Multi-source data analysis intelligent scheduling system for power grid safety guarantee

Abstract

The invention discloses an intelligent scheduling system for power grid safety guarantee based on multi-source data analysis, and the system comprises a space-air-ground sensing layer which is used for collecting meteorological data, equipment state data, and power grid operation data; the multi-source data fusion engine is connected with the air-space-ground sensing layer and is used for carrying out space-time alignment and feature extraction on the multi-source heterogeneous data and generating a power grid panoramic state matrix; the extreme weather decision closed-loop module deduces a disaster propagation path through federal learning based on the power grid panoramic state matrix, and outputs a scheduling control instruction; the power grid immune engine is used for receiving the power grid panoramic state matrix, simulating a hidden fault chain through a digital twinborn body and dynamically generating a protection device correction strategy; and the intelligent execution terminal is used for executing the scheduling control instruction and the protection device correction strategy and feeding back the physical power grid state to the digital twinborn body. Through multi-source data fusion and an intelligent algorithm, power grid panoramic perception, disaster pre-judgment, dynamic protection and strategy optimization are realized, and the safety guarantee capability is improved.

Description

Technical Field

[0001] The present invention relates to the field of power grid security protection, and in particular to an intelligent dispatching system for power grid security assurance based on multi-source data analysis. Background Art

[0002] Dispatching systems are essential for ensuring power grid security. Traditional power grid dispatching and management systems rely primarily on single data sources and static models. These systems face challenges in addressing modern power grid security threats, including insufficient data fusion capabilities, passive responses to extreme weather events, and delayed detection of hidden faults. Traditional monitoring methods struggle to detect equipment hazards early, often revealing faults that are already visible by the time they are discovered, resulting in high disposal costs. Therefore, an intelligent dispatching system for power grid security, leveraging multi-source data analysis, was proposed. Summary of the Invention

[0003] The present invention solves the technical problems existing in the prior art through the following technical solutions, which include:

[0004] The air-space-ground sensing layer is used to collect meteorological data, equipment status data, and power grid operation data;

[0005] The multi-source data fusion engine connects to the air-space-ground perception layer to perform spatiotemporal alignment and feature extraction on multi-source heterogeneous data, generating a panoramic power grid status matrix.

[0006] The extreme weather decision-making closed-loop module uses federated learning to deduce disaster propagation paths based on the grid's panoramic state matrix and outputs dispatch control instructions;

[0007] The grid immunity engine receives the grid's panoramic state matrix, simulates hidden fault chains through digital twins, and dynamically generates protection device correction strategies;

[0008] Intelligent execution terminal, used to execute dispatch control instructions and protection device correction strategies, and feedback the physical grid status to the digital twin;

[0009] The extreme weather decision closed-loop module and the power grid immunity engine interact with each other through a collaborative control interface to determine the device vulnerability index.

[0010] Furthermore, the air-space-ground perception layer includes:

[0011] LiDARs deployed in transmission corridors to monitor micro-terrain wind speed data in real time;

[0012] Passive wireless multi-parameter sensors installed on power equipment to collect vibration, temperature and partial discharge signals;

[0013] Wildfire monitoring data and line ice coverage data obtained through satellite remote sensing and drone inspections;

[0014] The multi-parameter sensor and the laser wind radar share a communication gateway.

[0015] Furthermore, the multi-source data fusion engine performs:

[0016] Use a time-space alignment algorithm to unify the timestamps of PMU microsecond-level data and SCADA second-level data;

[0017] Generate the device vulnerability index through the feature extraction formula:

[0018]

[0019] k1, k2, and k3 are weight coefficients, k1+k2+k3=1, which are fitted by historical disaster data;

[0020] W speed is the normalized wind speed impact value, ranging from [0, 1], where 0 means no wind and 1 means exceeding the design wind speed, obtained by laser wind radar;

[0021] R wind The wind resistance level of the equipment is obtained from the equipment technical parameter library;

[0022] F fire is the wildfire approach factor, ranging from [0,10], where 0 means no risk and 10 means the fire point is less than 500m from the equipment. It is obtained through satellite remote sensing and drone inspections;

[0023] ΔT aging The aging temperature rise value of the equipment is the difference between the current temperature and the initial operating temperature, which is obtained through a passive wireless multi-parameter sensor;

[0024] T threshold is the temperature rise safety threshold, which is obtained from the equipment technical specification library.

[0025] Furthermore, the extreme weather decision closed-loop module includes:

[0026] The federated learning agent submodule is used to train the disaster simulation model locally on the edge control terminals in each region;

[0027] Central aggregation server, used to aggregate model parameters in each region to generate cross-regional cascading failure plans;

[0028] Adaptive scheduling strategy library, used to match the optimal action combination according to the disaster type;

[0029] When the wildfire approach factor exceeds the preset threshold, it outputs a command to reduce the current carrying capacity of adjacent lines and triggers a fire linkage signal;

[0030] When the equipment vulnerability index exceeds the safety limit, a preset load shedding strategy is generated.

[0031] Furthermore, the grid immunity engine includes:

[0032] The real-time simulation unit receives equipment status data from the air-space-ground perception layer and builds a digital twin on the RT-LAB platform;

[0033] Fault propagation analysis unit, used to describe implicit fault chains based on the Petri net model:

[0034]

[0035] in is the set of protected nodes, T is the set of fault propagation events, F is the set of directed arcs, and M0 is the initial identifier;

[0036] A dynamic protection unit is used to generate protection constant corrections through reinforcement learning agents.

[0037] Furthermore, the dynamic protection unit performs:

[0038] Receiving equipment vulnerability index and load mutation rate λ;

[0039] Decision-making through a deep deterministic policy gradient agent to protect the fixed value adjustment:

[0040] a t =μ(s t |θ μ );

[0041] μ is the policy function, θ μ is the policy network parameter

[0042] s t is the state vector, s t =[V index ,λ];

[0043] Send the revised protection setting to the relay protection device.

[0044] Furthermore, the collaborative control interface performs:

[0045] When disaster simulation predicts that a line is threatened by wildfire, a high-risk area identifier is sent to the power grid immunity engine;

[0046] The power grid immunity engine dynamically improves protection sensitivity and expands the monitoring parameter range for lines in high-risk areas.

[0047] Furthermore, the intelligent execution terminal includes: a waterproof and anti-magnetic reinforced chassis and an instruction priority arbitration unit.

[0048] Furthermore, the water-resistant magnetic reinforced chassis has a built-in backup battery and satellite communication module;

[0049] The instruction priority arbitration unit processes parallel instructions according to the following rules:

[0050] Protection action instructions take precedence over load transfer instructions, and load transfer instructions take precedence over equipment parameter adjustment instructions.

[0051] Furthermore, the digital twin optimizes decision-making through a closed-loop verification mechanism:

[0052] Record the actual grid status y after the dispatch instruction is executed real ;

[0053] Calculate the digital twin predicted state y pred With the actual state y real The differences:

[0054]

[0055] Update the federated learning model parameters through the gradient descent algorithm:

[0056]

[0057] θ t is the federated learning model parameter, η is the learning rate, is the gradient of the loss function.

[0058] Compared with the existing technology, the present invention has the following advantages: the intelligent dispatching system for power grid security based on multi-source data analysis collects multi-dimensional data through the air-space-ground perception layer, combines the multi-source data fusion engine to achieve time-space alignment and feature extraction, generates a panoramic state matrix of the power grid, improves the comprehensive perception and comprehensive analysis capabilities of the power grid operation status, deduces the disaster propagation path based on federated learning, and dynamically outputs dispatching control instructions for extreme weather such as wildfires and strong winds, such as adjusting the line current carrying capacity and triggering fire linkage, thereby enhancing the power grid's rapid response and active defense capabilities in response to sudden disasters, and uses digital twins to simulate hidden fault chains through Petri net models. And reinforcement learning dynamically generates protection device correction strategies, identifies equipment vulnerabilities in advance and optimizes protection settings, realizes advanced warning and active immunity of fault hazards, and the extreme weather decision-making module and the power grid immunity engine interact with the equipment vulnerability index to form a collaborative protection mechanism; at the same time, through the digital twin closed-loop verification mechanism, the model parameters are optimized based on the actual power grid status, and the accuracy and reliability of system decisions are continuously improved. The intelligent execution terminal adopts a waterproof and anti-magnetic reinforcement design and a built-in backup power supply to ensure the stability of command execution in extreme environments. The command priority arbitration mechanism ensures that key protection actions are executed first, improving the real-time and safety of power grid control. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0060] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process. However, the protection scope of the present invention is not limited to the following embodiment.

[0061] like Figure 1 As shown, this embodiment provides a technical solution: an intelligent dispatching system for power grid security assurance based on multi-source data analysis, comprising:

[0062] The air-space-ground sensing layer is used to collect meteorological data, equipment status data, and power grid operation data;

[0063] The multi-source data fusion engine connects to the air-space-ground perception layer to perform spatiotemporal alignment and feature extraction on multi-source heterogeneous data, generating a panoramic power grid status matrix.

[0064] The extreme weather decision-making closed-loop module uses federated learning to deduce disaster propagation paths based on the grid's panoramic state matrix and outputs dispatch control instructions;

[0065] The grid immunity engine receives the grid's panoramic state matrix, simulates hidden fault chains through digital twins, and dynamically generates protection device correction strategies;

[0066] Intelligent execution terminal, used to execute dispatch control instructions and protection device correction strategies, and feedback the physical grid status to the digital twin;

[0067] The extreme weather decision closed-loop module and the power grid immunity engine interact with each other through a collaborative control interface to determine the device vulnerability index.

[0068] The air-space-ground perception layer includes:

[0069] LiDARs deployed in transmission corridors to monitor micro-terrain wind speed data in real time;

[0070] Passive wireless multi-parameter sensors installed on power equipment to collect vibration, temperature and partial discharge signals;

[0071] Wildfire monitoring data and line ice coverage data obtained through satellite remote sensing and drone inspections;

[0072] The multi-parameter sensor and the laser wind radar share a communication gateway.

[0073] The multi-source data fusion engine performs:

[0074] Use a time-space alignment algorithm to unify the timestamps of PMU microsecond-level data and SCADA second-level data;

[0075] Generate the device vulnerability index through the feature extraction formula:

[0076]

[0077] k1, k2, and k3 are weight coefficients, k1+k2+k3=1, which are fitted by historical disaster data;

[0078] W speed is the normalized wind speed impact value, ranging from [0, 1], where 0 means no wind and 1 means exceeding the design wind speed, obtained by laser wind radar;

[0079] R wind The wind resistance level of the equipment is obtained from the equipment technical parameter library;

[0080] F fire is the wildfire approach factor, ranging from [0,10], where 0 means no risk and 10 means the fire point is less than 500m from the equipment. It is obtained through satellite remote sensing and drone inspections;

[0081] ΔT aging The aging temperature rise value of the equipment is the difference between the current temperature and the initial operating temperature, which is obtained through a passive wireless multi-parameter sensor;

[0082] T threshold is the temperature rise safety threshold, obtained from the equipment technical specification library;

[0083] The above process solves the time synchronization problem of PMU and SCADA data with different frequencies by unifying the time benchmark of multi-source heterogeneous data, ensuring the spatiotemporal consistency of power grid status analysis, avoiding decision-making bias caused by time dislocation, and quantifying equipment vulnerability risks by combining multi-dimensional parameters such as wind speed, wildfire, and equipment aging temperature rise. The equipment vulnerability index is generated through a weighted formula, and the abstract equipment risk is converted into quantifiable and comparable values, which facilitates the rapid positioning of high-risk equipment. Dynamic risk assessment and early warning access data from laser wind radar, satellite remote sensing, sensors, etc. in real time, and the vulnerability index is dynamically updated to support power grid operation and maintenance personnel to intervene in potential equipment in advance to prevent failures.

[0084] For example, a power transmission line in a mountainous area encounters strong winds and a wildfire occurs nearby.

[0085] The laser wind radar detected a wind speed of 80% of the equipment's design speed;

[0086] Satellite remote sensing shows that the wildfire is only 800 meters away from the line;

[0087] The passive sensor detected that the temperature of a tower equipment increased by 25℃ compared with the initial value.

[0088] Assume that the weight coefficients k1 = 0.4, k2 = 0.5, k3 = 0.1;

[0089]

[0090] System according to V index =4.403 determines that the line equipment has a high risk level, triggers an early warning, and automatically adjusts the sensitivity of the protection device. At the same time, it pushes joint handling suggestions for approaching wildfires and abnormal equipment temperature rise to the operation and maintenance personnel to avoid equipment failures caused by the accumulation of single risks.

[0091] The extreme weather decision-making closed-loop module includes:

[0092] The federated learning agent submodule is used to train the disaster simulation model locally on the edge control terminals in each region;

[0093] Central aggregation server, used to aggregate model parameters in each region to generate cross-regional cascading failure plans;

[0094] Adaptive scheduling strategy library, used to match the optimal action combination according to the disaster type;

[0095] When the wildfire approach factor exceeds the preset threshold, it outputs a command to reduce the current carrying capacity of adjacent lines and triggers a fire linkage signal;

[0096] When the equipment vulnerability index exceeds the safety limit, a preset load shedding strategy is generated;

[0097] Distributed collaboration and cross-regional defense utilize federated learning to train disaster simulation models locally at edge terminals in each region, avoiding the communication bottlenecks and data privacy issues associated with centralized training. A central aggregation server generates cross-regional cascading failure scenarios, enhancing the grid's collaborative defense capabilities against large-scale disasters. Accurately matching disaster types and providing rapid response, the adaptive scheduling strategy library automatically matches the optimal action combination based on the real-time disaster type, shortening decision-making time. For example, in response to wildfire threats, current reduction and firefighting linkage can be simultaneously executed, achieving a closed-loop monitoring, analysis, and control. Tiered early warning and proactive load management set safety thresholds based on equipment vulnerability indexes and disaster parameters. When these thresholds are exceeded, preset load shedding strategies are automatically generated, prioritizing power supply to critical loads and reducing the risk of large-scale power outages.

[0098] If the power grid monitors that wildfires have occurred in multiple locations, the approach factor F of the wildfire in area A is fire =9, that is, the fire point is less than 500 meters away from the equipment, triggering the extreme weather decision closed-loop module to respond.

[0099] Federated learning deduction: Edge terminals in each region train disaster propagation models based on local wildfire historical data. After integration, the central aggregation server predicts that the wildfire may spread to areas B and C, forming a cross-regional chain failure plan, such as adjusting the line flow in areas B and C in advance.

[0100] Adaptive strategy execution: For area A, the system automatically executes: reducing the current carrying capacity of adjacent lines, reducing line heating, and reducing the risk of short circuits caused by wildfires;

[0101] Trigger a fire linkage signal: Send wildfire location information to nearby fire stations to initiate emergency firefighting operations.

[0102] Load management: If the vulnerability index of a substation equipment in area A exceeds the safety limit, the system immediately generates a preset load shedding strategy, prioritizing the shedding of non-critical user loads, such as industrial users, to ensure power supply to important loads such as hospitals and communication base stations.

[0103] The grid immunity engine includes:

[0104] The real-time simulation unit receives equipment status data from the air-space-ground perception layer and builds a digital twin on the RT-LAB platform;

[0105] Fault propagation analysis unit, used to describe implicit fault chains based on the Petri net model:

[0106]

[0107] in is the set of protected nodes, T is the set of fault propagation events, F is the set of directed arcs, and M0 is the initial identifier;

[0108] Dynamic protection unit, used to generate protection value correction through reinforcement learning agent;

[0109] Advanced simulation and early warning of hidden faults builds digital twins of power grid equipment based on the RT-LAB platform, mapping physical equipment states such as temperature, vibration, and discharge signals in real time. Hidden fault chains are simulated through Petri net models to identify potential faults that have not yet become apparent, shifting from post-fault repairs to pre-fault prevention.

[0110] Intelligent generation of dynamic protection strategies utilizes reinforcement learning agents to automatically optimize protection device settings based on equipment vulnerability indexes and operating data, adapting to changes in grid operation modes and avoiding false or refusal protection. It clearly displays the entire process of faults from origin to spread, assisting operation and maintenance personnel in quickly locating the source of faults and formulating isolation strategies.

[0111] For example, the winding insulation of a main transformer in a substation has aged due to long-term operation, resulting in abnormal partial discharge, but the traditional protection device has not yet been triggered.

[0112] The air-space-ground sensing layer collects the transformer's vibration, temperature, and partial discharge signals, transmits them to the real-time simulation unit of the grid immunity engine, and builds a digital twin of the transformer on the RT-LAB platform, dynamically displaying the degree of insulation aging, such as the aging temperature rise value ΔT aging =20℃.

[0113] Hidden fault chain simulation: The fault propagation analysis unit simulates the fault chain based on the Petri net model: initial identification M0: insulation aging node activation;

[0114] Event T: Partial discharge intensifies to insulation breakdown and winding short circuit;

[0115] Directed arc F: represents the propagation path of the fault from insulation aging to abnormal discharge and short circuit fault.

[0116] The system predicts that if no intervention is made, the transformer may trip, affecting power supply to surrounding areas.

[0117] Dynamic protection strategy generation: The dynamic protection unit uses reinforcement learning to calculate that the transformer overcurrent protection setting needs to be temporarily adjusted from 1.2 times the rated current to 1.1 times, and the action time needs to be shortened to 0.5 seconds. At the same time, insulation aging warnings and protection setting optimization suggestions are pushed to operation and maintenance personnel.

[0118] The intelligent execution terminal issues instructions to correct the protection settings, and the digital twin synchronously simulates the fault response effect under the new settings, verifying that the adjusted protection strategy can quickly eliminate the fault before the fault occurs, avoiding expansion into a station-wide power outage.

[0119] The dynamic protection unit performs:

[0120] Receiving equipment vulnerability index and load mutation rate λ;

[0121] Decision-making through a deep deterministic policy gradient agent to protect the fixed value adjustment:

[0122] a t =μ(s t |θ μ );

[0123] μ is the policy function, θ μ is the policy network parameter

[0124] s t is the state vector, s t =[V index ,λ];

[0125] Send the revised protection setting to the relay protection device;

[0126] The equipment vulnerability index and load mutation rate are used as state vectors to comprehensively reflect the equipment risk and grid operating conditions, avoiding the one-sidedness of single parameter decision-making.

[0127] Continuous action space optimization uses the DDPG algorithm, which is a continuous action decision model in reinforcement learning to generate protection constant adjustment quantities, such as continuous numerical corrections to current and voltage thresholds. Compared with traditional discrete strategies, this can achieve more refined protection parameter optimization.

[0128] The network parameters of the adaptive environmental change strategy are continuously trained and updated through grid operation data, enabling the system to automatically adapt to changes in equipment risks in different seasons and load modes, thereby improving the robustness of the protection strategy.

[0129] For example, during the high temperature period in summer, the load of a distribution transformer suddenly increases (load mutation rate λ = 20%), and the aging temperature rise value of the equipment ΔT aging = 28℃, close to the safety threshold of 30℃, resulting in a vulnerability index of V index =6.5, high;

[0130] Dynamic protection unit receives s t =[6.5,0.2]. DDPG agent analysis shows that the transformer currently faces the dual risks of overload and aging, and the overload protection setting needs to be adjusted.

[0131] Strategy function μ(s t |θ μ ) Output protection fixed value adjustment amount a t =-15%, that is, the original overload protection threshold is reduced from 1.1 times the rated current to 0.935 times, so as to trigger the alarm in advance and limit the load growth;

[0132] After the intelligent execution terminal issues the new set value, the transformer triggers an alarm before the load increases further. The operation and maintenance personnel promptly start the backup power supply to avoid the transformer from overheating and burning. At the same time, the system updates the strategic network parameter θ according to the actual load change data. μ , optimize the decision logic for subsequent similar scenarios.

[0133] The collaborative control interface performs:

[0134] When disaster simulation predicts that a line is threatened by wildfire, a high-risk area identifier is sent to the power grid immunity engine;

[0135] The grid immunity engine dynamically improves protection sensitivity and expands the range of monitoring parameters for lines in high-risk areas;

[0136] When disaster simulation predicts that a line is threatened, high-risk area identification is shared in real time, triggering the power grid immunity engine to specifically enhance local defense capabilities and dynamically adjust the accuracy of protection strategies. The power grid immunity engine dynamically improves protection sensitivity and expands the monitoring parameter range based on high-risk area identification, achieving refined protection for specific risk areas, avoiding resource waste caused by indiscriminate adjustment of protection strategies across the entire network, and shortening fault response time. Based on real-time interactive equipment vulnerability index and disaster prediction information, protection strategies are preset for lines in high-risk areas in advance, shortening decision-making delays when disasters occur, and improving the efficiency of power grid emergency response.

[0137] The extreme weather decision closed-loop module predicts through federated learning that a transmission line in a certain mountain area will be threatened by a wildfire, and the wildfire approach factor F fire =8, close to the threshold value of 10.

[0138] The extreme weather decision closed-loop module sends the high-risk area identification of the line to the power grid immunity engine through the collaborative control interface, such as the line ID, the risk type of wildfire, and the predicted threat time window.

[0139] After receiving the identification, the grid immunity engine performs the following actions on the line: Improving protection sensitivity: shortening the second-stage operation time of the distance protection from 0.5 seconds to 0.3 seconds, ensuring rapid fault removal in the event of a short circuit caused by a wildfire;

[0140] Expanded monitoring parameters: Start high-frequency partial discharge monitoring of the line insulators, increasing the frequency from once per minute to once per second, to capture insulation degradation signals caused by high temperatures caused by wildfires in real time.

[0141] Coordinated defense effect: When the subsequent wildfire approaches 400 meters from the line (F fire =10). Because the system had optimized its protection strategy in advance, it quickly tripped the moment a wire burned and short-circuited, isolating the fault point and preventing the wildfire from spreading and causing a cascading tripping of adjacent lines. The power outage was limited to a single line, without affecting power supply to surrounding areas.

[0142] Furthermore, the intelligent execution terminal includes:

[0143] Waterproof and anti-magnetic reinforced chassis with built-in backup battery and satellite communication module;

[0144] The instruction priority arbitration unit processes parallel instructions according to the following rules:

[0145] Protection action instructions take precedence over load transfer instructions, and load transfer instructions take precedence over equipment parameter adjustment instructions.

[0146] Furthermore, the digital twin optimizes decision-making through a closed-loop verification mechanism:

[0147] Record the actual grid status y after the dispatch instruction is executedreal ;

[0148] Calculate the digital twin predicted state y pred With the actual state y real The differences:

[0149]

[0150] Update the federated learning model parameters through the gradient descent algorithm:

[0151]

[0152] θ t is the federated learning model parameter, η is the learning rate, is the gradient of the loss function;

[0153] Based on the difference calculation between the actual grid status data after the execution of the dispatch instruction and the predicted status of the digital twin, the gradient descent algorithm is used to automatically update the parameters of the federated learning model, so that the system can continuously correct the prediction deviation according to the actual operation feedback, improve the accuracy of disaster simulation and dispatching strategies, and the self-evolution capability of the model avoids the prediction failure problem of traditional static models caused by factors such as changes in grid topology and equipment aging. The model iteration is driven by data closed loop, adapting to the dynamic evolution of the grid operating environment and maintaining the reliability of decision-making in the long term. The quantitative evaluation of fault handling effect intuitively reflects the effectiveness of the dispatching strategy in a numerical way, assisting operation and maintenance personnel to quickly locate model defects or strategy deficiencies, and realize the "execution-verification-optimization" management closed loop.

[0154] After the system issues a "reduce current carrying capacity" instruction in response to a wildfire threat on a certain line, it evaluates the effectiveness of the decision through a closed-loop verification mechanism.

[0155] Data collection and difference calculation: The intelligent execution terminal feedbacks the actual grid status: the line current capacity drops from 800A to 600A, and the temperature drops from 75°C to 65°C;

[0156] The digital twin predicts a current of 650A and a temperature of 68°C.

[0157] Calculate the difference:

[0158] Model parameter update: The system uses a gradient descent algorithm to adjust the wildfire-capacity correlation parameters in the federated learning model, making the predicted capacity for subsequent similar scenarios closer to the actual value.

[0159] Verification of optimization effect: When a similar wildfire threat occurred again, the difference between the model's predicted current capacity adjustment and the actual execution effect was reduced to less than 800, and the decision-making accuracy was significantly improved.

[0160] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In the description of the present invention, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

[0161] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0162] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. An intelligent dispatching system for power grid security based on multi-source data analysis, characterized in that: include: The air-space-ground sensing layer is used to collect meteorological data, equipment status data, and power grid operation data; The multi-source data fusion engine connects to the air-space-ground perception layer to perform spatiotemporal alignment and feature extraction on multi-source heterogeneous data, generating a panoramic power grid status matrix. The extreme weather decision-making closed-loop module uses federated learning to deduce disaster propagation paths based on the grid's panoramic state matrix and outputs dispatch control instructions; The grid immunity engine receives the grid's panoramic state matrix, simulates hidden fault chains through digital twins, and dynamically generates protection device correction strategies; Intelligent execution terminal, used to execute dispatch control instructions and protection device correction strategies, and feedback the physical grid status to the digital twin; The extreme weather decision closed-loop module and the power grid immunity engine interact with each other through a collaborative control interface to determine the device vulnerability index.

2. The intelligent dispatching system for power grid security assurance based on multi-source data analysis according to claim 1, characterized in that: The air-space-ground perception layer includes: LiDARs deployed in transmission corridors to monitor micro-terrain wind speed data in real time; Passive wireless multi-parameter sensors installed on power equipment to collect vibration, temperature and partial discharge signals; Wildfire monitoring data and line ice coverage data obtained through satellite remote sensing and drone inspections; The multi-parameter sensor and the laser wind radar share a communication gateway.

3. The intelligent dispatching system for power grid security assurance based on multi-source data analysis according to claim 1, characterized in that: The multi-source data fusion engine performs: Use a time-space alignment algorithm to unify the timestamps of PMU microsecond-level data and SCADA second-level data; Generate device vulnerability index V through feature extraction formula index .

4. The intelligent dispatching system for power grid security assurance based on multi-source data analysis according to claim 3, characterized in that: The extreme weather decision-making closed-loop module includes: The federated learning agent submodule is used to train the disaster simulation model locally on the edge control terminals in each region; Central aggregation server, used to aggregate model parameters in each region to generate cross-regional cascading failure plans; Adaptive scheduling strategy library, used to match the optimal action combination according to the disaster type; When the wildfire approach factor exceeds the preset threshold, it outputs a command to reduce the current carrying capacity of adjacent lines and triggers a fire linkage signal; When the equipment vulnerability index exceeds the safety limit, a preset load shedding strategy is generated.

5. The intelligent dispatching system for power grid security assurance based on multi-source data analysis according to claim 1, characterized in that: The grid immunity engine includes: The real-time simulation unit receives equipment status data from the air-space-ground perception layer and builds a digital twin on the RT-LAB platform; Fault propagation analysis unit, used to describe hidden fault chains based on the Petri net model; A dynamic protection unit is used to generate protection constant corrections through reinforcement learning agents.

6. The intelligent dispatching system for power grid security assurance based on multi-source data analysis according to claim 5, characterized in that: The dynamic protection unit performs: Receiving equipment vulnerability index and load mutation rate λ; Protecting fixed value adjustments through deep deterministic policy gradient agent decisions; Send the revised protection setting to the relay protection device.

7. The intelligent dispatching system for power grid security assurance based on multi-source data analysis according to claim 1, characterized in that: The collaborative control interface performs: When disaster simulation predicts that a line is threatened by wildfire, a high-risk area identifier is sent to the power grid immunity engine; The power grid immunity engine dynamically improves protection sensitivity and expands the monitoring parameter range for lines in high-risk areas.

8. The intelligent dispatching system for power grid security assurance based on multi-source data analysis according to claim 1, characterized in that: The intelligent execution terminal includes: a waterproof and anti-magnetic reinforced chassis and an instruction priority arbitration unit.

9. The intelligent dispatching system for power grid security assurance based on multi-source data analysis according to claim 8, characterized in that: The water-resistant magnetic reinforced chassis has a built-in backup battery and satellite communication module; The instruction priority arbitration unit processes parallel instructions according to the following rules: Protection action instructions take precedence over load transfer instructions, and load transfer instructions take precedence over equipment parameter adjustment instructions.

10. The intelligent dispatching system for power grid security assurance based on multi-source data analysis according to claim 1, characterized in that: The digital twin optimizes decision making through a closed-loop verification mechanism: Record the actual grid status y after the dispatch instruction is executed real ; Calculate the digital twin predicted state y pred With the actual state y real differences; Update the federated learning model parameters through the gradient descent algorithm.

Images