A method and system for adaptive pre-plan evolution and collaborative emergency control of a gravity energy storage system for extreme working conditions

Abstract

The present application belongs to the technical field of emergency management of gravity energy storage system, and particularly relates to a gravity energy storage system adaptive preplan evolution and collaborative emergency control method and system for extreme working conditions. The present application establishes a risk situation evolution prediction algorithm combined with the large inertia characteristics of the gravity energy storage system and a dynamic emergency disposal window adaptive algorithm, avoids the secondary mechanical impact and structural deformation caused by the rigid fixed triggering and sudden stop, and significantly reduces the physical damage of large inertia equipment under extreme working conditions; the risk classification accuracy is improved by using multi-source data fusion; the bus collapse and parking position collision caused by the concurrent execution of multiple preplans are eliminated by the constraint scheduling for multiple channel scenes; the preplan library has self-learning and safe rollback capabilities by the version evolution mechanism based on digital twinning and quantitative evaluation, and the life cycle artificial maintenance cost is reduced.

Description

Technical Field

[0001] This invention belongs to the field of emergency management technology for gravity energy storage systems, specifically relating to an adaptive contingency plan evolution and collaborative emergency control method and system for gravity energy storage systems under extreme conditions. Background Technology

[0002] Gravity energy storage systems are physical energy storage devices that utilize the lifting and lowering of heavy objects to achieve the mutual conversion of electrical energy and potential energy. The physical architecture of gravity energy storage systems generally has characteristics such as large inertia, strong electromechanical coupling, and high correlation of multiple faults. Under extreme operating conditions such as external environmental disturbances such as strong wind swaying, blizzards and low temperatures, earthquakes, or internal equipment abnormalities such as converter failure and mechanical transmission jamming, the system is very likely to evolve from early minor symptoms to structural failure or destruction in a very short time. With the expansion of the scale of gravity energy storage in new power systems, multi-channel and multi-unit coordinated operation has become the norm, which further increases the complexity of emergency response under extreme operating conditions.

[0003] Existing emergency protection for gravity energy storage and large-scale lifting equipment largely relies on programmable logic controllers and monitoring and data acquisition systems. These systems employ fixed single-point threshold alarms, mechanical and electrical hardware protection interlocks, and, after an alarm is triggered, heavily depend on manual shutdown checks and experience-based handling based on operating procedures. Therefore, this solution provides an adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions to address the aforementioned problems. Summary of the Invention

[0004] The purpose of this invention is to provide an adaptive contingency plan evolution and collaborative emergency control method and system for gravity energy storage systems under extreme conditions, which can solve the problems mentioned in the background art.

[0005] The specific technical solution adopted by this invention is as follows: An adaptive contingency plan evolution and collaborative emergency control method and system for gravity energy storage systems under extreme operating conditions includes: Acquire multi-source operating status data of the gravity energy storage system, perform real-time multimodal feature extraction and feature fusion on the multi-source operating status data to obtain real-time multimodal features, and construct a multimodal health profile of the gravity energy storage system based on the real-time multimodal features; The current extreme working condition type is identified based on the multimodal health profile, and the current feature value of the multimodal feature corresponding to the current extreme working condition type is obtained from the real-time multimodal features. The risk level corresponding to the current extreme working condition type is calculated by combining the current feature value of the multimodal feature. Extract multi-dimensional feature parameters corresponding to the current extreme working condition type, use a lightweight time series prediction model deployed at the edge to predict the real-time situation evolution, obtain the prediction results, and calculate the latest safe handling time based on the prediction results; Based on the current extreme working condition type, the risk level, and the latest safe handling time, multiple candidate plans are retrieved from the preset plan library, and conflict checks are performed on the multiple candidate plans to generate a conflict-free instantiated emergency plan. According to the instantiated emergency plan, multiple actuators of the gravity energy storage system are subjected to shock-resistant coordinated control, and when a communication or power supply abnormality is detected, the system switches to the edge-end link failure autonomous mode and executes preset autonomous protection actions. Obtain the system feedback status after executing the anti-impact collaborative control or the autonomous protection action, and quantitatively evaluate the control and treatment effect based on the feedback status to obtain a comprehensive treatment score. The system checks whether the comprehensive handling score meets the preset standard. If it does not meet the preset standard, the system feedback status is input into the digital twin model for virtual playback and cause localization to obtain the cause analysis result. Based on the cause analysis result, the version parameters of the candidate plans in the plan library are updated to generate a new version plan and stored in the plan library. The candidate plans before the update are retained as historical rollback points.

[0006] In a preferred embodiment, the steps of acquiring multi-source operating status data of the gravity energy storage system, performing real-time multimodal feature extraction and feature fusion on the multi-source operating status data to obtain real-time multimodal features, and constructing a multimodal health profile of the gravity energy storage system based on the real-time multimodal features include: Collect mechanical vibration data, acoustic data, electrical quantity data, displacement and attitude data, braking and locking status data, and environmental disaster data of the gravity energy storage system as multi-source operating status data; Using the unified clock of the edge controller in the gravity energy storage system as a reference, a unified timestamp is assigned to the multi-source operating status data, and a unified time window is used to align the data according to different sampling frequencies; Real-time multimodal features are extracted from the multi-source operating status data. The real-time multimodal features include at least: mechanical vibration features, acoustic features, electrical quantity features, displacement and attitude features, braking and locking status features, and environmental disaster features. The real-time multimodal features are organized according to the structure of component-mode-operating condition-time window to construct a multimodal health profile that includes at least the health features of drive motor, converter, hoisting mechanism, brake, wire rope / chain, guide rail and tower, load attitude, environmental disturbance, and multi-channel collaborative state.

[0007] In a preferred embodiment, the multimodal health profile includes at least the following feature dimensions: mechanical vibration features, acoustic features, electrical quantity features, displacement and attitude features, braking and locking state features, and environmental disaster features; the calculation of the risk level corresponding to the current extreme working condition type includes: An extreme operating condition type library is constructed, which includes strong wind oscillation, low temperature braking degradation, mechanical transmission chain jamming, wire rope abnormality, converter / grid disturbance, multi-unit concurrent conflict, and earthquake / structural disturbance. The current feature value of the multimodal feature is matched with the extreme operating condition type library to generate the current extreme operating condition type. Extract the mean and standard deviation of the health baseline for each feature dimension from the multimodal health profile; The current feature value of the multimodal feature for each feature dimension in the real-time multimodal feature is obtained respectively. Based on the current feature value of the multimodal feature, the mean of the health benchmark, and the standard deviation of the health benchmark, the anomaly degree of each feature dimension is calculated, and the anomaly degree of each feature dimension is recorded as the single-modal anomaly degree; wherein, the single-modal anomaly degree includes mechanical vibration anomaly degree, acoustic anomaly degree, electrical anomaly degree, displacement and attitude anomaly degree, braking and locking anomaly degree, and environmental anomaly degree; Based on the current extreme operating condition type, dynamic weights are assigned to each single-mode anomaly degree. The dynamic weights satisfy normalization constraints. A comprehensive risk value is calculated based on each single-mode anomaly degree and its corresponding dynamic weight. The corresponding risk level is determined based on the comparison between the comprehensive risk value and a preset graded risk threshold.

[0008] In a preferred embodiment, the step of extracting multi-dimensional feature parameters corresponding to the current extreme operating condition type, using a lightweight time-series prediction model deployed at the edge to predict the real-time situation evolution, obtaining prediction results, and calculating the latest safety response time based on the prediction results includes: The system acquires the mass of the heavy object, the lifting and lowering speed of the heavy object, the current risk level, the swing angle of the heavy object, the tension of the wire rope or chain, the braking response time, the DC bus voltage, the instantaneous wind speed, and the guide rail offset of the gravity energy storage system. These data are used as inputs to the lightweight time-series prediction model, which outputs the predicted values ​​of the swing angle of the heavy object, the tension of the wire rope or chain, the braking distance, the DC bus voltage, and the guide rail offset within a preset prediction time window. Based on the predicted values, calculate the sway safety margin, braking safety margin, busbar safety margin, and locking safety margin, and obtain the comprehensive safety margin based on the weighted combination of each safety margin. The minimum predicted time for each preset key risk indicator to reach the danger boundary is calculated based on the predicted value. The total time from receiving the disposal instruction to completing the emergency action of the gravity energy storage system is counted as the emergency action completion time, and the latest safe disposal time is calculated.

[0009] In a preferred embodiment, the contingency plan library is a pre-built contingency plan library, and its construction method includes: Based on the preset extreme operating conditions of the gravity energy storage system and the corresponding risk levels, multiple contingency plan templates are designed in advance. Establish a contingency plan library that includes a contingency plan template library, an action library, a parameter library, a constraint library, a conflict rule library, and a version management library, and store the multiple contingency plan templates in the contingency plan template library; The step of retrieving multiple candidate plans from a preset plan library includes: The matching degree is calculated based on the current extreme working condition type, risk level, and latest safe handling time, and the matching situation with the contingency plan templates in the contingency plan template library. Dynamic weights are assigned to various matching degrees, and a plan matching score is calculated. The plan template with the highest matching score and that meets the safety constraints is selected as the candidate plan.

[0010] In a preferred embodiment, the conflict verification of multiple candidate solutions includes: Determine whether there are conflicts in action timing, component permissions, parameter thresholds, docking positions, electrical constraints, or mechanical constraints among multiple candidate plans; Conflict resolution is carried out in the following priority order: ensuring personal safety and preventing loss of control of heavy objects, preventing structural damage, preventing electrical shocks, ensuring no docking conflicts, and reducing downtime. The rules for conflict resolution include: system-level safety actions take precedence over local equipment protection actions; controlled deceleration takes precedence over mechanical emergency stop; electrical power slope limitation takes precedence over multi-channel concurrent power reduction; braking actions take precedence over unloading actions; and multi-channel locking is performed in a queue manner to avoid peak times. The instantiated emergency plan includes at least: extreme operating condition type, risk level, latest safe handling time, action sequence, execution component, action start and duration, control parameters, electrical power slope constraint, mechanical braking and locking sequence, expected handling effect, failure rollback action and edge chain breakage autonomous backup action.

[0011] In a preferred embodiment, the step of performing shock-resistant coordinated control on multiple actuators of the gravity energy storage system according to the instantiated emergency plan includes: When performing electrical side control, the power reduction actions of each hoisting channel are sorted and staggered, the total power change rate is monitored in real time, and the concurrent power slope is limited to not exceed the upper limit of the total power slope allowed by the DC bus, so as to avoid bus voltage surges and sudden changes in motor torque. When performing mechanical control, the sequence of controlled deceleration, graded braking, emergency stop, and redundant locking is executed. The graded braking provides first-level braking, second-level braking, or third-level braking according to the risk level. The emergency stop selects the stopping position based on the current position of the load, the running speed, the nearest safe stopping point, and the predicted braking distance. The redundant locking sorts the locking order of multiple channels according to the risk level and the distance from the danger boundary, and restricts the locking time interval between adjacent channels to not less than the minimum locking time interval.

[0012] In a preferred embodiment, the edge controller of the gravity energy storage system is communicatively connected to the cloud platform, and the switching to the edge-end disconnected autonomous mode includes: When the communication interruption time between the edge controller and the cloud platform exceeds the threshold, the master station control command times out and does not respond, the upper-layer scheduling system is unavailable, the network quality is lower than the security control requirements, or the digital twin service of the cloud platform is unavailable and the local risk level reaches the preset level, the system enters the edge-end disconnection autonomous mode. In the aforementioned edge-end chain break autonomous mode, only locally pre-verified whitelist action sets are allowed to be executed. The whitelist action sets include: smooth power reduction, controlled deceleration, first-level braking, second-level braking, low-speed hovering, guide rail limit slot locking, redundant mechanical locking, prohibition of concurrent emergency stop, prohibition of high-impact unloading, and multi-channel staggered peak locking.

[0013] This invention also provides an adaptive contingency plan evolution and collaborative emergency control system for gravity energy storage systems under extreme conditions, used to implement the adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions as described in any of the above claims, comprising: The multimodal data acquisition module is used to acquire multi-source operating status data of the gravity energy storage system, perform real-time multimodal feature extraction and feature fusion on the multi-source operating status data to obtain real-time multimodal features, and construct a multimodal health profile of the gravity energy storage system based on the real-time multimodal features. The working condition identification and risk classification module is used to identify the current extreme working condition type based on the multimodal health profile, obtain the current feature value of the multimodal feature corresponding to the current extreme working condition type in the real-time multimodal features, and calculate the risk level corresponding to the current extreme working condition type by combining the current feature value of the multimodal feature. The risk situation evolution prediction module is used to extract multi-dimensional feature parameters corresponding to the current extreme working condition type, and to use a lightweight time series prediction model deployed at the edge to perform real-time situation evolution prediction, obtain prediction results, and calculate the latest safe disposal time based on the prediction results. The contingency plan retrieval and instantiation module is used to retrieve multiple candidate contingency plans from a preset contingency plan library by combining the current extreme working condition type, the risk level and the latest safe handling time, perform conflict verification on the multiple candidate contingency plans, and generate a conflict-free instantiated emergency plan. The collaborative control and edge chain disconnection autonomous module is used to perform anti-impact collaborative control on multiple actuators of the gravity energy storage system according to the instantiated emergency plan, and switch to the edge chain disconnection autonomous mode when a communication or power supply abnormality is detected, and execute preset autonomous protection actions. The quantitative evaluation module is used to obtain the system feedback status after the anti-impact collaborative control or the autonomous protection action is executed, and to quantitatively evaluate the control and treatment effect based on the feedback status to obtain a comprehensive treatment score. The digital twin and contingency plan update module is used to detect whether the comprehensive handling score meets the preset standard. If it does not meet the preset standard, the system feedback status is input into the digital twin model for virtual playback and cause location to obtain the cause analysis result. Based on the cause analysis result, the version parameters of the candidate contingency plans in the contingency plan library are updated to generate a new version contingency plan and stored in the contingency plan library. The candidate contingency plans before the update are retained as historical rollback points.

[0014] And, an electronic device, the electronic device comprising: At least one processor; and a memory that is communicatively connected to the processor; The memory stores a computer program, which, when executed by the processor, implements the adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions as described above.

[0015] The technical effects achieved by this invention are as follows: This invention establishes a risk situation evolution prediction model that incorporates the large inertia characteristics of gravity energy storage and calculates the latest safe handling time. This avoids secondary mechanical impacts and structural deformations caused by rigid fixed triggering and sudden stops, significantly reducing physical damage to large inertia equipment under extreme operating conditions. By employing multi-source operating status data fusion to solve the problem of false activation / failure to activate caused by single sensor failure or environmental noise, the accuracy of risk classification is improved, greatly enhancing the success rate of handling complex fault situations. Through electrical / mechanical dual-constraint scheduling for multi-channel scenarios, the invention eliminates bus collapse and docking position collisions caused by the concurrent execution of multiple candidate plans, thus eliminating the risk of multi-unit coordination conflicts at the site level. Through versioned evolution of plans based on digital twins and quantitative assessment, the plan library has self-learning and safe rollback capabilities, reducing the cost of manual maintenance throughout the entire life cycle. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the adaptive collaborative handling logic for strong wind oscillation conditions in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the multi-channel / multi-unit collaborative scheduling logic of Embodiment 3 of the present invention; Figure 4 This is a schematic diagram of the system modules of the present invention; Figure 5 This is a schematic diagram of the electronic device structure of the present invention. Detailed Implementation

[0017] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0018] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0019] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in a preferred embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that mutually excludes other embodiments.

[0020] Example 1

[0021] Please see Figure 1As shown, this invention provides an adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions, including: S1. Obtain multi-source operating status data of the gravity energy storage system, perform real-time multimodal feature extraction and feature fusion on the multi-source operating status data to obtain real-time multimodal features, and construct a multimodal health profile of the gravity energy storage system based on the real-time multimodal features.

[0022] In S1, the gravity energy storage system includes a drive motor, converter, hoisting mechanism, brake, wire rope or chain, guide rail and tower, and an edge controller deployed at the edge. The edge controller of the gravity energy storage system is communicatively connected to the cloud platform. The gravity energy storage system stores and releases potential energy using a heavy object as a discrete medium through the motor and hoisting mechanism. With the expansion of gravity energy storage scale in new power systems, multi-channel and multi-unit collaborative operation has become the norm. This increases the complexity of emergency response under extreme conditions, requiring multi-source sensing and feature construction. The steps of acquiring multi-source operating status data of the gravity energy storage system, performing real-time multimodal feature extraction and feature fusion on the multi-source operating status data to obtain real-time multimodal features, and constructing a multimodal health profile of the gravity energy storage system based on the real-time multimodal features include: S11. Collect mechanical vibration data, acoustic data, electrical quantity data, displacement and attitude data, braking and locking status data, and environmental disaster data of the gravity energy storage system as multi-source operating status data; As described above, vibration sensors are arranged in the winch bearing housing, reducer housing, reducer bearing housing, wire rope or chain, guide rail support structure, and tower connection nodes to collect mechanical vibration data, which includes vibration velocity, vibration acceleration, and vibration displacement. Acoustic sensors are installed in the winch room, near the brake, in the guide rail limit area, and inside the tower to collect acoustic data, including friction sound, impact sound, and abnormal howling sound. Voltage, current, and bus voltage sensors are installed at the converter, motor stator side, DC bus, and grid connection point to collect electrical quantity data, including the effective value of three-phase current, the effective value of three-phase voltage, the voltage across the converter, and the current passing through it. Displacement sensors, encoders, laser rangefinders, or visual recognition devices are installed at key locations on the weight block, guide rail, pulley block, and tower to collect displacement and attitude data. The displacement and attitude data includes the weight mass, weight height, weight lifting speed, weight swing angle, weight swing speed, wire rope or chain tension, and guide rail offset. Pressure, temperature, stroke, and lock-in position sensors are arranged in the hydraulic brake, mechanical brake, and locking mechanism to collect braking and locking status data. The braking and locking status data includes braking pressure, braking response time, braking temperature rise, braking distance, number of lock-in initiations, and lock-in position. Sensors for wind speed, wind direction, temperature, humidity, snow load, seismic acceleration, and rain / snow intensity are deployed outside the station to collect environmental disaster data, including instantaneous wind speed, average wind speed, ambient temperature, and seismic acceleration.

[0023] S12. Using the unified clock of the edge controller in the gravity energy storage system as a reference, assign a unified timestamp to the multi-source operating status data, and perform data alignment using a unified time window according to different sampling frequencies.

[0024] S13. Extract real-time multimodal features from the multi-source operating status data. The real-time multimodal features include at least: mechanical vibration features, acoustic features, electrical quantity features, displacement and attitude features, braking and locking status features, and environmental disaster features.

[0025] As mentioned above, the following features of mechanical vibration are extracted from mechanical vibration data: vibration velocity, vibration acceleration, vibration displacement, impact factor obtained by the ratio of vibration acceleration to gravitational acceleration, multi-band energy ratio after wavelet packet decomposition, and low-frequency sway energy of the tower obtained by frequency band screening. The following acoustic features are extracted from acoustic data: friction sound, impact sound, abnormal whistling sound, and number of abnormal impact sounds. The following electrical quantity characteristics are extracted from the electrical quantity data: effective value of three-phase current, effective value of three-phase voltage, d-axis voltage and q-axis voltage extracted by Park transform, motor torque fluctuation and negative sequence component characterizing electromechanical disturbance, DC bus voltage change rate, voltage across the converter terminals and the rate of change of converter power obtained by multiplying by the current; The following characteristics of displacement and attitude data are extracted from the monitoring data: mass of the object, height of the object, lifting speed of the object, acceleration of the object, swing angle of the object, swing speed of the object, tension of the wire rope or chain calculated by combining the mass of the object with the acceleration of the object and the acceleration due to gravity, guide rail offset, and distance of the object from the nearest safe stopping position. The following characteristics of braking and locking status data are extracted from the monitoring data: braking pressure, braking pressure rise time, braking response delay, braking temperature rise, braking distance, number of locking initiations, locking status, and the number of locking failures obtained by subtracting the number of locking statuses from the number of locking initiations. The following characteristics of environmental disaster data were extracted from the monitoring data: instantaneous wind speed, average wind speed, gust coefficient obtained by dividing instantaneous wind speed by average wind speed, ambient temperature, and peak seismic acceleration.

[0026] S14. Organize the real-time multimodal features according to the structure of component-mode-operating condition-time window to construct a multimodal health profile that includes at least the health features of drive motor, converter, hoisting mechanism, brake, wire rope / chain, guide rail and tower, heavy object posture, environmental disturbance, and multi-channel collaborative state.

[0027] As described above, the health characteristics of the guide rail and tower include at least instantaneous wind speed, gust coefficient, load swing angle, load swing speed, and low-frequency oscillation energy of the tower. The brake health characteristics include at least ambient temperature, brake pressure rise time, brake response delay, and braking distance; The health characteristics of the drive motor include at least the effective value of the three-phase current, motor torque fluctuation, lifting speed of the heavy object, impact sound, number of abnormal impact sounds, and impact factor. The health characteristics of the wire rope or chain include at least: wire rope or chain tension, impact factor, load height, and guide rail offset. The converter health characteristics include at least d-axis voltage, q-axis voltage, negative sequence component, DC bus voltage change rate, and converter power change rate. The multi-channel coordinated state characteristics include at least braking pressure and lock-in status; The environmental disturbance characteristics include at least the peak seismic acceleration, the low-frequency swaying energy of the tower, and the locked-in state. The health characteristics of the hoisting mechanism include at least the impact factor, the number of abnormal impact sounds, the braking response delay, and the braking distance. The posture characteristics of the heavy object include at least the mass of the heavy object, the height of the heavy object, the lifting speed of the heavy object, the acceleration of the heavy object, the swing angle of the heavy object, the swing speed of the heavy object, the friction sound, the impact sound, the abnormal whistling sound, and the distance of the heavy object from the nearest safe stopping position.

[0028] It should be noted that the feature value corresponding to the single-dimensional feature in the real-time multimodal features is the current feature value of the multimodal features. The range of the current feature value of the real-time multimodal features in the multimodal health profile is the set of the current feature values ​​of the real-time multimodal features when the gravity energy storage system is in a healthy steady state without extreme operating conditions.

[0029] For example, the brake health characteristics under the conditions of "standby steady state, charging upward, releasing downward, and start-stop transitional emergency braking" are described below: Component: Brake; Operating condition 1: Standby steady state; Unified standard time window: 2026-07-27 15:00:00.000~15:00:10.000; Extracted features: Ambient temperature (1Hz) feature value: [26.1℃, 26.5℃], Braking pressure rise time (100Hz) feature value: 0s, Braking response delay (100Hz) feature value: [0ms, 0ms], Braking distance (50Hz) feature value: 0mm; Operating Condition 2: Charging Upwards; Unified standard time window: 2026-07-27 15:05:00.000~15:05:10.000; Extracted features: Ambient temperature (1Hz) feature value: [26.9℃, 27.4℃], Braking pressure rise time (100Hz) feature value: 0.28s, Braking response delay (100Hz) feature value: [25ms, 31ms], Braking distance (50Hz) feature value: 42.6±0.8mm; Operating Condition 3: Energy Release Downward; Unified standard time window: 2026-07-27 15:10:00.000~15:10:10.000; Extracted features: Ambient temperature (1Hz) feature value: [27.5℃, 28.1℃], Braking pressure rise time (100Hz) feature value: 0.26s, Braking response delay (100Hz) feature value: [23ms, 27ms], Braking distance (50Hz) feature value: 47.3±1.0mm; Operating Condition 4: Emergency Braking During Start-Stop Transition; Unified standard time window: 2026-07-27 15:15:00.000~15:15:10.000; Extracted features: Ambient temperature (1Hz) feature value: [28.4℃, 29.3℃], Braking pressure rise time (100Hz) feature value: 0.21s, Braking response delay (100Hz) feature value: [16ms, 20ms], Braking distance (50Hz) feature value: 51.2±1.2mm.

[0030] S2. Identify the current extreme working condition type based on the multimodal health profile, and obtain the current feature value of the multimodal feature corresponding to the current extreme working condition type in the real-time multimodal features. Combine the current feature value of the multimodal feature to calculate the risk level corresponding to the current extreme working condition type.

[0031] In step S2, the steps of identifying the current extreme working condition type based on the multimodal health profile, obtaining the current feature value of the multimodal feature corresponding to the current extreme working condition type from the real-time multimodal features, and calculating the risk level corresponding to the current extreme working condition type by combining the current feature value of the multimodal feature, include: S21. Construct an extreme operating condition type library that includes strong wind oscillation, low temperature braking degradation, mechanical transmission chain jamming, wire rope abnormality, converter / grid disturbance, multi-unit concurrent conflict, and earthquake / structural disturbance, and match the current feature value of the multimodal feature with the extreme operating condition type library to generate the current extreme operating condition type.

[0032] As mentioned above, the extreme condition type library includes: E1 strong wind swaying condition: manifested as instantaneous wind speed exceeding the threshold, gust coefficient exceeding the threshold, weight sway angle exceeding the threshold, weight sway speed exceeding the threshold, and tower low-frequency swaying energy exceeding the threshold.

[0033] E2 Low Temperature Braking Degradation Condition: This condition is characterized by ambient temperature below the preset low temperature threshold, brake pressure rise time exceeding the threshold, brake response delay exceeding the threshold, and braking distance exceeding the healthy baseline.

[0034] E3 Mechanical transmission chain jamming condition: manifested as motor current exceeding the threshold, lifting speed of heavy objects below the threshold, number of abnormal impact sounds exceeding the threshold, and impact factor of reducer or bearing exceeding the threshold.

[0035] E4 Abnormal working conditions of wire rope or chain: manifested as wire rope or chain tension exceeding the threshold, wire rope or chain impact factor exceeding the threshold, load height exceeding or falling below the threshold, guide rail offset exceeding or falling below the threshold.

[0036] E5 converter / grid disturbance condition: manifested as d-axis or q-axis voltage exceeding or falling below the threshold, negative sequence component exceeding the threshold, DC bus voltage change rate exceeding the threshold, and motor torque fluctuation exceeding the threshold.

[0037] E6 Multi-unit concurrent conflict condition: This manifests as the hydraulic or mechanical brakes of multiple hoists simultaneously exceeding the threshold braking pressure and reaching the locked position.

[0038] E7 Seismic / Structural Disturbance Condition: This condition is characterized by peak seismic acceleration exceeding the threshold, low-frequency sway energy of the tower exceeding the threshold, guide rail offset exceeding the threshold, and locking position exceeding the threshold.

[0039] It should be noted that the range of current feature values ​​of the real-time multimodal features in the extreme operating condition type library is the set of current feature values ​​of the real-time multimodal features when the gravity energy storage system experiences extreme operating conditions. When an extreme operating condition occurs, if the current feature value of the real-time multimodal features and the current feature value of the real-time multimodal features corresponding to a certain extreme operating condition type in the extreme operating condition type library both satisfy the range of current feature values ​​of the multimodal features of that extreme operating condition type, then that extreme operating condition type is identified and the current extreme operating condition type is generated.

[0040] S22. Extract the mean and standard deviation of the health baseline for each feature dimension from the multimodal health profile; As mentioned above, the formula for calculating the health baseline mean is: ; Among them, u healthy Let n be the mean of the health baseline, and n be the number of feature values ​​in a single feature dimension of the multimodal health profile that fit its value range. i The i-th feature value in a single feature dimension of a multimodal health profile that conforms to its value range; Formula for calculating the standard deviation of a health baseline: ; Where, σ healthy Standard deviation of the health baseline; S23. Obtain the current feature value of the multimodal feature for each feature dimension in the real-time multimodal feature. Calculate the anomaly degree of each feature dimension based on the current feature value, the mean of the health benchmark, and the standard deviation of the health benchmark. Record the anomaly degree of each feature dimension as the single-modal anomaly degree. The single-modal anomaly degree includes mechanical vibration anomaly degree, acoustic anomaly degree, electrical anomaly degree, displacement and attitude anomaly degree, braking and locking anomaly degree, and environmental anomaly degree.

[0041] As described above, the formula for calculating the single-mode anomaly degree is: For multimodal features with upper and lower constraints ; For multimodal features that only have upper limit constraints ; For multimodal features that only have lower bound constraints ; Where z represents the unimodal anomaly degree, and x represents the current feature value of a single feature dimension in the real-time multimodal features. To prevent extremely small constants with a denominator of zero, the aforementioned X can be 0.000001. maxLet X be the upper limit threshold of x. min Let x be the lower limit threshold.

[0042] S24. Based on the current extreme working condition type, assign dynamic weights to each single-mode anomaly degree, wherein the dynamic weights satisfy normalization constraints, and calculate a comprehensive risk value based on each single-mode anomaly degree and the corresponding dynamic weights. Based on the comparison result between the comprehensive risk value and the preset graded risk threshold, determine the corresponding risk level. As mentioned above, the dynamic weighting rules for single-mode anomaly degrees include: The initial single-mode weights are determined according to the operating condition type: Under the discharge descent condition, the displacement attitude weight, braking lock-up weight, and electrical weight are relatively high; Under full-load lifting conditions, the electrical weight, mechanical vibration weight, braking lock-up weight, and displacement attitude weight are relatively high. Under strong wind conditions, the environmental weight, displacement attitude weight, and mechanical vibration weight are relatively high; Under low-temperature conditions, the weight of brake lock-up and the weight of ambient temperature are relatively high; Under power grid disturbance conditions, the electrical weight and braking lock-up weight are relatively high; When a mode exhibits high anomaly over multiple consecutive time windows, the weight of that mode is increased. When a modality's data is lost, noise is excessive, or self-testing is abnormal, the weight of that modality is reduced, and the weights of other reliable modalities are increased. The formula for calculating the comprehensive risk value is as follows: ; Where R is the overall risk value, z vib For mechanical vibration anomaly, z sound For acoustic anomaly, z elec For electrical anomaly degree, z disp For the displacement attitude anomaly, z brakelock For brake lock-up abnormality, z env For environmental anomaly, W vib For mechanical vibration weights, W sound For acoustic weights, W elec For electrical weights, W disp W represents the displacement attitude weights. brakelock For braking lock-up weight, W env Environmental weight; T1, T2, T3, and T4 are preset as graded risk thresholds. The risk level corresponding to the current extreme working condition type is obtained by comparing the comprehensive risk value with the graded risk thresholds. The risk level includes L1, L2, L3, and L4, where the risk level is L1 when T1≤R<T2; L2 when T2≤R<T3; L3 when T3≤R<T4; and L4 when R≥T4.

[0043] S3. Extract the multi-dimensional feature parameters corresponding to the current extreme working condition type, use a lightweight time-series prediction model deployed at the edge to predict the real-time situation evolution, obtain the prediction results, and calculate the latest safe disposal time based on the prediction results.

[0044] In step S3, the steps of extracting multi-dimensional feature parameters corresponding to the current extreme operating condition type, using a lightweight time-series prediction model deployed at the edge to predict the real-time situation evolution, obtaining prediction results, and calculating the latest safe handling time based on the prediction results include: S31. Obtain the mass of the heavy object, the lifting speed of the heavy object, the current risk level, the swing angle of the heavy object, the tension of the wire rope or chain, the braking response time, the DC bus voltage, the instantaneous wind speed, and the guide rail offset of the gravity energy storage system, and use them as input to the lightweight time series prediction model, and output the predicted values ​​of the swing angle of the heavy object, the tension of the wire rope or chain, the braking distance, the DC bus voltage, and the guide rail offset within the preset prediction time window.

[0045] As described above, multi-dimensional feature parameters of the gravity energy storage system are obtained. These multi-dimensional feature parameters include the mass of the load, the lifting and lowering speed of the load, the current risk level, the swing angle of the load, the tension of the wire rope or chain, the braking response time, the DC bus voltage, the instantaneous wind speed, and the guide rail offset. A lightweight time series prediction model based on Long Short-Term Memory (LSTM) network, deployed at the edge, is adopted. The lightweight time series prediction model includes an input layer, two LSTM hidden layers, and an output layer. The input layer inputs the multi-dimensional feature parameters. Each LSTM hidden layer contains 64 units. The number of model parameters is controlled within 100,000, and the inference latency is less than 100 milliseconds. The output layer outputs the predicted values ​​of the load swing angle, wire rope or chain tension, braking distance, DC bus voltage, and guide rail offset within a preset prediction time window.

[0046] S32. Based on the predicted values, calculate the sway safety margin, braking safety margin, busbar safety margin, and locking safety margin, and obtain the comprehensive safety margin based on the weighted combination of each safety margin. As mentioned above, the formula for calculating the safety margin of oscillation is: ; Among them, T swingFor oscillation safety margin, x angle X is the current characteristic value of the swing angle of the weight. anglemax The maximum allowable swing angle of the weight under the current working conditions is T. swing The closer it is to 0, the higher the risk of oscillation; Formula for calculating braking safety margin: ; Among them, T brake For braking safety margin, x distance P is a characteristic value representing the distance of the heavy object from the nearest safe stopping position. distance The predicted braking distance is given by T. brake A value less than 0 indicates that the current braking capacity is insufficient to guarantee a safe stop; Formula for calculating busbar safety margin: ; Among them, T bus For the busbar safety margin, x bus r represents the current characteristic value of the DC bus voltage. bus is the rated bus voltage, and k is the allowable deviation; Formula for calculating the locking safety margin: T lock =p lock； Among them, T lock To provide a safety margin for locking, p lock p represents the probability of successful locking. lock The lower the value, the less reliable the mechanical locking mechanism is; Dynamic weights are assigned to the shimmy safety margin, braking safety margin, busbar safety margin, and locking safety margin. These dynamic weights satisfy normalization constraints. The formula for calculating the overall safety margin is as follows: ; Among them, T total To provide a comprehensive safety margin, the dynamic weights include a preset W. swing W brake W bus W lock W swing For the oscillation safety margin weight, W brake W is the weight for braking safety margin. bus W is the bus safety margin weight. lock To lock the safety margin weight.

[0047] S33. Calculate the minimum predicted time for each preset key risk indicator to reach the danger boundary based on the predicted value, and calculate the total time from receiving the disposal instruction to completing the emergency action of the gravity energy storage system as the emergency action completion time, and calculate the latest safe disposal time.

[0048] As described above, the start and end points of the time window are obtained by determining the time windows corresponding to when the predicted values ​​of the weight swing angle, wire rope or chain tension, braking distance, DC bus voltage, and guide rail offset reach their respective thresholds. The difference between the end point and the start point is used as the prediction time, and the formula for calculating the minimum prediction time is as follows: ; Among them, T limit For the minimum prediction time, t angle The predicted time t is the time it takes for the predicted value of the weight's swing angle to reach the maximum permissible swing angle. tension The predicted time t is the time it takes for the predicted tension of the wire rope or chain to reach the maximum permissible tension. brake The predicted time t is the time when the predicted braking distance reaches the maximum operating braking distance. bus t is the predicted time for the DC bus voltage to reach the maximum permissible DC bus voltage. rail The predicted time for the predicted value of the guide rail offset to reach the maximum allowable guide rail offset; Formula for calculating the completion time of emergency actions: ; Among them, T action T represents the time required to complete an emergency action. control To control the delay time of command issuance, T drive T is the time for reducing power in the converter or for the deceleration of the drive system. brakeresponse T is the response time of the braking system. execute For the mechanical locking execution time, it should be noted that T control T drive T brakeresponse T execute All are preset nominal values; The formula for calculating the latest safe handling time is as follows: ; Among them, T last For the latest safe response time, T current This refers to the current moment.

[0049] S4. Combining the current extreme working condition type, the risk level, and the latest safe handling time, retrieve multiple candidate plans from the preset plan library, perform conflict verification on the multiple candidate plans, and generate a conflict-free instantiated emergency plan.

[0050] In step S4, the contingency plan library is a pre-built contingency plan library, and its construction method includes: Based on the preset extreme operating condition types of the gravity energy storage system and the risk levels corresponding to the extreme operating condition type library, multiple contingency plan templates are designed in advance; For example, for strong wind swaying conditions, the contingency plan template includes: reducing the lifting speed of the heavy object, limiting the power change rate of the converter, suppressing the low-frequency swaying of the tower, calculating the braking trigger point based on the distance between the nearest heavy object and the nearest safe stopping position, activating graded braking, and mechanically locking at the guide rail limit slot. For low-temperature braking degradation, the contingency plan template includes: increasing the frequency of braking response monitoring, triggering braking in advance, reducing the slope of braking pressure rise, extending the controlled deceleration time, and adding mechanical redundancy lock-up verification. For mechanical transmission chain jamming, the contingency plan template includes: stopping acceleration command, reducing drive torque, initiating low-impact braking, detecting jamming location, planning the nearest safe stopping point, and entering edge chain breakage autonomous mode if necessary. Establish a contingency plan library that includes a contingency plan template library, an action library, a parameter library, a constraint library, a conflict rule library, and a version management library, and store the multiple contingency plan templates in the contingency plan template library; The contingency plan database contains six sub-databases: Contingency plan template library: Stores basic handling procedures according to extreme working conditions and risk levels; Action library: Stores actions such as controlled deceleration, graded braking, graded unloading, emergency stop, guide rail locking, redundant locking, smooth power reduction, bus slope limiting, and edge autonomous protection; Parameter library: Stores the range of control parameters corresponding to each action, such as deceleration rate, braking pressure, braking duration, lock-up interval, power slope upper limit, and docking position priority; Constraint library: Stores electrical constraints, mechanical constraints, timing constraints, component access constraints, and safety boundary constraints; Conflict rule base: Stores conflict types and conflict resolution rules between actions; Version management library: Stores the version number, applicable working conditions, historical handling effects, update reasons, digital twin verification results, and rollback points for each contingency plan.

[0051] The step of retrieving multiple candidate plans from a preset plan library includes: The matching degree is calculated based on the current extreme working condition type, risk level, and latest safe handling time, and the matching situation with the contingency plan templates in the contingency plan template library. Dynamic weights are assigned to various matching degrees, the dynamic weights satisfy normalization constraints, and a plan matching score is calculated. The plan template with the highest plan matching score and that satisfies the safety constraints is selected as the candidate plan.

[0052] Specifically, the formula for calculating the matching degree of the current extreme operating condition type is as follows: ; Among them, M type Matching degree to the current extreme working condition type; Formula for calculating risk level matching degree: ; Among them, M risk For risk level matching degree; Formula for calculating the matching degree of the latest safe response time: ; Among them, M limit Matching the latest safe response time; Formula for calculating the contingency plan matching score: ; The dynamic weights include a preset W. type W risk W limit W type W represents the weight for extreme operating conditions. risk As a weight for risk level, W limit Weighted by the latest safe response time.

[0053] When multiple extreme operating conditions overlap in the system, multiple candidate solutions may be matched simultaneously. In this case, conflict verification is required. The conflict verification of multiple candidate solutions includes: Determine if the following conflicts exist among multiple candidate contingency plans: Action timing conflict: Two contingency plans require the same component to perform different actions at the same time; Component permission conflict: Two plans simultaneously request control of the same actuator, such as the same hydraulic brake simultaneously performing light braking and heavy braking; Parameter threshold conflict: Two plans give different control ranges for the same parameter, for example, one requires rapid power reduction, while the other requires limiting the power slope; Parking location conflict: Multiple heavy objects or multiple passageways are planned to be parked at the same safe parking point; Electrical constraint conflict: Multiple channels simultaneously reducing power or braking simultaneously causes bus voltage surges; Mechanical constraint conflict: The simultaneous locking of multiple channels leads to superimposed impacts on the tower or guide rail structure; Conflict resolution is carried out in the following priority order: ensuring personal safety and preventing loss of control of heavy objects, preventing structural damage, preventing electrical shocks, ensuring no docking conflicts, and reducing downtime. The conflict resolution rules include: system-level safety actions take precedence over local equipment protection actions, controlled deceleration takes precedence over mechanical emergency stop, electrical power slope limitation takes precedence over multi-channel concurrent power reduction, braking actions take precedence over unloading actions, and multi-channel locking is executed in a queue manner to avoid peak times. The instantiated emergency plan includes at least: extreme operating condition type, risk level, latest safe handling time, action sequence, execution component, action start and duration, control parameters, electrical power slope constraint, mechanical braking and locking sequence, expected handling effect, failure rollback action and edge chain breakage autonomous backup action.

[0054] S5. According to the instantiated emergency plan, perform anti-impact coordinated control on multiple actuators of the gravity energy storage system, and switch to the edge-end chain break autonomous mode when a communication or power supply abnormality is detected, and execute the preset autonomous protection action.

[0055] In step S5, according to the instantiated emergency plan, shock-resistant coordinated control is performed on multiple actuators of the gravity energy storage system, including: When performing electrical side control, the power reduction actions of each hoisting channel are sorted and staggered, the total power change rate is monitored in real time, and the concurrent power slope is limited to not exceed the upper limit of the total power slope allowed by the DC bus, so as to avoid bus voltage surges and sudden changes in motor torque. When performing mechanical control, the sequence of controlled deceleration, graded braking, emergency stop, and redundant locking is executed. The graded braking provides first-level braking, second-level braking, or third-level braking according to the risk level. The emergency stop selects the stopping position based on the current position of the load, the running speed, the nearest safe stopping point, and the predicted braking distance. The redundant locking sorts the locking order of multiple channels according to the risk level and the distance from the danger boundary, and restricts the locking time interval between adjacent channels to not less than the minimum locking time interval.

[0056] Specifically, the graded braking includes: Level 1 braking: Used for Level 2 risk level, with lower braking pressure, the goal is to smoothly reduce speed; Secondary braking: Used for L3 risk level, the braking pressure is higher than that of primary braking, and the goal is to complete the stop within a safe distance; Level 3 braking: Used for L4 risk level, it brakes rapidly under the condition of meeting the impact constraint, and simultaneously activates redundant lock-up preparation. The principles for selecting the parking location include: Prioritize selecting the guide rail limiting slot that is closest to the location and meets the structural strength requirements; If the nearest card slot is unavailable, select the next secure card slot; If all card slots are unavailable, perform a low-speed hover and enter edge autonomy mode; The redundant locking requirements, when used in multi-channel locking, are ordered according to the following rules: Channels with high risk levels will be locked first. Access routes closer to the danger zone should be locked first. Access routes that could impact the same tower area must not be locked simultaneously; The interlocking time interval between adjacent channels must not be less than the minimum interlocking time interval.

[0057] The switching to the edge-end disconnected autonomous mode includes: When the communication interruption time between the edge controller and the cloud platform exceeds the threshold, the master station control command times out and does not respond, the upper-layer scheduling system is unavailable, the network quality is lower than the security control requirements, or the digital twin service of the cloud platform is unavailable and the local risk level reaches the preset level, the system enters the edge-end disconnection autonomous mode. In the aforementioned edge-end chain break autonomous mode, only locally pre-verified whitelist action sets are allowed to be executed. The whitelist action sets include: smooth power reduction, controlled deceleration, first-level braking, second-level braking, low-speed hovering, guide rail limit slot locking, redundant mechanical locking, prohibition of concurrent emergency stop, prohibition of high-impact unloading, and multi-channel staggered peak locking.

[0058] S6. Obtain the system feedback status after the anti-impact collaborative control or the autonomous protection action is executed, and quantitatively evaluate the control and handling effect based on the feedback status to obtain a comprehensive handling score.

[0059] In step S6, the process of obtaining the system feedback status after executing the anti-impact coordinated control or the autonomous protection action, and quantitatively evaluating the control effect based on the feedback status to obtain a comprehensive handling score, includes: S61. Obtain the system feedback status after executing the anti-impact collaborative control or the autonomous protection action; As described above, the feedback status includes at least the risk level after handling, braking response time, acceleration of the heavy object, predicted value of braking distance, and braking distance after handling; S62. Quantitatively evaluate the control and treatment effect based on the feedback status, and obtain risk reduction score, response time score, impact suppression score, and action execution accuracy score based on the results of the quantitative evaluation. As mentioned above, the formula for calculating the risk reduction score is:

[0060] For example, if the risk level is L3 and the risk level after treatment is L1, then... , where c riskReduce risk score; Formula for calculating response time score: ; Among them, c time To score response time, x time x is the current characteristic value of the braking response time. ideal x represents the optimal response time under the preset ideal state. limit This is the maximum permissible value for the preset braking response time; Formula for calculating impact suppression score: ; Among them, c impact For impact suppression score, x a Let g be the acceleration of the weight, g be the acceleration due to gravity, and K be a preset value. Maximum allowed value; Formula for calculating the accuracy score of action execution: ; Among them, c accuracy To score the accuracy of action execution, D distance Δ is the braking distance after handling. distance This is the maximum permissible value for the preset braking distance deviation.

[0061] S63. Assign dynamic weights to various scores, with the dynamic weights satisfying normalization constraints. Calculate a comprehensive handling score based on risk reduction score, response time score, impact suppression score, action execution accuracy score, and dynamic weights. The calculation formula for the comprehensive handling score is as follows: Where S represents the comprehensive treatment score, and W... riskreduce To reduce the weight of risk, W time For response time weighting, W impact As the impact suppression weight, W accuracy Weights for the accuracy of action execution; The comprehensive handling scoring levels are divided as follows: When S is greater than or equal to 90, the treatment effect is excellent. If S is greater than or equal to 75 and less than 90, the treatment effect is good. If S is greater than or equal to 60 and less than 75, the treatment effect is considered moderate. If S is less than 60, the treatment effect is poor. The preset standard is that S is greater than or equal to 90. If a secondary accident occurs, the handling effect will be directly judged as poor.

[0062] S7. Detect whether the comprehensive handling score meets the preset standard. If it does not meet the preset standard, input the system feedback status into the digital twin model for virtual playback and cause location to obtain the cause analysis result. Update the version parameters of the candidate plans in the plan library according to the cause analysis result, generate a new version plan and store it in the plan library, and retain the candidate plan before the update as a historical rollback point.

[0063] In step S7, the process includes detecting whether the comprehensive handling score meets a preset standard. If the comprehensive handling score does not meet the preset standard, the system feedback status is input into a digital twin model for virtual playback and cause localization. Based on the analysis results of the virtual playback and cause localization, multiple candidate plans stored in the plan library are updated, a new version plan is generated and stored in the plan library, and the candidate plan before the update is retained as a historical rollback point. S71. Construct a digital twin model of the gravity energy storage system; S72. When the comprehensive handling score is less than the preset standard, the real-time multimodal characteristics, instantiated emergency plan, and feedback status are input into the digital twin model for virtual playback and cause localization to obtain the reasons for the poor handling effect. As mentioned above, the reasons for poor handling results include: risk identification bias, evolution prediction bias, calculation of the latest safe handling time being too late, unreasonable action sequence, braking parameters being too conservative or too aggressive, insufficient power slope limitation, locking sequence leading to structural impact, and equipment aging leading to delayed action response. S73. Based on the comprehensive handling score and the reasons for the poor handling effect, the emergency plan library is updated in a version. Bayesian optimization is used to locally optimize the parameters of the instantiated emergency plan in the digital twin model so that the comprehensive handling score is greater than or equal to the preset standard, thereby obtaining a new version of the emergency plan.

[0064] As mentioned above, the rules for version updates of the contingency plan are as follows: When the handling effect is excellent, the contingency plan is not updated, and only successful cases are recorded. When the treatment effect is good, only minor adjustments are made to local parameters, such as adjusting the braking trigger threshold, power slope threshold, or lock-up time interval by 1% to 5%. When the handling effect is medium, the parameters of the corresponding contingency plan module are updated and the action sequence is optimized. After the digital twin is verified, a new version is generated. When the handling effect is poor, the version of the contingency plan is frozen, manual review or high-level digital twin verification is initiated, and a new candidate contingency plan is generated. Only after the new candidate contingency plan passes the security verification can it be included in the callable contingency plan library.

[0065] S74. Store the new version of the contingency plan in the contingency plan library, record the version number of the new version of the contingency plan, and retain the candidate contingency plan before the update as a historical rollback point.

[0066] As mentioned above, when recording a new version of the contingency plan, the version record should include at least: contingency plan number, version number, applicable extreme working condition type, applicable risk level, parameters before update, parameters after update, reason for update, digital twin verification result, physical system trial result, version effective time, and version rollback point.

[0067] It should be noted that if the effectiveness of the new version of the contingency plan declines or becomes abnormal in actual operation, the system can automatically revert to the previous stable version.

[0068] Example 2

[0069] Please see Figure 2 During the discharge process, the tower gravity energy storage system encountered gusts of wind, causing the heavy object to sway significantly. The adaptive and coordinated handling process for strong wind swaying conditions is as follows: 1) The multimodal data acquisition module synchronously acquires instantaneous wind speed, gust coefficient, load swing angle, load swing speed, and tower low-frequency oscillation energy through sensors; 2) The risk level calculation outputs "L3 risk level". A lightweight time series prediction model is used to predict the real-time situation evolution and obtain the prediction results. The latest safe disposal time is calculated based on the prediction results. The latest safe disposal time is calculated. If no intervention is made, the swing angle will be too large after 15 seconds, which will cause the heavy object to hit the tower wall. The disposal window period is output as 10 seconds. 3) Search the contingency plan library and generate an instantiated emergency plan: constrain the frequency converter to perform controlled deceleration, activate the secondary hydraulic brake, and plan to descend to the nearest guide rail limit slot for mechanical locking within 5 seconds; 4) The quantitative assessment module found that the braking distance was 0.5 meters longer than expected. The comprehensive handling score was rated as good. The system performed a virtual replay in the digital twin model. By comparing the ideal timing (command issuance time, hydraulic pressure building time, and braking force establishment time) with the actual timing, it was found that the actual command issuance time was 0.4 seconds later than the standard value, while other links were normal. Therefore, it was determined that the cause was the control link communication delay. Based on this, a version evolution update was triggered. After the version evolution update was triggered, the braking trigger threshold for the same type of working condition was advanced from "weight swing angle > 15°" to "weight swing angle > 14°", and a new version plan was generated to compensate for the communication delay.

[0070] Example 3

[0071] Please see Figure 3 A power station with eight hoisting channels suddenly encountered a severe voltage drop disturbance in the power grid. At this time, three units were fully loaded while hoisting, and two units were unloaded while descent. The multi-channel / multi-unit coordinated dispatching process is as follows: 1) Electrical quantity data extraction revealed significant d-axis or q-axis voltage exceeding or falling below the threshold, negative sequence component exceeding the threshold, DC bus voltage change rate exceeding the threshold, and motor torque fluctuation exceeding the threshold; 2) If all 8 generating units simultaneously apply emergency mechanical brakes, the sudden drop in back electromotive force and the abrupt change in mechanical stress of their motors will destroy the DC bus and tower beams. 3) The collaborative control and edge chain disconnection autonomous module intervenes to implement smooth power reduction and hovering for the three fully loaded ascending units; and limits the concurrent power slope for the two unloaded descending units, and performs controlled deceleration during off-peak hours. 4) Mechanical side forced queuing and locking: Channel 1 is locked first, and Channel 2 is locked 200ms later. Through coordinated control, the execution is carried out to avoid resonance impact and successfully avoid global disaster.

[0072] Example 4

[0073] Please see Figure 4 An adaptive contingency plan evolution and collaborative emergency control system for gravity energy storage systems under extreme operating conditions, employing the aforementioned adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme operating conditions, includes: The multimodal data acquisition module is used to acquire multi-source operating status data of the gravity energy storage system, perform real-time multimodal feature extraction and feature fusion on the multi-source operating status data to obtain real-time multimodal features, and construct a multimodal health profile of the gravity energy storage system based on the real-time multimodal features. The working condition identification and risk classification module is used to identify the current extreme working condition type based on the multimodal health profile, obtain the current feature value of the multimodal feature corresponding to the current extreme working condition type in the real-time multimodal features, and calculate the risk level corresponding to the current extreme working condition type by combining the current feature value of the multimodal feature. The risk situation evolution prediction module is used to extract multi-dimensional feature parameters corresponding to the current extreme working condition type, and to use a lightweight time series prediction model deployed at the edge to perform real-time situation evolution prediction, obtain prediction results, and calculate the latest safe disposal time based on the prediction results. The contingency plan retrieval and instantiation module is used to retrieve multiple candidate contingency plans from a preset contingency plan library by combining the current extreme working condition type, the risk level and the latest safe handling time, perform conflict verification on the multiple candidate contingency plans, and generate a conflict-free instantiated emergency plan. The collaborative control and edge chain disconnection autonomous module is used to perform anti-impact collaborative control on multiple actuators of the gravity energy storage system according to the instantiated emergency plan, and switch to the edge chain disconnection autonomous mode when a communication or power supply abnormality is detected, and execute preset autonomous protection actions. The quantitative evaluation module is used to obtain the system feedback status after the anti-impact collaborative control or the autonomous protection action is executed, and to quantitatively evaluate the control and treatment effect based on the feedback status to obtain a comprehensive treatment score. The digital twin and contingency plan update module is used to detect whether the comprehensive handling score meets the preset standard. If it does not meet the preset standard, the system feedback status is input into the digital twin model for virtual playback and cause location to obtain the cause analysis result. Based on the cause analysis result, the version parameters of the candidate contingency plans in the contingency plan library are updated to generate a new version contingency plan and stored in the contingency plan library. The candidate contingency plans before the update are retained as historical rollback points.

[0074] Example 5

[0075] Please see Figure 5 This application also provides an electronic device, which includes: At least one processor; and memory that is communicatively connected to at least one processor; The memory stores a computer program, which, when executed by the processor, implements the adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions as described above.

[0076] The processor of the aforementioned electronic device can be a central processing unit (CPU), a graphics processing unit (GPU), or a digital signal processor (DSP), etc., and the memory can be random access memory (RAM), read-only memory (ROM), or solid-state drive (SSD), etc. The electronic device may also include an arithmetic logic unit (ALU), input devices, and output devices. The ALU can be an arithmetic logic unit (ALU) or a floating-point unit (FPU). Input devices include keyboards and mice, and output devices include monitors and printers. They work together to efficiently process and store large amounts of data and ensure stable system operation.

[0077] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0078] The above description is merely a preferred embodiment of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. Structures, devices, and operating methods not specifically described or explained in this invention are implemented according to conventional methods in the art unless otherwise specified or limited.

Claims

1. A method for adaptive contingency plan evolution and collaborative emergency control of gravity energy storage systems under extreme operating conditions, characterized in that, include: Acquire multi-source operating status data of the gravity energy storage system, perform real-time multimodal feature extraction and feature fusion on the multi-source operating status data to obtain real-time multimodal features, and construct a multimodal health profile of the gravity energy storage system based on the real-time multimodal features; The current extreme working condition type is identified based on the multimodal health profile, and the current feature value of the multimodal feature corresponding to the current extreme working condition type is obtained from the real-time multimodal features. The risk level corresponding to the current extreme working condition type is calculated by combining the current feature value of the multimodal feature. Extract multi-dimensional feature parameters corresponding to the current extreme working condition type, use a lightweight time series prediction model deployed at the edge to predict the real-time situation evolution, obtain the prediction results, and calculate the latest safe handling time based on the prediction results; Based on the current extreme working condition type, the risk level, and the latest safe handling time, multiple candidate plans are retrieved from the preset plan library, and conflict checks are performed on the multiple candidate plans to generate a conflict-free instantiated emergency plan. According to the instantiated emergency plan, multiple actuators of the gravity energy storage system are subjected to shock-resistant coordinated control, and when a communication or power supply abnormality is detected, the system switches to the edge-end link failure autonomous mode and executes preset autonomous protection actions. Obtain the system feedback status after executing the anti-impact collaborative control or the autonomous protection action, and quantitatively evaluate the control and treatment effect based on the feedback status to obtain a comprehensive treatment score. The system checks whether the comprehensive handling score meets the preset standard. If it does not meet the preset standard, the system feedback status is input into the digital twin model for virtual playback and cause localization to obtain the cause analysis result. Based on the cause analysis result, the version parameters of the candidate plans in the plan library are updated to generate a new version plan and stored in the plan library. The candidate plans before the update are retained as historical rollback points.

2. The adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions as described in claim 1, characterized in that: The steps of acquiring multi-source operating status data of the gravity energy storage system, performing real-time multimodal feature extraction and feature fusion on the multi-source operating status data to obtain real-time multimodal features, and constructing a multimodal health profile of the gravity energy storage system based on the real-time multimodal features include: Collect mechanical vibration data, acoustic data, electrical quantity data, displacement and attitude data, braking and locking status data, and environmental disaster data of the gravity energy storage system as multi-source operating status data; Using the unified clock of the edge controller in the gravity energy storage system as a reference, a unified timestamp is assigned to the multi-source operating status data, and a unified time window is used to align the data according to different sampling frequencies; Real-time multimodal features are extracted from the multi-source operating status data. The real-time multimodal features include at least: mechanical vibration features, acoustic features, electrical quantity features, displacement and attitude features, braking and locking status features, and environmental disaster features. The real-time multimodal features are organized according to the structure of component-mode-operating condition-time window to construct a multimodal health profile that includes at least the health features of drive motor, converter, hoisting mechanism, brake, wire rope / chain, guide rail and tower, load attitude, environmental disturbance, and multi-channel collaborative state.

3. The adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions as described in claim 1, characterized in that: The multimodal health profile includes at least the following feature dimensions: mechanical vibration features, acoustic features, electrical quantity features, displacement and posture features, braking and locking state features, and environmental disaster features; The calculation of the risk level corresponding to the current extreme operating condition type includes: An extreme operating condition type library is constructed, which includes strong wind oscillation, low temperature braking degradation, mechanical transmission chain jamming, wire rope abnormality, converter / grid disturbance, multi-unit concurrent conflict, and earthquake / structural disturbance. The current feature value of the multimodal feature is matched with the extreme operating condition type library to generate the current extreme operating condition type. Extract the mean and standard deviation of the health baseline for each feature dimension from the multimodal health profile; The current feature value of the multimodal feature for each feature dimension in the real-time multimodal feature is obtained respectively. Based on the current feature value of the multimodal feature, the mean of the health benchmark, and the standard deviation of the health benchmark, the anomaly degree of each feature dimension is calculated, and the anomaly degree of each feature dimension is recorded as the single-modal anomaly degree; wherein, the single-modal anomaly degree includes mechanical vibration anomaly degree, acoustic anomaly degree, electrical anomaly degree, displacement and attitude anomaly degree, braking and locking anomaly degree, and environmental anomaly degree; Based on the current extreme operating condition type, dynamic weights are assigned to each single-mode anomaly degree. The dynamic weights satisfy normalization constraints. A comprehensive risk value is calculated based on each single-mode anomaly degree and its corresponding dynamic weight. The corresponding risk level is determined based on the comparison between the comprehensive risk value and a preset graded risk threshold.

4. The adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions as described in claim 1, characterized in that: The process involves extracting multi-dimensional feature parameters corresponding to the current extreme operating condition type, using a lightweight time-series prediction model deployed at the edge to predict the real-time situation evolution, obtaining prediction results, and calculating the latest safe response time based on the prediction results, including: The system acquires the mass of the heavy object, the lifting and lowering speed of the heavy object, the current risk level, the swing angle of the heavy object, the tension of the wire rope or chain, the braking response time, the DC bus voltage, the instantaneous wind speed, and the guide rail offset of the gravity energy storage system. These data are used as inputs to the lightweight time-series prediction model, which outputs the predicted values ​​of the swing angle of the heavy object, the tension of the wire rope or chain, the braking distance, the DC bus voltage, and the guide rail offset within a preset prediction time window. Based on the predicted values, calculate the sway safety margin, braking safety margin, busbar safety margin, and locking safety margin, and obtain the comprehensive safety margin based on the weighted combination of each safety margin. The minimum predicted time for each preset key risk indicator to reach the danger boundary is calculated based on the predicted value. The total time from receiving the disposal instruction to completing the emergency action of the gravity energy storage system is counted as the emergency action completion time, and the latest safe disposal time is calculated.

5. The adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions as described in claim 1, characterized in that: The contingency plan database is a pre-built database, and its construction methods include: Based on the preset extreme operating condition types of the gravity energy storage system and the risk levels corresponding to the extreme operating condition type library, multiple contingency plan templates are designed in advance; Establish a contingency plan library that includes a contingency plan template library, an action library, a parameter library, a constraint library, a conflict rule library, and a version management library, and store the multiple contingency plan templates in the contingency plan template library; The step of retrieving multiple candidate plans from a preset plan library includes: The matching degree is calculated based on the current extreme working condition type, risk level, and latest safe handling time, and the matching situation with the contingency plan templates in the contingency plan template library. Dynamic weights are assigned to various matching degrees, and a plan matching score is calculated. The plan template with the highest matching score and that meets the safety constraints is selected as the candidate plan.

6. The adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions according to claim 1, characterized in that: The conflict verification of multiple candidate plans includes: Determine whether there are conflicts in action timing, component permissions, parameter thresholds, docking positions, electrical constraints, or mechanical constraints among multiple candidate plans; Conflict resolution is carried out in the following priority order: ensuring personal safety and preventing loss of control of heavy objects, preventing structural damage, preventing electrical shocks, ensuring no docking conflicts, and reducing downtime. The rules for conflict resolution include: system-level safety actions take precedence over local equipment protection actions; controlled deceleration takes precedence over mechanical emergency stop; electrical power slope limitation takes precedence over multi-channel concurrent power reduction; braking actions take precedence over unloading actions; and multi-channel locking is performed in a queue manner to avoid peak times. The instantiated emergency plan includes at least: extreme operating condition type, risk level, latest safe handling time, action sequence, execution component, action start and duration, control parameters, electrical power slope constraint, mechanical braking and locking sequence, expected handling effect, failure rollback action and edge chain breakage autonomous backup action.

7. The adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions according to claim 1, characterized in that: The step of performing shock-resistant coordinated control on multiple actuators of the gravity energy storage system according to the instantiated emergency plan includes: When performing electrical side control, the power reduction actions of each hoisting channel are sorted and staggered, the total power change rate is monitored in real time, and the concurrent power slope is limited to not exceed the upper limit of the total power slope allowed by the DC bus, so as to avoid bus voltage surges and sudden changes in motor torque. When performing mechanical control, the sequence of controlled deceleration, graded braking, emergency stop, and redundant locking is executed. The graded braking provides first-level braking, second-level braking, or third-level braking according to the risk level. The emergency stop selects the stopping position based on the current position of the load, the running speed, the nearest safe stopping point, and the predicted braking distance. The redundant locking sorts the locking order of multiple channels according to the risk level and the distance from the danger boundary, and restricts the locking time interval between adjacent channels to not less than the minimum locking time interval.

8. The adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions according to claim 2, characterized in that: The edge controller of the gravity energy storage system is communicatively connected to the cloud platform. The switching to the edge-end disconnection autonomous mode includes: When the communication interruption time between the edge controller and the cloud platform exceeds the threshold, the master station control command times out and does not respond, the upper-layer scheduling system is unavailable, the network quality is lower than the security control requirements, or the digital twin service of the cloud platform is unavailable and the local risk level reaches the preset level, the system enters the edge-end disconnection autonomous mode. In the aforementioned edge-end chain break autonomous mode, only locally pre-verified whitelist action sets are allowed to be executed. The whitelist action sets include: smooth power reduction, controlled deceleration, first-level braking, second-level braking, low-speed hovering, guide rail limit slot locking, redundant mechanical locking, prohibition of concurrent emergency stop, prohibition of high-impact unloading, and multi-channel staggered peak locking.

9. An adaptive contingency plan evolution and collaborative emergency control system for gravity energy storage systems under extreme conditions, characterized in that: A method for implementing the adaptive contingency plan evolution and coordinated emergency control of a gravity energy storage system for extreme operating conditions, as described in any one of claims 1 to 8, includes: The multimodal data acquisition module is used to acquire multi-source operating status data of the gravity energy storage system, perform real-time multimodal feature extraction and feature fusion on the multi-source operating status data to obtain real-time multimodal features, and construct a multimodal health profile of the gravity energy storage system based on the real-time multimodal features. The working condition identification and risk classification module is used to identify the current extreme working condition type based on the multimodal health profile, obtain the current feature value of the multimodal feature corresponding to the current extreme working condition type in the real-time multimodal features, and calculate the risk level corresponding to the current extreme working condition type by combining the current feature value of the multimodal feature. The risk situation evolution prediction module is used to extract multi-dimensional feature parameters corresponding to the current extreme working condition type, and to use a lightweight time series prediction model deployed at the edge to perform real-time situation evolution prediction, obtain prediction results, and calculate the latest safe disposal time based on the prediction results. The contingency plan retrieval and instantiation module is used to retrieve multiple candidate contingency plans from a preset contingency plan library by combining the current extreme working condition type, the risk level and the latest safe handling time, perform conflict verification on the multiple candidate contingency plans, and generate a conflict-free instantiated emergency plan. The collaborative control and edge chain disconnection autonomous module is used to perform anti-impact collaborative control on multiple actuators of the gravity energy storage system according to the instantiated emergency plan, and switch to the edge chain disconnection autonomous mode when a communication or power supply abnormality is detected, and execute preset autonomous protection actions. The quantitative evaluation module is used to obtain the system feedback status after the anti-impact collaborative control or the autonomous protection action is executed, and to quantitatively evaluate the control and treatment effect based on the feedback status to obtain a comprehensive treatment score. The digital twin and contingency plan update module is used to detect whether the comprehensive handling score meets the preset standard. If it does not meet the preset standard, the system feedback status is input into the digital twin model for virtual playback and cause location to obtain the cause analysis result. Based on the cause analysis result, the version parameters of the candidate contingency plans in the contingency plan library are updated to generate a new version contingency plan and stored in the contingency plan library. The candidate contingency plans before the update are retained as historical rollback points.

10. An electronic device, characterized in that: The electronic device includes: At least one processor; and a memory that is communicatively connected to the processor; The memory stores a computer program, which, when executed by the processor, implements the adaptive contingency plan evolution and collaborative emergency control method for gravity energy storage systems under extreme conditions, as described in any one of claims 1 to 8.

Images