Real-time state evaluation method and system for network spare power automatic switching

Abstract

The invention provides a real-time state evaluation method and system for network spare power automatic switching, and relates to the technical field of power distribution networks, and the method comprises the steps: carrying out the prediction based on a multi-source data set collected in real time through a digital twin-driven LSTM model, and loading a corresponding evaluation model according to a prediction result; according to the loaded evaluation model, the initialized threshold parameter and a health degree score set associated with all spare power automatic switching power supplies, reliability evaluation is carried out on the switching path of the feeder line, and a state evaluation report is generated; submitting the state evaluation report to a pre-configured block chain node, executing rule consensus verification through an intelligent contract, and outputting a control instruction of a block chain signature; according to the method, the control instruction is executed, the configuration parameters and the health degree score set of the digital twinning during prediction are iteratively updated according to the execution result, and through multi-source data fusion, dynamic modal prediction, block chain consensus verification and an iterative optimization mechanism, the system self-adaption and intelligence are realized, and the power supply safety and stability of the power distribution network are enhanced.

Description

Technical Field

[0001] This invention relates to the field of power distribution network technology, and in particular to a real-time status assessment method and system for automatic switching of backup power in a network. Background Technology

[0002] In flexible distribution network systems, automatic transfer switching (ATS) devices are crucial for ensuring power supply reliability. With the widespread application of distributed power sources (such as photovoltaics and wind turbines) and energy storage devices, the operating modes of distribution networks are becoming more complex and variable, potentially exhibiting grid-connected or islanded states. Traditional ATS methods are largely based on fixed rules and local information, making them ill-suited to dynamically changing grid environments. Furthermore, the lack of single data sources and multi-source data fusion analysis leads to inaccurate state assessments. In addition, existing methods lack transparency and traceability in the decision-making process, failing to meet the high security and reliability requirements of smart grids.

[0003] Therefore, it is necessary to provide a real-time status assessment method and system for network backup automatic transfer to solve the above-mentioned technical problems. Summary of the Invention

[0004] To address the aforementioned technical issues, this invention provides a real-time status assessment method and system for network backup automatic transfer. Through multi-source data fusion, dynamic modal prediction, blockchain consensus verification, and iterative optimization mechanisms, the system achieves self-adaptation, transparency, and intelligence, thereby enhancing the security and stability of power distribution networks.

[0005] On one hand, the present invention provides a real-time status assessment method for network automatic transfer switching, including at least two automatic transfer power sources, each of which is connected to the feeder of the distribution network through an automatic transfer switching node. The method includes the following steps:

[0006] Based on real-time acquired multi-source datasets, predictions are made using a digital twin-driven LSTM model, and the corresponding evaluation model is loaded based on the prediction results.

[0007] Based on the loaded evaluation model, the initialized threshold parameters, and the health score set of all associated backup power supplies, the reliability of the feeder switching path is evaluated, and a status evaluation report is generated.

[0008] The status assessment report is submitted to a pre-configured blockchain node, and the consensus verification of the execution rules is performed through a smart contract, outputting control instructions with blockchain signatures.

[0009] The control instructions are executed, and the configuration parameters of the digital twin and the health score set are iteratively updated based on the execution results.

[0010] Preferably, the multi-source dataset includes output fluctuation data of each backup automatic transfer power source, electrical quantities of the feeder, and status data of each backup automatic transfer switching node.

[0011] Preferably, the multi-source dataset acquired in real time is used to make predictions using a digital twin-driven LSTM model, and a corresponding evaluation model is loaded based on the prediction results, including:

[0012] The multi-source real-time state dataset is subjected to time series alignment processing to generate a standardized input vector;

[0013] The input vector is input into a pre-trained digital twin-driven LSTM model, which outputs the operating mode probability of the distribution network, wherein the operating mode probability includes grid-connected mode probability and islanded mode probability;

[0014] When the probability of grid connection mode or islanding mode exceeds a preset threshold, the corresponding evaluation model is loaded, wherein the evaluation model includes the corresponding grid connection evaluation model and islanding evaluation model.

[0015] Preferably, the initialization process of the threshold parameter and the health score set includes:

[0016] Historical health scores of each standby power source are extracted from historical blockchain data to form the health score set.

[0017] Initialize the corresponding threshold parameters according to the loaded evaluation model, including:

[0018] If it is a grid-connected evaluation model, the voltage support threshold of the distribution network is initialized to a preset voltage;

[0019] If it is an island assessment model, the initial penetration threshold of the backup power supply is the preset penetration ratio.

[0020] Preferably, the reliability assessment of the switching path of the feeder includes:

[0021] The real-time health of each backup power supply is calculated based on the health score set, and a health anomaly list is generated based on the real-time health. The health anomaly list includes backup power supplies whose real-time health is lower than a preset health threshold.

[0022] Execute the corresponding evaluation strategy based on the type of the evaluation model:

[0023] If it is a grid-connected evaluation model, verify whether the voltage of the feeder of the distribution network has reached the initial voltage support threshold, and calculate the reliability score of the switching path based on the real-time health of each backup power source.

[0024] If it is an island assessment model, verify whether the penetration rate of each backup power supply has reached the initial penetration rate threshold, and calculate the reliability score of the switching path based on the real-time health of each power supply.

[0025] A switching feasibility score is generated by combining the reliability score with the real-time collected electrical quantities of the feeder.

[0026] Based on the switching feasibility score and the list of health anomalies, a switching path reconstruction strategy is generated and integrated to form a status assessment report.

[0027] Preferably, the switching path reconstruction strategy includes at least one of the following:

[0028] The switching priority of each backup power supply is arranged in descending order of real-time health status;

[0029] When the reliability score of the primary switching path is lower than the preset score threshold, the predefined backup switching path is activated.

[0030] Set the closing time difference for different backup automatic transfer power supplies, wherein the closing time difference is dynamically calculated based on the electrical quantities of the feeder.

[0031] Preferably, the consensus verification through smart contract execution rules includes:

[0032] The status assessment report is analyzed to extract the switching feasibility score, health anomaly list, and switching path reconstruction strategy.

[0033] Select the corresponding consensus rule set based on the current operating mode:

[0034] If it is a grid-connected mode, the main network collaborative consensus rule is adopted: the main network connection node of the distribution network needs to sign and confirm, and the switching feasibility score exceeds the preset score;

[0035] If it is an isolated mode, the local node consensus rule is adopted: more than a preset proportion of backup self-investment switching nodes must agree and the switching feasibility score must exceed a preset score;

[0036] A draft control instruction is generated based on the aforementioned switching path reconstruction strategy;

[0037] When the corresponding consensus rules are met, a digital signature of the blockchain is attached to the draft control instruction, and the blockchain-signed control instruction is output.

[0038] Preferably, the iterative update based on the execution result includes:

[0039] Obtain the execution record of the control command, and extract the execution status code and the electrical quantity recovery curve of the feeder;

[0040] Based on the execution status code and the electrical quantity recovery curve of the feeder, the LSTM model weight matrix in the configuration parameters of the digital twin during prediction is corrected.

[0041] Based on the incremental mechanical wear data of each backup automatic transfer node, the health score of the corresponding backup automatic transfer power supply in the health score set is updated and stored.

[0042] On the other hand, the present invention also provides a real-time status assessment system for network automatic transfer switching, used to execute a real-time status assessment method for network automatic transfer switching, including at least two automatic transfer power sources, each automatic transfer power source being connected to the feeder of the distribution network through an automatic transfer switching node, the system comprising:

[0043] The modality prediction loading module is used to make predictions based on real-time acquired multi-source datasets using a digital twin-driven LSTM model, and load the corresponding evaluation model based on the prediction results.

[0044] The path assessment report module is used to assess the reliability of the feeder switching path based on the loaded assessment model, the initialized threshold parameters, and the health score set of all associated backup power supplies, and generate a status assessment report.

[0045] The blockchain verification instruction module is used to submit the status assessment report to a pre-configured blockchain node, verify the rule consensus through smart contract execution, and output control instructions with blockchain signatures.

[0046] The instruction execution iteration module is used to execute the control instructions and iteratively update the configuration parameters of the digital twin and the health score set during prediction based on the execution results.

[0047] Compared with related technologies, the real-time status assessment method and system for network backup automatic transfer provided by the present invention has the following beneficial effects:

[0048] This invention improves the accuracy and real-time performance of state assessment through multi-source data fusion analysis. It utilizes a digital twin-driven LSTM model to predict power grid operating modes and dynamically loads corresponding assessment models, enhancing the system's adaptability. Consensus verification of smart contract execution rules based on blockchain technology ensures the transparency and traceability of the decision-making process, improving system security and reliability. Simultaneously, iterative updates to digital twin configuration parameters and health score sets based on execution results enable continuous system optimization.

[0049] Furthermore, the switching path reconfiguration strategy proposed in this invention comprehensively considers the real-time health of the backup power supply and the electrical quantity of the feeder, effectively improving the reliability of the switching path. Attached Figure Description

[0050] Figure 1 A flowchart of a real-time status assessment method for network backup automatic transfer provided by the present invention;

[0051] Figure 2The present invention provides a module structure diagram of a real-time status assessment system for network backup automatic transfer. Detailed Implementation

[0052] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the drawings, not all structures. Moreover, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0053] It should also be noted that, for ease of description, the accompanying drawings show only the parts relevant to the invention and not all of them. Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe the operations (or steps) as sequential processes, many of the operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but it may also have additional steps not included in the drawings. The process may correspond to a method, function, procedure, subroutine, subprogram, etc.

[0054] Example 1

[0055] The physical architecture of this invention includes a backup power supply layer, a backup power transfer switching node, a distribution network and feeders, and a blockchain node layer. The backup power supply layer comprises multiple photovoltaic inverters, wind turbines, and energy storage devices. It collects power output fluctuation data in real time via edge sensors (≥10Hz update frequency) and provides emergency power (capacity covering 50kW-10MW) to the distribution network feeders through the backup power transfer switching node in the event of a main grid failure. The backup power transfer switching node includes intelligent circuit breakers (such as QF1 / QF2) and fast disconnect switches (operation time ≤200ms), physically connecting the main grid, the local backup power supply, and the distribution network feeders. Simultaneously, it collects switch status, vibration and temperature rise data, and precise operation timestamps in real time via mechanical wear sensors.

[0056] The distribution network and feeders are mainly composed of 110kV / 35kV substations, with power transmission channels (rated capacity ≤20MVA) formed by 10kV / 0.4kV overhead lines or cables. Voltage / current transformers (PT / CT) and frequency measurement units (PMU) are deployed on the feeders to continuously monitor electrical quantities with voltage accuracy of 0.1kV, current accuracy of 1A, and frequency accuracy of 0.01Hz. The blockchain node layer is deployed on the local controller (edge ​​layer) of the backup automatic transfer switching node and the regional cloud platform (regional layer) of the distribution substation. The edge node stores local power / switch data and executes islanded consensus, while the regional node manages the network topology data and coordinates smart contracts. The two work together to realize control command signature verification and historical data storage.

[0057] The digital twin platform runs on GPU server clusters and real-time databases, mapping the impedance matrix of power distribution network nodes 1:1. It carries LSTM prediction models driven by digital twins (input dimension ≥ 50) and dynamically adjusts the model weight matrix, federated learning rules and equipment degradation coefficients to form a closed-loop control from data acquisition (power layer) → state assessment (digital twin) → decision execution (blockchain).

[0058] On one hand, the present invention provides a real-time status assessment method for network automatic transfer switching, including at least two automatic transfer power sources, each of which is connected to the feeder of the distribution network through an automatic transfer switching node. The method includes the following steps:

[0059] S1: Based on real-time collected multi-source datasets, predictions are made using a digital twin-driven LSTM model, and the corresponding evaluation model is loaded based on the prediction results.

[0060] The multi-source dataset includes output fluctuation data of each backup power source, electrical quantities of the feeders, and status data of each backup power source switching node.

[0061] Specifically, step S1 includes the following steps:

[0062] S11: Perform time series alignment processing on the multi-source real-time state dataset to generate a standardized input vector.

[0063] In this embodiment, the core function of this step is to solve the problem of timestamp misalignment caused by the difference in sampling frequency of multi-source data (such as 10Hz for power output fluctuation data and 50Hz for feeder electrical quantity data), and to ensure the spatiotemporal consistency of the subsequent LSTM model input.

[0064] The specific implementation process is as follows: First, using the sampling time of the distribution network feeder voltage as the reference time axis (e.g., one sampling point every 20ms), the photovoltaic / wind turbine output fluctuation data is aligned to this time axis using a cubic spline interpolation algorithm; simultaneously, the discrete state change events of the standby automatic transfer switching nodes (e.g., QF1 tripping) are converted into timestamp-synchronized state vectors (filled with 1s for the state duration and 0s for the remaining time periods); finally, Z-score standardization is performed on all data streams to generate a standardized input vector with dimensions of [time step × number of features]. The technical effect of this step is to unify the three heterogeneous data types—electrical quantities, mechanical states, and output fluctuations—into a computable time series matrix, eliminating dimensional differences and providing a regularized input for modal prediction.

[0065] S12: Input the input vector into a pre-trained digital twin-driven LSTM model and output the operating mode probability of the distribution network, wherein the operating mode probability includes grid-connected mode probability and islanded mode probability.

[0066] In this embodiment, this step uses deep learning to capture the implicit correlation between the power grid's operating status and mode switching. Its core function is to quantify the power grid's operating risks and provide a basis for decision-making in evaluating model selection.

[0067] The specific implementation includes: the LSTM model adopts a three-layer hidden layer structure (128 / 64 / 32 neurons), and the time window length is set to 1 second (50 time steps); the input vector is updated through the forget gate and the input gate control unit state, and the output gate generates the hidden state at the current time step; finally, the fully connected layer is connected to the Softmax activation function to output two probability values.

[0068] The model training uses historical data stored on the blockchain, aggregates the gradients of edge nodes through federated learning, and updates the model weights every 24 hours to ensure prediction accuracy.

[0069] S13: When the grid-connected mode probability or the islanding mode probability exceeds a preset threshold, load the corresponding evaluation model, wherein the evaluation model includes the corresponding grid-connected evaluation model and the islanding evaluation model.

[0070] In this embodiment, this step enables dynamic switching of the evaluation strategy to adapt to the reliability evaluation requirements under different power grid operating conditions.

[0071] The specific implementation process is as follows: Real-time monitoring of the probability value output by the LSTM; if the grid-connected mode probability P... 并网 If the value is ≥0.7, then load the grid-connected evaluation model (including the main grid voltage support threshold determination module and the power output weighted scoring module) from the digital twin platform; if the islanding mode probability P 孤岛If the probability is ≥0.7, then load the islanded evaluation model; for special operating conditions where the probability values ​​are all below 70%, keep the currently activated evaluation model unchanged.

[0072] S2: Based on the loaded evaluation model, the initialized threshold parameters, and the health score set of all associated backup power supplies, perform a reliability assessment on the switching path of the feeder and generate a status assessment report.

[0073] In step S2, the initialization process of the threshold parameter and the health score set includes:

[0074] First, historical health scores for each standby power source are extracted from historical blockchain data to form the health score set.

[0075] In this embodiment, the implementation process is as follows: A smart contract deployed on the blockchain node layer queries the historical health database, extracting minute-by-minute health sampling data for each power source within the last 30 days by power source ID index. A time-degradation weighted average is calculated for each power source, specifically using an exponential decay model to assign weights, with the weight of data from the last 3 hours accounting for 70% (i.e., weights are calculated using an hourly decay coefficient of 0.05, with higher weights for more recent data). Finally, a structured dataset containing the mapping relationship between power source IDs and health scores is generated. For example, key-value pairs such as PV1 (health score 87), WT2 (health score 92), and ESS3 (health score 78) are generated. This step, by highlighting the importance of recent data, eliminates the impact of momentary equipment failures on the assessment, establishing an accurate health benchmark for reliability analysis.

[0076] Secondly, the corresponding threshold parameters are initialized according to the loaded evaluation model, including:

[0077] If it is a grid-connected evaluation model, the voltage support threshold of the distribution network is initialized to a preset voltage;

[0078] If it is an island assessment model, the initial penetration threshold of the backup power supply is the preset penetration ratio.

[0079] In this embodiment, the process of initializing the corresponding threshold parameters according to the type of evaluation model loaded is divided into a dual-path scheme: when the grid-connected evaluation model is activated, the voltage support threshold is dynamically set by calling the distribution network topology parameters, with 0.85 times the rated voltage as the base value, and is adjusted according to the real-time load rate (the threshold is increased to 0.9 times the rated voltage when the load rate exceeds 80%).

[0080] When the islanding assessment model is activated, the penetration rate threshold is calculated based on the percentage of healthy power supply capacity (the ratio of the total power supply capacity with a health score greater than 80 to the total power supply capacity of the entire system multiplied by a baseline coefficient of 0.4), while limiting the final threshold to a range of 30% to 50%. This mechanism adaptively adjusts the voltage threshold in grid-connected mode through load conditions and optimizes the islanding mode penetration rate index using the percentage of healthy power supply capacity, achieving precise matching between assessment parameters and system operating conditions.

[0081] In step S2, the reliability assessment of the switching path of the feeder includes:

[0082] First, the real-time health status of each backup power supply is calculated based on the health status score set, and a health status anomaly list is generated based on the real-time health status, wherein the health status anomaly list includes backup power supplies whose real-time health status is lower than a preset health threshold.

[0083] In this embodiment, based on the initialized health score set (including the historical weighted score of each power source) and combined with the real-time collected mechanical wear increment data (triggered when the vibration amplitude exceeds the baseline value), the real-time health of each standby power source is dynamically calculated. The calculation formula is: Real-time health = Historical health - Wear increment × Material attenuation coefficient, where the material attenuation coefficient for photovoltaic equipment is 0.8, for wind turbines it is 0.6, and for energy storage equipment it is 0.5. When the real-time health score is lower than the preset threshold of 70 points, the power source ID is added to the health anomaly list. By quantifying mechanical wear into a health attenuation index, weak power source nodes can be identified in real time (for example, when the circuit breaker vibration amplitude reaches 50μm, the health score of photovoltaic inverter PV1 drops from 85 points to 77 points).

[0084] Secondly, the corresponding evaluation strategy is executed according to the type of the evaluation model:

[0085] If it is a grid-connected evaluation model, verify whether the voltage of the feeder of the distribution network has reached the initial voltage support threshold, and calculate the reliability score of the switching path based on the real-time health of each backup power source.

[0086] If it is an island assessment model, verify whether the penetration rate of each backup power supply has reached the initial penetration rate threshold, and calculate the reliability score of the switching path based on the real-time health of each power supply.

[0087] In this embodiment, the grid-connected mode involves verifying whether the voltage of the distribution network feeders reaches the initialization threshold (0.85-0.95 times the rated voltage). If the voltage is qualified, the reliability score of the switching path is calculated using the following formula:

[0088]

[0089] Where S is the reliability score of the switching path, and H is... i For the real-time health of the i-th backup power supply, ω i The output weight of the i-th backup power source is allocated according to the power capacity ratio (e.g., if photovoltaic accounts for 60%, the weight is 0.6), and V is the voltage stability adjustment factor (calculated based on the absolute value of the voltage deviation; the smaller the deviation, the higher the factor value).

[0090] Islanding mode: Verify whether the penetration rate of the backup power supply exceeds the initialization threshold (30%-50%). If the penetration rate is qualified, the reliability score of the switching path is:

[0091]

[0092] Among them, β is the penetration rate target adjustment factor (the factor value increases linearly when the penetration rate is in the range of 30%-50%), which can ensure that large-capacity healthy power supply is selected first in grid-connected mode and avoid the chain failure caused by the lowest healthy power supply in islanded mode.

[0093] Next, a switching feasibility score is generated by combining the reliability score with the electrical quantities of the feeder collected in real time.

[0094] In this embodiment, the overall reliability score and real-time feeder electrical quantities (current harmonic distortion rate / frequency deviation) are combined using a weighted fusion algorithm to generate a switching feasibility score: Switching feasibility score = Reliability score × 60% + (100 - Electrical quantity risk value) × 40%.

[0095] The calculation logic for electrical quantity risk values ​​is as follows: when the current distortion rate is greater than 5%, an additional 10 points are added to the risk value for every 1% exceeding 5%; when the frequency deviation is greater than 0.2Hz, an additional 5 points are added to the risk value for every 0.1Hz exceeding 0.2Hz. The technical effect is to avoid high-risk switching operations during voltage fluctuations (for example, a current distortion rate of 7% results in a risk value of 20, reducing the feasibility score by 8%).

[0096] Finally, a switching path reconstruction strategy is generated based on the switching feasibility score and the list of health anomalies, and integrated to form a status assessment report.

[0097] In this embodiment, when the switching feasibility score is higher than 80 and there is no abnormal power supply, the priority policy generation process is triggered.

[0098] The process first extracts the real-time health scores of each backup power source (e.g., ESS1 has a score of 92, PV2 has a score of 88, and WT3 has a score of 85). Then, it strictly sorts the power sources from highest to lowest health score, generating a clear power switching priority sequence (e.g., ESS1 → PV2 → WT3). When a power source is detected with a health score difference exceeding 10 points (e.g., WT3's health score drops to 75, 17 points lower than the highest value), its ranking will be forcibly adjusted to the lowest position. Finally, standardized JSON instructions are output, clearly indicating the operation type and power source execution sequence.

[0099] When the switching feasibility score is below 60 or there is a power source listed in the health anomaly list, the backup path strategy generation mechanism is activated.

[0100] The mechanism first obtains the abnormal power source ID, then queries a pre-stored distribution network topology database containing detailed connection schemes for the primary path and several backup paths (e.g., primary path QF1→QF2→PV2, backup path QF3→QF5→ESS1 with a score of 90). It automatically selects the highest-scoring qualified backup path (e.g., "QF3→QF5→ESS1" with a score of 90) and outputs a standardized structure containing execution instructions, path identifiers, and excluded power sources.

[0101] When the total harmonic distortion (THD) of the feeder current is detected to exceed 10%, the timing adjustment strategy is automatically generated.

[0102] The process first performs a Fast Fourier Transform (FFT) on the feeder current to accurately extract the fundamental frequency (e.g., the fundamental period for a 50Hz system is 0.02 seconds). Then, the closing time difference is set differently based on the power type: energy storage devices are delayed by 1.2 times the fundamental period (24 milliseconds), photovoltaic devices by 0.8 times the fundamental period (16 milliseconds), and wind turbine devices maintain zero delay. Finally, a device-level command with precise timing parameters is output.

[0103] Once the above strategies are generated, the system automatically integrates them into a structured status assessment report, which includes a feasibility score (e.g., 76 points), a list of health anomalies, specific reconstruction strategies (including strategy type, path parameters, and delay intervals), and an ISO 8601 standard timestamp.

[0104] S3: Submit the status assessment report to the pre-configured blockchain node, verify the consensus through smart contract execution rules, and output control instructions with blockchain signatures.

[0105] Specifically, step S3 includes the following steps:

[0106] S31: Parse the status assessment report and extract the switching feasibility score, health anomaly list, and switching path reconstruction strategy.

[0107] In this embodiment, the core function of parsing the status assessment report and extracting key elements is to transform the assessment results into verifiable decision inputs.

[0108] The specific implementation process is as follows: The state assessment report is structured using a JSON parser in the smart contract. First, the digital signature of the report is verified, and after confirming the data integrity, three key elements are extracted:

[0109] 1) Switch the feasibility scoring field;

[0110] 2) An array of lists of individuals with abnormal health status;

[0111] Switch paths and reconstruct the policy object (including policy type / alternate path / time parameters).

[0112] S32: Select the corresponding consensus rule set based on the current operating mode:

[0113] If it is a grid-connected mode, the main network collaborative consensus rule is adopted: the main network connection node of the distribution network needs to sign and confirm, and the switching feasibility score exceeds the preset score;

[0114] If it is an isolated mode, the local node consensus rule is adopted: more than a preset proportion of backup self-investment switching nodes must agree and the switching feasibility score must exceed a preset score.

[0115] In this embodiment, the core function of selecting the consensus rule set based on the current operating mode is to achieve dynamic decision-making mechanism adaptation. The specific implementation involves dual-channel logic:

[0116] When in grid-connected mode, activate the main grid coordination rule: call the distribution network topology interface to identify the main grid connection node (such as QF1 circuit breaker), require the node to sign the instruction draft with its private key, and at the same time, the switching feasibility score must be ≥80 points;

[0117] In islanded mode, the local node initiation rules are as follows: Calculate the number of backup self-connection switching nodes participating in consensus (e.g., 5), with a preset ratio of 2 / 3, meaning at least 4 nodes must agree, and the score must be at least 70. Through modal adaptive mechanism, the safety of the large power grid is ensured in grid-connected mode, and distributed autonomous decision-making is achieved in islanded mode.

[0118] S33: Generate a draft control instruction based on the switching path reconstruction strategy.

[0119] In this embodiment, the core function of generating a draft control instruction based on the reconfiguration strategy is to complete the conversion of device-level executable instructions. The specific implementation process is as follows: Identify the strategy type → Map device operation parameters.

[0120] Priority strategy: Convert the power supply sequence into a circuit breaker action sequence (QF5_CLOSE→QF6_CLOSE);

[0121] Alternate path strategy: Query the topology library by path ID and generate switch instructions;

[0122] Timing strategy: Convert delay parameters into converter control commands;

[0123] The final assembly is a draft of structured instructions.

[0124] S34: When the corresponding consensus rules are met, attach a digital signature of the blockchain to the draft control instruction and output the control instruction with the blockchain signature.

[0125] In this embodiment, the core function of blockchain consensus verification and signature output is to enable the issuance of authoritative instructions that are tamper-proof.

[0126] The specific implementation process is as follows: When the consensus rules in step S32 are met: 1) In mainnet mode, the QF1 node attaches a digital signature using its private key; 2) In island mode, the local node layer executes PBFT consensus (≥2 / 3 nodes sign); the signature data generates a hash digest using SHA-256 and writes it into the blockchain transaction, while also attaching a timestamp. The final output is a control command with a complete blockchain signature.

[0127] S4: Execute the control command and iteratively update the configuration parameters of the digital twin and the health score set during prediction based on the execution result.

[0128] Specifically, step S4 includes the following steps:

[0129] S41: Obtain the execution record of the control command, and extract the execution status code and the electrical quantity recovery curve of the feeder.

[0130] In this embodiment, the core function of obtaining control command execution records and extracting key data is to provide a feedback basis for parameter iterative updates. The specific implementation process is as follows: Through the transaction backtracking interface of the blockchain node layer, the execution record log is retrieved based on the transaction hash in the control command; two types of key data are extracted from the log:

[0131] 1) Execution status codes, which are decimal codes representing specific operating conditions;

[0132] 2) Feeder electrical quantity recovery curve, including the change trajectory of voltage / current / frequency within 200ms after command execution (sampling rate 10kHz).

[0133] The immutability of blockchain ensures the authenticity of feedback data, providing high-precision time-domain signals for subsequent model correction.

[0134] S42: Based on the execution status code and the electrical quantity recovery curve of the feeder, correct the LSTM model weight matrix in the configuration parameters of the digital twin during prediction.

[0135] In this embodiment, the core function of modifying the LSTM model weight matrix is ​​to achieve dynamic optimization of the digital twin's predictive capability. The specific implementation process is as follows: First, the electrical quantity recovery curve and the predicted value are time-series aligned, and the mean squared error (MSE) is calculated as the loss function; second, the learning rate is set according to the execution status code (0.01 for normal success and 0.05 for failure to accelerate convergence); finally, the weight matrix is ​​updated through the backpropagation time time (BPTT) algorithm.

[0136] S43: Based on the incremental mechanical wear data of each backup automatic transfer node, update and store the health score of the corresponding backup automatic transfer power supply in the health score set.

[0137] In this embodiment, the core function of updating and storing the health score set is to establish a closed-loop device status tracking system. The specific implementation steps are as follows:

[0138] 1) Collect incremental mechanical wear data (200Hz sampling accuracy) from the vibration sensor of the standby automatic transfer switching node (QF1 / QF2, etc.);

[0139] 2) Calculate the health degradation value according to equipment type: Photovoltaic degradation = vibration increment × 0.8, fan degradation = temperature increment × 0.6;

[0140] 3) Update the health status of the corresponding power supply in the rating set;

[0141] 4) The updated data will be stored in the blockchain historical database via a smart contract (transaction hashes will be automatically generated).

[0142] Example 2

[0143] On the other hand, the present invention also provides a real-time status assessment system for network automatic transfer switching, used to execute a real-time status assessment method for network automatic transfer switching, including at least two automatic transfer power sources, each of which is connected to the feeder of the distribution network through an automatic transfer switching node, as referenced. Figure 2 As shown, the system includes:

[0144] The modality prediction loading module is used to make predictions based on real-time acquired multi-source datasets using a digital twin-driven LSTM model, and load the corresponding evaluation model based on the prediction results.

[0145] The path assessment report module is used to assess the reliability of the feeder switching path based on the loaded assessment model, the initialized threshold parameters, and the health score set of all associated backup power supplies, and generate a status assessment report.

[0146] The blockchain verification instruction module is used to submit the status assessment report to a pre-configured blockchain node, verify the rule consensus through smart contract execution, and output control instructions with blockchain signatures.

[0147] The instruction execution iteration module is used to execute the control instructions and iteratively update the configuration parameters of the digital twin and the health score set during prediction based on the execution results.

[0148] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0149] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium capable of carrying or storing data.

[0150] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

Claims

1. A real-time status assessment method for network automatic transfer switching, comprising at least two automatic transfer power sources, each automatic transfer power source being connected to a feeder of the distribution network through an automatic transfer switching node, characterized in that, The method includes the following steps: Based on real-time acquired multi-source datasets, predictions are made using a digital twin-driven LSTM model, and the corresponding evaluation model is loaded based on the prediction results. Based on the loaded evaluation model, the initialized threshold parameters, and the health score set of all associated backup power supplies, the reliability of the feeder switching path is evaluated, and a status evaluation report is generated. The status assessment report is submitted to a pre-configured blockchain node, and the consensus verification of the execution rules is performed through a smart contract, outputting control instructions with blockchain signatures. The control instructions are executed, and the configuration parameters of the digital twin and the health score set are iteratively updated based on the execution results.

2. The real-time status assessment method for network backup self-transfer as described in claim 1, characterized in that, The multi-source dataset includes output fluctuation data of each backup automatic transfer power source, electrical quantities of the feeders, and status data of each backup automatic transfer switching node.

3. The real-time status assessment method for network backup self-transfer as described in claim 2, characterized in that, The multi-source dataset acquired in real time is used to make predictions using a digital twin-driven LSTM model, and a corresponding evaluation model is loaded based on the prediction results, including: The multi-source real-time state dataset is subjected to time series alignment processing to generate a standardized input vector; The input vector is input into a pre-trained digital twin-driven LSTM model, which outputs the operating mode probability of the distribution network, wherein the operating mode probability includes grid-connected mode probability and islanded mode probability; When the probability of grid connection mode or islanding mode exceeds a preset threshold, the corresponding evaluation model is loaded, wherein the evaluation model includes the corresponding grid connection evaluation model and islanding evaluation model.

4. The real-time status assessment method for network backup self-transfer as described in claim 1, characterized in that, The initialization process for the threshold parameters and health score set includes: Historical health scores of each standby power source are extracted from historical blockchain data to form the health score set. Initialize the corresponding threshold parameters according to the loaded evaluation model, including: If it is a grid-connected evaluation model, the voltage support threshold of the distribution network is initialized to a preset voltage; If it is an island assessment model, the initial penetration threshold of the backup power supply is the preset penetration ratio.

5. The real-time status assessment method for network backup automatic transfer according to claim 4, characterized in that, The reliability assessment of the switching path of the feeder includes: The real-time health of each backup power supply is calculated based on the health score set, and a health anomaly list is generated based on the real-time health. The health anomaly list includes backup power supplies whose real-time health is lower than a preset health threshold. Execute the corresponding evaluation strategy based on the type of the evaluation model: If it is a grid-connected evaluation model, verify whether the voltage of the feeder of the distribution network has reached the initial voltage support threshold, and calculate the reliability score of the switching path based on the real-time health of each backup power source. If it is an island assessment model, verify whether the penetration rate of each backup power supply has reached the initial penetration rate threshold, and calculate the reliability score of the switching path based on the real-time health of each power supply. A switching feasibility score is generated by combining the reliability score with the real-time collected electrical quantities of the feeder. Based on the switching feasibility score and the list of health anomalies, a switching path reconstruction strategy is generated and integrated to form a status assessment report.

6. The real-time status assessment method for network backup automatic transfer according to claim 5, characterized in that, The switching path reconstruction strategy includes at least one of the following: The switching priority of each backup power supply is arranged in descending order of real-time health status; When the reliability score of the primary switching path is lower than the preset score threshold, the predefined backup switching path is activated. Set the closing time difference for different backup automatic transfer power supplies, wherein the closing time difference is dynamically calculated based on the electrical quantities of the feeder.

7. The real-time status assessment method for network backup self-transfer as described in claim 6, characterized in that, The consensus verification through smart contract execution rules includes: The status assessment report is analyzed to extract the switching feasibility score, health anomaly list, and switching path reconstruction strategy. Select the corresponding consensus rule set based on the current operating mode: If it is a grid-connected mode, the main network collaborative consensus rule is adopted: the main network connection node of the distribution network needs to sign and confirm, and the switching feasibility score exceeds the preset score; If it is an isolated mode, the local node consensus rule is adopted: more than a preset proportion of backup self-investment switching nodes must agree and the switching feasibility score must exceed a preset score; A draft control instruction is generated based on the aforementioned switching path reconstruction strategy; When the corresponding consensus rules are met, a digital signature of the blockchain is attached to the draft control instruction, and the blockchain-signed control instruction is output.

8. The real-time status assessment method for network backup self-transfer as described in claim 7, characterized in that, The iterative update based on the execution result includes: Obtain the execution record of the control command, and extract the execution status code and the electrical quantity recovery curve of the feeder; Based on the execution status code and the electrical quantity recovery curve of the feeder, the LSTM model weight matrix in the configuration parameters of the digital twin during prediction is corrected. Based on the incremental mechanical wear data of each backup automatic transfer node, the health score of the corresponding backup automatic transfer power supply in the health score set is updated and stored.

9. A real-time status assessment system for network automatic transfer switching, used to execute a real-time status assessment method for network automatic transfer switching as described in any one of claims 1 to 8, comprising at least two automatic transfer power sources, each automatic transfer power source being connected to a feeder of the distribution network through an automatic transfer switching node, characterized in that, The system includes: The modality prediction loading module is used to make predictions based on real-time acquired multi-source datasets using a digital twin-driven LSTM model, and load the corresponding evaluation model based on the prediction results. The path assessment report module is used to assess the reliability of the feeder switching path based on the loaded assessment model, the initialized threshold parameters, and the health score set of all associated backup power supplies, and generate a status assessment report. The blockchain verification instruction module is used to submit the status assessment report to a pre-configured blockchain node, verify the rule consensus through smart contract execution, and output control instructions with blockchain signatures. The instruction execution iteration module is used to execute the control instructions and iteratively update the configuration parameters of the digital twin and the health score set during prediction based on the execution results.

Images