A high-proportion new energy power grid extreme weather condition dispatching optimization system and method

Abstract

The application discloses a high-proportion new energy power grid extreme weather condition scheduling optimization system and method, and relates to the technical field of power system scheduling and control. The system obtains new energy output and load point prediction, and constructs a prediction error interval and an error uncertainty set based on distribution-independent calibration; on this basis, the system calculates the continuous shortage risk and power fluctuation risk and generates an early warning level; according to the early warning level, the system performs rolling optimization, configures conventional backup and extreme backup, and introduces backup deliverability constraints, determines the additional uplink and downlink backup capacity according to the most unfavorable disturbance situation under the uncertainty set constraints; the system executes the scheduling instruction and closes the loop, triggers the emergency, implements orderly power utilization and emergency resource calling according to the load priority. The application realizes risk early warning and deliverable backup configuration without relying on error distribution assumption, improves supply and demand guarantee and operation safety margin, and enhances the resilience and continuous power supply capacity of the power grid under extreme weather.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching and control technology, and in particular to a dispatching optimization system and method for a high-proportion renewable energy power grid under extreme weather conditions. Background Technology

[0002] Power system dispatch optimization is used to coordinate power generation output, adjustable resources, and load-side measures to ensure the safe and stable operation of the power grid while meeting power balance and security constraints. With the increasing proportion of new energy sources such as wind and solar power, system operational uncertainty has increased, making dispatch more sensitive to forecast deviations and fluctuations.

[0003] Under extreme weather conditions such as typhoons, torrential rains, cold waves, high temperatures, and icing, the output of new energy sources and load demand are prone to drastic changes, significantly increasing forecast deviations and potentially causing changes in statistical characteristics. This leads to a contraction of system safety margins and a decrease in resource availability, with operational risks manifesting as both persistent shortages and rapid fluctuations. Existing dispatching systems mostly employ rolling adjustments coupled with backup, risk assessment, and emergency response. However, in the aforementioned scenarios, uncertainties and inconsistencies between quantified risk results and actual operations may still occur. Risk levels are difficult to effectively drive dynamic adjustments to dispatching strategies and resource allocation, and there is insufficient consistency between planned decisions and execution responses, affecting the feasibility of rolling operations and power supply assurance capabilities. Furthermore, the requirements for real-time performance and availability are higher under extreme conditions, further increasing the risk of operational uncertainty transmission.

[0004] Therefore, the core technical problem that urgently needs to be solved is: under the condition that extreme weather leads to a significant increase in uncertainty and possible changes in statistical characteristics, how to reliably quantify and continuously assess the forecast deviation and operational risks, and enable the risk level to effectively drive the dynamic adjustment of scheduling strategies and resource allocation, so as to ensure consistency between decision-making and execution in rolling operation, thereby improving the safety and stability of the power grid and the power supply guarantee capability. Summary of the Invention

[0005] The purpose of this invention is to provide a scheduling optimization system and method for high-proportion renewable energy power grids under extreme weather conditions. This system addresses the limitations of existing technologies in scenarios involving extreme weather combined with high-proportion renewable energy, such as insufficient quantification reliability of prediction deviations and operational risks, inadequate linkage between risk assessment and scheduling strategies, difficulty in ensuring the executability and consistency of resource allocation during rolling operation, and poor coordination between emergency response and planning decisions. The invention aims to achieve continuous assessment and dynamic control of uncertainty and risk levels, improve the stability and executability of rolling scheduling decisions, enhance the guarantee capability of critical loads and power supply reliability, and improve the safe and stable operation level and system resilience of the power grid under extreme weather conditions.

[0006] To achieve the above objectives, the present invention employs the following technical solution:

[0007] On the one hand, this invention provides a dispatch optimization system for a high-proportion renewable energy power grid under extreme weather conditions, comprising:

[0008] The distribution-independent prediction module is used to output point prediction results of new energy output and load, and output distribution-independent uncertainty characterization results, which include prediction error interval and prediction error uncertainty set.

[0009] The continuous risk warning module is used to generate warning levels based on point prediction results and uncertainty characterization results. The warning level is determined according to a preset mapping rule based on the comparison results of the continuous shortage risk quantity and the first preset threshold and the comparison results of the power fluctuation risk quantity and the second preset threshold. The continuous shortage risk quantity is obtained by aggregating the power shortage quantity according to a preset aggregation rule, and the power fluctuation risk quantity is obtained by aggregating the power fluctuation overshoot quantity according to a preset aggregation rule.

[0010] A dual-layer backup rolling module is used to switch between normal mode and extreme mode and perform rolling optimization based on the warning level. When the warning level reaches or exceeds a preset switching threshold, it switches to extreme mode. The dual-layer backup rolling module configures dual-layer backup of normal backup and extreme backup in extreme mode, and sets backup deliverability constraints. Backup deliverability constraints include window connection constraints, which limit the backup adjustment amount of each schedulable resource unit at the rolling window connection to not exceed the corresponding ramping capacity. The extreme backup quota is determined based on the prediction error scenario that leads to the minimum system safety constraint margin within the prediction error uncertainty set, and selects the prediction error scenario with the largest supply-demand gap when the safety constraint margin is the same, including additional uplink backup capacity and additional downlink backup capacity.

[0011] The priority closed-loop module is used to execute the scheduling instructions output by the dual-layer backup rolling module and send back the execution feedback. When the preset emergency triggering conditions are triggered, it implements orderly power consumption and emergency resource call according to the preset load priority.

[0012] As a preferred embodiment of the present invention, the distribution-independent prediction module includes:

[0013] An environmental feature generation unit is used to generate environmental features based on meteorological data and power grid operation data.

[0014] The point prediction unit is used to output point prediction results of new energy output and load based on environmental characteristics, and to form net load prediction error samples for each region based on the point prediction results and historical observation data. The prediction error samples are used as calibration samples.

[0015] The distribution-independent calibration unit is used to generate extreme operating condition indications based on environmental characteristics, perform gating selection on prediction error samples and set sample weights according to the extreme operating condition indications, generate the lower limit of prediction error, the upper limit of prediction error, and the upper bound of the total net load prediction error of the entire network for each region, and construct a prediction error uncertainty set; the prediction error uncertainty set satisfies that the prediction error of each region is between the corresponding upper and lower limits, and the total net load prediction error of the entire network does not exceed the total upper bound.

[0016] The coverage closed-loop recalibration unit is used to count the proportion of actual prediction errors falling into the prediction error uncertainty set during rolling operation, and to perform recalibration operation when the proportion is lower than the preset coverage threshold, updating the lower limit of prediction error, the upper limit of prediction error, the upper limit of the total net load prediction error of the whole network, and the prediction error uncertainty set.

[0017] As a preferred embodiment of the present invention, the continuous risk warning module includes:

[0018] A scenario generation unit is used to generate a risk calculation scenario set, which includes scenarios generated by sampling based on a prediction error uncertainty set according to a preset sampling rule and scenarios matched from a preset extreme scenario library based on extreme working condition indications according to a preset matching rule.

[0019] The risk component calculation unit is used to assess the power shortage and power fluctuation overshoot based on point prediction results and uncertainty characterization results in various scenarios. The available output and ramping capacity of schedulable resources are reduced according to the availability parameters of extreme weather.

[0020] The risk quantity generation unit is used to perform time aggregation and scenario aggregation on power shortage quantity and power fluctuation overshoot quantity respectively to generate continuous shortage risk quantity and power fluctuation risk quantity.

[0021] The warning level determination unit is used to compare the continuous shortage risk with the first preset threshold, compare the power fluctuation risk with the second preset threshold, and output the warning level according to the preset mapping rule.

[0022] The risk component calculation unit is also used to incorporate planned load reduction measures during the assessment, which are taken from the most recent load reduction plan output by the dual-layer standby rolling module.

[0023] As a preferred embodiment of the present invention, the dual-layer backup rolling module includes:

[0024] The mode switching unit is used to switch between normal mode and extreme mode according to the warning level;

[0025] The backup configuration unit is used to configure a two-layer backup of normal backup and extreme backup in extreme mode. The extreme backup quota is determined based on the uncertainty characterization result and allocated to the uplink backup and downlink backup according to the preset duration level.

[0026] A deliverability constraint construction unit is used to construct backup deliverability constraints in the rolling optimization problem. These backup deliverability constraints include: a window connection constraint, used to limit the backup adjustment amount of each schedulable resource unit at the boundary of adjacent rolling cycles to not exceed the ramp-up capacity; a hierarchical consistency constraint, used to limit the backup capacity between hierarchical levels of different durations to meet the incremental constraint condition, wherein the backup capacity of the longer duration hierarchical level is not less than the backup capacity of the shorter duration hierarchical level; and a differential coverage constraint, used to limit the uplink and downlink backup of each schedulable resource unit within the same duration hierarchical level to cover the net load increase and decrease changes, respectively, based on the net load change demand within the rolling window; the differential coverage constraint is expressed using a capacity inequality, where the sum of the uplink backup capacity of each schedulable resource unit within the same duration hierarchical level is not less than the net load increase change demand, and the sum of the downlink backup capacity is not less than the net load decrease change demand.

[0027] The solution execution unit is used to solve the rolling optimization problem with deliverability constraints in each rolling cycle and output the scheduling instructions for the current rolling cycle.

[0028] In a preferred embodiment of the present invention, the dual-layer backup rolling module and the priority closed-loop module adopt the same load priority rule; the dual-layer backup rolling module includes:

[0029] The load classification unit is used to divide the load into critical loads, high priority loads, and general loads;

[0030] A critical load hard constraint generation unit is used to generate critical load hard constraints in the rolling optimization problem corresponding to the extreme mode, wherein the hard constraints limit the load reduction of the critical load to zero.

[0031] The tiered penalty configuration unit is used to configure non-power supply penalty weights for high-priority loads and general loads, with the weight of high-priority loads being greater than that of general loads.

[0032] The priority closed-loop module includes an orderly power consumption execution unit, which is used to execute a load reduction plan based on the load classification results, implement load reduction from low to high load priority, and implement load restoration from high to low load priority.

[0033] As a preferred embodiment of the present invention, the dual-layer backup rolling module further includes:

[0034] The dominance disturbance solution unit is used to solve the dominance disturbance case under the constraint of the uncertainty set of prediction error. The dominance disturbance case is the prediction error case that leads to the most unfavorable system security constraints. The solution is performed according to the sequential optimization rule of first minimizing the security constraint margin variable and then maximizing the supply-demand gap under the constraint that the security constraint margin variable reaches its minimum value. The security constraint margin variable is defined by the minimum value of the remaining margin of each security constraint. The supply-demand gap is the non-negative gap corresponding to the difference between load demand and the total output of schedulable resources and new energy sources.

[0035] The extreme reserve quota calculation unit is used to calculate the additional uplink reserve capacity and additional downlink reserve capacity that meet the power balance constraints and security constraints based on the dominant disturbance situation, and output them as the extreme reserve quota to the reserve configuration unit.

[0036] As a preferred embodiment of the present invention, the dual-layer backup rolling module further includes:

[0037] The strategy parameter template storage unit is used to store a mapping table between warning levels and strategy parameter configuration templates;

[0038] The template switching unit is used to read the strategy parameter configuration template according to the mapping table and perform the switching when the warning level changes. The strategy parameter configuration template is used to define the rolling optimization parameter set, which includes the rolling optimization step size, rolling prediction window length, load priority guarantee strategy, normal standby and extreme standby configuration ratio range, and emergency resource activation switch. The template switching unit outputs the rolling optimization parameter set to the solution execution unit.

[0039] As a preferred embodiment of the present invention, the dual-layer backup rolling module further includes:

[0040] A reward construction unit is used to construct a weighted negative cost reward for reinforcement learning strategy optimization. The reward is obtained by weighted summation of a power outage penalty term, an energy storage cycle degradation term, and an operating cost term.

[0041] The feasible domain calculation unit is used to calculate the allowable range of charging power and the allowable range of discharging power based on the energy storage state of charge, power upper and lower limits and ramping capability.

[0042] The action constraint unit is used to limit the charge and discharge power control quantity output by the reinforcement learning policy to within the allowable range of charge power and discharge power through a limiting mapping, and generate candidate control commands.

[0043] The candidate solution injection unit is used to input candidate control instructions as initial solutions or candidate solutions to the solution execution unit for solving the rolling optimization problem.

[0044] As a preferred embodiment of the present invention, the dual-layer backup rolling module includes a backup takeover unit, which outputs a backup control command when the following triggering conditions are met: the time taken to solve the rolling optimization problem exceeds a preset computation time threshold; the solution execution unit returns to an infeasible state; the solution execution unit returns to a non-converged state; the scheduling command output by the solution execution unit fails to pass the preset safety verification rules; the backup control command is output to the priority closed-loop module and executed by the priority closed-loop module; a preset conservative strategy is used to generate the backup control command, which includes limiting correction based on the scheduling command output in the previous rolling cycle that meets the power balance constraints and safety constraints, and priority power supply correction based on the critical load guarantee constraints corresponding to the critical load hard constraint generation unit.

[0045] On the other hand, the present invention also provides a scheduling optimization method for high-proportion renewable energy power grids under extreme weather conditions, applied to the scheduling optimization system for high-proportion renewable energy power grids under extreme weather conditions as described above, comprising the following steps:

[0046] S1: Acquire meteorological data and power grid operation data, and generate environmental characteristics based on the meteorological data and power grid operation data;

[0047] S2: Based on environmental characteristics, output point prediction results of new energy output and load, and output distribution-independent uncertainty characterization results, including prediction error interval and prediction error uncertainty set;

[0048] S3: Generate an early warning level based on the point prediction results and uncertainty characterization results. The early warning level is determined according to the comparison results of the continuous shortage risk quantity and the first preset threshold and the comparison results of the power fluctuation risk quantity and the second preset threshold, according to the preset mapping rules. The continuous shortage risk quantity is obtained by aggregating the power shortage quantity according to the preset aggregation rules, and the power fluctuation risk quantity is obtained by aggregating the power fluctuation overshoot quantity according to the preset aggregation rules.

[0049] S4: Switch between normal mode and extreme mode and perform rolling optimization based on the warning level. When the warning level reaches or exceeds the preset switching threshold, switch to extreme mode. In extreme mode, configure dual-layer backup of normal backup and extreme backup, and set backup deliverability constraints. The backup deliverability constraints include window connection constraints. The window connection constraints are used to limit the backup adjustment amount of each schedulable resource unit at the rolling window connection point to not exceed the corresponding ramping capacity.

[0050] S5: Determine the extreme reserve quota based on the prediction error scenario that minimizes the system security constraint margin within the prediction error uncertainty set, and select the prediction error scenario with the largest supply-demand gap to determine the extreme reserve quota under the same security constraint margin. The extreme reserve quota includes additional uplink reserve capacity and additional downlink reserve capacity.

[0051] S6: Based on rolling optimization, output scheduling instructions and execute scheduling instructions, transmit execution feedback back, and implement orderly power consumption and emergency resource call according to preset load priority when preset emergency trigger conditions are triggered.

[0052] The beneficial effects of this invention are as follows: This invention outputs point prediction results and provides the prediction error range and prediction error uncertainty set by the distribution-independent prediction module without relying on the assumption of prediction error distribution, enabling scheduling decisions to cope with the significantly enhanced uncertainty of the operating environment under extreme weather conditions based on robust uncertainty input; the continuous risk warning module quantifies and maps the two types of risks, power shortage and power fluctuation, into warning levels according to preset aggregation rules, realizing the hierarchical characterization and pre-triggered response of risk situation, reducing misjudgment and improving response timeliness. The dual-layer reserve rolling module switches between normal and extreme modes based on the warning level. In extreme mode, it employs a dual-layer configuration of normal and extreme reserves, and limits the adjustment of reserve settings at the boundary of the rolling window to not exceed the corresponding ramp-up capacity through window connection constraints. This elevates reserves from capacity-configurable to cross-cycle deliverable, reducing the risk of reserve failure due to execution limitations. Within the set of prediction error uncertainties, it selects the prediction error scenario that minimizes the system safety constraint margin and maximizes the supply-demand gap under the same safety constraint margin, and determines the additional uplink and downlink reserve capacity accordingly. This ensures that the extreme reserve quota directly corresponds to the safety and supply-demand weaknesses under the most unfavorable disturbance, improving the safety margin and supply-demand guarantee capability under extreme conditions while controlling the reserve scale. The priority closed-loop module executes the scheduling instructions output by the rolling optimization and sends back feedback to form a closed-loop control. When emergency conditions are triggered, it implements orderly power consumption and emergency resource allocation according to preset load priorities, balancing critical load protection, system safety constraint satisfaction, and resource utilization efficiency, thereby improving the operational resilience and continuous power supply capability of the high-proportion renewable energy grid under extreme weather conditions. Attached Figure Description

[0053] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of 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. Wherein:

[0054] Figure 1 This is a schematic diagram of the modular structure of the system of the present invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention are within the scope of protection of the present invention.

[0056] like Figure 1 As shown, this is an embodiment of the present invention, which provides a dispatch optimization system for a high-proportion renewable energy power grid under extreme weather conditions, comprising:

[0057] (1) Distribution-independent prediction module

[0058] In this embodiment, the distribution-independent prediction module outputs baseline predictions and uncertainty descriptions for scheduling in each rolling cycle, providing input for subsequent risk warnings and rolling optimization. It outputs point prediction results for new energy output and load, and outputs distribution-independent uncertainty characterization results, including prediction error ranges and prediction error uncertainty sets. The distribution-independent prediction module includes: an environmental feature generation unit, a point prediction unit, a distribution-independent calibration unit, and a coverage closed-loop recalibration unit.

[0059] An environmental feature generation unit is used to generate environmental features based on meteorological data and power grid operation data. In implementation, meteorological data and power grid operation data can be aligned at a preset time granularity and output as feature vectors for drive point prediction and extreme condition identification. The region can be a region obtained by scheduling partitioning or node aggregation, and the preset time granularity is consistent with the rolling scheduling cycle.

[0060] The point prediction unit outputs point prediction results of renewable energy output and load based on environmental characteristics, and forms net load prediction error samples for each region based on the point prediction results and historical observation data. These prediction error samples serve as calibration samples. To maintain consistency, net load can be defined as: Net Load = Load - Renewable Energy Output; Net Load Prediction Error = Actual Net Load Value - Point Prediction Net Load Value. The point prediction unit calculates the error samples for each region based on historical observations and point predictions and maintains the calibration sample set.

[0061] A distribution-independent calibration unit is used to generate extreme operating condition indications based on environmental characteristics. Based on these indications, it performs gating selection on prediction error samples and sets sample weights, generating lower and upper limits for prediction errors in each region, an upper bound for the total net load prediction error of the entire network, and constructing a prediction error uncertainty set. This set satisfies the condition that the prediction errors in each region are between their corresponding upper and lower limits, and the total net load prediction error of the entire network does not exceed the upper bound. In implementation, extreme operating condition indications (characterizing the degree of extremeness / risk level) are first obtained from environmental characteristics. Then, gating selection is performed according to these indications (e.g., selecting samples of the same or more extreme levels) and weights are set (giving higher weights to samples closer to the current operating conditions). The upper and lower limits and the total upper bound can be obtained through sorting and statistics: the weighted quantiles of the weighted error samples are used to obtain the lower / upper limits for each region; the weighted quantiles of the total error samples (the sum of the absolute values ​​of errors in each region at each moment) are used to obtain the total upper bound. This forms an uncertainty set containing regional upper and lower bound constraints plus the total upper bound constraint for the entire network, and outputs the prediction error interval and the uncertainty set.

[0062] The coverage closed-loop recalibration unit is used to statistically analyze the proportion of actual prediction errors falling into the prediction error uncertainty set during rolling operation. When this proportion falls below a preset coverage threshold, recalibration is performed, updating the lower limit of prediction error, the upper limit of prediction error, the upper bound of the total net load prediction error for the entire network, and the prediction error uncertainty set. Coverage determination can be based on the following criteria: if the error in each region is within its corresponding upper and lower limits at a given time and the total error for the entire network does not exceed the upper bound, then it is determined to fall into the set; coverage proportion = number of times falling into the set within the monitoring window / total number of times. If the coverage proportion falls below the threshold, recalibration is triggered, for example, by adjusting the confidence level parameter, updating the gating rule threshold, updating the weight allocation, or updating the sample window length, and outputting the updated upper and lower limits, upper bound, and uncertainty set. After recalibration is triggered, the lower limit of prediction error, the upper limit of prediction error, and the upper bound of the total error are regenerated based on the updated gating rules, sample weights, and confidence level, and the prediction error uncertainty set is updated accordingly for subsequent rolling cycles.

[0063] (2) Continuous risk warning module

[0064] In this embodiment, the continuous risk warning module is used to assess the supply and demand shortage risk and power fluctuation risk that may be caused by point prediction and uncertainty disturbances during rolling operation, and output a warning level to drive subsequent scheduling strategy adjustments. It is used to generate a warning level based on point prediction results and uncertainty characterization results. The warning level is determined according to a preset mapping rule based on the comparison results of the continuous shortage risk quantity and a first preset threshold, and the comparison results of the power fluctuation risk quantity and a second preset threshold. The continuous shortage risk quantity is obtained by aggregating the power shortage quantity according to a preset aggregation rule, and the power fluctuation risk quantity is obtained by aggregating the power fluctuation overshoot quantity according to a preset aggregation rule. The continuous risk warning module includes: a scenario generation unit, a risk component calculation unit, a risk quantity generation unit, and a warning level determination unit.

[0065] A scenario generation unit is used to generate a risk calculation scenario set. This risk calculation scenario set includes scenarios generated based on a prediction error uncertainty set sampled according to a preset sampling rule, and scenarios matched from a preset extreme scenario library based on extreme condition indications according to a preset matching rule. In implementation, the risk calculation scenario set can consist of two parts:

[0066] Sampling scenario: Generate multiple prediction error scenarios within the prediction error uncertainty set according to preset sampling rules, and superimpose these error scenarios onto the point prediction trajectory to form the corresponding net load scenario; the sampling rules can be random sampling, stratified sampling, or sampling by boundary / typical points, etc.

[0067] Library Matching Scenarios: When the extreme operating condition indicator represents the impact of extreme weather, a historical / simulation scenario similar to the current operating condition is selected from a preset extreme scenario library according to preset matching rules as the risk calculation scenario; the matching rules can be based on meteorological type, intensity level, regional characteristics, or similarity index. The above two types of scenarios can be used in parallel, or one can be selected according to a preset strategy or a weighted fusion can be performed.

[0068] The risk component calculation unit is used to assess power shortage and power fluctuation overshoot based on point prediction results and uncertainty characterization results in various scenarios. Available output and ramp-up capacity of schedulable resources are reduced according to extreme weather availability parameters. In specific implementation, the risk component calculation unit calculates risk components for each scenario and each time step, including at least power shortage and power fluctuation overshoot.

[0069] The power shortage can be calculated as the non-negative part of the supply-demand gap. For example, the sum of the output of new energy sources and the output of dispatchable resources can be subtracted from the load demand at this time step (which can be superimposed with the demand after the planned load reduction), and the part that is greater than zero is taken as the power shortage.

[0070] Power fluctuation overshoot can be calculated as the non-negative portion of the mismatch between net load change demand and available ramping capacity. For example, it can be calculated by subtracting the sum of available ramping capacity of dispatchable resources in that direction from the net load change (increase or decrease) in adjacent time steps, and taking the portion greater than zero as the power fluctuation overshoot. The available output and ramping capacity of dispatchable resources can be reduced based on extreme weather availability parameters, such as scaling the rated capacity according to a preset reduction factor or availability ratio, to reflect capacity reductions caused by equipment damage, fuel constraints, environmental impacts, etc.

[0071] The risk component calculation unit is also used to incorporate planned load reduction measures during the assessment. These planned load reduction measures are taken from the most recent load reduction plan output by the dual-layer standby rolling module. In implementation, the planned load reduction measures can be used as adjustments to load demand and directly applied to the calculation of the aforementioned power shortage and net load changes to ensure consistency between the risk assessment and the rolling scheduling plan.

[0072] The risk quantity generation unit is used to perform time-based and scenario-based aggregation of power shortage and power fluctuation overshoot, respectively, to generate persistent shortage risk and power fluctuation risk. In implementation, time aggregation can employ methods such as summation, mean, maximum value, or weighted cumulative aggregation to characterize the persistence or peak value of the risk. Scenario aggregation can employ methods such as maximum value, quantile, or expected value to characterize the uncertainty impact across scenarios. For example, the power shortage within each scenario can be summed over time within a rolling window to obtain the cumulative shortage for that scenario, and then the quantile or maximum value can be taken across all scenarios to obtain the persistent shortage risk. Similarly, the maximum overshoot of power fluctuation can be obtained by taking the maximum value over time to form the maximum overshoot for that scenario, and then the quantile or maximum value can be taken across all scenarios to obtain the power fluctuation risk. These aggregation methods can be uniformly configured by preset aggregation rules to adapt to different risk preferences and operating strategies.

[0073] The early warning level determination unit compares the continuous shortage risk with a first preset threshold and the power fluctuation risk with a second preset threshold, and outputs the early warning level according to a preset mapping rule. In implementation, the preset mapping rule can be a lookup table or a rule set: first, the comparison results of whether the continuous shortage risk exceeds the first preset threshold and whether the power fluctuation risk exceeds the second preset threshold are obtained separately; then, the comparison results are combined and mapped to the early warning level. The early warning level can be a multi-level discrete level, used for subsequent operation mode switching and resource strategy adjustment. The early warning level determination unit can update the early warning level once in each rolling cycle and use it as input for the subsequent two-layer backup rolling module.

[0074] (3) Double-layer spare rolling module

[0075] This module is used to switch between normal and extreme modes and perform rolling optimization based on the warning level. When the warning level reaches or exceeds the preset switching threshold, it switches to extreme mode. The dual-layer backup rolling module configures dual-layer backup of normal and extreme backup in extreme mode and sets backup deliverability constraints. Backup deliverability constraints include window connection constraints, which limit the backup adjustment amount of each schedulable resource unit at the rolling window connection to not exceed the corresponding ramping capacity. The extreme backup quota is determined based on the prediction error scenario that results in the minimum system safety constraint margin within the prediction error uncertainty set. Under the condition of the same safety constraint margin, the prediction error scenario with the largest supply-demand gap is selected, including additional uplink backup capacity and additional downlink backup capacity.

[0076] In one embodiment, to ensure that the mathematical expressions for solving subsequent deliverability constraints and dominance perturbations strictly correspond to the terminology of the claims, the region set is defined as follows: ,in Representing each region; the set of schedulable resource units is ,in Represents each schedulable resource unit; the preset duration level set is: ,in This indicates a preset duration level, and the duration level... Corresponding preset duration The rolling cycle index is (Corresponding to the junction of adjacent scrolling cycles / scrolling windows).

[0077] The dual-layer backup rolling module includes:

[0078] The mode switching unit is used to switch between normal mode and extreme mode according to the warning level;

[0079] The backup configuration unit is used to configure a two-layer backup of normal backup and extreme backup in extreme mode. The extreme backup quota is determined based on the uncertainty characterization result and allocated to the uplink backup and downlink backup according to the preset duration level.

[0080] A deliverability constraint construction unit is used to construct backup deliverability constraints in the rolling optimization problem. These backup deliverability constraints include: a window connection constraint, used to limit the backup adjustment amount of each schedulable resource unit at the boundary of adjacent rolling cycles to not exceed the ramp-up capacity; a hierarchical consistency constraint, used to limit the backup capacity between hierarchical levels of different durations to meet the incremental constraint condition, wherein the backup capacity of the longer duration hierarchical level is not less than the backup capacity of the shorter duration hierarchical level; and a differential coverage constraint, used to limit the uplink and downlink backup of each schedulable resource unit within the same duration hierarchical level to cover the net load increase and decrease changes, respectively, based on the net load change demand within the rolling window; the differential coverage constraint is expressed using a capacity inequality, where the sum of the uplink backup capacity of each schedulable resource unit within the same duration hierarchical level is not less than the net load increase change demand, and the sum of the downlink backup capacity is not less than the net load decrease change demand.

[0081] In one embodiment, uplink reserve and downlink reserve can be denoted as follows: and and resources The uphill / downhill climbing ability is denoted as follows: , The ramp-up capacity is measured on a single rolling cycle timescale and can be discounted according to extreme weather availability parameters to reflect deliverability. The net load increase change demand and net load decrease change demand can be denoted as follows: , This indicates that within the rolling prediction window, the duration is... The net load increase / decrease change demand is statistically obtained over a time span. This change demand can be obtained by differentiating the net load across adjacent time steps within a rolling window, taking the non-negative parts of the positive and negative differences respectively, and overlaying planned load reduction measures as needed to maintain consistency. To ensure consistency between reserve adjustment and ramp-up capacity, the window connection constraint adopts a bilateral constraint form.

[0082] (1) Window connection constraints:

[0083] ;

[0084] (2) Hierarchical consistency constraint (the hierarchy with a longer duration is not less than the hierarchy with a shorter duration):

[0085] ;

[0086] (3) Difference Cover Constraint:

[0087] ;

[0088] In one embodiment, to provide an executable requirement statistics metric, let the set of time indices within the scrolling window be... The net load of the entire network scenario is The following example can be used:

[0089] ;

[0090] The solution execution unit is used to solve the rolling optimization problem with deliverability constraints in each rolling cycle and output the scheduling instructions for the current rolling cycle.

[0091] The dual-layer backup rolling module and the priority closed-loop module adopt the same load priority rule; the dual-layer backup rolling module includes:

[0092] The load classification unit is used to divide the load into critical loads, high priority loads, and general loads;

[0093] A critical load hard constraint generation unit is used to generate critical load hard constraints in the rolling optimization problem corresponding to the extreme mode, wherein the hard constraints limit the load reduction of the critical load to zero.

[0094] The tiered penalty configuration unit is used to configure non-power supply penalty weights for high-priority loads and general loads, with the weight of high-priority loads being greater than that of general loads.

[0095] The priority closed-loop module includes an orderly power consumption execution unit, which is used to execute a load reduction plan based on the load classification results, implement load reduction from low to high load priority, and implement load restoration from high to low load priority.

[0096] The dual-layer backup rolling module also includes:

[0097] The dominance disturbance solution unit is used to solve the dominance disturbance case under the constraint of the uncertainty set of prediction error. The dominance disturbance case is the prediction error case that leads to the most unfavorable system security constraints. The solution is performed according to the sequential optimization rule of first minimizing the security constraint margin variable and then maximizing the supply-demand gap under the constraint that the security constraint margin variable reaches its minimum value. The security constraint margin variable is defined by the minimum value of the remaining margin of each security constraint. The supply-demand gap is the non-negative gap corresponding to the difference between load demand and the total output of schedulable resources and new energy sources.

[0098] In one embodiment, the regional net load forecasting error vector is denoted as... The set of prediction error uncertainties is denoted as And it must satisfy the following conditions: the regional prediction error is between the upper and lower limits and the total net load prediction error of the entire network does not exceed the upper limit of the total, specifically:

[0099] ;

[0100] in: , These correspond to the lower and upper limits of the regional prediction error, respectively. This corresponds to the upper bound of the total net load forecast error for the entire network. To avoid ambiguity regarding the point of error application, the error is superimposed on the predicted net load at that point to form the scenario net load, satisfying:

[0101] ;

[0102] in: Forecast net load for regional points, Let be the scenario value of the regional net load under disturbance conditions. Define the set of safety constraints as follows: Safety constraints The remaining margin is When the constraints are satisfied When the constraints are violated Define the safety constraint margin variable:

[0103] ;

[0104] The sequential optimization rule adopts a two-stage definition: the first stage is in Minimize within The second stage maximizes the supply-demand gap under the constraint of the optimal value in the first stage. :

[0105] ;

[0106] in: The supply-demand gap (non-negative) can be defined as the non-negative portion of the net load demand exceeding the available supply capacity under a given scenario. Available supply capacity includes at least the total output of dispatchable resources and the total output of new energy sources, and can be reduced by the availability parameter for extreme weather conditions. In engineering implementation, equivalent constraints can be implemented using numerical tolerance; the equivalent notation is used to define sequential rules.

[0107] The extreme reserve quota calculation unit is used to calculate the additional uplink reserve capacity and additional downlink reserve capacity that meet the power balance constraints and security constraints based on the dominant disturbance situation, and output them as the extreme reserve quota to the reserve configuration unit.

[0108] In one embodiment, the additional uplink reserve capacity and the additional downlink reserve capacity are respectively denoted as... and

[0109] It can also be defined using the minimum additional reserve method:

[0110] ;

[0111] in: At least it indicates that: in the case of dominant disturbance Under the corresponding net load scenario, there exists a set of schedulable resource output adjustments and reserve calls that ensure power balance constraints are met and all safety constraints are satisfied; and , The maximum available capacity in the uplink and downlink directions is limited respectively; the resulting extreme reserve quota is allocated to the uplink reserve and downlink reserve according to the preset duration level.

[0112] The dual-layer backup rolling module also includes:

[0113] The strategy parameter template storage unit is used to store a mapping table between warning levels and strategy parameter configuration templates;

[0114] The template switching unit is used to read the strategy parameter configuration template according to the mapping table and perform the switching when the warning level changes. The strategy parameter configuration template is used to define the rolling optimization parameter set, which includes the rolling optimization step size, rolling prediction window length, load priority guarantee strategy, normal standby and extreme standby configuration ratio range, and emergency resource activation switch. The template switching unit outputs the rolling optimization parameter set to the solution execution unit.

[0115] The dual-layer backup rolling module also includes:

[0116] A reward construction unit is used to construct a weighted negative cost reward for reinforcement learning strategy optimization. The reward is obtained by weighted summation of a power outage penalty term, an energy storage cycle degradation term, and an operating cost term.

[0117] The feasible domain calculation unit is used to calculate the allowable range of charging power and the allowable range of discharging power based on the energy storage state of charge, power upper and lower limits and ramping capability.

[0118] The action constraint unit is used to limit the charge and discharge power control quantity output by the reinforcement learning policy to within the allowable range of charge power and discharge power through a limiting mapping, and generate candidate control commands.

[0119] The candidate solution injection unit is used to input candidate control instructions as initial solutions or candidate solutions to the solution execution unit for solving the rolling optimization problem.

[0120] In one embodiment, candidate control commands are processed by actionable domain and limit mapping before injection to ensure that power upper and lower limits and ramping capability constraints are met; candidate control commands are only used as initial solutions or candidate solutions for rolling optimization, and the final scheduling command is based on the rolling optimization solution results that meet power balance constraints and safety constraints.

[0121] The dual-layer backup rolling module includes a backup takeover unit, which outputs backup control commands when the following triggering conditions are met: the time taken to solve the rolling optimization problem exceeds a preset computation time threshold; the solution execution unit returns to an infeasible state; the solution execution unit returns to a non-converged state; the scheduling command output by the solution execution unit fails to pass the preset safety verification rules; the backup control command is output to the priority closed-loop module and executed by the priority closed-loop module; a preset conservative strategy is used to generate backup control commands, which includes limiting correction based on the scheduling command output in the previous rolling cycle that meets the power balance constraints and safety constraints, and priority power supply correction based on the critical load guarantee constraints corresponding to the critical load hard constraint generation unit.

[0122] (4) Priority closed-loop module

[0123] It is used to execute the scheduling instructions output by the dual-layer backup rolling module and send back the execution feedback, and to implement orderly power consumption and emergency resource call according to the preset load priority when the preset emergency triggering conditions are triggered.

[0124] In one embodiment, the priority closed-loop module serves as a closed-loop execution layer for planning, execution, feedback, and re-optimization. It adopts the same load priority rules as the dual-layer backup rolling module and provides data input for the status update, window connection constraint construction, and resource availability reduction parameter update of the next rolling cycle through feedback fields.

[0125] 1) Instruction execution and feedback

[0126] The priority closed-loop module receives the scheduling instructions output by the dual-layer backup rolling module and issues them for execution. The scheduling instructions include at least: output setting, energy storage charging and discharging, interruptible load calling, backup setting, load reduction plan, etc., and collects and transmits the execution results.

[0127] The feedback cycle is the same as or a sub-cycle of the rolling cycle, in order to meet the real-time requirements of rolling optimization for state updates and window connection constraint construction.

[0128] To ensure closure with the standby deliverability constraint (window connection constraint) and the scope of the dominant disturbance / extreme standby calculation in (3), the feedback fields shall include at least:

[0129] Actual output, actual charging / discharging power, actual interruptible load call-up, actual load reduction, and graded load execution status;

[0130] Resource availability status information includes at least: available output, available ramping capacity, and standby settings and delivery information such as energy storage state of charge (SOC). It also includes at least: the uplink / downlink standby settings of each schedulable resource unit at each duration level in the previous rolling cycle. , And the corresponding actual standby call volume / occupancy or standby available reserve; among which Used in the constraints for connecting the next rolling cycle window Construction and verification.

[0131] Emergency triggering and response results indicators should include at least: whether it was triggered, the type of triggering reason, the amount of emergency resources called up and the execution result, which will be used for risk assessment and strategy parameter template switching in the next rolling cycle.

[0132] The feedback information serves as input for the next rolling cycle optimization, used to complete state updates, window connection constraint construction, and closed-loop correction of load reduction plans and resource availability parameters.

[0133] 2) Emergency Triggering and Orderly Power Consumption

[0134] When the emergency triggering conditions are met, the handling process is initiated; orderly power consumption is carried out according to the preset load priority: load reduction is implemented from low to high, and load restoration is implemented from high to low, and the results of each step are included in the feedback and transmission to ensure consistency between planning, execution and optimization.

[0135] In one embodiment, the preset emergency triggering conditions include at least one of the following types or a combination thereof: a continuous supply-demand gap exceeding a preset threshold, a critical safety constraint margin below a preset threshold, a critical equipment / communication anomaly causing instructions to be unreliably executed, a sudden drop in resource availability causing the original scheduling instructions to be undeliverable, etc.; after triggering, critical load protection related strategies are executed first, and coordinated with load reduction plans to quickly restore a safe operating state.

[0136] 3) Emergency resource mobilization

[0137] After an emergency is triggered, emergency resources (such as rapid start-up power, energy storage, demand response, etc.) are called up according to the preset activation switch and call priority rules. The amount called up can be determined by the scale of the supply and demand gap or the safety margin requirement, and is coordinated with orderly power consumption to restore the safe operating state as soon as possible.

[0138] In one embodiment, the execution results of emergency resource mobilization (including the mobilization object, mobilization power / capacity, response time and actual effective amount) and the implementation results of orderly power consumption are sent back to the dual-layer backup rolling module for updating the net load caliber, available supply capacity caliber and constraint feasibility in the next rolling cycle.

[0139] like Figure 2 As shown, another embodiment of the present invention provides a scheduling optimization method for a high-proportion renewable energy power grid under extreme weather conditions, applied to the scheduling optimization system for a high-proportion renewable energy power grid under extreme weather conditions as described above, including the following steps:

[0140] S1: Acquire meteorological data and power grid operation data, and generate environmental characteristics based on the meteorological data and power grid operation data;

[0141] S2: Based on environmental characteristics, output point prediction results of new energy output and load, and output distribution-independent uncertainty characterization results, including prediction error interval and prediction error uncertainty set;

[0142] S3: Generate an early warning level based on the point prediction results and uncertainty characterization results. The early warning level is determined according to the comparison results of the continuous shortage risk quantity and the first preset threshold and the comparison results of the power fluctuation risk quantity and the second preset threshold, according to the preset mapping rules. The continuous shortage risk quantity is obtained by aggregating the power shortage quantity according to the preset aggregation rules, and the power fluctuation risk quantity is obtained by aggregating the power fluctuation overshoot quantity according to the preset aggregation rules.

[0143] S4: Switch between normal mode and extreme mode and perform rolling optimization based on the warning level. When the warning level reaches or exceeds the preset switching threshold, switch to extreme mode. In extreme mode, configure dual-layer backup of normal backup and extreme backup, and set backup deliverability constraints. The backup deliverability constraints include window connection constraints. The window connection constraints are used to limit the backup adjustment amount of each schedulable resource unit at the rolling window connection point to not exceed the corresponding ramping capacity.

[0144] S5: Determine the extreme reserve quota based on the prediction error scenario that minimizes the system security constraint margin within the prediction error uncertainty set, and select the prediction error scenario with the largest supply-demand gap to determine the extreme reserve quota under the same security constraint margin. The extreme reserve quota includes additional uplink reserve capacity and additional downlink reserve capacity.

[0145] S6: Based on rolling optimization, output scheduling instructions and execute scheduling instructions, transmit execution feedback back, and implement orderly power consumption and emergency resource call according to preset load priority when preset emergency trigger conditions are triggered.

[0146] In summary, this invention constructs an integrated closed-loop scheduling framework for extreme weather, encompassing forecasting, early warning, rolling optimization, execution feedback, and more. This framework forms a complete technology chain from uncertainty characterization to backup configuration and emergency response, and can be integrated with existing scheduling automation, meteorological, and operational monitoring systems for online deployment. This solution is suitable for the operation and management of areas with high wind and solar power integration under extreme weather disturbances such as typhoons, cold waves, heavy rains, and heat waves. It provides more engineering-executable decision support for scheduling strategies and has good engineering applicability and promising prospects for widespread application.

[0147] 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 it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the claims of the present invention.

Claims

1. A dispatch optimization system for a high-proportion renewable energy power grid under extreme weather conditions, characterized in that, include: The distribution-independent prediction module is used to output point prediction results of new energy output and load, and output distribution-independent uncertainty characterization results, which include prediction error interval and prediction error uncertainty set. The continuous risk warning module is used to generate warning levels based on point prediction results and uncertainty characterization results. The warning level is determined according to the comparison results of the continuous shortage risk quantity and the first preset threshold and the comparison results of the power fluctuation risk quantity and the second preset threshold, according to the preset mapping rules. The sustained shortage risk is obtained by aggregating the power shortage amount according to a preset aggregation rule, and the power fluctuation risk is obtained by aggregating the power fluctuation overshoot amount according to a preset aggregation rule; A dual-layer backup rolling module is used to switch between normal mode and extreme mode and perform rolling optimization based on the warning level. When the warning level reaches or exceeds a preset switching threshold, it switches to extreme mode. The dual-layer backup rolling module configures dual-layer backup of normal backup and extreme backup in extreme mode, and sets backup deliverability constraints. Backup deliverability constraints include window connection constraints, which limit the backup adjustment amount of each schedulable resource unit at the rolling window connection to not exceed the corresponding ramping capacity. The extreme backup quota is determined based on the prediction error scenario that leads to the minimum system safety constraint margin within the prediction error uncertainty set, and selects the prediction error scenario with the largest supply-demand gap when the safety constraint margin is the same, including additional uplink backup capacity and additional downlink backup capacity. The priority closed-loop module is used to execute the scheduling instructions output by the dual-layer backup rolling module and send back the execution feedback. When the preset emergency triggering conditions are triggered, it implements orderly power consumption and emergency resource call according to the preset load priority.

2. The dispatch optimization system for high-proportion renewable energy power grids under extreme weather conditions according to claim 1, characterized in that, The distribution-independent prediction module includes: An environmental feature generation unit is used to generate environmental features based on meteorological data and power grid operation data. A point prediction unit is used to output point prediction results of new energy output and load based on environmental features, and to form net load prediction error samples for each region based on the point prediction results and historical observation data. The prediction error samples are used as calibration samples. A distribution-independent calibration unit is used to generate extreme operating condition indications based on environmental features, to perform gating selection and set sample weights for prediction error samples according to extreme operating condition indications, to generate the lower limit value, upper limit value, and upper bound of the total net load prediction error for each region, and to construct a prediction error uncertainty set. The prediction error uncertainty set satisfies that the prediction error for each region is between the corresponding upper and lower limits, and the total net load prediction error for the entire network does not exceed the total upper bound. A coverage closed-loop recalibration unit is used to count the proportion of actual prediction errors falling into the prediction error uncertainty set during rolling operation, and to perform a recalibration operation when the proportion is lower than a preset coverage threshold, updating the lower limit value, upper limit value, upper bound of the total net load prediction error for the entire network, and prediction error uncertainty set.

3. The dispatch optimization system for high-proportion renewable energy power grids under extreme weather conditions according to claim 1, characterized in that, The continuous risk warning module includes: The scenario generation unit generates a risk calculation scenario set, which includes scenarios generated by sampling based on the prediction error uncertainty set according to a preset sampling rule and scenarios matched from a preset extreme scenario library based on extreme operating condition indications according to a preset matching rule. The risk component calculation unit evaluates power shortage and power fluctuation overshoot based on point prediction results and uncertainty characterization results in each scenario, and reduces the available output and ramp-up capability of schedulable resources according to extreme weather availability parameters. The risk quantity generation unit performs time aggregation and scenario aggregation on the power shortage and power fluctuation overshoot, respectively, to generate a continuous shortage risk quantity and a power fluctuation risk quantity. The warning level determination unit compares the continuous shortage risk quantity with a first preset threshold and the power fluctuation risk quantity with a second preset threshold, and outputs a warning level according to a preset mapping rule. The risk component calculation unit also incorporates planned load reduction measures during the evaluation, which are taken from the most recent load reduction plan output by the dual-layer standby rolling module.

4. The dispatch optimization system for high-proportion renewable energy power grids under extreme weather conditions according to claim 1, characterized in that, The dual-layer backup rolling module includes: The system includes a mode switching unit for switching between normal and extreme modes based on the warning level; a backup configuration unit for configuring a dual-layer backup system (normal and extreme backups) in extreme mode, where the extreme backup quota is determined based on uncertainty characterization results and allocated to uplink and downlink backups according to preset duration levels; and a deliverability constraint construction unit for constructing backup deliverability constraints in the rolling optimization problem. These constraints include: window connection constraints, limiting the backup adjustment amount of each schedulable resource unit at the boundary of adjacent rolling cycles to not exceed the ramp-up capacity; and hierarchy consistency constraints, ensuring that the backup capacity between different duration levels meets the incremental constraint condition. The constraints are as follows: the reserve capacity of the longer duration level is not less than the reserve capacity of the shorter duration level; differential coverage constraint is used to limit the uplink and downlink reserves of each schedulable resource unit within the same duration level to cover the net load increase and decrease changes, respectively, based on the net load change demand within the rolling window; the differential coverage constraint is expressed by the capacity inequality, whereby the sum of the uplink reserve capacity of each schedulable resource unit within the same duration level is not less than the net load increase change demand, and the sum of the downlink reserve capacity is not less than the net load decrease change demand; the solution execution unit is used to solve the rolling optimization problem with deliverability constraints in each rolling cycle and output the scheduling instruction for the current rolling cycle.

5. The dispatch optimization system for high-proportion renewable energy power grids under extreme weather conditions according to claim 1, characterized in that, The dual-layer backup rolling module and the priority closed-loop module adopt the same load priority rule. The dual-layer backup rolling module includes: a load classification unit for dividing the load into critical loads, high-priority loads, and general loads; a critical load hard constraint generation unit for generating critical load hard constraints in the rolling optimization problem corresponding to the extreme mode, wherein the hard constraints limit the load reduction of critical loads to zero; and a graded penalty configuration unit for configuring no-power penalty weights for high-priority loads and general loads, wherein the weight of high-priority loads is greater than that of general loads. The priority closed-loop module includes an ordered power consumption execution unit for executing the load reduction plan according to the load classification results, implementing load reduction from low to high load priority, and implementing load restoration from high to low load priority.

6. The dispatch optimization system for high-proportion renewable energy power grids under extreme weather conditions according to claim 1, characterized in that, The dual-layer backup rolling module also includes: The dominant disturbance solution unit is used to solve the dominant disturbance case under the constraint of the uncertainty set of prediction error. The dominant disturbance case is the prediction error case that leads to the most unfavorable system security constraints. The solution is performed according to the sequential optimization rule of first minimizing the security constraint margin variable and then maximizing the supply-demand gap under the constraint that the security constraint margin variable reaches its minimum value. The security constraint margin variable is defined by the minimum value of the remaining margin of each security constraint. The supply-demand gap is the non-negative gap corresponding to the difference between load demand and the total output of dispatchable resources and new energy sources. The extreme reserve quota calculation unit is used to calculate the additional uplink reserve capacity and additional downlink reserve capacity that meet the power balance constraints and security constraints based on the dominant disturbance case, and output them as the extreme reserve quota to the reserve configuration unit.

7. The dispatch optimization system for high-proportion renewable energy power grids under extreme weather conditions according to claim 4, characterized in that, The dual-layer backup rolling module further includes: a strategy parameter template storage unit for storing a mapping table between warning levels and strategy parameter configuration templates; and a template switching unit for reading the strategy parameter configuration template according to the mapping table and performing a switch when the warning level changes. The strategy parameter configuration template is used to define the rolling optimization parameter set, which includes the rolling optimization step size, rolling prediction window length, load priority guarantee strategy, the ratio range of normal backup and extreme backup configurations, and emergency resource activation switch. The template switching unit outputs the rolling optimization parameter set to the solution execution unit.

8. The dispatch optimization system for high-proportion renewable energy power grids under extreme weather conditions according to claim 4, characterized in that, The dual-layer backup rolling module also includes: The reward construction unit is used to construct a weighted negative cost reward for reinforcement learning strategy optimization. The reward is obtained by weighted summation of the non-power supply penalty term, the energy storage cycle degradation term, and the operating cost term. The action feasible region calculation unit is used to calculate the allowable charging power range and the allowable discharging power range based on the energy storage state of charge, power upper and lower limits, and ramping capability. The action constraint unit is used to limit the charging and discharging power control quantities output by the reinforcement learning strategy to within the allowable charging power range and the allowable discharging power range through amplitude limiting mapping, and generate candidate control commands. The candidate solution injection unit is used to input the candidate control commands as the initial solution or candidate solution for solving the rolling optimization problem into the solution execution unit.

9. The dispatch optimization system for high-proportion renewable energy power grids under extreme weather conditions according to claim 5, characterized in that, The dual-layer backup rolling module includes: The backup takeover unit outputs backup control commands when the following triggering conditions are met: the time taken to solve the rolling optimization problem exceeds a preset computation time threshold; the solution execution unit returns to an infeasible state; the solution execution unit returns to a non-converged state; the scheduling command output by the solution execution unit fails to pass the preset safety verification rules; the backup control command is output to the priority closed-loop module and executed by the priority closed-loop module; a preset conservative strategy is used to generate backup control commands. The preset conservative strategy includes limiting correction based on the scheduling command output in the previous rolling cycle that meets the power balance constraints and safety constraints, and priority power supply correction based on the critical load guarantee constraints corresponding to the critical load hard constraint generation unit.

10. A method for scheduling optimization of a high-proportion renewable energy power grid under extreme weather conditions, applied to the scheduling optimization system for a high-proportion renewable energy power grid under extreme weather conditions as described in any one of claims 1-9, characterized in that, Includes the following steps: S1: Acquire meteorological data and power grid operation data, and generate environmental characteristics based on the meteorological data and power grid operation data; S2: Based on environmental characteristics, output point prediction results of new energy output and load, and output distribution-independent uncertainty characterization results, including prediction error interval and prediction error uncertainty set; S3: Generate an early warning level based on the point prediction results and uncertainty characterization results. The early warning level is determined according to the comparison results of the continuous shortage risk quantity and the first preset threshold and the comparison results of the power fluctuation risk quantity and the second preset threshold, according to the preset mapping rules. The continuous shortage risk quantity is obtained by aggregating the power shortage quantity according to the preset aggregation rules, and the power fluctuation risk quantity is obtained by aggregating the power fluctuation overshoot quantity according to the preset aggregation rules. S4: Switch between normal mode and extreme mode and perform rolling optimization based on the warning level. When the warning level reaches or exceeds the preset switching threshold, switch to extreme mode. In extreme mode, configure dual-layer backup of normal backup and extreme backup, and set backup deliverability constraints. The backup deliverability constraints include window connection constraints. The window connection constraints are used to limit the backup adjustment amount of each schedulable resource unit at the rolling window connection point to not exceed the corresponding ramping capacity. S5: Determine the extreme reserve quota based on the prediction error scenario that minimizes the system security constraint margin within the prediction error uncertainty set, and select the prediction error scenario with the largest supply-demand gap to determine the extreme reserve quota under the same security constraint margin. The extreme reserve quota includes additional uplink reserve capacity and additional downlink reserve capacity. S6: Based on rolling optimization, output scheduling instructions and execute scheduling instructions, transmit execution feedback back, and implement orderly power consumption and emergency resource call according to preset load priority when preset emergency trigger conditions are triggered.

Images