Method, system and computer device for selective ride-through protection of gravity energy storage system electromechanical decoupling and coupling faults

Abstract

The application relates to a method, system and computer device for selectively crossing over mechanical and electrical decoupling and coupling faults of a gravity energy storage system, wherein mechanical and electrical state data of a single unit is synchronously acquired, when an abnormality occurs, a mutation time sequence and a change rate of mechanical and electrical characteristics are compared in depth by using causal evolution Bayesian inference. In this way, the physical evolution path of the abnormality can be accurately analyzed, whether the fault source is mechanical and physical damage can be accurately determined, and the severity thereof can be evaluated. When preset isolation conditions are met, physical isolation of the damaged single unit is realized through a soft switch buffer circuit, meanwhile, a power shortage caused by off-grid is dynamically calculated, and available compensation margins of healthy single units in the same cluster are real-time checked. Power increment instructions are accurately distributed according to actual margins of the healthy units, the healthy cluster is driven to cooperatively improve output to make up for the position, and blind large-area off-grid of the whole station or the whole machine is avoided. The smooth transition of total output power of a grid connection point of the gravity energy storage system is maintained, and the grid connection resilience of the power station is improved.

Description

Technical Field

[0001] This application relates to the field of gravity energy storage system technology, and in particular to a method, system and computer equipment for selective ride-through protection against electromechanical decoupling and coupling faults in gravity energy storage systems. Background Technology

[0002] Gravity energy storage systems, as large-scale electromechanical coupling systems, play a crucial role in power grid peak shaving and frequency regulation. Because they include mechanical transmission chains such as winches and pulley systems, as well as high-power electrical equipment such as generators and power converters, physical damage such as jamming and track wear on the mechanical side during long-term, high-frequency charging and discharging operation can cause mechanical disturbances that propagate to the electrical side, leading to stator current distortion, voltage fluctuations, or sudden drops in output, thus forming electromechanical coupling faults. The core challenge facing the grid-connected safety of gravity energy storage is how to quickly locate the root cause of such faults and take protective measures without affecting the main power grid.

[0003] Please see Figure 1 and Figure 2 In existing technologies, the protection and control process of energy storage systems is as follows: Figure 2 As shown, a sudden mechanical disturbance (such as jamming) occurs first, due to... Figure 1 The local equipment layer lacks electromechanical internal decoupling sensing, therefore the system can only rely on voltage / current transformers deployed at the power converter outlet for single-dimensional grid-side electrical quantity data acquisition. Under normal operating conditions, the system uses methods such as... Figure 1 The standard industrial control network (WAN / LAN) uploads data to a remote cloud-based central controller for processing. This long-distance transmission introduces significant communication latency, which is the first bottleneck causing slow system response.

[0004] Secondly, when a sudden mechanical disturbance (such as jamming) propagates across the network to the grid side and causes deviations in VT / CT measurements, the system must sequentially undergo a cumbersome process involving the acquisition of grid-side electrical quantities (V / I / f), centralized data upload, processing by the cloud-based central controller (SCADA), and centralized command issuance. This inevitably introduces communication delays. Only after this lengthy process will the system initiate a post-event adjustment mechanism based on the error, activating the overall trip isolation mode and driving the physical layer isolation actuator to disconnect the main circuit. This not only causes the entire unit to be directly disconnected from the grid but also results in a sudden drop in power at the PCC point to zero. This data must then be transmitted back to the local equipment layer again after undergoing bidirectional communication delays. Figure 2 As shown, the system can ultimately only trigger error-triggered adjustment (post-event adjustment) because the entire decision-making path is too long (e.g., Figure 1 (As shown by the bidirectional delay arrow in the image), the data sent from the cloud central controller is as follows: Figure 1The coarse "system trip" command shown by the arrow already exhibits a response lag when it propagates to the actuator. This coarse command drives a conventional isolating contactor ( Figure 1 The electrical protection actuator directly and physically disconnects the main circuit, causing the entire unit to be disconnected from the grid and isolated, and the power at the PCC point to drop to zero. Figure 2 (See the bottom label) It will also generate severe back EMF impact ripple on the DC bus, resulting in serious secondary electrical impact losses.

[0005] Therefore, traditional protection systems focus only on electrical characteristics (such as voltage and current) and cannot identify whether the root cause of fluctuations is pure grid disturbance or internal mechanical and physical damage to the system, which can easily lead to misjudgments. Furthermore, once protection is triggered, a "one-size-fits-all" strategy of disconnecting the entire unit or station from the grid is typically adopted, resulting in significant energy losses due to unnecessary shutdowns and severely weakening the grid-connected resilience of the power station. Even if a single unit is disconnected, directly disconnecting the mechanically loaded motor contactor will generate a strong back electromotive force, causing a secondary impact on the DC bus. Summary of the Invention

[0006] In view of this, in order to at least partially improve the above-mentioned problems, this application provides a method, system and computer device for electromechanical decoupling and selective ride-through protection of coupling faults in a gravity energy storage system.

[0007] Firstly, this application provides a method for selective ride-through protection against electromechanical decoupling and coupling faults in a gravity energy storage system. The gravity energy storage system comprises multiple clusters, and each cluster includes multiple individual units, including: Acquire mechanical status data and electrical operation data for each of the single units, wherein the mechanical status data includes at least the mechanical vibration signal of the winch gearbox collected by the mechanical sensing layer, and the electrical operation data includes at least the voltage and current signals of the power converter grid side collected by the electrical sensing layer. Mechanical features of the mechanical state data and electrical features of the electrical operation data are extracted respectively, and the coupling weight between the mechanical features and the electrical features is calculated to construct a directed deep fault association map; When abnormal fluctuations are detected in the gravity energy storage system, causal evolution Bayesian inference is initiated based on the deep fault correlation map. The temporal sequence of the abrupt change times of the mechanical features and the electrical features is compared, and the first rate of change corresponding to the mechanical features is compared with the second rate of change corresponding to the electrical features, so as to decouple and determine whether the abnormal fluctuations are electromechanical coupling faults caused by physical damage to mechanical equipment. If the abnormal fluctuation is determined to be an electromechanical coupling fault, the severity of the electromechanical coupling is assessed based on the deep fault correlation map. When the severity level meets the preset conditions, an isolation command is sent to the isolation execution layer, and physical isolation is achieved through a soft-switching buffer isolation circuit based on the isolation command. While performing physical isolation, the damaged unit that experienced the electromechanical coupling fault and the healthy units in the same cluster as the damaged unit are identified. The power deficit caused by the disconnection of the damaged unit from the grid is calculated. The power deficit is the active power undertaken by the damaged unit before disconnection from the grid. At the same time, the available power compensation margin of each healthy unit is checked in real time. When the sum of the available power compensation margins of all healthy single-units in the same cluster is greater than or equal to the power deficit, power increments are allocated to each healthy single-unit according to each available power compensation margin and corresponding power increment commands are issued to control each healthy single-unit to work together to compensate for the power deficit, so that the total output power of the gravity energy storage system grid connection point is smoothly transitioned.

[0008] In one embodiment, the step of extracting mechanical features from the mechanical state data and electrical features from the electrical operation data, and calculating the coupling weights between the mechanical features and the electrical features to construct a directed deep fault association map, includes: The mechanical vibration features are extracted from the mechanical state data as the mechanical features; The total harmonic distortion (THD) distortion features in the electrical operating data are extracted as the electrical features. The edge weights between the mechanical vibration characteristics and the total harmonic distortion (THD) distortion characteristics are calculated using the mutual information method, the Pearson correlation coefficient method, or the Jacobian matrix based on the mechanism equation as coupling weights. A directed deep fault association graph is constructed based on the coupling weights.

[0009] In one embodiment, causal evolution Bayesian inference is initiated based on the deep fault correlation map, comparing the temporal sequence of abrupt changes in the mechanical features and the electrical features, and comparing the first rate of change corresponding to the mechanical features with the second rate of change corresponding to the electrical features, to decouple and determine whether the abnormal fluctuation is an electromechanical coupling fault caused by physical damage to mechanical equipment, including: The abrupt change times of the mechanical features and the electrical features are identified by using wavelet modulus maxima detection or CUSUM abrupt change detection, respectively. If the abrupt change of the mechanical feature precedes the abrupt change of the electrical feature and the time difference between the two is greater than the preset discrimination dead zone, and the first rate of change is greater than the second rate of change, then the abnormal fluctuation is determined to be an electromechanical coupling fault caused by physical damage to the mechanical equipment, and a first electromechanical decoupling protection command is output. If the abrupt change of the mechanical feature precedes the abrupt change of the electrical feature and the time difference between the two is greater than the preset discrimination dead zone, but the first rate of change is less than or equal to the second rate of change, then the abnormal fluctuation is determined to be an electrical side abnormality, and a second electromechanical decoupling protection command is output.

[0010] In one embodiment, it also includes: If the abrupt change of the electrical characteristic occurs before the abrupt change of the mechanical characteristic, and the second rate of change corresponding to the electrical characteristic is greater than the first rate of change corresponding to the mechanical characteristic, then the decoupling determination indicates that the abnormal fluctuation is caused by an external power grid fault. The isolation command issued to the isolation execution layer is blocked, the soft switch buffer isolation circuit is prohibited from operating, all the single units in the station are kept in physical grid-connected state, and each single unit is controlled to switch to the preset voltage ride-through control mode to provide voltage support to the power grid.

[0011] In one embodiment, assessing the severity of the electromechanical coupling fault based on the deep fault correlation map includes: Based on the coupling weights corresponding to the deep fault association map, the electromechanical coupling faults are classified into mild faults, moderate faults, or severe faults. When the severity reaches the single-machine isolation threshold, the electromechanical coupling fault is classified as a moderate fault. The method further includes: If it is a minor fault, derating ride-through control is executed to reduce the power limit of the damaged unit and allow it to continue operating; In the case of a severe fault, the damaged unit and all healthy units in the same cluster are synchronously isolated, and a station-level interlock is triggered.

[0012] In one embodiment, real-time verification of the available power compensation margin of the healthy stand-alone unit includes: The current output, maximum allowable output, ramp rate limit, mechanical travel, temperature rise constraint, and reserve capacity of the healthy single unit are obtained to jointly determine the available power compensation margin.

[0013] In one embodiment, the method further includes: If the sum of the available power compensation margins of all the healthy single-units in the same cluster is less than the power deficit, then the healthy single-units and / or system backup units in the adjacent clusters are called up to perform power compensation in order to make up for the power deficit. If the power deficit is still not made up after compensation, the total active power output of the gravity energy storage system will be reduced by a limited slope according to the active power change rate constraint specified by the grid-connected dispatch, and sending excess output commands to healthy single units that are already at full load will be prohibited.

[0014] In one embodiment, the soft-switching buffer isolation circuit includes a main isolation contactor and a buffer energy dissipation branch. The main isolation contactor is connected in series between the generator stator three-phase output terminal of the single unit and the AC side of the power converter. The buffer energy dissipation branch is connected in parallel with the generator stator three-phase output terminal. Physical isolation is achieved through a soft-switching buffer isolation circuit, including: following the timing sequence of first turning on the buffer energy dissipation branch and then turning off the main isolation contactor, so that the stator-side transient current of the damaged unit is dissipated through the buffer energy dissipation branch to achieve smooth physical isolation; The buffer energy dissipation branch includes a three-phase rectifier bridge, a DC-side chopper switch, an energy dissipation resistor, and an RC snubber branch, wherein the DC-side chopper switch is an insulated gate bipolar transistor (IGBT), a solid-state relay, or a thyristor. Alternatively, the buffer energy dissipation branch may include an AC-side bidirectional controllable energy dissipation branch consisting of bidirectional IGBT switches, anti-parallel thyristors, solid-state relays, or hybrid circuit breakers connected in series in each opposite direction.

[0015] This application also provides an electromechanical decoupling and coupling fault selective ride-through protection system for a gravity energy storage system, applied to a gravity energy storage system comprising multiple clusters, each cluster comprising multiple single-unit modules, for implementing the method described in any of the preceding claims, including: The data acquisition module is used to acquire the mechanical status data and electrical operation data of each of the single units, wherein the mechanical status data includes at least the mechanical vibration signal of the winch gearbox collected by the mechanical sensing layer, and the electrical operation data includes at least the voltage signal and current signal of the power converter grid side collected by the electrical sensing layer. The graph construction module is used to extract the mechanical features of the mechanical state data and the electrical features of the electrical operation data respectively, and calculate the coupling weight between the mechanical features and the electrical features to construct a directed deep fault association graph. The decoupling determination module is used to initiate causal evolution Bayesian inference based on the deep fault correlation map after abnormal fluctuations are detected in the gravity energy storage system. It compares the temporal relationship between the abrupt change times of the mechanical features and the abrupt change times of the electrical features, and compares the first rate of change corresponding to the mechanical features with the second rate of change corresponding to the electrical features, so as to decouple and determine whether the abnormal fluctuations are electromechanical coupling faults caused by physical damage to mechanical equipment. The severity assessment module is used to assess the severity of the electromechanical coupling fault based on the deep fault correlation map when the abnormal fluctuation is determined to be an electromechanical coupling fault. An isolation control module is used to issue an isolation command to the isolation execution layer when the severity meets a preset condition, so that the isolation execution layer performs physical isolation through a soft-switching buffer isolation circuit based on the isolation command; A soft-switching buffer isolation circuit is provided, comprising a main isolation contactor and a buffer energy dissipation branch. The main isolation contactor is connected in series between the three-phase output terminal of the generator stator of the single unit and the AC side of the power converter. The buffer energy dissipation branch is connected in parallel with the three-phase output terminal of the generator stator. The isolation execution layer controls the soft-switching buffer isolation circuit to perform physical isolation based on the isolation command. The power compensation module is used to determine the damaged single unit that has experienced the electromechanical coupling fault and the healthy single unit in the same cluster as the damaged single unit while performing the physical isolation, calculate the power deficit caused by the disconnection of the damaged single unit from the grid, the power deficit being the active power undertaken by the damaged single unit before disconnection, and check the available power compensation margin of each healthy single unit in real time. And when the sum of the available power compensation margins of all healthy single units in the same cluster is greater than or equal to the power deficit, power increments are allocated to each healthy single unit according to the available power compensation margin of each healthy single unit and corresponding power increment commands are issued to control each healthy single unit to work together to increase output and compensate for the power deficit, so that the total output power of the gravity energy storage system grid connection point is smoothly transitioned.

[0016] This application also provides a computer device including a memory and a processor, the memory storing a computer application program, and the processor executing the computer application program, wherein the computer application program is configured to implement the method described in any of the preceding claims when executed by the processor.

[0017] Beneficial Effects: By synchronously acquiring electromechanical status data of individual units, a directed deep fault correlation map is constructed. When an anomaly occurs, causal evolution Bayesian inference is used to deeply compare the abrupt change sequence and rate of change of electromechanical characteristics. This enables precise analysis of the physical evolution path of the anomaly, accurate determination of whether the root cause of the fault is mechanical or physical damage, and intelligent assessment of its severity, thereby avoiding false or blind operation of system protection. When the preset isolation conditions are met, the damaged individual unit is physically isolated through a soft-switching buffer circuit. This effectively absorbs the back electromotive force of the motor, eliminates the arcing phenomenon caused by directly disconnecting a large inductive load, suppresses voltage surges on the DC bus, and ensures the safety of the power converter and main circuit. While physically isolating the damaged individual unit, the system dynamically calculates the power deficit caused by grid disconnection and verifies the available compensation margin of healthy individual units in the same cluster in real time. When the total margin meets the requirements, the power increment command is precisely allocated according to the actual margin of each healthy unit, driving the healthy cluster to work together to increase output to make up for the shortfall. This avoids the blind large-scale disconnection of the entire station or the whole unit. Under the safety boundary of ensuring that the healthy single unit is not overloaded, the total output power of the gravity energy storage system grid connection point (PCC) is maintained smoothly, and the grid connection resilience of the power station is improved. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 This is a schematic diagram of the centralized control and protection structure of a conventional energy storage system using existing technology.

[0019] Figure 2 This is a flowchart of existing technology protection and control procedures.

[0020] Figure 3 This is a hardware topology diagram of the electromechanical decoupling and coupling fault selective ride-through protection method for gravity energy storage systems proposed in this application.

[0021] Figure 4 This is a flowchart illustrating the electromechanical decoupling and selective fault ride-through protection method for gravity energy storage systems proposed in this application.

[0022] Figure 5 This is a structural diagram of a soft-switching buffer isolation circuit according to an embodiment of this application.

[0023] Figure 6 This is a schematic diagram of the electromechanical decoupling and selective fault ride-through protection system of a gravity energy storage system according to another embodiment of this application.

[0024] Figure 7 This is a block diagram of a computer device according to an embodiment of this application. Detailed Implementation

[0025] The terms "first," "second," and "third," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the term "comprising," and any variations thereof, is intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or modules is not limited to the listed steps or modules, but may optionally include steps or modules not listed, or may optionally include other steps or modules inherent to these processes, methods, products, or apparatuses.

[0026] Please see Figure 3-7 This application provides a method for selective ride-through protection against electromechanical decoupling and coupling faults in a gravity energy storage system. The gravity energy storage system adopts a hierarchical distributed architecture in its hardware topology, including a multi-cluster local equipment layer, a distributed edge control layer, and a station-level central decision-making system. The local equipment layer is divided into multiple clusters, each cluster comprising multiple parallel single-unit devices, and each single-unit device is controlled by an edge controller unit in the distributed edge control layer. The method is specifically executed by the station-level central decision-making system in coordination with each edge controller unit, and includes: S1. Obtain mechanical status data and electrical operation data for each of the single units, wherein the mechanical status data includes at least the mechanical vibration signal of the winch gearbox collected by the mechanical sensing layer, and the electrical operation data includes at least the voltage signal and current signal of the power converter grid side collected by the electrical sensing layer. S2. Extract the mechanical features of the mechanical state data and the electrical features of the electrical operation data respectively, and calculate the coupling weight between the mechanical features and the electrical features to construct a directed deep fault association map; S3. When abnormal fluctuations are detected in the gravity energy storage system, causal evolution Bayesian inference is initiated based on the deep fault correlation map. The temporal sequence of the abrupt change times of the mechanical features and the abrupt change times of the electrical features is compared, and the first rate of change corresponding to the mechanical features is compared with the second rate of change corresponding to the electrical features, so as to decouple and determine whether the abnormal fluctuations are electromechanical coupling faults caused by physical damage to mechanical equipment. S4. If the abnormal fluctuation is determined to be an electromechanical coupling fault, the severity of the electromechanical coupling is assessed based on the deep fault correlation map. S5. When the severity meets the preset conditions, an isolation command is sent to the isolation execution layer, and physical isolation is achieved through a soft-switching buffer isolation circuit based on the isolation command. S6. While performing physical isolation, identify the damaged unit that has experienced the electromechanical coupling fault and the healthy unit in the same cluster as the damaged unit, calculate the power deficit caused by the disconnection of the damaged unit from the grid, the power deficit being the active power undertaken by the damaged unit before disconnection from the grid; and simultaneously check the available power compensation margin of each healthy unit in real time. S7. When the sum of the available power compensation margins of all the healthy single-unit units in the same cluster is greater than or equal to the power deficit, power increments are allocated to each of the healthy single-unit units according to the available power compensation margins and corresponding power increment commands are issued to control each of the healthy single-unit units to work together to compensate for the power deficit, so that the total output power of the gravity energy storage system grid connection point is smoothly transitioned.

[0027] As described above, this application quantifies the inherent coupling mapping relationship between mechanical and electrical characteristics by synchronously acquiring the mechanical state data and electrical operation data of each individual unit, and constructs a directed deep fault correlation graph. Upon detecting abnormal fluctuations, causal evolution Bayesian inference is used to compare not only the temporal sequence of abrupt changes in the mechanical and electrical characteristics, but also the first rate of change corresponding to the mechanical characteristics and the second rate of change corresponding to the electrical characteristics. This accurately identifies and analyzes the physical evolution path of abnormal fluctuations, thereby precisely determining whether an electromechanical coupling fault is caused by physical damage to the mechanical equipment. If so, the severity is assessed using the deep fault correlation graph, and when preset conditions are met, a specific soft-switching buffer isolation circuit is controlled by the isolation execution layer to perform an action.

[0028] This soft-switching buffer isolation circuit employs a "first-on, then-off" buffer control sequence. First, the buffer energy-dissipating branch connected in parallel with the stator is turned on, allowing the transient high current generated when the damaged unit disconnects from the grid, as well as the residual induced electromagnetic energy on the generator stator side, to be controllably dissipated through this branch. The main isolation contactor is then safely disconnected once the stator current drops to a safe threshold or approaches zero-crossing. This control sequence absorbs the motor's back electromotive force, eliminates the arcing phenomenon caused by directly disconnecting a large inductive load, suppresses voltage surges and secondary electrical stresses on the DC bus during isolation, improves the safety of the power converter and main circuit, and achieves shock-free, smooth, and safe physical isolation of the damaged unit.

[0029] In addition to physically isolating the damaged individual units, the system precisely locates the damaged units and dynamically calculates the power deficit caused by their disconnection from the grid. It also verifies the available power compensation margin of other healthy units in the same cluster in real time and quantitatively. When the sum of the available power compensation margins of all healthy units in the same cluster is greater than or equal to the power deficit, instead of a simple, coarse average distribution, the system allocates power increments based on the available power compensation margin of each healthy unit and issues commands to drive the surrounding healthy units in the cluster to rapidly and collaboratively increase their output to compensate for the power deficit left by the disconnection of the damaged unit. This avoids the high energy losses caused by blindly disconnecting the entire station or the entire unit from the grid. While ensuring the safety boundary of preventing overload and cascading trips of healthy units, the total output power of the gravity energy storage system's grid connection point (PCC) remains smoothly transitioning without sudden changes throughout the fault ride-through, thus improving the grid resilience of the gravity energy storage power station.

[0030] In one embodiment, step S2, which involves extracting mechanical features from the mechanical state data and electrical features from the electrical operation data, and calculating the coupling weights between the mechanical features and the electrical features to construct a directed deep fault association map, includes: S21. Extract the mechanical vibration features from the mechanical state data as the mechanical features; As described above, during the operation of the gravity energy storage unit, piezoelectric vibration sensors deployed in the mechanical sensing layer collect real-time temporal mechanical vibration signals (such as mechanical vibration acceleration signals) of the winch gearbox during rotation and transmission at a preset sampling frequency (e.g., 10kHz). The collected temporal mechanical vibration signals are then processed using frequency domain / time-frequency domain signal processing methods such as discrete wavelet transform, fast Fourier transform (FFT), or power spectral density estimation to extract the fundamental frequency amplitude, spectral distortion rate, or vibration energy change rate. These are used as the mechanical vibration characteristics characterizing the physical damage or internal friction state of the mechanical transmission chain, and are thus considered as mechanical features.

[0031] S22. Extract the total harmonic distortion (THD) distortion features from the electrical operation data as the electrical features; As described above, synchronously with the extraction of the mechanical features, voltage transformers and Hall current sensors deployed in the electrical sensing layer perform high-frequency synchronous sampling of the voltage and current signals at the AC terminal of the power converter grid side corresponding to the single unit, using the same unified timing reference. The current signal can be a three-phase AC stator current sequence, and the voltage signal can be a grid-side voltage sequence. The fundamental and harmonic separation calculations are performed on the sampled voltage and current signals, and the total harmonic distortion (THD) rate or amplitude variation characteristics of specific lower-order harmonics of the three-phase AC power are statistically extracted and used as the electrical features characterizing the electrical side fluctuations.

[0032] S23. Using the mutual information method, Pearson correlation coefficient method, or Jacobian matrix based on the mechanism equation, calculate the edge weight between the mechanical vibration characteristics and the total harmonic distortion (THD) distortion characteristics as the coupling weight. As described above, after extracting the mechanical feature sequence and the electrical feature sequence, any one of the mathematical correlation analysis algorithms, such as Mutual Information, Pearson Correlation Coefficient, or Jacobian Matrix based on mechanistic equations, is used to quantify and measure the cross-domain intrinsic mapping relationship between the mechanical features and the electrical features.

[0033] Taking the mutual information method as an example: the mutual information value of mechanical and electrical features is calculated by statistically analyzing their joint probability distribution and independent marginal probability distribution within a preset sliding window time. This mutual information value is used to quantitatively characterize the tightness of energy mapping of physical disturbances in the mechanical transmission chain propagating across the electrical domain. This mutual information value is defined as the edge weight between the two feature nodes, i.e., the coupling weight. Under normal grid-connected full-power operation conditions with no jamming and no grid disturbances, the calculated electromechanical feature mutual information weight remains below a preset low coupling threshold (e.g., 0.1 or 0.65).

[0034] S24. Construct a directed deep fault association graph based on the coupling weights; As described above, a directed mathematical graph model structure is first established. The extracted mechanical vibration features are set as mechanical layer feature nodes in the graph, and the total harmonic distortion (THD) distortion features are set as electrical layer feature nodes. Based on the inherent mechanical physical damage evolution of the gravity energy storage system to the physical fault propagation direction of the electrical grid anomaly, a directed edge topology structure is set from the mechanical layer feature nodes to the electrical layer feature nodes. Finally, the mutual information value (or Pearson correlation coefficient, Jacobian matrix element value) calculated in real time in step S23 is assigned to the directed edge as its edge weight, thereby completing the dynamic construction of the directed deep fault correlation graph.

[0035] Even better, it also includes: S25, after initiating the causal evolution Bayesian inference, feeding back the causal inference results and updating the edge weights in the deep fault association graph in real time.

[0036] When an abnormal fluctuation occurs in the energy storage system and initiates causal evolution Bayesian inference, the algorithm performs conditional probability derivation based on the directed topology of the deep fault correlation graph, quantitatively outputting the root cause confidence score. This root cause confidence score indicates whether the abnormal fluctuation belongs to "electromechanical coupling faults caused by physical damage to mechanical equipment" or "electrical-side anomalies." After each Bayesian inference cycle, the root cause confidence score output from that inference is used as feedback and input into the mathematical model of the deep fault correlation graph. By adjusting the probability distribution benchmark for the next inference cycle, correcting the correlation coefficient calculation window length, or introducing adaptive multiplicative coefficients for the Jacobian matrix elements, the edge weights of the directed edges in the graph are adaptively corrected and updated in real time. This allows the deep fault correlation graph to dynamically learn and adaptively adjust as the system's operating state changes over time, providing dynamically adaptable decision boundaries for quantitative comparison of multi-source features and accurate output of electromechanical decoupling protection commands in subsequent cycles.

[0037] In one embodiment, step S3, which involves initiating causal evolution Bayesian inference based on the deep fault correlation map, comparing the temporal sequence of abrupt changes in the mechanical features and the electrical features, and comparing the first rate of change corresponding to the mechanical features with the second rate of change corresponding to the electrical features, to decouple and determine whether the abnormal fluctuation is an electromechanical coupling fault caused by physical damage to mechanical equipment, includes: S31. Using wavelet modulus maxima detection or CUSUM mutation detection, identify the mutation time of the mechanical feature and the mutation time of the electrical feature respectively; and extract the absolute rate of change of the mechanical feature within the mutation sliding window and the absolute rate of change of the electrical feature within the corresponding sliding window respectively.

[0038] As described above, by locating the modulus maxima at different scales using wavelet transform, or by monitoring the temporal sliding window statistical changes of the signal mean / variance using the CUSUM algorithm, the abrupt moment when the amplitude of the mechanical vibration characteristic first deviates from the steady-state value can be accurately identified, defined as the abrupt moment t of the mechanical characteristic. m Accurately identify the moment when the total harmonic distortion (THD) distortion characteristic of the grid side exceeds the limit or when the waveform changes abruptly, and define this as the abrupt change moment t of the electrical characteristic. e Meanwhile, within a preset time window following the abrupt change, the system calculates the absolute change amplitude (i.e., absolute change rate) of the mechanical and electrical characteristic fluctuations.

[0039] S32. The absolute rate of change corresponding to the extracted mechanical feature and the absolute rate of change corresponding to the electrical feature are normalized based on their respective steady-state reference values ​​during normal system operation to eliminate the differences between different physical dimensions, thereby obtaining the dimensionless first relative rate of change and the second relative rate of change.

[0040] As mentioned above, mechanical characteristics (e.g., acceleration amplitude) and electrical characteristics (e.g., voltage / current distortion rate) have completely different physical dimensions and cannot be directly compared. Therefore, the system extracts the steady-state vibration amplitude reference and steady-state THD reference value of each unit under fault-free full-power operation as reference denominators. Through normalization calculation, the absolute rate of change is converted into a dimensionless percentage form, thus obtaining the first relative rate of change and the second relative rate of change, thereby providing a unified mathematical scale for the comparison of cross-domain characteristics.

[0041] S33. If the abrupt change of the mechanical feature precedes the abrupt change of the electrical feature, and the time difference between the two is greater than the preset discrimination dead zone, then the temporal sequence relationship and the comparison result of the first relative change rate and the second relative change rate are used as input conditions. Based on the directed topology of the deep fault association map, conditional probability derivation is performed, and the confidence of the fault root cause indicating the root cause of abnormal fluctuations is quantitatively output.

[0042] As described above, the system has a preset discrimination dead zone (in this embodiment, the preset discrimination dead zone is set to 5ms). If the abrupt change time t of the mechanical feature... m The abrupt change t prior to the electrical characteristic e (that is, satisfying t) m <t eIf the time difference between the two is greater than the preset discrimination dead zone (5ms), then the current system fluctuation is determined to meet the temporal causal precondition of "mechanical source causing electrical change". Based on this, the system abandons the coarse binary logic judgment and directly packages the comparison results of the relative change rate after unification of dimensions along with the temporal relationship as input data. This data is then substituted into the deep fault correlation map for Bayesian conditional probability derivation, thereby obtaining the probability value that specifically points to each potential fault source, i.e., the confidence level of the fault root cause.

[0043] S34. Perform branch discrimination on the output confidence scores of the fault root causes: If the first relative rate of change is greater than the second relative rate of change, such that the confidence of the fault root cause of physical damage to mechanical equipment output by the causal evolution Bayesian inference is greater than the preset confidence threshold, then the abnormal fluctuation is determined to be an electromechanical coupling fault caused by physical damage to mechanical equipment, and a first electromechanical decoupling protection command is output. If the first relative rate of change is less than or equal to the second relative rate of change, then based on the deep fault correlation map, the conditional probability distribution is corrected. When the confidence level of reasoning that points to the electrical side's own abnormality is greater than the preset confidence threshold, the abnormal fluctuation is determined to be the electrical side's own abnormality, and the second electromechanical decoupling protection command is output. When the confidence level of the fault root cause indicating physical damage to mechanical equipment and the confidence level of the fault indicating electrical side abnormality are both less than the preset confidence threshold, it is determined that the current electromechanical coupling causal evolution is in an uncertain state, a review instruction is output, and the current output confidence level is used as a feedback quantity to be input into the deep fault association map to adaptively correct the coupling weight, and the next inference cycle is entered for review.

[0044] As mentioned above, the system internally presets an objective confidence threshold (e.g., set to 85%). If the first relative rate of change dominates, it indicates that the severity of mechanical fluctuations is absolutely dominant in the physical evolution path. The confidence level of mechanical fault output by Bayesian inference will naturally be higher than 85%, and the system outputs the first electromechanical decoupling protection command accordingly, triggering subsequent graded assessment and soft-switching isolation. If the second relative rate of change dominates, it indicates that the abrupt change intensity of electrical fluctuations is high, and the initial weak fluctuations on the mechanical side are actually the initial manifestation of the reverse coupling of the electrical transient process. The Bayesian inference corrects the probability distribution accordingly, making the electrical anomaly confidence level higher than 85%. The system outputs the second electromechanical decoupling protection command accordingly to prioritize the electrical side's strategy of first buffering isolation and then smoothing power transition. Furthermore, if the two features are not significant due to system background noise or composite disturbances, that is, neither confidence level reaches the 85% discrimination threshold, the "uncertainty / verification" branch mechanism is triggered. The system utilizes dynamic self-learning feedback characteristics to update the edge weights of directed edges in the graph using the lower confidence level as feedback, and re-evaluates them in the next cycle. This improves the robustness of intelligent decoupling judgment of electromechanical faults under complex working conditions.

[0045] In one embodiment, the method further includes: If the abrupt change of the electrical characteristic occurs before the abrupt change of the mechanical characteristic, and the second rate of change corresponding to the electrical characteristic is greater than the first rate of change corresponding to the mechanical characteristic, then the decoupling determination indicates that the abnormal fluctuation is caused by an external power grid fault. The isolation command issued to the isolation execution layer is blocked, the soft switch buffer isolation circuit is prohibited from operating, all the single units in the station are kept in physical grid-connected state, and each single unit is controlled to switch to the preset voltage ride-through control mode to provide voltage support to the power grid.

[0046] As described above, if the abrupt change time t of the electrical characteristic e The abrupt change t prior to the mechanical feature m (that is, satisfying t) e < t m If the time difference between the two is greater than the preset dead zone (5ms), the current fluctuation is determined to meet the timing characteristics of a forced mechanical response induced by an electrical source. Based on this, if the second rate of change is greater than the first rate of change—for example, the instantaneous voltage drop rate at the grid connection point caused by an external transmission line fault is significantly higher than the torque pulsation and gearbox vibration rate transmitted to the mechanical side via the drive shaft—it indicates that the intensity of the electrical disturbance dominates the physical evolution path, and the fluctuation on the mechanical side is forced by the electrical side. The causal evolution Bayesian inference thus excludes internal mechanical physical damage and infers that the root cause of the abnormal fluctuation is an external power grid fault.

[0047] In response to the determination result, anti-maloperation interlocking control is executed: the isolation command issued to the isolation execution layer is blocked, and the soft-switching buffer isolation circuit is prohibited from operating (including prohibiting the switching of the main isolation contactor and the chopper switch in the buffer energy dissipation branch), keeping all single units in the station in a physically grid-connected state, and avoiding the erroneous disconnection of healthy single units due to external grid disturbances; at the same time, the power converter of each single unit is controlled to switch to a preset voltage ride-through control mode. This preset voltage ride-through control mode is, for example, a low voltage ride-through (LVRT) mode, in which the power converter adjusts the control strategy within a preset very short time (e.g., within 20ms), suspends active power output or limits active current, and injects a preset proportion of reactive current into the grid (e.g., injecting a reactive support current of 1.0 pu of rated current) to provide transient reactive power support, assist in grid voltage recovery, and realize the gravity energy storage system's anti-maloperation high-toughness ride-through under external grid faults.

[0048] In one embodiment, assessing the severity of the electromechanical coupling fault based on the deep fault correlation map includes: Based on the coupling weights corresponding to the deep fault association map, the electromechanical coupling faults are classified into mild faults, moderate faults, or severe faults. Specifically, the system pre-stores a first weight threshold and a second weight threshold that characterize the severity of the system's stress, wherein the second weight threshold is greater than the first weight threshold. If the extracted coupling weight is less than or equal to the first weight threshold, the system classifies the current electromechanical coupling fault as a mild fault. If the extracted coupling weight is between the first weight threshold and the second weight threshold (for example, the mutual information coupling weight obtained from the graph is 0.8), the system classifies the current electromechanical coupling fault as a moderate fault. If the extracted coupling weight is equal to or greater than the second weight threshold, the system classifies the current electromechanical coupling fault as a severe fault.

[0049] When the severity reaches the single-machine isolation threshold, the electromechanical coupling fault is classified as a moderate fault. When the severity is classified as a moderate fault, the system performs the following actions: it issues an isolation command to the isolation execution layer to drive the soft-switching buffer isolation circuit to perform smooth physical isolation on the damaged unit; at the same time, it triggers power compensation calculation to calculate the power deficit caused by the disconnection of the damaged unit from the grid, determines the power increment allocation corresponding to each healthy unit in the same cluster, issues the corresponding power increment command to each healthy unit, and controls each healthy unit to work together to increase output in order to compensate for the power deficit and achieve smooth collaborative breakthrough under moderate fault conditions.

[0050] In one embodiment, when the severity of the electromechanical coupling fault is assessed as a minor fault, it indicates that the damaged unit is partially obstructed but has not yet completely lost its operational capability. In this case, minor bypass control logic is executed: A derating operation command is issued to the damaged unit to reduce its power limit, enabling it to maintain physical grid-connected operation under limited power conditions. At the same time, the output status of other healthy units in the same cluster remains unchanged, and they are not called upon to participate in additional power compensation.

[0051] When the severity of the electromechanical coupling fault is assessed as a severe fault, it indicates severe physical damage such as rigid jamming at the mechanical transmission end, and the fault impact range exceeds the tolerance limit of a single converter, potentially triggering a cascading failure of other single units within the same cluster. In this case, severe isolation protection logic is executed: Skip the power compensation step, control the damaged unit and all healthy units in the same cluster to synchronously perform isolation and disconnect from the network, and trigger a station-level interlock command to the station-level controller to put the entire station into a heavily isolated state to protect the underlying hardware equipment from further damage.

[0052] In one embodiment, step S6, which involves real-time verification of the available power compensation margin of the healthy stand-alone unit, includes: S61. Obtain the current output, maximum allowable output, ramp rate limit, mechanical stroke, temperature rise constraint, and reserve capacity of the healthy single unit to jointly determine the available power compensation margin.

[0053] As mentioned above, when a moderate fault occurs and the damaged unit is physically isolated, to ensure that the healthy units participating in the collaborative compensation do not experience chain reactions such as mechanical overload, electrical overcurrent, or overtemperature tripping during the output increase process, it is necessary to dynamically verify the multi-dimensional physical boundary constraints of each healthy unit. Specifically, the following parameters of each healthy unit in the same cluster can be obtained in real time through the station-level communication network: current output, maximum allowable output, ramp rate limit, mechanical travel constraint, temperature rise constraint, and reserve capacity. A multi-constraint minimum dynamic programming method is used to determine the available power compensation margin H of the i-th healthy unit. i : ; Among them, P i,0 The current output (actual active power) of the i-th healthy single unit is sampled and reported in real time by the power converter control unit or high-frequency power sensor; P i,max R represents the maximum allowable output of this healthy single-unit, which is a preset rated constant. iThe ramp rate limit for this healthy single-unit is characterized by the maximum rate at which its active power can be adjusted per unit time. It is determined by the mechanical inertia of the winch drive train and the control bandwidth of the converter's transient current loop. Δt is the calculation step size for a single active power dispatch response, and R... i ×Δt is used to limit the maximum transient power increment allowed due to physical inertia within a single scheduling cycle, M i,lim The maximum power increment allowed to be released by this healthy single-unit under the current mechanical travel constraints is calculated in real time by the station-level controller based on the position of the winch drum encoder, the current height of the gravity block, and the remaining safety turns of the wire rope. T i,lim The allowable transient overload power space for this healthy single-unit generator under the current temperature rise constraint is dynamically calculated by the local temperature rise control model based on temperature sensor data from the generator stator windings, gearbox main bearings, and power converter IGBT modules. C i,res The current allocated power plant reserve capacity for this healthy single-unit is determined based on the total station spinning reserve capacity. By minimizing the power constraint values ​​across the aforementioned multiple dimensions, the absolutely safe available power compensation margin H for each healthy single-unit is determined. i This facilitates the limiting allocation of subsequent power increment commands, eliminating the risk of component overload, cascading trips, or system crashes caused by the healthy cluster's supplementary output at the physical boundary.

[0054] In one embodiment, the method further includes: If the sum of the available power compensation margins of all the healthy single-units in the same cluster is less than the power deficit, then the healthy single-units and / or system backup units in the adjacent clusters are called up to perform power compensation in order to make up for the power deficit. If the power deficit is still not made up after compensation, the total active power output of the gravity energy storage system will be reduced by a limited slope according to the active power change rate constraint specified by the grid-connected dispatch, and sending excess output commands to healthy single units that are already at full load will be prohibited.

[0055] As described above, if the sum of the available power compensation margins is less than the power deficit, it indicates that the available compensation capacity of the healthy single-units in the same cluster is insufficient to completely fill the power deficit. Then, through the station-level communication network, the healthy single-units and / or system standby units in the cluster adjacent to the cluster where the damaged single-unit is located are determined. The available power compensation margin of the healthy single-units or standby units in the adjacent clusters is calculated according to the multivariate constraint minimum dynamic programming method, and corresponding power increment instructions are allocated to them to make up for the unfilled power deficit in the same cluster.

[0056] In addition, if the power deficit is still not fully compensated after compensation, the active power change rate constraint specified by the grid-connected dispatch is obtained. Based on the active power change rate constraint, the total active power output of the gravity energy storage system is adjusted downward by a limit slope to smoothly reduce the total output target value and avoid power surges from impacting the power grid.

[0057] At the same time, the current output status of each healthy unit is monitored in real time; when it is determined that the current output of any healthy unit has reached its maximum allowable output, it is prohibited to send an over-output command to that healthy unit, so as to limit the operation of each healthy unit within its rated safety boundary, prevent overcurrent and overtemperature protection tripping caused by overload compensation, and cut off the propagation chain of the fault range expansion.

[0058] In one embodiment, see Figure 5 The soft-switching buffer isolation circuit includes a main isolation contactor and a buffer energy dissipation branch. In terms of physical wiring, the main isolation contactor KM is connected in series to the three-phase output terminal of the generator stator of the single unit. Figure 5 On the main power path between the U-phase, V-phase, and W-phase of the generator and the AC side of the power converter, the buffer energy dissipation branch is connected in parallel with the three-phase output terminal of the generator stator.

[0059] like Figure 5 As shown, the buffer energy dissipation branch includes a three-phase rectifier bridge BR and a DC-side chopper switch T. br Energy-consuming resistor R br And the RC absorption branch (composed of resistor R) sn and capacitor C sn (connected in series), wherein the three-phase AC power from the generator stator is converted by the three-phase rectifier bridge BR and then connected to the DC power consumption circuit; the DC-side chopper switch T br With the energy-consuming resistor R br After being connected in series, it is connected in parallel between the positive and negative terminals of the DC output of the three-phase rectifier bridge BR; the RC absorption branch is connected in parallel across the chopper switch and the energy-consuming resistor to absorb high-frequency voltage spikes. The DC-side chopper switch T br Controllable turn-on and turn-off can be achieved by using insulated gate bipolar transistors (IGBTs), solid-state relays (SSRs), or thyristors.

[0060] When the isolation execution layer receives an isolation command for the damaged unit, the specific steps for achieving physical isolation through the soft-switching buffer isolation circuit are executed in the following sequence: first, the buffer power-dissipating branch is turned on, and then the main isolation contactor is turned off. First, the isolation execution layer sends a signal to the DC-side chopper switch T. brA closing trigger signal is sent to activate the buffer energy dissipation branch. At this time, the transient current on the generator stator side and the remaining induced electromagnetic energy are transferred to the buffer energy dissipation path, enter the DC side via the three-phase rectifier bridge BR, and then pass through the energy dissipation resistor R. br It is converted into heat energy and dissipated in a controlled manner.

[0061] After the buffer energy dissipation branch is turned on, the system monitors the stator current flowing through the main power path in real time. Once the current flowing through the main isolation contactor KM drops to a preset safety threshold (or approaches the current zero-crossing range), the isolation execution layer controls the disconnection of the main isolation contactor KM. After the main isolation contactor KM is completely disconnected and smooth disconnection is achieved, the isolation execution layer shuts down the DC-side chopper switch T. br This allows for the smooth physical isolation of damaged individual units.

[0062] As an alternative embodiment, the buffer energy dissipation branch can also dissipate energy directly on the AC side. Specifically, the buffer energy dissipation branch includes a bidirectional controllable energy dissipation branch on the AC side, consisting of bidirectional IGBT switches connected in series in each phase, anti-parallel thyristors, solid-state relays, or hybrid circuit breakers. This topology eliminates the need for a three-phase rectifier bridge and, by directly controlling the conduction angle or duty cycle of the bidirectional semiconductor switches on each phase AC branch, can also achieve bidirectional conduction and controllable dissipation of stator transient high current.

[0063] It is understood that although the steps in the flowcharts of the above embodiments are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the above embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0064] Based on the same inventive concept, this application also provides a selective ride-through protection system for electromechanical decoupling and coupling faults in a gravity energy storage system as described above. The solution provided by this system is similar to the solution described in the above method; therefore, the specific limitations in one or more system embodiments provided below can be found in the limitations of the method described above, and will not be repeated here.

[0065] In one exemplary embodiment, such as Figure 6As shown, a selective ride-through protection system for electromechanical decoupling and coupling faults in a gravity energy storage system is provided, including: a data acquisition module 1, a map construction module 2, a decoupling determination module 3, a severity assessment module 4, an isolation control module 5, a soft-switching buffer isolation circuit 6, and a power compensation module 7.

[0066] The above system corresponds to the aforementioned method embodiments and can implement the corresponding method steps. Its implementation principle and technical effect are similar, and will not be repeated here.

[0067] Please see Figure 7 This application also provides a computer device. The computer device 90 may include a processor 91, a memory 92, and computer applications, wherein: The memory 92 is used to store the computer application, and the memory may also be flash memory. The computer application is, for example, an application that implements the various method embodiments described above.

[0068] Processor 91 is configured to execute the computer application stored in the memory to implement the steps in the various method embodiments described above. For details, please refer to the relevant descriptions in the preceding method embodiments.

[0069] Alternatively, the memory 92 can be either standalone or integrated with the processor 91.

[0070] When the memory 92 is a device independent of the processor 91, the computer device 90 may further include: Bus 93 is used to connect the memory 92 and the processor 91.

[0071] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0072] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein, and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A method for selective ride-through protection against electromechanical decoupling and coupling faults in a gravity energy storage system, wherein the gravity energy storage system comprises multiple clusters, each cluster comprising multiple single-unit devices, characterized in that... include: Acquire mechanical status data and electrical operation data for each of the single units, wherein the mechanical status data includes at least the mechanical vibration signal of the winch gearbox collected by the mechanical sensing layer, and the electrical operation data includes at least the voltage and current signals of the power converter grid side collected by the electrical sensing layer. Mechanical features of the mechanical state data and electrical features of the electrical operation data are extracted respectively, and the coupling weight between the mechanical features and the electrical features is calculated to construct a directed deep fault association map; When abnormal fluctuations are detected in the gravity energy storage system, causal evolution Bayesian inference is initiated based on the deep fault correlation map. The temporal sequence of the abrupt changes in the mechanical features and the electrical features is compared. The first rate of change corresponding to the mechanical features and the second rate of change corresponding to the electrical features are normalized based on their respective steady-state reference values ​​to obtain dimensionless first and second relative rates of change. The causal evolution Bayesian inference uses the temporal sequence and the comparison results of the first and second relative rates of change as input conditions. Based on the directed topology of the deep fault correlation map, conditional probability derivation is performed to quantitatively output the fault root cause confidence level indicating the root cause of the abnormal fluctuations. When the fault root cause confidence level reaches a preset confidence threshold, a decoupling determination is made to determine whether the abnormal fluctuations are electromechanical coupling faults caused by physical damage to mechanical equipment. If the abnormal fluctuation is determined to be an electromechanical coupling fault, the severity of the electromechanical coupling is assessed based on the deep fault correlation map. When the severity reaches the single-machine isolation threshold, an isolation command is sent to the isolation execution layer, and physical isolation is achieved through a soft-switching buffer isolation circuit based on the isolation command. While performing physical isolation, the damaged unit that experienced the electromechanical coupling fault and the healthy units in the same cluster as the damaged unit are identified. The power deficit caused by the disconnection of the damaged unit from the grid is calculated. The power deficit is the active power undertaken by the damaged unit before disconnection from the grid. At the same time, the available power compensation margin of each healthy unit is checked in real time. When the sum of the available power compensation margins of all healthy single-units in the same cluster is greater than or equal to the power deficit, power increments are allocated to each healthy single-unit according to each available power compensation margin and corresponding power increment commands are issued to control each healthy single-unit to work together to compensate for the power deficit, so that the total output power of the gravity energy storage system grid connection point is smoothly transitioned.

2. The electromechanical decoupling and selective ride-through protection method for gravity energy storage systems according to claim 1, characterized in that, The mechanical features of the mechanical state data and the electrical features of the electrical operation data are extracted respectively, and the coupling weights between the mechanical features and the electrical features are calculated to construct a directed deep fault association map, including: The mechanical vibration features are extracted from the mechanical state data as the mechanical features; The total harmonic distortion (THD) distortion features in the electrical operating data are extracted as the electrical features. The edge weights between the mechanical vibration characteristics and the total harmonic distortion (THD) distortion characteristics are calculated using the mutual information method, the Pearson correlation coefficient method, or the Jacobian matrix based on the mechanism equation as coupling weights. A directed deep fault association graph is constructed based on the coupling weights.

3. The electromechanical decoupling and selective ride-through protection method for gravity energy storage systems according to claim 1, characterized in that, Using the temporal sequence relationship and the comparison results of the first relative change rate and the second relative change rate as input conditions, conditional probability derivation is performed based on the directed topology of the deep fault association map, and the confidence of the fault root cause indicating the root cause of abnormal fluctuations is quantitatively output. When the confidence level of the fault root cause reaches a preset confidence threshold, the decoupling determination is made to determine whether the abnormal fluctuation is an electromechanical coupling fault caused by physical damage to mechanical equipment, including: The abrupt change times of the mechanical features and the electrical features are identified by using wavelet modulus maxima detection or CUSUM abrupt change detection, respectively. If the abrupt change of the mechanical feature precedes the abrupt change of the electrical feature and the time difference between the two is greater than a preset discrimination dead zone: If the first relative rate of change is greater than the second relative rate of change, such that the confidence of the fault root cause of physical damage to mechanical equipment output by the causal evolution Bayesian inference is greater than the preset confidence threshold, then the abnormal fluctuation is determined to be an electromechanical coupling fault caused by physical damage to mechanical equipment, and a first electromechanical decoupling protection command is output. If the first relative rate of change is less than or equal to the second relative rate of change, then based on the deep fault correlation map, the conditional probability distribution is corrected. When the confidence level of reasoning that points to the electrical side's own abnormality is greater than the preset confidence threshold, the abnormal fluctuation is determined to be the electrical side's own abnormality, and the second electromechanical decoupling protection command is output. When the confidence level of the fault root cause indicating physical damage to mechanical equipment and the confidence level of the fault indicating electrical side abnormality are both less than the preset confidence threshold, it is determined that the current electromechanical coupling causal evolution is in an uncertain state, a review instruction is output, and the current output confidence level is used as a feedback quantity to be input into the deep fault association map to adaptively correct the coupling weight, and the next inference cycle is entered for review.

4. The electromechanical decoupling and selective ride-through protection method for gravity energy storage systems according to claim 3, characterized in that, Also includes: If the abrupt change of the electrical characteristic occurs before the abrupt change of the mechanical characteristic, and the second rate of change corresponding to the electrical characteristic is greater than the first rate of change corresponding to the mechanical characteristic, then the decoupling determination indicates that the abnormal fluctuation is caused by an external power grid fault. The isolation command issued to the isolation execution layer is blocked, the soft switch buffer isolation circuit is prohibited from operating, all the single units in the station are kept in physical grid-connected state, and each single unit is controlled to switch to the preset voltage ride-through control mode to provide voltage support to the power grid.

5. The electromechanical decoupling and selective ride-through protection method for gravity energy storage systems according to claim 1, characterized in that, The severity of the electromechanical coupling fault is assessed based on the deep fault correlation map, including: Based on the coupling weights corresponding to the deep fault association map, the electromechanical coupling faults are classified into mild faults, moderate faults, or severe faults. When the severity reaches the single-machine isolation threshold, the electromechanical coupling fault is classified as a moderate fault. The method further includes: If it is a minor fault, derating ride-through control is executed to reduce the power limit of the damaged unit and allow it to continue operating; In the case of a severe fault, the damaged unit and all healthy units in the same cluster are synchronously isolated, and a station-level interlock is triggered.

6. The electromechanical decoupling and selective ride-through protection method for gravity energy storage systems according to claim 1, characterized in that, The real-time verification of the available power compensation margin of the healthy stand-alone unit includes: The current output, maximum allowable output, ramp rate limit, mechanical travel, temperature rise constraint, and reserve capacity of the healthy single unit are obtained to jointly determine the available power compensation margin.

7. The electromechanical decoupling and selective ride-through protection method for gravity energy storage systems according to claim 1, characterized in that, The method further includes: If the sum of the available power compensation margins of all the healthy single-units in the same cluster is less than the power deficit, then the healthy single-units and / or system backup units in the adjacent clusters are called up to perform power compensation in order to make up for the power deficit. If the power deficit is still not made up after compensation, the total active power output of the gravity energy storage system will be reduced by a limited slope according to the active power change rate constraint specified by the grid-connected dispatch, and sending excess output commands to healthy single units that are already at full load will be prohibited.

8. The electromechanical decoupling and selective ride-through protection method for gravity energy storage systems according to claim 1, characterized in that, The soft-switching buffer isolation circuit includes a main isolation contactor and a buffer energy dissipation branch. The main isolation contactor is connected in series between the generator stator three-phase output terminal and the AC side of the power converter of the single unit. The buffer energy dissipation branch is connected in parallel with the generator stator three-phase output terminal. Physical isolation is achieved through a soft-switching buffer isolation circuit, including: following the timing sequence of first turning on the buffer energy dissipation branch and then turning off the main isolation contactor, so that the stator-side transient current of the damaged unit is dissipated through the buffer energy dissipation branch to achieve smooth physical isolation; The buffer energy dissipation branch includes a three-phase rectifier bridge, a DC-side chopper switch, an energy dissipation resistor, and an RC snubber branch, wherein the DC-side chopper switch is an insulated gate bipolar transistor (IGBT), a solid-state relay, or a thyristor. Alternatively, the buffer energy dissipation branch may include an AC-side bidirectional controllable energy dissipation branch consisting of bidirectional IGBT switches, anti-parallel thyristors, solid-state relays, or hybrid circuit breakers connected in series in each opposite direction.

9. A selective ride-through protection system for electromechanical decoupling and coupling faults in a gravity energy storage system, applied to a gravity energy storage system comprising multiple clusters, each cluster comprising multiple individual units, characterized in that, For implementing the method as described in any one of claims 1-8, comprising: The data acquisition module is used to acquire the mechanical status data and electrical operation data of each of the single units, wherein the mechanical status data includes at least the mechanical vibration signal of the winch gearbox collected by the mechanical sensing layer, and the electrical operation data includes at least the voltage signal and current signal of the power converter grid side collected by the electrical sensing layer. The graph construction module is used to extract the mechanical features of the mechanical state data and the electrical features of the electrical operation data respectively, and calculate the coupling weight between the mechanical features and the electrical features to construct a directed deep fault association graph. The decoupling determination module is used to initiate causal evolution Bayesian inference based on the deep fault correlation map when abnormal fluctuations are detected in the gravity energy storage system. It compares the temporal sequence of abrupt changes in the mechanical features and the electrical features, and normalizes the first rate of change corresponding to the mechanical features and the second rate of change corresponding to the electrical features based on their respective steady-state reference values ​​to obtain dimensionless first and second relative rates of change. Using the temporal sequence and the comparison results of the first and second relative rates of change as input conditions, it performs conditional probability derivation based on the directed topology of the deep fault correlation map, quantitatively outputting the confidence level of the fault root cause indicating the source of the abnormal fluctuations. When the confidence level of the fault root cause reaches a preset confidence threshold, it decouples and determines whether the abnormal fluctuations are caused by electromechanical coupling faults resulting from physical damage to mechanical equipment. The severity assessment module is used to assess the severity of the electromechanical coupling fault based on the deep fault correlation map when the abnormal fluctuation is determined to be an electromechanical coupling fault. An isolation control module is used to issue an isolation command to the isolation execution layer when the severity meets a preset condition, so that the isolation execution layer performs physical isolation through a soft-switching buffer isolation circuit based on the isolation command; A soft-switching buffer isolation circuit is provided, comprising a main isolation contactor and a buffer energy dissipation branch. The main isolation contactor is connected in series between the three-phase output terminal of the generator stator of the single unit and the AC side of the power converter. The buffer energy dissipation branch is connected in parallel with the three-phase output terminal of the generator stator. The isolation execution layer controls the soft-switching buffer isolation circuit to perform physical isolation based on the isolation command. The power compensation module is used to determine the damaged single unit that has experienced the electromechanical coupling fault and the healthy single unit in the same cluster as the damaged single unit while performing the physical isolation, calculate the power deficit caused by the disconnection of the damaged single unit from the grid, the power deficit being the active power undertaken by the damaged single unit before disconnection, and check the available power compensation margin of each healthy single unit in real time. And when the sum of the available power compensation margins of all healthy single units in the same cluster is greater than or equal to the power deficit, power increments are allocated to each healthy single unit according to the available power compensation margin of each healthy single unit and corresponding power increment commands are issued to control each healthy single unit to work together to increase output and compensate for the power deficit, so that the total output power of the gravity energy storage system grid connection point is smoothly transitioned.

10. A computer device, characterized in that, It includes a memory and a processor, the memory storing a computer application, and the processor executing the computer application, wherein the computer application is used to implement the method as described in any one of claims 1-8 when executed by the processor.

Images