Multi-level fault strategy intelligent decomposition method and system based on federated learning and block chain

Abstract

The invention discloses a multi-level fault strategy intelligent decomposition method and system based on federated learning and a block chain. The method comprises the steps of fault data acquisition and preprocessing; generating a whole network fault strategy driven by federated learning; strategy decomposition and decision collaboration based on federated learning and a block chain; a real-time adjustment and feedback mechanism; the system comprises a data acquisition module, a fault diagnosis and disposal module and a feedback mechanism module. By introducing the federated learning technology, intelligent decomposition and decision-making cooperation of a multi-level power grid system are realized, a fault processing strategy can be automatically generated and refined, it is ensured that each power grid level executes specific operation according to local characteristics, manual intervention is avoided, the real-time performance and accuracy of fault response are improved, and the fault processing efficiency is improved. The adaptability and the automation degree of the system are improved through a real-time adjustment and feedback mechanism, efficient, safe and stable operation of power grid fault handling is achieved, and dependence on manual operation is reduced.

Description

Technical Field

[0001] This invention relates to the field of power grid fault handling technology, and in particular to a multi-level fault strategy intelligent decomposition method and system based on federated learning and blockchain. Background Technology

[0002] With the continuous expansion of power grid scale and the increasing complexity of its operation, traditional power grid fault handling methods face many challenges. Currently, power grid fault handling relies on human experience and fixed operating procedures, especially in complex and dynamically changing power grid environments, where existing methods exhibit significant shortcomings. First, traditional methods lack dynamic adaptability and cannot adjust processing strategies in real time, especially when the power grid topology changes or new equipment is connected, resulting in significant response delays. Second, multi-level coordination in the current power grid system also has problems; insufficient information sharing between different levels of the power grid system leads to delayed responses in multi-level fault situations, affecting overall processing efficiency and accuracy. Furthermore, existing intelligent auxiliary decision-making technologies have limitations. Some intelligent auxiliary decision-making systems based on large language models have emerged, but these existing models still suffer from insufficient accuracy and content randomness when facing complex power grid fault scenarios. These models typically require significant human intervention for fine-tuning, especially when facing new equipment and fault modes, lacking efficient adaptive capabilities.

[0003] Chinese patent CN117709461A discloses a method, system, device, and medium for generating auxiliary decision-making for power grid fault handling. This method is based on a large power language model with a Transformer architecture. It generates auxiliary decision-making prompts for power grid fault handling by inputting power dispatch fault handling information and query questions. This method optimizes the large power language model through training aligned with human feedback, making it more adaptable and accurate in handling power grid faults. However, this method still has problems. First, there is the limitation of relying on offline training. Although the decision system based on the large language model improves the generation speed, it relies on offline training and static rules, which cannot quickly respond to the dynamic changes in power grid faults, resulting in a response lag. Second, there is the limitation of human feedback alignment. Although the human feedback alignment mechanism can improve decision accuracy, it still relies on human intervention. In complex and rapidly changing power grid fault scenarios, it may not be able to adapt in a timely manner, affecting the decision-making effect. Summary of the Invention

[0004] Purpose of the invention: The purpose of this invention is to provide a multi-level fault strategy intelligent decomposition method and system based on federated learning and blockchain; to address the shortcomings of existing power grid fault handling technologies, particularly in fault diagnosis, strategy decomposition, cross-level collaboration, and real-time adjustment.

[0005] Technical Solution: The intelligent decomposition method for multi-level fault strategies based on federated learning and blockchain described in this invention includes the following steps:

[0006] (1) Fault data acquisition and preprocessing;

[0007] (2) Federated learning-driven generation of network-wide fault policies;

[0008] (3) Policy decomposition and decision collaboration based on federated learning and blockchain;

[0009] (4) Real-time adjustment and feedback mechanism.

[0010] Furthermore, step (1) fault data acquisition includes collecting real-time operating data and historical fault data from equipment and sensors at different levels of the power grid.

[0011] Furthermore, step (2) includes determining local models, local model aggregation, and generating network-wide fault strategies.

[0012] Furthermore, determining the local model includes:

[0013] Traditional rule base: Based on power grid operation rules and emergency response plans, it simulates power grid fault handling;

[0014] Deep learning models, such as convolutional neural networks or recurrent neural networks, are trained using historical big data to predict power grid equipment failures and load fluctuations.

[0015] Furthermore, the local model aggregation includes:

[0016] Through federated learning, local models from multiple power grid levels are aggregated to generate a unified fault diagnosis and handling strategy model for the entire network. Each level uploads its locally trained model parameters instead of directly transmitting raw data. The central server performs a weighted average of the model parameters from each level and generates a fault diagnosis and handling strategy shared across the entire network. The aggregation formula for the local models is as follows:

[0017]

[0018] Where wl represents the local model parameters of the Lth level, Nl is the amount of data at the Lth level, and N is the total amount of data in the entire network. global It is the aggregated whole network model.

[0019] Furthermore, step (3) includes strategy decomposition, blockchain-based collaborative decision execution, and collaborative decision-making and execution.

[0020] Furthermore, the strategy decomposition includes automatic completion based on the generated network-wide strategy and through learning from local data at each power grid level. Each level will generate detailed execution steps suitable for the local environment. The strategy decomposition formula is as follows:

[0021]

[0022] Among them, S global (t) represents the network-wide unified fault policy, S l (t) represents the local fault strategy at level l, N l Let N be the data volume of the l-th level, and N be the total data volume of the entire network.

[0023] Furthermore, the blockchain-based collaborative decision-making execution includes automatically executing specific strategies at each level through smart contracts in the blockchain, without human intervention. The execution process of the blockchain smart contract is represented by the following formula:

[0024] C l (t)=F SmartContract (S l (t), device_action)

[0025] Among them, C l (t) is the smart contract executed at time t in the l-th level, F SmartContract It is the execution function of the smart contract, S l (t) is the fault policy at level l, and device_action is the device operation command triggered by the smart contract.

[0026] Furthermore, the collaborative decision-making and execution, including strategy decomposition and the collaboration of blockchain technology, enables different levels of the power grid to work together efficiently. The formula for collaborative decision-making and execution is expressed as follows:

[0027]

[0028] Among them, S exec (t) represents the total fault strategy implemented by each level of the power grid, λ l The weight of the execution step at each level, S l (t) represents the local fault strategy at level l.

[0029] The multi-level fault strategy intelligent decomposition system based on federated learning and blockchain described in this invention includes a data acquisition module, a fault diagnosis and handling module, and a feedback mechanism module. The data acquisition module collects status and operation data from different levels of power grid equipment in real time and performs preprocessing. Then, using federated learning technology, based on the local data of each power grid level, the fault diagnosis and handling module automatically generates fault diagnosis and handling strategies for the entire network. Finally, through the feedback mechanism module, the power grid system can adjust the fault strategy of the entire network in a timely manner according to dynamic changes, ensuring that the power grid can respond quickly and maintain its stable and efficient operation.

[0030] Beneficial Effects: Compared with existing technologies, this invention has the following significant advantages: By introducing federated learning technology, this invention achieves intelligent decomposition and decision-making collaboration in multi-level power grid systems. It can automatically generate and refine fault handling strategies, ensuring that each power grid level executes specific operations according to local characteristics, avoiding manual intervention and improving the real-time performance and accuracy of fault response. Blockchain technology further ensures the transparency and security of information sharing. Through decentralized and tamper-proof mechanisms, it improves the trustworthiness and execution efficiency of the system, ensuring efficient collaboration between different power grid levels and reducing human error and information inconsistency. In addition, the real-time adjustment and feedback mechanism improves the adaptability and automation of the system, achieving efficient, safe, and stable operation of power grid fault handling and reducing reliance on manual operation. Attached Figure Description

[0031] Figure 1 This is a flowchart of the present invention;

[0032] Figure 2 A flowchart is generated for the network-wide fault policy based on federated learning. Detailed Implementation

[0033] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0034] like Figure 1 As shown, the intelligent decomposition method for multi-level fault policies based on federated learning and blockchain described in this invention includes the following steps:

[0035] 1. Fault Data Acquisition and Preprocessing

[0036] Fault data acquisition and preprocessing are fundamental to power grid fault diagnosis systems. First, real-time operational data (such as current, voltage, and frequency) and historical fault data are collected from equipment and sensors at different levels of the power grid (grid level, provincial level, and prefecture level). This data provides essential information for subsequent model training and fault diagnosis. To ensure data consistency and high quality, multi-level data preprocessing is crucial.

[0037] The first step in preprocessing is noise reduction, which involves smoothing the data using methods such as moving averages, removing outliers and noise, and ensuring that the data accurately reflects the operating trend of the power grid.

[0038] Secondly, missing value imputation is a common approach, typically using the mean or median to fill in missing data, ensuring data integrity. Regarding data standardization, converting the data to a zero-mean, unit-variance format ensures that dimensional differences between different features do not affect subsequent analysis.

[0039] Finally, normalization scales the data to a uniform range, typically [0, 1], to prevent features with excessively large differences in magnitude from unnecessarily affecting model training. These steps ensure the consistency and high quality of power grid data, providing a reliable data foundation for fault diagnosis and strategy generation.

[0040] 2. Federated Learning-Driven Generation of Network-wide Fault Response Strategies

[0041] Federated learning-driven network-wide fault strategy generation utilizes distributed learning methods to automatically generate a fault diagnosis and handling strategy adapted to the entire power grid, based on data from different power grid levels (such as grid-level, provincial-level, and prefecture-level). This process protects data privacy through federated learning technology while ensuring that different levels of the power grid system can collaboratively generate a consistent fault handling plan. The flowchart for federated learning-based network-wide fault strategy generation is shown below. Figure 2 As shown,

[0042] (1) Determine the local model

[0043] Local models are models trained using local data from various levels of the power grid (e.g., grid-level, provincial-level, prefecture-level). Each power grid level trains its model based on its unique data characteristics and equipment status to generate fault diagnosis and handling strategies suitable for its local environment. The goal of a local model is to accurately diagnose fault types and provide targeted solutions by learning from historical and real-time data at its level. Common types of local models include:

[0044] 1) Traditional rule base: Based on power grid operation rules and emergency response plans, simulates power grid fault handling.

[0045] 2) Deep learning models: such as convolutional neural networks (CNN) or recurrent neural networks (RNN), which are trained using historical big data to predict power grid equipment failures, load fluctuations, etc.

[0046] The local models are trained based on the data characteristics and fault history of the hierarchical devices. Through these local models, the power grid can effectively analyze data from different levels and perform real-time fault diagnosis.

[0047] (2) Local model aggregation

[0048] Through federated learning, local models from multiple power grid levels are aggregated to generate a unified fault diagnosis and handling strategy model for the entire network. Each level uploads its locally trained model parameters instead of directly transmitting raw data, thus avoiding data leakage. The central server performs a weighted average of the model parameters from each level and generates a network-wide shared fault diagnosis and handling strategy. The aggregation formula for the local models is as follows:

[0049]

[0050] Where wl represents the local model parameters of the Lth level, Nl is the amount of data at that level, and N is the total amount of data in the entire network. l This is the aggregated network model. Through this weighted aggregation method, each power grid level contributes different impacts based on its data volume, ensuring that the generated network model is representative and can effectively handle fault data from different power grid levels.

[0051] (3) Generation of network-wide fault strategies

[0052] After the entire network model is aggregated, the resulting unified fault diagnosis strategy will be used for fault diagnosis and handling at each level of the power grid. Based on real-time power grid data (such as voltage, current, and equipment status), the entire network model will automatically generate fault diagnosis and handling strategies adapted to the current state of the power grid. In this way, the power grid system can automatically identify faults and quickly provide solutions.

[0053] The network-wide model can not only generate fault handling strategies, but also dynamically adjust and optimize them according to real-time changes in the power grid. For example, when the power grid topology changes or new equipment is connected, the network-wide model can automatically adapt and generate corresponding fault response strategies to ensure the stable operation of the power grid.

[0054] 3. Policy decomposition and decision collaboration based on federated learning and blockchain

[0055] The execution of power grid fault handling strategies requires effective coordination among multiple levels of the power grid system (such as grid-level, provincial-level, and prefecture-level) to ensure information sharing and consistency and efficiency in execution. A strategy decomposition and decision-making collaboration technology based on federated learning and blockchain aims to decompose the fault strategies generated across the entire network according to the characteristics of each level and ensure collaborative execution between levels. Federated learning, through distributed learning methods, ensures that each power grid level can share models and generate strategies suitable for local needs without disclosing privacy data. Blockchain technology ensures the transparency and security of decision execution, reduces human intervention, and improves the efficiency of system execution.

[0056] (1) Strategy decomposition

[0057] After the overall network fault strategy is generated, it needs to be decomposed according to the characteristics of each power grid level. Different levels of power grid systems (such as grid level, provincial level, and prefecture level) may have different equipment, load demands, and fault handling capabilities. Therefore, the fault strategy needs to be adjusted and refined according to these different characteristics to ensure that each level of the power grid can make the most suitable response based on the local conditions.

[0058] Within the framework of federated learning, policy decomposition is automatically completed based on the generated network-wide policy, through learning from local data at each power grid level. Each level generates detailed execution steps suitable for its local environment. For example, the network-level policy includes overall load adjustment and backup power dispatch, while provincial and prefecture-level policies further refine these steps, implementing them down to the fault handling of specific equipment. The policy decomposition formula is as follows:

[0059]

[0060] Among them, S global (t) represents the network-wide unified fault policy, S l (t) represents the local fault strategy at level l, N l The data volume is at level l. N is the total data volume of the entire network.

[0061] (2) Blockchain-based collaborative decision-making and execution

[0062] Blockchain technology plays a crucial role in ensuring coordinated execution across all levels of the power grid. After the overall network strategy is broken down, blockchain provides a decentralized mechanism to ensure that the execution process between different grid levels is transparent and tamper-proof. This not only guarantees the fairness of decision-making but also improves the efficiency of the execution process.

[0063] Smart contracts in the blockchain enable the automated execution of specific strategies at each level without human intervention. For example, after a fault handling plan is released at the grid level, provincial and municipal systems automatically receive and refine the execution steps. Ultimately, through the blockchain's security mechanisms, specific operations such as fault isolation and equipment recovery are executed on field equipment. Smart contracts ensure that the execution strategies of each level of the power grid proceed as expected, and the blockchain provides transparent records, preventing any potential tampering.

[0064] The execution process of a blockchain smart contract can be represented by the following formula:

[0065] C l (t)=F SmartContract (S l (t), device_action)

[0066] Among them, C l (t) is the smart contract executed at time t in the l-th level, FSmartContract S represents the execution function of a smart contract. l (t) represents the fault policy at level l, and device_action is the device operation command triggered by the smart contract. This formula guarantees the consistency between the policy and the device operation, ensuring that the fault handling execution steps can be executed accurately across different levels of the power grid system.

[0067] (3) Collaborative decision-making and execution

[0068] The synergy between strategy decomposition and blockchain technology enables efficient collaboration across different levels of the power grid. Once the overall grid strategy is decomposed and transmitted via blockchain, each level of the grid executes specific actions based on local conditions. Through blockchain, the execution process at each level of the power grid is transparent and tamper-proof, ensuring synchronized execution and consensus, thus preventing the escalation of faults and system delays.

[0069] The formula for collaborative decision-making and execution is expressed as follows:

[0070]

[0071] Among them, S exec (t) represents the total fault strategy implemented by each level of the power grid, λ l The weight of the execution step at each level, S l (t) represents the local fault strategy at level l. This weighted mechanism ensures that the execution steps at different levels are coordinated and unified, ultimately guaranteeing the stability of the power grid and the effectiveness of fault handling.

[0072] 4. Real-time adjustment and feedback mechanism

[0073] During power grid fault handling, the operating state of the power grid is dynamically changing and may be affected by various factors such as changes in grid topology and load fluctuations. Therefore, ensuring that the power grid can adjust its fault response strategy in a timely manner according to these changes and maintain grid stability is crucial to the entire power grid fault handling system. The core objective of the real-time adjustment and feedback mechanism is to optimize the entire network fault strategy based on real-time data flow and system feedback, ensuring that it automatically adapts to changes in the power grid and generates corresponding fault response strategies.

[0074] (1) Adaptive adjustment of the whole network model

[0075] After fault data is collected and preprocessed in real time, the power grid's overall fault strategy model (a unified model generated based on federated learning) will automatically adjust according to the feedback from the real-time data. Through real-time feedback, the power grid system can dynamically optimize fault diagnosis and handling strategies and respond quickly to changes in the power grid. For example, when the power grid topology changes, the overall model can automatically adapt and calculate the optimal fault isolation and recovery steps to ensure stable system operation.

[0076] The adjustment formula for the entire network model is as follows:

[0077]

[0078] θ t This represents the model parameters at time step t, where η is the learning rate. It is the gradient of the loss function with respect to the model parameters, X. t yt is the real-time data at time step t, and yt is the actual fault label at time step t.

[0079] By dynamically updating the model, the power grid system can adaptively adjust its fault handling strategy based on fault information.

[0080] (2) Feedback learning and adaptive optimization

[0081] The feedback learning mechanism utilizes a combination of real-time and historical data to continuously optimize fault handling strategies and improve the power grid system's response capability when faults occur. Each time the system receives new real-time data, the power grid model automatically updates based on the new feedback and adjusts the fault response strategy.

[0082] In this process, feedback learning adjusts the model parameters by minimizing the loss function, enabling the system to respond more quickly and accurately to similar future failures. The specific optimization process can be represented by the following formula:

[0083]

[0084] Among them, S adjusted (t) is the adjusted fault response strategy, S global (t) represents the network-wide unified fault policy, and α is the adjustment coefficient. The gradient of the strategy adjustment represents the optimization direction of the fault response strategy, S. i (t) is the fault policy at the i-th level, X i (t) represents real-time power grid data, and N represents the number of power grid levels.

Claims

1. A multi-level fault policy intelligent decomposition method based on federated learning and blockchain, characterized in that, Includes the following steps: (1) Fault data acquisition and preprocessing; (2) Federated learning-driven generation of network-wide fault policies; (3) Policy decomposition and decision collaboration based on federated learning and blockchain; (4) Real-time adjustment and feedback mechanism.

2. The intelligent decomposition method for multi-level fault strategies based on federated learning and blockchain according to claim 1, characterized in that, Step (1) Fault data acquisition includes collecting real-time operating data and historical fault data from equipment and sensors at different levels of the power grid.

3. The intelligent decomposition method for multi-level fault strategies based on federated learning and blockchain according to claim 1, characterized in that, Step (2) includes determining the local model, local model aggregation, and generating the network-wide fault strategy.

4. The intelligent decomposition method for multi-level fault strategies based on federated learning and blockchain according to claim 3, characterized in that, The determination of the local model includes: Traditional rule base: Based on power grid operation rules and emergency response plans, it simulates power grid fault handling; Deep learning models, such as convolutional neural networks or recurrent neural networks, are trained using historical big data to predict power grid equipment failures and load fluctuations.

5. The intelligent decomposition method for multi-level fault strategies based on federated learning and blockchain according to claim 3, characterized in that, The local model aggregation includes: Through federated learning, local models from multiple power grid levels are aggregated to generate a unified fault diagnosis and handling strategy model for the entire network. Each level uploads its locally trained model parameters instead of directly transmitting raw data. The central server performs a weighted average of the model parameters from each level and generates a fault diagnosis and handling strategy shared across the entire network. The aggregation formula for the local models is as follows: Among them, w l N represents the local model parameters at level L. l Let L be the amount of data at level L, N be the total amount of data in the entire network, and w be the amount of data at level L. global It is the aggregated whole network model.

6. The intelligent decomposition method for multi-level fault strategies based on federated learning and blockchain according to claim 1, characterized in that, Step (3) includes strategy decomposition, blockchain-based collaborative decision execution, and collaborative decision-making and execution.

7. The intelligent decomposition method for multi-level fault strategies based on federated learning and blockchain according to claim 6, characterized in that, The strategy decomposition includes automatic completion based on the generated network-wide strategy and through learning from local data at each power grid level. Each level will generate detailed execution steps suitable for the local environment. The strategy decomposition formula is as follows: Among them, S global (t) represents the network-wide unified fault policy, S l (t) represents the local fault strategy at level l, N l Let N be the data volume of the l-th level, and N be the total data volume of the entire network.

8. The intelligent decomposition method for multi-level fault strategies based on federated learning and blockchain according to claim 6, characterized in that, The blockchain-based collaborative decision-making execution includes automatically executing specific strategies at each level through smart contracts in the blockchain, without human intervention. The execution process of the blockchain smart contract is represented by the following formula: C l (t)=F SmartContract (S l (t),device_action) Among them, C l (t) is the smart contract executed at time t in the l-th level, F SmartContract It is the execution function of the smart contract, S l (t) is the fault policy at level l, and device_action is the device operation command triggered by the smart contract.

9. The intelligent decomposition method for multi-level fault strategies based on federated learning and blockchain according to claim 6, characterized in that, The collaborative decision-making and execution, including strategy decomposition and the collaboration of blockchain technology, enables different levels of the power grid to work together efficiently. The formula for collaborative decision-making and execution is expressed as follows: Among them, S exec (t) represents the total fault strategy implemented by each level of the power grid, λ l The weight of the execution step at each level, S l (t) represents the local fault strategy at level l.

10. A multi-level fault policy intelligent decomposition system based on federated learning and blockchain, characterized in that, It includes a data acquisition module, a fault diagnosis and handling module, and a feedback mechanism module. The data acquisition module collects real-time status and operation data from different levels of power grid equipment and performs preprocessing. Then, using federated learning technology, based on the local data of each power grid level, the fault diagnosis and handling module automatically generates fault diagnosis and handling strategies for the entire network. Finally, through the feedback mechanism module, the power grid system can adjust the fault strategy of the entire network in a timely manner according to dynamic changes, ensuring that the power grid can respond quickly and maintain its stable and efficient operation.

Images