Distributed optical fiber sensing crack monitoring and service life prediction system of energy storage cabinet shell

Abstract

The invention discloses a distributed optical fiber sensing crack monitoring and service life prediction system for an energy storage cabinet shell, which comprises a data acquisition module, a signal processing module, a state diagnosis module and a service life prediction module, and is characterized in that the data acquisition module is used for synchronously acquiring sensing data of a distributed optical fiber sensor and working condition data of an energy storage cabinet; a physical information neural network model is arranged in the signal processing module, and the signal processing module is used for decoupling and separating pure mechanical strain field data from the coupling sensing data. Through deep fusion of the physical information neural network and deep learning, intelligent decoupling of the temperature and the strain effect is realized, the problem of false alarm and missing alarm caused by cross sensitivity in a traditional monitoring method is solved, accurate identification, quantitative tracking and life prediction can be carried out on the energy storage cabinet shell crack, and the energy storage cabinet shell crack detection method is suitable for popularization and application. And the operation safety and reliability of the energy storage system are obviously improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage safety, in particular to a distributed optical fiber sensing crack monitoring and life prediction system for an energy storage cabinet shell. BACKGROUND

[0002] Energy storage technology is a key supporting technology for building a new power system and achieving the "double carbon" goal. Among them, the electrochemical energy storage cabinet as an important energy storage unit, its safety and reliability is directly related to the stable operation of the entire energy storage power station. The energy storage cabinet shell, as the first physical barrier to protect the internal battery module and electrical elements, its structural integrity is crucial. However, during long-term service, the energy storage cabinet shell will be subjected to the combined action of multiple complex loads such as thermal stress generated by internal battery charging and discharging cycles, external environmental temperature difference changes, and mechanical vibration. These factors may induce fatigue damage of the shell material and eventually form cracks. The initiation and propagation of cracks will weaken the load-bearing capacity and sealing performance of the shell, and once out of control, it may lead to exposure of internal components, insulation failure, and even cause serious safety accidents such as thermal runaway.

[0003] Currently, for the health monitoring of large structural components, traditional methods mostly rely on periodic offline non-destructive testing such as ultrasonic flaw detection and penetration testing. These methods not only require professional personnel to operate on site, interrupting normal equipment operation, but also have long detection cycles, making it impossible to capture the initiation and early propagation of cracks in real time, and difficult to meet the scene requirements of energy storage cabinets which require high frequency and all-weather online monitoring. Distributed optical fiber sensing technology provides a new possibility for online monitoring of energy storage cabinet shells due to its distributed measurement, anti-electromagnetic interference, and intrinsic safety. However, in practical applications, this technology faces a core challenge: the sensing optical fiber has cross-sensitivity to temperature and strain. The energy storage cabinet will produce significant temperature changes during charging and discharging. This thermal strain (or temperature frequency shift) caused by normal operating conditions will be coupled with the pure mechanical strain caused by structural deformation (such as cracks), forming a complex sensing signal. Existing monitoring methods cannot effectively distinguish between the two effects, often misjudging normal temperature fluctuations as structural abnormalities, resulting in a large number of false alarms; or the small mechanical strain signal is overwhelmed by the strong temperature signal, causing missed reports of cracks. This coupling effect of temperature and strain greatly limits the accuracy and reliability of distributed optical fiber sensing technology in the application of energy storage cabinet crack monitoring.

[0004] Therefore, the present application provides a distributed optical fiber sensing crack monitoring and life prediction system for an energy storage cabinet shell to solve the above problems. SUMMARY

[0005] In view of the deficiencies of the prior art, the present application provides a distributed optical fiber sensing crack monitoring and life prediction system for an energy storage cabinet shell, which solves the above problems.

[0006] To achieve the above object, the application is implemented by the following technical solutions: a distributed optical fiber sensing crack monitoring and life prediction system for an energy storage cabinet shell, comprising:

[0007] a data acquisition module, configured to synchronously acquire sensing data of the distributed optical fiber sensors arranged along the energy storage cabinet shell and working condition data of the energy storage cabinet;

[0008] a signal processing module, connected with the data acquisition module, configured to receive the sensing data and the working condition data, and decouple and separate pure mechanical strain field data representing shell structure deformation from the sensing data and the working condition data;

[0009] a state diagnosis module, connected with the signal processing module, configured to identify crack events on the shell surface according to the pure mechanical strain field data, and perform state quantification and evolution tracking on the identified cracks;

[0010] a life prediction module, connected with the state diagnosis module and the data acquisition module, configured to predict the remaining service life of the cracks based on the evolution history of the cracks and the working condition data of the energy storage cabinet.

[0011] Preferably, the sensing data of the distributed optical fiber sensors acquired by the data acquisition module is distributed frequency shift data containing temperature and strain coupling effects; and the working condition data at least includes state of charge data and charge and discharge current data of the energy storage cabinet.

[0012] Preferably, the signal processing module is provided with a physical information neural network model, the input of the physical information neural network model is time and spatial position information, the sensing data and the working condition data, and the output is the pure mechanical strain field data and surface temperature field data of the energy storage cabinet shell.

[0013] Preferably, the physical information neural network model is trained by a hybrid loss function, the hybrid loss function includes a data-driven term and a physical constraint term, the data-driven term is used to minimize the error between the coupling effect predicted by the network and the actually acquired sensing data, and the physical constraint term is constructed based on a heat conduction partial differential equation, and is used to constrain the temperature field output by the network to comply with physical laws, wherein the charge and discharge current data is used as a heat source term.

[0014] Preferably, the state diagnosis module comprises:

[0015] a crack identification unit, configured to identify a strain concentrated area in the pure mechanical strain field data by setting a threshold or a clustering algorithm, and determine the strain concentrated area as a crack event;

[0016] a state quantification unit configured to map the features of the strain concentration region to the equivalent physical parameters of the crack through a pre-trained surrogate model;

[0017] an evolution tracking unit configured to establish a time series profile of the location and the quantified state of each identified crack over time.

[0018] Preferably, the life prediction module employs a long short-term memory (LSTM) model; the input of the LSTM model is the state evolution sequence of the specified crack within a past time window and the corresponding working condition data sequence.

[0019] Preferably, the life prediction module utilizes the trained LSTM model to predict the future state evolution trend of the crack in a rolling manner until the predicted crack state reaches a preset failure threshold, and the time difference from the current time to the time when the failure threshold is reached is taken as the predicted value of the remaining service life.

[0020] Preferably, the method further comprises:

[0021] a visual alarm module configured to render the pure mechanical strain field and the identified crack information on the three-dimensional digital twin model of the energy storage cabinet in real time, and trigger a hierarchical alarm when the crack state deteriorates or the remaining service life is lower than a preset safety threshold.

[0022] Preferably, the surrogate model in the state quantification unit is a multilayer perceptron network trained based on finite element simulation data, which establishes a nonlinear mapping relationship between the features such as peak value, width, and integral area of the strain concentration region and the equivalent length and depth of the crack.

[0023] The application also provides a distributed optical fiber sensing crack monitoring and life prediction method for an energy storage cabinet shell, comprising the following steps:

[0024] S1: synchronously collecting the coupled sensing data of the distributed optical fiber sensors arranged along the energy storage cabinet shell and the real-time working condition data of the energy storage cabinet;

[0025] S2: inputting the coupled sensing data and the working condition data into a pre-trained physical information neural network model for intelligent decoupling to obtain pure mechanical strain field data;

[0026] S3: analyzing the pure mechanical strain field data to identify and quantify the state of the crack and record the evolution history thereof over time;

[0027] S4: inputting the evolution history of the crack and the corresponding working condition data into a pre-trained long short-term memory (LSTM) model to predict the remaining service life of the crack.

[0028] Advantages

[0029] The application provides a distributed optical fiber sensing crack monitoring and life prediction system for an energy storage cabinet shell.

[0030] The application inputs the operating conditions (such as charging and discharging current) of the energy storage cabinet as a heat source of a physical model by using a physical information neural network and a signal processing module, so that accurate decoupling of a temperature field and a pure mechanical strain field is realized at a data level. This method not only overcomes the shortcomings of insufficient precision or complex deployment of traditional temperature compensation methods, but also fundamentally eliminates false positives and false negatives caused by normal thermal effects. Based on the high signal-to-noise ratio strain data after decoupling, the system can accurately identify and quantitatively track early micro-cracks, and scientifically predict the remaining life of the cracks by combining a deep learning model. Finally, through the combination of visual alarm and digital twin model, timely, accurate and intuitive decision support is provided for predictive maintenance of the energy storage cabinet, and the operation safety and reliability of the energy storage system are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 A functional block diagram of the distributed optical fiber sensing crack monitoring and life prediction system for the energy storage cabinet shell according to the embodiment of the application is shown.

[0032] Figure 2 A flowchart of the distributed optical fiber sensing crack monitoring and life prediction method for the energy storage cabinet shell according to the embodiment of the application is shown.

[0033] Figure 3 A functional block diagram of the state diagnosis module according to the embodiment of the application is shown.

[0034] In the figure, 100 is a data acquisition module, 200 is a signal processing module, 300 is a state diagnosis module, 310 is a crack identification unit, 320 is a state quantification unit, 330 is an evolution tracking unit, 400 is a life prediction module, and 500 is a visual alarm module. DETAILED DESCRIPTION

[0035] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, rather than all the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application.

[0036] Embodiment:

[0037] Figure 1A functional block diagram of a distributed optical fiber sensing crack monitoring and life prediction system for an energy storage cabinet shell according to an embodiment of the present application. The system includes a data acquisition module 100, a signal processing module 200, a state diagnosis module 300, a life prediction module 400, and an optional visual alarm module 500.

[0038] Data acquisition module 100: for synchronously acquiring sensing data of the distributed optical fiber sensor arranged along the energy storage cabinet shell and working condition data of the energy storage cabinet. It should be understood that in order to accurately decouple temperature and strain, the measurement results containing coupling effects and the causes of temperature changes must be known at the same time. Therefore, this module not only acquires the original data of the optical fiber sensor, but also synchronously acquires working condition data reflecting the heat production state of the energy storage cabinet.

[0039] In one specific embodiment, a distributed optical fiber sensor (such as a sensing optical fiber based on Brillouin optical time domain analysis BOTDA) is used to be pasted and laid along the key areas (such as welds, corner points, reinforcing ribs, etc.) of the energy storage cabinet shell. The optical fiber demodulator collects Brillouin frequency shift data distributed along the length of the optical fiber at a set time interval (such as once per minute), and this data is the sensing data containing temperature and strain coupling effects. At the same time, the working condition data is read in real time from the battery management system (BMS) or energy management system (EMS) of the energy storage cabinet through a communication interface (such as CAN bus or Modbus protocol), and these data at least include the state of charge (SOC) of the battery, real-time charging and discharging current / power, etc. All data are marked with accurate time stamps for subsequent spatio-temporal alignment.

[0040] Signal processing module 200, connected with data acquisition module 100, for receiving sensing data and working condition data, and separating pure mechanical strain field data representing shell structure deformation therefrom.

[0041] In one specific embodiment, a physical information neural network (PINN) model is provided in the signal processing module 200. The model is essentially a multi-layer perceptron, with time and spatial coordinates (t, x) as inputs, and predicted pure mechanical strain and temperature [ε m (t, x), T(t, x)] at the time and space point as outputs. More specifically, the PINN model is trained by a hybrid loss function L total , which is calculated as follows:

[0042] L total = λ data ·L data + λ phys ·L phys

[0043] Where λ data and λ phys are weight coefficients.

[0044] Data-driven term L data To ensure the model's prediction is consistent with the actual measurement data. The total frequency shift Δν of the fiber sensor is a linear superposition of pure mechanical strain ε m and temperature change ΔT, i.e., Δν = C ε ·ε m +C T ·ΔT, where C ε and C T are the strain and temperature sensitivity coefficients of the fiber, respectively. The total frequency shift predicted by the model is:

[0045] Δν pred =C ε ·ε m,pred +C T ·(T pred -T ref )

[0046] The data-driven term is the mean square error (MSE) between the predicted frequency shift and the actual measured frequency shift:

[0047]

[0048] where the summation goes through all N measurement data points.

[0049] Physical constraint term L phys To ensure that the model output temperature field T pred (t, x) follows the heat conduction physical law, for a one-dimensional simplified model of the energy storage cabinet shell, its heat conduction partial differential equation (PDE) is:

[0050]

[0051] where α is the thermal diffusivity, ρ is the material density, and c p is the specific heat capacity. The heat source term Q(t, x) is mainly determined by the battery charging and discharging heat, which is proportional to the square of the current I(t), i.e., Q(t, x) ≈ k · I(t) 2 , k is the proportional coefficient. The physical constraint term is the mean square error of the PDE residual, calculated at a series of collocation points in the space-time domain:

[0052]

[0053] where the summation goes through all M collocation points. The partial derivative of T pred in the network is accurately calculated through automatic differentiation technology. By minimizing L total , the PINN model can learn a solution that satisfies both the measurement data and the physical law, thereby accurately decomposing Δν pred into ε m,pred and Tpred .

[0054] The state diagnosis module 300 is connected with the signal processing module 200, and is used to identify, quantify and track the crack according to the pure mechanical strain field data. It can be understood that after obtaining the pure strain field, an automatic process is needed to interpret the data and convert it into a clear description of the crack state.

[0055] In one specific embodiment, as shown in FIG. 3, the module includes: Figure 3

[0056] The crack identification unit 310: This unit continuously scans the pure mechanical strain field data. Since the emergence of cracks will cause stress concentration at the tip and nearby areas, which is manifested as local mutation of strain values. By setting a reasonable strain threshold, or using clustering algorithms such as DBSCAN, these strain concentrated areas can be automatically identified and determined as crack events.

[0057] The state quantification unit 320: For each identified strain concentrated area, this unit aims to map its features to the equivalent physical parameters of the crack (such as length, depth). It should be understood that this mapping relationship is very complex. Therefore, this unit uses a pre-trained proxy model. Specifically, a multi-layer perceptron (MLP) network is used to map the features of the strain concentrated area to the equivalent physical parameters of the crack. The calculation method is as follows:

[0058] First, for an identified strain concentrated area ε m (x) (where x belongs to the interval [x start , x end ]), a set of feature vectors F is extracted:

[0059] Peak strain F peak = max(ε m (x))

[0060] Area width F width = x end -x start

[0061] Strain integral

[0062] Then, the feature vector F = [F peak , F width , F area ] is input to the MLP network. The network undergoes nonlinear transformation through multiple hidden layers (for example, two hidden layers, 64 neurons per layer, using ReLU activation function), and finally outputs the equivalent physical parameters of the crack, such as equivalent length L eq and equivalent depth D eq ​The entire mapping process can be represented as:

[0063] [L eq D eq ] = MLP(F; θ MLP )

[0064] Where, θ MLP The network weight parameters are obtained through supervised learning on a large number of finite element simulation datasets.

[0065] Evolution Tracking Unit 330: This unit establishes a time-series profile for each identified and quantified crack. The profile records the fixed location information of the crack, as well as the historical sequence of its quantified state (such as equivalent length and depth) over time. This provides crucial input data for subsequent lifetime prediction.

[0066] The life prediction module 400, connected to the condition diagnosis module 300 and the data acquisition module 100, is used to predict the remaining service life (RUL) of a crack. It should be understood that knowing only the current state of the crack is insufficient; predicting its future development trend is crucial for developing maintenance plans.

[0067] In one specific embodiment, this module employs a Long Short-Term Memory (LSTM) network model. Its computation method is as follows:

[0068] First, construct the input sequence. At time t, take data from the past W time steps to form an input sequence X. t Each element in the sequence is the crack state S at that moment. i =[L eq (i),D eq (i)] and the corresponding operating condition data C i = concatenated vector x of [SOC(i), I(i)] i =[S i C i Therefore, the input sequence is X. t =[x t-W+1 ,…,x t ].

[0069] The sequence is input into the LSTM network. At each time step i, the LSTM unit adjusts the current input x. i and the hidden state h from the previous moment i-1 Cell state c i-1 The cell state is updated and a new hidden state is calculated through internal forget gates, input gates, and output gates:

[0070] Forgotten Gate: f i =σ(W f [h i-1 ,xi ]+b f )

[0071] Input gate: i i = σ(W i [h i-1 , x i ]+b i )

[0072] Candidate cell state:

[0073] New cell state:

[0074] Output gate: o i = σ(W o [h i-1 , x i ]+b o )

[0075] New hidden state: h i = o i ⊙ tanh(c i )

[0076] where σ is the Sigmoid function, tanh is the hyperbolic tangent function, W and b are learnable weights and biases, and ⊙ denotes element-wise multiplication.

[0077] The hidden state h t at the last time step t contains the dynamic information of the entire input sequence. It is fed into a fully connected layer (regression head) to predict the crack state at the next time step t+1:

[0078]

[0079] In the prediction phase, this module predicts forward in a rolling manner:

[0080] 1. Use the last W real data points to form the initial input sequence, and predict

[0081] 2. Concatenate with a predicted or assumed future working condition C t+1 to form

[0082] 3. Roll the input sequence forward, remove the oldest element x t-W+1 , and add a new predicted element

[0083] 4. Repeat this process to generate until the predicted crack state first reaches the preset failure threshold Sfail (e.g., L eq beyond the critical length).

[0084] The remaining useful life (RUL) is RUL = k · Δt, where Δt is the time step.

[0085] The visualization alarm module 500 is used to render the pure mechanical strain field cloud map and the identified crack information (location, size, RUL, etc.) on the three-dimensional digital twin model of the energy storage cabinet in real time. When the crack state is significantly deteriorated (such as the expansion rate is accelerated) or the predicted remaining useful life is lower than the preset safety threshold, the system triggers a hierarchical alarm (such as yellow warning, red alarm), and notifies the operation and maintenance personnel through sound and light, short message or upper computer software, etc., to provide intuitive and timely decision basis for them.

[0086] Figure 2 The flowchart of the distributed optical fiber sensing crack monitoring and life prediction method for the energy storage cabinet shell according to the embodiment of the present application. The method comprises the following steps:

[0087] S1: Synchronously collecting data, synchronously collecting the coupled sensing data (such as frequency shift) of the distributed optical fiber sensor arranged along the energy storage cabinet shell and the real-time working condition data (such as SOC, charging and discharging current) of the energy storage cabinet.

[0088] S2: Intelligent decoupling, inputting the collected coupled sensing data and working condition data together with time and space information into a pre-trained physical information neural network (PINN) model. The model uses data and physical double constraints to output high-precision pure mechanical strain field data.

[0089] S3: Diagnosis and tracking, analyzing the pure mechanical strain field data, identifying the strain concentrated area representing the crack through threshold or clustering algorithm. Then, using an MLP-based proxy model to quantify the strain feature into the physical size of the crack, and recording its evolution history archive over time.

[0090] S4: Life prediction, inputting the evolution history sequence of the specified crack and its corresponding working condition data sequence into a pre-trained long short-term memory network (LSTM) model. The model deduces the future state evolution trend of the crack until it reaches the failure threshold in a rolling prediction manner, thereby calculating its remaining useful life.

[0091] In summary, the embodiments of the present application are illustrated. By introducing a physical information neural network, it skillfully utilizes the working condition data as the known conditions of the physical model, realizes the high-precision intelligent decoupling of the temperature and strain effects in the optical fiber sensing signal, and lays a solid foundation for the subsequent accurate crack diagnosis and life prediction. The whole system constitutes a closed-loop intelligent monitoring system from data acquisition, signal processing, state diagnosis to life prediction, which can effectively improve the operation safety of the energy storage cabinet and realize the transformation from passive maintenance to predictive maintenance.

[0092] Meanwhile, the contents not described in detail in the specification all belong to the prior art known to those skilled in the art.

[0093] It should be noted that, in this paper, relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment.

[0094] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A distributed optical fiber sensing crack monitoring and life prediction system for an energy storage tank enclosure, characterized by, The application relates to a state diagnosis and life prediction method for an energy storage cabinet, which comprises the following steps: a data acquisition module (100) is used for synchronously acquiring sensing data of distributed optical fiber sensors arranged along an energy storage cabinet shell and working condition data of the energy storage cabinet; a signal processing module (200) is connected with the data acquisition module (100) and is used for receiving the sensing data and the working condition data and decoupling pure mechanical strain field data representing shell structure deformation from the sensing data and the working condition data; a state diagnosis module (300) is connected with the signal processing module (200) and is used for identifying crack events on the shell surface according to the pure mechanical strain field data and quantifying the states of the identified cracks and tracking crack evolution; a life prediction module (400) is connected with the state diagnosis module (300) and the data acquisition module (100) and is used for predicting the remaining service life of the cracks based on crack evolution history and working condition data of the energy storage cabinet.

2. The distributed optical fiber sensing crack monitoring and life prediction system for the energy storage tank shell of claim 1, wherein, The sensing data of the distributed optical fiber sensors acquired by the data acquisition module (100) is distributed frequency shift data containing temperature and strain coupling effects; and the working condition data at least includes state of charge data and charging and discharging current data of the energy storage cabinet.

3. The distributed optical fiber sensing crack monitoring and life prediction system for the energy storage tank shell of claim 2, wherein, A physical information neural network model is arranged in the signal processing module (200), the input of the physical information neural network model is time and space position information, the sensing data and the working condition data, and the output is the pure mechanical strain field data and surface temperature field data of the energy storage cabinet shell.

4. The distributed optical fiber sensing crack monitoring and life prediction system for the energy storage tank shell of claim 3, wherein, The physical information neural network model is trained by a hybrid loss function, the hybrid loss function comprises a data driving item and a physical constraint item, the data driving item is used for minimizing the error between the coupling effect predicted by the network and the actually acquired sensing data, the physical constraint item is constructed based on a heat conduction partial differential equation and is used for constraining the temperature field output by the network to comply with physical laws, wherein the charging and discharging current data is used as a heat source item.

5. The distributed optical fiber sensing crack monitoring and life prediction system for the energy storage tank shell of claim 1, wherein, The state diagnosis module (300) comprises: a crack identification unit (310) is used for identifying a strain concentrated area in the pure mechanical strain field data by setting a threshold value or a clustering algorithm and judging the strain concentrated area as a crack event; a state quantification unit (320) is used for mapping the features of the strain concentrated area into equivalent physical parameters of the crack through a pre-trained proxy model; an evolution tracking unit (330) is used for establishing a time sequence file containing the position and quantitative state of each identified crack changing with time.

6. The distributed optical fiber sensing crack monitoring and life prediction system for the energy storage tank shell of claim 1, wherein, The life prediction module (400) adopts a long short-term memory (LSTM) model; the input of the LSTM model is the state evolution sequence of a specified crack in a past time window and a corresponding working condition data sequence.

7. The distributed optical fiber sensing crack monitoring and life prediction system for the energy storage tank shell of claim 6, wherein, The life prediction module (400) uses the trained LSTM model to forward predict the future state evolution trend of the crack in a rolling manner until the predicted crack state reaches a preset failure threshold, and the time difference from the current time to the time when the failure threshold is reached is taken as the predicted value of the remaining service life.

8. The distributed optical fiber sensing crack monitoring and life prediction system for the energy storage tank shell of claim 1, wherein, The application further comprises the following steps: A visualized alarm module (500) is configured to render the pure mechanical strain field and the identified crack information on a three-dimensional digital twin model of the energy storage cabinet in real time, and trigger a hierarchical alarm when the crack state deteriorates or the remaining service life is lower than a preset safety threshold.

9. The distributed optical fiber sensing crack monitoring and life prediction system for the energy storage tank shell of claim 5, wherein, The surrogate model in the state quantification unit (320) is a multilayer perceptron network trained based on finite element simulation data, which establishes a nonlinear mapping relationship between the peak value, width, integral area and other features of the strain concentration area and the equivalent length and depth of the crack.

10. A method for distributed optical fiber sensing crack monitoring and life prediction of an energy storage tank shell, characterized in that, The method comprises the following steps: S1: synchronously collecting coupled sensing data of distributed optical fiber sensors arranged along the energy storage cabinet shell and real-time working condition data of the energy storage cabinet; S2: inputting the coupled sensing data and the working condition data into a pre-trained physical information neural network model for intelligent decoupling to obtain pure mechanical strain field data; S3: analyzing the pure mechanical strain field data to identify and quantify the state of the crack and record the evolution history of the crack over time; S4: inputting the evolution history of the crack and the corresponding working condition data into a pre-trained long short-term memory network model to predict the remaining service life of the crack.

Images