Power intelligent dispatching system and method for extreme weather events

Abstract

The application discloses a power intelligent scheduling system and method for extreme climate events, and belongs to the field of extreme climate data analysis, and comprises the following steps: a risk characterization module is used for accessing and preprocessing extreme climate multi-source data, and each partition risk characterization facing extreme event supply guarantee constraints is formed based on the multi-source data; an upper limit checking module is used for correcting each partition net load prediction according to each partition risk characterization to obtain a conservative upper limit of each partition net load, and performing safety constraint checking based on the conservative upper limit of each partition net load to obtain a power grid key constraint state index; and a decision feedback module is used for generating a structured scheduling decision set based on each partition risk characterization, the conservative upper limit of the net load and the power grid key constraint state index. The technical problems that the existing technology causes the scheduling strategy to have response lag when the extreme event comes, the supply-demand gap of a high upper limit of a key period is not prepared, and the cross-region support plan is not sufficient in realizability are solved.

Description

Technical Field

[0001] This invention relates to the field of extreme climate data analysis technology, and in particular to a power intelligent dispatching system and method for extreme climate events. Background Technology

[0002] As global climate change intensifies, the frequency and intensity of extreme weather events such as heat waves, cold waves, torrential rains, and typhoons are constantly increasing, significantly impacting the safe and stable operation of power systems. Extreme weather events are often characterized by their suddenness, wide range of impact, and uncertain duration, easily causing rapid fluctuations in electricity load within a short period of time, and having a cumulative impact on power generation capacity, grid transmission capacity, and load-side response capabilities.

[0003] Against this backdrop, power system operation control is gradually evolving from traditional static dispatching to intelligent dispatching for extreme weather scenarios. By comprehensively considering meteorological information, power grid operating status, load change characteristics, and adjustable resource capabilities, it enables early identification and proactive control of power system operation risks, thereby ensuring safe power supply under extreme weather conditions.

[0004] For example, Chinese invention patent CN112330099B discloses a resource scheduling method for a power distribution system under extreme natural disaster weather. The power distribution system includes a power distribution network and a producer-consumer cluster. The power distribution network contains multiple nodes, and the producer-consumer cluster is connected to the power distribution network through at least one node. Mobile generators are also deployed in the power distribution system. The method includes the following steps: generating an objective function for the operating cost of the power distribution system under extreme natural disaster weather; constructing a meteorological model for extreme natural disaster weather; based on the meteorological model, constructing a damage model of extreme natural disaster weather on the power distribution system; combining the damage model, constructing an uncertain model of power distribution system failure under the influence of extreme natural disaster weather; setting constraints for the mobile generators and the loads of lines and nodes in the power distribution network; solving the objective function based on the constraints and the uncertain model to obtain the solution result; and scheduling resources in the power distribution system according to the solution result.

[0005] For example, Chinese invention patent CN119026842A discloses a multi-level collaborative dual-layer optimization scheduling method for a new type of distribution network under extreme weather conditions, including the following steps: S1, predicting icing based on weather data to obtain first data; obtaining line fault probability based on the first data; the first data includes line icing thickness data; S2, performing Monte Carlo scenario simulation based on the line fault probability to obtain a first fault scenario set; the first fault scenario set includes a preset number of fault scenarios; S3, reducing scenarios based on the first fault scenario set to obtain a second fault scenario set; the second fault scenario set includes fault scenarios with higher fault probabilities; S4, building a first model based on preset constraints and objective function; the first model includes an upper-level planning model and a lower-level operation model; S5, solving the first model based on the second fault scenario set to obtain the distribution network reconfiguration and mobile energy storage pre-configuration scheme and scheduling plan.

[0006] However, in the process of implementing the inventive technical solutions in the embodiments of this application, it was found that the above-mentioned technologies have at least the following technical problems: the above-mentioned solutions mainly focus on the disaster damage of the distribution network, the probability of line node failure, scenario set simulation and two-layer optimization, and emphasize the reconstruction, configuration and operation plan of mobile power / mobile energy storage under the fault scenario of the distribution side; however, in the extreme climate response scenarios of provincial and higher-level power grids, extreme events often exhibit the spatial characteristics of cross-regional propagation and multi-zone synchronous enhancement, as well as the time characteristics of rapid load surge or fall and rapid fluctuation of new energy output in a short period of time. Moreover, different sub-districts are coupled with key sections through tie lines, and the mutual assistance capability is affected by the changes in the safety boundary of the grid and the operation mode, and has significant time-varying and realizability constraints. In such scenarios, if modeling and scenario solving approaches still primarily rely on single-region or single-fault mechanisms, the following key gaps are likely to emerge: it is difficult to simultaneously incorporate the constraints of rapid load changes in multiple zones, weakening or fluctuating renewable energy output, and the availability of cross-regional channels into a consistent decision-making basis within the same rolling time domain. Consequently, it is difficult to form a provincial-level risk that can be characterized in advance and executed in terms of scheduling output, such as higher peak electricity demand, more sudden power supply gaps, and inability of inter-regional mutual assistance. This results in scheduling strategies exhibiting lag in response to extreme events, insufficient preparation for the upper limit of supply and demand gaps during critical periods, and insufficient feasibility of cross-regional support plans. Summary of the Invention

[0007] To address the technical problems existing in current technologies, such as delayed response of dispatch strategies during extreme events, insufficient preparation for the upper limit of supply-demand gaps during critical periods, and insufficient feasibility of cross-regional support plans, this invention provides a smart power dispatch system and method for extreme weather events. The technical solution is as follows:

[0008] On the one hand, a smart power dispatching system for extreme weather events is provided, which includes:

[0009] The 009 Risk Characterization Module is used to access and preprocess multi-source extreme climate data, and form risk characterizations for each region based on the multi-source data to meet supply constraints for extreme events. The Upper Bound Verification Module is used to correct the net load forecast of each region based on the risk characterization of each region to obtain the conservative upper bound of the net load of each region, and perform safety constraint verification based on the conservative upper bound of the net load of each region to obtain the key constraint status indicators of the power grid. The Decision Feedback Module is used to generate a structured dispatch decision set based on the risk characterization of each region, the conservative upper bound of the net load, and the key constraint status indicators of the power grid.

[0010] On the other hand, a smart power dispatching method for extreme weather events is provided, which includes:

[0011] Step 1: Access and preprocess multi-source extreme climate data to form risk characterizations for each region under extreme event supply constraints. Step 2: Correct the net load forecast for each region based on the risk characterizations to obtain conservative upper limits for net load. Perform safety constraint verification based on the conservative upper limits for net load to obtain key power grid constraint status indicators. Step 3: Generate a structured dispatch decision set based on the risk characterizations for each region, the conservative upper limits for net load, and the key power grid constraint status indicators.

[0012] The beneficial effects of the technical solutions provided by the embodiments of the present invention include at least the following:

[0013] 1. The intelligent power dispatching system and method for extreme weather events provided by this invention spatially aggregates multi-source extreme weather data by partitioning and introducing a partitioned threshold system to generate extreme intensity indicators, short-term variation amplitude, and spatial synchronicity indicators, and hierarchically forms risk representations for each partition. These risk representations are then converted into load correction coefficients according to a preset coefficient mapping relationship, and the net load forecast for each partition is corrected using a step-by-step maximization method. This yields a conservative upper bound for the net load of each partition that can cover high-level scenarios, thereby simultaneously injecting three types of risks—intensity level, enhancement rate, and simultaneous impact on multiple partitions—into the upper bound for safety, thus improving the coverage of upper-tail net load risks under extreme events and reducing reserve shortages and trigger lags caused by low upper bounds during critical periods. Compared to existing technologies that use fixed margins, single thresholds, or rely solely on single error statistics to provide safety margins, this invention uses the synergy of risk representation and correction rules to construct an upper bound that better reflects the evolutionary characteristics of extreme events, effectively solving the problem of systematically underestimating the upper tail of net load under extreme events.

[0014] 2. After obtaining the conservative upper bound of net load for each region, this invention further maps the conservative upper bound to the region-injected boundary conditions of the power grid model and performs safety constraint verification, outputting key constraint state indicators of the power grid. This allows the worst-case reasonable supply and demand scenario to be implemented on the physical boundary of the grid structure, identifying the risk of exceeding limits, approaching limits, and margin levels of key transmission channels in advance, and constraining the operational boundary and feasible range of the scheduling scheme accordingly. Simultaneously, the risk characterization of each region, the conservative upper bound of net load, and the key constraint state indicators are collaboratively input into preset decision generation rules to form a structured scheduling decision set with a fixed structure, outputting planned decisions, limit-based decisions, and trigger-based decisions. This ensures that resource pre-scheduling, boundary constraints, and trigger handling are self-consistently linked under the same logical link. Compared to the prior art where planned scheduling, limit setting, and trigger handling are fragmented and may result in situations where predictions are feasible but the grid structure cannot fulfill them, this invention achieves a seamless connection from the upper bound to the grid constraints and then to the decision output, making the scheduling results more physically executable and consistent.

[0015] 3. Under the condition of executing a structured scheduling decision set, this invention determines the execution deviation by obtaining execution receipts corresponding to each scheduling measure and comparing them with target parameters. When the execution deviation meets preset trigger conditions, the scheduling safety boundary is adaptively tightened and the scheduling decision set is updated. This incorporates the potential deviation between the planned value of the scheduling measure and the actual status of the execution receipt into safety boundary management, enabling online identification and protective tightening of insufficient response to critical resources. This avoids risk spread due to inadequate energy storage, demand response, or mutual aid plans during the rapid evolution of extreme events. Compared to existing technologies that lack structured fulfillment feedback and still operate based on ideal assumptions, this invention forms a continuous link of risk characterization—upper boundary protection—grid structure verification—structured decision-making—receipt closed-loop tightening, producing a comprehensive effect of predictive protection, physical fulfillment, and controllable execution, thereby improving the reliability and controllability of supply during extreme weather events. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0017] Figure 1 A schematic diagram of the structure of a power intelligent dispatching system for extreme weather events provided in this application embodiment;

[0018] Figure 2 This application provides a schematic diagram illustrating the annual variation of peak daily electricity load in a typical zone over many years, as part of an embodiment of the present application.

[0019] Figure 3 A schematic diagram of a power intelligent dispatching method for extreme weather events provided in an embodiment of this application;

[0020] Figure 4 A flowchart for risk characterization and structured scheduling decision generation provided in the embodiments of this application. Detailed Implementation

[0021] Embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of the present disclosure are shown in the drawings, it should be understood that embodiments of the present disclosure may be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure.

[0022] It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure. In the description of the embodiments of this disclosure, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "this embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects.

[0023] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0024] Specific Implementation Example 1: This example focuses on provincial power grid dispatching for supply assurance during extreme weather events. The provincial dispatching system divides the province into N zones, denoted as D = {1, 2, ..., N}. The system operates on a rolling basis according to a fixed dispatching cycle, with the cycle number denoted as t. The rolling prediction time domain includes the next H cycles.

[0025] like Figure 1 The diagram shown is a schematic of the structure of a smart power dispatching system for extreme weather events provided in this application embodiment. The system includes: a risk characterization module, an upper bound verification module, a decision feedback module, and a dispatching database.

[0026] The risk characterization module is connected to the upper bound verification module, which in turn is connected to the decision feedback module. The risk characterization module, the upper bound verification module, and the decision feedback module are all connected to the scheduling database. The scheduling database is used to store various parameters involved in the power intelligent scheduling system for extreme weather events.

[0027] The risk characterization module is used to access and preprocess multi-source extreme climate data, and to form risk characterizations for each region based on the multi-source data in response to supply constraints for extreme events.

[0028] In this embodiment, the extreme climate multi-source data is a data set related to the identification of extreme climate events and the scheduling of supply. It is accessed through existing business interfaces and managed according to a unified standard. The extreme climate multi-source data includes weather forecast data, weather data, and partition boundary data for partition spatial aggregation. Among them, the weather forecast data and weather data include time-series observations or rasterized products of temperature, rainfall, and wind speed within the province. The partition boundary data is used to determine the coverage of each partition to support the partition spatial aggregation of representative meteorological values.

[0029] Preprocessing is the process of converting the accessed data into standardized inputs that can be used for risk characterization calculations. Preprocessing includes: aligning timestamps of different data sources and mapping them to scheduling cycles; uniformly converting the units and dimensions of different data sources; spatially matching meteorological data by region according to the region boundaries and generating aggregateable raster indexes and area weights; regularizing missing, duplicate, out-of-bounds, and anomalous jump data to form continuous time series; and writing the preprocessing results into the data storage area for subsequent use in generating regional meteorological representative values, extreme intensity indicators, short-term variation amplitude, and spatial synchronicity indicators.

[0030] A zonal threshold system T is established, which is a preset threshold configuration set. It is used to standardize the impact of different zonal regions under different operating conditions to obtain classification indicators, and at the same time, it performs classification judgment and outputs the level results: regional meteorological representative value, extreme intensity index, short-term change amplitude, and spatial synchronicity index.

[0031] The generation method of T is as follows: in the offline stage, historical meteorological data and historical load operation data of the province are collected, historical supply pressure windows are selected as calibration samples, and an initial threshold range is given in combination with meteorological classification standards. The threshold is calibrated and solidified at the zonal scale. T is updated on a seasonal basis and remains fixed and effective when running online.

[0032] Among them, historical meteorological data are actual meteorological data or meteorological forecast data for historical periods in the province. The actual meteorological data are the observed values ​​of temperature, rainfall, wind speed and relative humidity recorded at a fixed time granularity. The meteorological forecast data are meteorological forecast products for the corresponding historical periods and include the forecast release time, forecast lead time and forecast value. In the offline stage, the historical meteorological data are converted into the meteorological representative value time series of each region according to the regional spatial aggregation rules, which are used to characterize the extreme intensity level and evolution process of each region under different operating conditions. Historical load operation data consists of operation records generated and stored by the provincial dispatching network system during historical periods. These records include measured load sequences for each region, measured renewable energy output sequences for each region, deviation records between dispatch forecasts and actual measurements, records of limitations and alarms for key constraints, and parameters and execution receipts for dispatch measures. After offline processing, historical load operation data is used to generate characteristic quantities such as regional net load, supply-demand gap, degree of approach to key constraints, and implementation status of measures. These quantities are used to identify historical supply pressure windows and calibrate the initial threshold range, enabling the threshold system to effectively distinguish between supply pressure windows and non-pressure windows at the regional scale. This provides an implementable and traceable threshold basis for determining extreme intensity levels, short-term change levels, and spatial synchronization levels.

[0033] To facilitate understanding of the offline calibration process of the zoning threshold system T, this embodiment provides a traceable calibration example: In the offline stage, multi-year historical hourly meteorological-load data of a typical zone are selected as samples to form a load peak sequence and a multi-meteorological element sequence. The sample data are then cleaned and time-granularity standardized. Annual statistical analysis of the load peak sequence reveals a clear seasonal high-level window, such as... Figure 2 The diagram shows the annual variation of peak daily electricity load in a typical zone over many years. The peak load exhibits segmented characteristics of a high-level window in winter and a second-highest window in summer. Based on this, the threshold configuration is fixed by season at the zone scale to avoid misjudgment and threshold drift caused by using a uniform threshold throughout the year.

[0034] To ensure that subsequent calculations can be implemented, T in this embodiment includes at least the following parameters (all obtained from offline calibration and embedded in the configuration): a m,d b m,d The standardized two-end anchor points of meteorological element m in partition d (used to map representative values ​​to the 0-1 interval), a m,d Corresponding to the standardized lower bound benchmark, b m,d Corresponding to the standardized upper bound benchmark; σ m,d This parameter represents the risk direction, taking a value of +1 or -1, and is used to characterize whether the increase or decrease of this element corresponds to an enhanced risk; w m,d The element fusion weights represent the weights that satisfy the condition for partition d. Extreme intensity classification threshold, short-term change classification threshold, spatial synchronization classification threshold, and preset threshold θ for synchronization statistics. sy,d θ sy,d This represents the spatial synchronization statistics threshold for partition d. sy is simply a marker for the purpose of synchronization statistics, indicating the threshold used for spatial synchronization statistics to determine whether the partition should be included in the high-risk partition count.

[0035] By transforming extreme event risks from descriptive to calculable, gradable, and table-based discrete inputs, a unified basis is provided for subsequent conservative upper bound adjustments and safety boundary setting, thereby reducing upper-tail underestimation and strategy inconsistency caused by subjective judgment from the source.

[0036] Real-time access to weather forecast or real-time raster data. The weather forecast or real-time raster data is regular grid data divided according to a preset spatial resolution. Each raster cell in the grid corresponds to a unique raster cell identifier g and has a corresponding meteorological element value z at each time t. m,g,t The boundary vectors of each partition are pre-obtained and overlaid on the regular grid in the execution space to determine the set of raster cell identifiers Gd that intersect with partition d, and the intersection area A between partition d and raster cell g is calculated. d,g As an area weight; during online operation, meteorological element values ​​of the zone coverage raster are read based on Gd and spatially aggregated according to the area weight to obtain the zone meteorological representative value x. m,d,t : .

[0037] To unify meteorological or operational elements with different dimensions into comparable risk profiles, a common processing approach of element standardization, weighted fusion, and threshold grading is typically adopted. Specifically, for each element m, based on a in T... m,d b m,d σ m,d x m,d,t Standardized to s m,d,t : ;where clip(·) is the truncation operator, restricting the result to [0, 1]; the standardized scores of each element are weighted according to w m,d By merging, the extreme intensity index I of the partition is obtained. d,t : Subsequently, based on the extreme intensity grading threshold of the partitioned threshold system, I was... d,t Grading, output extreme intensity level L I,d,t .

[0038] Define a differential window τ (determined and fixed by the scheduling cycle and business requirements, for example, two cycles), and calculate the short-term variation amplitude ΔI. d,t : Based on the short-term change graded threshold of the partitioned threshold system, ΔI d,tGrading, outputting short-term change level L Δ,d,t .

[0039] Based on the preset threshold θ used for synchronous statistics in T sy,d Construct a set of high-risk partitions: The statistical proportion generates the spatial synchronicity index S. t : , Indicates the number of high-risk partitions. This represents the total number of partitions, and S is determined based on the spatial synchronization hierarchical threshold of T. t Hierarchical, output space synchronization level L S,t , which represents the level of spatial synchronization of high-risk zones across the province at the current time t.

[0040] The three levels are used to form the risk characterization for each region: R d,t =(L I,d,t L Δ,d,t L S,t ), where R d,t This serves as the direct input for subsequent upper bound adjustments and decision generation.

[0041] Taking a province experiencing a sustained high-temperature period at 14:00 on July 15th as an example, the province is divided into three zones: d1, d2, and d3. Meteorological raster data, after area-weighted aggregation, yields representative temperature values ​​of 38.5℃, 39.2℃, and 37.8℃, respectively. The zone threshold system stipulates that 30℃ corresponds to an intensity score of 0, and 40℃ corresponds to an intensity score of 1. Based on this, extreme intensity indices are calculated to be 0.85, 0.92, and 0.78, respectively. The intensity scores are then further classified into intervals according to the threshold system: when 0 ≤ I... d,t Level 1 is defined as <0.60, and Level 2 is defined as 0.60 ≤ I. d,t Level 2 is defined as <0.80, and Level 3 is defined as 0.80 ≤ I. d,t A value ≤1.00 corresponds to Level 3. Based on this, d1 and d2 are classified as Level 3, and d3 as Level 2, resulting in output intensity levels of 3, 3, and 2. Compared to 13:00, the intensity indices of the three zones increased from 0.72, 0.80, and 0.63 to 0.85, 0.92, and 0.78, respectively, with changes of 0.13, 0.12, and 0.15. According to the threshold system, all are classified as Level 3. The threshold system stipulates that a synchronous statistical threshold of at least 0.80 is considered a high-risk zone. Therefore, d1 and d2 are included, and the proportion of high-risk zones is 2 / 3 = 0.67, resulting in a synchronous level of 2 according to the threshold system. Thus, the risk characterization for each zone is as follows: d1 is (Intensity 3, Change 3, Synchronous 2), d2 is (Intensity 3, Change 3, Synchronous 2), and d3 is (Intensity 2, Change 3, Synchronous 2).

[0042] The upper bound verification module is used to correct the net load forecast of each zone based on the risk characterization of each zone to obtain the conservative upper bound of the net load of each zone, and to perform safety constraint verification based on the conservative upper bound of the net load of each zone to obtain the key constraint status indicators of the power grid.

[0043] Obtain load forecast L for each zone d,t With the forecast of new energy output in each region (G) d,t Both are output from the provincial dispatch network forecasting system, used to calculate the net load forecast (NL) for each region. d,t : Based on the three risk levels in the risk characterization and the preset coefficient mapping relationship (fixed as a lookup table configuration after offline calibration), the first correction coefficient is determined. Second correction factor Third correction factor Where f1, f2, and f3 are preset mapping relationships, which are derived from the offline statistical distribution of net load prediction errors under extreme events. Conservative coefficients that can cover the upper tail deviation are selected for tiered configuration, so that the corrected upper bound can cover the actual high net load.

[0044] Sort α1, α2, and α3 in descending order, and use the correction coefficient that ranks first in the sort as the load correction coefficient. By correcting the net load forecast using a load correction factor, a conservative upper bound NLu for the net load in zone d and period t is obtained. d,t : When load surges and renewable energy weaken in the same direction, conventional separate forecasts followed by subtraction can easily underestimate the upper limit of net load. This step uses risk grading to drive conservative adjustments and adds the maximum value for each item according to the most unfavorable dimension, making it easier for the upper bound to cover the actual high net load during extreme periods, thereby reducing the risk of insufficient reserves and false triggering delays.

[0045] The power grid model parameters, including network topology, node set, and line and transformer parameters, are provided by the provincial dispatch network model library; the operating mode parameters, including maintenance status, tie line status, and operating mode switching information, are provided by the operating mode system; and the key constraint limit parameter L is obtained from the fixed operating procedures and maintained according to the mode change. k The limit for each critical constraint object k is any object in the pre-configured set of critical constrained objects of the power grid, which is used to characterize the physical constraint carrier that may be restricted in advance during the supply guarantee process of extreme events and affect the cross-regional support and power grid security boundary. The critical constraint object k includes: critical transmission lines, critical transformers, inter-regional tie lines, and inter-regional critical transmission interfaces composed of multiple transmission lines.

[0046] The upper bound of the net load for each zone is conservatively set to NLu. d,tThis is mapped to the corresponding zonal injection boundary conditions of the power grid model. The allocation coefficient Mn,d from zonal to node is used, where Mn,d represents the proportion of net load in zonal d distributed across node n. Historical load measurements are used to statistically analyze the load proportion of each node within zonal d over different time periods, and a stable caliber (e.g., by season / working day type) is adopted to obtain Mn,d. n,d Offline configuration is hardened to meet the requirements. The equivalent net injection change (or equivalent load injection) P of node n in time period t is obtained. n,t : Based on node injection volume P n,t Perform power flow calculations to obtain the constraint quantity F for each critical constraint object k. k,t ; F k,t With limit L k Compare and calculate the utilization and margin of key constraint objects: , Based on this, the constraint status of key constraint objects is determined, and the key constraint status index C of the power grid is output. k,t In this embodiment, C k,t The output is structured and consists of the following fields: constraint object identifier k, utilization rate u. k,t , margin k,t By projecting the worst-case reasonable net load upper limit onto the physical constraints of the grid structure, a quantifiable key constraint state can be obtained, thereby avoiding the potential problem that the load gap can be contained within the upper limit, but the grid structure channel exceeds the limit first, making the solution unworkable.

[0047] In this embodiment, power flow calculation refers to calculating the power flow based on the node injection quantity P, given the power grid model parameters and operating mode parameters. n,t As boundary conditions, the calculation process for voltage at each node and power flow in each branch of the power grid is solved, thereby obtaining the power flow results for lines, transformers, tie lines, and key transmission interfaces between sections during period t. Power flow calculation can be performed using either AC power flow algorithms or DC power flow algorithms.

[0048] Taking a province experiencing a sustained high-temperature period at 14:00 on July 15th as an example, the provincial power grid provides the following forecasts for 17:00–18:00: d1 load 7800MW, renewable energy 1200MW; d2 load 9200MW, renewable energy 900MW; d3 load 6500MW, renewable energy 1400MW. The net load forecasts are 6600MW, 8300MW, and 5100MW respectively. The coefficient mapping table is configured as follows: Intensity 3 → 0.10, Intensity 2 → 0.06; Change 3 → 0.08; Synchronous 2 → 0.06. Using the method of maximizing each term: d1 is set to 0.10, d2 to 0.10, and d3 to 0.08. The conservative upper bounds for the net load are obtained: d1 is 7260MW, d2 is 9130MW, and d3 is 5508MW. The conservative upper bound of net load for each zone from 17:00 to 18:00 is mapped to the zone-injected boundary conditions of the power grid model. Power flow calculations are performed to obtain the constraint quantities of key constraint objects, and these are compared with the limits to output key constraint status indicators. Example: Key channel k1 has a limit of 5000MW, with a calculated constraint quantity of 5200MW (over-limit); key channel k2 has a limit of 4000MW, with a calculated constraint quantity of 3850MW (approaching limit). Therefore, the key constraint status indicators for "k1 over-limit risk" and "k2 high approach risk" and their corresponding time periods are output.

[0049] The decision feedback module is used to generate a structured scheduling decision set based on the risk characterization of each zone, the conservative upper limit of net load, and the key constraint status indicators of the power grid.

[0050] Based on the risk characterization of each partition R d,t Generate scheduling control strength parameter K t In this embodiment, K t The risk level is mapped through lookup table rules and used to control the reserve level of planned items, the tightening of quota items, and the sensitivity of trigger items.

[0051] The supply-demand gap constraint parameter G is formed based on the conservative upper bound of net load. t The system calculates the total upper limit of net load for the entire province or key areas of concern based on a conservative upper limit, and reads the total available capacity of adjustable resources within the same period. The available capacity of adjustable resources is formed by the resource ledger and the availability status on the execution side, reflecting the available charging and discharging capacity of energy storage, available reserve capacity, available demand response capacity, and verifiable cross-regional support capacity. The system compares the total upper limit of net load with the total available capacity of adjustable resources: if the total upper limit of net load is higher than the total available capacity of adjustable resources, it determines that there is a supply-demand gap pressure, and converts the difference and its proportion into a supply-demand gap constraint parameter G. t If the upper limit of net load is not higher than the total available capacity of adjustable resources, then G will be... t Set to a low notch pressure level. G tIt is used to characterize the degree of supply and demand imbalance during critical periods, and is subsequently used in decision generation to determine the conservatism of backup structure, energy storage reservation, demand response pre-allocation, and inter-regional exchange plans.

[0052] Power grid operation boundary constraint parameters B are formed based on key power grid constraint state indicators. t In this embodiment, all key constraints are summarized and statistically analyzed, and the most stressful constraint is extracted as the power grid operation boundary representation for that period. Specifically, the key constraint with the smallest margin or the highest utilization rate is identified, and its margin and utilization rate are used as the representative of the overall grid boundary stress level. When multiple key constraints are approaching their limits simultaneously, the boundary stress level is determined according to the most unfavorable one. The system converts the boundary stress level into the power grid operation boundary constraint parameter B. t This causes B to... t Increase; when the margin is sufficient or the utilization rate is low, B t Decrease. B t It is used to constrain cross-regional exchange boundaries and quota strategies in subsequent decision generation.

[0053] K t G t B t Input the preset decision generation rules to generate a structured scheduling decision set A, and fix the output into three categories: planning decisions output parameters for resource reservation and backup arrangements during critical periods; quota decisions output parameters for the operation boundary of key constraint objects; and trigger decisions output trigger conditions, trigger actions, effective periods and release condition identifiers.

[0054] To ensure the executability and consistency of dispatch decisions, the pre-defined decision generation rules are not generated ad hoc, but rather are a set of rules pre-built and solidified by the dispatching side. These rules originate from: extracting templates of typical dispatch measures applicable to different risk levels and operational scenarios based on current power dispatching operation procedures, safety and stability control requirements, and historical operating experience; mapping different level combinations to corresponding dispatching strategy categories and parameter value ranges by combining extreme intensity levels, short-term change levels, and spatial synchronization levels defined in the threshold system T; pre-configuring the feasible parameter ranges, applicable objects, and time constraints of various dispatching measures based on grid structure constraints, critical equipment operating limits, and historical executability; and solidifying these rules in a structured form as a decision generation rule library, which automatically matches and outputs corresponding structured dispatch decisions based on real-time input risk characterization results during operation. Through this method, the pre-defined decision generation rules are configured during the system deployment phase and invoked as deterministic rules during operation, thus avoiding the uncertainty caused by ad-hoc manual judgment.

[0055] Decision set A is output as a structured field, which includes: decision type, object identifier, target parameter, applicable time period, execution constraint, trigger condition identifier, and release condition identifier, and retains the basis for generation.

[0056] The parameters for reserve arrangements in planned decisions include reserve type, target reserved capacity, arrival time limit, and duration; the parameters for energy storage reservations in planned decisions include the upper limit of reserved power, the lower limit of available electricity during critical periods, and the pre-critical period status requirements; the parameters for operational boundaries in quota-based decisions include the identifier of key constraints, operational boundary values, and effective periods; the triggering conditions for triggering decisions include risk level conditions, supply-demand gap pressure conditions, and margin conditions for key constraints, and the triggering actions include increasing reserves, calling up energy storage, initiating demand response, tightening operational boundaries, and tightening inter-regional exchange boundaries, with corresponding effective periods and release conditions set.

[0057] Under the condition that the structured scheduling decision set is issued and executed, a closed-loop correction is performed on the structured scheduling decision set. Specifically, in each rolling cycle, execution receipts corresponding one-to-one with the structured scheduling decision set are obtained from the scheduling execution side. The execution receipts are execution result information returned by the execution system. The execution result information includes the object identifier of the corresponding scheduling measure, the applicable time period, the execution confirmation information, and the execution measurement information, which are used to characterize the actual completion and arrival status of each scheduling measure in the structured scheduling decision set within the corresponding time period.

[0058] The execution receipt provides corresponding execution result fields for different types of scheduling measures: When the scheduling measure is a standby arrangement, the execution receipt provides whether the standby resources are put into operation as required, the capacity level put into operation, and whether the arrival time meets the requirements; when the scheduling measure is an energy storage charging and discharging arrangement, the execution receipt provides the actual charging and discharging output level of the energy storage during the applicable period and whether it is completed as required by the scheduling; when the scheduling measure is a demand response arrangement, the execution receipt provides the actual response capacity, the response start time, and the duration; when the scheduling measure is an adjustment to the inter-regional mutual assistance or exchange plan, the execution receipt provides whether the plan has been confirmed to be executed, the actual power realization status on the communication channel, and the degree of deviation; when the scheduling measure is an adjustment to the limit or operating boundary, the execution receipt provides whether the limit has taken effect and whether there are any over-limit or approaching operating phenomena during the effective period.

[0059] Simultaneously, the target parameters corresponding to each scheduling measure in the structured scheduling decision set are read. These target parameters are executable target fields that are fixed and output along with the decision set. They specify the target level and constraints that each scheduling measure should achieve within the corresponding time period. Target parameters include target capacity, target output, target arrival time, target duration, and effective time period information. Using the scheduling measure object identifier and applicable time period as the association key, the actual execution result in the execution receipt is compared with the target parameters in the structured scheduling decision set to determine the execution deviation. The execution deviation characterizes the degree to which the actual execution result fails to meet the target parameters.

[0060] After determining the execution deviation, the system further determines whether to tighten the scheduling safety boundary based on pre-set trigger conditions. These pre-set trigger conditions include deviation threshold conditions and persistence conditions: the deviation threshold condition determines that the implementation level of a key measure is below the allowable range, and the persistence condition determines that the deviation has not improved within a continuous rolling cycle. When the pre-set trigger conditions are met, adaptive tightening is performed on the scheduling safety boundary. This adaptive tightening includes increasing the reserve target level, raising energy storage reservation requirements, increasing the pre-arranged quota for demand response, tightening the boundaries for inter-regional mutual assistance and exchange, and further tightening the operational boundaries of key constraints. This ensures that even when key resource responses are insufficient or not fully realized, the system can still guarantee grid security and supply reliability with a more conservative boundary.

[0061] The degree of implementation is used to characterize the degree of conformity between the actual execution results of the key measures and the target parameters. Specifically, it is obtained by comparing the actual execution volume and implementation status provided by the execution receipt with the target parameters: for measures with target values, the degree of implementation reflects the proportion of the actual execution volume to the target value; for measures with status requirements, the degree of implementation reflects whether the actual status has reached the target status or is calculated based on a partial degree of implementation.

[0062] The aforementioned persistence condition is used to determine whether the insufficient arrival or execution deviation is persistent. That is, if the arrival level is continuously lower than the deviation threshold or the execution deviation is continuously exceeded within the allowable range for x consecutive rolling cycles, it is considered that the deviation has not been improved within the consecutive rolling cycles. The value of X is determined by those skilled in the art.

[0063] After adaptive tightening is completed, an updated structured scheduling decision set is generated and issued for execution. The updated structured scheduling decision set includes at least the adjusted decision items, the adjusted target parameters, the effective period, and the adjustment reason identifier, thus forming a closed-loop correction process of issuance-receipt-deviation judgment-boundary tightening-policy update. This ensures that the scheduling strategy under extreme weather events can be dynamically corrected as the execution and fulfillment capabilities change, and maintains controllability and traceability.

[0064] Taking a province experiencing a sustained high-temperature period at 14:00 on July 15th as an example, based on risk characteristics (high intensity, rapid enhancement, medium-high synchronicity), conservative upper limit of net load (high level during critical periods), and key constraint states (k1 exceeding limits, k2 approaching), a structured scheduling decision set is generated according to preset rules, and three fixed outputs are generated: Planned: 1000MW of rapid standby from 17:00 to 18:00, 600MW of planned energy storage discharge, and 300MW of pre-arranged demand response; Quota-based: the k1 operating boundary is tightened to 4800MW, and the upper limit of the inter-regional exchange boundary is 300MW; Trigger-based: if the k1 utilization rate is not lower than 0.95 for two consecutive cycles, then an additional 200MW of rapid standby is added and the k1 boundary is further tightened. The release condition is that the utilization rate continuously falls below 0.90 for two consecutive cycles. Execution receipts are the actual execution volume and status reported back from the execution side. The 17:00 feedback report showed that the actual energy storage target of 600MW was 420MW (achievement rate 0.70%), and the actual demand response target of 300MW was 240MW (achievement rate 0.80%). Both were below the preset achievement rate threshold of 0.90 and had not improved for two consecutive cycles, thus the execution deviation was determined to be valid. The system triggered adaptive tightening of the safety boundary: 300MW of reserve was added, the k1 boundary was further tightened from 4800MW to 4700MW, the inter-regional exchange boundary was lowered from 300MW to 200MW, and a new version of the structured dispatch decision set and the logbook were updated and output.

[0065] like Figure 4 As shown in the flowchart of risk characterization and structured scheduling decision generation provided in this application embodiment, the process first involves accessing and preprocessing multi-source extreme climate data. After this step, extreme intensity levels, short-term variation levels, and spatial synchronization levels are generated, and these levels are combined to obtain the risk characterization for each region. After obtaining the risk characterization for each region, the net load forecast for each region is corrected based on the load correction coefficient, thereby obtaining the conservative upper bound of the net load for each region. Subsequently, a safety constraint check is performed on the conservative upper bound of the net load for each region to obtain the key constraint status indicators of the power grid. After obtaining the key constraint status indicators of the power grid, a structured scheduling decision set is generated, which includes planned decisions, quota decisions, and trigger decisions.

[0066] In Specific Implementation Example 2, based on Specific Implementation Example 1, a partition stage identification process is introduced after the risk characterization module outputs the risk characterization of each partition and before the upper bound verification and decision generation. This process is used to divide the evolution process of the same partition in extreme events into continuous stages, and uses the stage identifier as an additional input to the decision generation rules, enabling the scheduling strategy to switch smoothly by stage and reduce frequent jitter near the threshold.

[0067] The zoning stage identification uses the three types of quantities already obtained in Specific Implementation 1 as inputs. The extreme intensity index of the zone is obtained by standardizing and fusing the meteorological representative values ​​through the threshold system. The short-term change amplitude is obtained by calculating the difference between the extreme intensity index in adjacent periods. The duration of continuous over-threshold is obtained by continuously counting whether the extreme intensity index is not lower than the preset threshold in the threshold system. The counting threshold for the duration of continuous over-threshold is directly adopted from the preset threshold used for synchronous statistics in the zoning threshold system.

[0068] Maintain a continuous threshold counter for each partition. In each scheduling cycle, first determine whether the extreme intensity index of the partition is not lower than the preset threshold of the corresponding partition in the threshold system. If it is not lower, the counter is incremented by one based on the previous cycle; if it is lower, the counter is cleared. The initial value of the counter is zero, and it is updated on a rolling basis with the scheduling cycle.

[0069] The phase division rules are provided by a partitioned threshold system. The rule parameters are calibrated offline and fixed into the configuration. These include at least the extreme intensity phase thresholds used to distinguish between low, medium, and high intensity segments; the threshold for identifying the magnitude of change between rapid enhancement and the start of decline; the duration threshold for distinguishing between short-term surges and sustained high levels; and the hysteresis bandwidth parameter used to separate the entry and exit thresholds, preventing phases from jumping back and forth around the thresholds. All of these parameters are statistically derived from the correspondence between intensity level, enhancement rate, duration, and supply pressure in historical extreme event samples and are updated seasonally.

[0070] Among them, the low intensity segment, the medium intensity segment, and the high intensity segment refer to the three intensity ranges specified by the threshold system when the extreme intensity index of the zone falls below the low threshold: when the extreme intensity index is below the low threshold, it is determined to be the low intensity segment; when the extreme intensity index is between the low threshold and the high threshold, it is determined to be the medium intensity segment; when the extreme intensity index reaches or exceeds the high threshold, it is determined to be the high intensity segment.

[0071] Each period outputs a stage identifier for each partition. The stage identifier contains four stages, which are used to characterize the risk evolution in chronological order. When the intensity begins to enter the rising range and the change is positive, it is the warm-up stage. When the intensity reaches the median or the continuous over-threshold count reaches the preset duration, it is the arrival stage. When the intensity is at a high level or the change is characterized by rapid enhancement, it is the peak stage. When the intensity falls from a high level and the change is negative and the exit condition is continuously met, it is the decay stage.

[0072] Phase markers do not replace the original risk characterization, but serve as additional inputs to decision generation rules. In the warm-up phase, the advance nature of trigger-based decisions is improved, and priority is given to preparing quick reserves and energy storage reserves. In the arrival phase, the reserve intensity of planned decisions is increased, and the protection boundaries of quota-based decisions are solidified in advance. In the peak phase, priority is given to ensuring the safety margin of key constraints, and trigger-based actions tend to tighten boundaries and increase reserves. In the decay phase, tightening is gradually lifted according to hysteresis and persistence judgments to avoid strategy jitter.

[0073] By identifying stages, the switching of scheduling strategies changes from being triggered by a single threshold to a smooth switching based on stage evolution. This reduces the frequent tightening and loosening caused by the fluctuation of intensity indicators around the threshold under extreme events, thereby improving the stability of scheduling output and the realizability of execution, and enhancing the ability to interpret whether the risk is rapidly increasing or remains at a high level.

[0074] Taking a province experiencing a sustained high-temperature period at 14:00 on July 15th as an example, at 13:00, the intensity index d1 was 0.72, d2 was 0.80, and d3 was 0.63; at 14:00, d1 was 0.85, d2 was 0.92, and d3 was 0.78. According to the counting rule of whether it is ≥0.80: d1 did not reach the threshold at 13:00 but reached the threshold at 14:00, so the sustained over-threshold count at 4:00 was 1; d2 had already reached the threshold at 13:00 and continued to reach the threshold at 14:00, so the sustained over-threshold count at 14:00 was 2; d3 did not reach the threshold at either 13:00 or 14:00, so the sustained over-threshold count at 14:00 was 0. The criteria for determining the solidification stage in the threshold system are as follows: In the preheating stage, the intensity level is no lower than 2 and the change level is 3, with a continuous over-threshold count of less than 2; in the arrival stage, the intensity level is 3 and the continuous over-threshold count is no lower than 2; in the peak stage, the intensity level is 3 and the change level is 3, with a continuous over-threshold count of no lower than 3; in the decay stage, the intensity level decreases and the change level is 0 or 1, continuously meeting the exit conditions. Based on this, the 14:00 stage identification output is: d1 is the preheating stage (intensity 3, change 3 but count is 1), d2 is the arrival stage (intensity 3 and count is 2), and d3 is the preheating stage (intensity 2, change 3, count is 0). The spatial synchronization level remains unchanged at 2, used to characterize the simultaneous impact on multiple zones across the province.

[0075] Specific Implementation Example 3 enhances the process of determining the load correction coefficient based on Specific Implementation Example 1. This allows different zones to adopt different correction intensities under the same risk level, thereby reducing overly conservative or missed judgments caused by a one-size-fits-all approach. Implementation Example 3 does not introduce a complex online learning model; instead, it achieves this through a mapping relationship between a zone type table and differential coefficients, making maintenance costs manageable.

[0076] An offline partition type table is constructed, which assigns a type identifier to each partition. The type identifier is determined by two types of structural features: load sensitivity features and new energy structure features.

[0077] Load sensitivity features are used to characterize the sensitivity of a region's load to meteorological changes. Historical representative temperature values ​​and historical measured load data are used to statistically analyze the load's sensitivity to temperature changes seasonally, and this data is then solidified into regional load sensitivity indicators. The platform utilizes historical representative temperature values ​​and historical measured load data to perform time-granularity alignment and anomaly cleaning on both types of data offline, and groups the samples by season. Within each season, the typical correspondence between temperature changes and load changes is statistically analyzed to obtain the "average load change caused by a unit temperature change" or an equivalent normalized response strength. The statistical results for each season are solidified into regional load sensitivity indicators and written into a configuration table for risk condition correction of load forecasts and expanded assessment of net load uncertainty during the operational phase. Historical representative temperature values ​​are obtained from the regional spatial aggregation of the risk characterization module, and historical measured loads are obtained from the scheduling measurement system.

[0078] In Specific Implementation One, all three types of correction coefficients are determined by the risk level through a preset coefficient mapping relationship. This Implementation Three extends the mapping relationship to one differentiated by partition type; for the same extreme intensity level, the first correction coefficient values ​​differ for different partition types; for the same short-term change level, the second correction coefficient values ​​differ for different partition types; and for the same spatial synchronization level, the third correction coefficient values ​​differ for different partition types. The differentiated mapping relationship is obtained by offline statistical analysis of the upper tail of the net load prediction error for the same partition type at each risk level, and then fixed after selecting conservative coefficients that can cover the upper tail deviation and configuring them in tiers.

[0079] When running online, the system first reads the type identifier of each partition according to the partition type table, and then, in combination with the extreme intensity level, short-term change level and spatial synchronization level in the risk characterization, extracts three types of correction coefficients from the three coefficient mapping tables of the corresponding types.

[0080] The same synthesis rule as in Specific Implementation Example 1 is still adopted: the three types of correction coefficients are compared, and the largest one is taken as the load correction coefficient. The net load forecast is corrected using this load correction coefficient to obtain the conservative upper limit of the net load.

[0081] Differential adjustments make the conservative upper bound of net load more closely match the characteristics of the zoning structure: load-sensitive zoning is adjusted more fully under the same risk level, and new energy fluctuation zoning is more cautious about the upper tail under the same risk level. This allows for earlier exposure of key constraints that may actually approach the limit during safety constraint verification, reducing missed judgments. In the generation of structured scheduling decisions, the decision intensity of planned and triggered types is more consistent with the zoning differences, reducing unnecessary over-conservatism. In the execution receipt closed loop, if a certain type of zoning continues to show execution deviations, targeted calibration can be performed on that type of zoning when updating the mapping relationship offline, without affecting the parameter configuration of other types of zoning, facilitating continuous iteration.

[0082] Taking a province experiencing a sustained high-temperature period at 14:00 on July 15th as an example, the offline construction of the zoning type table is based on historical statistics: d2 belongs to the air conditioning sensitive zone where load surges are more pronounced during high temperatures; d1 belongs to the comprehensive sensitive zone (somewhat temperature sensitive but weaker than d2); d3 belongs to the zone with a high proportion of renewable energy (net load is more significantly affected by renewable energy fluctuations). In the coefficient mapping table, different tiers are used for different types of zoning. To reflect differentiation while remaining practical, this example provides differentiated coefficient values ​​for the same level (these values ​​are obtained and fixed through offline calibration based on historical error): 0.12 corresponds to intensity level 3 for the air conditioning sensitive zone (d2) (more conservative than the general 0.10); 0.10 still corresponds to intensity level 3 for the comprehensive sensitive zone (d1); and 0.09 corresponds to change level 3 for the zone with a high proportion of renewable energy (d3) (more conservative than the general 0.08). The method of maximizing each item is still used. d1 is revised from 6600MW to 7260MW (unchanged); d2 is revised from 8300MW to 9296MW (higher than 9130MW in Example 1, reflecting a stronger upward pressure due to air conditioning sensitivity); d3 is revised from 5100MW to 5559MW (slightly higher than 5508MW in Example 1, reflecting a stronger uncertainty in areas with a high proportion of new energy).

[0083] Specific embodiment four, such as Figure 3 The diagram shown is a schematic flowchart of a power intelligent dispatching method for extreme weather events provided in an embodiment of this application, including:

[0084] Step 1: Access and preprocess multi-source extreme climate data to form risk characterizations for each region based on the multi-source data and supply guarantee constraints for extreme events; Step 2: Correct the net load forecast for each region according to the risk characterizations to obtain the conservative upper bound of the net load for each region, and perform safety constraint verification based on the conservative upper bound of the net load for each region to obtain the key constraint status indicators of the power grid; Step 3: Generate a structured dispatch decision set based on the risk characterizations for each region, the conservative upper bound of the net load, and the key constraint status indicators of the power grid.

[0085] Through the above description of the implementation methods, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the above functions can be divided into different functional modules to complete all or part of the functions described above.

[0086] In the embodiments provided in this application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, or indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.

[0087] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units, located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0088] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0089] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, or the parts that contribute to the solution, or all or part of the technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0090] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A smart power dispatching system for extreme weather events, characterized in that, include: The risk characterization module is used to access and preprocess multi-source extreme climate data, and to form risk characterizations for each region based on the multi-source data in response to supply constraints for extreme events. The specific process for forming risk representations for each region based on multi-source data to address supply constraints during extreme events is as follows: The zoning threshold system is a pre-configured set of thresholds used to standardize and classify the meteorological representative values, extreme intensity indicators, short-term variation amplitudes, and spatial synchronicity indicators of different zones under different operating conditions, and output the corresponding level results. Multi-source extreme climate data are spatially aggregated by region to obtain representative meteorological values ​​for each region, and extreme intensity indices for each region are generated based on the representative meteorological values ​​for each region. Based on the aforementioned partition threshold system, the extreme intensity index of each partition is classified and determined to output the extreme intensity level of each partition; The short-term change amplitude is obtained by calculating the difference between adjacent periods based on the extreme intensity index, and the short-term change amplitude is classified and judged by the partition threshold system to output the short-term change level. Based on the preset threshold in the partition threshold system, the number of partitions with extreme intensity indicators not lower than the preset threshold is counted, and the ratio with the total number of partitions is calculated to generate a spatial synchronization index. The spatial synchronization index is then graded and judged through the partition threshold system to output the spatial synchronization level. The extreme intensity level, the short-term change level, and the spatial synchronization level constitute the risk characterization of each zone; The upper bound verification module is used to correct the net load forecast of each zone based on the risk characterization of each zone to obtain the conservative upper bound of the net load of each zone, and to perform safety constraint verification based on the conservative upper bound of the net load of each zone to obtain the key constraint state indicators of the power grid. The process of correcting the net load forecast for each region based on the risk characterization of each region to obtain a conservative upper bound for the net load of each region is as follows: Obtain the load forecast and the renewable energy output forecast for each zone, and obtain the net load forecast for each zone based on the load forecast and the renewable energy output forecast for each zone; Based on the risk characterization of each zone, the load correction coefficient is determined, and the net load forecast of each zone is corrected by the load correction coefficient, thereby obtaining the conservative upper bound of the net load of each zone. The process for determining the load correction coefficient based on the risk characterization of each zone is as follows: Based on the extreme intensity level in the risk characterization of each zone, the first correction coefficient is determined according to the preset coefficient mapping relationship; Based on the short-term change levels in the risk characterization of each zone, the second correction coefficient is determined according to the preset coefficient mapping relationship; Based on the spatial synchronization level in the risk characterization of each zone, the third correction coefficient is determined according to the preset coefficient mapping relationship; The first, second, and third correction factors are sorted from largest to smallest. Based on the sorting result, the correction factor ranked first is taken as the load correction factor. The process of performing safety constraint verification based on the conservative upper bound of the net load of each zone to obtain the key constraint state indicators of the power grid is as follows: Obtain power grid model parameters, operation mode parameters, and key constraint limit parameters; Map the conservative upper bound of the net load of each zone to the zone injection boundary conditions corresponding to the power grid model; Based on the partition injection boundary conditions, power flow calculations are performed to obtain the constraint quantities of each key constraint object; The constraint quantity is compared with the key constraint limit parameter to determine the constraint status of each key constraint object, and the key constraint status index of the power grid is output. The decision feedback module is used to generate a structured scheduling decision set based on the risk characterization of each zone, the conservative upper bound of net load, and the key constraint status indicators of the power grid. The specific process for generating a structured dispatch decision set based on risk characterization of each region, conservative upper bound of net load, and key power grid constraint state indicators is as follows: The scheduling control strength parameters are determined based on the risk characterization of each partition. The supply-demand gap constraint parameters for key periods are determined based on the conservative upper bound of the net load. The power grid operation boundary constraint parameters are determined based on the aforementioned key power grid constraint status indicators. The scheduling control strength parameters, the supply-demand gap constraint parameters, and the power grid operation boundary constraint parameters are input into a preset decision generation rule to generate the structured scheduling decision set, and the structured scheduling decision set is output as planning decisions, quota decisions, and trigger decisions. Under the condition that the structured scheduling decision set is issued and executed, the structured scheduling decision set is subjected to closed-loop correction, including the following steps: Obtain the execution receipt corresponding to the structured scheduling decision set; The execution receipt is the execution result information returned by the scheduling execution side, which is used to characterize the actual execution volume and status of each scheduling measure in the structured scheduling decision set in the corresponding time period; The execution feedback is compared with the target parameters in the structured scheduling decision set to determine the execution deviation; And when the execution deviation meets the preset triggering conditions, the scheduling safety boundary is adaptively tightened, and the structured scheduling decision set is updated accordingly.

2. The intelligent power dispatching system for extreme weather events as described in claim 1, characterized in that, Based on the risk characterization of each partition, risk stage identification is performed, and the specific steps are as follows: The risk stage of each zone is determined based on the extreme intensity index, the short-term change amplitude, and the persistence characterization quantity. The persistence characteristic quantity is obtained by statistically analyzing the duration for which the extreme intensity index satisfies the preset threshold in the partition threshold system. It outputs stage identifiers according to preset stage determination rules, and stage switching adopts a hysteresis determination mechanism; The stage identifier is used for strategy selection or parameter switching during scheduling decision generation.

3. The intelligent power dispatching system for extreme weather events as described in claim 1, characterized in that, In determining the load correction coefficient based on the risk characterization of each zone, the differentiated response of each zone is analyzed. The specific steps are as follows: Construct a partition type table, and configure differentiated coefficient mapping relationships for different partitions based on the partition type table; The partition type table is used to characterize the partition load sensitivity characteristics and the partition new energy structure characteristics. Based on the partition type and the risk characterization of each partition, the load correction coefficient is determined from the differentiated coefficient mapping relationship to generate a differentiated conservative upper bound for net load.

4. A method applied to a power intelligent dispatching system for extreme weather events as described in any one of claims 1-3, characterized in that, include: Step 1: Access and preprocess multi-source extreme climate data, and form risk characterizations for each region based on the multi-source data to meet supply constraints for extreme events; Step 2: Based on the risk characterization of each zone, the net load forecast of each zone is corrected to obtain the conservative upper limit of the net load of each zone. Based on the conservative upper limit of the net load of each zone, the safety constraint verification is performed to obtain the key constraint state indicators of the power grid. Step 3: Generate a structured dispatch decision set based on the risk characterization of each zone, the conservative upper bound of net load, and the key constraint state indicators of the power grid.

Images