optimization model, a distribution mutation identification index is set. When the observation exceeds the uncertainty envelope, a safe order reduction response is triggered, resulting in operational results with safe order reduction capability for sudden disturbances, including: Based on the operation results of the operable sub-bar optimization model, the operating state parameters and key disturbance variables of each layer are continuously monitored, and the mutation signs in their time series changes are identified, resulting in a mutation sign dataset reflecting the dynamic fluctuation characteristics of multiple layers. Based on the mutation symptom dataset, the distribution deviation rate, time-varying gradient, and abnormal growth coefficient were calculated. Combined with historical statistical patterns, a multidimensional distribution mutation identification index was formed, resulting in a distribution mutation identification index set that can quantify the degree of envelope breakthrough. By comparing the distribution mutation identification index set with the aforementioned uncertainty envelope boundary, setting a hierarchical trigger threshold, and activating a response signal when the index exceeds the critical value, an envelope breakthrough determination mechanism for security response triggering is obtained. Based on the envelope breakthrough determination mechanism and combined with the verified recoverable control strategy, the operation modes of the source, network, load and storage layers are reduced in order and the load is reduced and optimized, resulting in a safe reduction response scheme with multi-layer linkage characteristics. The aforementioned safe order reduction response scheme was applied to the real-time operation control process to perform source power adjustment, network impedance reconstruction, load shedding and energy storage compensation operations, resulting in system operation results with safe order reduction capability for sudden disturbances.

9. A high-order uncertainty collaborative control method for power system source-grid-load integration based on distributed bar optimization according to claim 1, characterized in that, The step of feeding back the running results to the high-order dependency feature dataset, updating the uncertainty envelope and risk budget, and forming a rolling learning and self-healing scheduling loop includes: Based on the operational results of the system with the ability to safely reduce the order of sudden disturbances, the actual operating status, control response and risk deviation characteristics are identified. The time series data and spatial coordination information are organized and feature-encoded to obtain an operational feedback dataset that reflects the dynamic response characteristics of the system. The running feedback dataset is input into the higher-order dependency feature dataset, the feature distribution is updated in a rolling manner, abnormal drift and model mismatch behavior are identified, and the bias features are corrected by incremental learning method, resulting in a dynamically calibrated higher-order dependency feature update set. Based on the higher-order dependency feature update set, the distribution of perturbation variables and the amplitude of boundary fluctuations are re-estimated, envelope adaptive expansion and contraction adjustment is performed, and the dynamic robust boundary is updated to obtain the self-corrected uncertain envelope update result. Based on the uncertainty envelope update results, the risk allocation ratio and coupling coefficient are re-evaluated and adjusted to obtain the cross-layer risk budget update results that reflect the latest system state; The uncertainty envelope update result and the cross-layer risk budget update result are input into the sub-Bruker optimization model to achieve iterative correction of the strategy parameters and scheduling strategy.

10. A high-order uncertainty collaborative control system for power system source-grid-load integration based on distributed bar optimization, characterized in that, include: The acquisition module is used to collect multi-layer operation data from the power supply side, network side, load side and energy storage side, extract key features reflecting skewness, sharpness and related dependencies, and obtain a high-order dependency feature dataset. An uncertainty envelope construction module is used to construct a hybrid uncertainty envelope with high-order statistical constraints based on the high-order dependency feature dataset, so as to define the credible boundary of the perturbation variable; A boundary construction module is used to self-calibrate the mixed uncertainty envelope using real-time running deviations, so that the envelope range can automatically expand and contract with data drift, resulting in a dynamically adjusted robust uncertainty boundary. The risk budget result calculation module is used to establish the risk budget allocation relationship between the power supply side, network side, load side and energy storage side under the dynamically adjusted uncertainty robust boundary, and realize the linkage and balance of risks between layers through inter-layer sensitive mapping to obtain a coordinated cross-layer risk budget result. The recoverable control strategy setting module is used to generate a real-time correction strategy that satisfies monotonicity and smoothness constraints based on the cross-layer risk budget results, thereby obtaining a verified recoverable control strategy. The sub-Brubar optimization model construction module is used to combine the verified recoverable control strategy with the cross-layer risk budget results to equivalently reconstruct the collaborative optimization problem, transforming the structure containing high-order constraints into a single-layer solvable form, so as to construct an operable sub-Brubar optimization model. The operation result calculation module is used to set the distribution mutation identification index based on the operation results of the operable sub-Bruker optimization model. When the observation breaks through the uncertainty envelope, it triggers the safe order reduction response and obtains the operation results with the ability to safely reduce order during sudden disturbances. The update module is used to feed back the running results to the high-order dependency feature dataset, update the uncertainty envelope and risk budget, and form a rolling learning and self-healing scheduling loop.

Images