Full life cycle performance analysis method and system based on battery replacement battery and medium

Abstract

The invention provides a full life cycle performance analysis method and system based on a battery replacement battery and a medium, and the method comprises the steps: firstly, carrying out the comprehensive collection and preprocessing of battery charging and discharging data and environment data, constructing a structured multi-dimensional feature vector through feature extraction, inputting the structured multi-dimensional feature vector into a pre-trained random forest prediction model, and carrying out the calculation to obtain a first prediction SOH value; then, performing grading judgment on the predicted value based on a preset triple threshold comparison mechanism, generating different grades of early warning signals and maintenance suggestions, and automatically outputting a visual SOH attenuation curve and a text report; finally, the performance of the model is monitored regularly through the deviation between the actually measured SOH value and the predicted value, and a model iteration and threshold dynamic optimization mechanism is triggered when the deviation exceeds the limit. According to the invention, through combination of multi-source data fusion and machine learning prediction, the accuracy and reliability of battery health state evaluation are improved; and a multi-stage early warning and dynamic feedback mechanism is introduced, so that sudden faults of the battery are avoided.

Description

Technical Field

[0001] This invention relates to the field of battery swapping, and more specifically, to a method, system, and medium for full life-cycle performance analysis of battery swapping. Background Technology

[0002] Currently, with the promotion of electric vehicles and battery swapping, the accurate assessment of the State of Health (SOH) and the full life-cycle management of batteries, as core components, have become a key focus of the industry. Traditional battery performance analysis methods often rely on single types of data, such as linear extrapolation based solely on the number of charge-discharge cycles or capacity decay curves. These methods fail to effectively integrate environmental data and real-time operating parameters, resulting in incomplete data dimensions, coarse feature extraction, and an inability to fully reflect the complex degradation mechanisms of batteries. Furthermore, most existing technologies use fixed thresholds or simple logical judgments for battery life termination warnings, lacking intelligent discrimination mechanisms based on multi-level dynamic thresholds. This makes them unsuitable for different battery types, usage scenarios, and individual differences, often leading to false alarms or missed alarms. This increases the risk of sudden failures and may result in premature battery retirement, causing resource waste. Regarding predictive models, although some methods incorporate machine learning algorithms, such as support vector machines and simple neural networks, these models are mostly static, lacking continuous learning and online calibration capabilities. Over time, they are prone to increased prediction bias and decreased generalization ability, and they fail to achieve closed-loop linkage between prediction results and operational decisions. More importantly, existing systems often handle monitoring, prediction, and maintenance in isolation, failing to build a fully automated system encompassing data acquisition, feature extraction, model prediction, threshold alerts, and feedback optimization. Their outputs are mostly presented in numerical or simple chart formats, lacking actionable natural language suggestions and intuitive visualization aids, increasing the decision-making burden on maintenance personnel and reducing management efficiency. These shortcomings collectively limit the applicability and reliability of existing technologies in actual battery swapping operation scenarios. Therefore, there is an urgent need for a battery lifecycle performance analysis method that can integrate multi-source data, possess dynamic optimization capabilities, and support closed-loop management. Summary of the Invention

[0003] In view of the above problems, the purpose of this invention is to provide a method, system, and medium for full life cycle performance analysis of battery swapping systems. This method achieves accurate prediction and intelligent management of battery health status through multi-source data fusion and machine learning algorithms. First, it systematically collects battery charging / discharging data and environmental data. After preprocessing and feature extraction, a structured multi-dimensional feature vector is constructed, and a pre-trained random forest model is used to output the predicted state of health (SOH). Then, a triple threshold comparison mechanism is introduced to trigger tiered early warnings and generate differentiated maintenance suggestions based on the predicted values, while automatically generating a visualized degradation curve and text report. Furthermore, by periodically comparing the deviation between predicted and measured values, the model iteration and threshold optimization process is dynamically triggered to achieve closed-loop self-correction. This invention improves the accuracy and reliability of battery health status assessment, significantly reduces the risk of sudden failures, optimizes maintenance resource allocation, and extends battery service life. It represents a digital management technology for the entire life cycle of battery swapping systems.

[0004] The first aspect of this invention provides a method for full life-cycle performance analysis based on battery swapping, the method comprising: The system acquires battery charging and discharging data and environmental data, and obtains structured multidimensional feature vectors based on preset preprocessing and feature extraction logic. The multidimensional feature vector is input into a pre-trained random forest prediction model to obtain the first predicted SOH value; Based on the preset triple threshold comparison mechanism, according to the first predicted SOH value, a graded early warning signal is output and a graded maintenance suggestion is generated; Based on the first predicted SOH value and the graded maintenance recommendations, a maintenance text recommendation is generated and a visual SOH decay curve is output. Based on the preset optimization cycle, the first measured SOH value is obtained, and the SOH deviation value is obtained by combining the first predicted SOH value. If the SOH deviation value exceeds the preset deviation threshold, the random forest prediction model and the triple threshold are updated based on the preset model iteration and threshold optimization mechanism.

[0005] In this scheme, the structured multidimensional feature vector is obtained based on the preset preprocessing and feature extraction logic, specifically including: Acquire raw data, including the charge / discharge data and environmental data; Based on the preset integrity verification and outlier removal mechanism, processed data is obtained from the original data; Based on a preset timestamp alignment and data fusion mechanism, aligned data is obtained according to the processed data; Based on the preset feature index extraction logic, a feature vector is obtained according to the aligned data; Based on the feature vector, a multidimensional feature vector is obtained using a preset structured operation.

[0006] In this scheme, the step of inputting the multidimensional feature vector into a pre-trained random forest prediction model to obtain the first predicted SOH value specifically includes: Based on historical datasets, the random forest algorithm is trained and cross-validated to generate the pre-trained random forest prediction model. The multidimensional feature vector is input into a pre-trained random forest prediction model, and the first predicted SOH value is output. Based on a preset time series analysis algorithm, the first predicted SOH value is serialized to generate an SOH value decay trend curve.

[0007] In this scheme, the preset triple threshold comparison mechanism, based on the first predicted SOH value, outputs a graded early warning signal and generates graded maintenance suggestions, specifically including: Based on battery type and application scenario, three thresholds are obtained, including warning threshold, replacement threshold and emergency threshold; If the first predicted SOH value is higher than the preset warning threshold, monitoring will continue. If the first predicted SOH value is between the warning threshold and the replacement threshold, a first-level warning signal is output, and a suggestion to increase the detection frequency is generated. If the first predicted SOH value is between the replacement threshold and the emergency threshold, a secondary replacement signal is output and a battery replacement plan is generated. If the first predicted SOH value is lower than the emergency threshold, a level three emergency signal is output and a battery replacement command is generated.

[0008] In this solution, the step of generating maintenance text suggestions and outputting a visualized SOH decay curve based on the first predicted SOH value and graded maintenance suggestions specifically includes: Based on the hierarchical maintenance recommendations, a natural language description template is selected from the preset text template library; The maintenance text suggestion is generated by filling the selected natural language description template with the first predicted SOH value, battery identifier, and timestamp information; Based on the graphics processing component, an SOH decay curve is plotted according to the first predicted SOH value; Based on the aforementioned triple threshold, a reference line is marked on the SOH decay curve; According to the SOH value decay trend curve, add it to the tail of the SOH decay curve.

[0009] In this scheme, if the SOH deviation value exceeds a preset deviation threshold, the random forest prediction model and the triple threshold are updated based on a preset model iteration and threshold optimization mechanism, specifically including: When the SOH deviation value exceeds the preset deviation threshold, the model iteration and threshold optimization mechanism are triggered. The multidimensional feature vector and the first measured SOH value are incorporated into the historical dataset to perform incremental training or complete retraining of the random forest prediction model. Based on the SOH deviation value, the triple threshold is adjusted and updated according to the preset fine-tuning step. The updated model and thresholds are deployed to the online prediction system, and once the accuracy is verified, they are deployed locally.

[0010] A second aspect of the present invention provides a full life-cycle performance analysis system based on a battery swapping system, including a full life-cycle performance analysis method program based on a battery swapping system. When the processor executes the full life-cycle performance analysis method program based on a battery swapping system, it performs the following steps: The system acquires battery charging and discharging data and environmental data, and obtains structured multidimensional feature vectors based on preset preprocessing and feature extraction logic. The multidimensional feature vector is input into a pre-trained random forest prediction model to obtain the first predicted SOH value; Based on the preset triple threshold comparison mechanism, according to the first predicted SOH value, a graded early warning signal is output and a graded maintenance suggestion is generated; Based on the first predicted SOH value and the graded maintenance recommendations, a maintenance text recommendation is generated and a visual SOH decay curve is output. Based on the preset optimization cycle, the first measured SOH value is obtained, and the SOH deviation value is obtained by combining the first predicted SOH value. If the SOH deviation value exceeds the preset deviation threshold, the random forest prediction model and the triple threshold are updated based on the preset model iteration and threshold optimization mechanism.

[0011] In this scheme, the structured multidimensional feature vector is obtained based on the preset preprocessing and feature extraction logic, specifically including: Acquire raw data, including the charge / discharge data and environmental data; Based on the preset integrity verification and outlier removal mechanism, processed data is obtained from the original data; Based on a preset timestamp alignment and data fusion mechanism, aligned data is obtained according to the processed data; Based on the preset feature index extraction logic, a feature vector is obtained according to the aligned data; Based on the feature vector, a multidimensional feature vector is obtained using a preset structured operation.

[0012] In this scheme, the step of inputting the multidimensional feature vector into a pre-trained random forest prediction model to obtain the first predicted SOH value specifically includes: Based on historical datasets, the random forest algorithm is trained and cross-validated to generate the pre-trained random forest prediction model. The multidimensional feature vector is input into a pre-trained random forest prediction model, and the first predicted SOH value is output. Based on a preset time series analysis algorithm, the first predicted SOH value is serialized to generate an SOH value decay trend curve.

[0013] A third aspect of the present invention provides a computer-readable storage medium comprising a method program for full life cycle performance analysis based on a battery swapping system. When the method program is executed by a processor, it implements the steps of the full life cycle performance analysis method for a battery swapping system as described in any of the preceding claims.

[0014] This invention provides a method, system, and medium for full lifecycle performance analysis of swappable batteries. First, it comprehensively collects and preprocesses battery charging / discharging data and environmental data. A structured multidimensional feature vector is constructed through feature extraction and input into a pre-trained random forest prediction model to calculate the first predicted State of Health (SOH) value. Then, based on a preset triple threshold comparison mechanism, the predicted value is graded and judged, generating different levels of early warning signals and maintenance suggestions, and automatically outputting a visualized SOH decay curve and text report. Finally, the model performance is monitored periodically by comparing the deviation between the measured SOH value and the predicted value, and a model iteration and threshold dynamic optimization mechanism is triggered when the deviation exceeds the limit. This invention improves the accuracy and reliability of battery health status assessment by combining multi-source data fusion and machine learning prediction; the introduced multi-level early warning and dynamic feedback mechanism avoids sudden battery failures. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope.

[0016] Figure 1 A flowchart of a method for full life-cycle performance analysis based on a battery swapping device according to the present invention is shown; Figure 2 This invention provides a flowchart of a data preprocessing and feature extraction process according to an embodiment of the invention. Figure 3A flowchart illustrating a prediction process for SOH values ​​provided by an embodiment of the present invention is shown. Figure 4 A block diagram of a full life-cycle performance analysis system based on a battery swapping device according to the present invention is shown. Detailed Implementation

[0017] 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.

[0018] Unless otherwise defined, all terms (including technical and scientific terms) used in embodiments of this invention shall have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. It should also be understood that terms such as those defined in a common dictionary shall be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and not as being interpreted in an idealized or highly formalized sense, unless expressly defined in this embodiment of the invention.

[0019] The terms "first," "second," and similar words used in the embodiments of this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "an," "a," or "the" do not indicate a quantity limitation, but rather indicate the presence of at least one. Similarly, terms such as "including" or "comprising" mean that the element or object preceding the word encompasses the elements or objects listed after the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The steps preceding or following the steps in the method of the embodiments of this invention are not necessarily performed precisely in sequence. Instead, various steps can be processed in reverse order or simultaneously. Furthermore, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0020] In addition, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0021] Figure 1 A flowchart of a method for full life-cycle performance analysis based on swappable batteries according to the present invention is shown.

[0022] like Figure 1As shown, the first aspect of this invention discloses a method for full life-cycle performance analysis based on battery swapping, the method comprising: S120: Acquire battery charging and discharging data and environmental data, and obtain structured multidimensional feature vectors based on preset preprocessing and feature extraction logic; S140, The multidimensional feature vector is input into the pre-trained random forest prediction model to obtain the first predicted SOH value; S160, based on a preset triple threshold comparison mechanism, outputs a graded early warning signal and generates graded maintenance suggestions according to the first predicted SOH value; S180, Based on the first predicted SOH value and the graded maintenance recommendations, generate maintenance text recommendations and output a visualized SOH decay curve; S110, based on the preset optimization cycle, obtain the first measured SOH value, and combine it with the first predicted SOH value to obtain the SOH deviation value; S112, if the SOH deviation value exceeds the preset deviation threshold, then based on the preset model iteration and threshold optimization mechanism, update the random forest prediction model and the triple threshold.

[0023] It should be noted that the charge and discharge data includes real-time data streams and operating status data of the battery management system. The real-time data streams include at least the number of cycles, depth of charge, and charge / discharge rate. The operating status data includes at least the operating voltage, operating current, and operating internal resistance. The environmental data includes at least the battery temperature and battery humidity values. The SOH value is the battery health status value, which indicates the degree of degradation of the battery's current state relative to its initial state.

[0024] This embodiment provides a method for full lifecycle performance analysis based on swappable batteries. First, it systematically acquires multi-source heterogeneous data generated during battery operation. This data includes real-time charge / discharge cycle parameters provided by the battery management system and environmental sensor data such as temperature and humidity. After successfully acquiring the raw data, a pre-defined preprocessing and feature extraction logic is executed. This includes cleaning the raw data to eliminate noise interference, aligning the timestamps of the multi-source data for data fusion, and extracting key feature indicators that deeply reflect the battery's degradation state from the fused data. This yields a structured multi-dimensional feature vector. Subsequently, the multi-dimensional feature vector is input into a random forest prediction model pre-trained using a large amount of historical data. This model performs collaborative reasoning and calculation through multiple integrated decision trees, ultimately outputting a first predicted State of Health (SOH) value for the battery. Then, a pre-defined triple threshold comparison mechanism is activated. By comparing the obtained first predicted SOH value with pre-defined multi-level thresholds, different levels of urgency-based warning signals are output, and graded maintenance recommendations are generated accordingly. Finally, maintenance text is automatically generated based on the graded comparison prediction results and maintenance recommendations. Simultaneously, a visualized SOH decay curve integrating historical data and future trends is output for user decision-making reference. By obtaining the battery's actual measured SOH value and comparing it with the predicted value to obtain the SOH deviation value, if this deviation value exceeds the preset deviation threshold, it will trigger the model iteration and threshold optimization mechanism to update the parameters of the prediction model and adaptively adjust the specific values ​​of the three thresholds, thereby forming a closed-loop system for full life cycle performance analysis with self-learning capabilities.

[0025] Figure 2 A flowchart illustrating a data preprocessing and feature extraction process provided by an embodiment of the present invention is shown.

[0026] According to embodiments of the present invention, such as Figure 2 As shown, the structured multidimensional feature vector obtained based on the preset preprocessing and feature extraction logic specifically includes: S202, Obtain raw data, including the charge / discharge data and environmental data; S204, Based on the preset integrity verification and outlier removal mechanism, the processed data is obtained according to the original data; S206, Based on a preset timestamp alignment and data fusion mechanism, aligned data is obtained according to the processed data; S208, Based on the preset feature index extraction logic, the feature vector is obtained according to the alignment data; S210, Based on the feature vector, a multidimensional feature vector is obtained using a preset structured operation.

[0027] It should be noted that the feature vector includes at least the average temperature within the cycle, the temperature fluctuation variance, the duration of the constant voltage charging phase, the capacity decay rate, and the internal resistance change rate. This embodiment provides a preprocessing and feature extraction logic, which is a multi-stage data refinement and feature extraction process. In this embodiment, firstly, raw data streams including voltage, current, internal resistance, temperature, humidity, cycle count, depth of charge, and charge / discharge rate are obtained from the underlying sensors and battery management system. Secondly, based on a preset integrity verification mechanism, the raw data is checked to identify and process missing values. Simultaneously, based on statistical methods or threshold rules, an outlier removal mechanism is activated to filter out abnormal data points that significantly deviate from the normal range, thereby obtaining processed data. Then, a preset timestamp alignment and data fusion mechanism is invoked to synchronize and interpolate the processed data from different sources, ensuring that all data points are fully aligned in the time dimension, forming an aligned dataset. Based on this, a preset feature index extraction logic is executed. Key feature indicators characterizing the battery performance degradation trajectory are calculated and extracted from the aligned data. These key feature indicators include, but are not limited to, the average temperature and temperature fluctuation variance within the cycle to reflect thermal management status, the duration of the constant voltage charging phase to assess charging behavior characteristics, and the capacity decay rate and internal resistance change rate to indicate health levels. Finally, all extracted feature indicators are normalized and vectorized to obtain a multi-dimensional feature vector, providing standardized input for subsequent machine learning models.

[0028] Figure 3 A flowchart illustrating a prediction process for SOH values ​​provided by an embodiment of the present invention is shown.

[0029] According to embodiments of the present invention, such as Figure 3 As shown, the step of inputting the multidimensional feature vector into a pre-trained random forest prediction model to obtain the first predicted SOH value specifically includes: S302, Based on historical datasets, the random forest algorithm is trained and cross-validated to generate the pre-trained random forest prediction model; S304, input the multidimensional feature vector into the pre-trained random forest prediction model and output the first predicted SOH value; S306, based on a preset time series analysis algorithm, the first predicted SOH value is serialized to generate an SOH value decay trend curve.

[0030] It should be noted that the historical dataset includes historical multidimensional feature vectors and their corresponding measured SOH values. In this embodiment, during the model training phase, a historical dataset with a large number of historical samples is used, where each sample consists of historical multidimensional feature vectors and measured SOH values. Using this dataset, a random forest algorithm is employed for model training, and cross-validation is used to evaluate model performance and prevent overfitting, ultimately generating a random forest prediction model. During the online prediction phase, the current multidimensional feature vector is input into this pre-trained model, and the first predicted SOH value is output. Furthermore, to achieve a trend-based understanding of future battery performance, a preset time series analysis algorithm is used to model and analyze the current and historical first predicted SOH values ​​as a time series. By fitting its decay trajectory, an SOH value decay trend curve is generated, thereby providing users with a forward-looking insight into the remaining usable life of the battery.

[0031] According to an embodiment of the present invention, the step of outputting a graded early warning signal and generating graded maintenance suggestions based on a preset triple threshold comparison mechanism and a first predicted SOH value specifically includes: Based on battery type and application scenario, three thresholds are obtained, including warning threshold, replacement threshold and emergency threshold; If the first predicted SOH value is higher than the preset warning threshold, monitoring will continue. If the first predicted SOH value is between the warning threshold and the replacement threshold, a first-level warning signal is output, and a suggestion to increase the detection frequency is generated. If the first predicted SOH value is between the replacement threshold and the emergency threshold, a secondary replacement signal is output and a battery replacement plan is generated. If the first predicted SOH value is lower than the emergency threshold, a level three emergency signal is output and a battery replacement command is generated.

[0032] It should be noted that this embodiment provides a triple threshold comparison and early warning mechanism, which is a precise decision-making process based on multi-level threshold values. In this embodiment, firstly, based on the specific chemical type of the battery, the application scenario, and the operation strategy, a set of three key thresholds are dynamically or statically set: an early warning threshold, a replacement threshold, and an emergency threshold. Then, the first predicted SOH value is sequentially compared and logically judged with these three thresholds. If the first predicted SOH value is higher than the early warning threshold, the battery is determined to be in a healthy state, and the normal monitoring mode is maintained without generating any early warning signal. If the first predicted SOH value drops to the range between the early warning threshold and the replacement threshold, a first-level early warning signal is output, indicating that the battery performance has shown signs of degradation, and a maintenance suggestion to increase the detection frequency is automatically generated to monitor its degradation trend. If the first predicted SOH value further drops to the range between the replacement threshold and the emergency threshold, a second-level replacement signal is output, indicating that the battery performance no longer meets the normal usage requirements, and a maintenance suggestion for a battery replacement plan is automatically formulated to initiate a planned replacement process. If the first predicted SOH value eventually falls below the emergency threshold, a level 3 emergency signal with the highest priority will be output, indicating that the battery has an extremely high risk of failure. An instruction will be generated to immediately stop using the battery and perform an emergency replacement, thereby ensuring safety and reliability.

[0033] According to an embodiment of the present invention, the step of generating maintenance text suggestions and outputting a visualized SOH decay curve based on the first predicted SOH value and graded maintenance suggestions specifically includes: Based on the hierarchical maintenance recommendations, a natural language description template is selected from the preset text template library; The maintenance text suggestion is generated by filling the selected natural language description template with the first predicted SOH value, battery identifier, and timestamp information; Based on the graphics processing component, an SOH decay curve is plotted according to the first predicted SOH value; Based on the aforementioned triple threshold, a reference line is marked on the SOH decay curve; According to the SOH value decay trend curve, add it to the tail of the SOH decay curve.

[0034] It should be noted that this embodiment provides a process for generating maintenance text suggestions and visual output, specifically a process for converting machine decisions into human-readable information. In this embodiment, a graded maintenance suggestion signal is first received, and a corresponding natural language description template is automatically selected from a pre-built text template library. The templates in the library are constructed using professional domain knowledge to ensure the professionalism and accuracy of the output suggestions. After selecting a template, the current first predicted SOH value, the unique identifier of the monitored battery, and the current timestamp information are used as variables and automatically filled into the reserved positions in the template. Through this data and text fusion, a semantically clear and directly executable maintenance text suggestion is generated. Simultaneously, an embedded graphics processing component is invoked to plot an SOH decay curve covering historical data points, with time as the horizontal axis and the estimated SOH value as the vertical axis. At the same time, horizontal reference lines representing warning thresholds, replacement thresholds, and emergency thresholds are drawn on the curve with distinct colors or line styles, thereby indicating the current battery state relative to the safety boundary. Finally, the generated future SOH value decay trend curve is added to the end of the current SOH decay curve in the form of a dashed line or different colors, realizing an integrated presentation of history, current status and future trends, thus providing users with an intuitive and information-rich decision dashboard.

[0035] According to an embodiment of the present invention, if the SOH deviation value exceeds a preset deviation threshold, the random forest prediction model and the triple threshold are updated based on a preset model iteration and threshold optimization mechanism, specifically including: When the SOH deviation value exceeds the preset deviation threshold, the model iteration and threshold optimization mechanism are triggered. The multidimensional feature vector and the first measured SOH value are incorporated into the historical dataset to perform incremental training or complete retraining of the random forest prediction model. Based on the SOH deviation value, the triple threshold is adjusted and updated according to the preset fine-tuning step. The updated model and thresholds are deployed to the online prediction system, and once the accuracy is verified, they are deployed locally.

[0036] It should be noted that this embodiment provides a model and threshold update mechanism. In this embodiment, by continuously monitoring the deviation between the predicted and measured SOH values, once the SOH deviation value exceeds the deviation threshold representing the tolerance for model performance degradation, the built-in model iteration and threshold optimization mechanism is activated. In the model iteration stage, the multidimensional feature vector that caused the deviation and the first measured SOH value are used as new training samples and incorporated into the historical dataset used for model training. Based on the expanded dataset, the original random forest prediction model is incrementally trained or completely retrained. This allows the model to learn the latest degradation patterns of battery performance from the new data, thereby refreshing its internal parameters and improving the accuracy of its future predictions. In the threshold optimization stage, based on the magnitude and direction of the calculated SOH deviation value and based on preset fine-tuning step rules, the specific values ​​of the three thresholds—warning, replacement, and emergency—are adaptively adjusted. For example, if the model continuously overestimates SOH, the threshold may be moderately increased to provide an early warning. After updating the model and thresholds, the updated model and thresholds are first deployed in an isolated online prediction sandbox environment. Only after the test and verification are passed will they be finally deployed to the online prediction system in the production environment to maintain the credibility of the system's analysis results.

[0037] It is worth mentioning that it also includes: After implementing the tiered maintenance recommendations, record the battery's subsequent operating status information and fault occurrence information; Based on the fault occurrence information, the subsequent operating status information is correlated with the executed maintenance recommendations to generate a maintenance effectiveness evaluation report; The maintenance effectiveness evaluation report is fed back to the model iteration and threshold optimization mechanism.

[0038] It should be noted that this embodiment provides a closed-loop feedback mechanism to verify and optimize the initial maintenance decisions through practical effects, thereby improving the overall intelligence level of the system. In this embodiment, after executing the graded maintenance recommendations, a long-term effect tracking phase is initiated. The subsequent operating status information of the battery is continuously recorded, including its charging and discharging behavior, environmental conditions, and detailed information on whether sudden failures occur. A correlation analysis process is initiated to analyze the causal relationship between the executed maintenance recommendations and the actual failure status and performance degradation rate of the battery; for example, if a battery recommended for "planned replacement" quickly fails, it indicates that the original warning may have been insufficient. Based on this deep correlation analysis, an evaluation report is automatically generated, assessing the effectiveness and accuracy of the previous warning decisions. Finally, the evaluation report is fed back to the aforementioned model iteration and threshold optimization mechanism, serving as an important auxiliary decision-making basis for optimizing the parameters of the random forest prediction model and adjusting the triple threshold values. This forms a continuous learning closed loop from decision-making to execution to feedback and ultimately to decision optimization, enhancing the practicality and reliability of the system.

[0039] It is worth mentioning that after generating the hierarchical maintenance recommendations, the following is also included: In response to a battery replacement plan or battery replacement instruction, a battery replacement work order is generated, which includes at least the replacement priority and battery identifier. The replacement work order is pushed to the work order pool queue of the operation and maintenance management system. Based on the work order pool queue, a maintenance work order is generated, which includes at least the location of the battery swapping station and the recommended replacement time. After the battery is replaced, its decommissioned status is marked based on the SOH value of the old battery.

[0040] It should be noted that in this embodiment, when the system generates a tiered maintenance suggestion containing a "battery replacement plan" or "battery replacement instruction," it automatically triggers a work order creation process. Based on the urgency of the suggestion, the geographical location of the battery swapping station, the battery model, and other information, a structured battery replacement work order is automatically generated. The work order includes at least a replacement priority for scheduling decisions and a battery identifier for accurate identification. Subsequently, the work order is pushed to the work order pool queue of the operation and maintenance management system through a preset interface, awaiting scheduling and allocation. The operation and maintenance management system then automatically assigns the work order to the most suitable operation and maintenance personnel based on the priority of all work orders in the work order pool, geographical location clustering, and the current location and workload of the operation and maintenance personnel, using a built-in scheduling algorithm. A detailed maintenance work order is generated, including the specific location of the battery swapping station, the recommended replacement time window, and other key execution information. After the operation and maintenance personnel arrive on-site and complete the battery replacement operation, they confirm the work order completion via their mobile terminal and determine whether the removed battery is decommissioned based on the measured SOH value. This not only completes the closed loop of a single maintenance task but also provides a complete record of the battery's entire lifecycle.

[0041] It is worth mentioning that it also includes: Obtain the real-time first predicted SOH value and its degradation trend curve of multiple batteries in the battery swapping station; Based on the SOH value, remaining capacity, historical cycle count, and current battery swapping service demand of each battery, a battery scheduling optimization objective function is constructed. Combining a pre-set reinforcement learning algorithm, the optimal battery allocation and scheduling strategy is output to match replacement work orders; The charging strategy and maintenance plan for batteries within the battery swapping station are dynamically adjusted based on the scheduling results.

[0042] It should be noted that in this embodiment, the first predicted State of Health (SOH) value and its future degradation trend of all batteries in the battery swapping station are monitored in real time, and the real-time remaining power, cumulative cycle count, and current battery swapping service request queue of each battery are simultaneously acquired. Based on this multi-source information, a scheduling optimization objective function is constructed with the goals of maximizing battery utilization efficiency, minimizing user waiting time, balancing battery degradation, and prioritizing high-demand services. Subsequently, a reinforcement learning algorithm based on deep networks or policy gradient methods is used to solve this objective function online or near-line. This algorithm learns through continuous interaction and trial and error with the battery swapping station environment, and finally outputs a set of optimal battery allocation strategies. The battery allocation strategy intelligently allocates batteries with excellent health and sufficient power to vehicles with strict range requirements or prioritizes urgent orders, while guiding batteries with average health but normal function to short-distance travel or vehicles that can accept slightly lower performance. This satisfies user needs while slowing down the degradation rate of high SOH batteries and fully utilizing the remaining value of low SOH batteries. At the same time, it will dynamically generate adaptive charging strategies and maintenance plans based on scheduling strategies and SOH prediction results, thereby realizing dynamic allocation of battery resources at the battery swapping station level and improving operational efficiency and the economic value of the battery throughout its entire life cycle.

[0043] It is worth mentioning that it also includes: It aggregates charging and discharging data, environmental data, and SOH decay records of batteries of the same batch or model from multiple battery swapping stations; Based on the aggregated multi-station data, the pre-trained random forest prediction model is trained using federated learning to generate a globally optimized prediction model.

[0044] It should be noted that in this embodiment, a data sharing platform covering multiple battery swapping stations is constructed. This platform collects full lifecycle operational data of batteries of the same batch or model from each station through standardized interfaces, including but not limited to charge and discharge cycle parameters, environmental temperature and humidity change history, historical SOH monitoring values, and final decommissioning records, thereby forming a larger and more diverse enhanced dataset. Without directly sharing the original data, the global initial model is distributedly trained using local data from each station. By aggregating the parameter updates of each local model, a more generalizable globally optimized random forest prediction model is iteratively generated. This global model is then distributed and deployed to the local analysis systems of each battery swapping station to replace or assist the original local prediction models, especially when facing scarce local data or a lack of historical data for new batches of batteries, it can still make accurate predictions.

[0045] Figure 4 A block diagram of a full life-cycle performance analysis system based on a battery swapping device according to the present invention is shown.

[0046] like Figure 4 As shown, the second aspect of the present invention discloses a full life cycle performance analysis system 4 based on a battery swapping system, including a memory 41 and a processor 42. The memory includes a full life cycle performance analysis method program based on a battery swapping system. When the processor executes the full life cycle performance analysis method program based on a battery swapping system, it performs the following steps: The system acquires battery charging and discharging data and environmental data, and obtains structured multidimensional feature vectors based on preset preprocessing and feature extraction logic. The multidimensional feature vector is input into a pre-trained random forest prediction model to obtain the first predicted SOH value; Based on the preset triple threshold comparison mechanism, according to the first predicted SOH value, a graded early warning signal is output and a graded maintenance suggestion is generated; Based on the first predicted SOH value and the graded maintenance recommendations, a maintenance text recommendation is generated and a visual SOH decay curve is output. Based on the preset optimization cycle, the first measured SOH value is obtained, and the SOH deviation value is obtained by combining the first predicted SOH value. If the SOH deviation value exceeds the preset deviation threshold, the random forest prediction model and the triple threshold are updated based on the preset model iteration and threshold optimization mechanism.

[0047] It should be noted that the charge and discharge data includes real-time data streams and operating status data of the battery management system. The real-time data streams include at least the number of cycles, depth of charge, and charge / discharge rate. The operating status data includes at least the operating voltage, operating current, and operating internal resistance. The environmental data includes at least the battery temperature and battery humidity values. The SOH value is the battery health status value, which indicates the degree of degradation of the battery's current state relative to its initial state.

[0048] This embodiment provides a method for full lifecycle performance analysis based on swappable batteries. First, it systematically acquires multi-source heterogeneous data generated during battery operation. This data includes real-time charge / discharge cycle parameters provided by the battery management system and environmental sensor data such as temperature and humidity. After successfully acquiring the raw data, a pre-defined preprocessing and feature extraction logic is executed. This includes cleaning the raw data to eliminate noise interference, aligning the timestamps of the multi-source data for data fusion, and extracting key feature indicators that deeply reflect the battery's degradation state from the fused data. This yields a structured multi-dimensional feature vector. Subsequently, the multi-dimensional feature vector is input into a random forest prediction model pre-trained using a large amount of historical data. This model performs collaborative reasoning and calculation through multiple integrated decision trees, ultimately outputting a first predicted State of Health (SOH) value for the battery. Then, a pre-defined triple threshold comparison mechanism is activated. By comparing the obtained first predicted SOH value with pre-defined multi-level thresholds, different levels of urgency-based warning signals are output, and graded maintenance recommendations are generated accordingly. Finally, maintenance text is automatically generated based on the graded comparison prediction results and maintenance recommendations. Simultaneously, a visualized SOH decay curve integrating historical data and future trends is output for user decision-making reference. By obtaining the battery's actual measured SOH value and comparing it with the predicted value to obtain the SOH deviation value, if this deviation value exceeds the preset deviation threshold, it will trigger the model iteration and threshold optimization mechanism to update the parameters of the prediction model and adaptively adjust the specific values ​​of the three thresholds, thereby forming a closed-loop system for full life cycle performance analysis with self-learning capabilities.

[0049] According to an embodiment of the present invention, the process of obtaining a structured multidimensional feature vector based on a preset preprocessing and feature extraction logic specifically includes: Acquire raw data, including the charge / discharge data and environmental data; Based on the preset integrity verification and outlier removal mechanism, processed data is obtained from the original data; Based on a preset timestamp alignment and data fusion mechanism, aligned data is obtained according to the processed data; Based on the preset feature index extraction logic, a feature vector is obtained according to the aligned data; Based on the feature vector, a multidimensional feature vector is obtained using a preset structured operation.

[0050] It should be noted that the feature vector includes at least the average temperature within the cycle, the temperature fluctuation variance, the duration of the constant voltage charging phase, the capacity decay rate, and the internal resistance change rate. This embodiment provides a preprocessing and feature extraction logic, which is a multi-stage data refinement and feature extraction process. In this embodiment, firstly, raw data streams including voltage, current, internal resistance, temperature, humidity, cycle count, depth of charge, and charge / discharge rate are obtained from the underlying sensors and battery management system. Secondly, based on a preset integrity verification mechanism, the raw data is checked to identify and process missing values. Simultaneously, based on statistical methods or threshold rules, an outlier removal mechanism is activated to filter out abnormal data points that significantly deviate from the normal range, thereby obtaining processed data. Then, a preset timestamp alignment and data fusion mechanism is invoked to synchronize and interpolate the processed data from different sources, ensuring that all data points are fully aligned in the time dimension, forming an aligned dataset. Based on this, a preset feature index extraction logic is executed. Key feature indicators characterizing the battery performance degradation trajectory are calculated and extracted from the aligned data. These key feature indicators include, but are not limited to, the average temperature and temperature fluctuation variance within the cycle to reflect thermal management status, the duration of the constant voltage charging phase to assess charging behavior characteristics, and the capacity decay rate and internal resistance change rate to indicate health levels. Finally, all extracted feature indicators are normalized and vectorized to obtain a multi-dimensional feature vector, providing standardized input for subsequent machine learning models.

[0051] According to an embodiment of the present invention, the step of inputting the multidimensional feature vector into a pre-trained random forest prediction model to obtain a first predicted SOH value specifically includes: Based on historical datasets, the random forest algorithm is trained and cross-validated to generate the pre-trained random forest prediction model. The multidimensional feature vector is input into a pre-trained random forest prediction model, and the first predicted SOH value is output. Based on a preset time series analysis algorithm, the first predicted SOH value is serialized to generate an SOH value decay trend curve.

[0052] It should be noted that the historical dataset includes historical multidimensional feature vectors and their corresponding measured SOH values. In this embodiment, during the model training phase, a historical dataset with a large number of historical samples is used, where each sample consists of historical multidimensional feature vectors and measured SOH values. Using this dataset, a random forest algorithm is employed for model training, and cross-validation is used to evaluate model performance and prevent overfitting, ultimately generating a random forest prediction model. During the online prediction phase, the current multidimensional feature vector is input into this pre-trained model, and the first predicted SOH value is output. Furthermore, to achieve a trend-based understanding of future battery performance, a preset time series analysis algorithm is used to model and analyze the current and historical first predicted SOH values ​​as a time series. By fitting its decay trajectory, an SOH value decay trend curve is generated, thereby providing users with a forward-looking insight into the remaining usable life of the battery.

[0053] According to an embodiment of the present invention, the step of outputting a graded early warning signal and generating graded maintenance suggestions based on a preset triple threshold comparison mechanism and a first predicted SOH value specifically includes: Based on battery type and application scenario, three thresholds are obtained, including warning threshold, replacement threshold and emergency threshold; If the first predicted SOH value is higher than the preset warning threshold, monitoring will continue. If the first predicted SOH value is between the warning threshold and the replacement threshold, a first-level warning signal is output, and a suggestion to increase the detection frequency is generated. If the first predicted SOH value is between the replacement threshold and the emergency threshold, a secondary replacement signal is output and a battery replacement plan is generated. If the first predicted SOH value is lower than the emergency threshold, a level three emergency signal is output and a battery replacement command is generated.

[0054] It should be noted that this embodiment provides a triple threshold comparison and early warning mechanism, which is a precise decision-making process based on multi-level threshold values. In this embodiment, firstly, based on the specific chemical type of the battery, the application scenario, and the operation strategy, a set of three key thresholds are dynamically or statically set: an early warning threshold, a replacement threshold, and an emergency threshold. Then, the first predicted SOH value is sequentially compared and logically judged with these three thresholds. If the first predicted SOH value is higher than the early warning threshold, the battery is determined to be in a healthy state, and the normal monitoring mode is maintained without generating any early warning signal. If the first predicted SOH value drops to the range between the early warning threshold and the replacement threshold, a first-level early warning signal is output, indicating that the battery performance has shown signs of degradation, and a maintenance suggestion to increase the detection frequency is automatically generated to monitor its degradation trend. If the first predicted SOH value further drops to the range between the replacement threshold and the emergency threshold, a second-level replacement signal is output, indicating that the battery performance no longer meets the normal usage requirements, and a maintenance suggestion for a battery replacement plan is automatically formulated to initiate a planned replacement process. If the first predicted SOH value eventually falls below the emergency threshold, a level 3 emergency signal with the highest priority will be output, indicating that the battery has an extremely high risk of failure. An instruction will be generated to immediately stop using the battery and perform an emergency replacement, thereby ensuring safety and reliability.

[0055] According to an embodiment of the present invention, the step of generating maintenance text suggestions and outputting a visualized SOH decay curve based on the first predicted SOH value and graded maintenance suggestions specifically includes: Based on the hierarchical maintenance recommendations, a natural language description template is selected from the preset text template library; The maintenance text suggestion is generated by filling the selected natural language description template with the first predicted SOH value, battery identifier, and timestamp information; Based on the graphics processing component, an SOH decay curve is plotted according to the first predicted SOH value; Based on the aforementioned triple threshold, a reference line is marked on the SOH decay curve; According to the SOH value decay trend curve, add it to the tail of the SOH decay curve.

[0056] It should be noted that this embodiment provides a process for generating maintenance text suggestions and visual output, specifically a process for converting machine decisions into human-readable information. In this embodiment, a graded maintenance suggestion signal is first received, and a corresponding natural language description template is automatically selected from a pre-built text template library. The templates in the library are constructed using professional domain knowledge to ensure the professionalism and accuracy of the output suggestions. After selecting a template, the current first predicted SOH value, the unique identifier of the monitored battery, and the current timestamp information are used as variables and automatically filled into the reserved positions in the template. Through this data and text fusion, a semantically clear and directly executable maintenance text suggestion is generated. Simultaneously, an embedded graphics processing component is invoked to plot an SOH decay curve covering historical data points, with time as the horizontal axis and the estimated SOH value as the vertical axis. At the same time, horizontal reference lines representing warning thresholds, replacement thresholds, and emergency thresholds are drawn on the curve with distinct colors or line styles, thereby indicating the current battery state relative to the safety boundary. Finally, the generated future SOH value decay trend curve is added to the end of the current SOH decay curve in the form of a dashed line or different colors, realizing an integrated presentation of history, current status and future trends, thus providing users with an intuitive and information-rich decision dashboard.

[0057] According to an embodiment of the present invention, if the SOH deviation value exceeds a preset deviation threshold, the random forest prediction model and the triple threshold are updated based on a preset model iteration and threshold optimization mechanism, specifically including: When the SOH deviation value exceeds the preset deviation threshold, the model iteration and threshold optimization mechanism are triggered. The multidimensional feature vector and the first measured SOH value are incorporated into the historical dataset to perform incremental training or complete retraining of the random forest prediction model. Based on the SOH deviation value, the triple threshold is adjusted and updated according to the preset fine-tuning step. The updated model and thresholds are deployed to the online prediction system, and once the accuracy is verified, they are deployed locally.

[0058] It should be noted that this embodiment provides a model and threshold update mechanism. In this embodiment, by continuously monitoring the deviation between the predicted and measured SOH values, once the SOH deviation value exceeds the deviation threshold representing the tolerance for model performance degradation, the built-in model iteration and threshold optimization mechanism is activated. In the model iteration stage, the multidimensional feature vector that caused the deviation and the first measured SOH value are used as new training samples and incorporated into the historical dataset used for model training. Based on the expanded dataset, the original random forest prediction model is incrementally trained or completely retrained. This allows the model to learn the latest degradation patterns of battery performance from the new data, thereby refreshing its internal parameters and improving the accuracy of its future predictions. In the threshold optimization stage, based on the magnitude and direction of the calculated SOH deviation value and based on preset fine-tuning step rules, the specific values ​​of the three thresholds—warning, replacement, and emergency—are adaptively adjusted. For example, if the model continuously overestimates SOH, the threshold may be moderately increased to provide an early warning. After updating the model and thresholds, the updated model and thresholds are first deployed in an isolated online prediction sandbox environment. Only after the test and verification are passed will they be finally deployed to the online prediction system in the production environment to maintain the credibility of the system's analysis results.

[0059] It is worth mentioning that it also includes: After implementing the tiered maintenance recommendations, record the battery's subsequent operating status information and fault occurrence information; Based on the fault occurrence information, the subsequent operating status information is correlated with the executed maintenance recommendations to generate a maintenance effectiveness evaluation report; The maintenance effectiveness evaluation report is fed back to the model iteration and threshold optimization mechanism.

[0060] It should be noted that this embodiment provides a closed-loop feedback mechanism to verify and optimize the initial maintenance decisions through practical effects, thereby improving the overall intelligence level of the system. In this embodiment, after executing the graded maintenance recommendations, a long-term effect tracking phase is initiated. The subsequent operating status information of the battery is continuously recorded, including its charging and discharging behavior, environmental conditions, and detailed information on whether sudden failures occur. A correlation analysis process is initiated to analyze the causal relationship between the executed maintenance recommendations and the actual failure status and performance degradation rate of the battery; for example, if a battery recommended for "planned replacement" quickly fails, it indicates that the original warning may have been insufficient. Based on this deep correlation analysis, an evaluation report is automatically generated, assessing the effectiveness and accuracy of the previous warning decisions. Finally, the evaluation report is fed back to the aforementioned model iteration and threshold optimization mechanism, serving as an important auxiliary decision-making basis for optimizing the parameters of the random forest prediction model and adjusting the triple threshold values. This forms a continuous learning closed loop from decision-making to execution to feedback and ultimately to decision optimization, enhancing the practicality and reliability of the system.

[0061] It is worth mentioning that after generating the hierarchical maintenance recommendations, the following is also included: In response to a battery replacement plan or battery replacement instruction, a battery replacement work order is generated, which includes at least the replacement priority and battery identifier. The replacement work order is pushed to the work order pool queue of the operation and maintenance management system. Based on the work order pool queue, a maintenance work order is generated, which includes at least the location of the battery swapping station and the recommended replacement time. After the battery is replaced, its decommissioned status is marked based on the SOH value of the old battery.

[0062] It should be noted that in this embodiment, when the system generates a tiered maintenance suggestion containing a "battery replacement plan" or "battery replacement instruction," it automatically triggers a work order creation process. Based on the urgency of the suggestion, the geographical location of the battery swapping station, the battery model, and other information, a structured battery replacement work order is automatically generated. The work order includes at least a replacement priority for scheduling decisions and a battery identifier for accurate identification. Subsequently, the work order is pushed to the work order pool queue of the operation and maintenance management system through a preset interface, awaiting scheduling and allocation. The operation and maintenance management system then automatically assigns the work order to the most suitable operation and maintenance personnel based on the priority of all work orders in the work order pool, geographical location clustering, and the current location and workload of the operation and maintenance personnel, using a built-in scheduling algorithm. A detailed maintenance work order is generated, including the specific location of the battery swapping station, the recommended replacement time window, and other key execution information. After the operation and maintenance personnel arrive on-site and complete the battery replacement operation, they confirm the work order completion via their mobile terminal and determine whether the removed battery is decommissioned based on the measured SOH value. This not only completes the closed loop of a single maintenance task but also provides a complete record of the battery's entire lifecycle.

[0063] It is worth mentioning that it also includes: Obtain the real-time first predicted SOH value and its degradation trend curve of multiple batteries in the battery swapping station; Based on the SOH value, remaining capacity, historical cycle count, and current battery swapping service demand of each battery, a battery scheduling optimization objective function is constructed. Combining a pre-set reinforcement learning algorithm, the optimal battery allocation and scheduling strategy is output to match replacement work orders; The charging strategy and maintenance plan for batteries within the battery swapping station are dynamically adjusted based on the scheduling results.

[0064] It should be noted that in this embodiment, the first predicted State of Health (SOH) value and its future degradation trend of all batteries in the battery swapping station are monitored in real time, and the real-time remaining power, cumulative cycle count, and current battery swapping service request queue of each battery are simultaneously acquired. Based on this multi-source information, a scheduling optimization objective function is constructed with the goals of maximizing battery utilization efficiency, minimizing user waiting time, balancing battery degradation, and prioritizing high-demand services. Subsequently, a reinforcement learning algorithm based on deep networks or policy gradient methods is used to solve this objective function online or near-line. This algorithm learns through continuous interaction and trial and error with the battery swapping station environment, and finally outputs a set of optimal battery allocation strategies. The battery allocation strategy intelligently allocates batteries with excellent health and sufficient power to vehicles with strict range requirements or prioritizes urgent orders, while guiding batteries with average health but normal function to short-distance travel or vehicles that can accept slightly lower performance. This satisfies user needs while slowing down the degradation rate of high SOH batteries and fully utilizing the remaining value of low SOH batteries. At the same time, it will dynamically generate adaptive charging strategies and maintenance plans based on scheduling strategies and SOH prediction results, thereby realizing dynamic allocation of battery resources at the battery swapping station level and improving operational efficiency and the economic value of the battery throughout its entire life cycle.

[0065] It is worth mentioning that it also includes: It aggregates charging and discharging data, environmental data, and SOH decay records of batteries of the same batch or model from multiple battery swapping stations; Based on the aggregated multi-station data, the pre-trained random forest prediction model is trained using federated learning to generate a globally optimized prediction model.

[0066] It should be noted that in this embodiment, a data sharing platform covering multiple battery swapping stations is constructed. This platform collects full lifecycle operational data of batteries of the same batch or model from each station through standardized interfaces, including but not limited to charge and discharge cycle parameters, environmental temperature and humidity change history, historical SOH monitoring values, and final decommissioning records, thereby forming a larger and more diverse enhanced dataset. Without directly sharing the original data, the global initial model is distributedly trained using local data from each station. By aggregating the parameter updates of each local model, a more generalizable globally optimized random forest prediction model is iteratively generated. This global model is then distributed and deployed to the local analysis systems of each battery swapping station to replace or assist the original local prediction models, especially when facing scarce local data or a lack of historical data for new batches of batteries, it can still make accurate predictions.

[0067] A third aspect of the present invention provides a computer-readable storage medium comprising a method program for full life cycle performance analysis based on a battery swapping system. When the method program is executed by a processor, it implements the steps of the full life cycle performance analysis method for a battery swapping system as described in any of the preceding claims.

[0068] In summary, this invention provides a method, system, and medium for full lifecycle performance analysis of swappable batteries. First, it comprehensively collects and preprocesses battery charging / discharging data and environmental data. A structured multidimensional feature vector is constructed through feature extraction and input into a pre-trained random forest prediction model to calculate the first predicted SOH value. Then, based on a preset triple threshold comparison mechanism, the predicted value is graded and judged, generating different levels of early warning signals and maintenance suggestions, and automatically outputting a visualized SOH decay curve and text report. Finally, the model performance is monitored periodically by comparing the deviation between the measured SOH value and the predicted value, and a model iteration and threshold dynamic optimization mechanism is triggered when the deviation exceeds the limit. This invention improves the accuracy and reliability of battery health status assessment by combining multi-source data fusion and machine learning prediction; the introduced multi-level early warning and dynamic feedback mechanism avoids sudden battery failures.

[0069] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0070] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for full life-cycle performance analysis based on swappable batteries, characterized in that, The method includes: The system acquires battery charging and discharging data and environmental data, and obtains structured multidimensional feature vectors based on preset preprocessing and feature extraction logic. The multidimensional feature vector is input into a pre-trained random forest prediction model to obtain the first predicted SOH value; Based on the preset triple threshold comparison mechanism, according to the first predicted SOH value, a graded early warning signal is output and a graded maintenance suggestion is generated; Based on the first predicted SOH value and the graded maintenance recommendations, a maintenance text recommendation is generated and a visual SOH decay curve is output. Based on the preset optimization cycle, the first measured SOH value is obtained, and the SOH deviation value is obtained by combining the first predicted SOH value. If the SOH deviation value exceeds the preset deviation threshold, the random forest prediction model and the triple threshold are updated based on the preset model iteration and threshold optimization mechanism.

2. The method for full life-cycle performance analysis based on swappable batteries according to claim 1, characterized in that, The preprocessing and feature extraction logic based on a preset method yields a structured multidimensional feature vector, specifically including: Acquire raw data, including the charge / discharge data and environmental data; Based on the preset integrity verification and outlier removal mechanism, processed data is obtained from the original data; Based on a preset timestamp alignment and data fusion mechanism, aligned data is obtained according to the processed data; Based on the preset feature index extraction logic, a feature vector is obtained according to the aligned data; Based on the feature vector, a multidimensional feature vector is obtained using a preset structured operation.

3. The method for full life-cycle performance analysis based on swappable batteries according to claim 1, characterized in that, The step of inputting the multidimensional feature vector into a pre-trained random forest prediction model to obtain the first predicted SOH value specifically includes: Based on historical datasets, the random forest algorithm is trained and cross-validated to generate the pre-trained random forest prediction model. The multidimensional feature vector is input into a pre-trained random forest prediction model, and the first predicted SOH value is output. Based on a preset time series analysis algorithm, the first predicted SOH value is serialized to generate an SOH value decay trend curve.

4. The method for full life-cycle performance analysis based on swappable batteries according to claim 1, characterized in that, The preset triple threshold comparison mechanism outputs a graded early warning signal and generates graded maintenance suggestions based on the first predicted SOH value, specifically including: Based on battery type and application scenario, three thresholds are obtained, including warning threshold, replacement threshold and emergency threshold; If the first predicted SOH value is higher than the preset warning threshold, monitoring will continue. If the first predicted SOH value is between the warning threshold and the replacement threshold, a first-level warning signal is output, and a suggestion to increase the detection frequency is generated. If the first predicted SOH value is between the replacement threshold and the emergency threshold, a secondary replacement signal is output and a battery replacement plan is generated. If the first predicted SOH value is lower than the emergency threshold, a level 3 emergency signal is output and a battery replacement command is generated.

5. The method for full life-cycle performance analysis based on swappable batteries according to claim 3, characterized in that, The step of generating maintenance text suggestions and outputting a visualized SOH decay curve based on the first predicted SOH value and graded maintenance suggestions specifically includes: Based on the hierarchical maintenance recommendations, a natural language description template is selected from the preset text template library; The maintenance text suggestion is generated by filling the selected natural language description template with the first predicted SOH value, battery identifier, and timestamp information; Based on the graphics processing component, an SOH decay curve is plotted according to the first predicted SOH value; Based on the aforementioned triple threshold, a reference line is marked on the SOH decay curve; According to the SOH value decay trend curve, add it to the tail of the SOH decay curve.

6. The method for full life-cycle performance analysis based on swappable batteries according to claim 1, characterized in that, If the SOH deviation value exceeds a preset deviation threshold, the random forest prediction model and the triple threshold are updated based on a preset model iteration and threshold optimization mechanism, specifically including: When the SOH deviation value exceeds the preset deviation threshold, the model iteration and threshold optimization mechanism are triggered. The multidimensional feature vector and the first measured SOH value are incorporated into the historical dataset to perform incremental training or complete retraining of the random forest prediction model. Based on the SOH deviation value, the triple threshold is adjusted and updated according to the preset fine-tuning step. The updated model and thresholds are deployed to the online prediction system, and once the accuracy is verified, they are deployed locally.

7. A full life-cycle performance analysis system based on swappable batteries, characterized in that, The system includes a memory and a processor. The memory includes a program for a full life-cycle performance analysis method based on a battery swapping system. When the processor executes the program, the full life-cycle performance analysis method based on the battery swapping system performs the following steps: The system acquires battery charging and discharging data and environmental data, and obtains structured multidimensional feature vectors based on preset preprocessing and feature extraction logic. The multidimensional feature vector is input into a pre-trained random forest prediction model to obtain the first predicted SOH value; Based on the preset triple threshold comparison mechanism, according to the first predicted SOH value, a graded early warning signal is output and a graded maintenance suggestion is generated; Based on the first predicted SOH value and the graded maintenance recommendations, a maintenance text recommendation is generated and a visual SOH decay curve is output. Based on the preset optimization cycle, the first measured SOH value is obtained, and the SOH deviation value is obtained by combining the first predicted SOH value. If the SOH deviation value exceeds the preset deviation threshold, the random forest prediction model and the triple threshold are updated based on the preset model iteration and threshold optimization mechanism.

8. The full life-cycle performance analysis system based on a battery swapping system according to claim 7, characterized in that, The preprocessing and feature extraction logic based on a preset method yields a structured multidimensional feature vector, specifically including: Acquire raw data, including the charge / discharge data and environmental data; Based on the preset integrity verification and outlier removal mechanism, processed data is obtained from the original data; Based on a preset timestamp alignment and data fusion mechanism, aligned data is obtained according to the processed data; Based on the preset feature index extraction logic, a feature vector is obtained according to the aligned data; Based on the feature vector, a multidimensional feature vector is obtained using a preset structured operation.

9. The full life-cycle performance analysis system based on a battery swapping system according to claim 7, characterized in that, The step of inputting the multidimensional feature vector into a pre-trained random forest prediction model to obtain the first predicted SOH value specifically includes: Based on historical datasets, the random forest algorithm is trained and cross-validated to generate the pre-trained random forest prediction model. The multidimensional feature vector is input into a pre-trained random forest prediction model, and the first predicted SOH value is output. Based on a preset time series analysis algorithm, the first predicted SOH value is serialized to generate an SOH value decay trend curve.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The computer-readable storage medium includes a method program for full life cycle performance analysis based on a battery swapping system. When the method program is executed by a processor, it implements the steps of the method for full life cycle performance analysis based on a battery swapping system as described in any one of claims 1 to 6.

Images