A Real-Time Optimized Dispatch System and Method for Power Grids Based on Digital Twins

Abstract

This invention discloses a real-time power grid optimization scheduling system and method based on digital twins, belonging to the field of power grid scheduling technology. By deploying sensors to collect environmental parameters of key nodes in real time, a dynamic environmental condition coefficient is constructed. This coefficient is combined with frequency stability and relative strength index analysis to determine the power grid status. A multi-model fusion prediction mechanism is established to achieve accurate spatiotemporal prediction of load and generation. A multi-objective optimization model is adopted with "maximizing renewable energy consumption + minimizing scheduling costs" as the core objectives. A genetic algorithm is introduced to solve for the optimal scheduling scheme, and feasible solutions are screened using forward simulation of the digital twin model. Through anomaly warning triggering, environmental correlation analysis, and model iterative optimization, the scheduling strategy is dynamically adjusted to improve power supply efficiency and emergency response capabilities under extreme scenarios. This method effectively integrates multi-source heterogeneous data, achieving real-time perception of power grid operating status, collaborative optimization of multiple energy resources, and adaptive iteration of the scheduling model.

Description

Technical Field

[0001] This invention belongs to the field of power grid dispatching technology, specifically relating to a real-time optimized dispatching system and method for power grids based on digital twins. Background Technology

[0002] With the large-scale integration of new energy sources (photovoltaics, wind power, etc.) into the power grid, traditional power grid dispatch faces challenges such as the high proportion of fluctuating power sources connected to the grid, increased uncertainty in load demand, and more complex equipment operating environments. Existing dispatch methods mostly rely on static models or single prediction algorithms, making it difficult to perceive the dynamic impact of environmental parameters (such as temperature, solar radiation, etc.) on grid load / generation in real time, resulting in problems such as insufficient frequency stability, limited absorption of new energy sources, and delayed response of reserve resources.

[0003] While digital twin technology has been initially applied in the power grid sector (e.g., equipment condition monitoring and fault prediction), existing solutions largely focus on modeling historical or current data, lacking the ability to predict disturbances over future time intervals, and failing to deeply integrate the correlation analysis of multi-source heterogeneous data (e.g., environmental parameters, renewable energy output, and load fluctuations). Furthermore, traditional optimization algorithms (e.g., linear programming and particle swarm optimization) are prone to getting trapped in local optima when dealing with multi-objective coordinated scheduling (maximizing renewable energy consumption + minimizing cost), making it difficult to meet the comprehensive requirements of new power systems for real-time performance, economy, and reliability.

[0004] Therefore, there is an urgent need for a real-time power grid optimization and scheduling method based on digital twins. This method should dynamically integrate environmental parameters, multi-model prediction (spatiotemporal dual dimensions), intelligent optimization algorithms (genetic algorithms combined with digital twin simulation), and adaptive model iteration mechanisms to achieve early warning of power grid operation anomalies, multi-energy collaborative scheduling, and dynamic strategy optimization, thereby improving the absorption capacity of new energy sources, frequency stability, and power supply efficiency. This invention proposes a real-time power grid optimization and scheduling system and method based on digital twins. Summary of the Invention

[0005] In order to overcome the shortcomings and deficiencies of the existing technology, the first objective of this invention is to provide a real-time power grid optimization scheduling system based on digital twins; the second objective of this invention is to provide a real-time power grid optimization scheduling method based on digital twins.

[0006] The first objective of this invention is achieved through the following technical solution:

[0007] A real-time power grid optimization and dispatching system based on digital twins includes:

[0008] Digital twin module: used to build a high-fidelity virtual image of the power grid, supporting full-process simulation and deduction;

[0009] Environmental perception and anomaly early warning module: used for multi-parameter fusion-based quantitative assessment and early warning of environmental risks;

[0010] Power grid status analysis module: used for multi-dimensional quantitative assessment of power grid stability;

[0011] Intelligent forecasting module: used for accurate load and power generation forecasting through spatiotemporal fusion;

[0012] Multi-energy collaborative scheduling module: used for real-time scheduling decision generation for global optimization;

[0013] Closed-loop optimization module: Continuously improves scheduling strategies based on feedback.

[0014] The second objective of this invention is achieved through the following technical solution:

[0015] A method for real-time optimal scheduling of power grids based on digital twins is described below for use in a real-time optimal scheduling system for power grids based on digital twins.

[0016] S1: Power Grid Operation Anomaly Early Warning: Collect environmental parameters of key power grid nodes and calculate environmental condition coefficients; based on the anomaly threshold and duration of the environmental condition coefficients, trigger graded early warnings and call digital twin models to simulate risks;

[0017] S2: Power Grid Status Analysis: Calculate the power grid frequency stability and relative strength index (RSI) to determine frequency stability anomalies; when the RSI deviates from the threshold, calculate the abnormal correlation between frequency stability and environmental condition coefficients to trigger pre-adjustment of scheduling strategies;

[0018] S3: Power Generation and Load Forecasting: Employs a multi-model fusion approach to generate spatiotemporal dual-dimensional load forecasts; based on the deviation between the forecasts and actual power generation capacity, triggers standby unit pre-start or energy storage emergency preparation;

[0019] S4: Multi-energy coordinated scheduling: Construct an objective function that maximizes new energy consumption and minimizes scheduling costs, and solve it by combining power balance and equipment constraints; use a genetic algorithm to generate scheduling schemes, and screen feasible solutions through digital twin forward simulation;

[0020] S5: Scheduling Model Optimization: Define power supply efficiency coefficients to evaluate scheduling effectiveness; achieve iterative optimization of the model based on knowledge graphs, transfer learning, and genetic algorithm weight adjustments.

[0021] Preferably, the calculation method for the power grid environmental condition coefficient is as follows: ;

[0022] in, This is the temperature reference standard value; and These are the maximum and minimum temperatures within the monitoring period, respectively. This is the reference standard value for solar radiation; and These represent the maximum and minimum values ​​of solar radiation during the monitoring period, respectively. This is a temperature weighting coefficient, emphasizing the impact of load-side temperature on the power grid; This is the solar radiation weighting coefficient, which emphasizes the impact of fluctuations in renewable energy output on the power grid. Improvement in load-sensitive scenarios Improvement in new energy scenarios .

[0023] Preferably, the abnormal warning triggering logic is as follows:

[0024] Preset abnormal threshold for environmental condition coefficients (For example, in high-temperature and high-load scenarios, the value is set to 0.8). When the conditions are met for three consecutive monitoring cycles... hour:

[0025] A red alert is triggered, sending an equipment overload risk warning to the dispatch center;

[0026] The system automatically retrieves digital twin models of the warning area to simulate the impact of continuous load growth on power grid topology, power flow distribution, and voltage stability.

[0027] Preferably, the method for calculating the power grid frequency stability and the relative strength index (RSI) is as follows:

[0028] Sampling points were set up on the bus at each voltage level to calculate frequency stability in real time. : ;in, For real-time frequency; The rated frequency;

[0029] Statistics on the number of frequency increase periods in the most recent 24 monitoring periods With the number of downward cycles ,calculate: ;when or At that time, the frequency stability was determined to be abnormal. and For a preset threshold, and .

[0030] Preferably, the method for calculating the anomaly correlation degree and pre-adjusting the scheduling strategy is as follows:

[0031] When RSI deviates from the normal threshold range At that time, calculate frequency stability Environmental condition coefficient abnormal correlation : ;

[0032] in, The frequency stability of the i-th cycle; The environmental condition coefficient for the i-th period; The mean environmental condition coefficients under the baseline environmental pressure level; This represents the average frequency stability. The number of periods used when calculating the degree of abnormal correlation;

[0033] like The pre-adjustment mechanism for the scheduling strategy includes advance allocation of reserve capacity and pre-optimization of unit output curves; among which, This indicates the threshold for strong correlation.

[0034] Preferably, the multi-model fusion method is used to generate spatiotemporal dual-dimensional load forecast values ​​as follows:

[0035] Digital twin-based prediction generates the first predicted value. Based on the digital twin model of the power grid, real-time operating data including renewable energy output, energy storage SOC, and load distribution are input, and load prediction values ​​for each node are generated through graph neural networks.

[0036] ARIMA-corrected predictions generate a second prediction value. The historical load data is differentially processed to construct an ARIMA(p,d,q) model, outputting short-term load correction values ​​and corrected forecast values. : ;in, This is the correction amount for the ARIMA model; These are the weighting coefficients. Based on dynamic adjustment of prediction accuracy of two models, a balance is achieved between digital twin and time trend prediction ARIMA;

[0037] When detected This indicates that the load forecast is lower than the actual power generation capacity, posing a risk of load shortfall. It is necessary to verify whether this is due to an underestimation of load or an overestimation of power generation.

[0038] Initiate the preheating procedure for the standby conventional generating unit; send an emergency discharge preparation command to the energy storage system, among which... The current available power generation capacity of the power grid includes new energy output, traditional unit backup, and available discharge of energy storage.

[0039] when When this occurs, it indicates that the actual load exceeds the predicted load, resulting in insufficient power generation. This triggers the standby unit to preheat and discharge energy storage, quickly supplementing the power deficit. This represents the current available power generation capacity.

[0040] Preferably, multi-energy coordinated dispatch includes:

[0041] Objective function: ;in, This is the fuel cost item for traditional generating units, used to minimize the economic cost of traditional energy power generation; This is the charging and discharging cost item for the energy storage system, used to balance the peak shaving and valley filling functions of energy storage with operating costs; This is a penalty factor for the renewable energy consumption rate, used to maximize renewable energy consumption. For new energy power, Total power generation;

[0042] Constraints: Power balance: ; , , These are respectively the power of new energy sources, the power of energy storage, and the power of traditional generating units;

[0043] Equipment constraints: ; ;in, For new energy power; For conventional unit power; This refers to the power range of traditional generator units, with the lower limit being the minimum technical output and the upper limit being the rated power. This represents the energy storage power range; a negative sign indicates charging, a positive sign indicates discharging, and the upper limit is the maximum charging and discharging power.

[0044] Scheduling scheme generation: The objective function is solved using a genetic algorithm, with chromosomes encoding the output ratio of each power source. Combined with forward simulation using a digital twin model, feasible solutions with no line overload and qualified voltage are selected, and a three-dimensional scheduling scheme is output, including the scheduling amount of new energy sources, the energy storage charging and discharging plan, and the start-up and shutdown strategy of traditional units.

[0045] Preferably, the scheduling model optimization includes:

[0046] Evaluate power supply efficiency coefficient : ; ;in, The actual power supply to region i; Predict the electricity demand for region i; The total number of regions participating in the power supply efficiency coefficient assessment;

[0047] when Optimization schemes for similar scenarios are extracted from the power grid dispatch knowledge graph. The actual power supply to region i; Predict the electricity demand for region i:

[0048] Transfer learning is applied to the load prediction module of the digital twin model to enhance the recognition of features in extreme scenarios;

[0049] Adjust the weights of the fitness function in the genetic algorithm.

[0050] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0051] 1. This invention quantifies load / generation side risks (such as air conditioning surges and photovoltaic fluctuations caused by high temperatures) through an environmental perception module, and combines digital twin simulation to achieve minute-level early warning of anomalies (such as triggering a red warning when the coefficient exceeds the threshold for three consecutive cycles), and provides the dispatch center with a visualized risk evolution path (such as simulation of the overload impact on equipment due to continuous load growth), significantly shortening the time for manual judgment.

[0052] 2. This invention is based on spatiotemporal fusion prediction (GNN spatial modeling + ARIMA time correction), which reduces load forecasting errors and dynamically adjusts the renewable energy absorption rate and energy storage charging and discharging strategies in conjunction with the supply and demand balance model. This reduces the curtailment rate during peak photovoltaic power generation periods, decreases the number of start-ups and shutdowns of traditional units, and lowers the annualized dispatch cost (through joint optimization of fuel cost and energy storage loss).

[0053] 3. This invention links scheduling effect evaluation with a knowledge graph, enabling the system to automatically extract optimization strategies from historical extreme scenarios (such as "high load, low renewable energy" conditions), and enhance the extreme feature recognition capability of the digital twin model through transfer learning. In scenarios of sudden load surges, the speed of scheduling scheme generation is improved after model iteration, and the voltage qualification rate is increased (through dynamic adjustment of fitness weights using a genetic algorithm). Attached Figure Description

[0054] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art 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.

[0055] Figure 1 A block diagram of the real-time power grid optimization and dispatching system based on digital twins of the present invention is shown.

[0056] Figure 2 A flowchart of the real-time power grid optimization scheduling method based on digital twins of the present invention is shown;

[0057] Figure 3 A flowchart of the power grid state analysis process of the present invention is shown. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more exemplary embodiments. Numerous specific details are provided in the following description to give a full understanding of exemplary embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more of the specific details omitted, or other methods, components, steps, etc., can be employed. In other instances, well-known structures, methods, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this disclosure. Example

[0060] See Figure 1 As shown, the real-time power grid optimization and dispatching system based on digital twins in this embodiment includes...

[0061] Digital twin module: used to build a high-fidelity virtual image of the power grid, supporting full-process simulation and deduction.

[0062] Environmental perception and anomaly early warning module: used for multi-parameter fusion-based quantitative assessment and early warning of environmental risks.

[0063] Power Grid Status Analysis Module: Used for multi-dimensional quantitative assessment of power grid stability.

[0064] Intelligent forecasting module: used for accurate load and power generation forecasting through spatiotemporal integration.

[0065] Multi-energy collaborative scheduling module: used for generating real-time scheduling decisions for global optimization.

[0066] Closed-loop optimization module: Continuously improves scheduling strategies based on feedback.

[0067] The beneficial effects of this embodiment are as follows: the system realizes high-precision simulation and real-time extrapolation of the power grid through digital twins, and enhances risk prevention and control capabilities by combining environmental perception and anomaly early warning; multi-dimensional state analysis ensures power grid stability; intelligent prediction optimizes resource allocation, collaborative scheduling promotes the consumption of new energy sources, and closed-loop optimization continuously improves strategies, thereby comprehensively improving the safety, economy and environmental protection of power grid operation. Example

[0068] See Figure 2 As shown in the figure, the real-time power grid optimization and scheduling method based on digital twins in this embodiment has the following process:

[0069] Step 1: Early warning of abnormal power grid operation.

[0070] S11. Power grid environmental condition detection and coefficient generation.

[0071] Deploy sensors at key nodes of the power grid (substations / power plants / load centers) to collect environmental parameters affecting power grid operation in real time:

[0072] Load-related parameters: temperature (affects air conditioning load) and humidity (affects equipment insulation performance) in industrial / residential areas.

[0073] Power generation related parameters: solar radiation intensity in photovoltaic fields, wind speed / direction in wind farms (affecting the output of new energy sources).

[0074] Calculation of power grid environmental coefficient: The environmental condition coefficient is constructed by linearly normalizing the temperature T and solar radiation intensity S (mapping to the [0,1] interval). ;

[0075] in, This is the temperature reference standard value; and These are the maximum and minimum temperatures within the monitoring period, respectively. This is the reference standard value for solar radiation; and These represent the maximum and minimum values ​​of solar radiation during the monitoring period, respectively. This is a temperature weighting coefficient, emphasizing the impact of load-side temperature on the power grid (such as the risk of equipment overload caused by a surge in air conditioning load). This is the solar radiation weighting coefficient, which emphasizes the impact of fluctuations in renewable energy output on the power grid (such as frequency / voltage fluctuations caused by sudden changes in photovoltaic power). (Improvement in load-sensitive scenarios) New energy scenarios are being improved. ).

[0076] S12, Set the abnormal warning trigger logic.

[0077] Anomaly warning triggering logic: preset environmental coefficient Abnormal threshold (For example, in high-temperature and high-load scenarios, the value is set to 0.8). This applies to three consecutive monitoring cycles. hour:

[0078] A red alert is triggered, sending an equipment overload risk warning to the dispatch center;

[0079] The system automatically retrieves the digital twin model of the area to simulate the impact of continuous load growth on the power grid.

[0080] Step 2: Power grid status analysis.

[0081] See Figure 3 As shown, the power grid state analysis process is as follows:

[0082] S21. Calculation of power grid frequency stability and relative strength index.

[0083] Sampling points were set up on the bus at each voltage level to calculate frequency stability in real time. : ;

[0084] in, For real-time frequency; This is the rated frequency.

[0085] Relative Strength Index (RSI) calculation:

[0086] Statistics on the number of frequency increase periods in the most recent 24 monitoring periods With the number of downward cycles ,calculate: ;

[0087] when or When this occurs, the frequency stability is determined to be abnormal.

[0088] S22, Calculation of abnormal correlation degree.

[0089] When the RSI exceeds the normal range (30-70), calculate the frequency stability. Environmental condition coefficient abnormal correlation : ;

[0090] in, The frequency stability of the i-th period (time series data, reflecting the frequency fluctuation trajectory); This represents the average frequency stability. The number of periods used when calculating the degree of abnormal correlation; The environmental condition coefficient for the i-th period (from step one, reflecting the environmental pressure on the load / generation side, such as high temperature and new energy fluctuations); This represents the average environmental condition coefficient (baseline environmental pressure level).

[0091] like This indicates a strong correlation threshold (frequency anomalies are directly related to environmental pressures, such as high-temperature loads causing frequency drops, or large-scale photovoltaic power generation causing frequency increases), triggering a pre-adjustment mechanism for the scheduling strategy.

[0092] Step 3: Power generation and load forecasting.

[0093] S31, Multi-model fusion prediction.

[0094] Digital Twin Basic Prediction (First Prediction Value) Based on the digital twin model of the power grid, real-time operating data (new energy output, energy storage SOC, load distribution) are input, and load prediction values ​​for each node are generated through graph neural network (GNN).

[0095] ARIMA revised forecast (second forecast value) ): Differentiate historical load data (time series) to construct an ARIMA(p,d,q) model, outputting short-term load correction values ​​and corrected forecast values. : ;

[0096] in, This is the correction amount for the ARIMA model; The weighting coefficients (dynamically adjusted based on the prediction accuracy of the two models) balance spatial dynamic prediction (digital twin) and temporal trend prediction (ARIMA) to achieve spatiotemporal dual-dimensional fusion.

[0097] When detected This indicates that the load forecast is lower than the actual power generation capacity, posing a risk of load shortfall (it needs to be verified whether the load is underestimated or the power generation is overestimated):

[0098] Initiate the preheating procedure for the standby conventional generating unit; send an emergency discharge preparation command to the energy storage system. Among these, This represents the current available power generation capacity of the power grid (new energy output + traditional unit reserve + available energy storage discharge).

[0099] when (Actual load > forecast, insufficient power generation) triggers standby unit preheating + energy storage discharge preparation to quickly make up for the power deficit.

[0100] Step 4: Multi-energy coordinated scheduling.

[0101] S41. The supply and demand balance objective function aims to "maximize the absorption of new energy sources and minimize scheduling costs." An optimization model is constructed as follows: ;in, This is the fuel cost item for traditional units (such as coal-fired and gas-fired units), used to minimize the economic cost of traditional energy power generation and encourage low-fuel-consumption dispatch strategies (such as prioritizing the use of high-efficiency units). This is the charge and discharge cost item (including equipment wear and tear and lifespan degradation cost) for energy storage systems (such as lithium batteries and pumped hydro storage), used to balance the "peak shaving and valley filling" function of energy storage with operating costs, and to avoid the decline in economic efficiency caused by overcharging and discharging; Penalty item for renewable energy consumption rate ( For new energy power, (Total power generation), used to maximize the consumption of new energy sources.

[0102] The constraints are as follows:

[0103] Power balance: ; , , These are respectively the power of new energy sources, the power of energy storage, and the power of traditional generating units.

[0104] Equipment constraints: ; ;in, For new energy power; This is the power output of a traditional generator unit; This refers to the power range of traditional generating units (the lower limit is the minimum technical output, such as the minimum combustion load required for coal-fired units; the upper limit is the rated power). This refers to the energy storage power range (a negative sign indicates charging, a positive sign indicates discharging, and the upper limit is the maximum charging and discharging power, such as the charge and discharge rate limit of lithium batteries).

[0105] S42. Optimal scheduling scheme generation.

[0106] The objective function is solved using a genetic algorithm, with chromosomes encoded as the output ratio of each power source;

[0107] Forward simulation was performed using a digital twin model of the power grid to screen feasible solutions with no line overload and qualified voltage.

[0108] The output includes a three-dimensional scheduling scheme that includes new energy dispatch volume, energy storage charging and discharging plan, and traditional unit start-up and shutdown strategy.

[0109] After solving the objective function using a genetic algorithm, the final solution is selected through the following steps:

[0110] Feasibility verification: The digital twin module is used to simulate candidate solutions, and solutions that cause line overload or voltage exceedance are eliminated;

[0111] Multi-objective evaluation: Calculate a comprehensive score for the remaining feasible solutions;

[0112] Optimal solution selection: Output the solution with the highest overall score.

[0113] Step 5: Optimize the scheduling model.

[0114] Scheduling effectiveness evaluation metrics:

[0115] Water supply coefficient mapping → power supply efficiency coefficient : ; ;in, The actual power supply to region i; (Predict the electricity demand for region i). The total number of regions participating in the power supply efficiency coefficient assessment.

[0116] Model iterative optimization mechanism:

[0117] when (Power supply efficiency threshold)

[0118] Extract optimization solutions for similar scenarios from the power grid dispatch knowledge graph (such as the strategy of "high load and low new energy scenario → priority discharge of energy storage").

[0119] Transfer learning is applied to the load prediction module of the digital twin model to enhance the recognition of features in extreme scenarios;

[0120] Adjusting the fitness function weights of the genetic algorithm prioritizes the response speed of traditional units.

[0121] The beneficial effects of this embodiment are as follows: This method realizes real-time perception and simulation of the power grid operation status through digital twin technology, and combines multi-model fusion prediction and optimized scheduling algorithms to improve the frequency stability and supply-demand balance of the power grid, enhance the absorption of new energy sources, reduce scheduling costs, improve power supply efficiency and emergency response capabilities, and ensure the safe, economical and efficient operation of the power grid.

[0122] The weighting coefficients of this invention are used to measure the degree of influence of different factors or variables on a certain outcome or decision. The weighting coefficient is defined as the numerical value assigned to each factor when comparing and evaluating multiple factors, reflecting their importance or priority. These weighting coefficients can be determined according to specific circumstances and needs, and are usually jointly formulated and confirmed by professionals or relevant stakeholders. By reasonably setting the weighting coefficients, programs or systems can be helped to make decisions or predictions more accurately.

[0123] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0124] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to specific implementations. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A real-time power grid optimization dispatching system based on digital twins, characterized in that, The system includes: Digital twin module: used to build a high-fidelity virtual image of the power grid, supporting full-process simulation and deduction; Environmental perception and anomaly early warning module: used for multi-parameter fusion-based quantitative assessment and early warning of environmental risks; Power grid status analysis module: used for multi-dimensional quantitative assessment of power grid stability; Intelligent forecasting module: used for accurate load and power generation forecasting through spatiotemporal fusion, specifically: generating the first forecast value based on digital twin-based forecasting. Based on a digital twin model of the power grid, real-time operational data including renewable energy output, energy storage SOC, and load distribution are input, and a graph neural network is used to generate load forecasts for each node; ARIMA-corrected forecasts generate a second forecast value. The historical load data is differentially processed to construct an ARIMA(p,d,q) model, outputting short-term load correction values. The corrected forecast values ​​are... ; ; These are weighting coefficients used to balance spatial dynamics prediction and temporal trend prediction. When detected When this occurs, it indicates that the load forecast is lower than the actual power generation capacity, posing a risk of load shortfall. It is necessary to verify whether this is due to an underestimation of load or an overestimation of power generation: Initiate the preheating procedure for the standby conventional generating units; send an emergency discharge preparation command to the energy storage system, wherein... The current available generating capacity of the power grid includes renewable energy output, traditional unit reserves, and available discharge from energy storage; when When the actual load exceeds the predicted load, insufficient power generation is triggered, and the standby unit is prepared for preheating and energy storage discharge to quickly make up for the power deficit. Multi-energy collaborative scheduling module: used for real-time scheduling decision generation for global optimization; Closed-loop optimization module: Continuously improves scheduling strategies based on feedback.

2. A method for real-time power grid optimization scheduling based on digital twins, used in the real-time power grid optimization scheduling system based on digital twins as described in claim 1, characterized in that, The process of the real-time power grid optimization scheduling method based on digital twins is as follows: S1: Power Grid Operation Anomaly Early Warning: Collects environmental parameters of key power grid nodes and calculates environmental condition coefficients. ; ;in, This is the temperature reference standard value; and These are the maximum and minimum temperatures within the monitoring period, respectively. This is the reference standard value for solar radiation; and These represent the maximum and minimum values ​​of solar radiation during the monitoring period, respectively. This is a temperature weighting coefficient, emphasizing the impact of load-side temperature on the power grid; This is a solar radiation weighting coefficient, emphasizing the impact of fluctuations in renewable energy output on the power grid. Based on the abnormal threshold and duration of environmental condition coefficients, a graded early warning is triggered and a digital twin model is invoked to simulate the risk. S2: Power Grid State Analysis: Calculate the power grid frequency stability and Relative Strength Index (RSI) to determine frequency stability anomalies; when the RSI deviates from the threshold, calculate the abnormal correlation between frequency stability and environmental condition coefficients to trigger pre-adjustment of the scheduling strategy; power grid frequency stability ; ;in, For real-time frequency; The rated frequency; Relative Strength Index (RSI): ;in, , These represent the number of periods with increasing frequency and the number of periods with decreasing frequency over the most recent 24 monitoring periods; abnormal correlation. : ;in, The frequency stability of the i-th cycle; This represents the average frequency stability. The number of periods used when calculating the degree of abnormal correlation; The environmental condition coefficient for the i-th period; This represents the average environmental condition coefficient. S3: Power Generation and Load Forecasting: Employs a multi-model fusion approach to generate spatiotemporal dual-dimensional load forecasts; based on the deviation between the forecasts and actual power generation capacity, triggers standby unit pre-start or energy storage emergency preparation; S4: Multi-energy coordinated scheduling: Construct an objective function that maximizes new energy consumption and minimizes scheduling costs, and solve it by combining power balance and equipment constraints; use a genetic algorithm to generate scheduling schemes, and screen feasible solutions through digital twin forward simulation; S5: Scheduling Model Optimization: Define a power supply efficiency coefficient to evaluate scheduling effectiveness; implement iterative model optimization based on knowledge graphs, transfer learning, and genetic algorithm weight adjustments; power supply efficiency coefficient ; ;in, The actual power supply to region i; (Predict the electricity demand for region i). The total number of regions participating in the power supply efficiency coefficient assessment.

3. The real-time power grid optimization scheduling method based on digital twins according to claim 2, characterized in that, The method for calculating the environmental condition coefficient is as follows: Based on a temperature reference standard, a temperature sensitivity index is constructed by measuring the actual temperature fluctuation range within the monitoring period. Based on a solar radiation reference standard, a power generation fluctuation index is constructed by measuring the actual radiation intensity fluctuation range within the monitoring period. The environmental condition coefficient is obtained by weighted summing of the temperature and solar radiation influence components. .

4. The real-time power grid optimization scheduling method based on digital twins according to claim 2, characterized in that, The abnormal warning triggering logic is as follows: Preset abnormal threshold for environmental condition coefficients When the conditions are met for three consecutive monitoring cycles hour: A red alert is triggered, sending an equipment overload risk warning to the dispatch center; The system automatically retrieves digital twin models of the warning area to simulate the impact of continuous load growth on power grid topology, power flow distribution, and voltage stability.

5. The real-time power grid optimization scheduling method based on digital twins according to claim 2, characterized in that, The calculation method for the power grid frequency stability and the relative strength index (RSI) is as follows: Sampling points were set up on the bus at each voltage level to calculate frequency stability in real time. ; Statistical analysis of the number of frequency increase cycles within the most recent 24 monitoring periods With the number of downward cycles ,calculate ;when or At that time, the frequency stability was determined to be abnormal. and For a preset threshold, and .

6. The real-time power grid optimization scheduling method based on digital twins according to claim 5, characterized in that, The method for calculating abnormal correlation and pre-adjusting scheduling strategy is as follows: When RSI deviates from the normal threshold range At that time, calculate frequency stability Environmental condition coefficient abnormal correlation ;like The pre-adjustment mechanism for the scheduling strategy includes advance allocation of reserve capacity and pre-optimization of unit output curves; among which, This indicates the threshold for strong correlation.

7. The real-time power grid optimization scheduling method based on digital twins according to claim 2, characterized in that, The multi-energy coordinated scheduling includes: Objective function: ;in, This is the fuel cost item for traditional generating units, used to minimize the economic cost of traditional energy power generation; This is the charging and discharging cost item for the energy storage system, used to balance the peak shaving and valley filling functions of energy storage with operating costs; This is a penalty factor for the renewable energy consumption rate, used to maximize renewable energy consumption. For new energy power, Total power generation; Constraints: Power balance: ; , , These are respectively the power of new energy sources, the power of energy storage, and the power of traditional generating units; Equipment constraints: ; ;in, For new energy power; For conventional unit power; This refers to the power range of traditional generator units, with the lower limit being the minimum technical output and the upper limit being the rated power. This represents the energy storage power range; a negative sign indicates charging, a positive sign indicates discharging, and the upper limit is the maximum charging and discharging power. Scheduling scheme generation: The objective function is solved using a genetic algorithm, with chromosomes encoding the output ratio of each power source. Combined with forward simulation using a digital twin model, feasible solutions with no line overload and qualified voltage are selected, and a three-dimensional scheduling scheme is output, including the scheduling amount of new energy sources, the energy storage charging and discharging plan, and the start-up and shutdown strategy of traditional units. After solving the objective function using a genetic algorithm, the final solution is selected through the following steps: Feasibility verification: The digital twin module is used to simulate candidate solutions, and solutions that cause line overload or voltage exceedance are eliminated; Multi-objective evaluation: Calculate a comprehensive score for the remaining feasible solutions; Optimal solution selection: Output the solution with the highest overall score.

8. The real-time power grid optimization scheduling method based on digital twins according to claim 2, characterized in that, The scheduling model optimization includes: Evaluate power supply efficiency coefficient : ; ;in, The actual power supply to region i; Predict the electricity demand for region i; The total number of regions participating in the power supply efficiency coefficient assessment; when At that time, optimization solutions for similar scenarios are extracted from the power grid dispatch knowledge graph, among which, The actual power supply to region i; Predict the electricity demand for region i: Transfer learning is applied to the load prediction module of the digital twin model to enhance the recognition of features in extreme scenarios; Adjust the weights of the fitness function in the genetic algorithm.

Images