A support vector machine-based energy storage battery non-destructive testing and evaluation method and system

Abstract

The application discloses a kind of based on support vector machine's energy storage battery nondestructive testing and evaluation method and system.The method of the application includes: using ultrasonic nondestructive scanning equipment, battery electrolyte wetting state, gas production, lithium deposition and pole surface are imaged, and the ultrasonic image of energy storage battery is obtained;Extract the core features in ultrasonic image, including battery liquid wetting degree feature, gas aggregation area feature, interface and material continuity feature, lithium metal deposition area feature;Based on the multidimensional features extracted, form five-dimensional evaluation index system considering wettability, gas density, surface quality, internal structure and lithium deposition detection, determine the weight of each index using the technique for order preference by similarity to ideal solution, and apply the weight to the feature vector to obtain the weighted feature vector;The weighted feature vector is input into the optimized SVM model to realize intelligent diagnosis and evaluation of the state of energy storage battery.The application realizes the rapid detection, accurate evaluation and forward-looking early warning of the state of energy storage battery.

Description

Technical Field

[0001] This invention belongs to the field of battery health status assessment technology, specifically a non-destructive testing and evaluation method and system for energy storage batteries based on support vector machines. Background Technology

[0002] Energy storage batteries will face various forms and types of defects at different stages of their life cycle. Among them, lithium plating and gas production are key causes of serious safety accidents such as battery capacity decay, increased internal resistance, and even thermal runaway. However, these defects are often difficult to quantify and accurately locate in the early stages of failure through external macroscopic electrical parameters. Therefore, battery testing technology is urgently needed for rapid screening and monitoring.

[0003] Currently, battery monitoring technologies are categorized into two types based on the destructive nature of the sample: destructive testing and non-destructive testing. Destructive testing primarily refers to disassembly and analysis, which can directly obtain the internal physicochemical information of the battery, but it results in the complete scrapping of the sample, making further testing impossible. Traditional non-destructive testing technologies, such as X-ray imaging, infrared thermography, electrochemical impedance spectroscopy, and optical microscopy, have inherent limitations in terms of detection resolution, speed, cost, and safety. Because traditional detection technologies and evaluation methods lack the ability to quantitatively identify and evaluate microscopic and hidden defects with high precision, fault warnings are delayed, often leaving battery management in a passive "post-mortem analysis" situation, making true predictive maintenance difficult.

[0004] Against this backdrop, ultrasonic scanning-based non-destructive testing technology can balance performance indicators in terms of resolution, speed, cost, and safety, making it particularly suitable for real-time monitoring of battery mechanical performance and interface status. Meanwhile, Support Vector Machine (SVM), a machine learning algorithm used for classification and regression, aims to construct an optimal decision boundary that clearly distinguishes different categories of samples with the maximum margin, thereby enhancing the model's predictive performance on unknown data. This has made SVM an ideal algorithm for evaluating the state of energy storage batteries.

[0005] Therefore, in order to achieve "proactive early warning" and "full life cycle management" of batteries, there is an urgent need for a battery state detection and evaluation scheme that deeply integrates ultrasonic non-destructive testing technology and support vector machine algorithm. Summary of the Invention

[0006] This invention aims to address the problems of low efficiency, inaccurate condition evaluation, and delayed safety warnings in existing battery health status detection methods. It provides a non-destructive testing and evaluation method and system for energy storage batteries that integrates ultrasonic scanning and machine learning algorithms. This system can perform non-destructive and efficient detection of subtle internal physical structural changes throughout the entire lifecycle of a battery, from grid connection to operation, maintenance, and retirement. This enables rapid detection, accurate evaluation, and proactive early warning of the energy storage battery's condition. Furthermore, based on the detection results, this invention can intelligently diagnose and evaluate the energy storage battery's "normal," "slightly damaged," and "severely damaged" states, thereby providing technical assurance for the reliable operation of energy storage battery systems.

[0007] Therefore, the present invention adopts the following technical solution: a non-destructive testing and evaluation method for energy storage batteries based on support vector machines, comprising: Ultrasonic non-destructive scanning equipment was used to image the electrolyte wetting state, gas production, lithium deposition and electrode characterization of the battery, and to obtain reflected / transmitted ultrasonic images of the energy storage battery. Core features were extracted from the ultrasound images, including features of battery fluid wetting, gas accumulation areas, interface and material continuity, and lithium metal deposition areas. Based on the extracted multi-dimensional features, a five-dimensional evaluation index system is formed, which includes wettability, gas density, surface quality, internal structure, and lithium plating detection. The weight of each index is determined by the approximation ideal solution ranking method, and the weights are applied to the feature vector to obtain the weighted feature vector. The weighted feature vector is input into the optimized SVM model to achieve intelligent diagnosis and evaluation of the energy storage battery status.

[0008] Furthermore, the non-destructive testing and evaluation method for energy storage batteries also includes: generating a visual interface for non-destructive testing and evaluation of energy storage batteries, transforming complex analysis results into intuitive information; setting four options in the control panel: "Select Image," "Retest," "Save Results," and "Clear Image"; outputting clear three-level status classification results, namely, intelligent diagnostic evaluation results indicating whether the battery is in a "normal" state, a "slightly damaged" state, or a "severely damaged" state; the intelligent diagnostic evaluation results are linked to the operation and maintenance strategies of the energy storage power station to automatically generate differentiated maintenance plans.

[0009] Furthermore, the differentiated maintenance plan is as follows: based on the intelligent diagnostic evaluation results, routine inspections are carried out on batteries in "normal" condition, planned maintenance is performed on batteries with "minor damage" by cleaning the surface, tightening connections, and calibrating parameters, and emergency shutdown maintenance is carried out on batteries with "severe damage" by isolating the battery pack, replacing core components, and conducting system-level testing.

[0010] This risk-level-based intelligent decision-making avoids over-maintenance or under-maintenance, enables precise allocation and optimized configuration of operation and maintenance resources, and significantly improves operation and maintenance efficiency and economic benefits.

[0011] Furthermore, while acquiring ultrasonic images, the typical characteristics of energy storage batteries in four key stages—grid access, operation, maintenance, and decommissioning—are analyzed. A mapping model between battery status and the entire life cycle stages is established, forming an image status database of energy storage batteries based on ultrasonic scanning and covering the entire life cycle.

[0012] Furthermore, the steps of the sorting method for approximating the ideal solution include: 1) Positive orientation of indicator attributes: (The following is a list of parameters, not part of the main text) i One evaluation object, j One evaluation indicator, which transforms all low-quality and neutral indicators into high-quality indicators; 2) Constructing the initial normalized matrix: Normalize the forward-oriented data to obtain the normalized matrix. Z ; 3) Determine the positive and negative ideal solutions: Based on the normalized matrix Z, the cosine method is used to find the positive and negative ideal solutions; 4) Calculate the distance between each evaluation object and the positive and negative ideal solutions: 5) Calculate the degree of closeness between each evaluation object and the positive ideal solution. C i ; 6) Based on proximity C i The ranking is based on the weights of wettability, gas, surface quality, internal structure, and lithium plating, with the index corresponding to the maximum degree of proximity being designed, and the state evaluation is performed based on this.

[0013] Furthermore, the ideal solution Z + It is composed of the maximum value of each column's elements, and the negative ideal solution Z. - It is composed of the minimum value in each column.

[0014] Furthermore, the SVM model forms a sample matrix and label vector based on the energy storage battery image state database; completes data preprocessing and calls the extracted core features of the ultrasound image; labels the battery state to be evaluated, and labels each sample; selects the radial basis function kernel function as the kernel function, and applies the grid search method to optimize the penalty factor C and kernel function parameters.

[0015] This invention also provides a support vector machine-based non-destructive testing and evaluation system for energy storage batteries, used to implement the above-mentioned non-destructive testing and evaluation method for energy storage batteries, comprising: The image data acquisition layer uses ultrasonic non-destructive scanning equipment to image the electrolyte wetting state, gas production, lithium plating and electrode characterization of the battery, and to acquire reflected / transmitted ultrasonic images of the energy storage battery. The battery status evaluation layer is used to extract core features from ultrasonic images and form a five-dimensional evaluation index system of wettability, gas density, surface quality, internal structure, and lithium plating detection. The weight of each index is determined by the approximation ideal solution ranking method, and the weights are applied to the feature vector to obtain a weighted feature vector. The weighted feature vector is then input into the optimized SVM model to realize intelligent diagnosis and evaluation of the energy storage battery status. The visualization output layer is used to generate a visualization interface for non-destructive testing and evaluation of energy storage batteries, transforming complex analysis results into intuitive information.

[0016] Compared with existing technologies, this invention has the advantages of being fast, accurate, and safe, enabling early warning, precise performance prediction, and intelligent operation and maintenance of energy storage battery systems, significantly improving the safety and economic benefits of energy storage systems. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the non-destructive testing and evaluation method for energy storage batteries according to the present invention; Figure 2 This is a flowchart of the steps for forming the image state database of the energy storage battery according to the present invention; Figure 3 This is a flowchart of the battery state evaluation method based on the SVM algorithm of the present invention; Figure 4 This is a flowchart illustrating the generation process of the differentiated maintenance solution of the present invention. Figure 5 This is a diagram showing the composition of the non-destructive testing and evaluation system for energy storage batteries according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Example 1 This embodiment presents a non-destructive testing and evaluation method for energy storage batteries based on support vector machines. Figure 1 As shown, the steps are as follows: Step 1: Using ultrasonic non-destructive scanning equipment, images are captured of the battery's electrolyte wetting state, gas production, lithium plating, and electrode characterization. Reflected / transmitted ultrasonic images of the energy storage battery are acquired to visually reveal its physical structural defects, such as delamination, bubbles, impacts, and scratches. Simultaneously, the typical characteristics of the energy storage battery during four key stages—grid connection, operation, maintenance, and decommissioning—are analyzed. A mapping model between battery state and all lifecycle stages is established, forming an image state database of energy storage batteries based on ultrasonic scanning and covering the entire lifecycle. Figure 2 As shown.

[0021] Step 2: Extract the core features from the ultrasound images, including features of battery fluid wetting, gas accumulation areas, interface and material continuity, and lithium metal deposition areas.

[0022] Step 3: Based on the extracted multi-dimensional features, a five-dimensional evaluation index system is formed, including wettability, gas density, surface quality, internal structure, and lithium plating detection. The weight of each index is determined by the approximation ideal solution ranking method, and the weights are applied to the feature vector to obtain the weighted feature vector.

[0023] Step 4: Input the weighted feature vector into the optimized SVM model to achieve intelligent diagnostic evaluation of the energy storage battery status. The SVM model forms a sample matrix and label vector based on the energy storage battery image status database. Data preprocessing is completed and the extracted core features of the ultrasound image are called. Data labeling is performed on the battery status to be evaluated, and each sample is labeled. The radial basis function kernel function is selected as the kernel function, and the penalty factor C and kernel function parameters are tuned by applying the grid search method.

[0024] The flowcharts for steps 2 to 4 are as follows: Figure 3 As shown.

[0025] Step 5: Generate a visual interface for non-destructive testing and evaluation of energy storage batteries. The control panel has four options: "Select Image," "Retest," "Save Results," and "Clear Image." It outputs clear three-level status classification results: intelligent diagnostic evaluation results indicating whether the battery is in a "normal," "slightly damaged," or "severely damaged" state. These intelligent diagnostic evaluation results are linked to the operation and maintenance strategies of the energy storage power station, automatically generating differentiated maintenance plans, such as... Figure 4 As shown.

[0026] In step 1, an ultrasonic non-destructive scanning device with millimeter-level resolution, capable of accurately transmitting and receiving ultrasonic pulses, is employed. This device can image the electrolyte wetting state, gas generation, lithium plating, and electrode characterization within the battery, making it an ideal tool for detecting changes in the battery's microstructure. By analyzing the signal changes during the propagation, reflection, or transmission of ultrasonic waves inside the battery, the device can acquire reflected / transmitted ultrasonic images of the battery, thereby visually revealing the interface conditions between different media such as electrode layers, separators, and electrolytes, as well as potential physical defects such as voids, delamination, and bubbles.

[0027] In step 1, the typical characteristics and key tasks of energy storage batteries in the grid connection, operation, maintenance, and decommissioning stages are analyzed. Through operations such as ① charging and discharging in silicone oil at room temperature (0.5C); ② high-temperature aging at 50℃ and 1C; ③ low-temperature aging at -5℃ and 1C; ④ cyclic aging with iron powder at 25℃ and 1C to achieve micro-short circuit by opening the valve; ⑤ cyclic aging with iron powder at -2.5℃ by opening the valve to achieve lithium plating; ⑥ impact, crushing, and squeezing; and ⑦ scratching by friction and abrasion, battery state models for different stages are designed. A battery image state database based on ultrasonic scanning and covering the entire life cycle is constructed to lay a complete image data foundation for subsequent battery state evaluation.

[0028] In step 2, the changes in acoustic parameters when ultrasound propagates in different media are used to extract the core features of the image, and to infer and characterize the internal state of the battery.

[0029] (1) Wetting characteristics: This refers to whether the electrolyte fully wets the porous structure of the electrode plates (positive and negative electrodes), and whether air is present in the insufficiently wetted areas. The speed of sound in air is much lower than that of the electrolyte and electrode materials, and the overall equivalent speed of sound in the unwetted areas will be significantly reduced. The degree of wetting is quantified by measuring the absolute value of the speed of sound in a specific area of ​​the battery or by comparing it with the speed of sound in a known fully wetted area. A speed of sound distribution map is plotted; the lower the speed of sound, the worse the wettability.

[0030] (2) Characteristics of gas accumulation regions: Due to their extremely low acoustic impedance, gases cause near total reflection or severe acoustic attenuation of ultrasound waves at the gas-liquid or gas-solid interface, which manifests as signal loss or acoustic shadowing areas on ultrasound images. By setting specific signal intensity thresholds, these areas can be identified as gas accumulation regions, and their areas can be calculated. The size of the gas-producing region directly reflects the degree of abnormality of the internal chemical reaction of the battery and the potential safety risks.

[0031] (3) Interface and material continuity characteristics: Any defects (such as cracks, delamination, impurities) during the inspection of battery surface quality and internal structure will generate additional echo signals during ultrasonic scanning. By measuring the number of echoes, the number of unwanted echoes appearing between the main interface echoes can indicate the quantity and degree of internal and external defects.

[0032] (4) Characteristics of lithium metal deposition regions: This refers to the uneven deposition of lithium metal on the surface of the negative electrode, which is a major side reaction in fast charging and low-temperature charging. The density and elastic modulus of lithium metal are different from those of graphite negative electrodes, and its deposition will change the equivalent sound velocity of the local material. By setting the relative change in sound velocity, scanning is performed at the same location before and after charging or before and after cycling. Regions with significantly reduced sound velocity can be identified as lithium deposition regions, and the area and severity of lithium deposition regions are characterized by the average change in sound velocity.

[0033] In step 3, in order to comprehensively evaluate the health status of the battery throughout its entire life cycle, this invention designs a five-dimensional evaluation index system that considers wettability, gas density, surface quality, internal structure and lithium plating detection, and uses the Approximation Ideal Solution Ranking Method (TOPSIS) to determine the weight of each index, and applies the weights to the feature vector to obtain a weighted feature vector.

[0034] TOPSIS is a typical multi-attribute decision analysis method. Its core idea is that the selected solution should be closest to the "ideal optimal solution" and furthest from the "ideal worst solution." Its algorithm flow is as follows: (1) Positive orientation of indicator attributes: assuming i One evaluation object, j The evaluation indicators (quantified wettability, gas content, surface quality, internal structure, and lithium plating region) are used to convert all low-performing and neutral indicators into high-performing indicators. The method is as follows:

[0035] in, M It is the expected value of the neutral indicator.

[0036] (2) Constructing the normalized initial matrix: Normalize the forward-directed data to obtain the normalized matrix. Z To eliminate the influence of dimensions.

[0037] (3) Determining the positive and negative ideal solutions: Based on the normalized Z, the cosine method is used to find the positive and negative ideal solutions. Positive ideal solution Z + It is composed of the maximum value of each column's elements, and the negative ideal solution Z. - It is composed of the minimum value in each column.

[0038] (4) Calculate the distance between each evaluation object and the positive and negative ideal solutions: ,

[0039] In the formula, To evaluate the distance between the object and the positive and negative ideal solutions, This is the distance between the evaluation object and the negative ideal solution.

[0040] (5) Calculate the degree of closeness between each evaluation object and the positive ideal solution:

[0041] (6) According to C i Sort, C i The closer the value is to 1, the better the solution. C i The maximum value corresponds to the weights of wettability, gas, surface quality, internal structure, and lithium plating, and the state evaluation is based on these.

[0042] In step 4, the weighted feature vector is input into the optimized SVM model to achieve intelligent diagnostic evaluation of the energy storage battery. SVM, as a supervised learning algorithm with good generalization ability, can effectively handle nonlinear and multi-classification problems, achieving accurate evaluation of the battery's health status. The SVM classification function formula is:

[0043] in, For kernel function, For Lagrange multipliers, The sample labels are (+1 indicates normal status, -1 indicates damaged status). This is the bias term (the intercept term of the decision surface, which determines the distance between the decision surface and the origin).

[0044] (1) Model construction: ① A full life cycle battery image database based on ultrasonic scanning was constructed to form a sample matrix and label vector; ② Data preprocessing including data cleaning, data alignment, and standardization was completed and the extracted core features of the ultrasonic images were called; ③ Data labeling was performed on the battery state to be evaluated and each sample was labeled; ④ The radial basis function (RBF) was selected as the kernel function and the grid search method was applied to optimize the penalty factor C and the RBF kernel parameters.

[0045] (2) Status evaluation: ① Perform model inference, load the constructed and trained SVM model, and input the preprocessed feature vector into the model; ② Perform result output and evaluation, output the category label based on the classification model, and send the result to the upper layer system.

[0046] In step 5, the differentiated maintenance plan is as follows: based on the intelligent diagnostic evaluation results, routine inspections are carried out on batteries in the "normal" state; planned maintenance is performed on batteries with "minor damage" by cleaning the surface, tightening connections, and calibrating parameters; and emergency shutdown maintenance is carried out on batteries with "severe damage" by isolating the battery pack, replacing core components, and conducting system-level testing. This reduces maintenance costs, reduces downtime, extends system life, and improves overall safety, realizing the transformation from "post-fault maintenance" to "predictive maintenance" and establishing a complete closed loop for battery life cycle management.

[0047] This risk-level-based intelligent decision-making avoids over-maintenance or under-maintenance, enables precise allocation and optimized configuration of operation and maintenance resources, and significantly improves operation and maintenance efficiency and economic benefits.

[0048] The SVM-based non-destructive testing and evaluation method for energy storage batteries proposed in this invention brings significant and profound benefits in many aspects. These benefits together constitute its core competitiveness and value in the field of battery health management: Breakthrough in Efficiency: Achieving "Lightning Speed" in Battery Testing. This invention achieves a significant efficiency improvement in battery health status assessment, reducing evaluation time to just 8 minutes. This speed improvement enables the invention to handle routine rapid inspections of large-scale energy storage power stations, real-time quality control on production lines, and rapid diagnostics of electric vehicle battery packs. It drastically shortens the diagnostic cycle, significantly reduces the time and labor costs associated with testing, and provides unprecedented possibilities for the efficient utilization and rapid maintenance of battery assets.

[0049] Enhanced Accuracy: "Precise Vision" and "Intelligent Classification" of Microscopic Defects. This invention leverages the unique advantages of integrating ultrasonic scanning and the SVM intelligent algorithm to improve evaluation accuracy. First, at the physical level, the system can accurately locate gas-generating areas inside the battery with a precision of 0.5mm. This allows for the accurate identification and visualization of even minute bubbles or early gas generation phenomena, providing maintenance personnel with a precise "navigation map." Second, at the intelligent diagnostic level, after rigorous validation on large-scale datasets (especially complex and easily degraded high-temperature cycling datasets for pouch batteries), the SVM classification accuracy consistently exceeds 90%. This means that the system exhibits extremely high reliability and consistency in distinguishing between "normal," "minor damage," and "severe damage" battery states, significantly reducing the risks of false negatives and missed positives common in traditional methods, ensuring accurate assessment of battery health.

[0050] Safety Assurance: A Fundamental Leap from "Post-Event Crisis" to "Pre-Event Prevention." The most significant benefit of this invention lies in its groundbreaking improvement in battery safety. Through real-time sensing of the battery's internal microstructure using high-resolution ultrasound, and the intelligent identification and prediction of these microscopic changes using an SVM model, this invention can provide early warnings of critical precursors to potential thermal runaway, such as lithium plating and gas production. These early, hidden defects are often difficult to detect using traditional methods, until they develop into irreversible catastrophic events. This invention's "pre-event prevention" capability provides operators with a valuable early warning window, enabling them to take effective intervention measures before thermal runaway occurs, such as emergency discharge, local isolation, shutdown for maintenance, or replacement. This marks a complete shift in battery safety management philosophy from passive "post-event analysis" and "crisis response" to proactive "pre-event prevention" and "risk avoidance," greatly ensuring the safety of personnel and property, as well as the stability and reliability of the power grid.

[0051] Example 2 This embodiment provides a non-destructive testing and evaluation system for energy storage batteries based on support vector machines, used to implement the non-destructive testing and evaluation method for energy storage batteries described in Embodiment 1, such as... Figure 5 As shown, it consists of an image data acquisition layer based on ultrasonic nondestructive testing, a battery state evaluation layer based on the SVM algorithm, and a visualization result output layer for battery state evaluation.

[0052] The image data acquisition layer uses ultrasonic non-destructive scanning equipment to image the electrolyte wetting state, gas production, lithium plating, and electrode characterization of the battery, thereby acquiring reflected / transmitted ultrasonic images of the energy storage battery.

[0053] The battery status evaluation layer is used to extract core features from ultrasonic images and form a five-dimensional evaluation index system that considers wettability, gas density, surface quality, internal structure, and lithium plating detection. The TOPSIS method is used to determine the weight of each index and apply the weight to the feature vector to obtain a weighted feature vector. The weighted feature vector is then input into the optimized SVM model to realize intelligent diagnostic evaluation of the energy storage battery status.

[0054] The visualization output layer is used to generate a visualization interface for non-destructive testing and evaluation of energy storage batteries, transforming complex analysis results into intuitive information. The control panel has four options: "Select Image," "Retest," "Save Results," and "Clear Image." It outputs clear three-level status classification results, namely, intelligent diagnostic evaluation results indicating whether the battery is in a "normal," "slightly damaged," or "severely damaged" state. The intelligent diagnostic evaluation results are linked to the operation and maintenance strategies of the energy storage power station, automatically generating differentiated maintenance plans.

[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation of the present invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications and substitutions should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. A non-destructive testing and evaluation method for energy storage batteries based on support vector machines, characterized in that, include: Ultrasonic non-destructive scanning equipment was used to image the electrolyte wetting state, gas production, lithium deposition and electrode characterization of the battery, and to obtain reflected / transmitted ultrasonic images of the energy storage battery. Core features were extracted from the ultrasound images, including features of battery fluid wetting, gas accumulation areas, interface and material continuity, and lithium metal deposition areas. Based on the extracted multi-dimensional features, a five-dimensional evaluation index system is formed, which includes wettability, gas density, surface quality, internal structure, and lithium plating detection. The weight of each index is determined by the approximation ideal solution ranking method, and the weights are applied to the feature vector to obtain the weighted feature vector. The weighted feature vector is input into the optimized SVM model to achieve intelligent diagnosis and evaluation of the energy storage battery status.

2. The method for non-destructive testing and evaluation of energy storage batteries according to claim 1, characterized in that, Also includes: A visualization interface for non-destructive testing and evaluation of energy storage batteries is generated, transforming complex analysis results into intuitive information. The control panel has four options: "Select Image," "Retest," "Save Results," and "Clear Image." It outputs clear three-level status classification results, namely, intelligent diagnostic evaluation results indicating whether the battery is in a "normal," "slightly damaged," or "severely damaged" state. The intelligent diagnostic evaluation results are linked to the operation and maintenance strategies of the energy storage power station, automatically generating differentiated maintenance plans.

3. The method for non-destructive testing and evaluation of energy storage batteries according to claim 2, characterized in that, The differentiated maintenance plan is as follows: based on the intelligent diagnostic evaluation results, routine inspections are carried out on batteries in "normal" condition, planned maintenance is performed on batteries with "minor damage" by cleaning the surface, tightening connections, and calibrating parameters, and emergency shutdown maintenance is carried out on batteries with "severe damage" by isolating the battery pack, replacing core components, and conducting system-level testing.

4. The method for non-destructive testing and evaluation of energy storage batteries according to any one of claims 1-3, characterized in that, While acquiring ultrasonic images, the typical characteristics of energy storage batteries in four key stages—grid access, operation, maintenance, and decommissioning—are analyzed. A mapping model between battery status and the entire life cycle stages is established, forming an image status database of energy storage batteries based on ultrasonic scanning and covering the entire life cycle.

5. The method for non-destructive testing and evaluation of energy storage batteries according to any one of claims 1-3, characterized in that, The steps of the sorting method for approximating the ideal solution include: 1) Positive orientation of indicator attributes: (The following is a list of parameters, not part of the main text) i One evaluation object, j One evaluation indicator, which transforms all low-quality and neutral indicators into high-quality indicators; 2) Constructing the initial normalized matrix: Normalize the forward-oriented data to obtain the normalized matrix. Z ; 3) Determine the positive and negative ideal solutions: Based on the normalized matrix Z, the cosine method is used to find the positive and negative ideal solutions; 4) Calculate the distance between each evaluation object and the positive and negative ideal solutions: 5) Calculate the degree of closeness between each evaluation object and the positive ideal solution. C i ; 6) Based on proximity C i The ranking is based on the weights of wettability, gas, surface quality, internal structure, and lithium plating, with the index corresponding to the maximum degree of proximity being designed, and the state evaluation is performed based on this.

6. The method for non-destructive testing and evaluation of energy storage batteries according to claim 5, characterized in that, The ideal solution Z + It is composed of the maximum value of each column's elements, and the negative ideal solution Z. - It is composed of the minimum value in each column.

7. The method for non-destructive testing and evaluation of energy storage batteries according to any one of claims 1-3, characterized in that, The SVM model described above forms a sample matrix and label vector based on the image state database of energy storage batteries; it completes data preprocessing and calls the extracted core features of ultrasound images.

8. The method for non-destructive testing and evaluation of energy storage batteries according to any one of claims 1-3, characterized in that, The SVM model described above labels the battery state to be evaluated, giving each sample a label.

9. The method for non-destructive testing and evaluation of energy storage batteries according to any one of claims 1-3, characterized in that, The SVM model described above selects the radial basis function kernel as the kernel function and applies a grid search method to fine-tune the penalty factor C and the kernel function parameters.

10. A non-destructive testing and evaluation system for energy storage batteries based on support vector machines, used to implement the non-destructive testing and evaluation method for energy storage batteries according to any one of claims 1-9, characterized in that, include: The image data acquisition layer uses ultrasonic non-destructive scanning equipment to image the electrolyte wetting state, gas production, lithium plating and electrode characterization of the battery, and to acquire reflected / transmitted ultrasonic images of the energy storage battery. The battery status evaluation layer is used to extract core features from ultrasonic images and form a five-dimensional evaluation index system of wettability, gas density, surface quality, internal structure, and lithium plating detection. The weight of each index is determined by the approximation ideal solution ranking method, and the weights are applied to the feature vector to obtain a weighted feature vector. The weighted feature vector is then input into the optimized SVM model to realize intelligent diagnosis and evaluation of the energy storage battery status. The visualization output layer is used to generate a visualization interface for non-destructive testing and evaluation of energy storage batteries, transforming complex analysis results into intuitive information.

Images