A method and device for safety monitoring of energy storage battery compartments based on multi-source data

Abstract

This application provides a method and apparatus for safety monitoring of energy storage battery compartments based on multi-source data, relating to the field of battery energy storage systems. The method includes: acquiring multi-source parameter data corresponding to the energy storage battery compartment and environmental parameter data surrounding the battery compartment through multiple types of sensors configured in the battery compartment; inputting the multi-source parameter data and environmental parameter data into a graph neural network model, and performing cross-sensor data repair based on the graph neural network model; constructing the coupling relationship between the repaired source parameter data and environmental parameter data under multi-physical quantity constraints, and constructing an anomaly distribution prediction model based on the coupling relationship; obtaining safety anomaly characteristics corresponding to the energy storage battery compartment based on the coupling relationship and a multi-scale time-series trend extraction analysis method; inputting the safety anomaly characteristics into the anomaly distribution prediction model, and outputting the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly distribution prediction model. This application solves the problem that manual inspections are difficult to identify potential safety hazards in energy storage battery compartments in a timely manner, significantly increasing the possibility of fault propagation.

Description

Technical Field

[0001] This application relates to the field of battery energy storage systems, and in particular to a method and device for safety monitoring of energy storage battery compartments based on multi-source data. Background Technology

[0002] With the continuous expansion of offshore wind farm layout and the sustained increase in individual unit capacity, energy storage systems, as a core supporting technology for improving the grid connection stability of offshore wind power, reducing power fluctuations, and enhancing dispatch flexibility, are being rapidly deployed in offshore wind power platforms. Compared to terrestrial application environments, offshore energy storage battery compartments are exposed to extreme conditions such as high humidity, high salt spray, strong winds, and unstable communication links for extended periods, resulting in an extremely complex operating environment with numerous uncontrollable factors. Therefore, to ensure the safe operation of energy storage battery compartments at sea, safety monitoring has become a current research trend.

[0003] Traditional safety monitoring methods largely rely on manual inspections of energy storage battery compartments. This approach typically involves inspectors collaborating using offshore communication equipment to conduct routine checks at fixed time intervals, observing for abnormal temperature rises, structural deformation, gas leaks, or control unit malfunctions. However, this method alone is insufficient due to the harsh environment of offshore wind power platforms, limited on-site operating conditions, unstable communication links, and the long and infrequent nature of manual inspections. It fails to cover the continuous dynamic processes of system operation, often making it difficult to identify potential safety hazards in the energy storage battery compartment in a timely manner, significantly increasing the possibility of fault propagation.

[0004] Therefore, there is an urgent need for a method and device for safety monitoring of energy storage battery compartments based on multi-source data. Summary of the Invention

[0005] This application provides a method and device for safety monitoring of energy storage battery compartments based on multi-source data, which solves the problem that manual inspections are difficult to identify potential safety hazards in energy storage battery compartments in a timely manner, greatly increasing the possibility of fault spread.

[0006] The first aspect of this application provides a safety monitoring method for an energy storage battery compartment based on multi-source data. The method includes: acquiring multi-source parameter data corresponding to the energy storage battery compartment and environmental parameter data around the energy storage battery compartment through multiple types of sensors configured in the energy storage battery compartment; inputting the multi-source parameter data and environmental parameter data into a graph neural network model, and performing cross-sensor data repair based on the graph neural network model; constructing the coupling relationship between the repaired source parameter data and environmental parameter data under multiple physical quantity constraints, and constructing an anomaly distribution prediction model based on the coupling relationship; acquiring safety anomaly features corresponding to the energy storage battery compartment based on the coupling relationship and a multi-scale time-series trend extraction analysis method; inputting the safety anomaly features into the anomaly distribution prediction model, and outputting the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly distribution prediction model.

[0007] Optionally, multi-source parameter data corresponding to the energy storage battery compartment and environmental parameter data around the energy storage battery compartment are acquired through multiple types of sensors configured in the energy storage battery compartment. Specifically, this includes: acquiring multi-source parameter data corresponding to the energy storage battery compartment through multiple types of sensors configured in the energy storage battery compartment, including electrical parameter data, thermal parameter data, gas composition parameter data, and structural response parameter data; and acquiring environmental parameter data around the energy storage battery compartment, including humidity data, salt spray concentration data, wind speed and direction data, and air pressure change rate data.

[0008] Optionally, multi-source parameter data and environmental parameter data are input into a graph neural network model, and cross-sensor data repair is performed based on the graph neural network model. Specifically, this includes: constructing a heterogeneous graph structure based on the physical layout and data type associations of each sensor node in the energy storage battery compartment. In the heterogeneous graph structure, nodes represent data sources of multiple types of sensors, and edges represent cross-domain structural associations between nodes. Cross-domain structural associations include spatial adjacency, functional coupling, and historical covariance. The feature vectors corresponding to nodes in the heterogeneous graph structure are defined as the temporal features of the data within the time window, and environmental parameter data is introduced as a dynamic adjustment factor for edge weights. Feature propagation and aggregation operations are performed on the graph structure through the graph neural network model to perform cross-sensor data repair.

[0009] Optionally, the coupling relationship between the repaired source parameter data and environmental parameter data under multiple physical quantity constraints is constructed, and an anomaly distribution prediction model is constructed based on the coupling relationship. Specifically, this includes: constructing the coupling relationship between the repaired source parameter data and environmental parameter data under multiple physical quantity constraints, including the power excitation path of the physical field to the thermal field, the diffusion driving path of the thermal field to the gas field, and the performance feedback path of the gas field to the physical field; and generating an anomaly distribution prediction model by training a variational Bayesian inference structure based on the coupling relationship and combined with the distribution function.

[0010] Optionally, based on the coupling relationship and using a multi-scale time-series trend extraction analysis method, the safety anomaly characteristics corresponding to the energy storage battery compartment are obtained. Specifically, this includes: performing multi-scale decomposition of various parameter data corresponding to the coupling relationship in the time dimension to obtain parameter change sequences at different time scales; extracting intrinsic mode components from the parameter change sequences based on wavelet transform or empirical mode decomposition methods to obtain multi-scale mode components; calculating residual evolution tensors for the multi-scale mode components, and establishing a trend deviation index set in conjunction with the driving path in the coupling relationship, so as to use the trend deviation index set as safety anomaly characteristics.

[0011] Optionally, the safety anomaly characteristics are input into the anomaly distribution prediction model, and the safety monitoring results corresponding to the energy storage battery compartment are output based on the anomaly distribution prediction model. Specifically, this includes: calculating the anomaly probability density value corresponding to the safety anomaly characteristics based on the anomaly distribution prediction model; constructing a risk index function reflecting the degree of state deviation under multi-physical coupling based on the anomaly probability density value; performing a sliding integral on the risk index function within a preset time window to form an anomaly intensity cumulative curve; and outputting the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly intensity cumulative curve and preset multi-level safety state threshold values.

[0012] Optionally, the preset multi-level safety status thresholds include a first threshold and a second threshold. Based on the cumulative anomaly intensity curve and the preset multi-level safety status thresholds, the safety monitoring results corresponding to the energy storage battery compartment are output. Specifically, this includes: determining whether the cumulative anomaly intensity curve is less than or equal to the first threshold; if the cumulative anomaly intensity curve is less than or equal to the first threshold, the safety monitoring result is confirmed as a safe state; determining whether the cumulative anomaly intensity curve is greater than the first threshold and less than or equal to the second threshold; if the cumulative anomaly intensity curve is greater than the first threshold and less than or equal to the second threshold, the safety monitoring result is confirmed as a warning state; determining whether the cumulative anomaly intensity curve is greater than the second threshold; if the cumulative anomaly intensity curve is greater than the second threshold, the safety monitoring result is confirmed as a high-risk state.

[0013] A second aspect of this application provides a safety monitoring device for an energy storage battery compartment based on multi-source data. The device includes an acquisition module and a processing module, wherein...

[0014] The acquisition module is used to acquire multi-source parameter data corresponding to the energy storage battery compartment through multiple types of sensors configured in the energy storage battery compartment, as well as environmental parameter data around the energy storage battery compartment.

[0015] The processing module is used to input the multi-source parameter data and the environmental parameter data into a graph neural network model, and perform cross-sensor data repair based on the graph neural network model; construct the coupling relationship between the repaired source parameter data and the environmental parameter data under multi-physical quantity constraints, and construct an anomaly distribution prediction model based on the coupling relationship; obtain the safety anomaly characteristics corresponding to the energy storage battery compartment based on the coupling relationship and a multi-scale time-series trend extraction analysis method; input the safety anomaly characteristics into the anomaly distribution prediction model, and output the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly distribution prediction model.

[0016] A third aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described above.

[0017] A fourth aspect of this application provides a computer-readable storage medium storing a computer program, which is executed by a processor using the method described in any of the foregoing descriptions.

[0018] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0019] 1. This system acquires multi-source parameter data corresponding to the energy storage battery compartment and environmental parameter data surrounding the battery compartment using multiple types of sensors configured in the battery compartment. The multi-source parameter data and environmental parameter data are input into a graph neural network model, and cross-sensor data repair is performed based on the graph neural network model. The system constructs the coupling relationship between the repaired source parameter data and environmental parameter data under multiple physical quantity constraints, and builds an anomaly distribution prediction model based on this coupling relationship. Based on the coupling relationship and using a multi-scale time-series trend extraction analysis method, the system obtains the safety anomaly characteristics corresponding to the energy storage battery compartment. These safety anomaly characteristics are input into the anomaly distribution prediction model, and the model outputs the corresponding safety monitoring results for the energy storage battery compartment. This achieves highly robust and timely safety identification and early warning reasoning for the operating status of the energy storage battery compartment under complex operating conditions such as limited multi-source data quality, heterogeneous sampling nodes, and incomplete physical quantities. It improves the overall anomaly perception accuracy and response sensitivity under non-ideal sampling conditions such as local distortion, sensor degradation, or data loss, significantly reducing the probability of potential safety hazards occurring in the energy storage battery compartment.

[0020] 2. Based on the physical layout and data type relationships of the sensor nodes in the energy storage battery compartment, a heterogeneous graph structure is constructed. In the heterogeneous graph structure, nodes represent data sources of multiple types of sensors, and edges represent cross-domain structural relationships between nodes, including spatial adjacency, functional coupling, and historical covariance. The feature vectors corresponding to the nodes in the heterogeneous graph structure are defined as the temporal features of the data within the time window, and environmental parameter data is introduced as a dynamic adjustment factor for edge weights. Feature propagation and aggregation operations are performed on the graph structure through a graph neural network model to perform cross-sensor data repair. This enables structural perception and context reconstruction of low-quality data such as abnormal packet loss, data drift, and sampling distortion in a multi-source heterogeneous sensor network, improving the completeness and availability of key operating state variables collected by the energy storage battery compartment in a dynamic operating environment, and enhancing the robustness and accuracy of the model driven by multi-source data.

[0021] 3. Construct the coupling relationship between the repaired source parameter data and environmental parameter data under multi-physical quantity constraints. The coupling relationship includes the power excitation path of the physical field to the thermal field, the diffusion driving path of the thermal field to the gas field, and the performance feedback path of the gas field to the physical field. Based on the coupling relationship and combined with the distribution function, an anomaly distribution prediction model is generated by training the variational Bayesian inference structure. This enables high-precision modeling and forward uncertainty inference of potential non-stationary anomalies in the multi-physical coupling behavior inside the energy storage battery compartment. It effectively identifies the probability of occurrence and spatial distribution trend of abnormal states under complex coupling conditions and improves the ability to identify early abnormal signals and the speed of early warning response. Attached Figure Description

[0022] Figure 1 This is a flowchart illustrating a method for safety monitoring of an energy storage battery compartment based on multi-source data, provided in an embodiment of this application.

[0023] Figure 2 This is a schematic diagram of a module of a safety monitoring device for an energy storage battery compartment based on multi-source data, provided in an embodiment of this application.

[0024] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0025] Explanation of reference numerals in the attached figures: 21. Acquisition module; 22. Processing module; 301. Processor; 302. Communication bus; 303. User interface; 304. Network interface; 305. Memory. Detailed Implementation

[0026] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0027] The terminology used in the following embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to include the plural expressions as well, unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this application refers to and includes any or all possible combinations of one or more of the listed items.

[0028] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature, and in the description of the embodiments of this application, unless otherwise stated, "multiple" means two or more.

[0029] To enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0030] Please refer to Figure 1 The diagram illustrates a flowchart of a safety monitoring method for an energy storage battery compartment based on multi-source data, provided in an embodiment of this application. The flowchart mainly includes the following steps: S101 to S105.

[0031] Step S101: Acquire multi-source parameter data corresponding to the energy storage battery compartment through multiple types of sensors configured in the energy storage battery compartment, and acquire environmental parameter data around the energy storage battery compartment.

[0032] Specifically, this application proposes a safety monitoring method for the complex operating environment of energy storage battery compartments in offshore wind power platforms, characterized by high humidity and salt spray, strong wind-vibration coupling, unstable communication, and enclosed structures. This method features multi-source heterogeneous data fusion, adaptive modeling, and risk prediction capabilities. In this complex scenario, the safety status of the energy storage battery compartment is affected by multiple physical field disturbances. A single parameter or isolated signal cannot accurately reflect its true operational risk. Therefore, comprehensive, layered, and asynchronous multi-source parameter acquisition from both inside and outside the compartment is necessary to support subsequent spatiotemporal coupling modeling and dynamic safety assessment.

[0033] Multi-source parameter data corresponding to the energy storage battery compartment is acquired by various types of sensors configured inside and outside the battery compartment. These sensors include, but are not limited to, electrical parameter acquisition units (for acquiring current, voltage, power factor, voltage fluctuation rate, etc.), thermal parameter acquisition units (for acquiring temperature distribution, temperature rise rate, heat flux density, etc. at multiple locations), gas detection modules (for monitoring combustible gas concentration, humidity-induced gas leakage diffusion rate, and changes in volatile chemical components), structural response sensors (such as triaxial accelerometers, strain gauges, impact response detectors, etc.), and external environment monitoring modules that work in conjunction with these sensors (such as wind speed and direction sensors, sea fog concentration detectors, external temperature and humidity sensors, and air pressure monitoring devices). Therefore, the multi-source parameter data includes, but is not limited to, electrical parameter data, thermal parameter data, gas composition parameter data, and structural response parameter data. The environmental parameter data includes, but is not limited to, humidity data, salt spray concentration data, wind speed and direction data, and air pressure change rate data. Among them, the physical deployment of various types of sensors takes key units (such as cell arrangement area, heat dissipation unit, high voltage bus interface, door sealing boundary, etc.) as the core deployment area, and combines the principle of multi-location spatial redundancy deployment to enhance the overall monitoring redundancy and robustness.

[0034] Meanwhile, environmental parameter data acquisition spans the physical boundary area between the cabin and the outside world, supporting the modeling requirements for the indirect impact of environmental disturbances on the cabin's operational status. For example, wind speed changes causing cabin micro-vibrations and gas disturbance path reconstruction, and increased salt spray concentration inducing short-term sensor inaccuracies or electrical corrosion effects, all need to be included as environmental parameter inputs within a unified modeling framework. Furthermore, to ensure stable operation of the acquisition system under conditions of high electromagnetic interference and weak signal obstruction, sensor nodes support local data caching, multi-period acquisition, and asynchronous communication upload mechanisms to guarantee the integrity and temporal consistency of multi-source data in high-noise scenarios.

[0035] Step S102: Input multi-source parameter data and environmental parameter data into the graph neural network model, and perform cross-sensor data repair based on the graph neural network model.

[0036] Specifically, the multi-source parameter data and environmental parameter data obtained in step S101 are input together into the graph neural network model. The graph neural network model not only considers the changing trends of the sensor's own data, but also uses the structural and functional relationships between it and other sensors to fuse contextual information, thereby achieving high-fidelity repair of data anomalies or missing regions.

[0037] In one possible implementation, step S102 further includes: constructing a heterogeneous graph structure based on the physical layout and data type association of each sensor node in the energy storage battery compartment. In the heterogeneous graph structure, nodes represent data sources of multiple types of sensors, and edges represent cross-domain structural associations between nodes. Cross-domain structural associations include spatial adjacency, functional coupling, and historical covariance. The feature vectors corresponding to the nodes in the heterogeneous graph structure are defined as data time-series features within a time window, and environmental parameter data are introduced as dynamic adjustment factors for edge weights. Feature propagation and aggregation operations are performed on the graph structure through a graph neural network model to perform cross-sensor data repair.

[0038] Specifically, in the graph structure construction phase, a heterogeneous graph structure G = (V, E) is constructed based on the actual deployment location information and sensing task type of each sensor in the energy storage battery compartment. The node set V contains all configured sensor units, such as voltage, current, temperature, gas concentration, vibration, and humidity types, while the edge set E describes the cross-domain structural relationships between any two nodes. These relationships are established according to three mechanisms: first, spatial adjacency, where two sensors deployed in physically close compartment structural areas or hot gas channels are linked by a spatial connection edge; second, functional coupling, such as physical coupling between current nodes and temperature nodes, or voltage nodes and gas nodes, under thermal dissipation mechanisms, resulting in functional edges; and third, historical covariance, where significant covariance relationships are extracted and covariance edges are formed by analyzing the time-series correlation coefficients, cross-information, or synchronization disturbance frequencies between two nodes in historical operating data.

[0039] Secondly, in the feature representation definition stage, for each node v i ∈V defines its node feature vector The feature vector is a combination of multidimensional data time-series features collected within a sliding time window, such as mean, variance, temporal gradient, and frequency domain energy spectral density. In this process, a labeling encoding and missing value masking mechanism is used for data segments containing missing or outliers, so that the graph model can identify uncertain regions and assign different weights in the aggregation operation.

[0040] During the dynamic adjustment phase of edge weights, external environmental parameters (such as wind speed w, air pressure p, salt spray concentration s, and relative humidity h) are introduced as adjustment factors for the edge weight function, and the adjustment is applied to each edge e. ij ∈E calculate its weight:

[0041] α ij =f(sim(h) i ,h j ),D ij ,φ(w,p,s,h))

[0042] Where sim(h) i ,h j ) represents the similarity of node features (such as Euclidean distance or cosine similarity), D ij Represents the physical distance between nodes. φ(w,p,s,h) is a perturbation function that is adaptively adjusted according to the environmental conditions. It is used to amplify or suppress the propagation capability of edges, ensuring that critical structural connections are preserved even in harsh environments.

[0043] Finally, in the graph neural propagation stage, models such as Graph Convolutional Networks (GCN) or Graph Attention Networks (GAT) are applied to execute feature propagation mechanisms on the constructed graph structure. This involves reconstructing and predicting anomalous data through message passing between adjacent nodes. Taking Graph Attention Networks as an example, its propagation update rule can be expressed as:

[0044]

[0045] in Let W be the edge attention coefficient of the l-th layer. (l) Let σ be the weight matrix, and σ be the activation function. This represents the set of neighbors of node i. This process is executed iteratively across multiple graph network layers, gradually aggregating information at the nodes, ultimately achieving the repair and interpolation of abnormal or missing data segments.

[0046] The output consists of a set of repaired multi-source parameter data and environmental parameter data. The data of each sensor node has been enhanced in accuracy based on global association and local semantics of the graph structure, laying the data foundation for subsequent multiphysics coupling relationship modeling. This implementation method has strong robustness, topological interpretability, and cross-domain fault tolerance, and is particularly suitable for offshore energy storage scenarios with high disturbance, high noise, and high failure rate.

[0047] Step S103: Construct the coupling relationship between the repaired source parameter data and environmental parameter data under multiple physical quantity constraints, and construct an anomaly distribution prediction model based on the coupling relationship.

[0048] Specifically, based on the restored, complete, and highly reliable multi-source parameter data and environmental parameter data, a coupling relationship that can truly reflect the multi-physics interaction mechanism of the energy storage battery compartment is established. On this basis, an anomaly distribution prediction model with probabilistic reasoning capabilities is constructed to dynamically quantify the safety risks that may occur under complex coupling conditions.

[0049] In one possible implementation, step S103 further includes: constructing the repaired source parameter data and environmental parameter data, and the coupling relationship under multiple physical quantity constraints, the coupling relationship including the power excitation path of the physical field to the thermal field, the diffusion driving path of the thermal field to the gas field, and the performance feedback path of the gas field to the physical field; based on the coupling relationship and combined with the distribution function, to generate an anomaly distribution prediction model by training a variational Bayesian inference structure.

[0050] Specifically, based on the repaired parameter data set output in step S102, a nonlinear coupling path is first constructed between three key physical variables: the physical field (electrical parameters), the thermal field (temperature), and the gas field (gas concentration). This coupling modeling not only captures single-dimensional trends but also reflects the dynamic causal interactions between the three types of physical variables. Especially in highly turbulent marine environments, this path exhibits significant nonlinearity, hysteresis, and local amplification.

[0051] In one possible implementation, the coupling relationship includes the following three types of path mechanisms:

[0052] The Joule heating mechanism, which represents the heat generated by the action of current and voltage on a structure, is characterized by the following integral model:

[0053]

[0054]

[0055] Among them, I i (t): Current data of the i-th sensor node at time t, derived from the current-type sensor node features repaired in S102. R i (τ): The equivalent resistance of the i-th node at time τ. If it cannot be directly measured, it can be estimated by voltage-current relationship, or it can be fitted by historical data. It supports graph neural network output. Q i (t): The cumulative Joule heat generated by the current within the time window [t-Δ,t], measured in Joules. T i (t): The temperature response at the i-th point, representing the temperature rise caused by heat acting on a local part of the cabin. T env (t): Ambient temperature, obtained from an ambient temperature sensor, is part of environmental parameter data. m i The equivalent mass of a local structure within the cabin is used for heat capacity calculations and is determined based on the battery structure materials. i Equivalent specific heat capacity, expressed in J / (kg·K), represents the heating energy required per unit mass. i The heat flow disturbance term represents the temperature change from sources other than the physical field, such as sunlight and wind-cooled disturbances.

[0056] The thermogenic volatilization mechanism, which illustrates how temperature fields influence gas release and diffusion, is represented by the following gas concentration model:

[0057]

[0058] Where C(x,t): the concentration of the gas at position x and time t, the unit can be ppm or mg / m³. 3 The value originates from a gas sensor. D(T(x,t)): diffusion coefficient, which varies with temperature and is a temperature-dependent function that physically reflects the driving effect of molecular thermal motion on diffusion behavior. The spatial Laplace term of gas concentration reflects the rate of change of the spatial distribution of gas. S(x,t): Gas source term, representing the rate of new gas addition caused by thermal decomposition, electrolysis, or chemical reaction, in units of concentration / time.

[0059] Changes in gas concentration cause fluctuations in electrical performance (such as insulation degradation and voltage interference), which can be represented by the following model:

[0060] V i obs (t)=V i ideal (t)-α·ln(1+C i (t))

[0061] Among them, V i obs (t): Measured voltage value, derived from the voltage sensor node after S102 repair. V i ideal (t): Theoretical expected voltage value, derived from control strategy, steady-state modeling, or historical average reference. C i (t): Gas concentration at the i-th node, derived from gas sensor node data. α: Concentration influence coefficient, representing the intensity of voltage interference caused by a unit change in concentration, typically obtained through scene data fitting, with units of V.

[0062] After completing the above coupling modeling, the three types of variables are uniformly combined into a joint state tensor:

[0063] Z(t)={I i (t),T i (t),C i (t),V i obs (t),H i (t),P i (t)}

[0064] Here, Z(t) represents the state vector tensor at time t, which contains a combination of multiple physical field features.

[0065] Subsequently, its corresponding multidimensional joint probability distribution function is constructed:

[0066] P(Z(t)|θ)=P(I|H,P)·P(T|I)·P(C|T)·P(V|C)

[0067] Where P(Z(t)|θ): conditional joint probability density function, representing the probability value of the current state vector given model parameters θ, measuring its "abnormality". θ is the parameter set of the joint distribution model, including the mean, variance, covariance matrix, and structural coefficients of each sub-distribution, which are statistics that need to be learned. To address the problems of high parameter dimensionality and difficulty in obtaining the true posterior, a variational Bayesian inference structure is introduced, with an approximate posterior q(θ)≈P(θ|Z). q(θ) is used to approximate the variational distribution of the true posterior distribution P(θ|Z). The structure generally adopts a differentiable Gaussian distribution, a normal mixture, or a neural network transformation generator. By maximizing the lower bound of evidence (ELBO):

[0068]

[0069] in, The variational evidence lower bound is the objective function for variational inference. Maximizing this function approximates the true posterior. D represents the expected log-likelihood of the current data in the model, given an approximate posterior. KL [q(θ)||P(θ)]: KL divergence between the variational distribution and the prior, a regularization term that constrains the learned model from deviating excessively from the prior distribution. The final output variational approximation distribution q * (θ) serves as the core structure of the anomaly distribution prediction model, enabling continuous estimation of the probability interval of anomalies in the energy storage battery compartment under the background of multi-physics coupling, thus providing a predictive basis for subsequent risk quantification and graded response.

[0070] Step S104: Based on the coupling relationship and using the multi-scale time-series trend extraction and analysis method, obtain the safety anomaly features corresponding to the energy storage battery compartment.

[0071] Specifically, based on coupled modeling, multi-scale trend extraction and anomaly offset identification are performed on the temporal evolution behavior of various key physical parameters. By revealing the local anomaly evolution trajectory of the state response between various subsystems of the energy storage battery compartment at different time scales, the deep-seated safety risks in the multi-physics interaction system can be mined and characterized.

[0072] In one possible implementation, step S104 further includes: performing multi-scale decomposition on the various parameter data corresponding to the coupling relationship in the time dimension to obtain parameter change sequences at different time scales; extracting intrinsic mode components from the parameter change sequences based on wavelet transform or empirical mode decomposition methods to obtain multi-scale mode components; calculating residual evolution tensors on the multi-scale mode components, and establishing a trend deviation index set in combination with the driving path in the coupling relationship, so as to use the trend deviation index set as a safety anomaly feature.

[0073] Specifically, each component of the state tensor Z(t) constructed in step S103 is decomposed into a multi-scale form within a selected time window [t-Δ,t]. This process aims to decompose the original data sequence into several information components with different time resolutions to capture different physical change mechanisms such as rapid disturbances (e.g., instantaneous current jumps), mesoscale fluctuations (e.g., thermal diffusion processes), and slow trends (e.g., gas accumulation processes).

[0074] The Discrete Wavelet Transform (DWT) or Empirical Mode Decomposition (EMD) methods are used to process each parameter sequence x. i Processing is performed on (t)∈Z(t). If EMD is used, the processing procedure is as follows:

[0075]

[0076] in Let r represent the k-th eigenmode component of the i-th parameter. (i) (t) represents the residual trend term, and K represents the total number of modes. Each IMF k The components reflect the characteristics of state changes at different frequency scales.

[0077] Subsequently, for each type of physical parameter component (such as T) i (t), C i (t), V i (t) Extract its cross-scale residual evolution behavior, and define the residual evolution tensor as:

[0078]

[0079] The residual evolution tensor is used to measure the difference in response amplitude between different frequency components, thereby revealing possible anomalous local perturbations or trend abrupt changes.

[0080] Based on this, and combining the three types of coupling paths (electro-induced heat, thermally driven gas, and gas-induced electrical disturbance) defined in step S103, a trend deviation index set is constructed to describe the strength of the deviation between the actual evolution process and the predicted path of the coupling model. For example:

[0081] For the path of the thermal field induced by the rational field, construct:

[0082]

[0083] Where, δ T (t) represents the path of the thermal field induced by the ideal field. To calculate the temperature rise based on the power model in step S103, T i (t) represents the actual observed value.

[0084] For the thermally driven gas field path, construct:

[0085]

[0086] Where, δ C (t) represents the path of the gas field driven by the thermal field. This represents the gas concentration predicted based on a diffusion model.

[0087] For the gas field feedback field path, construct:

[0088]

[0089] Where, δ V (t) represents the gas field feedback path.

[0090] Each of the above δ * The δ(t) values ​​all constitute trend deviation indicators, reflecting the degree of deviation of the system state from the expected model of the coupling path, and are physical anomalies caused by coupling mismatch. Ultimately, all δ... T (t), δ C (t), δ V (t) and residual tensor E (i) (t) together form an eigenvector:

[0091] F abn (t)=[δ T (t),δ C (t),δ V (t),‖E (T) (t)‖,‖E (C) (t)‖,‖E (V) (t)‖,…]

[0092] Among them, F abn (t) is the jointly constructed feature vector, which is the safety anomaly feature of the energy storage battery compartment in the multi-scale physical evolution process at the current moment. It is used to characterize the degree and direction of the state evolution deviating from the model expectation, and provides input for subsequent anomaly risk scoring and level determination.

[0093] Step S105: Input the safety anomaly characteristics into the anomaly distribution prediction model, and output the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly distribution prediction model.

[0094] Specifically, the security anomaly feature vector extracted in step S104 is used as input, and the anomaly distribution prediction model established in step S103 is used for inference calculation to finally output the real-time security status level.

[0095] In one possible implementation, step S105 further includes: calculating the anomaly probability density value corresponding to the safety anomaly feature based on the anomaly distribution prediction model; constructing a risk index function reflecting the degree of state deviation under multi-physical coupling based on the anomaly probability density value; performing a sliding integral on the risk index function within a preset time window to form an anomaly intensity cumulative curve; and outputting the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly intensity cumulative curve and preset multi-level safety state threshold values.

[0096] Specifically, firstly, the security anomaly feature vector F constructed in step S104 is... abn (t) Input the anomaly distribution prediction model trained in step S103. This model is a joint probability model based on variational Bayesian inference, and its goal is to calculate the anomaly probability density value of the current state based on the input features. This process can be described as follows:

[0097] p θ (F abn (t))=P(F abn (t)∣θ)

[0098] Where θ is the parameter set of the anomaly distribution prediction model, obtained through variational inference training; p θ This represents the density value of the current input feature under normal distribution; the lower the value, the higher the degree of abnormality.

[0099] Subsequently, a risk index function reflecting the degree of deviation from the multi-physics coupling state is constructed based on this probability density value. This function can be defined in one possible way as its negative logarithmic form:

[0100] R(t) = -log(p) θ (F abn (t)))

[0101] This function is monotonic and nonnegative. The more the feature deviates from the training distribution, the smaller its density and the larger the risk value R(t), thus it can be used to express the overall safety deviation of the system at the current moment.

[0102] Further, to eliminate short-term fluctuation interference and reflect the persistence of anomalies in the time dimension, a sliding integration operation is performed on the risk index function R(t) within the time window [t - Δ, t] to construct the cumulative anomaly intensity curve S(t):

[0103]

[0104] where w(τ) is a time decay function (such as exponential weight or Gaussian window), which is used to enhance the contribution of the latest risk.

[0105] In a possible implementation manner, step S105 further includes: determining whether the cumulative anomaly intensity curve is less than or equal to the first threshold value; if the cumulative anomaly intensity curve is less than or equal to the first threshold value, then confirming that the safety monitoring result is in a safe state; determining whether the cumulative anomaly intensity curve is greater than the first threshold value and less than or equal to the second threshold value; if the cumulative anomaly intensity curve is greater than the first threshold value and less than or equal to the second threshold value, then confirming that the safety monitoring result is in a warning state; determining whether the cumulative anomaly intensity curve is greater than the second threshold value; if the cumulative anomaly intensity curve is greater than the second threshold value, then confirming that the safety monitoring result is in a high-risk state.

[0106] Specifically, according to the preset multi-level safety state threshold values {T1, T2}, a hierarchical determination logic is executed:

[0107] If S(t) < T1, then output "normal state";

[0108] If T1 ≤ S(t) < T2, then output "warning state";

[0109] If S(t) ≥ T2, then output "high-risk state".

[0110] Among them, T1 is the first threshold value and T2 is the second threshold value. The threshold values can be determined through retrospective analysis of known accidents and normal operation conditions in historical monitoring data, or can be dynamically self-learned and adjusted in combination with Bayesian quantile inference.

[0111] This application employs the aforementioned method to acquire multi-source parameter data corresponding to the energy storage battery compartment and environmental parameter data surrounding the battery compartment using multiple types of sensors configured in the battery compartment. The multi-source parameter data and environmental parameter data are input into a graph neural network model, and cross-sensor data repair is performed based on the graph neural network model. The repaired source parameter data and environmental parameter data are then constructed, establishing a coupling relationship under multiple physical quantity constraints. An anomaly distribution prediction model is built based on this coupling relationship. Based on this coupling relationship and a multi-scale time-series trend extraction analysis method, safety anomaly characteristics corresponding to the energy storage battery compartment are obtained. These safety anomaly characteristics are input into the anomaly distribution prediction model, and the model outputs the corresponding safety monitoring results for the energy storage battery compartment. This achieves highly robust and timely safety identification and early warning reasoning for the operating status of the energy storage battery compartment under complex operating conditions such as limited multi-source data quality, heterogeneous sampling nodes, and incomplete physical quantities. It improves the overall anomaly perception accuracy and response sensitivity under non-ideal sampling conditions such as local distortion, sensor degradation, or data loss, significantly reducing the probability of potential safety hazards occurring in the energy storage battery compartment.

[0112] Please refer to Figure 2 This illustration shows a schematic diagram of a multi-source data-based energy storage battery compartment safety monitoring device according to an embodiment of this application. The device includes an acquisition module 21 and a processing module 22, wherein...

[0113] The acquisition module 21 is used to acquire multi-source parameter data corresponding to the energy storage battery compartment through multiple types of sensors configured in the energy storage battery compartment, and to acquire environmental parameter data around the energy storage battery compartment.

[0114] Processing module 22 is used to input the multi-source parameter data and the environmental parameter data into a graph neural network model, and perform cross-sensor data repair based on the graph neural network model; construct the coupling relationship between the repaired source parameter data and the environmental parameter data under multi-physical quantity constraints, and construct an anomaly distribution prediction model based on the coupling relationship; obtain the safety anomaly characteristics corresponding to the energy storage battery compartment based on the coupling relationship and a multi-scale time-series trend extraction analysis method; input the safety anomaly characteristics into the anomaly distribution prediction model, and output the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly distribution prediction model.

[0115] In one possible implementation, the acquisition module 21 is used to acquire multi-source parameter data corresponding to the energy storage battery compartment through multiple types of sensors configured in the energy storage battery compartment, and to acquire environmental parameter data around the energy storage battery compartment. Specifically, this includes: acquiring multi-source parameter data corresponding to the energy storage battery compartment through multiple types of sensors configured in the energy storage battery compartment, the multi-source parameter data including electrical parameter data, thermal parameter data, gas composition parameter data, and structural response parameter data; and acquiring environmental parameter data around the energy storage battery compartment, the environmental parameter data including humidity data, salt spray concentration data, wind speed and direction data, and air pressure change rate data.

[0116] In one possible implementation, the processing module 22 is used to input multi-source parameter data and environmental parameter data into a graph neural network model, and perform cross-sensor data repair based on the graph neural network model. Specifically, this includes: constructing a heterogeneous graph structure based on the physical layout and data type association relationships of each sensor node in the energy storage battery compartment. In the heterogeneous graph structure, nodes represent data sources of multiple types of sensors, and edges represent cross-domain structural association relationships between nodes. Cross-domain structural association relationships include spatial adjacency relationships, functional coupling relationships, and historical covariance relationships. On the heterogeneous graph structure, the feature vector corresponding to the node is defined as the data temporal characteristics within the time window, and environmental parameter data is introduced as a dynamic adjustment factor for edge weights. The graph neural network model is used to perform feature propagation and aggregation operations on the graph structure to perform cross-sensor data repair.

[0117] In one possible implementation, the processing module 22 is used to construct the coupling relationship between the repaired source parameter data and environmental parameter data under multiple physical quantity constraints, and to construct an anomaly distribution prediction model based on the coupling relationship. Specifically, this includes: constructing the coupling relationship between the repaired source parameter data and environmental parameter data under multiple physical quantity constraints, the coupling relationship including the power excitation path of the physical field to the thermal field, the diffusion driving path of the thermal field to the gas field, and the performance feedback path of the gas field to the physical field; and generating an anomaly distribution prediction model by training a variational Bayesian inference structure based on the coupling relationship and in combination with the distribution function.

[0118] In one possible implementation, the processing module 22 is used to obtain the safety anomaly features corresponding to the energy storage battery compartment based on the coupling relationship and the multi-scale time series trend extraction analysis method. Specifically, it includes: performing multi-scale decomposition of various parameter data corresponding to the coupling relationship in the time dimension to obtain parameter change sequences at different time scales; extracting intrinsic mode components from the parameter change sequences based on wavelet transform or empirical mode decomposition methods to obtain multi-scale mode components; calculating residual evolution tensors for the multi-scale mode components, and establishing a trend deviation index set in combination with the driving path in the coupling relationship, so as to use the trend deviation index set as safety anomaly features.

[0119] In one possible implementation, the processing module 22 is used to input the safety anomaly characteristics into the anomaly distribution prediction model and output the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly distribution prediction model. Specifically, this includes: calculating the anomaly probability density value corresponding to the safety anomaly characteristics based on the anomaly distribution prediction model; constructing a risk index function reflecting the degree of state deviation under multi-physical coupling based on the anomaly probability density value; performing a sliding integral on the risk index function within a preset time window to form an anomaly intensity cumulative curve; and outputting the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly intensity cumulative curve and preset multi-level safety state threshold values.

[0120] In one possible implementation, the processing module 22 is used to preset multi-level safety state thresholds, including a first threshold and a second threshold. Based on the cumulative abnormal intensity curve and the preset multi-level safety state thresholds, it outputs the safety monitoring result corresponding to the energy storage battery compartment. Specifically, this includes: determining whether the cumulative abnormal intensity curve is less than or equal to the first threshold; if the cumulative abnormal intensity curve is less than or equal to the first threshold, then confirming the safety monitoring result as a safe state; determining whether the cumulative abnormal intensity curve is greater than the first threshold and less than or equal to the second threshold; if the cumulative abnormal intensity curve is greater than the first threshold and less than or equal to the second threshold, then confirming the safety monitoring result as a warning state; determining whether the cumulative abnormal intensity curve is greater than the second threshold; if the cumulative abnormal intensity curve is greater than the second threshold, then confirming the safety monitoring result as a high-risk state.

[0121] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0122] This application also provides an electronic device. (See reference...) Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: at least one processor 301, at least one communication bus 302, a user interface 303, at least one network interface 304, and a memory 305.

[0123] The communication bus 302 is used to enable communication between these components.

[0124] The user interface 303 may include a display screen and a camera. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0125] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0126] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 305, and by calling data stored in memory 305. Optionally, the processor 301 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 301 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor 301.

[0127] The memory 305 may include random access memory (RAM) or read-only memory. Optionally, the memory 305 may include a non-transitory computer-readable storage medium. The memory 305 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. (Refer to...) Figure 3 The memory 305, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and a safety monitoring application for the energy storage battery compartment based on multi-source data.

[0128] exist Figure 3 In the illustrated electronic device, the user interface 303 is primarily used to provide an input interface for the user and acquire user input data; while the processor 301 can be used to call the energy storage battery compartment safety monitoring application stored in the memory 305 based on multi-source data. When executed by one or more processors 301, the electronic device performs one or more of the methods described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0129] This application also provides a computer-readable storage medium storing instructions. When executed by one or more processors, these instructions cause an electronic device to perform one or more of the methods described in the above embodiments.

[0130] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0131] In the various embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between apparatuses or units may be electrical or other forms.

[0132] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0133] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0134] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0135] The above description is merely an exemplary embodiment disclosed in this application and should not be construed as limiting the scope of this application. Any equivalent changes and modifications made in accordance with the teachings of this application shall still fall within the scope of this application.

[0136] This application is intended to cover any variations, uses, or adaptations disclosed herein that follow the general principles disclosed herein and include common knowledge or customary technical means in the art that are not described in this application.

Claims

1. A method for safety monitoring of energy storage battery compartments based on multi-source data, characterized in that, The method includes: Multi-source parameter data corresponding to the energy storage battery compartment is acquired by multiple types of sensors configured in the energy storage battery compartment, as well as environmental parameter data around the energy storage battery compartment. The multi-source parameter data and the environmental parameter data are input into the graph neural network model, and cross-sensor data repair is performed based on the graph neural network model; The process involves constructing the coupling relationship between the repaired source parameter data and the environmental parameter data under multiple physical quantity constraints, and then building an anomaly distribution prediction model based on this coupling relationship. Specifically, this includes: constructing the coupling relationship between the repaired source parameter data and the environmental parameter data under multiple physical quantity constraints, whereby the coupling relationship includes the power excitation path of the electric field on the thermal field, the diffusion driving path of the thermal field on the gas field, and the performance feedback path of the gas field on the electric field; and generating the anomaly distribution prediction model based on this coupling relationship and in conjunction with a distribution function by training a variational Bayesian inference structure. Based on the coupling relationship, and using a multi-scale time-series trend extraction and analysis method, the safety anomaly characteristics corresponding to the energy storage battery compartment are obtained. Specifically, this includes: performing multi-scale decomposition of various parameter data corresponding to the coupling relationship in the time dimension to obtain parameter change sequences at different time scales; extracting intrinsic mode components from the parameter change sequences based on wavelet transform or empirical mode decomposition methods to obtain multi-scale mode components; calculating residual evolution tensors for the multi-scale mode components, and establishing a trend deviation index set in conjunction with the driving path in the coupling relationship, so as to use the trend deviation index set as the safety anomaly characteristics. The safety anomaly characteristics are input into the anomaly distribution prediction model, and the safety monitoring results corresponding to the energy storage battery compartment are output according to the anomaly distribution prediction model.

2. The method according to claim 1, characterized in that, The process of acquiring multi-source parameter data corresponding to the energy storage battery compartment through multiple types of sensors configured in the energy storage battery compartment, and acquiring environmental parameter data around the energy storage battery compartment, specifically includes: Multi-source parameter data corresponding to the energy storage battery compartment is acquired by multiple types of sensors configured in the energy storage battery compartment. The multi-source parameter data includes electrical parameter data, thermal parameter data, gas composition parameter data, and structural response parameter data. The environmental parameter data around the energy storage battery compartment is acquired, including humidity data, salt spray concentration data, wind speed and direction data, and air pressure change rate data.

3. The method according to claim 1, characterized in that, The step of inputting the multi-source parameter data and the environmental parameter data into a graph neural network model, and performing cross-sensor data repair based on the graph neural network model, specifically includes: Based on the physical layout and data type association of each sensor node in the energy storage battery compartment, a heterogeneous graph structure is constructed. In the heterogeneous graph structure, nodes represent the data sources of the multiple types of sensors, and edges represent the cross-domain structural associations between the nodes. The cross-domain structural associations include spatial adjacency, functional coupling, and historical covariance. On the heterogeneous graph structure, the feature vector corresponding to the node is defined as the data temporal feature within the time window, and the environmental parameter data is introduced as a dynamic adjustment factor for edge weights; The graph neural network model performs feature propagation and aggregation operations on the graph structure to perform cross-sensor data repair.

4. The method according to claim 1, characterized in that, The step of inputting the safety anomaly characteristics into the anomaly distribution prediction model and outputting the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly distribution prediction model specifically includes: Calculate the anomaly probability density value corresponding to the security anomaly feature based on the anomaly distribution prediction model; Based on the aforementioned anomaly probability density value, a risk index function reflecting the degree of state deviation under multi-physical coupling states is constructed; The risk index function is integrated over a preset time window to form a cumulative curve of abnormal intensity. Based on the cumulative curve of abnormal intensity and the preset multi-level safety state threshold, the safety monitoring results corresponding to the energy storage battery compartment are output.

5. The method according to claim 4, characterized in that, The preset multi-level safety state threshold values ​​include a first threshold value and a second threshold value. The step of outputting the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly intensity cumulative curve and the preset multi-level safety state threshold values ​​specifically includes: Determine whether the cumulative curve of the abnormal intensity is less than or equal to the first threshold value; If the cumulative curve of the abnormal intensity is less than or equal to the first threshold value, then the safety monitoring result is confirmed to be in a safe state. Determine whether the cumulative curve of the abnormal intensity is greater than the first threshold value and less than or equal to the second threshold value; If the cumulative curve of the abnormal intensity is greater than the first threshold value and less than or equal to the second threshold value, then the safety monitoring result is confirmed to be in an early warning state. The cumulative curve of the anomaly intensity is determined to be greater than the second threshold value; If the cumulative curve of the abnormal intensity is greater than the second threshold value, then the safety monitoring result is confirmed to be a high-risk state.

6. A safety monitoring device for an energy storage battery compartment based on multi-source data, characterized in that, The device includes an acquisition module and a processing module, wherein, The acquisition module is used to acquire multi-source parameter data corresponding to the energy storage battery compartment through multiple types of sensors configured in the energy storage battery compartment, and to acquire environmental parameter data around the energy storage battery compartment. The processing module is used to input the multi-source parameter data and the environmental parameter data into a graph neural network model, and perform cross-sensor data repair based on the graph neural network model; construct the coupling relationship between the repaired source parameter data and the environmental parameter data under multi-physical quantity constraints, and construct an anomaly distribution prediction model based on the coupling relationship, specifically including: constructing the coupling relationship between the repaired source parameter data and the environmental parameter data under multi-physical quantity constraints, the coupling relationship including the power excitation path of the electric field to the thermal field, the diffusion driving path of the thermal field to the gas field, and the performance feedback path of the gas field to the electric field; generating the anomaly distribution prediction model by training a variational Bayesian inference structure based on the coupling relationship and in combination with the distribution function; and constructing an anomaly distribution prediction model based on the coupling relationship. The system establishes a safety anomaly feature for the energy storage battery compartment based on a multi-scale time-series trend extraction analysis method. Specifically, this includes: decomposing various parameter data corresponding to the coupling relationship into multi-scale parameters over time to obtain parameter change sequences at different time scales; extracting intrinsic mode components from the parameter change sequences using wavelet transform or empirical mode decomposition methods to obtain multi-scale mode components; calculating residual evolution tensors for the multi-scale mode components and establishing a trend deviation index set based on the driving path in the coupling relationship, using the trend deviation index set as the safety anomaly feature; inputting the safety anomaly feature into the anomaly distribution prediction model, and outputting the safety monitoring results corresponding to the energy storage battery compartment based on the anomaly distribution prediction model.

7. An electronic device, characterized in that, The device includes a processor, a communication bus, a user interface, a network interface, and a memory. The memory is used to store instructions. The user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1 to 5.

Images