Energy storage power station early warning method and device based on particle fault detection

Abstract

The present invention discloses an energy storage power station early warning method and device based on microparticle fault detection, which relates to the technical field related to energy storage power stations. The method includes: setting a deployment network of microparticle probes for the power station energy storage equipment and installing a detection array; generating a fault discriminator by constructing a fault particle fingerprint library and integrating it into the energy storage management system; performing microparticle detection based on the detection array, triggering a two-stage decision to determine the fault data chain, and the energy storage management system responding to the fault data chain to perform directional early warning management; wherein, based on the fault data chain, multiple detection guidance based on the detection array is performed, and fault verification is performed by introducing associated gas. This solves the technical problems existing in the prior art, such as insufficient detection network coverage, weak correlation of fault mechanisms, and lack of dynamic verification means, which lead to poor accuracy and reliability of energy storage power station fault diagnosis and early warning, and achieves the technical effect of improving the accuracy and reliability of energy storage power station fault diagnosis and early warning.

Description

Technical Field

[0001] The present application relates to the technical field related to energy storage power stations, and specifically to an energy storage power station early warning method and device based on particle fault detection. Background Art

[0002] Energy storage power stations are critical facilities for peak shaving, frequency regulation, and emergency power supply in power systems, and their safety and reliability are crucial. However, energy storage devices, especially lithium-ion batteries, can fail during operation due to problems such as thermal runaway, internal short circuits, and mechanical aging, potentially leading to fires or explosions. Traditional fault monitoring relies primarily on threshold alarms for macroscopic parameters such as voltage, current, and temperature. These warnings often only trigger after the fault has reached an irreversible stage, making it difficult to detect early hidden dangers. Furthermore, fault warnings based on particle detection struggle to cover the complex spatial distribution within the battery compartment, resulting in incomplete capture of particle motion characteristics. The mapping relationship between particle data and fault types lacks systematic modeling, resulting in low fault monitoring accuracy and high misjudgment rates.

[0003] Therefore, at the current stage, relevant technologies have technical problems such as insufficient detection network coverage, weak correlation of fault mechanisms, and lack of dynamic verification methods, which lead to poor accuracy and reliability of fault diagnosis and early warning of energy storage power stations. Summary of the Invention

[0004] This application provides an energy storage power station early warning method and device based on particle fault detection, which solves the technical problems existing in the prior art, such as insufficient detection network coverage, weak correlation of fault mechanisms, and lack of dynamic verification means, which lead to poor accuracy and reliability of energy storage power station fault diagnosis and early warning, and achieves the technical effect of improving the accuracy and reliability of energy storage power station fault diagnosis and early warning.

[0005] The present application provides an energy storage power station early warning method based on particle fault detection. The method includes: setting a deployment network of particle probes for the power station energy storage equipment, installing a detection array in the battery compartment of the power station energy storage equipment, wherein a first deployment is performed based on the particle motion characteristics of the top-sidewall-bottom, and a second deployment is performed based on the division of spatial safety boundaries; by constructing a fault particle fingerprint library, defining a first judgment node based on the deployment network, and defining a second inference node based on the reverse inference of particle characteristics and fault mechanisms, generating a fault discriminator and integrating it into an energy storage management system; performing particle detection based on the detection array, triggering a two-order decision based on the fault discriminator, determining a fault data chain, and the energy storage management system performing directional early warning management in response to the fault data chain; wherein, based on the fault data chain, performing complex detection guidance based on the detection array, and introducing associated gas for fault verification.

[0006] In a possible implementation, the energy storage power station early warning method based on particle fault detection further performs the following processing: defining the number of particle nodes based on the installed capacity of the energy storage equipment, wherein the number of particle nodes is an interval dimension; performing deployment based on the number of particle nodes by top scanning, sidewall networking, and bottom deposition to determine a first deployment method; determining a second deployment method based on a particle safety boundary based on the fault source; and determining the deployment network by fusing the first deployment method and the second deployment method.

[0007] In a possible implementation, the energy storage power station early warning method based on particle fault detection further performs the following processing: retrieving fault records of the power station energy storage equipment, performing homologous clustering to mine intra-class fault sequences, wherein the intra-class fault sequences at least include a fault source and a fault mechanism; defining a relative safety boundary based on the intra-class fault sequence, wherein there is a corresponding relationship between the relative safety boundary and the intra-class fault sequence; and deploying particle nodes within the relative safety boundary to determine the second deployment method.

[0008] In a possible implementation, the energy storage power station early warning method based on particle fault detection further performs the following processing: according to the fault mechanism, determining the intrinsic characteristics and evolutionary characteristics of the particles, and defining the node window size; wherein the intrinsic characteristics of the particles at least include the granularity range; taking the intersection of the first deployment method and the second deployment method, determining the node network and performing the detection domain expansion based on the node window size as the deployment network.

[0009] In a possible implementation, the energy storage power station early warning method based on particle fault detection also performs the following processing: installing a detection array in the battery compartment of the power station energy storage equipment, wherein the detection array is composed of nano-particle detectors; based on the deployment network, the non-intersection of the detection path is used to constrain the deployment position of the detection array in the compartment.

[0010] In a possible implementation, the energy storage power station early warning method based on particle fault detection further performs the following processing: determining the fault particle fingerprint elements by integrating the intra-class fault sequence, wherein the fault particle fingerprint elements at least include the fault type-characteristic particles-size distribution-motion characteristics-associated gas; extracting the first threshold feature and the second mechanism feature based on the fault particle fingerprint elements, wherein the first threshold feature is a normal critical value based on the direct detection information of the particles; constructing the fault particle fingerprint library based on the fault particle fingerprint elements, the first threshold feature and the second mechanism feature.

[0011] In a possible implementation, the energy storage power station early warning method based on particle fault detection further performs the following processing: constructing a discrimination block by building in the fault particle fingerprint library; and performing supervised training on the discrimination block based on the adversarial principle to generate the fault discriminator, wherein the particle detection data is used as input, the threshold discrimination-mechanism deduction trigger-mechanism deduction execution is used as the intermediate logic, and the fault data is used as the output.

[0012] In a possible implementation, the energy storage power station early warning method based on particle fault detection further performs the following processing: the fault discriminator includes three-order training; wherein the three-order training method includes: using a discriminant-generative architecture, performing first-order training based on threshold discrimination, and using the discriminant architecture as the first judgment node; using a discriminant-generative architecture, performing second-order training based on mechanism deduction, and using the generative architecture as the second inference node; performing migration calls and third-order training based on node cascade triggering on the first judgment node and the second inference node to determine the fault discriminator.

[0013] In a possible implementation, the energy storage power station early warning method based on particle fault detection further performs the following processing: directional particle detection of the power station energy storage equipment is performed according to the detection array to determine the detection array; the energy storage management system receives the detection array, triggers the first judgment node to perform threshold feature matching and binary classification judgment, and determines the first judgment result; if the first judgment result is abnormal, triggers the second inference node to perform knowledge reasoning based on fault mechanism tracing, determines the fault sequence, and adds it to the fault data chain.

[0014] The present application also provides an energy storage power station early warning device based on particle fault detection, the device comprising: a detection array installation module, configured to set a deployment network of particle probes for power station energy storage equipment, and install a detection array within a battery compartment of the power station energy storage equipment, wherein a first deployment is performed based on the particle motion characteristics of the top-sidewall-bottom, and a second deployment is performed based on the division of spatial safety boundaries; a fault discriminator generation module, configured to define a first determination node based on the deployment network by constructing a fault particle fingerprint library, and a second inference node based on the reverse inference of particle characteristics and fault mechanisms, thereby generating a fault discriminator and integrating it into an energy storage management system; a fault data chain determination module, configured to perform particle detection based on the detection array, trigger a two-order decision based on the fault discriminator, determine a fault data chain, and the energy storage management system perform directional early warning management in response to the fault data chain; and a fault verification module, configured to perform complex detection guidance based on the detection array based on the fault data chain, and perform fault verification by introducing associated gas.

[0015] The energy storage power station early warning method and device based on microparticle fault detection proposed in this application is intended to set up a deployment network of microparticle probes and install a detection array for the power station energy storage equipment; by constructing a fault particle fingerprint library, a fault discriminator is generated and built into the energy storage management system; microparticle detection is performed based on the detection array, triggering a two-stage decision to determine the fault data chain, and the energy storage management system responds to the fault data chain to perform directional early warning management; wherein, based on the fault data chain, repeated detection guidance based on the detection array is performed, and fault verification is performed by introducing associated gas. This solves the technical problems existing in the prior art, such as insufficient detection network coverage, weak correlation of fault mechanisms, and lack of dynamic verification methods, which lead to poor accuracy and reliability of energy storage power station fault diagnosis and early warning, and achieves the technical effect of improving the accuracy and reliability of energy storage power station fault diagnosis and early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] To more clearly illustrate the technical solutions of the embodiments of the present disclosure, the accompanying drawings of the embodiments of the present disclosure are briefly introduced below. Flowcharts are used in this application to illustrate the operations performed by the apparatus according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in precise order. Instead, various steps may be processed in reverse order or simultaneously as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0017] Figure 1 A flow chart of an energy storage power station early warning method based on particle fault detection provided in an embodiment of the present application.

[0018] Figure 2 Schematic diagram of the structure of an energy storage power station early warning device based on particle fault detection provided in an embodiment of the present application.

[0019] Description of reference numerals: detection array installation module 10 , fault discriminator generation module 20 , fault data link determination module 30 , fault verification module 40 . DETAILED DESCRIPTION

[0020] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0021] In order to make the purpose, technical solutions and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limiting this application. All other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0022] In the following description, reference is made to “some embodiments”, which describes a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms “first\second” involved are merely to distinguish similar objects and do not represent a specific ordering of the objects. The terms “including” and “having” and any variations are intended to cover non-exclusive inclusions, for example, a process, method, device, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs. The terms used herein are for the purpose of describing the embodiments of this application only.

[0023] The embodiment of the present application provides an energy storage power station early warning method based on particle fault detection, such as Figure 1 As shown, the method includes:

[0024] In step S100, a deployment network of microparticle probes is set for the power station energy storage equipment, and a detection array is installed in the battery compartment of the power station energy storage equipment. The first deployment is based on the particle movement characteristics of the top-sidewall-bottom, and the second deployment is based on the division of the spatial safety boundary.

[0025] Preferably, for the battery compartment of the power station energy storage equipment, a deployment network of microparticle probes is set up through a first deployment based on the top-sidewall-bottom particle movement characteristics and a second deployment based on the spatial safety boundary division, and a detection array is installed in the battery compartment to achieve all-round and high-precision particle fault monitoring. The deployment network is composed of multiple microparticle detection units, which are arranged in the battery compartment according to a specific rule to form a detection array covering the entire compartment. The probe of each detection node is an independent sensing unit that can detect the concentration, particle size, composition and other characteristics of the particles. The probes are connected by wired / wireless means to transmit data to the central management system in real time, and each probe has a clear spatial coordinate to facilitate fault tracing.

[0026] Preferably, the first deployment is mainly based on the movement laws of particles under different fault modes, such as diffusion direction, sedimentation velocity, airflow influence, etc., and detection arrays are arranged at key positions in the battery compartment to ensure that particle signals from different sources can be captured. Among them, the detection array deployed on the top is used to monitor light particles, such as nano-particles and aerosols produced by the decomposition of electrode materials, which can identify early thermal runaway, electrolyte volatilization and other faults; the detection array deployed on the side wall is used to monitor medium and low-speed diffusion particles, such as metal debris and diaphragm damaged particles, which can identify mechanical aging, internal short circuit, local overheating and other faults; the detection array deployed on the bottom is used to monitor heavy sedimentation particles, such as large particles detached from the electrode and metal lithium dendrite debris, which can identify faults such as battery structure damage. In addition, the combination of probes at different heights and directions can capture particles with different physical properties and cover the entire movement trajectory of particles to avoid missed detection.

[0027] Preferably, the second deployment arranges detection arrays at key locations in the battery compartment and adjusts probe density and sensitivity based on the spatial safety level of the battery compartment, such as high-risk areas, buffer zones, and safe zones, to optimize detection efficiency and capture particle signals in real time. The high-risk area refers to the area near the battery module. High-density probes and high-sensitivity monitoring are used to quickly capture particles abnormally released by battery cells, achieving early warning. The buffer zone refers to the gaps or ventilation paths of power station energy storage equipment. Medium-density probes are used to monitor the diffusion trend of particles to determine whether the fault has spread to adjacent battery packs or other areas in the compartment. The safe zone refers to the walls or entrances and exits of the battery compartment. Low-density probes are used to monitor the environment to eliminate external dust interference and ensure data reliability. By deploying detection in different zones for zoned monitoring, the cost of global uniform deployment is avoided. Changes in particle distribution can also be used to determine the scope of fault impact, thereby providing graded warnings for the battery compartments of power station energy storage equipment and improving the safety management level of energy storage power stations.

[0028] Furthermore, step S100 also includes step S110, defining the number of microparticle nodes based on the installed capacity of the energy storage device, wherein the number of microparticle nodes is an interval dimension; step S120, performing deployment based on the number of microparticle nodes by top scanning, sidewall networking, and bottom deposition to determine a first deployment method; step S130, determining a second deployment method based on a microparticle safety boundary based on a fault source; and step S140, fusing the first deployment method and the second deployment method to determine the deployment network.

[0029] Preferably, the battery compartment size of the energy storage power station, such as 1MWh / 10MWh / 100MWh, directly affects the diffusion range of fault particles and the monitoring complexity. The number of particle node probes is quantified by the installed capacity, and the number of particle nodes is configured using the interval dimension to adapt to the battery compartment size. For example, small energy storage equipment corresponds to 20 to 30 probes per compartment in the low-density interval, medium-sized energy storage equipment corresponds to 50 to 100 probes per compartment in the medium-density interval, and large energy storage equipment corresponds to 100 to 200 probes in the high-density interval. Then, based on the particle movement, the nodes are deployed hierarchically according to their functions. , including a top scanning layer, a sidewall networking layer and a bottom deposition layer. The top scanning layer uses a wide-angle, high-sensitivity probe to cover the cabin roof with a sparse array to capture light particles that rise during thermal runaway, such as smoke and aerosols; the sidewall networking layer forms a grid-like, dense deployment in the gaps between battery modules to monitor horizontally diffused fault particles, such as metal dust and electrolyte vapor; the bottom deposition layer is a high-precision deposition sensor laid on the bottom of the cabin to detect settled heavy particles, such as electrode debris and precipitated metal, and the particle node density is dynamically adjusted by the interval dimension corresponding to the installed capacity.

[0030] Preferably, the battery compartment is partitioned based on the microparticle safety boundaries of the fault source, resulting in core risk areas (high-risk areas such as battery modules and high-voltage connections); secondary risk areas (such as heat dissipation channels and cabin support structures); and low-risk areas (such as hatches and maintenance passages). A second deployment method is then determined based on the fault source risk zoning of the battery compartment. When a risk zone experiences frequent failures, its safety boundary level is automatically upgraded and microparticle nodes are added, achieving spatial risk-adaptive deployment of energy storage equipment and ensuring that microparticle probe resources are tilted towards high-risk areas. Finally, the first and second deployment methods are integrated, i.e., basic nodes are allocated according to the top-sidewall-bottom layout of the first deployment, and node density and type are adjusted according to the safety boundaries of the second deployment. This determines the deployment network, achieving full spatial monitoring coverage of the battery compartment of the power station's energy storage equipment, and improving the accuracy and speed of fault location.

[0031] Furthermore, step S130 also includes step S131, retrieving fault records of the power station energy storage equipment and performing homologous clustering to mine intra-class fault sequences, wherein the intra-class fault sequences at least include a fault source and a fault mechanism; step S132, defining a relative safety boundary based on the intra-class fault sequence, wherein there is a corresponding relationship between the relative safety boundary and the intra-class fault sequence; and step S133, deploying micro-nodes within the relative safety boundary to determine the second deployment mode.

[0032] Preferably, historical fault records of the energy storage equipment of the power station are retrieved from the energy storage power station operation and maintenance database, such as thermal runaway, short circuit, electrolyte leakage and other events. Each fault record contains the fault source and fault mechanism, wherein the fault source includes the fault location and equipment type, and the fault mechanism refers to the root cause of the fault, such as dendrite puncture, diaphragm aging, and cooling failure; then the fault records are subjected to homologous clustering, that is, unsupervised clustering is used to classify the faults according to homologous features such as fault location, triggering conditions, and accompanying particle types to obtain intra-class fault sequences, wherein each intra-class fault sequence is associated with particle features. Next, based on the intra-class fault sequences, the spatial distribution patterns of the fault sources are analyzed. For example, 80% of thermal runaways occur in the third module at the rear of the battery compartment. Combined with the propagation paths of the fault mechanisms, such as the diffusion of thermal runaway particles along the airflow toward the top, particle concentration exceeding a certain range, or particle dissipation exceeding a specific range, dynamic safety boundaries (relative safety boundaries) are defined. These boundaries may include a core boundary (high-frequency fault occurrence area), a secondary boundary (fault impact diffusion path), an observation boundary, and a potential chain reaction area. Each relative safety boundary is associated with an intra-class fault sequence. For example, the core boundary corresponds to the cathode lithium plating-lithium particle release fault sequence, while the secondary boundary corresponds to the electrolyte leakage-vapor diffusion fault sequence. Finally, particle nodes are deployed based on the relative safety boundaries. For example, high-density multimodal probes are deployed at the core boundary, airflow tracking probes are deployed at the secondary boundary, and low-power monitoring nodes are deployed at the observation boundary. This determines the second deployment method, optimizing probe resource utilization while improving the accuracy and reliability of fault warnings for power plant energy storage equipment.

[0033] Furthermore, step S140 also includes step S141, determining the intrinsic characteristics and evolution characteristics of the particles according to the fault mechanism, and defining the node window size; wherein the intrinsic characteristics of the particles at least include the granularity range; step S142, taking the intersection of the first deployment method and the second deployment method, determining the node network and performing the detection domain expansion based on the node window size as the deployment network.

[0034] Preferably, the intrinsic characteristics of the particles are determined based on the type of particles associated with the fault mechanism, including at least the particle size range. For example, soot particles released from thermal runaway are 0.1-10 μm, metal debris generated by diaphragm rupture is 1-100 μm, and aerosols from electrolyte leakage are 0.01-1 μm. The density of metal particles is greater than that of carbon particles, affecting the settling velocity. Electrode material debris often carries static charge, affecting adsorption behavior. The diffusion path is then determined as the particle evolution characteristic, including the rise of light particles such as aerosols with the airflow and the sedimentation of heavy particles such as metal chips to the bottom. The probe's detection domain size is then dynamically adjusted based on the intrinsic and particle evolution characteristics, that is, the node window size is defined. Specifically, a small window of 0.1-1 μm is deployed in the top scanning layer to monitor early aerosol faults; a medium window of 1-50 μm is covered in the sidewall network layer for routine metal dust monitoring; and a large window of 50-500 μm is configured in the bottom deposition layer to capture electrode debris deposits, thereby improving the signal capture efficiency at different fault stages.

[0035] Preferably, the intersection of the first deployment mode and the second deployment mode is taken, that is, high-priority nodes are deployed in the overlapping area of ​​the two types of deployment, such as the intersection of the top scanning layer and the core boundary, and the window size is adapted, wherein the window size of different sources is different; then the node network is formed with the intersection nodes of the first and second deployments, and then the detection domain expansion based on the node window size is performed. Specifically, the spatial detection domain expansion of the node window size is performed according to the window size requirements, that is, additional nodes are added on the particle diffusion path, and the functional detection domain is expanded, and multiple window probes are superimposed at the same position to finally form a deployment network. It can also be dynamically adjusted according to the number of abnormal particles or the environment. For example, in a high temperature environment, the sampling frequency of the top small window node is increased to capture early aerosols, thereby providing underlying perception protection for the safety of power station energy storage equipment.

[0036] Furthermore, step S100 further includes step S150, installing a detection array in the battery compartment of the power station energy storage equipment, wherein the detection array is composed of nano-particle detectors; step S160, based on the deployment network, constraining the deployment position of the detection array in the compartment with the non-intersection of the detection path.

[0037] Preferably, multiple nanoparticle detectors are assembled into a detection array and deployed within the battery compartment of a power plant's energy storage equipment. These nanoparticle detectors, including laser scattering and electrostatic adsorption types, interact with particles by emitting detection signals (light waves or electric fields). Their effective detection range takes the form of directional paths, such as conical beams or fan-shaped electric fields. If the paths of two nanoparticle detectors intersect, particles in the intersection area may be double-counted, leading to concentration misjudgments and signal intensities canceling out in the intersecting paths. A deployment network, a fusion of the first and second deployment methods, then constrains the in-cabin deployment positions of the detection array based on the non-intersecting detection paths. Specifically, detectors on the same layer must ensure non-overlapping beams, and detectors on different layers must be staggered to avoid vertical path intersections. Furthermore, in the core risk area of ​​the second deployment method, the spacing can be appropriately reduced, but dynamic power adjustment is required. If intersecting paths cannot be avoided, a time-slice polling mechanism is used for monitoring. This ensures the fidelity of the detection signal and the accuracy of fault location in the power plant's energy storage equipment.

[0038] Step S200 : By constructing a fault particle fingerprint library, defining a first determination node based on the deployment network, defining a second inference node based on particle characteristics-fault mechanism inversely, generating a fault discriminator and embedding it in the energy storage management system.

[0039] Step S200 further includes step S210, determining the fault particle fingerprint elements by integrating the intra-class fault sequence, wherein the fault particle fingerprint elements at least include fault type-characteristic particles-size distribution-motion characteristics-associated gas; step S220, extracting the first threshold feature and the second mechanism feature using the fault particle fingerprint elements, wherein the first threshold feature is a normal critical value based on direct detection information of particles; step S230, constructing the fault particle fingerprint library using the fault particle fingerprint elements, the first threshold feature and the second mechanism feature.

[0040] Preferably, the fault sequences within the class are integrated, that is, the data of the same type of fault sequences are aligned, and the five-dimensional characteristics of the fault are extracted to form the fault particle fingerprint elements, wherein the fault particle fingerprint elements at least include the fault type-characteristic particles-size distribution-motion characteristics-associated gases. Specifically, the fault type is determined according to the cluster classification label, such as thermal runaway, short circuit, and electrolyte leakage; characteristic particles are the signature particles released by the fault, such as carbon soot particles released by thermal runaway and aluminum / copper debris released by diaphragm rupture; size distribution refers to the particle size range and distribution of the particles, such as 0.1-1μm nanoparticles in the early stage and 1-100μm agglomerates in the later stage; motion characteristics refer to the diffusion direction and speed of the particles, such as vertical rise, horizontal diffusion or sedimentation; associated gases refer to the associated gas type and concentration threshold.

[0041] Preferably, the first threshold feature and the second mechanism feature are extracted from the fault particle fingerprint element, wherein the first threshold feature is a normal critical value based on the direct detection information of the particle, that is, a quantitative boundary value for distinguishing normal and abnormal conditions based on the concentration, particle size, and velocity data of the particle detector, which is used for rapid screening of abnormalities; the second mechanism feature is the association rule between particle behavior and fault mechanism extracted in combination with the fault evolution law, which is used for deep fault diagnosis. Finally, the fault particle fingerprint element, the first threshold feature, and the second mechanism feature are integrated into a structured database, namely the fault particle fingerprint library, wherein the base layer stores the original fingerprint elements, and the rule layer stores the associated threshold and mechanism features, which are used to dynamically update new fault modes in real time. When the detector captures the particle signal, it quickly matches the fault particle fingerprint library to achieve high-precision fault diagnosis and early warning.

[0042] Preferably, by constructing a fault particle fingerprint library, a first determination node is defined based on the deployment network, that is, the first determination node is configured according to the key monitoring points of the detection network as a rapid screening layer, and the detector data is compared with the first threshold characteristics in the fault particle fingerprint library in real time to output a binary judgment of normal or abnormal; then the second inference node is defined based on the reverse reasoning of particle characteristics-fault mechanism. Specifically, when the first determination node detects an abnormality, the second inference node is defined to perform reverse reasoning of particle characteristics-fault mechanism, for example, integrating particle size distribution, motion trajectory characteristics and associated gas combination, and then matching the fault mechanism and performing fault type probability calculation to locate the fault source; then a fault discriminator is generated and used as a functional module of the energy storage management system, which may decide to output four levels of warning signals, such as observation level, single indicator exceeds the threshold; warning level, multiple indicator combination is abnormal; emergency level, mechanism matching confirms the fault; disaster level, chain reaction prediction, thereby ensuring the accuracy of energy storage equipment fault diagnosis and fault recognition rate.

[0043] Furthermore, step S200 also includes step S240, constructing a discrimination block by building the fault particle fingerprint library; step S250, supervising the discrimination block with the adversarial principle to generate the fault discriminator, wherein the particle detection data is used as input, the threshold discrimination-mechanism deduction trigger-mechanism deduction execution is used as the intermediate logic, and the fault data is used as the output.

[0044] Preferably, a feature matrix block is constructed based on the fault particle fingerprint library as a discrimination block, wherein the row vector stores the fingerprint features of multiple fault modes, and the column vector records the physical constraints of each feature, such as the concentration gradient threshold and the time delay parameter; the discrimination block is then supervised and trained based on the adversarial principle, that is, a generator and a discriminator are configured, the generator is used to simulate the variation of fault particle data, including abnormal data under extreme working conditions and adversarial samples of new faults; the discriminator is used to judge the authenticity of the data and identify the real fault mode; specifically, the generator attempts to construct a single feature anomaly, such as only the particle size exceeds the standard, and the discriminator directly intercepts it through threshold discrimination; the generator then generates a composite feature attack, and the discriminator starts the mechanism deduction and decomposition, while introducing a time dimension attack, such as a slowly increasing particle concentration, and finally the discriminator identifies the gradual change pattern through sequence analysis. The fault discriminator takes particle detection data as input, that is, it receives the particle data stream from the detection array in real time, compares it with the threshold of each indicator to make a quick judgment, and then performs weighted combination of abnormal features to automatically select the reasoning path. It then executes mechanism deduction to deeply diagnose the fault, including calculating the probability of the fault type and calculating the probability distribution of the fault source in combination with the detection network. Finally, it outputs the fault data of the power station energy storage equipment and ensures that the false alarm rate of faults is reduced and the recognition rate of new faults is improved.

[0045] Furthermore, step S250 also includes that the fault discriminator includes three-order training; wherein the three-order training method includes: using a discriminant-generative architecture, performing first-order training based on threshold discrimination, and using the discriminant architecture as the first judgment node; using a discriminant-generative architecture, performing second-order training based on mechanism deduction, and using the generative architecture as the second inference node; performing migration calls and node cascade triggering-based third-order training on the first judgment node and the second inference node to determine the fault discriminator.

[0046] Preferably, the fault discriminator includes three-order training. Specifically, a discriminant-generator architecture is configured, including constructing a feature extraction module based on a convolutional neural network and simulating particle distribution data under various normal / abnormal working conditions. Then, a first-order training based on threshold discrimination is performed, that is, the generator simulates abnormal particle data with random noise, and the discriminator learns to identify normal / abnormal states that conform to the threshold law of the fingerprint library, and embeds the discriminant architecture as the first judgment node into the edge computing unit of the detection array, thereby establishing a rapid response capability based on physical thresholds, forming the first line of defense for fault detection, and being able to output a normal / abnormal binary decision.

[0047] Preferably, a discriminant-generative architecture is configured, the generator constructs a feature combination that conforms to the fault mechanism, the discriminator builds a graph neural network (GNN), the nodes correspond to the detection positions, and the edges represent the particle diffusion paths. Then, second-order training based on mechanism deduction is performed. Specifically, the generator generates adversarial samples that are confusing to the mechanism, such as combining the short-circuit particle features with the leaked gas. Then, the discriminator learns the mechanism mapping rules through multi-label classification and uses the generated architecture as the second inference node to improve the accuracy of fault type identification. Finally, the first decision node and the second inference node are seamlessly connected and dynamically coordinated to build a complete fault discrimination decision, including migration calls and three-order training based on node cascade triggering. Migration calls refer to using the output of the first decision node as the trigger condition for the second inference node, and the gradient can be transferred across nodes during backpropagation; three-order training based on node cascade triggering refers to designing a composite loss function and shifting the focus of the adversary to timing behavior. The generator generates progressive fault samples with latent period characteristics, and the discriminator learns the association mining of early weak signals. Finally, a complete fault discriminator is formed to ensure that the false alarm rate of fault identification is reduced while improving the accuracy and real-time performance of fault warning of power station energy storage equipment.

[0048] Step S300 , performing particle detection according to the detection array, triggering a two-stage decision based on the fault discriminator, determining a fault data chain, and the energy storage management system performing directional early warning management in response to the fault data chain.

[0049] Step S300 further includes step S310, performing directional particle detection on the power station energy storage equipment according to the detection array to determine the detection array; step S320, the energy storage management system receives the detection array, triggers the first judgment node to perform threshold feature matching and binary classification judgment, and determines the first judgment result; step S330, if the first judgment result is abnormal, triggers the second inference node to perform knowledge reasoning based on fault mechanism tracing, determines the fault sequence and adds it to the fault data chain.

[0050] Preferably, a detection array is used to perform directional particle detection of power station energy storage equipment, that is, probes based on the deployed network collect multi-dimensional particle data at a preset frequency to form a three-dimensional data matrix aligned in time and space, that is, to determine the detection array. After the energy storage management system receives the detection array, it triggers the first judgment node to perform threshold feature matching. Specifically, key particle indicators are selected and compared with the preset thresholds in the fault particle fingerprint library. At the same time, the thresholds are dynamically adjusted according to the temperature and humidity in the battery compartment; then a binary classification judgment is performed to determine the first judgment result, including normal judgment and abnormal judgment, wherein the normal judgment means that all particle indicators are lower than the dynamic threshold, and the abnormal judgment means that any particle indicator exceeds the dynamic threshold, thereby activating the second inference node and marking the coordinates of the abnormal area. If the first judgment result is abnormal, to avoid wasting system resources, the second reasoning node is triggered to perform knowledge reasoning based on fault mechanism tracing. Specifically, the particle size distribution and movement pattern are matched first and the associated gas combination rules are verified. Then, based on the time series of the detection array, the fault starting point is reversed. For example, the earliest abnormal position is determined by the concentration gradient, and the battery compartment airflow model is combined to predict the scope of the fault impact. Finally, a structured diagnostic report is output, that is, the fault sequence is determined and the confirmed fault sequence is added to the fault data chain, realizing real-time early warning of the power station energy storage equipment while ensuring the accuracy and efficiency of fault diagnosis.

[0051] Preferably, the energy storage management system responds to the fault data chain for targeted early warning management. Core elements include precise risk location, hierarchical dynamic response, and closed-loop action verification. Specifically, the energy storage management system automatically decomposes key elements in the fault data chain, including the energy storage device's fault type, location coordinates, development stage, and impact range. Then, combined with real-time data, it uses a fuzzy logic algorithm to calculate the risk diffusion probability and output a four-level warning level. Based on the analysis results of the fault data chain, differentiated response strategies are initiated. The battery management system then precisely controls the power connections of the target module for spatial targeting. For progressive faults, such as slowly developing internal short circuits, temporal targeting is implemented. Work orders are automatically generated and navigated to the fault point, with simultaneous delivery of action plans to facilitate targeted O&M resource scheduling. After implementation, enhanced detection of the fault area is automatically triggered to assess the effectiveness of the action in real time. If indicators continue to deteriorate, the warning level is upgraded. If indicators return to normal, the system switches to observation mode. Finally, the entire process data from this early warning processing is added to the fault particle fingerprint database to optimize future response strategies. This ensures the precise spatial location of faults in power plant energy storage equipment, significantly improving the safety and O&M efficiency of the power plant.

[0052] Step S400 , wherein, based on the fault data link, repeated detection guidance based on the detection array is performed, and fault verification is performed by introducing associated gas.

[0053] Preferably, after the initial fault determination, a closed-loop verification process is started, and the accuracy of the fault diagnosis is ensured by dynamically adjusting the detection strategy and cross-verifying the multimodal data. Specifically, the working mode of the detection array is automatically adjusted according to the position information in the fault data chain, such as enhanced scanning or path optimization, and repeated detection guidance based on the detection array is performed, that is, special monitoring is started for the identified suspicious particle characteristics. For example, if 0.5-1μm particles are detected for the first time, high-resolution spectral analysis is enabled during repeated detection. If accompanied by abnormal vibration, the acoustic wave detection mode is activated synchronously; then the fault verification is performed by introducing associated gas. Specifically, based on the fault The particle fingerprint library establishes a correspondence between gases and particles. For example, in the early stages of thermal runaway, particles are 0.1-1μm metal oxides, and in the event of electrolyte leakage, particles are 1-10μm droplets. When the particle detection and judgment are made, if the judgment probability is relatively low, the corresponding gas sensor is automatically activated. If nanoparticles are detected, CO monitoring is initiated, and the gas concentration is checked to see if it reaches the threshold within the expected time window. At the same time, it is confirmed that the gas concentration peak position coincides with the particle-rich area, thereby achieving auxiliary fault judgment, ensuring a lower failure rate for energy storage equipment and improving the accuracy of fault identification and diagnosis in energy storage power stations, thereby enhancing the scientificity and reliability of safety management of energy storage equipment in power stations.

[0054] In the above, refer to Figure 1 The energy storage power station early warning method based on particle fault detection according to an embodiment of the present invention is described in detail. Figure 2 An energy storage power station early warning device based on particle fault detection according to an embodiment of the present invention is described.

[0055] The energy storage power station early warning device based on particle fault detection according to the embodiment of the present invention is used to solve the technical problems existing in the prior art, such as insufficient detection network coverage, weak correlation of fault mechanisms, and lack of dynamic verification means, which lead to poor accuracy and reliability of energy storage power station fault diagnosis and early warning, and achieve the technical effect of improving the accuracy and reliability of energy storage power station fault diagnosis and early warning. Figure 2 As shown, the energy storage power station early warning device based on particle fault detection includes: a detection array installation module 10, a fault discriminator generation module 20, a fault data link determination module 30, and a fault verification module 40.

[0056] The detection array installation module 10 is used to set up a deployment network of microparticle probes for power station energy storage equipment and install the detection array in the battery compartment of the power station energy storage equipment, wherein the first deployment is based on the particle motion characteristics of the top-sidewall-bottom, and the second deployment is based on the division of the spatial safety boundary. The fault discriminator generation module 20 is used to define a first judgment node based on the deployment network by constructing a fault particle fingerprint library, and define a second inference node based on the reverse inference of particle characteristics and fault mechanism, to generate a fault discriminator and integrate it into the energy storage management system. The fault data chain determination module 30 is used to perform microparticle detection based on the detection array, trigger a two-order decision based on the fault discriminator, determine the fault data chain, and the energy storage management system responds to the fault data chain for directional early warning management. The fault verification module 40 is used to perform complex detection guidance based on the detection array based on the fault data chain and perform fault verification by introducing associated gas.

[0057] The specific configuration of the detection array installation module 10 will be described in detail below. The detection array installation module 10 further includes: defining the number of microparticle nodes based on the installed capacity of the energy storage device, where the number of microparticle nodes is an interval dimension; performing a deployment based on the number of microparticle nodes using top scanning, sidewall networking, and bottom deposition to determine a first deployment method; determining a second deployment method based on a microparticle safety boundary based on a fault source; and fusing the first and second deployment methods to determine the deployment network.

[0058] The specific configuration of the detection array installation module 10 will be described in detail below. The detection array installation module 10 further includes: retrieving fault records of power plant energy storage equipment and performing homology clustering to mine intra-class fault sequences, where the intra-class fault sequences include at least the fault source and fault mechanism; defining a relative safety boundary based on the intra-class fault sequence, where the relative safety boundary corresponds to the intra-class fault sequence; and deploying micro-nodes within the relative safety boundary to determine the second deployment method.

[0059] The specific configuration of the detection array installation module 10 will be described in detail below. The detection array installation module 10 further includes: determining particle intrinsic characteristics and particle evolution characteristics based on the failure mechanism, and defining a node window size; wherein the particle intrinsic characteristics include at least a granularity range; and determining a node network by intersecting the first and second deployment methods and performing a detection domain expansion based on the node window size, as the deployment network.

[0060] The specific configuration of the detection array installation module 10 will be described in detail below. The detection array installation module 10 further includes: installing a detection array within a battery compartment of a power plant energy storage device, wherein the detection array is composed of nano-particle detectors; and constraining the deployment position of the detection array within the compartment based on the non-intersecting detection paths according to the deployment network.

[0061] The specific configuration of the fault discriminator generation module 20 will be described in detail below. The fault discriminator generation module 20 further includes: determining fault particle fingerprint elements by integrating the intra-class fault sequence, wherein the fault particle fingerprint elements at least include fault type, characteristic particles, size distribution, motion characteristics, and associated gases; extracting a first threshold feature and a second mechanism feature based on the fault particle fingerprint elements, wherein the first threshold feature is a normal critical value based on direct particle detection information; and constructing the fault particle fingerprint library based on the fault particle fingerprint elements, the first threshold feature, and the second mechanism feature.

[0062] The specific configuration of the fault discriminator generation module 20 will be described in detail below. The fault discriminator generation module 20 further includes: constructing a discrimination block by embedding the fault particle fingerprint library; and performing supervised training on the discrimination block using the adversarial principle to generate the fault discriminator. The fault discriminator uses particle detection data as input, threshold discrimination, mechanism deduction triggering, and mechanism deduction execution as intermediate logic, and outputs fault data.

[0063] The specific configuration of the fault discriminator generation module 20 will be described in detail below. The fault discriminator generation module 20 further includes: The fault discriminator includes three-stage training; wherein the three-stage training method includes: performing first-stage training based on threshold discrimination using a discriminant-generative architecture, with the discriminant architecture as the first decision node; performing second-stage training based on mechanism deduction using a discriminant-generative architecture, with the generative architecture as the second inference node; and performing migration calls and third-stage training based on node cascade triggering on the first decision node and the second inference node to determine the fault discriminator.

[0064] The specific configuration of the fault data chain determination module 30 will be described in detail below. The fault data chain determination module 30 further includes: performing directional particle detection on the power plant energy storage equipment based on the detection array to determine a detection array; the energy storage management system receiving the detection array and triggering a first determination node to perform threshold feature matching and binary classification to determine a first determination result; if the first determination result is abnormal, triggering a second inference node to perform knowledge inference based on fault mechanism tracing, determine the fault sequence, and add it to the fault data chain.

[0065] The energy storage power station early warning device based on particle fault detection provided by the embodiment of the present invention can execute the energy storage power station early warning method based on particle fault detection provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0066] Although the present application makes various references to certain modules in the apparatus according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other and are not used to limit the scope of protection of the present invention.

[0067] The above specific embodiments do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application shall be included within the scope of protection of this application.

Claims

1. An energy storage power station early warning method based on particle fault detection is characterized in that: The method comprises: For power plant energy storage equipment, a deployment network of microparticle probes is set up, and detection arrays are installed in the battery compartments of the power plant energy storage equipment. The first deployment is based on the particle movement characteristics of the top-sidewall-bottom, and the second deployment is based on the division of spatial safety boundaries. By building a fault particle fingerprint library, defining a first determination node based on the deployment network, and defining a second inference node based on the reverse inference of particle characteristics and fault mechanisms, a fault discriminator is generated and built into the energy storage management system; Performing particle detection according to the detection array triggers a two-stage decision based on the fault discriminator to determine a fault data chain, and the energy storage management system performs directional early warning management in response to the fault data chain; Among them, complex detection guidance based on the detection array is carried out based on the fault data link, and fault verification is carried out by introducing associated gas.

2. The energy storage power station early warning method based on particle fault detection according to claim 1, characterized in that: Set up the deployment network of particle probes, including: Defining the number of micro-nodes based on the installed capacity of the energy storage device, wherein the number of micro-nodes is an interval dimension; Performing a deployment based on the number of the particle nodes by top scanning, sidewall networking, and bottom deposition to determine a first deployment method; Determine the second deployment method based on the particle safety boundary based on the fault source; The first deployment mode and the second deployment mode are integrated to determine the deployment network.

3. The energy storage power station early warning method based on particle fault detection according to claim 2, characterized in that: The second deployment method is determined based on the micro-particle safety boundary based on the fault source, including: Retrieving fault records of power station energy storage equipment and performing homology clustering to mine intra-class fault sequences, wherein the intra-class fault sequences at least include the fault source and the fault mechanism; Defining a relative safety boundary according to the intra-class fault sequence, wherein the relative safety boundary corresponds to the intra-class fault sequence; Deploy micro-nodes within the relatively safe boundary and determine the second deployment mode.

4. The energy storage power station early warning method based on particle fault detection according to claim 3 is characterized in that: According to the fault mechanism, determining the intrinsic characteristics and evolutionary characteristics of the particles, and defining the node window size; wherein the intrinsic characteristics of the particles at least include the particle size range; An intersection of the first deployment mode and the second deployment mode is taken to determine a node network and perform detection domain expansion based on the node window size, which serves as the deployment network.

5. The energy storage power station early warning method based on particle fault detection according to claim 4 is characterized in that: Installing a detection array in a battery compartment of a power station energy storage device, wherein the detection array is composed of nano-particle detectors; According to the deployment network, the deployment position of the detection array in the cabin is constrained by the non-intersection of the detection paths.

6. The energy storage power station early warning method based on particle fault detection according to claim 3 is characterized in that: Build a fault particle fingerprint library, including: Determine the fault particle fingerprint elements by integrating the intra-class fault sequence, wherein the fault particle fingerprint elements at least include fault type-characteristic particles-size distribution-motion characteristics-associated gas; Extracting a first threshold feature and a second mechanism feature using the fault particle fingerprint element, wherein the first threshold feature is a constant critical value based on direct particle detection information; The faulty particle fingerprint library is constructed using the faulty particle fingerprint element, the first threshold feature, and the second mechanism feature.

7. The energy storage power station early warning method based on particle fault detection according to claim 1, characterized in that: Generate a fault discriminator, including: By building the fault particle fingerprint library, a discrimination block is constructed; The discrimination block is supervised and trained based on the adversarial principle to generate the fault discriminator, wherein the particle detection data is used as input, the threshold discrimination-mechanism deduction trigger-mechanism deduction execution is used as the intermediate logic, and the fault data is used as the output.

8. The energy storage power station early warning method based on particle fault detection according to claim 7 is characterized in that: The fault discriminator includes three-stage training; The three-stage training method includes: Using a discriminative-generative architecture, we perform first-order training based on threshold discrimination, using the discriminative architecture as the first decision node. Performing second-order training based on mechanism inference with a discriminative-generative architecture, using the generative architecture as the second inference node; The first determination node and the second inference node are migrated and called, and three-stage training based on node cascade triggering is performed to determine the fault discriminator.

9. The energy storage power station early warning method based on particle fault detection according to claim 1, characterized in that: Identify the faulty data link, including: Performing directional particle detection on the power station energy storage device according to the detection array to determine the detection number array; The energy storage management system receives the detection data array, triggers a first determination node to perform threshold feature matching and binary classification determination, and determines a first determination result; If the first determination result is abnormal, the second reasoning node is triggered to perform knowledge reasoning based on fault mechanism tracing, determine the fault sequence and add it to the fault data chain.

10. An energy storage power station early warning device based on particle fault detection is characterized in that: The device is used to implement the energy storage power station early warning method based on particle fault detection according to any one of claims 1 to 9, and the device includes: A detection array installation module is used to set up a deployment network of particle probes for power station energy storage equipment and install the detection arrays in the battery compartments of the power station energy storage equipment, wherein a first deployment is based on the particle movement characteristics of the top-sidewall-bottom, and a second deployment is based on the division of spatial safety boundaries; A fault discriminator generation module is configured to construct a fault particle fingerprint library, define a first determination node based on the deployment network, define a second inference node based on the reverse deduction of particle characteristics and fault mechanisms, generate a fault discriminator, and embed it into the energy storage management system; a fault data link determination module, configured to perform particle detection based on the detection array, trigger a two-stage decision based on the fault discriminator, determine a fault data link, and the energy storage management system performs directional early warning management in response to the fault data link; The fault verification module is used to perform complex detection guidance based on the detection array based on the fault data link and to perform fault verification by introducing associated gas.

Images