Island operation and data synchronization method for intelligent operation and inspection platform on edge side of wind power plant

Abstract

The invention discloses an island operation and data synchronization method for an intelligent operation and maintenance platform at the edge side of a wind power plant, particularly relates to the technical field of intelligent operation and maintenance of the wind power plant, is applied to an edge server deployed at the local of the wind power plant, and comprises the following steps: S1, normal online monitoring and preparation; s2, autonomous operation in an island mode; s3, network recovery detection; s4, performing intelligent difference synchronization and conflict resolution; and S5, recovering and optimizing the mode. Through deep integration of three technologies of edge side autonomous operation, block chain credible evidence storage and lightweight digital twinborn prediction, it is ensured that the wind power plant can still maintain core inspection, monitoring and safety protection functions under the condition of extreme network interruption, and the operation toughness and reliability of the wind power plant in remote and severe environments are improved; the fault risk is reduced, the trust problem of offline data during synchronization is solved, the high consistency and logic correctness of the final states of cloud and edge data are ensured, and data confusion or loss is avoided.

Description

Technical Field

[0001] This invention relates to the field of intelligent operation and maintenance technology for wind farms, and more specifically, to a method for islanded operation and data synchronization of an intelligent operation and maintenance platform on the edge side of a wind farm. Background Technology

[0002] With the deepening of the "carbon peaking and carbon neutrality" strategy, wind power, as the main force of clean energy, has seen its installed capacity continue to grow rapidly. In order to improve operational efficiency and reduce the levelized cost of electricity, the intelligent and unmanned operation and maintenance of wind farms has become an inevitable trend. At present, the smart wind farm architecture based on "cloud-edge-device" collaboration is the mainstream direction. Among them, the edge-side intelligent operation and maintenance platform located at the local site is responsible for real-time data collection, preprocessing and rapid response, which is the key to reducing cloud load and achieving agile closed loop.

[0003] However, in existing technological practices, this model faces several prominent challenges that severely restrict its reliability and practicality: 1. Strong network dependence and gaps in operation and maintenance under extreme conditions: Wind farms are mostly located in remote areas with harsh natural environments. Communication networks (fiber optic, wireless) are easily interrupted by natural disasters such as strong winds, heavy rain, and blizzards. Most existing edge systems are designed with "online priority". Once the connection with the cloud is lost, its intelligent analysis and task scheduling functions often stagnate or can only execute very simple preset logic. As a result, during the most severe weather, wind farms are in an "operation and maintenance blind spot", and the risks to equipment safety and site safety increase sharply. 2. Offline data has low credibility and its value is questionable after synchronization. During network outages, even if edge devices can cache some data, these data lack effective integrity proof and source authentication when synchronized afterward. The cloud center has difficulty in determining whether these data have been tampered with or whether they accurately correspond to specific devices and time points. This lack of credibility greatly reduces the value of offline data, making it unsuitable for serious fault analysis, responsibility identification, or performance evaluation, thus limiting the comprehensive utilization of data assets. 3. Lack of proactive predictive capabilities and a passive operation and maintenance model. The existing edge computing nodes mainly focus on real-time monitoring and alarms based on fixed rules, which is a passive "perception-response" model. Although complex digital twins and predictive maintenance models may be deployed in the cloud, these models cannot serve the edge side when the network is interrupted. Due to the lack of localized intelligent predictive capabilities, the edge side cannot anticipate the degradation trend of equipment and adjust the operation and maintenance strategy in advance during network isolation, thus missing the golden window for preventing failures. 4. Weak conflict resolution capability: After network recovery, existing systems typically use simple full data overwrite or timestamp overwrite methods for synchronization, which is inefficient and prone to data conflicts. When the cloud and edge have different records of the same device status, there is a lack of a fair, automatic, and verifiable conflict resolution rule. Manual intervention is often required for judgment, which is cumbersome and prone to errors, affecting the accuracy and timeliness of operation and maintenance decisions.

[0004] To address the aforementioned issues, while some studies have explored solutions from individual perspectives such as improving edge computing reliability, applying blockchain for evidence storage, or developing digital twin technology, no technical solution has yet been found that can deeply integrate edge autonomous operation, blockchain trusted evidence storage, and lightweight digital twin prediction to form a complete, self-consistent technical system that can systematically solve the problems of network reliability, data credibility, and operational intelligence simultaneously. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, this invention provides a method for islanded operation and data synchronization of an intelligent operation and maintenance platform on the edge side of a wind farm.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for islanded operation and data synchronization of an intelligent operation and maintenance platform on the edge side of a wind farm, applied to an edge server deployed locally at a wind farm site, comprising the following steps: S1. Routine online monitoring and preparation: Under normal network conditions, maintain communication with the cloud-based operation and maintenance management platform and synchronize the lightweight digital twin model of the equipment; continuously monitor the network link, and trigger mode switching when the network interruption exceeds the preset threshold. S2. Autonomous Operation in Isolated Mode: Switching to isolated mode, computing resources are scheduled according to the pre-set local task priority strategy library to execute core operation and maintenance tasks; at the same time, the lightweight digital twin model is enabled to perform health status simulation and fault prediction for key equipment, and task priorities are dynamically adjusted according to the prediction results; key data generated during operation is stored in a local temporary database, and a hash value of the key data is generated synchronously. The hash value and timestamp are signed using the edge server's private key to form a timestamped signed data block as a proof of storage to be synchronized; S3, Network Recovery Detection: Periodically attempts to restore network connection, and enters synchronization mode after the connection is stable; S4. Intelligent Difference Synchronization and Conflict Resolution: A secure channel is established with the cloud, and the intelligent synchronization engine is activated. The engine identifies and extracts incremental data, status update data, digital twin-derived data, and potential conflict data from the local temporary database. The timestamped signature data block, along with the corresponding key data, is transmitted to the cloud. The cloud verifies the signature validity and data integrity, and then submits the verified signature data block to the blockchain network for notarization. For conflicting data, it is resolved according to preset rules, which prioritize data versions supported by valid blockchain notarization or earlier timestamped signature data blocks. S5. Mode Recovery and Optimization: After data synchronization is complete, switch back to online mode; report the performance data during the isolated operation period for optimization of the task priority strategy library and digital twin model.

[0007] As a further improvement to the technical solution of the present invention, the generation of the timestamped signature data block in S2 specifically includes: aggregating the hash values ​​of multiple key data generated within a preset period using a Merkle tree, signing the root hash value of the Merkle tree and the period timestamp, and forming an aggregated signature data block for that period.

[0008] As a further improvement to the technical solution of the present invention, the lightweight digital twin model in S2 includes a performance degradation model based on the physical characteristics of the equipment and / or a pruned and quantized deep learning prediction model, which is used to predict the remaining useful life (RUL) of the equipment or calculate the probability of failure in islanded mode.

[0009] As a further improvement to the technical solution of the present invention, the local task priority strategy library in S2 defines at least the monitoring tasks involving equipment security and fire risk, the illegal intrusion detection tasks, and the predictive inspection tasks for equipment with rapidly declining health triggered by the digital twin model as the highest priority tasks.

[0010] As a further improvement to the technical solution of the present invention, in step S4, the verified signature data block is submitted to the blockchain network for notarization, specifically to a permissioned consortium blockchain network built for notarization of energy equipment operation and maintenance data.

[0011] As a further improvement to the technical solution of the present invention, the preset conflict data resolution rules in S4 also include: for the same device status data, if neither party has blockchain evidence, then the data with the updated timestamp is adopted; if the timestamps are the same or close, then the data generated by the AI ​​model with higher accuracy is adopted.

[0012] As a further improvement to the technical solution of the present invention, in step S4, after the intelligent difference synchronization and conflict resolution, the method further includes: using real device status data and fault records synchronized from the edge side, the cloud performs feedback training and parameter updates on the main digital twin model in the cloud and the lightweight digital twin model sent to the edge side.

[0013] A smart operation and maintenance platform for the edge of a wind farm, deployed locally at the wind farm site, specifically includes: The status monitoring and mode switching module is used to communicate with the cloud under normal conditions and trigger mode switching when the network is interrupted. The island autonomous operation module is used to schedule tasks according to preset strategies in island mode, call the lightweight digital twin model to predict equipment status, and generate key data and its signature for storage. A network detection module is used to periodically detect network recovery. The intelligent synchronization engine module is used to identify discrepancies in data after network recovery, transmit signed evidence for cloud verification and on-chain storage, and resolve data conflicts according to rules. The configuration optimization module is used to report performance data after synchronization is complete in order to optimize strategies and models.

[0014] This invention achieves the following significant synergistic effects through the deep integration of three technologies: edge-side autonomous operation, blockchain-based trusted evidence storage, and lightweight digital twin prediction: 1. By building a complete autonomous task scheduling and execution system at the edge, the wind farm can maintain core inspection, monitoring and security protection functions even under extreme network interruption conditions. This transforms the traditional operation and maintenance model of "operation and maintenance stop when the network is interrupted" into "intelligent operation and maintenance continues even when the network is interrupted", fundamentally improving the operational resilience and reliability of the wind farm in remote and harsh environments. 2. By deploying lightweight digital twin models to the edge, enabling them to perform real-time health simulations and failure probability predictions for critical equipment in isolated environments, and dynamically adjusting operation and maintenance strategies accordingly, the system has been upgraded from a passive response relying on threshold alarms to proactive intervention and preventative maintenance based on state prediction, significantly improving equipment availability and reducing the risk of major failures. 3. This invention introduces a blockchain anchoring mechanism, providing tamper-proof, verifiable, and traceable trusted evidence for key operation and maintenance data generated locally in an isolated environment. This not only solves the trust problem of offline data synchronization, but also makes these data trusted assets with legal effect and high analytical value, providing a solid basis for fault attribution, insurance claims, and performance evaluation. 4. By adopting intelligent differential synchronization and multi-layer conflict resolution strategies, only incremental and key data are synchronized after the network is restored. Data conflicts are automatically resolved according to rules such as "blockchain evidence priority", which reduces network bandwidth consumption and synchronization time, and ensures a high degree of consistency and logical correctness of the final state of data between the cloud and the edge, avoiding data chaos or loss.

[0015] 5. A complete closed loop of "edge execution - data synchronization - cloud optimization - model update" has been constructed. Real data from isolated operation is fed back to the cloud for continuous optimization of task priority strategies and digital twin models. The optimized models are then distributed to the edge, enabling the entire system to continuously learn and improve itself in practice, and the level of intelligent operation and maintenance continues to improve over time. Attached Figure Description

[0016] Figure 1 This is a flowchart of the method of the present invention.

[0017] Figure 2 This is a schematic diagram of the conflict resolution rules of the present invention. Detailed Implementation

[0018] 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.

[0019] As attached Figure 1-2 The method for islanded operation and data synchronization of an intelligent operation and maintenance platform on the edge side of a wind farm, as shown, is applied to an edge server deployed locally at a wind farm site and includes the following steps: S1. Routine online monitoring and preparation: Under normal network conditions, maintain communication with the cloud-based operation and maintenance management platform and synchronize the lightweight digital twin model of the equipment; continuously monitor the network link, and trigger mode switching when the network interruption exceeds the preset threshold. S2, Autonomous Operation in Isolated Mode: Switching to isolated mode, computing resources are scheduled according to the pre-set local task priority strategy library to execute core operation and maintenance tasks; at the same time, a lightweight digital twin model is enabled to perform health status simulation and fault prediction for key equipment, and task priorities are dynamically adjusted according to the prediction results; key data generated during operation is stored in a local temporary database, and a hash value of the key data is generated synchronously. The hash value and timestamp are signed using the edge server's private key to form a timestamped signed data block as a proof of storage to be synchronized; S3, Network Recovery Detection: Periodically attempts to restore network connection, and enters synchronization mode after the connection is stable; S4. Intelligent Difference Synchronization and Conflict Resolution: Establish a secure channel with the cloud and activate the intelligent synchronization engine; the engine identifies and extracts incremental data, state update data, digital twin-derived data, and potential conflicting data from the local temporary database; transmits the timestamped signature data block along with the corresponding key data to the cloud; the cloud verifies the signature validity and data integrity, and then submits the verified signature data block to the blockchain network for notarization; for conflicting data, it is resolved according to preset rules, which prioritize the data version supported by a valid blockchain notarization or an earlier timestamped signature data block; S5. Mode Recovery and Optimization: After data synchronization is complete, switch back to online mode; report the performance data during the isolated operation period for optimization of the task priority strategy library and digital twin model.

[0020] Furthermore, generating a timestamped signature data block in S2 specifically includes: aggregating the hash values ​​of multiple key data generated within a preset period using a Merkle tree, signing the Merkle tree root hash value and the period timestamp, and forming an aggregated signature data block for that period.

[0021] Furthermore, the lightweight digital twin model in S2 includes a performance degradation model based on the physical characteristics of the equipment and / or a pruned and quantized deep learning prediction model, used to predict the remaining useful life (RUL) of the equipment or calculate the probability of failure in islanded mode.

[0022] Furthermore, the local task priority policy library in S2 defines at least the monitoring tasks involving equipment security and fire risk, illegal intrusion detection tasks, and predictive inspection tasks for devices with rapidly declining health triggered by the digital twin model as the highest priority tasks.

[0023] Furthermore, in S4, the verified signature data block is submitted to the blockchain network for notarization, specifically to a permissioned consortium blockchain network built for notarizing energy equipment operation and maintenance data.

[0024] Furthermore, the default conflict data resolution rules in S4 also include: for the same device status data, if neither party has blockchain evidence, the data with the updated timestamp will be adopted; if the timestamps are the same or close, the data generated by the more accurate AI model will be adopted.

[0025] Furthermore, in S4, after intelligent difference synchronization and conflict resolution, it also includes: using real device status data and fault records synchronized from the edge side to perform feedback training and parameter updates on the main digital twin model in the cloud and the lightweight digital twin model sent to the edge side.

[0026] A smart operation and maintenance platform for the edge of a wind farm, deployed locally at the wind farm site, specifically includes: The status monitoring and mode switching module is used to maintain communication with the cloud-based operation and maintenance management platform under normal network conditions, synchronize the lightweight digital twin model of the equipment, continuously monitor the network link, and trigger mode switching when the network interruption exceeds a preset threshold. The isolated autonomous operation module is used in isolated mode to schedule computing resources according to a pre-set local task priority policy library, execute core operation and maintenance tasks, enable a lightweight digital twin model to perform health status simulation and fault prediction for critical equipment, dynamically adjust task priorities based on prediction results, store key data generated during operation in a local temporary database, synchronously generate hash values ​​for this key data, and sign the hash value and timestamp using the edge server's private key to form a timestamped signed data block as proof of storage to be synchronized; the isolated autonomous operation unit is configured to: handle multiple data generated within a preset period The hash values ​​of key data are aggregated using a Merkle tree, and the root hash value of the Merkle tree and the periodic timestamp are signed to form an aggregated signed data block for that period; the lightweight digital twin model includes a performance degradation model based on the physical characteristics of the equipment and / or a pruned and quantized deep learning prediction model, which is used to predict the remaining useful life of the equipment or calculate the probability of failure in islanded mode; the local task priority policy library defines at least the monitoring tasks involving equipment security and fire risk, illegal intrusion detection tasks, and predictive inspection tasks for equipment with rapidly declining health triggered by the digital twin model as the highest priority tasks; The network detection module is used to periodically attempt to restore the network connection and enter synchronization mode after the connection is stable; The intelligent synchronization engine module establishes a secure channel with the cloud and initiates the intelligent synchronization engine. This engine identifies and extracts incremental data, status update data, digital twin-derived data, and potentially conflicting data from the local temporary database. It then transmits timestamped signed data blocks along with corresponding key data to the cloud. The cloud verifies the signature validity and data integrity, subsequently submitting the verified signed data blocks to a permissioned consortium blockchain network for notarization. For conflicting data, it resolves conflicts according to preset rules, prioritizing data versions with valid blockchain notarization or earlier timestamped signed data blocks. These preset rules also include: for the same device status data, if neither party has blockchain notarization, the timestamped data is adopted; if the timestamps are the same or close, data generated by a more accurate AI model is adopted. The configuration optimization module is used to switch back to online mode after data synchronization is completed; it reports the performance data during the isolated operation period to optimize the task priority strategy library and digital twin model; among them, the cloud uses real device status data and fault records synchronized from the edge to perform feedback training and parameter updates on the main digital twin model in the cloud and the lightweight digital twin model sent to the edge.

[0027] The lightweight digital twin model is as follows: 1. Model building and lightweight methods: The lightweight digital twin model is built through the following two complementary paths: Path 1: Performance degradation model based on physical characteristics; taking the main transformer of a wind farm as an example, this model is built based on its key failure mechanism (such as insulation thermal aging). Specifically, the Arrhenius reaction rate equation is used to describe the aging process of the degree of polymerization (DP) of the insulating paper with operating temperature and time: DP(t) = DP0 * exp(-A * t * exp(-Ea / (k *T))), where DP0 is the initial degree of polymerization, A is the pre-exponential factor, Ea is the activation energy, k is the Boltzmann constant, and T is the absolute operating temperature of the winding (estimated by load and cooling conditions). Through real-time monitoring of load current and ambient temperature data, the model extrapolates the current DP value at the edge and predicts the remaining time (RUL) for it to drop to the fault threshold (e.g., DP=200). This model is encapsulated as a calculation module containing several parametric equations, with low computational complexity and minimal memory usage. Path 2: Deep learning prediction model based on pruning and quantization; For gearbox vibration fault prediction, firstly, a hybrid model of a one-dimensional convolutional neural network (1D-CNN) and a long short-term memory network (LSTM) is trained in the cloud using complete historical vibration time series data. After training, channel pruning is used to remove unimportant convolutional kernels in the network, reducing the model size to 30%-40% of the original version. Subsequently, post-training quantization is used to convert the model's weights and activation values ​​from 32-bit floating-point numbers (FP32) to 8-bit integers (INT8). This process can be automatically completed using edge inference frameworks such as TensorRT and OpenVINO. The quantized model improves inference speed by 2-3 times and reduces memory usage by about 75% while maintaining controllable accuracy loss (e.g., less than 2%). Finally, the generated lightweight model file in .engine or .tflite format is distributed to the edge server. 2. Model Deployment and Invocation: The edge server has a dedicated model container responsible for loading and managing the aforementioned lightweight model. In silo mode, when the task scheduler triggers a predictive task, it calls the model container API. For example, it inputs the gearbox vibration spectrum features of the last 24 hours into the quantized 1D-CNN-LSTM model. The model outputs a probability value (a scalar between 0 and 1) of a specific fault (such as tooth surface pitting) occurring in the next 12 hours. This probability value is compared with a preset threshold as the basis for dynamically adjusting the inspection task.

[0028] The specific implementation of the blockchain-based evidence storage process is as follows: 1. Data signing and generation of blocks to be stored: The edge server has a built-in cryptography module that uses the Elliptic Curve Digital Signature Algorithm (ECDSA) with the secp256k1 curve to generate and securely store its unique public-private key pair. For critical data Di requiring notarization (such as a "main transformer bushing temperature over-limit" alarm), the notarization process is as follows: a. Hash calculation: Calculate the hash value H(Di) of data Di using the SHA-256 algorithm; b. Timestamp Acquisition: Obtain the accurate Coordinated Universal Time (UTC) timestamp Ti from a local trusted clock; c. Data block construction: Concatenate H(Di) and Ti to form the message to be signed, M=H(Di)||Ti; d. Digital signature: Use the private key Priv_Edge of the edge server to perform an ECDSA signature on M, and obtain the signature value Sig(M); e. Generation of evidence storage package: Package the triple {Di,Ti,Sig(M)} as a basic evidence storage unit and store it in the "waiting to be uploaded to the chain cache area"; f. Aggregated Signature: To improve efficiency, every time N basic evidence storage units are accumulated (e.g., N=100) or a time window is reached (e.g., 1 hour), the system constructs a Merkle Tree from all H(Di) and signs the root hash value RootHash of the Merkle Tree and the end timestamp Twindow of the time window to generate an aggregated signature data block Blockagg={RootHash,Twindow,Sig(RootHash||Twindow)}; 2. Cloud-based verification and on-chain evidence storage: The cloud platform pre-stores the public keys of all legitimate edge servers (PubEdge). Upon receiving synchronization data: a. Signature verification: For a single evidence storage unit, the cloud uses PubEdge to verify the received Sig(M) and recalculate H'(Di), comparing whether it is consistent with H(Di), thereby verifying the integrity of the data and the authenticity of the source; b. Submitting to the blockchain: After successful verification, the cloud service calls the chaincode (smart contract) of the consortium blockchain network (such as one built on Hyperledger Fabric with the channel name "wind-farm-maintenance"). This chaincode provides a recordEvidence transaction interface, and the cloud constructs a transaction proposal containing the following key fields: Evidence ID (UUID), Edge Node ID, Data Hash (H(Di) or RootHash), Timestamp (Ti or Twindow), and Signature (Sig(M) or Aggregate Signature). c. On-chain storage: After transactions are sorted by consensus, they are written into blocks. Only the metadata and signature of the evidence are stored on the chain, not the original data Di itself, in order to protect privacy and save storage. After a transaction is successfully executed, a unique transaction ID (TxID) is returned. d. Association and Confirmation: The cloud associates the TxID with the corresponding original data record and stores it in the cloud database. At the same time, the TxID can be sent to the edge server as proof of the completion of the evidence storage. The edge server can then clear the corresponding local cache.

[0029] The conflict resolution rules follow a clear hierarchical decision tree logic, as detailed in the attached diagram. Figure 2 As shown, the rule description is as follows: ΔT is a configurable time tolerance parameter used to handle small time deviations caused by network latency. The confidence level can be predefined using metadata such as model version number, training data volume, and historical accuracy.

[0030] In summary, this invention, through the deep integration of three technologies—edge-side autonomous operation, blockchain-based trusted evidence storage, and lightweight digital twin prediction—ensures that wind farms can maintain core inspection, monitoring, and security functions even under extreme network outage conditions. This enhances the resilience and reliability of wind farms in remote and harsh environments, reduces the risk of major failures, solves the trust problem of offline data synchronization, provides a solid basis for fault attribution, insurance claims, and performance evaluation, and guarantees a high degree of consistency and logical correctness of the final state of cloud and edge data, avoiding data chaos or loss. This enables the entire system to continuously learn and self-improve in practice, and the level of intelligent operation and maintenance continuously improves over time.

[0031] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for islanded operation and data synchronization of an intelligent operation and maintenance platform on the edge side of a wind farm, characterized in that, For edge servers deployed locally at wind farm sites, the following steps are included: S1. Routine online monitoring and preparation: Under normal network conditions, maintain communication with the cloud-based operation and maintenance management platform and synchronize the lightweight digital twin model of the equipment; continuously monitor the network link, and trigger mode switching when the network interruption exceeds the preset threshold. S2. Autonomous Operation in Isolated Mode: Switching to isolated mode, computing resources are scheduled according to the pre-set local task priority strategy library to execute core operation and maintenance tasks; at the same time, the lightweight digital twin model is enabled to perform health status simulation and fault prediction for key equipment, and task priorities are dynamically adjusted according to the prediction results; key data generated during operation is stored in a local temporary database, and a hash value of the key data is generated synchronously. The hash value and timestamp are signed using the edge server's private key to form a timestamped signed data block as a proof of storage to be synchronized; S3, Network Recovery Detection: Periodically attempts to restore network connection, and enters synchronization mode after the connection is stable; S4. Intelligent Difference Synchronization and Conflict Resolution: A secure channel is established with the cloud, and the intelligent synchronization engine is activated. The engine identifies and extracts incremental data, status update data, digital twin-derived data, and potential conflict data from the local temporary database. The timestamped signature data block, along with the corresponding key data, is transmitted to the cloud. The cloud verifies the signature validity and data integrity, and then submits the verified signature data block to the blockchain network for notarization. For conflicting data, it is resolved according to preset rules, which prioritize data versions supported by valid blockchain notarization or earlier timestamped signature data blocks. S5. Mode Recovery and Optimization: After data synchronization is complete, switch back to online mode; report the performance data during the isolated operation period for optimization of the task priority strategy library and digital twin model.

2. The method for islanded operation and data synchronization of an intelligent operation and maintenance platform on the edge side of a wind farm according to claim 1, characterized in that: The generation of timestamped signature data blocks in S2 specifically includes: aggregating the hash values ​​of multiple key data generated within a preset period using a Merkle tree, signing the root hash value of the Merkle tree and the period timestamp, and forming an aggregated signature data block for that period.

3. The method for islanded operation and data synchronization of an intelligent operation and maintenance platform on the edge side of a wind farm according to claim 1, characterized in that: The lightweight digital twin model in S2 includes a performance degradation model based on the physical characteristics of the equipment and / or a pruned and quantized deep learning prediction model, which is used to predict the remaining useful life (RUL) of the equipment or calculate the probability of failure in islanded mode.

4. The method for islanded operation and data synchronization of an intelligent operation and maintenance platform on the edge side of a wind farm according to claim 1, characterized in that: The local task priority policy library in S2 defines at least the monitoring tasks involving equipment security and fire risk, illegal intrusion detection tasks, and predictive inspection tasks for equipment with rapidly declining health triggered by the digital twin model as the highest priority tasks.

5. The method for islanded operation and data synchronization of an intelligent operation and maintenance platform on the edge side of a wind farm according to claim 1, characterized in that: In step S4, the verified signature data block is submitted to the blockchain network for notarization, specifically to a permissioned consortium blockchain network built for notarization of energy equipment operation and maintenance data.

6. The method for islanded operation and data synchronization of an intelligent operation and maintenance platform on the edge side of a wind farm according to claim 1, characterized in that: The conflict data resolution rules preset in S4 also include: for the same device status data, if neither party has blockchain evidence, then the data with the updated timestamp is adopted; if the timestamps are the same or close, then the data generated by the more accurate AI model is adopted.

7. The method for islanded operation and data synchronization of an intelligent operation and maintenance platform on the edge side of a wind farm according to claim 1, characterized in that: In step S4, after the intelligent difference synchronization and conflict resolution, the method further includes: using real device status data and fault records synchronized from the edge side, the cloud performs feedback training and parameter updates on the main digital twin model in the cloud and the lightweight digital twin model sent to the edge side.

8. A smart operation and maintenance platform for the edge side of a wind farm, deployed locally at the wind farm site, operating based on the method described in any one of claims 1-7, characterized in that, Specifically, it includes: The status monitoring and mode switching module is used to communicate with the cloud under normal conditions and trigger mode switching when the network is interrupted. The island autonomous operation module is used to schedule tasks according to preset strategies in island mode, call the lightweight digital twin model to predict equipment status, and generate key data and its signature for storage. A network detection module is used to periodically detect network recovery. The intelligent synchronization engine module is used to identify discrepancies in data after network recovery, transmit signed evidence for cloud verification and on-chain storage, and resolve data conflicts according to rules. The configuration optimization module is used to report performance data after synchronization is complete in order to optimize strategies and models.

Images