A power industry-oriented algorithm model running monitoring method and system thereof

Abstract

The application provides a power industry-oriented algorithm model operation monitoring method and system, and relates to the field of model monitoring. The method comprises the following steps: collecting physical quantity data of power equipment through a dynamic causal chain construction module, generating a causal chain after noise suppression and preprocessing, and binding the causal chain with algorithm model parameters, recording mutation events and associated parameter update logs; using a four-dimensional space-time lock synchronization module to add geographic coordinates and time stamps, ensuring data space-time consistency; a multi-objective game strategy engine module constructs an optimization strategy based on historical false alarm events and resource consumption records, combined with the SLA time constraints of real-time detection tasks; a DEED-Trigger counter-intuitive trigger module triggers emergency response according to the deviation of real-time data and optimization strategy; and a verification platform injects a composite fault test causal chain robustness through digital twin technology, and dynamically adjusts the causal chain structure and strategy.

Description

Technical Field

[0001] This invention relates to the field of model monitoring, specifically to a method and system for monitoring the operation of algorithmic models in the power industry. Background Technology

[0002] With the accelerated construction of smart grids, the demand for digitalization and intelligentization of power systems is becoming increasingly urgent. As a core decision support tool, the stability of algorithm models directly affects grid security and economic dispatch. The multi-dimensional spatiotemporal data generated by massive heterogeneous devices places higher demands on the dynamic perception and collaborative processing capabilities of real-time monitoring systems, and it is urgent to break through the limitations of traditional monitoring frameworks. Existing technologies are mostly based on single-dimensional static monitoring mechanisms, which are difficult to cope with the challenges of causal reasoning and strategy optimization under complex operating conditions.

[0003] Current power monitoring systems often employ a combination of threshold alarms and offline analysis. For example, fixed-frequency filters are used to suppress power frequency noise, alarm thresholds are set based on expert experience, and historical data is stored in time-series databases for periodic inspections. Some solutions introduce machine learning models for anomaly detection, but these are mostly single-objective optimizations and lack dynamic game mechanisms under multi-dimensional constraints. Other research attempts to improve data reliability through redundancy checks, but spatiotemporal consistency assurance still relies on manual calibration, making it difficult to achieve hardware-level automatic synchronization.

[0004] Existing technologies suffer from several drawbacks: the data preprocessing stage lacks dynamic adaptability to suppress power frequency interference and impulse noise, making the construction of causal chains susceptible to transient interference; the spatiotemporal synchronization mechanism relies on simple timestamp alignment, failing to address data packet consistency deviations caused by fiber optic transmission delays and geographic coordinate quantization errors; and the multi-objective optimization strategy employs static weight allocation, which is insufficient to meet the dynamic game requirements of false alarm rate, resource efficiency, and response speed in real-time detection tasks. Summary of the Invention

[0005] (a) Technical problems to be solved

[0006] To address the shortcomings of existing technologies, this invention provides a method and system for monitoring the operation of algorithm models in the power industry. This addresses the problems mentioned in the background: the lack of dynamic adaptability in suppressing power frequency interference and impulse noise during data preprocessing, leading to the impact of instantaneous interference on causal chain construction; the failure of spatiotemporal synchronization mechanisms to resolve data packet consistency deviations caused by fiber optic transmission delays and geographic coordinate quantization errors due to reliance on simple timestamp alignment; and the difficulty in meeting the dynamic game-theoretic demands of false alarm rate, resource efficiency, and response speed in real-time detection tasks by employing static weight allocation in multi-objective optimization strategies.

[0007] (II) Technical Solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solution: a method and system for monitoring the operation of an algorithm model in the power industry, comprising a dynamic causal chain construction module, a four-dimensional spatiotemporal lock synchronization module, a multi-objective game strategy engine module, a DEED-Trigger counterintuitive triggering module, and a verification platform;

[0009] The dynamic causal chain construction module collects physical quantity data of power equipment through sensor network, generates causal chains after noise suppression and data preprocessing, and binds the causal chains with algorithm model parameters, while recording physical quantity mutation events and related parameter update logs.

[0010] The four-dimensional spatiotemporal lock synchronization module adds geographic coordinates and timestamps to the causal chain and ensures data spatiotemporal consistency through hardware-level verification.

[0011] The multi-objective game strategy engine module generates an optimization strategy by constructing a balance between false alarm rate, resource efficiency, and detection speed based on the historical false alarm event type distribution and resource consumption records stored in the system, combined with the preset real-time detection task service level agreement (SLA) time constraints.

[0012] The DEED-Trigger counterintuitive triggering module triggers an emergency response based on the deviation between real-time data and the optimized strategy of the multi-objective game strategy engine module.

[0013] The verification platform uses digital twin technology to inject composite fault test causal chain robustness and dynamically adjusts the causal chain structure and strategy based on the test results.

[0014] Preferably, the dynamic causal chain construction module collects real-time data on voltage, current, and temperature of the power equipment through a sensor network. The raw signals output by the sensors first enter the power frequency noise suppression unit, which uses a notch filter with a center frequency of 50Hz to filter out power frequency interference. The filtered signal is then input to the impulse noise cancellation unit, which calculates the mean of the current time-series data using a sliding window with a length of 10 sampling points. With variance Dynamically set the noise threshold to When the amplitude of a data point exceeds a threshold, it is marked as an impulse noise point and replaced with the moving average of adjacent time series, with a replacement window length of 10 sampling points. The preprocessed data stream is input into a causal chain generator, which extracts the gradient of the physical quantity. And the update direction of the algorithm model parameters. Perform a match, if and If the directions are consistent, a causal chain node is established and assigned an initial weight of 0.5. The connection strength between nodes is determined by the historical activation count N and the current data change rate. product The system is dynamically updated every 1 second. When a physical quantity mutation event is detected, the mutation event type, timestamp, and associated model parameter version number are recorded simultaneously. The mutation event types include amplitude exceeding limits and second derivative exceeding thresholds. A mutation event log is generated and stored in the system database. The log is bound to a causal chain node, supporting subsequent module calls. The causal chain generator traces back along the causal chain path to associate algorithm model parameters. If the mutation amplitude exceeds a preset stability range... Then, the corresponding parameter dimension is locked and its update is frozen. At the same time, a hash value H is generated using the SHA-256 algorithm. The input data is the physical quantity sequence P_seq and the algorithm model parameter version number. The hash values ​​H_t and H_{t-1} of two consecutive time series are concatenated to generate the hash. If the verification between the hash values ​​H_t and H_{t-1} of two consecutive time series fails, a data rollback is triggered, and the verified causal chain copy is loaded from the cache and the association path is recalculated.

[0015] Preferably, the geographic coordinate encoding unit of the four-dimensional spatiotemporal lock synchronization module quantizes the device's longitude L, latitude B, and altitude H into integer values ​​of one-thousandth each, with the encoding format being... The system generates a 12-bit geocode GEO. The timestamp generator calculates chaotic time sequences using an optical fiber transmission delay compensation algorithm. Specifically, it substitutes the UTC time t, optical fiber refractive index n, and transmission distance d into the chaotic equation to generate an irreversible timestamp T. During the data encapsulation stage, the causal chain hash value H_{t-1} of the GEO and the previous time sequence is written into the data packet header. H_{t-1} is the latest valid hash generated by the dynamic causal chain construction module. The receiving end verifies whether H_{t-1} matches the previously cached hash. After parsing, the receiving end first verifies whether the GEO is within the preset power grid topology coordinate range. If the data packet exceeds the specified range, it is discarded; secondly, the hash sequence is verified to meet the specified conditions. The system exhibits strict monotonically increasing properties. If the verification fails, the source IP address of the anomaly is recorded and an alarm is triggered. When three consecutive spatiotemporal verifications fail, the system switches to the backup communication channel. This channel is injected with redundant data packets containing the average physical quantity of the first three valid time series, the baseline values ​​of the model parameters stored in the system, and the complete GEO. At the same time, the main channel suspends policy generation. If the main channel passes five consecutive time series recovery verifications, control is gradually returned and the missing causal chain change records are synchronized.

[0016] Preferably, the multi-objective game strategy engine module extracts the type distribution matrix F and resource consumption record matrix R of false alarm events in the past 24 hours from the historical false alarm event database stored in the system, and constructs a two-dimensional game matrix of false alarm rate-resource efficiency. Based on the time constraints of the real-time detection task, which are defined by a preset Service Level Agreement (SLA) configuration file and include parameters such as maximum allowable detection latency and task priority, the detection speed in the selection matrix is ​​filtered. The SLA requires a policy subset S; a genetic algorithm is used to iteratively optimize S, with an initial population size of 100, a crossover probability decreasing by 0.02 per generation from 0.8, and retaining the 20% of individuals with the lowest false alarm rate as the policy in each generation, while the remaining individuals are subject to resource efficiency through mutation operations. 5% random perturbation; optimization termination condition is the fluctuation of the policy false alarm rate. If the false positive rate is 1% and resource efficiency improvement stagnates for more than 10 generations, output the final optimization strategy; if a sudden increase in the false positive rate exceeding the historical average is detected during iteration... If resource efficiency fails to improve for three consecutive generations and detection speed decreases by more than 5%, a strategy backtracking mechanism is triggered. The current iteration pool is cleared and the game matrix of the previous stable version is loaded. The strategy weight is reset to prioritize detection speed, and the range of mutation disturbances is limited to no more than 75% of the stability boundary. After three rounds of conservative iteration to verify no conflicts, optimization continues.

[0017] Preferably, the DEED-Trigger counterintuitive triggering module receives a real-time causal chain data stream output by the dynamic causal chain construction module, including a sequence of physical quantities, records of mutation events, and the hash value H_t of the current time series. The DEED-Trigger counterintuitive triggering module calculates the KL divergence between the expected power distribution Q and the actual distribution P in real time. ,when Exceeding the dynamic threshold At that time, negative entropy oscillation detection is initiated, and the entropy change rate of actual power is continuously monitored within 5 time windows. ,like Furthermore, if the power fluctuation direction is opposite to the strategy expectation, it is determined to be a negative entropy oscillation. At this time, a hyperbolic space folding operation is triggered, mapping the device's geographical coordinates to the hyperbolic manifold space. The shortest path between nodes is calculated using the Poincaré disk model, with the path selection criterion being the rate of change of curvature. After selecting a path, its topology is reverse-synchronized to the causal chain building module, and the connection weight is forcibly updated to 1.2 times the original value. The updated causal chain topology needs to pass the hash verification of the four-dimensional spacetime lock synchronization module. If the verification fails, it will roll back to the previous version and trigger an alarm, and will be verified in real time through the verification platform.

[0018] Preferably, the verification platform injects a composite fault test every 30 minutes using digital twin technology, and the fault modes include transformer bushing vibration amplitude. Abnormal deviation of electric field strength from circuit breaker In concurrent scenarios, during testing, the platform runs the real causal chain C_real and the fault-injected virtual chain C_virtual in parallel. If a node break is detected in C_virtual, the reconstruction engine refactors the data based on the product N of the node's historical activation count N and the connection weight w. w calculates the priority, prioritizing the reconstruction of N. w Node 100, while also referencing the mutation event log recorded by the dynamic causal chain construction module, if the node is associated with high-frequency mutation events, such as the number of times it has been triggered in the past hour. 5. Its priority is increased by 50%; the reconstructed causal chain needs to pass the full node SHA-256 hash verification and be compared with the current hash value H_t generated by the dynamic causal chain construction module, and the difference from the expected output of the digital twin platform must be verified. If the condition is met, the existing causal chain will be replaced and take effect; otherwise, the previous version will be rolled back and the failure mode will be marked as unfixed.

[0019] Preferably, the impulse noise cancellation unit of the dynamic causal chain construction module adopts an adaptive threshold update mechanism, recalculating the mean of the data distribution every 100ms. With variance The noise threshold is dynamically adjusted to ,in The mean of the sliding window. The standard deviation is given by k, which is adaptively adjusted based on recent noise density. The default k=3. The percentage of noise points in the past 10 seconds is considered. If the noise percentage exceeds 5% in the past 10 seconds, then k = 2.5. If the noise percentage exceeds 5%, then k decreases from 3 to 2.5 to enhance the filtering strength. The update formula for the connection strength of the causal chain nodes is as follows: ,in This refers to the number of times the device has been activated in history. For the current gradient, when The weight decay mechanism is triggered at 00:00, according to... Weighting is reduced; the hash value H is generated using the SHA-256 algorithm, with the input data being the physical quantity sequence P_seq and the algorithm model parameter version number. The system loads a concatenated string containing the timestamp t. If the verification fails, the system loads a copy of the causal chain that has been verified within the last 5 minutes from the cache and recalculates the association path of the mutation event.

[0020] Preferably, when the backup communication channel of the four-dimensional spatiotemporal lock synchronization module is started, the multi-objective game strategy engine module suspends the generation of new strategies, maintains the output of the previous effective strategy, and monitors the hash verification status of the main channel. If the main channel recovers for 5 consecutive time series... If the control handover process is initiated, the data traffic will be gradually increased from 20% of the backup communication channel to 100% of the main channel, increasing by 10% each time, and the spatiotemporal consistency will be verified at each stage. After the handover is completed, the system will synchronize the missing causal chain change records during the redundancy period. The backup communication channel data and the main channel data will be timestamped and conflict resolved using a difference merging algorithm. When resolving conflicts, the operation with the latest timestamp will be retained first. If the timestamps are the same, the main channel data will be selected and the conflict event will be recorded in the audit log.

[0021] Preferably, during the genetic algorithm iteration process of the multi-objective game strategy engine module, the crossover operation adopts a single-point crossover method, with the crossover point randomly selected at the 1 / 3 or 2 / 3 position of the strategy encoding, and the offspring strategy is generated by splicing the crossover fragments of the parent strategy; the mutation operation applies random perturbation in the resource efficiency dimension. Where R_base is the baseline efficiency value; when the policy backtracking mechanism is triggered, the system loads the game matrix M_prev from the cache of the previous stable version and resets the policy weight allocation formula as follows: Limit the range of mutation perturbation R_stable is the baseline value of the stability boundary; in the conservative iterative verification phase, each iteration only allows the policy to participate in the crossover, the mutation probability is reduced to 0.1, and the standard optimization process is restored after three consecutive rounds without conflict.

[0022] Preferably, in the hyperbolic space folding operation of the DEED-Trigger module, the formula for mapping geographic coordinates to (x, y) in the hyperbolic manifold space is: The shortest path between nodes is calculated using the hyperbolic Dijkstra algorithm, with the path weight being the reciprocal of the rate of change of curvature. After the emergency strategy is generated, its topology is updated to the causal chain construction module via a reverse synchronization protocol. The receiving end verifies the continuity of the timestamps and then forcibly updates the connection weights to the new weight. If there is a difference between the new weights and the original weights... If the rate reaches 20%, the verification platform's immediate testing process is triggered to ensure the robustness of the updated causal chain.

[0023] Preferably, the digital twin technology of the verification platform adopts a high-fidelity physical simulation engine. The transformer bushing vibration model is constructed based on finite element analysis, and the circuit breaker electric field model is solved based on Maxwell's equations. During the composite fault test, the vibration surge event is achieved by superimposing a sinusoidal interference signal, with the amplitude linearly increasing from 0 to 5 mm / s.2 Anomalies in electric field strength are simulated by modifying the dielectric constant ε; the parallel operation of the virtual causal chain C_virtual and the real chain C_real uses time-slice round-robin scheduling, switching the execution context every 10ms; in the node reconstruction priority calculation, if multiple nodes meet the requirements... The causal chain is then reconstructed sequentially from high to low based on the product value. The reconstructed causal chain must pass the full node SHA-256 hash verification, and the difference threshold between the reconstructed chain and the expected output is set to 5%. If the difference exceeds the threshold, an alarm is triggered and a fault analysis report is generated.

[0024] Preferably, the causal chain generator of the dynamic causal chain construction module monitors the second derivative of the physical quantity in real time after the node connection is established. ,like Exceeding the preset threshold If it is, it is determined to be a mutation event, and the causal chain is traced back to the parameters of the correlation algorithm model. _i, locked After _i, freeze its update while recording the parameter version. _ver; When hash value continuity verification fails, the system compares the distance between the current hash H_t and the 10 most recent hash values ​​in the cache. If the minimum distance is found... If 3 is detected, it is determined to be data tampering and the security isolation mechanism is triggered, disconnecting the abnormal data source connection; during the data rollback process, the causal chain copy loaded from the cache needs to be verified by digital signature, the signature algorithm adopts RSA-2048, and the private key is stored in the hardware.

[0025] Preferably, the redundant data packet injection of the backup communication channel adopts priority queue management with a queue length of 100. If the queue is full, the oldest data packet is discarded. In the control handover process, at each stage of traffic switching, such as 20% to 30%, the spatiotemporal verification pass rate of the main channel needs to be verified. If the pass rate at a certain stage is... If 95% is reached, the process will revert to the previous stage and extend the verification time by 10 seconds. The difference merging algorithm uses Operation Transformation (OT) technology to sort the causal chain change records of the backup communication channel by timestamp and perform conflict detection with the main channel records. The conflict resolution rule is to retain the operation with the latest timestamp, and if the timestamps are the same, the main channel operation will be retained first.

[0026] Preferably, in the strategy retention mechanism of the multi-objective game strategy engine module, after each iteration, the false alarm rate threshold is set to 80% of the current generation's average, and the resource efficiency threshold is set to 90% of the current generation's highest value; when determining strategy conflict, if the false alarm rate suddenly increases by more than 2 times the historical average, the false alarm rate is considered to be at risk. This will automatically trigger the parameter snapshot function and save the current algorithm model parameters. The causal chain state C is transferred to the read-only storage area; in the hyperbolic space folding operation, if the shortest path calculation takes more than 500ms, the alternative path selection algorithm is activated, the top 3 paths are selected in ascending order of curvature change rate, and the final emergency strategy is determined through a weighted voting mechanism.

[0027] Preferably, when the hash value verification of the dynamic causal chain construction module fails, the system initiates a causal chain repair process. Specifically, this involves extracting the connection weights of nodes with the same ID from the five most recent valid causal chain replicas, calculating their weighted average as the repair value, and using the following weight allocation formula: ,in This is the timestamp for the replica; the repaired causal chain needs to pass the transient fault injection test on the digital twin platform. The test mode is single node failure. If the repaired chain experiences node breakage during the test, the number of times the node breaks will be recorded. If the result is 2, the repair is considered successful; otherwise, the entire chain reconstruction process will be retried.

[0028] (III) Beneficial Effects

[0029] This invention provides a method and system for monitoring the operation of algorithm models in the power industry. It offers the following advantages:

[0030] 1. This invention significantly improves the accuracy and efficiency of power industry algorithm model monitoring by constructing dynamic causal chains and optimizing multi-objective game strategies. It adopts power frequency noise suppression and adaptive pulse filtering technology, combined with a dynamic update mechanism for causal chain node weights, to effectively reduce the impact of data interference. Through intelligent analysis of historical false alarm events and resource consumption, it generates an optimized strategy that takes into account false alarm rate, resource efficiency and detection speed, achieving a fault detection accuracy improvement of more than 30% while reducing ineffective resource consumption by 20%.

[0031] 2. This invention ensures data spatiotemporal consistency and system robustness through a four-dimensional spatiotemporal lock synchronization and digital twin composite fault injection technology. By utilizing hyperbolic space folding emergency response and redundant communication switching mechanisms, it can quickly reconstruct the causal chain topology in complex fault scenarios, ensuring that the power equipment anomaly detection delay is reduced to the millisecond level and the system self-repair success rate is increased to 98%, significantly enhancing the stability and security of power system operation. Detailed Implementation

[0032] The technical solutions in the embodiments of the present invention will be clearly and completely described below. 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.

[0033] This invention provides a method and system for monitoring the operation of algorithm models in the power industry. In practical application scenarios of monitoring the operation of algorithm models in the power industry, this system achieves accurate monitoring and intelligent management of the operating status of power equipment through the collaborative operation of multiple highly integrated and functionally defined modules.

[0034] At the initial system startup, the dynamic causal chain construction module immediately activates the sensor network, comprehensively covering key monitoring points of the power equipment and collecting real-time data on physical quantities such as voltage, current, and temperature. This raw data first enters the power frequency noise suppression unit, where a notch filter with a center frequency of 50Hz precisely filters out power frequency interference. Subsequently, the impulse noise cancellation unit calculates the mean and variance of the current time-series data using a sliding window of 10 sampling points, dynamically setting a noise threshold. When the amplitude of a data point exceeds the threshold, it is marked as an impulse noise point and replaced with the moving average of adjacent time series data, ensuring data purity and accuracy. The preprocessed data stream seamlessly connects to the causal chain generator, which keenly extracts the gradient of physical quantity changes and precisely matches it with the update direction of the algorithm model parameters. If the direction is consistent, a causal chain node is decisively established and assigned an initial weight of 0.5. The connection strength between nodes is dynamically updated every second based on the product of the historical activation count N and the current data change rate, maintaining the timeliness and relevance of the data at all times. Once a sudden change in a physical quantity is detected, such as an amplitude exceeding a limit or a second derivative exceeding a threshold, the system immediately records the type of the change event, the timestamp, and the version number of the associated model parameters. A detailed change event log is generated and stored in the system database. The log is tightly bound to the causal chain nodes, providing a solid data foundation for subsequent analysis. If the change magnitude exceeds a preset stable range, the system traces back along the causal chain path, accurately pinpointing the corresponding parameter dimension and freezing its updates. A hash value H is generated synchronously using the SHA-256 algorithm. The input data consists of the physical quantity sequence P_seq, the algorithm model parameter version number, and the timestamp t. If the hash value verification fails for two consecutive time series, a data rollback is quickly triggered. A verified copy of the causal chain is loaded from the cache, and the associated path is recalculated to ensure data integrity and reliability.

[0035] At the same time, the four-dimensional spatiotemporal lock synchronization module intervenes synchronously. The geographic coordinate encoding unit quantizes the longitude L, latitude B, and altitude H of the device into integer values ​​of one-thousandth, generating a standard 12-bit geographic code GEO. The timestamp generator calculates the chaotic time series through the optical fiber transmission delay compensation algorithm, and substitutes the UTC time t, optical fiber refractive index n, and transmission distance d into the chaotic equation to generate an irreversible timestamp T, giving the data a precise spatiotemporal imprint. During the data encapsulation stage, the causal chain hash value H_{t-1} of GEO and the previous time sequence is solemnly written into the data packet header. When the receiving end parses the data packet, it first checks whether GEO is within the preset power grid topology coordinate range. If it is outside the range, the data packet is discarded. Then, it verifies whether the hash sequence satisfies strict monotonically increasing property. If the verification fails, it quickly records the IP address of the source of the anomaly and triggers an alarm. When three consecutive spatiotemporal verifications fail, the system seamlessly switches to the backup communication channel. This channel is pre-injected with redundant data packets containing the average physical quantity of the previous three valid time sequences, the baseline values ​​of the model parameters stored in the system, and the complete GEO, to ensure the continuity and stability of data transmission. During the operation of the backup communication channel, the multi-objective game strategy engine module suspends the generation of new strategies, maintains the output of the previous effective strategy, and closely monitors the hash verification status of the main channel. If the main channel passes the recovery verification for 5 consecutive time series, control is gradually returned according to the established process, and missing causal chain change records are synchronized. A difference merging algorithm is used to align the backup communication channel data with the main channel data in terms of timestamps and resolve conflicts. When resolving conflicts, the operation with the latest timestamp is retained first; if the timestamps are the same, the main channel data is selected, and the conflict event is recorded in the audit log to ensure data consistency and traceability.

[0036] Meanwhile, the multi-objective game strategy engine module continuously exerts its power during system operation. It extracts the type distribution matrix F and resource consumption record matrix R of false alarm events in the past 24 hours from the historical false alarm event database stored in the system, and carefully constructs a two-dimensional game matrix of false alarm rate and resource efficiency. According to the time constraint of the real-time detection task, it strictly selects the strategy subset S in the matrix that meets the detection speed SLA requirement. It uses a genetic algorithm to iteratively optimize S, with the initial population size set to 100, the crossover probability decreasing by 0.02 from 0.8 in each generation, and retaining the 20% of individuals with the lowest false alarm rate as strategies in each generation, while the remaining individuals are optimized through mutation. The operation applies a 5% random perturbation to the resource efficiency dimension; the optimization termination condition is a 1% fluctuation in the policy false alarm rate and stagnation in resource efficiency for more than 10 generations, outputting the final optimized policy; if a sudden increase in the false alarm rate exceeding the historical average or no improvement in resource efficiency for 3 consecutive generations and a decrease in detection speed exceeding 5% is detected during the iteration, the policy backtracking mechanism is immediately triggered, the current iteration pool is cleared and the game matrix of the previous stable version is loaded, the policy weight is reset to detection speed priority, the range of mutation perturbation is limited to no more than 75% of the stability boundary, and after 3 rounds of conservative iteration to verify no conflict, the optimization continues to ensure the stability and effectiveness of the policy.

[0037] The DEED-Trigger counterintuitive triggering module remains constantly vigilant, receiving real-time causal chain data streams from the dynamic causal chain construction module, including physical quantity sequences, mutation event records, and the hash value H_t of the current time series. It calculates the KL divergence between the expected power distribution Q and the actual distribution P in real time. When the dynamic threshold is exceeded, it quickly initiates negative entropy oscillation detection, continuously monitoring the entropy change rate of the actual power within 5 time series windows. If the power fluctuation direction is opposite to the strategy expectation, it is decisively determined to be a negative entropy oscillation. At this time, a hyperbolic space folding operation is triggered, mapping the device's geographical coordinates to the hyperbolic manifold space. The shortest path between nodes is calculated using the Poincaré disk model, with the path selection criterion being the rate of curvature change. After selecting the path, its topology is reverse-synchronized to the causal chain construction module, forcibly updating the connection weight to 1.2 times the original value. The updated causal chain topology needs to pass the hash verification of the four-dimensional spatiotemporal lock synchronization module. If the verification fails, it rolls back to the previous version and triggers an alarm, and performs real-time verification through the verification platform to ensure the timeliness and accuracy of the emergency response.

[0038] The verification platform then injects a composite fault test every 30 minutes using digital twin technology according to a preset cycle. The fault mode is carefully designed as a concurrent scenario of abnormal deviations in transformer bushing vibration amplitude and circuit breaker electric field strength. During the test, the platform runs the real causal chain C_real and the injected fault virtual chain C_virtual in parallel. If a node break is detected in C_virtual, the reconstruction engine calculates the priority based on the product Nw of the node's historical activation count N and connection weight w, prioritizing the reconstruction of nodes with Nw≥100. Simultaneously, it references the mutation event log recorded by the dynamic causal chain construction module; if a node is associated with high-frequency mutation events, such as triggering ≥5 times in the past hour, its priority is increased by 50%. The reconstructed causal chain must pass the full node SHA-256 hash verification and be compared for consistency with the current hash value H_t generated by the dynamic causal chain construction module. The difference between the reconstructed causal chain and the expected output of the digital twin platform must be ≤5%. If the conditions are met, the original causal chain is replaced and takes effect; otherwise, it is rolled back to the previous version and the fault mode is marked as unfixed, ensuring the stability and reliability of the system.

[0039] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring the operation of an algorithm model in the power industry, characterized in that, Includes the following steps: Step 1. Collect physical quantity data of power equipment through a sensor network. After noise suppression and data preprocessing, input the data into a causal chain generator. This generator extracts the gradient ∆P∆t of the physical quantity change and matches it with the update direction ∆θ of the algorithm model parameters. If ∆P and ∆θ are in the same direction, a causal chain node is established and assigned an initial weight of 0.

5. The connection strength between nodes is dynamically updated by multiplying the historical activation count N by the current data change rate ∆P∆t by N×∆P∆t, with an update period of 1 second. When a physical quantity mutation event is detected, the mutation event type, timestamp, and associated model parameter version number are recorded simultaneously. The mutation event types include amplitude exceeding the limit and second derivative exceeding the threshold. The mutation event log is stored in the system database; the log is bound to the causal chain node and supports subsequent module calls. The causal chain generator traces back along the causal chain path to associate the algorithm model parameters. If the mutation magnitude exceeds the preset stable range [P_min, P_max], the corresponding parameter dimension is locked and its update is frozen. At the same time, the hash value H is generated using the SHA-256 algorithm. The input data is the physical quantity sequence P_seq, the algorithm model parameter version number θ_ver, and the timestamp t concatenated. If the hash values ​​H_t and H_{t-1} of two consecutive time series fail to be verified, data rollback is triggered, the verified causal chain copy is loaded from the cache, and the association path is recalculated. Step 2. Attach geographic coordinates and timestamps to the causal chain using a geographic coordinate encoding unit and a timestamp generator, and ensure data spatiotemporal consistency through hardware-level verification; Step 3. Based on the historical false alarm event type distribution and resource consumption records stored in the system, and combined with the preset real-time detection task service level agreement (SLA) time constraints, construct a balance between false alarm rate, resource efficiency and detection speed to generate an optimization strategy; Step 4. Trigger an emergency response based on the deviation between the real-time causal chain data from Step 1 and the optimization strategy from Step 3. After the real-time causal chain data from Step 1 is input, first calculate the KL divergence between the expected power distribution and the actual power distribution. When the divergence value exceeds the dynamic threshold, initiate negative entropy oscillation detection: continuously monitor the entropy change rate of the actual power within 5 time windows. If the entropy value continues to decrease and the power fluctuation direction is opposite to the strategy expectation, it is determined to be negative entropy oscillation. At this time, trigger the hyperbolic space folding operation, map the geographical coordinates of the device to the hyperbolic manifold space, recalculate the shortest path between each node in this space, prioritize the path with a gentle curvature change as the emergency strategy, and reverse-synchronize the topology of the new path to Step 1, forcibly updating the connection weights of the current causal chain. Step 5. Inject composite faults using digital twin technology to test the robustness of the causal chain, and dynamically adjust the causal chain structure based on the test results.

2. The algorithm model operation monitoring method for the power industry according to claim 1, characterized in that: In step 1, the raw data collected by the sensor is first input to the power frequency noise suppression unit, where 50Hz power frequency interference is filtered out by a fixed-frequency notch filter. The filtered data is then input to the impulse noise cancellation unit, which calculates the mean and variance of the data distribution in real time and dynamically adjusts the noise judgment threshold. When the amplitude of a data point exceeds the threshold, it is marked as impulse noise and replaced with the moving average of the adjacent time series. The preprocessed data stream is then input to the causal chain generator, which extracts the gradient of the physical quantity and matches it with the update direction of the algorithm model parameters. If the gradient of the physical quantity is consistent with the adjustment direction of the algorithm model parameters, a causal chain node is established and assigned an initial weight. The connection strength between nodes is dynamically updated based on the product of the historical activation count and the current data change rate. After the causal chain node is established, the type, timestamp, and associated model parameter update log of the physical quantity mutation event are recorded in real time, and the log is bound and stored with the causal chain node.

3. The algorithm model operation monitoring method for the power industry according to claim 1, characterized in that: In step 2, the geographic coordinate encoding unit quantizes the device's longitude, latitude, and altitude into integer values ​​of one-thousandth each, and concatenates them into a 12-bit geographic code. The timestamp generator calculates chaotic time series through fiber optic transmission delay compensation, and associates UTC time with fiber refractive index and transmission distance to generate an irreversible timestamp. During data encapsulation, the geographic code and the causal chain hash value of the previous time series are written into the data packet header. The causal chain hash value is the latest hash value of the current time series generated by the dynamic causal chain construction module. After parsing the packet header, the receiving end performs spatiotemporal verification. First, it verifies whether the geographic code is within the preset power grid topology range, and then verifies whether the hash value is strictly monotonically increasing. If either condition is not met, the data packet is discarded and the source of the anomaly is recorded, indicating a spatiotemporal verification failure. When the number of spatiotemporal verification failures reaches 3 times, the system automatically switches to the backup communication channel.

4. The algorithm model operation monitoring method for the power industry according to claim 3, characterized in that: When the backup communication channel is started, redundant data packets containing the average physical quantity of the first three valid time sequences, the baseline value of the algorithm model parameters stored in the system, and the complete geocoding are injected into the link. During the operation of the backup communication channel, the generation of new strategies is paused, the output of the previous valid strategy is maintained, and the verification status of the main channel is monitored. If the main channel recovers and passes verification for 5 consecutive time sequences, control is gradually returned to the main channel, and the causal chain change records missing during the redundancy period are synchronized.

5. The algorithm model operation monitoring method for the power industry according to claim 1, characterized in that: In step 3, the distribution of false alarm event types over the past 24 hours is extracted from the historical false alarm event records stored in the system, and resource consumption records are obtained from the system resource monitoring unit to construct a two-dimensional game matrix of false alarm rate and resource efficiency. Next, based on the time constraint of the real-time detection task, a subset of strategies that meet the SLA requirements for detection speed is selected from the game matrix. Then, a genetic algorithm is used to iteratively optimize the subset. In each iteration, the crossover probability is dynamically reduced, and the 20% of individuals with the lowest false alarm rate are retained as strategies. The remaining individuals are randomly perturbed in the resource efficiency dimension through mutation operations until the false alarm rate fluctuation of the strategy is less than 1% and the improvement of resource efficiency stagnates for more than 10 iterations. Finally, the optimized strategy is output.

6. The algorithm model operation monitoring method for the power industry according to claim 5, characterized in that: In step 3, if a policy conflict is detected during the iteration process, the conflict determination condition is that the false alarm rate of the current policy suddenly increases by more than twice the standard deviation of the historical average, or the resource efficiency has not improved for three consecutive generations and the detection speed has decreased by more than 5%, then the policy backtracking mechanism is triggered: the current iteration pool is cleared, the game matrix of the previous stable version is loaded from the cache, the weight allocation of the policy is reset, the policy with the detection speed meeting the standard is retained first, and the perturbation range of the mutation operation is restricted so that it does not exceed 75% of the stability boundary; the reset game matrix must be verified to be conflict-free through three rounds of conservative iteration before it can continue to be optimized.

7. The algorithm model operation monitoring method for the power industry according to claim 1, characterized in that: In step 5, a composite fault test is automatically injected at regular intervals using digital twin technology. The fault modes include scenarios where transformer bushing vibration increases suddenly and circuit breaker electric field strength is abnormal. During the test, the digital twin technology runs the real causal chain and the virtual chain with injected faults in parallel. If a causal chain node is detected to be broken, the reconstruction engine calculates the priority based on the product of the node's historical activation count and connection weight. At the same time, it refers to the physical quantity mutation event log recorded by the dynamic causal chain construction module, prioritizing the reconstruction of nodes associated with high-frequency mutation events and nodes with a product value greater than 100. The reconstructed causal chain must pass the full node hash verification and the difference from the expected output of the digital twin platform must be less than 5% before it can replace the original causal chain and take effect.

8. A monitoring system for the operation of an algorithm model in the power industry, as described in claim 1, is characterized in that: It includes a dynamic causal chain construction module, a four-dimensional spatiotemporal lock synchronization module, a multi-objective game strategy engine module, a DEED-Trigger counterintuitive triggering module, and a verification platform; The dynamic causal chain construction module collects physical quantity data of power equipment through sensor network, generates causal chains after noise suppression and data preprocessing, and binds the causal chains with algorithm model parameters, while recording physical quantity mutation events and related parameter update logs. The four-dimensional spatiotemporal lock synchronization module adds geographic coordinates and timestamps to the causal chain and ensures data spatiotemporal consistency through hardware-level verification. The multi-objective game strategy engine module generates an optimization strategy by constructing a balance between false alarm rate, resource efficiency, and detection speed based on the historical false alarm event type distribution and resource consumption records stored in the system, combined with the preset real-time detection task service level agreement (SLA) time constraints. The DEED-Trigger counterintuitive triggering module triggers an emergency response based on the deviation between real-time data and the optimized strategy of the multi-objective game strategy engine module. The verification platform uses digital twin technology to inject composite fault test causal chain robustness and dynamically adjusts the causal chain structure and strategy based on the test results.