Method and system for evaluating multi-dimensional safety performance of sodium-ion battery energy storage system

Abstract

The invention is applicable to the technical field of energy storage performance evaluation, and provides a multi-dimensional safety performance evaluation method and system for a sodium-ion battery energy storage system, and the method comprises the steps: obtaining physical image data and electrochemical sensor data of the sodium-ion battery energy storage system; processing by using the target detection model to obtain a physical safety risk refinement index, and processing by using the time sequence analysis model and the risk prediction model to obtain an electrochemical safety risk refinement index; adopting subjective and objective dynamic collaborative weighting to distribute weights, and performing first-layer fusion to obtain a physical safety risk index and an electrochemical safety risk index; carrying out second-layer weighted fusion to obtain a system safety performance index; and generating a safety control instruction for the sodium ion battery energy storage system based on the physical safety risk index, the electrochemical safety risk index and the system safety performance index. According to the invention, the comprehensiveness of safety monitoring of the energy storage system, the real-time performance of early warning, the accuracy of evaluation and the intelligence of control are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of energy storage performance evaluation technology, and in particular to a method and system for multi-dimensional safety performance evaluation of sodium-ion battery energy storage systems. Background Technology

[0002] In safety performance evaluation studies of sodium-ion battery energy storage systems, data acquisition is generally based on electrochemical sensors such as temperature, voltage, and current sensors deployed within the system. By monitoring this sensor data in real time and setting fixed safety thresholds, the system's safety status can be determined and warnings issued. Some methods further employ expert-based weighting methods or statistical models based on historical data to perform simple weighted fusion of multiple sensor indicators to obtain a comprehensive safety score.

[0003] However, the data stream from electrochemical sensors is relatively simple, resulting in a narrow data dimension and difficulty in capturing changes in the mechanical and physical state of the battery casing, such as deformation and loosening of connectors. This leads to incomplete risk perception coverage, and the response based on threshold triggering is lagging, failing to achieve early warning. Secondly, in terms of evaluation models, existing weight allocation methods either rely excessively on expert subjective experience, making them susceptible to personal judgment bias, or rely entirely on statistical regularities, making them susceptible to interference from extreme data. They lack a collaborative mechanism that can dynamically integrate subjective and objective information and adapt to the actual operating state of the system, which can easily lead to insufficient stability and adaptability of risk assessment results, making it difficult to meet the high reliability requirements of large-scale energy storage systems.

[0004] In view of this, a multi-dimensional safety performance evaluation method and system for sodium-ion battery energy storage systems is proposed. Summary of the Invention

[0005] This invention provides a multi-dimensional safety performance evaluation method and system for sodium-ion battery energy storage systems, which addresses the problem of difficulty in meeting the high reliability requirements of large-scale energy storage systems.

[0006] The first aspect of this invention provides a method for multi-dimensional safety performance evaluation of a sodium-ion battery energy storage system, comprising: Acquire physical image data and electrochemical sensor data of sodium-ion battery energy storage system; The physical image data is processed using a target detection model to calculate detailed physical safety risk indicators; the electrochemical sensor data is processed using a time series analysis model and a risk prediction model to calculate detailed electrochemical safety risk indicators. A dynamic collaborative weighting mechanism combining subjective and objective factors is adopted, which combines the analytic hierarchy process (AHP) and grey relational analysis, to assign weights to the physical safety risk refinement indicators and the electrochemical safety risk refinement indicators, respectively. Using the assigned weights, the physical safety risk refinement indicators are fused at the first layer to obtain the physical safety risk index, and the electrochemical safety risk refinement indicators are fused at the first layer to obtain the electrochemical safety risk index. The physical safety risk index and the electrochemical safety risk index are then weighted and fused in a second layer to obtain the system safety performance index. Based on the physical safety risk index, the electrochemical safety risk index, and the system safety performance index, safety control commands for the sodium-ion battery energy storage system are generated.

[0007] Furthermore, the physical image data is processed using a target detection model to calculate detailed physical security risk indicators, including: The sodium-ion battery energy storage system uses deployed industrial cameras to collect image data of the battery cell casing, wire solder joints, and cooling fan mechanical components. The image data is input into the target detection model to obtain the bounding box detection results of the battery casing, wire solder joints and cooling fan; Based on the bounding box detection results, the following calculations are performed: By comparing the width of the battery casing bounding box at the current moment with that at a historical moment, and combining the statistical characteristics of the historical width, the deformation detection index of the battery casing is calculated. By analyzing the spatial overlap and angular changes between the bounding boxes of the wire and the solder joint in images from different perspectives, the detection index for wire solder joint detachment is calculated. The coordinates of feature points of the cooling fan component are obtained by using a key point detection model, and the normalized offset between the feature point coordinates and the reference coordinates is calculated to obtain the fault monitoring index of the mechanical components of the cooling system.

[0008] Furthermore, the battery casing deformation detection index The calculation formula is: in: The width of the battery casing detected at the current moment. This is the arithmetic mean of the historical battery casing width. The standard deviation of the historical battery casing width; The detection index for wire solder joint detachment The calculation formula is: in: To effectively detect the total number of viewpoints, In the first The overlap status determination value between the solder joint and the wire boundary box under each detection perspective; when the overlap area is less than a preset threshold ratio. It is 0 if it is not 1 otherwise; The mechanical component failure monitoring indicators of the cooling system The calculation formula is: in: The total number of feature points defined for the keypoint detection model. For the current image, the first The coordinate vector of each feature point This represents the coordinate vector of the corresponding feature point under the baseline normal state. These are the reference dimensions for the components used in normalization.

[0009] Furthermore, the electrochemical sensor data is processed using a time series analysis model and a risk prediction model to calculate detailed electrochemical safety risk indicators, including: Collect battery temperature, voltage, and current data, and determine the temperature rise rate, voltage fluctuation sequence, and current anomaly sequence; The temperature rise rate, voltage fluctuation sequence, and current anomaly sequence are used as input features and fed into a temporal convolutional network model. The temporal convolutional network model extracts features from the input sequence through a dilated causal convolutional layer. The dilated causal convolution operation is defined as follows: in: For the current time step, The kernel size is [size]. It is the expansion factor. The output feature value calculated at the current time step. Given the input sequence, For convolution kernel function, For the input sequence in time The value at; The features extracted by the temporal convolutional network model are mapped to... The range is used to output the probability of abnormal temperature rise.

[0010] Furthermore, the process of using time series analysis models and risk prediction models to process the electrochemical sensor data and calculate detailed electrochemical safety risk indicators also includes: Construct a feature vector for predicting thermal runaway risk, the feature vector including battery temperature, temperature rise rate and voltage change characteristics; The feature vector is input into the Cox proportional hazards model, and the risk function of the Cox proportional hazards model is defined as: in: To in a given feature vector At any moment Instantaneous risk rate As the benchmark risk function, , , These are the regression coefficients corresponding to each feature obtained through model training; Calculate the current moment based on the instantaneous risk rate. To the Future The conditional probability of thermal runaway occurring within a given time period.

[0011] Furthermore, the refined indicators of electrochemical safety risks include temperature anomaly indicators. Probability of abnormal temperature rise and the probability of thermal runaway The calculation formulas are as follows: in: The battery surface temperature is collected in real time. This is the lower limit for the low temperature alarm. This is the upper limit for the over-temperature alarm. It is the Sigmoid activation function. The weight parameter matrix, Feature vectors extracted from a temporal convolutional network model. This is a bias term.

[0012] Furthermore, the subjective and objective dynamic collaborative weighting mechanism, which combines the analytic hierarchy process (AHP) and grey relational analysis, assigns weights to the detailed physical safety risk indicators and the detailed electrochemical safety risk indicators, respectively, including: A hierarchical analysis system for evaluating system safety performance is constructed. The target layer of the hierarchical analysis system is to optimize the safety performance of the sodium-ion battery energy storage system. The criterion layer includes physical safety risk criteria and electrochemical safety risk criteria. The scheme layer consists of the battery casing deformation detection index, the wire solder joint detachment detection index, the cooling system mechanical component failure monitoring index, the temperature anomaly index, the temperature rise anomaly probability, and the thermal runaway probability. Based on the aforementioned hierarchical analysis system, domain experts conduct pairwise importance comparisons of the elements in the criterion layer and the elements in the scheme layer to construct judgment matrices respectively; consistency checks are performed on each judgment matrix, and the subjective weights of each refined indicator are calculated after passing the consistency check. Ideal values ​​are set for the battery casing deformation detection index, the wire solder joint detachment detection index, the cooling system mechanical component fault monitoring index, the temperature anomaly index, the temperature rise anomaly probability, and the thermal runaway probability under an absolutely safe state, and an ideal reference sequence is constructed based on the ideal values; Based on the real-time operating data of the sodium-ion battery energy storage system, the grey correlation degree between the real-time data sequence of each of the refined indicators and the ideal reference sequence is calculated.

[0013] Furthermore, the aforementioned subjective and objective dynamic collaborative weighting mechanism also includes the step of synthesizing combined weights: The subjective weights of each of the refined indicators are calculated and then fused with the grey relational degree. Specifically, the grey relational degree is multiplied by the subjective weights, and the product of all indicators is normalized to obtain the final combined weight of each indicator.

[0014] Furthermore, the step of generating safety control commands for the sodium-ion battery energy storage system based on the physical safety risk index, the electrochemical safety risk index, and the system safety performance index includes: Set early warning thresholds and high-risk thresholds for the physical safety risk index, the electrochemical safety risk index, and the system safety performance index, respectively; When the electrochemical safety risk index exceeds its warning threshold, the generated instruction is used to perform power reduction operation or charging current limiting on the corresponding battery module. When the electrochemical safety risk index exceeds its high-risk threshold, the generated instruction is used to initiate enhanced cooling for the corresponding battery module. When the physical security risk index exceeds its warning threshold, the generated instruction is used to trigger a platform alarm and generate a maintenance work order; When the system security performance index exceeds its warning threshold, the generated instruction is used to trigger a system-level status warning prompt. When the system safety performance index exceeds its high-risk threshold, the generated instructions are used to cut off the faulty circuit and activate the fire alarm.

[0015] A second aspect of this invention provides a multi-dimensional safety performance evaluation system for sodium-ion battery energy storage systems, comprising: The data acquisition unit is used to acquire physical image data and electrochemical sensor data of the sodium-ion battery energy storage system. The physical safety risk refinement index and electrochemical safety risk refinement index determination unit are used to process the physical image data using a target detection model to calculate the physical safety risk refinement index; and to process the electrochemical sensor data using a time series analysis model and a risk prediction model to calculate the electrochemical safety risk refinement index. The weight allocation unit is used to assign weights to the physical safety risk refinement index and the electrochemical safety risk refinement index respectively by adopting a subjective and objective dynamic collaborative weighting mechanism that combines the analytic hierarchy process and the grey relational analysis method. The physical safety risk index and electrochemical safety risk index determination unit is used to perform a first-level fusion of the physical safety risk refinement indicators using the allocated weights to obtain the physical safety risk index, and to perform a first-level fusion of the electrochemical safety risk refinement indicators to obtain the electrochemical safety risk index. The system safety performance index determination unit is used to perform a second-level weighted fusion of the physical safety risk index and the electrochemical safety risk index to obtain the system safety performance index. The safety control command generation unit is used to generate safety control commands for the sodium-ion battery energy storage system based on the physical safety risk index, the electrochemical safety risk index, and the system safety performance index.

[0016] As can be seen from the above technical solutions, the present invention has the following advantages: This invention acquires multimodal data from physical images and electrochemical sensors, and calculates detailed physical and electrochemical safety risk indicators using target detection models, time series models, and risk prediction models, respectively. It employs a dynamic collaborative weighting mechanism combining analytic hierarchy process (AHP) and grey relational analysis to assign weights to each indicator. Then, through a first-level fusion, the detailed indicators are synthesized into a physical safety risk index and an electrochemical safety risk index. A second-level weighted fusion yields a system safety performance index, and safety control commands are generated based on these indices. This invention effectively improves the breadth of risk identification and early warning capabilities by fusing image and sensor data. The dynamic collaborative weighting mechanism dynamically adjusts weights based on real-time system operation data, ensuring that the evaluation results are both theoretically sound and adaptable to operating conditions, effectively overcoming the instability of evaluation results caused by subjective judgment bias and extreme data interference. Through two-level fusion, it achieves multi-level, refined evaluation from precise local anomaly localization to macro-level understanding of system-level risks, providing a precise data foundation for differentiated and graded safety control strategies. Attached Figure Description

[0017] Figure 1 This is a schematic flowchart of an embodiment of a method for evaluating the multi-dimensional safety performance of a sodium-ion battery energy storage system according to the present invention. Detailed Implementation

[0018] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “corresponding to,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0019] Example 1 The implementation method in this embodiment can be implemented in a system, on a server, or on a terminal; no specific limitation is made. The method in this application will be described below from the perspective of system implementation. Please refer to... Figure 1 The method provided in this application includes the following steps: S1. Acquire physical image data and electrochemical sensor data of the sodium-ion battery energy storage system; Physical image data is acquired by deploying several industrial cameras at key locations within the energy storage system. These cameras are positioned to target the surface of the battery cell casing, the solder joints between battery modules, and mechanical components of the cooling system, such as cooling fans. The industrial cameras periodically capture images at preset time intervals or capture images when a specific trigger signal is received. To comprehensively capture the morphological features of the target object, the system acquires images from multiple perspectives, including frontal, side, and top-down views, thus forming a multi-view image dataset.

[0020] Electrochemical sensor data is acquired through a sensor array mounted on the surface of individual battery cells or inside the module. This array includes a temperature sensor for real-time monitoring of the battery surface temperature, a voltage sensor for acquiring battery terminal voltage data, and a current sensor for measuring the charging and discharging current in the circuit. These sensors continuously and synchronously acquire the electrochemical operating parameters of the system at a specific frequency, such as once or ten times per second, forming continuous time-series data.

[0021] S2. The physical image data is processed using a target detection model to calculate detailed physical safety risk indicators; the electrochemical sensor data is processed using a time series analysis model and a risk prediction model to calculate detailed electrochemical safety risk indicators. S21. Process the physical image data using a target detection model to calculate detailed physical security risk indicators; the specific implementation process of this step is as follows: S211. By deploying industrial cameras, image data of the battery cell casing, wire solder joints, and cooling fan mechanical components in the sodium-ion battery energy storage system are acquired; S212. Input the image data into the target detection model to obtain the bounding box detection results of the battery casing, wire solder joints and cooling fan; S213. Based on the bounding box detection results, perform the following calculations: 1. By comparing the width of the battery casing boundary box at the current moment with that at a historical moment, and combining the statistical characteristics of the historical width, the deformation detection index of the battery casing is calculated; Battery casing deformation detection indicators are used to detect early signs of bulging or denting in the battery casing caused by internal gas generation, external pressure, or long-term aging. This deformation is a crucial precursor to serious accidents such as thermal runaway. The principle is to quantify the degree of deformation abnormality by comparing the current and historical widths of the battery casing boundary frame and combining the statistical characteristics of the historical widths. Battery casing deformation detection indicators The calculation formula is: in: The width of the battery casing detected at the current moment. This is the arithmetic mean of the historical battery casing width. This represents the standard deviation of the historical battery casing width.

[0022] 2. By analyzing the spatial overlap and angular changes between the wire and solder joint bounding boxes in images from different perspectives, the detection index for wire solder joint detachment is calculated; The conductor solder joint detachment detection index identifies the risk of loosening or disconnection of electrical connections due to vibration, fatigue, or manufacturing defects. Such risks can lead to localized overheating, arcing, or even fire. It works by analyzing the spatial overlap and angular changes between the conductor and solder joint bounding boxes in images from different viewpoints to comprehensively assess the reliability of the connection. The calculation formula is: in: To effectively detect the total number of viewpoints, In the first The overlap status determination value between the solder joint and the wire boundary box under each detection perspective; when the overlap area is less than a preset threshold ratio. It is 0 if it is not 0, otherwise it is 1.

[0023] 3. Obtain the feature point coordinates of the cooling fan component through the key point detection model, and calculate the normalized offset between the feature point coordinates and the reference coordinates to obtain the fault monitoring index of the mechanical components of the cooling system.

[0024] Fault monitoring indicators for mechanical components in cooling systems are used to monitor abnormal operating conditions of critical heat dissipation components such as cooling fans, including jamming, misalignment, or deformation, to ensure heat dissipation efficiency. The principle is to obtain the coordinates of feature points on the cooling fan component through a key point detection model, calculate the normalized offset between the feature point coordinates and the reference coordinates, and quantify the degree of fault in the mechanical component. Cooling System Mechanical Component Fault Monitoring Indicators The calculation formula is: in: The total number of feature points defined for the keypoint detection model. For the current image, the first The coordinate vector of each feature point This represents the coordinate vector of the corresponding feature point under the baseline normal state. These are the reference dimensions for the components used in normalization.

[0025] S22. Using time series analysis models and risk prediction models, electrochemical sensor data are processed to calculate detailed indicators of electrochemical safety risks; This step transforms raw sensor data into quantitative risk indicators with clear physical meaning by constructing a multi-layered electrochemical safety assessment system. This system comprises three key technical components: basic state monitoring, dynamic process analysis, and future risk prediction.

[0026] Specifically, the temperature rise rate is obtained by calculating the rate of temperature change between adjacent sampling points, reflecting the rate of heat accumulation in the battery. The voltage fluctuation sequence is calculated using the standard deviation within a sliding window, characterizing the stability of the electrical system. The current anomaly sequence is calculated using the deviation between real-time measurements and rated values, used to identify overload or internal short-circuit risks. These three features together form the input basis for electrochemical safety analysis. In the dynamic process analysis stage, a temporal convolutional network model is used to calculate the probability of temperature rise anomalies. This model is specifically designed for long-sequence dependency problems in energy storage system monitoring, and its dilated causal convolutional structure achieves efficient feature extraction through the following mechanism: 1. Collect battery temperature, voltage, and current data, and determine the temperature rise rate, voltage fluctuation sequence, and current anomaly sequence; 2. The temperature rise rate, voltage fluctuation sequence, and current anomaly sequence are used as input features and fed into a temporal convolutional network model. The temporal convolutional network model extracts features from the input sequence through dilated causal convolutional layers. The dilated causal convolution operation is defined as follows: in: For the current time step, The kernel size is [size]. It is the expansion factor. The output feature value calculated at the current time step. Given the input sequence, For convolution kernel function, For the input sequence in time The value at; 3. Map the features extracted by the temporal convolutional network model to... The range is used to output the probability of abnormal temperature rise.

[0027] In the future risk prediction phase, a thermal runaway probability assessment model is constructed based on survival analysis theory. This model quantifies risk through the following steps: 1. Construct a feature vector for predicting thermal runaway risk, which includes battery temperature, temperature rise rate, and voltage change characteristics; 2. Input the feature vector into the Cox proportional hazards model. The risk function of the Cox proportional hazards model is defined as: in: To in a given feature vector At any moment Instantaneous risk rate As the benchmark risk function, , , These are the regression coefficients corresponding to each feature obtained through model training; 3. Calculate the current moment based on the instantaneous risk rate. To the Future The conditional probability of thermal runaway occurring within a given time period.

[0028] The final output of detailed electrochemical safety risk indicators includes three core parameters: Detailed indicators for electrochemical safety risks include temperature anomaly indicators. Probability of abnormal temperature rise and the probability of thermal runaway The calculation formulas are as follows: in: The battery surface temperature is collected in real time. This is the lower limit for the low temperature alarm. This is the upper limit for the over-temperature alarm. It is the Sigmoid activation function. The weight parameter matrix, Feature vectors extracted from a temporal convolutional network model. This is a bias term.

[0029] S3. A dynamic collaborative weighting mechanism combining subjective and objective factors is adopted, which combines the analytic hierarchy process and the grey relational analysis method, to assign weights to the detailed indicators of physical safety risks and the detailed indicators of electrochemical safety risks respectively. By establishing a weight allocation system that integrates subjective and objective information, the problems of singular and static weight determination are solved. This method retains the prior knowledge of expert experience while incorporating the dynamic characteristics of real-time operational data, making the weight allocation both theoretically reasonable and adaptable to operating conditions. Specifically: S31. Construct a hierarchical analysis system for evaluating system safety performance; the target layer of the hierarchical analysis system is to optimize the safety performance of the sodium-ion battery energy storage system, the criterion layer includes physical safety risk criteria and electrochemical safety risk criteria, and the scheme layer consists of battery shell deformation detection index, wire solder joint detachment detection index, cooling system mechanical component fault monitoring index, temperature anomaly index, temperature rise anomaly probability and thermal runaway probability. S32. Based on the hierarchical analysis system, domain experts compare the importance of elements in the criterion layer and the scheme layer pairwise to construct judgment matrices; consistency checks are performed on each judgment matrix, and the subjective weights of each detailed indicator are calculated after passing the consistency check. The subjective weights of each detailed indicator are calculated and then integrated with the grey relational degree. The grey relational degree is multiplied by the subjective weights, and the product of all indicators is normalized to obtain the final combined weights of each indicator.

[0030] S33. Set the ideal values ​​of battery casing deformation detection index, wire solder joint detachment detection index, cooling system mechanical component fault monitoring index, temperature anomaly index, temperature rise anomaly probability and thermal runaway probability under an absolutely safe state, and construct an ideal reference sequence based on the ideal values; S34. Based on the real-time operating data of the sodium-ion battery energy storage system, calculate the grey correlation degree between the real-time data sequence of each refined index and the ideal reference sequence.

[0031] In the specific implementation of subjective weight calculation, experts in the field of energy storage safety were invited to conduct pairwise importance comparisons of elements at the criterion level and the scheme level using the 1-9 scaling method. At the criterion level, experts needed to determine the relative importance of physical safety risks and electrochemical safety risks; at the scheme level, they needed to compare the importance of three indicators under physical safety risks and three indicators under electrochemical safety risks. A judgment matrix was constructed based on the comparison results, and the initial weights of each indicator were calculated using the eigenvector method. To ensure the logical consistency of the expert judgments, a consistency test was performed, and the consistency ratio (CR) was calculated. When the CR was less than 0.10, the judgment matrix was considered to have satisfactory consistency. For the weight results that passed the test, the final subjective weight vector was formed.

[0032] When constructing the ideal reference sequence, ideal baseline values ​​for each indicator under an absolutely safe state are set for the safety characteristics of the sodium-ion battery energy storage system. The battery casing deformation detection index, wire solder joint detachment detection index, and cooling system mechanical component fault monitoring index are all zero under the ideal state, indicating no deformation, detachment, or fault. The temperature anomaly index is zero under the ideal state, representing that the temperature is within the optimal operating range. The probability of abnormal temperature rise and the probability of thermal runaway are also zero under the ideal state, indicating no risk of abnormal temperature rise or thermal runaway. The ideal reference sequence X0 = [0,0,0,0,0,0] constructed based on these ideal values ​​provides a benchmark for subsequent objective weight adjustments.

[0033] Based on real-time operational data of the energy storage system, actual observation values ​​of various indicators at the current moment are collected to construct a sample sequence. The minimum and maximum differences at two levels are determined by calculating the absolute difference between the sample sequence and the ideal reference sequence. Appropriate resolution coefficients are selected, and the correlation coefficients of each indicator at the current moment are calculated according to the correlation coefficient formula. For multiple samples, the arithmetic mean of the correlation coefficients of each indicator is taken to obtain its grey relational degree. The correlation degree ranges from 0 to 1; the closer the value is to 1, the closer the current state of the indicator is to the ideal safe state; the closer the value is to 0, the greater the deviation from the ideal state.

[0034] After completing the subjective weight calculation and grey relational analysis, the final combined weights are synthesized using specific fusion rules. The subjective weight of each indicator is multiplied by its grey relational degree, and the product of all indicators is normalized. Indicators with high grey relational degrees indicate that their current state is close to the ideal safety state, and they should be given higher weights in the final evaluation to reflect their reliability. Indicators with low grey relational degrees indicate that their current state deviates from the safety benchmark, and their weights should be reduced in the final evaluation to suppress the influence of unreliable data.

[0035] This embodiment uses six key safety indicators of a battery system as examples to illustrate in detail the complete process of subjective weight calculation and objective weight correction, and synthesizes the final combined weights used for safety evaluation. These six indicators are: battery casing deformation detection indicator, wire solder joint detachment detection indicator, cooling system mechanical component fault monitoring indicator, temperature anomaly probability, temperature rise anomaly probability, and thermal runaway probability.

[0036] I. Construct the judgment matrix; Several experts in the field of energy storage system safety were invited to conduct pairwise importance comparisons of the above n=6 indicators using the "1-9 scale method" (see Table 1 for the judgment matrix scale and its meaning). The experts concluded that the early warning importance of electrochemical risk indicators is higher than that of physical and mechanical risk indicators, and the importance of the six indicators from highest to lowest is as follows: thermal runaway probability > abnormal temperature rise probability > abnormal temperature index > cooling system mechanical failure > wire solder joint detachment > battery casing deformation. The resulting judgment matrix is ​​as follows: Table 1 Scheme Layer Judgment Matrix II. Calculate the weight vector; The weight vector of the above judgment matrix is ​​calculated using the eigenvector method.

[0037] 1. Calculate the geometric mean of the elements in each row of the judgment matrix. : The geometric mean is normalized to obtain the weighting coefficients. : Therefore, the subjective weight vector of the criterion layer is: Specifically, the risk weights are as follows: battery casing deformation 0.043, wire solder joint detachment 0.062, cooling system mechanical failure 0.082, abnormal temperature 0.176, abnormal temperature rise 0.261, and thermal runaway 0.376.

[0038] III. Consistency Check; To ensure the logical consistency of expert judgments, a consistency check is required.

[0039] Calculate the largest eigenvalue : Calculate the consistency index : Query the average random consistency index (6th order matrix RI=1.24), and calculate the consistency ratio. : because =0.056 < 0.1, the judgment matrix passes the consistency test, indicating that the expert judgments have good consistency, and the calculated subjective weights are... The allocation is reasonable.

[0040] IV. Objective weighting adjustment; Taking monitoring data at a certain moment as an example, the above six indicators are objectively weighted and adjusted: 1. Constructing an Ideal Reference Sequence and Real-Time Data: Ideal values ​​for each indicator under absolutely safe conditions are set, and an ideal reference sequence is constructed. In this embodiment, the ideal values ​​for the battery casing deformation detection indicator, wire solder joint detachment detection indicator, cooling system mechanical component fault monitoring indicator, temperature anomaly probability, temperature rise anomaly probability, and thermal runaway probability are all set to 0. Therefore, the ideal reference sequence is: 2. Collect real-time data for each indicator at the current moment to construct sample data: This sample indicates minor deformation and mechanical failure, with a high probability of temperature anomalies and temperature rise.

[0041] 3. Calculate the correlation coefficient: Calculate the absolute difference between each indicator and the ideal value for the sample. = Find the minimum and maximum differences between the two levels: , ; Take the resolution coefficient =0.5, calculate the correlation coefficient: The correlation coefficients of the six indicators in this sample are as follows: , , , , , .

[0042] 4. Generating Relationship Degrees: If there are multiple samples, the average of the correlation coefficients for each indicator needs to be calculated to obtain the grey relational degree for the corresponding indicator. This example uses only one sample, so the grey relational degree is obtained directly: , , , , .

[0043] Due to correlation The value range is [0, 1]. The closer the value is to 1, the closer the current state of the indicator is to the ideal safe state, and the lower the risk; the closer the value is to 0, the further it deviates from the ideal state, and the higher the risk. Therefore, the current battery casing deformation, wire solder joint detachment, cooling system mechanical component failure, and thermal runaway indicators have a very high correlation with their ideal values, indicating that the system's physical and mechanical state and core protection functions are at an excellent level. However, the temperature anomaly and temperature rise anomaly indicators have a low correlation with their ideal values. If the sample data is normal, it means that the temperature sensor itself may have experienced widespread failure, drift, or interference, leading to generally inaccurate temperature data. Therefore, the effectiveness of temperature as a highly discriminative risk indicator should be reduced.

[0044] The comparison results show that, in the current system, apart from potential risks in temperature management, the other indicators are quite stable. This result will be directly used for subsequent weight adjustments, thereby dynamically adjusting the importance ratio of these indicators in the final safety evaluation.

[0045] S4. Using the assigned weights, the detailed physical safety risk indicators are fused at the first layer to obtain the physical safety risk index, and the detailed electrochemical safety risk indicators are fused at the first layer to obtain the electrochemical safety risk index. S5. The physical safety risk index and the electrochemical safety risk index are weighted and fused in a second layer to obtain the system safety performance index; In the first layer of fusion, based on the combined weights obtained through a dynamic collaborative weighting mechanism of subjective and objective factors, the detailed indicators within the two dimensions of physical safety and electrochemical safety are linearly weighted and fused. The physical safety risk index is calculated by weighting and summing three indicators: battery casing deformation, wire solder joint detachment, and mechanical component failure in the cooling system. The electrochemical safety risk index is calculated by weighting and summing three indicators: temperature anomaly, probability of temperature rise anomaly, and probability of thermal runaway. This forms a specific risk index. In the second layer of fusion, based on the criterion layer weights determined by the analytic hierarchy process (AHP), the two specific risk indices are fused at the system level, and the system safety performance index is obtained through weighted summation. This two-layer fusion ensures the precise quantification of indicators within each risk dimension.

[0046] The following examples illustrate the implementation of these two steps, assuming that the final combined weights of the six indicators have been obtained through the above combined weighting method: Calculating the safety risk index: Due to the different dimensions of the various indicators, standardization is required to unify them to the range of [0,1]. Deviation standardization is employed. in, As an indicator j The original value, and Indicators j The minimum and maximum allowed values.

[0047] Calculating the specific safety risk index: The physical safety risk index integrates all physical monitoring indicators. ; The closer the value is to 1, the higher the physical safety risk. The electrochemical safety risk index integrates all electrochemical monitoring indicators: ; The closer the value is to 1, the higher the electrochemical safety risk.

[0048] Computer system security performance refers to: For a comprehensive evaluation of the overall system security, by and The weighted sum is obtained. The weights are calculated similarly to those for the criterion layer weights in Example 1, assuming a physical risk weight of 0.25 for B1 and an electrochemical risk weight of 0.75 for B2. The closer the value is to 1, the higher the overall security risk of the system.

[0049] S6. Based on the physical safety risk index, electrochemical safety risk index, and system safety performance index, generate safety control commands for the sodium-ion battery energy storage system.

[0050] Set early warning thresholds and high-risk thresholds for the physical safety risk index, electrochemical safety risk index, and system safety performance index, respectively; 1. When the electrochemical safety risk index exceeds its warning threshold, the generated instructions are used to perform power reduction operation or charging current limiting on the corresponding battery module; 2. When the electrochemical safety risk index exceeds its high-risk threshold, the generated command is used to initiate enhanced cooling for the corresponding battery module; 3. When the physical safety risk index exceeds its warning threshold, the generated instruction is used to trigger a platform alarm and generate a maintenance work order; 4. When the system security performance index exceeds its warning threshold, the generated command is used to trigger a system-level status warning. 5. When the system safety performance index exceeds its high-risk threshold, the generated instructions are used to cut off the faulty circuit and activate the fire alarm.

[0051] In this embodiment, a risk response system is constructed by setting the warning threshold for the electrochemical safety risk index to 0.5 and the high-risk threshold to 0.7, the warning threshold for the physical safety risk index to 0.7, and the warning threshold for the system safety performance index to 0.5 and the high-risk threshold to 0.8. When the electrochemical risk index exceeds 0.5, the system sends a power reduction or charging current limiting command to the battery management system to suppress the heat of electrochemical reaction by reducing the charging and discharging current. When the index exceeds 0.7, a cooling enhancement command is simultaneously sent to the thermal management system to increase the coolant flow rate or fan speed for rapid heat dissipation. When the physical risk index exceeds 0.7, the system triggers an audible and visual alarm on the monitoring platform and automatically generates a maintenance work order, prompting maintenance personnel to conduct on-site inspections of the mechanical structure. When the system safety performance index exceeds 0.5, a system-level status warning is activated, providing a global reminder through interface color changes and SMS notifications. When the index exceeds 0.8, a command to cut off the fault circuit is sent to the battery management system, and a pre-alarm for the fire protection system is activated, achieving full-process safety control from risk warning to emergency intervention.

[0052] Example 2 An embodiment of the multi-dimensional safety performance evaluation system for sodium-ion battery energy storage systems in this invention includes the following: The data acquisition unit is used to acquire physical image data and electrochemical sensor data of the sodium-ion battery energy storage system. The physical safety risk refinement index and electrochemical safety risk refinement index determination unit are used to process physical image data using a target detection model to calculate physical safety risk refinement index; and to process electrochemical sensor data using a time series analysis model and a risk prediction model to calculate electrochemical safety risk refinement index. The weight allocation unit is used to assign weights to the physical safety risk refinement indicators and the electrochemical safety risk refinement indicators respectively by adopting a dynamic collaborative weighting mechanism that combines the analytic hierarchy process and the grey relational analysis method. The physical safety risk index and electrochemical safety risk index determination unit is used to perform a first-level fusion of the physical safety risk refinement indicators to obtain the physical safety risk index, and to perform a first-level fusion of the electrochemical safety risk refinement indicators to obtain the electrochemical safety risk index. The system safety performance index determination unit is used to perform a second-level weighted fusion of the physical safety risk index and the electrochemical safety risk index to obtain the system safety performance index. The safety control command generation unit is used to generate safety control commands for the sodium-ion battery energy storage system based on the physical safety risk index, electrochemical safety risk index, and system safety performance index.

[0053] For specific limitations regarding the system, please refer to the method limitations described above, which will not be repeated here. Each module in the above system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0054] It is understood that those skilled in the art can combine various implementation methods in the above embodiments under the guidance of the above examples to obtain technical solutions with multiple implementation methods.

[0055] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for multi-dimensional safety performance evaluation of a sodium-ion battery energy storage system, characterized in that, include: Acquire physical image data and electrochemical sensor data of sodium-ion battery energy storage system; The physical image data is processed using a target detection model to calculate detailed physical safety risk indicators; the electrochemical sensor data is processed using a time series analysis model and a risk prediction model to calculate detailed electrochemical safety risk indicators. A dynamic collaborative weighting mechanism combining subjective and objective factors is adopted, which combines the analytic hierarchy process (AHP) and grey relational analysis, to assign weights to the physical safety risk refinement indicators and the electrochemical safety risk refinement indicators, respectively. Using the assigned weights, the physical safety risk refinement indicators are fused at the first layer to obtain the physical safety risk index, and the electrochemical safety risk refinement indicators are fused at the first layer to obtain the electrochemical safety risk index. The physical safety risk index and the electrochemical safety risk index are then weighted and fused in a second layer to obtain the system safety performance index. Based on the physical safety risk index, the electrochemical safety risk index, and the system safety performance index, safety control commands for the sodium-ion battery energy storage system are generated.

2. The multi-dimensional safety performance evaluation method for sodium-ion battery energy storage systems according to claim 1, characterized in that, The physical image data is processed using a target detection model to calculate detailed physical security risk indicators, including: The sodium-ion battery energy storage system uses deployed industrial cameras to collect image data of the battery cell casing, wire solder joints, and cooling fan mechanical components. The image data is input into the target detection model to obtain the bounding box detection results of the battery casing, wire solder joints and cooling fan; Based on the bounding box detection results, the following calculations are performed: By comparing the width of the battery casing bounding box at the current moment with that at a historical moment, and combining the statistical characteristics of the historical width, the deformation detection index of the battery casing is calculated. By analyzing the spatial overlap and angular changes between the bounding boxes of the wire and the solder joint in images from different perspectives, the detection index for wire solder joint detachment is calculated. The coordinates of feature points of the cooling fan component are obtained by using a key point detection model, and the normalized offset between the feature point coordinates and the reference coordinates is calculated to obtain the fault monitoring index of the mechanical components of the cooling system.

3. The method for multi-dimensional safety performance evaluation of sodium-ion battery energy storage systems according to claim 2, characterized in that: The battery casing deformation detection index The calculation formula is: in: The width of the battery casing detected at the current moment. This is the arithmetic mean of the historical battery casing width. The standard deviation of the historical battery casing width; The detection index for wire solder joint detachment The calculation formula is: in: To effectively detect the total number of viewpoints, In the first The overlap status determination value between the solder joint and the wire boundary box under each detection perspective; when the overlap area is less than a preset threshold ratio. It is 0 if it is not 1 otherwise; The mechanical component failure monitoring indicators of the cooling system The calculation formula is: in: The total number of feature points defined for the keypoint detection model. For the current image, the first The coordinate vector of each feature point This represents the coordinate vector of the corresponding feature point under the baseline normal state. These are the reference dimensions for the components used in normalization.

4. The multi-dimensional safety performance evaluation method for sodium-ion battery energy storage systems according to claim 1, characterized in that, The electrochemical sensor data is processed using a time series analysis model and a risk prediction model to calculate detailed electrochemical safety risk indicators, including: Collect battery temperature, voltage, and current data, and determine the temperature rise rate, voltage fluctuation sequence, and current anomaly sequence; The temperature rise rate, voltage fluctuation sequence, and current anomaly sequence are used as input features and fed into a temporal convolutional network model. The temporal convolutional network model extracts features from the input sequence through a dilated causal convolutional layer. The dilated causal convolution operation is defined as follows: in: For the current time step, The kernel size is [size]. It is the expansion factor. The output feature value calculated at the current time step. Given the input sequence, For convolution kernel function, For the input sequence in time The value at; The features extracted by the temporal convolutional network model are mapped to... The range is used to output the probability of abnormal temperature rise.

5. The multi-dimensional safety performance evaluation method for sodium-ion battery energy storage systems according to claim 4, characterized in that, The process of processing the electrochemical sensor data using time series analysis and risk prediction models to calculate detailed electrochemical safety risk indicators also includes: Construct a feature vector for predicting thermal runaway risk, the feature vector including battery temperature, temperature rise rate and voltage change characteristics; The feature vector is input into the Cox proportional hazards model, and the risk function of the Cox proportional hazards model is defined as: in: To in a given feature vector At any moment Instantaneous risk rate As the benchmark risk function, , , These are the regression coefficients corresponding to each feature obtained through model training; Calculate the current moment based on the instantaneous risk rate. To the Future The conditional probability of thermal runaway occurring within a given time period.

6. The method for multi-dimensional safety performance evaluation of sodium-ion battery energy storage systems according to claim 5, characterized in that, The detailed indicators for electrochemical safety risks include temperature anomaly indicators. Probability of abnormal temperature rise and the probability of thermal runaway The calculation formulas are as follows: in: The battery surface temperature is collected in real time. This is the lower limit for the low temperature alarm. This is the upper limit for the over-temperature alarm. It is the Sigmoid activation function. The weight parameter matrix, Feature vectors extracted from a temporal convolutional network model. This is a bias term.

7. The multi-dimensional safety performance evaluation method for sodium-ion battery energy storage systems according to claim 1, characterized in that, The aforementioned dynamic collaborative weighting mechanism, combining the analytic hierarchy process (AHP) and grey relational analysis, assigns weights to the detailed physical safety risk indicators and the detailed electrochemical safety risk indicators, including: A hierarchical analysis system for evaluating system safety performance is constructed. The target layer of the hierarchical analysis system is to optimize the safety performance of the sodium-ion battery energy storage system. The criterion layer includes physical safety risk criteria and electrochemical safety risk criteria. The scheme layer consists of the battery casing deformation detection index, the wire solder joint detachment detection index, the cooling system mechanical component failure monitoring index, the temperature anomaly index, the temperature rise anomaly probability, and the thermal runaway probability. Based on the aforementioned hierarchical analysis system, domain experts conduct pairwise importance comparisons of the elements in the criterion layer and the elements in the scheme layer to construct judgment matrices respectively; consistency checks are performed on each judgment matrix, and the subjective weights of each refined indicator are calculated after passing the consistency check. Ideal values ​​are set for the battery casing deformation detection index, the wire solder joint detachment detection index, the cooling system mechanical component fault monitoring index, the temperature anomaly index, the temperature rise anomaly probability, and the thermal runaway probability under an absolutely safe state, and an ideal reference sequence is constructed based on the ideal values; Based on the real-time operating data of the sodium-ion battery energy storage system, the grey correlation degree between the real-time data sequence of each of the refined indicators and the ideal reference sequence is calculated.

8. The method for multi-dimensional safety performance evaluation of sodium-ion battery energy storage systems according to claim 7, characterized in that, The aforementioned subjective and objective dynamic collaborative weighting mechanism also includes the step of synthesizing combined weights: The subjective weights of each of the refined indicators are calculated and then fused with the grey relational degree. Specifically, the grey relational degree is multiplied by the subjective weights, and the product of all indicators is normalized to obtain the final combined weight of each indicator.

9. The method for multi-dimensional safety performance evaluation of sodium-ion battery energy storage systems according to claim 8, characterized in that, The generation of safety control commands for the sodium-ion battery energy storage system based on the physical safety risk index, the electrochemical safety risk index, and the system safety performance index includes: Set early warning thresholds and high-risk thresholds for the physical safety risk index, the electrochemical safety risk index, and the system safety performance index, respectively; When the electrochemical safety risk index exceeds its warning threshold, the generated instruction is used to perform power reduction operation or charging current limiting on the corresponding battery module. When the electrochemical safety risk index exceeds its high-risk threshold, the generated instruction is used to initiate enhanced cooling for the corresponding battery module. When the physical security risk index exceeds its warning threshold, the generated instruction is used to trigger a platform alarm and generate a maintenance work order; When the system security performance index exceeds its warning threshold, the generated instruction is used to trigger a system-level status warning prompt. When the system safety performance index exceeds its high-risk threshold, the generated instructions are used to cut off the faulty circuit and activate the fire alarm.

10. A multi-dimensional safety performance evaluation system for sodium-ion battery energy storage systems, characterized in that, include: The data acquisition unit is used to acquire physical image data and electrochemical sensor data of the sodium-ion battery energy storage system. The physical safety risk refinement index and electrochemical safety risk refinement index determination unit is used to process the physical image data using the target detection model and calculate the physical safety risk refinement index. The electrochemical sensor data was processed using a time series analysis model and a risk prediction model to calculate a detailed index of electrochemical safety risks. The weight allocation unit is used to assign weights to the physical safety risk refinement index and the electrochemical safety risk refinement index respectively by adopting a subjective and objective dynamic collaborative weighting mechanism that combines the analytic hierarchy process and the grey relational analysis method. The physical safety risk index and electrochemical safety risk index determination unit is used to perform a first-level fusion of the physical safety risk refinement indicators using the allocated weights to obtain the physical safety risk index, and to perform a first-level fusion of the electrochemical safety risk refinement indicators to obtain the electrochemical safety risk index. The system safety performance index determination unit is used to perform a second-level weighted fusion of the physical safety risk index and the electrochemical safety risk index to obtain the system safety performance index. The safety control command generation unit is used to generate safety control commands for the sodium-ion battery energy storage system based on the physical safety risk index, the electrochemical safety risk index, and the system safety performance index.

Images