A load bearing capacity self-adaptive scheduling system and method for a high-proportion new energy power grid

Abstract

The application discloses a load capacity adaptive scheduling system and method for a high-proportion new energy power grid, and relates to the technical field of power system scheduling. The system adopts a three-layer structure: a multi-dimensional load capacity state generation layer constructs a rollable update power grid digital twin, models the load capacity as a multi-dimensional load capacity state variable and maps it to a load capacity constraint parameter to construct an operation feasible region; a rolling scheduling optimization decision layer establishes a rolling scheduling optimization model coupled across time scales under new energy output uncertainty to generate a scheduling scheme that satisfies the operation feasible region constraints and consistency constraints; and an online safety checking and feedback layer uses a proxy model based on the digital twin for online safety checking, outputs corrected load capacity constraint parameters to update the operation feasible region and trigger re-solving, and issues execution instructions and writes back operation feedback for continuous updating after the checking is passed. The application can improve the real-time performance of safety boundary description and scheduling executability, and improve new energy consumption capacity.

Description

Technical Field

[0001] This invention relates to the field of power system dispatching technology, and in particular to a capacity-adaptive dispatching system and method for power grids with a high proportion of renewable energy. Background Technology

[0002] With the rapid growth of renewable energy installed capacity, power grid operation is characterized by frequent source-load fluctuations, significant prediction errors, and dynamic changes in operating boundaries. Renewable energy output is significantly affected by meteorological factors, and its randomness and volatility can cause power flow redistribution, changes in reserve demand, and shrinkage of frequency safety margins in a short period. Under conditions of high renewable energy penetration, traditional scheduling methods relying on fixed operating limits and static safety checks are prone to coarse-grained expression and delayed updates of operating boundaries. Existing scheduling systems mostly use fixed thermal stability limits, static equipment capacity, and safety margins calculated offline as constraints. Faced with changes in meteorological conditions, equipment status, and topology switching, operating boundaries often require manual revision or low-frequency offline verification updates, making it difficult for boundary parameters to reflect the true capacity in a timely manner, leading to risks of overestimation or misjudgment. Furthermore, renewable energy fluctuations and load changes exhibit significant temporal coupling characteristics, requiring consistency across different time scales in day-ahead, intraday, and real-time scheduling. However, existing methods often suffer from loose connections between the planning and execution layers and inconsistent boundary constraints across different time scales, easily leading to feasible plans but frequent adjustments or even instability during the execution phase.

[0003] Faced with the uncertainty of new energy output, some methods extend the scheduling model using scenario-based, probabilistic, or robust approaches. However, safety verification often relies on post-hoc or sampling verification, resulting in high computational costs and difficulty in achieving rapid online verification and correction of key constraints. When the operating environment changes rapidly, scheduling schemes may require frequent re-solution, leading to insufficient real-time performance. Existing systems typically use operational feedback data obtained at the execution layer for monitoring, alarms, or post-hoc analysis, making it difficult to establish a continuous calibration mechanism for the boundary parameters of the scheduling model. The long-term accumulation of problems such as prediction bias, model errors, and equipment parameter drift gradually widens the deviation between the scheduling model and the actual power grid, affecting scheduling quality and safety.

[0004] Therefore, under the background of high proportion of new energy power grid operation, there is an urgent need for a dispatching system that can cope with uncertainty, can update the operating boundary according to the operating status, and can be continuously calibrated through online verification and feedback mechanisms, so as to balance safety, economy and feasibility. Summary of the Invention

[0005] The purpose of this invention is to provide a capacity-adaptive scheduling system and method for high-proportion renewable energy power grids. By constructing a digitally updated operation model, expressing the power grid capacity in a multi-dimensional state form and forming a constraint set that can be directly used for optimization, and introducing an online safety verification and feedback calibration mechanism, the invention achieves dynamic characterization of the power grid operation boundary and closed-loop optimization of scheduling.

[0006] The present invention achieves the above objectives through the following technical solutions:

[0007] A capacity-adaptive dispatching system for high-proportion renewable energy power grids, the system adopts a three-layer structure, including:

[0008] A multidimensional carrying capacity state generation layer is used to acquire power grid operation data and prediction data to construct a power grid digital twin that can be updated on a rolling basis. In the power grid digital twin, the power grid carrying capacity is modeled as a multidimensional carrying capacity state variable, the multidimensional carrying capacity state variable is mapped to carrying capacity constraint parameters, and an operational feasible domain is constructed to describe the boundary of safe operation of the power grid. The operational feasible domain represents the allowed operating state space of the power grid in each time period within the rolling scheduling time domain in the form of a constraint set.

[0009] The rolling scheduling optimization decision layer is used to establish a rolling scheduling optimization model coupled across time scales under the condition of uncertainty in new energy output, and to solve the rolling scheduling optimization model to generate a scheduling scheme that satisfies the operational feasible domain constraint and the cross-time scale consistency constraint.

[0010] The online safety verification and feedback layer is used to receive the scheduling scheme, perform online safety verification of the scheduling scheme based on the power grid digital twin using a proxy model coupled with the rolling scheduling optimization model, output the verification result and the corrected carrying capacity constraint parameters, update the operational feasible region based on the corrected carrying capacity constraint parameters, and trigger the rolling scheduling optimization model to resolve within the updated operational feasible region; when the verification is passed, the scheduling scheme is converted into an execution scheduling instruction and issued, and the operational feedback data of the execution scheduling instruction is obtained. The operational feedback data is written into the power grid digital twin to update the multidimensional carrying capacity state variables and the operational feasible region.

[0011] A further improvement of the present invention is that the multidimensional bearing capacity state generation layer includes:

[0012] The data acquisition and preprocessing module is used to acquire power grid operation data and forecast data, and perform at least one preprocessing operation among time synchronization, outlier removal, and missing value completion; the power grid operation data includes SCADA measurement data, PMU measurement data, power grid topology data, equipment parameter data, and protection setting data, and the forecast data includes load forecast data, new energy output forecast data, and meteorological forecast data;

[0013] The power grid digital twin rolling update module is used to perform hybrid state estimation and parameter identification based on preprocessed power grid operation data, and generate power grid digital twin state variables and power grid digital twin parameter calibration variables.

[0014] The multidimensional carrying capacity state calculation module is used to calculate multidimensional carrying capacity state variables based on the state quantities of the power grid digital twin, the parameter calibration quantities of the power grid digital twin, and the prediction data. The multidimensional carrying capacity state variables include the dynamic heat capacity value of the transmission line, the upper limit of the distributed new energy access carrying capacity, the power exchange boundary of the transmission and distribution interface, the frequency security carrying capacity boundary, and the asset health use budget boundary.

[0015] The carrying capacity constraint parameterization module is used to map the multidimensional carrying capacity state variables into carrying capacity constraint parameters, including: mapping the dynamic heat capacity value of the transmission line into a dynamic limit parameter; mapping the upper limit of distributed new energy access carrying capacity into node carrying capacity boundary parameters and regional carrying capacity boundary parameters; mapping the power exchange boundary of the transmission and distribution interface into power exchange range parameters and flexibility ramping capability parameters; mapping the frequency safety carrying capacity boundary into spinning reserve adequacy constraint parameters, frequency response constraint parameters, and inertia constraint parameters; and mapping the asset health use budget boundary into health use limit parameters.

[0016] The feasible domain construction module is used to determine the network state base value and parameter coefficients corresponding to each constraint based on the state variables and parameter calibration variables of the power grid digital twin, and to construct the feasible domain in combination with the carrying capacity constraint parameters, including transmission power flow constraints, node carrying capacity constraints, regional carrying capacity constraints, transmission and distribution interface switching power range constraints, flexibility ramping constraints, spinning reserve adequacy constraints, frequency response constraints, inertia constraints, and health usage limit constraints.

[0017] A further improvement of the present invention is that the multi-dimensional bearing capacity state calculation module, based on meteorological forecast data and equipment parameter data, calculates the predicted average value of the dynamic heat capacity of the transmission line according to the heat balance of the transmission line. and the predicted standard deviation The bearing capacity constraint parameterization module generates dynamic limit parameters according to the uncertainty adaptive reduction rule. The formula is:

[0018] ;

[0019] In the formula, Indicates the transmission line number. Indicates the time period number within the rolling scheduling time domain. , Preset non-negative coefficients;

[0020] Running the feasible domain construction module will dynamically limit parameters Write the upper bound of the power flow constraints in the feasible domain.

[0021] A further improvement of the present invention is that the bearing capacity constraint parameterization module generates the power range parameters for the transmission and distribution interface. , and parameters of flexibility and climbing ability ,satisfy:

[0022] ;

[0023] ;

[0024] In the formula, , These represent time periods within the rolling scheduling time domain. Time period The power switching setting value of the transmission and distribution interface;

[0025] The feasible domain construction module runs to exchange the power range parameters of the transmission and distribution interfaces. , Write the power range constraints of the transmission and distribution interface switching in the feasible operating domain, and incorporate the flexibility ramp-up capability parameters. Write flexibility ramp-up constraints into the feasible domain of operation.

[0026] A further improvement of the present invention is that the rolling scheduling optimization decision layer includes:

[0027] The model building module is used to receive the feasible domain as the constraint set input and establish a rolling scheduling optimization model coupled across time scales. The rolling scheduling optimization model includes day-ahead planning variables and intraday rolling adjustment variables.

[0028] The output variable introduction module is used to introduce the controllable dispatch output variable of the new energy power station as a decision variable in the rolling scheduling optimization model, and to set the output boundary constraints of the controllable dispatch output variable of the new energy power station.

[0029] The consistency constraint module is used to construct cross-timescale consistency constraints in the rolling scheduling optimization model, including unit ramping constraints, minimum start-up and shutdown constraints, energy storage state of charge (SOC) evolution constraints, and transmission and distribution interface switching power consistency constraints. The transmission and distribution interface switching power consistency constraints are used to limit time periods within the same rolling scheduling time domain. Power switching settings of the transmission and distribution interface satisfy: In the formula, For rolling scheduling of time periods within the time domain The planned power switching capacity of the transmission and distribution interface is as follows. For rolling scheduling of time periods within the time domain Intraday rolling adjustment volume;

[0030] The scenario reduction module is used to perform scenario reduction based on the set of candidate scenarios with uncertainties in new energy output and the sensitivity index of safety constraints, forming a scenario set for online solution;

[0031] The robust rolling solver module is used to perform rolling solver on the rolling scheduling optimization model under the constraints of the scenario set, and output the scheduling scheme corresponding to the daily plan and the intraday rolling scheduling plan.

[0032] A further improvement of the present invention is that the controllable scheduling output variable of the new energy power station is the time period of each new energy power station within the rolling scheduling time domain. Controllable scheduling output The output boundary constraint is the upper limit of the carrying capacity output of the new energy power station determined based on the node carrying capacity boundary parameters and / or the regional carrying capacity boundary parameters. and the time period of rolling dispatch of new energy power stations Predictable available output Set together and satisfy: ;

[0033] The security constraint sensitivity index Determine using the following formula:

[0034] ;

[0035] In the formula, Candidate scenarios Time period within the rolling scheduling time domain Corresponding to the Class constraint utilization For the first Preset warning thresholds for class constraints, For the first Weights of class constraints; This represents the total number of constraint types included in the calculation within the feasible domain.

[0036] The scenario reduction module sorts the scenarios from largest to smallest according to the security constraint sensitivity index and selects the top N candidate scenarios to form the scenario set.

[0037] A further improvement of the present invention is that the online security verification and feedback layer includes:

[0038] The verification feature construction module is used to receive the scheduling scheme, construct and output a security verification feature vector, wherein the security verification feature vector includes at least: the state variables of the power grid digital twin and the power switching settings of the transmission and distribution interfaces in the scheduling scheme. With controllable and scalable power output And the utilization rate of at least one type of constraint in the feasible domain;

[0039] The proxy verification module is used to perform online security verification based on the security verification feature vector and the proxy model, and output the verification result and uncertainty measure. and initial values ​​of bearing capacity constraint parameters;

[0040] The high-fidelity verification module is used to verify the uncertainty measure. The high-fidelity verification calculation based on the power grid digital twin is triggered according to the preset uncertainty threshold, and the high-fidelity verification result and the verification correction amount of the bearing capacity constraint parameter are output. The high-fidelity verification calculation includes one or more of AC power flow verification and N-1 verification.

[0041] The constraint calibration module is used to generate corrected bearing capacity constraint parameters based on the verification results, the high-fidelity verification results and the verification correction amount of bearing capacity constraint parameters when high-fidelity verification calculation is triggered; and to generate corrected bearing capacity constraint parameters based on the verification results and the initial values ​​of the bearing capacity constraint parameters when high-fidelity verification calculation is not triggered. The module writes the corrected bearing capacity constraint parameters into the feasible region construction module to update the feasible region and outputs a re-solution trigger signal to the rolling scheduling optimization model.

[0042] The instruction decomposition module is used to convert the scheduling scheme into an execution scheduling instruction when the verification result meets the preset pass conditions, and if a high-fidelity verification calculation is triggered, the high-fidelity verification result also meets the preset pass conditions. Under the conditions of satisfying the power range constraints and flexibility ramp-up constraints of the transmission and distribution interface, the module decomposes the power setting value of the transmission and distribution interface in the execution scheduling instruction according to the preset decomposition rules. It is decomposed into distributed power supply power commands, adjustable load power commands, and energy storage charging and discharging power commands.

[0043] The feedback event module is used to receive the operation feedback data of the execution scheduling instruction, write the operation feedback data into the power grid digital twin to update the multidimensional carrying capacity state variables and the operational feasible domain, and calculate the execution deviation and prediction error distribution drift based on the operation feedback data. When the execution deviation exceeds a preset deviation threshold or the prediction error distribution drift exceeds a preset drift threshold, a re-solution trigger signal is output to the rolling scheduling optimization model.

[0044] A further improvement of the present invention is that the proxy verification module adopts an integrated proxy model, wherein the integrated proxy model consists of at least Each sub-model is constructed, and the security verification feature vector is inferred to obtain the first... Bearing capacity constraint parameters in time period Predicted initial values ​​of bearing capacity constraint parameters The initial value of the bearing capacity constraint parameter is defined as the integrated mean value output by the proxy. ,Right now ,in Number the sub-models;

[0045] And measure the uncertainty Defined as: ;

[0046] When the uncertainty measure High-fidelity verification calculation is triggered when the uncertainty exceeds a preset threshold; otherwise, it is not triggered.

[0047] A further improvement of the present invention is that the feedback event module will time period Execution deviation Defined as the switching power setting value of the transmission and distribution interface. The difference between the actual power exchange interface switching power measured in the operational feedback data and the actual power exchange interface switching power. : ;

[0048] And the drift of the prediction error distribution Define as a scrolling window Mean of the prediction error series of internal new energy sources with standard deviation Offset relative to the reference window: ;

[0049] Among them, the prediction error of new energy , For the time period measured in the operation feedback data The actual output of the new energy power station , These represent the mean and standard deviation of the new energy prediction error sequence within the preset benchmark window, respectively. These are preset non-negative weighting coefficients.

[0050] An adaptive scheduling method for high-proportion renewable energy power grids, the method comprising:

[0051] The power grid operation data and forecast data are acquired to construct a power grid digital twin that can be updated on a rolling basis. In the power grid digital twin, the power grid carrying capacity is modeled as a multi-dimensional carrying capacity state variable, the multi-dimensional carrying capacity state variable is mapped to carrying capacity constraint parameters, and an operational feasible domain is constructed to describe the boundary of safe operation of the power grid.

[0052] Under the condition of uncertainty in new energy output, a rolling scheduling optimization model coupled across time scales is established, and the rolling scheduling optimization model is solved to generate a scheduling scheme that satisfies the operational feasible domain constraint and the cross-time scale consistency constraint.

[0053] The system receives the scheduling scheme and the state variables of the power grid digital twin. Based on the power grid digital twin, it uses a proxy model coupled with the rolling scheduling optimization model to perform online safety verification of the scheduling scheme, outputs the verification results and the corrected carrying capacity constraint parameters, updates the operational feasible region based on the corrected carrying capacity constraint parameters, and triggers the rolling scheduling optimization model to resolve within the updated operational feasible region. When the verification passes, the scheduling scheme is converted into an execution scheduling command and issued, and the operational feedback data of the execution scheduling command is obtained. The operational feedback data is written into the power grid digital twin to update the multidimensional carrying capacity state variables and the operational feasible region.

[0054] Compared with existing technologies, the beneficial effects of this invention are as follows: By expressing the grid carrying capacity in a multi-dimensional state form and further forming a constraint set, it can unify multi-dimensional constraints such as power flow carrying capacity, absorption capacity, interface switching capacity, frequency security margin, and asset health budget into a computable operational boundary expression. This transforms the operational boundary from empirical limits / discrete rules into a constraint space that can be directly used for scheduling solutions, which is beneficial to improving the boundary consistency and interpretability of the scheduling model. When there is uncertainty in the output of new energy sources, the rolling scheduling solution mechanism can continuously revise the scheduling scheme under the condition of continuous prediction updates, and maintain the connection and stability between planning and execution under the action of cross-timescale consistency constraints, reducing large adjustments caused by prediction deviations and improving the executability and continuity of scheduling. The introduction of an online safety verification mechanism, combined with the real-time status of the grid operation model, allows for rapid verification of the scheduling scheme. When risk trends are detected, calibration results can be generated in a timely manner and resolving can be triggered, enabling the scheduling scheme to be updated and iterated within the boundary that is closer to the actual operating constraints. This reduces unnecessary conservative margins while ensuring safety, and improves the new energy absorption capacity and system operating efficiency. By writing the execution layer feedback data back to the operating model, the operating boundary can be continuously corrected according to the statistical characteristics of equipment status, topology changes and prediction errors. This avoids the model parameters from gradually deviating from the actual operation, enhances the system's adaptability to long-term operating environment changes, and improves the stability and sustainability of scheduling performance. Attached Figure Description

[0055] 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:

[0056] Figure 1 This is a schematic diagram of the modular structure of the system of the present invention;

[0057] Figure 2 This is a flowchart of the method in an embodiment of the present invention. Detailed Implementation

[0058] 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.

[0059] In this specification, unless otherwise specified, "time period" refers to a discrete scheduling time unit within the rolling scheduling time domain; "new energy power station" may include one or more of wind farms, photovoltaic power stations, and integrated wind-solar-storage power stations; "transmission and distribution interface" refers to the power exchange interface formed by the interconnection node, interconnection line, interconnection transformer, or combination thereof between the transmission network and the distribution network. All "power" quantities are in MW, "energy" quantities are in MWh, and "utilization rate" is a dimensionless ratio. If different time resolutions coexist, time granularity mapping is performed first before calculating relevant constraints and indicators.

[0060] like Figure 1 The illustration shows an embodiment of the present invention, which provides a capacity-adaptive dispatching system for high-proportion renewable energy power grids, applicable to a regional power grid dispatching scenario with high renewable energy access. The regional power grid includes at least:

[0061] A power transmission network consisting of a set of transmission lines, transformers, busbars and switchgear;

[0062] At least one power distribution area, and each power distribution area is connected to distributed new energy sources, adjustable loads, energy storage devices and power distribution lines;

[0063] Multiple new energy power stations, including centralized wind farms, centralized photovoltaic power stations, or distributed new energy clusters;

[0064] At least one power supply interface;

[0065] At least one scheduling master server and at least one edge computing node.

[0066] In this embodiment, the system is deployed in the scheduling master station server, which communicates with the following systems:

[0067] SCADA systems are used to acquire operational measurements such as voltage, current, power, and switch status.

[0068] The PMU system is used to acquire synchronization phasors, frequency, and rate of frequency change.

[0069] Distribution automation systems are used to acquire information on distribution-side lines, switches, distributed energy sources, and loads.

[0070] Meteorological forecasting systems are used to acquire forecast data such as ambient temperature, wind speed, wind direction, solar irradiance, and humidity.

[0071] The equipment management system is used to acquire equipment parameters, maintenance plans, asset health status, and life assessment data.

[0072] The new energy power prediction system is used to obtain the predicted available power output of new energy power plants at different times.

[0073] In one example configuration, the day-ahead planning time resolution is 1 hour, the intraday rolling adjustment time resolution is 15 minutes, the online verification refresh cycle is 5 minutes or 15 minutes, the SCADA sampling cycle is 2 to 10 seconds, the PMU sampling cycle is 20 to 100 milliseconds, the rolling scheduling time domain length is the next 24 hours, the weather forecast update cycle is 15 minutes to 1 hour, and the new energy power forecast update cycle is 15 minutes.

[0074] The system in this embodiment generally comprises a three-layer structure: a multi-dimensional bearing capacity state generation layer, a rolling scheduling optimization decision layer, and an online safety verification and feedback layer. The data flow between the three layers is as follows:

[0075] The multidimensional carrying capacity state generation layer receives power grid operation data and prediction data, outputs multidimensional carrying capacity state variables and carrying capacity constraint parameters, and forms an operational feasible domain.

[0076] The rolling scheduling optimization decision layer optimizes the scheduling scheme by using the feasible domain as the constraint set.

[0077] The online safety verification and feedback layer performs online verification of the scheduling scheme, outputs the corrected bearing capacity constraint parameters, and feeds the corrected bearing capacity constraint parameters back to the multi-dimensional bearing capacity state generation layer, thereby updating the feasible domain and triggering a re-solution.

[0078] The approved scheduling scheme is converted into an execution scheduling instruction and issued. The execution feedback data is written back to the digital twin for status updates in the next rolling cycle.

[0079] The multidimensional carrying capacity state generation layer is used to acquire power grid operation data and forecast data, construct a power grid digital twin that can be updated on a rolling basis, and model the power grid carrying capacity as a multidimensional carrying capacity state variable in the power grid digital twin, and further construct the operational feasible domain.

[0080] In this embodiment, the execution process of this layer is as follows:

[0081] Step S11: Collect power grid operation data and forecast data. Power grid operation data includes at least SCADA measurement data, PMU measurement data, power grid topology data, equipment parameter data, and protection setting data. Forecast data includes at least load forecast data, renewable energy output forecast data, and meteorological forecast data.

[0082] Step S12: Perform data preprocessing, including time synchronization, outlier removal, and missing value completion. Time synchronization is used to unify data from different sampling frequencies to the rolling scheduling calculation time scale. Outlier removal can be performed using the 3σ principle, residual threshold test, or consistency judgment based on state estimation residuals. Missing value completion can be performed using linear interpolation, Kalman filtering, or mean completion based on similar operating conditions.

[0083] Step S13: Perform hybrid state estimation and parameter identification based on the preprocessed data to obtain the state quantities and parameter calibration quantities of the power grid digital twin. The state quantities of the power grid digital twin include at least the bus voltage amplitude, phase angle, line power flow, unit output, interface switching power, energy storage state of charge, and frequency state; the parameter calibration quantities of the power grid digital twin include at least the line impedance correction, transformer parameter correction, load model parameter correction, and equipment aging correction.

[0084] Step S14: Based on the state variables of the power grid digital twin, the parameter calibration values ​​of the power grid digital twin, and the prediction data, calculate the multidimensional carrying capacity state variables. The multidimensional carrying capacity state variables include at least the dynamic heat capacity value of transmission lines, the upper limit of distributed renewable energy access, the power exchange boundary of transmission and distribution interfaces, the frequency security carrying capacity boundary, and the asset health use budget boundary.

[0085] Step S15: Obtain bearing capacity constraint parameters by mapping multidimensional bearing capacity state variables, and construct the operational feasible domain based on the bearing capacity constraint parameters.

[0086] In one specific embodiment, the multidimensional carrying capacity state generation layer includes: a data acquisition and preprocessing module, a power grid digital twin rolling update module, a multidimensional carrying capacity state calculation module, a carrying capacity constraint parameterization module, and an operational feasible domain construction module.

[0087] The data acquisition and preprocessing module collects raw data from multiple data sources and transforms it into a unified data format suitable for subsequent state estimation, parameter identification, carrying capacity calculation, and optimization modeling. Specifically, for SCADA measurement data, it acquires bus voltage, active power, reactive power, line power flow, and switch status; for PMU measurement data, it acquires synchronous phasors, frequency, and rate of change of frequency; for power grid topology data, it acquires line connection relationships, equipment location areas, and switch topology status; for equipment parameter data, it acquires line impedance, conductor type, transformer capacity, rated capacity and rated power of energy storage, unit ramp-up capability, protection settings, and equipment health parameters; and for forecast data, it acquires load forecasts for each time period, predicted available output of new energy sources, and meteorological forecast results. After preprocessing, a dataset with a unified time scale is output. .

[0088] The power grid digital twin rolling update module is based on Perform hybrid state estimation and parameter identification. In one implementation, hybrid state estimation employs a joint weighted least squares method, considering the state variables of the power grid digital twin. At least including:

[0089] ;

[0090] In the formula, Indicates the amplitude of the bus voltage. Indicates the phase angle of the bus voltage. , These represent the active and reactive power outputs of the generator, respectively. Indicates the switching power of the transmission and distribution interface. Indicates the state of charge of the energy storage;

[0091] Parameter identification can employ least squares fitting, recursive least squares, or extended Kalman filtering to correct line parameters, transformer parameters, and equivalent load parameters, thereby forming the parameter calibration quantities of the power grid digital twin.

[0092] The multidimensional bearing capacity state calculation module is used to calculate multidimensional bearing capacity state variables, specifically including:

[0093] Dynamic heat capacity value of transmission lines: For lines Based on meteorological forecast data and conductor parameters, its duration during the specified time period is calculated. The dynamic heat capacity value. The circuit heat balance relationship satisfies:

[0094] Heat input includes Joule heat generated by the current in the conductor and heat generated by solar radiation;

[0095] Heat output includes convection heat dissipation and radiation heat dissipation.

[0096] The maximum allowable current corresponding to the steady-state temperature constraint of the conductor is solved iteratively and then converted into a dynamic heat capacity value.

[0097] Distributed renewable energy access capacity limit: for nodes or region The upper limit for new energy access capacity is calculated based on the following factors:

[0098] Does the node voltage deviation meet the allowable range?

[0099] Does the power flow exceed the limit?

[0100] Has the protective coordination been compromised?

[0101] The level of reverse power flow allowed by the upstream power grid;

[0102] Reactive power support and energy storage regulation capabilities on the distribution side.

[0103] The maximum load-bearing capacity of a node is denoted as The upper limit of regional carrying capacity is denoted as .

[0104] Power boundary of transmission and distribution interface: Based on the capacity of the transmission and distribution interface equipment, the power flow margin of surrounding lines, voltage stability constraints, and the availability of flexible resources on the distribution side, the power boundary of the transmission and distribution interface during the specified time period is calculated. The switching power boundary is denoted as and .

[0105] Frequency safety bearing boundary: Based on the system's equivalent inertia, spinning reserve capacity, primary frequency regulation capacity, and minimum frequency point constraint, the frequency safety bearing boundary is calculated and further mapped to spinning reserve adequacy constraint parameters, frequency response constraint parameters, and inertia constraint parameters.

[0106] Asset health usage budget boundary: Establish a health consumption model for critical equipment. Assume the equipment... During the period The health consumption is Allow health budgets to The health usage limit parameter is determined by the equipment's allowable budget and planned maintenance cycle.

[0107] Bearing capacity constraint parameterization module: Maps multidimensional bearing capacity state variables to bearing capacity constraint parameters, with the following mapping relationship:

[0108] The dynamic heat capacity value of transmission lines is mapped to dynamic limit parameters. ;

[0109] The upper limit of distributed new energy access capacity is mapped to the node carrying capacity boundary parameter. With regional bearing capacity boundary parameters ;

[0110] The power boundary of the transmission and distribution interface is mapped to the power range parameter of the transmission and distribution interface. , and flexibility climbing ability parameters ;

[0111] The frequency safety bearing boundary is mapped to the spinning reserve adequacy constraint parameters, frequency response constraint parameters, and inertia constraint parameters.

[0112] The asset health usage budget boundary is mapped to the health usage limit parameter.

[0113] The feasible region construction module determines the network state base value and parameter coefficients corresponding to each constraint based on the state variables and parameter calibration variables of the power grid digital twin, and constructs the feasible region in combination with the bearing capacity constraint parameters;

[0114] In this embodiment, the feasible domain is run. Includes the following constraints:

[0115] ;

[0116] In the formula, Indicates power flow constraints. Indicates the node bearing capacity constraint. Indicates regional bearing capacity constraints. This indicates the power range constraint of the transmission and distribution interface. Indicates flexibility in hill climbing constraints. This indicates a spindle reserve adequacy constraint. Indicates frequency response constraints. Indicates inertia constraint. This indicates that health status uses quota constraints.

[0117] The coefficients of each constraint in the feasible domain are derived from the current state of the digital twin, thus reflecting the actual operating conditions rather than fixed offline limits.

[0118] In this embodiment, the multi-dimensional load-bearing capacity calculation module calculates the predicted average value of the dynamic heat capacity of the transmission line based on meteorological forecast data and equipment parameter data, according to the heat balance calculation of the transmission line. and the predicted standard deviation The bearing capacity constraint parameterization module generates dynamic limit parameters according to the uncertainty adaptive reduction rule. .

[0119] For the line During the period Based on multiple meteorological forecast samples, historical error distributions, and model outputs, the predicted mean value of dynamic heat capacity is statistically obtained. and the predicted standard deviation Dynamic limit parameters The calculation formula is:

[0120] ;

[0121] In the formula, Indicates the transmission line number. Indicates the time period number within the rolling scheduling time domain; , For the preset non-negative coefficients, where, Used to reflect the basic safety margin Used to amplify the reduction range when uncertainty is high.

[0122] Running the feasible domain construction module will dynamically limit parameters Write the upper bound of the power flow constraint for the line During the period trend ,have: It is used to automatically increase the safety margin when there are large weather changes or high forecast errors, and to reduce conservatism when there is low uncertainty, thereby increasing the space for new energy consumption.

[0123] The load-bearing capacity constraint parameterization module generates the power range parameters for the transmission and distribution interface. , The calculation takes into account the following factors: the capacity of the transmission and distribution interface transformer, the power flow margin of the transmission-side associated lines, the distribution-side's ability to absorb new energy sources, the distribution-side's adjustable load and energy storage regulation capacity, voltage stability and reactive power support capacity.

[0124] Flexibility and climbing ability parameters This parameter characterizes the maximum allowable change in exchange power at the transmission and distribution interface between adjacent time periods. It can be determined by the following factors: the allowable rate of change of the interconnecting transformer, the network's ability to track flexible resources, the load-side response capability to interruptible or transferable loads, and the rated power and rate of change limits of the energy storage device.

[0125] For time periods within the rolling scheduling time domain Power switching settings of the transmission and distribution interface Write the following constraints into the feasible domain:

[0126] ;

[0127] ;

[0128] In the formula, Indicates the time period within the rolling scheduling time domain The power switching setting value of the transmission and distribution interface;

[0129] This constraint limits both the amplitude of the interface power and the rate of change between adjacent time periods, preventing intraday rolling optimization from yielding an interface power trajectory that is mathematically feasible but practically impossible for the transmission and distribution system to smoothly track.

[0130] The rolling scheduling optimization decision layer is used to establish a rolling scheduling optimization model coupled across time scales under the condition of uncertainty in new energy output, and to solve the model to generate a scheduling scheme that satisfies the operational feasible domain constraint and the cross-time scale consistency constraint.

[0131] In this embodiment, the rolling scheduling optimization model aims to minimize the overall operating cost and risk. The overall operating cost and risk may include one or more of the following: conventional unit fuel cost, unit start-up and shutdown cost, energy storage charging and discharging loss cost, wind and solar curtailment cost, power fluctuation penalty cost at transmission and distribution interfaces, and safety limit violation penalty cost.

[0132] The rolling solution adopts a control method of "current execution, future prediction, and periodic rolling". That is, in each rolling cycle, only the control quantity of the current execution period is extracted and issued, and the remaining periods are used as prediction references and recalculated in the next rolling cycle.

[0133] In one specific embodiment, the rolling scheduling optimization decision layer includes: a model building module, an output variable introduction module, a consistency constraint module, a scenario reduction module, and a robust rolling solution module.

[0134] The model building module uses the feasible region as the constraint set input to establish a rolling scheduling optimization model coupled across time scales. In one implementation, the rolling scheduling optimization model includes at least the following decision variables:

[0135] The conventional generating units are scheduled to operate at a recent time. and intraday adjustment volume ;

[0136] The planned switching power of the transmission and distribution interface is currently... and intraday rolling adjustment volume ;

[0137] Power switching settings for transmission and distribution interfaces ;

[0138] Energy storage charging power Discharge power and SOC;

[0139] Controllable dispatchable output variables of new energy power stations ;

[0140] Unit start-up and shutdown status variables .

[0141] The optimization objective can be expressed as:

[0142] ;

[0143] In the formula, Indicates the cost of electricity generation. Indicates start-up and shutdown costs. Indicates the cost of using energy storage. This signifies punishment for abandoning wind and light. This indicates the fluctuation cost of switching at the transmission and distribution interface. This indicates a penalty for exceeding safety limits.

[0144] The output variable introduction module is used to introduce the controllable dispatchable output variables of new energy power plants and set their output boundary constraints. The controllable dispatchable output variables of new energy power plants are the time periods of each new energy power plant within the rolling dispatch time domain. Controllable power output scheduling, with power output boundary constraints based on the upper limit of the carrying capacity of the new energy power station. and the time period of rolling dispatch of new energy power stations Predictable available output Set together and satisfy: .

[0145] For wind farms, the predicted available power output can be calculated by combining wind speed forecasts, wind turbine power curves, the number of available units, and curtailment plans; for photovoltaic power plants, the predicted available power output can be calculated by combining solar irradiance forecasts, module temperature forecasts, inverter efficiency, and module availability.

[0146] Maximum Capacity of New Energy Power Stations It is obtained through allocation of nodal bearing capacity boundary parameters and / or regional bearing capacity boundary parameters. In one implementation, for new energy power stations... Node and region Its maximum load-bearing capacity is:

[0147] ;

[0148] In the formula, These are the boundary parameters for the bearing capacity of the nodes; These are the boundary parameters for the regional bearing capacity. Rated capacity for new energy power stations; and The allocation coefficient can be determined according to one or more of the following rules:

[0149] Allocation based on the proportion of rated capacity of new energy power stations;

[0150] Allocation based on node or regional power flow sensitivity;

[0151] Allocation is based on priority for new energy power stations;

[0152] Allocation is based on historical absorption weight.

[0153] If capacity allocation is adopted, then the following conditions are met: , ;

[0154] in, For nodes The following is a collection of related new energy power stations. For the region The following is a collection of related new energy power stations.

[0155] The consistency constraint module is used to construct consistency constraints across time scales in the rolling scheduling optimization model, including:

[0156] Unit ramp-up constraints: ,in, , These represent the unit's maximum uphill and downhill climbing capabilities, respectively. , These represent conventional generating units. During the period Time period Those who have made meritorious contributions;

[0157] Minimum start-stop constraints: through start-stop state variables And implement constraints on cumulative power-on and power-off duration;

[0158] Constraints on the evolution of the state of charge (SOC) of energy storage: And satisfy: ,in, , These represent the time periods of the energy storage devices. Time period The state of charge; , These are charging efficiency and discharging efficiency, respectively. The duration of the time period; , These are the minimum allowable state of charge and the maximum allowable state of charge, respectively.

[0159] Power consistency constraints for transmission and distribution interfaces: ,in, For rolling scheduling of time periods within the time domain The planned power switching capacity of the transmission and distribution interface is as follows. For rolling scheduling of time periods within the time domain Intraday rolling adjustment volume; For rolling scheduling of time periods within the time domain The power switching setting value of the transmission and distribution interface.

[0160] If the time granularity of the day before and the day before is different, the day before schedule is first mapped to the same time resolution according to the rolling scheduling time domain, and then the above consistency relationship is executed.

[0161] The scenario reduction module performs scenario reduction based on the candidate scenario set for uncertainty in renewable energy output and the sensitivity index of safety constraints, forming a scenario set for online solution. Each scenario in the candidate scenario set contains the renewable energy output trajectory for each time period within the rolling scheduling time domain. Scenarios can be obtained through the following methods:

[0162] Monte Carlo sampling is performed based on the prediction error distribution;

[0163] Typical trajectories were extracted based on historical data and meteorological conditions.

[0164] Construct multi-branch trajectories based on the scene tree method.

[0165] In this embodiment, each candidate scenario in the set of candidate scenarios for uncertainty in new energy output is considered. A corresponding security constraint sensitivity index The calculation formula is as follows:

[0166] ;

[0167] In the formula, Candidate scenarios Time period within the rolling scheduling time domain Corresponding to the Class constraint utilization For the first Preset warning thresholds for class constraints, For the first Weights of class constraints; This represents the total number of constraint types included in the calculation within the feasible domain.

[0168] If the utilization rate of a certain constraint does not exceed the corresponding warning threshold, it will not penalize the scenario sensitivity index; if it exceeds the threshold, the greater the degree of exceedance, the greater the penalty. The greater the contribution, the greater the impact of high-risk constraints can be achieved through the squared penalty term.

[0169] For example, this embodiment provides the following calculation method for the utilization rate of various constraints:

[0170] Power Flow Constraint Utilization ;

[0171] Node bearing capacity constraint utilization rate ;

[0172] Regional carrying capacity constraint utilization rate ;

[0173] Power range constraints of transmission and distribution interface switching utilization

[0174] Flexibility and hill-climbing constraint utilization ;

[0175] Spinning reserve adequacy constraint utilization rate ;

[0176] Frequency response constraint utilization ;

[0177] Inertia constraint utilization ;

[0178] Health usage limit constraint utilization rate ;

[0179] In the formula, Candidate scenarios Downline During the period Line flow; Candidate scenarios Next node During the period The new energy injection power or equivalent carrying power; Candidate scenarios Lower region During the period The new energy aggregation injection power or equivalent carrying power; , Candidate scenarios Next period Time period The power switching setting value of the transmission and distribution interface; For time period The upper limit parameter of the power exchange capacity of the transmission and distribution interface; For time period The lower bound parameter of the power exchange capacity of the transmission and distribution interface; Candidate scenarios Next period Required spindle reserve capacity; For time period The available spinning reserve capacity is obtained by summing up available unit reserves, energy storage reserves, interruptible load reserves, etc. Candidate scenarios Next period Required frequency response capability; For time period It can provide frequency response capability, which is obtained by combining primary frequency regulation resources, energy storage rapid response resources, etc. Candidate scenarios Next period Required system equivalent inertia level; For time period The system's available equivalent inertia level is obtained by summing up the inertia of grid-connected synchronous generator units and virtual inertia resources. Candidate scenarios Next period Health budget consumption. For time period Allowed health usage limits.

[0180] In one implementation, a preset warning threshold is used. The weight can be 0.85, 0.90, or 0.95; The constraints can be set according to their importance, the severity of the consequences of exceeding the limit, and the statistical results of historical accidents. Larger weights can be set for frequency response constraints and inertia constraints, and different weights can be set for local node bearing capacity constraints according to the importance level of the nodes.

[0181] Scene reduction module according to Sort the scenarios from largest to smallest, and select the top N candidate scenarios to form a scenario set, where N is a preset positive integer. For example, when online solution time is limited, N can be set to 10 or N=20; when computing resources are sufficient, N can be appropriately increased. By reducing the number of scenarios, high-risk scenarios that are most sensitive to key constraints in the feasible domain can be retained first, thereby improving the robustness of the scheduling scheme to uncertainty while controlling the scale of online solution.

[0182] The robust rolling solver module is used to perform rolling solves on the rolling scheduling optimization model under scenario set constraints, and outputs the scheduling schemes corresponding to the daily plan and the intraday rolling scheduling plan. In one implementation, the following process is performed for each rolling cycle: input the current digital twin state and the updated feasible domain; input the scenario set; establish the robust rolling scheduling optimization model; solve for the full-time domain scheduling scheme; extract the scheduling instructions for the current execution period and issue them; wait for the next rolling cycle to reacquire data and update the solution.

[0183] The online safety verification and feedback layer is executed in each rolling scheduling cycle. It is used to verify the scheduling scheme output by the rolling scheduling optimization decision layer online, trigger high-fidelity verification and constraint calibration when necessary, and generate execution scheduling instructions after the verification is passed. At the same time, it triggers the solution again based on the running feedback data.

[0184] In this embodiment, the execution process of the online security verification and feedback layer is as follows:

[0185] Step S31: Obtain the scheduling scheme output by the rolling scheduling optimization decision layer and the current rolling cycle state of the power grid digital twin;

[0186] Step S32: Input the scheduling scheme and the current digital twin state into the proxy model, and the proxy model outputs the corrected bearing capacity constraint parameters;

[0187] Step S33: Update the operational feasible region based on the revised bearing capacity constraint parameters;

[0188] Step S34: Trigger the rolling scheduling optimization model to be re-solved within the updated feasible region;

[0189] Step S35: If the re-solved scheduling scheme passes the online safety verification, then generate and issue the execution scheduling instruction; otherwise, repeat steps S32 to S34 until the safety verification requirements are met or the preset conservative control strategy is triggered.

[0190] Step S36: Collect the operation feedback data after executing the scheduling command, and write the operation feedback data into the power grid digital twin to update the multi-dimensional carrying capacity state variables and the operational feasible domain; wherein, the operation feedback data includes at least: actual line power flow, actual interface switching power, actual new energy power plant output, actual energy storage charging and discharging status, actual frequency response, and equipment health consumption.

[0191] In this embodiment, the proxy model and the rolling scheduling optimization model work together in an outer loop coupling manner. Specifically, the rolling scheduling optimization model first obtains a scheduling scheme based on the power grid digital twin state variables, prediction data, and operational feasible region of the current rolling cycle; then, the online safety verification and feedback layer calls the proxy model to verify the scheduling scheme and generates initial or corrected values ​​of the carrying capacity constraint parameters; then, the operational feasible region construction module updates the operational feasible region based on the corrected carrying capacity constraint parameters and triggers the rolling scheduling optimization model to re-solve within the updated operational feasible region.

[0192] In one implementation, the external loop coupling process can be configured with a maximum number of iterations; if the preset pass condition is not met even after reaching the maximum number of iterations, a preset conservative control strategy is triggered. The preset conservative control strategy includes one or more of the following: tightening the power switching range of the transmission and distribution interface, increasing reserve requirements, reducing the upper limit of adjustable output of new energy sources, and increasing the safety margin of the health budget.

[0193] By using the above external loop coupling method, a closed-loop linkage between the rolling scheduling optimization model and the online safety verification can be achieved without changing the internal structure of the optimization solver, thereby improving the solution solving efficiency and engineering deployability.

[0194] The inputs to the online security verification and feedback layer include:

[0195] Scheduling scheme: at least includes , Conventional unit output, energy storage capacity, and SOC, etc.

[0196] The state quantities of the power grid digital twin include at least bus voltage / phase angle, line power flow, switch topology, equipment parameter calibration quantities, measured interface power, SOC, frequency indicators, etc.

[0197] Feasible domain constraint parameters and constraint set: including dynamic limit parameters, node / region bearing boundaries, interface exchange range and ramp parameters, spinning reserve / frequency response / inertia parameters, health limit, etc.

[0198] The output includes:

[0199] The revised bearing capacity constraint parameters (written into the running feasible domain construction module to update the running feasible domain).

[0200] Output a re-solution trigger signal to the rolling scheduling optimization model;

[0201] Execute scheduling commands (distributed power sources, adjustable loads, energy storage charging and discharging power commands, etc.);

[0202] Feedback event trigger signal (based on deviation and drift).

[0203] In one specific embodiment, the online security verification and feedback layer includes: a verification feature construction module, a proxy verification module, a high-fidelity verification module, a constraint calibration module, an instruction decomposition module, and a feedback event module.

[0204] The verification feature construction module uniformly encodes "digital twin status + scheduling scheme + feasible domain stress level" into a security verification feature vector. It is used to verify the inference of the proxy model. It should include at least the following components (which can be concatenated as column vectors):

[0205] Digital twin state components Bus voltage magnitude and phase angle (or its statistics, such as maximum / minimum / mean), power flow on critical lines. Its utilization rate, topology status (critical switches, tie-line status, which can be encoded with 0 / 1), and energy storage. The upper limit of available charging and discharging power for energy storage, and frequency indicators (frequency, RoCoF or equivalent frequency safety indicators).

[0206] Scheduling scheme components : Power setting value of transmission and distribution interface Controllable dispatch of power output from new energy power stations (Can be a station-level vector or aggregate value), conventional unit output and standby arrangements (or aggregate value), energy storage charging / discharging power settings;

[0207] The stress level component of the feasible domain Utilization rate of at least one type of constraint or margin In this embodiment, the constraint utilization rate preferentially adopts the defined caliber (such as power flow constraint utilization rate, interface switching power range constraint utilization rate, flexibility ramp constraint utilization rate, frequency / inertia / health budget constraint utilization rate, etc.), and takes the maximum value or Top-K mean as the aggregation feature for multiple lines / multiple nodes / multiple regions to ensure that the dimension is fixed.

[0208] To avoid instability in model training and inference caused by differences in units of measurement, this embodiment... Continuous variables are normalized in one of the following ways: normalized to rated values ​​(e.g., power divided by rated capacity); normalized to historical statistical mean and variance (z-score); or normalized to upper bound parameters (e.g., power flow divided by dynamic limit).

[0209] The proxy verification module uses an integrated proxy model to verify... Reasoning, outputting the verification result (pass / fail) and uncertainty measurement. Initial values ​​of bearing capacity constraint parameters (for subsequent calibration and updates).

[0210] Number the bearing capacity constraint parameters by type. Organization ("one number for each type of parameter"), for example:

[0211] Dynamic quota parameters (Critical path sets can be aggregated);

[0212] Boundary parameters of nodal bearing capacity (Aggregation or sub-nodes);

[0213] Boundary parameters of regional bearing capacity ;

[0214] Parameters of power range for transmission and distribution interfaces ;

[0215] Parameters of flexibility and climbing ability ;

[0216] Rotational reserve adequacy constraint parameters;

[0217] Frequency response constraint parameters;

[0218] Inertia constraint parameters;

[0219] Health status usage limit parameter.

[0220] In this embodiment, using Indicates the first Bearing capacity constraint parameters in time period The scalar representation (for vector parameters, the maximum / minimum / critical section value / scalar aggregated according to rules can be used).

[0221] The integrated agent model consists of at least Sub-model composition (typically taken) ), No. The sub-model outputs the first Bearing capacity constraint parameters in time period Predicted initial values ​​of bearing capacity constraint parameters The initial values ​​of the bearing capacity constraint parameters are defined as the integrated mean values ​​output by the proxy. ,Right now ,in Number the sub-models;

[0222] The uncertainty measure is determined based on the degree of dispersion of the output results of each sub-model. Defined as: ;

[0223] To achieve uniform triggering across the entire time domain, you can define... ,when A high-fidelity verification is triggered when the preset uncertainty threshold is exceeded; this embodiment uses time-based triggering, meaning that as long as any time period exists... If the triggering conditions are met, a high-fidelity check will be triggered for the entire time domain or the current window.

[0224] The proxy verification module outputs a "pass / fail" verification result, which can be implemented according to one of the following rules: the proxy model directly outputs the classification result; or the proxy model outputs the constraint margin prediction, and if all key constraint margins are ≥0, it is judged as pass; otherwise, it fails. This embodiment adopts the second method, where the proxy model outputs the constraint margin predictions for each type of constraint. If all execution windows are in the current execution window All satisfy (And if the utilization rate of critical constraints is ≤1), then the proxy verification result is passed.

[0225] The high-fidelity verification module is used to perform high-fidelity calculations based on the power grid digital twin when the uncertainty is high, and outputs high-fidelity verification results and verification corrections for bearing capacity constraint parameters.

[0226] Let the preset uncertainty threshold be... (Given by historical statistics or configuration files, typically taking...) (95th percentile)

[0227] like This triggers a high-fidelity check.

[0228] like High-fidelity verification is not triggered; only the initial values ​​of the bearing capacity constraint parameters output by the proxy verification are used for updating.

[0229] High-fidelity verification includes at least one or more of the following:

[0230] AC-PF (Alternating Current Power Flow Verification): Calculate voltage, power flow, reactive power, etc., using the parameter calibration values ​​of the digital twin and the current dispatch scheme, and check for line / transformer over-limit and voltage over-limit.

[0231] N-1 verification: In the preset fault set Next, disconnect one line / transformer / component one by one, re-execute AC power flow, and check for over-limit situations.

[0232] To ensure feasibility and avoid introducing uncertainty, this embodiment adopts a determination strategy of "preset set + risk screening":

[0233] Preset Fault Set : Fixed by the dispatching department or system configuration (e.g., critical lines / main transformer sets);

[0234] Risk screening: When high-fidelity verification is triggered, the power flow utilization rate in the current feasible operating domain is used as a basis. Construct a subset of Top-Q lines / sections (e.g., Q=10). ;

[0235] Final verification set: .

[0236] High-fidelity verification results should include at least:

[0237] Whether it passed (passed / failed);

[0238] The true margin or over-limit of critical constraints (e.g., line over-limit MW, voltage over-limit pu).

[0239] Correction amount for corresponding bearing capacity constraint parameters .

[0240] To enable those skilled in the art to implement this directly, this embodiment provides a general "over-limit → tightening" correction method:

[0241] For upper bound parameters (such as dynamic limits, interface upper limits, node / region capacity limits), if high-fidelity verification results in exceeding the limit... ,but ,in This is the preset correction step size coefficient.

[0242] For lower bound parameters (e.g., interface lower bound) (If the value is negative), if the limit is exceeded, it will be adjusted according to the consistent sign (fixed sign convention in the instruction manual);

[0243] If the limit is not exceeded, then .

[0244] The constraint calibration module integrates the proxy verification output with the high-fidelity verification output to generate the corrected bearing capacity constraint parameters and triggers the optimization re-solution.

[0245] For each type of parameter The revised bearing capacity constraint parameters are defined as follows:

[0246] If high-fidelity verification is triggered: ;

[0247] If high-fidelity verification is not triggered: ;

[0248] It can also perform boundary truncation to ensure physical rationality (e.g., parameters are not less than 0, and do not exceed the rated value): ,in Preset upper and lower bounds.

[0249] Will Map back to various carrying capacity constraint parameters (dynamic limit, boundary, interface range, reserve, frequency response, inertia, health limit, etc.), update the set of feasible domain constraints, and send a re-solve trigger signal to the rolling scheduling optimization model.

[0250] When both the proxy verification result and the high-fidelity verification result meet the preset passing conditions, the instruction decomposition module issues an execution scheduling instruction. The preset passing conditions include at least:

[0251] All key constraints within the current execution window have not exceeded the limits (utilization ≤ 1 or margin ≥ 0).

[0252] If the N-1 check is triggered, then in The set does not exceed the limit (or the exceedance does not exceed the preset tolerance threshold).

[0253] Exchange power setting value of the transmission and distribution interface Decomposed into distributed power command vectors (Set output of each distributed power source), adjustable load power command vector (Power reduction / transfer for each adjustable load, defined according to agreed direction) and energy storage charging / discharging power command vector (Positive and negative signs indicate charging / discharging, or can be represented by separate variables);

[0254] To ensure that the instruction decomposition can be implemented, this embodiment uses a defined solvable optimization rule as the decomposition rule:

[0255] Under the constraints of resource upper / lower limits, ramp-up, and SOC, minimize the sum of squares of the deviations between the decomposed interface power and the set value: ,in The equivalent switching power of the interface is calculated from the decomposition command (which can be given by the power balance and interface measurement model).

[0256] The constraints must include at least:

[0257] Distributed power generation output limits and ramping constraints;

[0258] Adjustable load limits and ramp constraints;

[0259] Upper and lower limits of energy storage charging and discharging power, upper and lower limits of SOC, and SOC evolution constraints;

[0260] Power range constraints and flexibility ramping constraints for transmission and distribution interfaces (ensuring that the decomposition results do not violate interface constraints).

[0261] After solving , , Send the data to the corresponding control terminal or aggregation controller.

[0262] The feedback event module receives operational feedback data from the execution of scheduling instructions, writes it back to the power grid digital twin to update the multidimensional carrying capacity state variables and the operational feasible region, and simultaneously calculates the execution deviation. With the drift of the prediction error distribution The formula is:

[0263] ;

[0264] ;

[0265] In the formula, For the time period measured in the operation feedback data The actual power exchange capacity of the transmission and distribution interface (in MW) is obtained from the SCADA / PMU power measurement aggregation using a time-period average. The power setting value is exchanged at the transmission and distribution interface corresponding to the execution of the scheduling command;

[0266] New energy prediction error Defined as , For the time period measured in the operation feedback data The actual output of new energy power stations (the average value of the time period from the station's metering / SCADA aggregation). For time period The predicted available output (from the prediction system, the predicted value is frozen at the time of generating the scheduling scheme).

[0267] In one implementation method and All values ​​are aggregated across all power stations (summed over all new energy power stations), thus This is the system-level prediction error; another implementation method is to calculate the maximum drift for each station separately. This specification uses the "aggregate value" caliber to ensure the uniqueness of the symbol.

[0268] Set a scrolling window Given the most recent L time periods (e.g., L=16 corresponds to 4 hours), and the baseline window is a historical reference window (e.g., a fixed window for the same type of weather days in the same season or a stable window based on rolling statistics from the most recent week), then:

[0269] , :window Inside The mean and standard deviation;

[0270] , Within the baseline window The mean and standard deviation;

[0271] Prediction error distribution drift Defined as: ,in To preset non-negative weighting coefficients, You can choose 1, or set the weights according to the impact of deviation and variance on system risk.

[0272] Let the deviation threshold be (A fixed percentage of the rated capacity of the transmission and distribution interface can be used, such as 2%–5%, or the 95th percentile of historical deviations), the drift threshold is... (The 95th percentile of historical drift can be used, or a preset value can be used based on operational experience), when the following conditions are met: or Then, a re-solution trigger signal is output to the rolling scheduling optimization model.

[0273] like Figure 2 As shown, another embodiment of the present invention provides a capacity-adaptive scheduling method for high-proportion renewable energy power grids, comprising the following steps:

[0274] S1: Obtain power grid operation data and forecast data to construct a power grid digital twin that can be updated on a rolling basis. In the power grid digital twin, the power grid carrying capacity is modeled as a multi-dimensional carrying capacity state variable, the multi-dimensional carrying capacity state variable is mapped to the carrying capacity constraint parameter, and an operational feasible domain is constructed to describe the boundary of safe operation of the power grid.

[0275] S2: Under the condition of uncertainty in new energy output, establish a rolling scheduling optimization model coupled across time scales, solve the rolling scheduling optimization model to generate a scheduling scheme that satisfies the operational feasible domain constraint and the cross-time scale consistency constraint;

[0276] S3: Receive the scheduling scheme and the state variables of the power grid digital twin. Based on the power grid digital twin, use a surrogate model trained under self-supervision and embedded in the optimization solution process to perform online safety verification of the scheduling scheme and output the corrected carrying capacity constraint parameters. Update the operational feasible domain based on the corrected carrying capacity constraint parameters and trigger the rolling scheduling optimization model to re-solve within the updated operational feasible domain. When the verification is passed, convert the scheduling scheme into an execution scheduling command and issue it, and obtain the operational feedback data of the execution scheduling command. Write the operational feedback data into the power grid digital twin to update the multi-dimensional carrying capacity state variables and the operational feasible domain.

[0277] In summary, this invention can achieve dynamic expression and rolling updates of the operating boundary under high-proportion renewable energy grid operation conditions. Combined with online safety verification and feedback calibration, it forms a scheduling closed loop, improving scheduling feasibility and renewable energy absorption capacity while ensuring safety constraints. It has good engineering application value.

[0278] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various variations or substitutions within the technical scope disclosed in this application, and these should all 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 capacity-adaptive dispatching system for high-proportion renewable energy power grids, characterized in that, The system adopts a three-layer structure, including: A multidimensional carrying capacity state generation layer is used to acquire power grid operation data and prediction data to construct a power grid digital twin that can be updated on a rolling basis. In the power grid digital twin, the power grid carrying capacity is modeled as a multidimensional carrying capacity state variable, the multidimensional carrying capacity state variable is mapped to carrying capacity constraint parameters, and an operational feasible domain is constructed to describe the boundary of safe operation of the power grid. The operational feasible domain represents the allowed operating state space of the power grid in each time period within the rolling scheduling time domain in the form of a constraint set. The rolling scheduling optimization decision layer is used to establish a rolling scheduling optimization model coupled across time scales under the condition of uncertainty in new energy output, and to solve the rolling scheduling optimization model to generate a scheduling scheme that satisfies the operational feasible domain constraint and the cross-time scale consistency constraint. An online safety verification and feedback layer is used to receive the scheduling scheme, perform online safety verification of the scheduling scheme based on the power grid digital twin using a proxy model coupled with the rolling scheduling optimization model, output the verification result and the corrected carrying capacity constraint parameters, update the operational feasible domain based on the corrected carrying capacity constraint parameters, and trigger the rolling scheduling optimization model to re-solve within the updated operational feasible domain; when the verification is passed, the scheduling scheme is converted into an execution scheduling instruction and issued, and the operational feedback data of the execution scheduling instruction is obtained. The operational feedback data is written into the power grid digital twin to update the multidimensional carrying capacity state variables and the operational feasible domain. The multidimensional bearing capacity state generation layer includes: The data acquisition and preprocessing module is used to acquire power grid operation data and forecast data, and perform at least one preprocessing operation among time synchronization, outlier removal, and missing value completion; the power grid operation data includes SCADA measurement data, PMU measurement data, power grid topology data, equipment parameter data, and protection setting data, and the forecast data includes load forecast data, new energy output forecast data, and meteorological forecast data; The power grid digital twin rolling update module is used to perform hybrid state estimation and parameter identification based on preprocessed power grid operation data, and generate power grid digital twin state variables and power grid digital twin parameter calibration variables. The multidimensional carrying capacity state calculation module is used to calculate multidimensional carrying capacity state variables based on the state quantities of the power grid digital twin, the parameter calibration quantities of the power grid digital twin, and the prediction data. The multidimensional carrying capacity state variables include the dynamic heat capacity value of the transmission line, the upper limit of the distributed new energy access carrying capacity, the power exchange boundary of the transmission and distribution interface, the frequency security carrying capacity boundary, and the asset health use budget boundary. The carrying capacity constraint parameterization module is used to map the multidimensional carrying capacity state variables into carrying capacity constraint parameters, including: mapping the dynamic heat capacity value of the transmission line into a dynamic limit parameter; mapping the upper limit of distributed new energy access carrying capacity into node carrying capacity boundary parameters and regional carrying capacity boundary parameters; mapping the power exchange boundary of the transmission and distribution interface into power exchange range parameters and flexibility ramping capability parameters; mapping the frequency safety carrying capacity boundary into spinning reserve adequacy constraint parameters, frequency response constraint parameters, and inertia constraint parameters; and mapping the asset health use budget boundary into health use limit parameters. The feasible domain construction module is used to determine the network state base value and parameter coefficients corresponding to each constraint based on the state variables and parameter calibration variables of the power grid digital twin, and to construct the feasible domain in combination with the carrying capacity constraint parameters, including transmission power flow constraints, node carrying capacity constraints, regional carrying capacity constraints, transmission and distribution interface switching power range constraints, flexibility ramping constraints, spinning reserve adequacy constraints, frequency response constraints, inertia constraints, and health usage limit constraints. The multi-dimensional load-bearing capacity state calculation module calculates the predicted average value of the dynamic heat capacity of the transmission line based on meteorological forecast data and equipment parameter data, according to the heat balance calculation of the transmission line. and the predicted standard deviation The bearing capacity constraint parameterization module generates dynamic limit parameters according to the uncertainty adaptive reduction rule. The formula is: ； In the formula, Indicates the transmission line number. Indicates the time period number within the rolling scheduling time domain. , Preset non-negative coefficients; Running the feasible domain construction module will dynamically limit parameters Write the upper bound of the power flow constraints in the feasible region; The bearing capacity constraint parameterization module generates the power range parameters for the power transmission and distribution interface. , and parameters of flexibility and climbing ability ,satisfy: ； ； In the formula, , These represent time periods within the rolling scheduling time domain. Time period The power switching setting value of the transmission and distribution interface; The feasible domain construction module runs to exchange the power range parameters of the transmission and distribution interfaces. , Write the power range constraints of the transmission and distribution interface switching in the feasible operating domain, and incorporate the flexibility ramp-up capability parameters. Write flexibility ramp-up constraints into the feasible domain of operation; The rolling scheduling optimization decision layer includes: The model building module is used to receive the feasible domain as the constraint set input and establish a rolling scheduling optimization model coupled across time scales. The rolling scheduling optimization model includes day-ahead planning variables and intraday rolling adjustment variables. The output variable introduction module is used to introduce the controllable dispatch output variable of the new energy power station as a decision variable in the rolling scheduling optimization model, and to set the output boundary constraints of the controllable dispatch output variable of the new energy power station. The consistency constraint module is used to construct cross-timescale consistency constraints in the rolling scheduling optimization model, including unit ramping constraints, minimum start-up and shutdown constraints, energy storage state of charge (SOC) evolution constraints, and transmission and distribution interface switching power consistency constraints. The transmission and distribution interface switching power consistency constraints are used to limit time periods within the same rolling scheduling time domain. Power switching settings of the transmission and distribution interface satisfy: In the formula, For rolling scheduling of time periods within the time domain The planned power switching capacity of the transmission and distribution interface is as follows. For rolling scheduling of time periods within the time domain Intraday rolling adjustment volume; The scenario reduction module is used to perform scenario reduction based on the set of candidate scenarios with uncertainties in new energy output and the sensitivity index of safety constraints, forming a scenario set for online solution; The robust rolling solution module is used to perform rolling solution on the rolling scheduling optimization model under the constraints of the scenario set, and output the scheduling scheme corresponding to the daily plan and the intraday rolling scheduling plan. The controllable scheduling output variable of the new energy power station is the time period of each new energy power station within the rolling scheduling time domain. Controllable scheduling output The output boundary constraint is the upper limit of the carrying capacity output of the new energy power station determined based on the node carrying capacity boundary parameters and / or the regional carrying capacity boundary parameters. and the time period of rolling dispatch of new energy power stations Predictable available output Set together and satisfy: ; The security constraint sensitivity index Determine using the following formula: ； In the formula, Candidate scenarios Time period within the rolling scheduling time domain Corresponding to the Class constraint utilization For the first Preset warning thresholds for class constraints, For the first Weights of class constraints; This represents the total number of constraint types included in the calculation within the feasible domain. The scenario reduction module sorts the scenarios from largest to smallest according to the security constraint sensitivity index and selects the top N candidate scenarios to form the scenario set.

2. The capacity-adaptive dispatching system for high-proportion renewable energy power grids according to claim 1, characterized in that, The online security verification and feedback layer include: The verification feature construction module is used to receive the scheduling scheme, construct and output a security verification feature vector, wherein the security verification feature vector includes at least: the state variables of the power grid digital twin and the power switching settings of the transmission and distribution interfaces in the scheduling scheme. With controllable and scalable power output And the utilization rate of at least one type of constraint in the feasible domain; The proxy verification module is used to perform online security verification based on the security verification feature vector and the proxy model, and output the verification result and uncertainty measure. and initial values ​​of bearing capacity constraint parameters; The high-fidelity verification module is used to verify the uncertainty measure. The high-fidelity verification calculation based on the power grid digital twin is triggered according to the preset uncertainty threshold, and the high-fidelity verification result and the verification correction amount of the bearing capacity constraint parameter are output. The high-fidelity verification calculation includes one or more of AC power flow verification and N-1 verification. The constraint calibration module is used to generate corrected bearing capacity constraint parameters based on the verification results, the high-fidelity verification results and the verification correction amount of bearing capacity constraint parameters when high-fidelity verification calculation is triggered; and to generate corrected bearing capacity constraint parameters based on the verification results and the initial values ​​of the bearing capacity constraint parameters when high-fidelity verification calculation is not triggered. The module writes the corrected bearing capacity constraint parameters into the feasible region construction module to update the feasible region and outputs a re-solution trigger signal to the rolling scheduling optimization model. The instruction decomposition module is used to convert the scheduling scheme into an execution scheduling instruction when the verification result meets the preset pass conditions, and if a high-fidelity verification calculation is triggered, the high-fidelity verification result also meets the preset pass conditions. Under the conditions of satisfying the power range constraints and flexibility ramp-up constraints of the transmission and distribution interface, the module decomposes the power setting value of the transmission and distribution interface in the execution scheduling instruction according to the preset decomposition rules. It is decomposed into distributed power supply power commands, adjustable load power commands, and energy storage charging and discharging power commands. The feedback event module is used to receive the operation feedback data of the execution scheduling instruction, write the operation feedback data into the power grid digital twin to update the multidimensional carrying capacity state variables and the operational feasible domain, and calculate the execution deviation and prediction error distribution drift based on the operation feedback data. When the execution deviation exceeds a preset deviation threshold or the prediction error distribution drift exceeds a preset drift threshold, a re-solution trigger signal is output to the rolling scheduling optimization model.

3. The capacity-adaptive dispatching system for high-proportion renewable energy power grids according to claim 2, characterized in that, The proxy verification module uses an integrated proxy model, which is integrated into the proxy model by at least The system consists of several sub-models, which are used to infer the security verification feature vectors to obtain the first sub-model. Class bearing capacity constraint parameters in time period Predicted initial values ​​of bearing capacity constraint parameters The initial value of the bearing capacity constraint parameter is defined as the integrated mean value output by the proxy. ,Right now ,in Number the sub-models; And measure the uncertainty Defined as: ; When the uncertainty measure High-fidelity verification calculation is triggered when the uncertainty exceeds a preset threshold; otherwise, it is not triggered.

4. The capacity-adaptive dispatching system for high-proportion renewable energy power grids according to claim 3, characterized in that, The feedback event module will time period Execution deviation Defined as the switching power setting value of the transmission and distribution interface. The difference between the actual power exchange interface switching power measured in the operational feedback data and the actual power exchange interface switching power. : ; And the drift of the prediction error distribution Define as a scrolling window Mean of the prediction error series of internal new energy sources with standard deviation Offset relative to the reference window: ; Among them, the prediction error of new energy , For the time period measured in the operation feedback data The actual output of the new energy power station , These represent the mean and standard deviation of the new energy prediction error sequence within the preset benchmark window, respectively. These are preset non-negative weighting coefficients.

5. A capacity-adaptive scheduling method for high-proportion renewable energy power grids, based on the capacity-adaptive scheduling system for high-proportion renewable energy power grids as described in any one of claims 1-4, characterized in that, The method includes: The power grid operation data and forecast data are acquired to construct a power grid digital twin that can be updated on a rolling basis. In the power grid digital twin, the power grid carrying capacity is modeled as a multi-dimensional carrying capacity state variable, the multi-dimensional carrying capacity state variable is mapped to carrying capacity constraint parameters, and an operational feasible domain is constructed to describe the boundary of safe operation of the power grid. Under the condition of uncertainty in new energy output, a rolling scheduling optimization model coupled across time scales is established, and the rolling scheduling optimization model is solved to generate a scheduling scheme that satisfies the operational feasible domain constraint and the cross-time scale consistency constraint. The system receives the scheduling scheme and the state variables of the power grid digital twin. Based on the power grid digital twin, it uses a proxy model coupled with the rolling scheduling optimization model to perform online safety verification of the scheduling scheme, outputs the verification results and the corrected carrying capacity constraint parameters, updates the operational feasible region based on the corrected carrying capacity constraint parameters, and triggers the rolling scheduling optimization model to resolve within the updated operational feasible region. When the verification passes, the scheduling scheme is converted into an execution scheduling command and issued, and the operational feedback data of the execution scheduling command is obtained. The operational feedback data is written into the power grid digital twin to update the multidimensional carrying capacity state variables and the operational feasible region.

Images