Digital twin-driven dynamic mapping and protection method for charging risks of electric two-wheelers

Abstract

This invention provides a method for dynamic mapping and protection against risks during extreme charging of electric two-wheelers using digital twin technology, belonging to the field of optimization control technology. The method includes: collecting extreme charging-related data and vehicle system data of the electric two-wheeler, and constructing a digital twin extreme charging model by combining the microscopic characteristics of the battery cells. A real-time correction mechanism is established to correct errors in the digital twin extreme charging model, resulting in a corrected extreme charging model. Early runaway characteristics are captured using a multi-scale feature fusion method to generate risk mapping data. Based on the risk mapping data and using a rough set-grey relational coupling method, a hierarchical protection strategy is evaluated and generated. The optimal protection path is dynamically selected, parsed into an execution instruction set, and simulated in the corrected extreme charging model to correct the hierarchical protection strategy. This invention can dynamically adjust model parameters and protection strategies according to different extreme charging scenarios and data changes, enhancing the model's adaptability to different situations and effectively reducing the safety risks of electric two-wheelers during extreme charging.

Description

Technical Field

[0001] This invention relates to the field of optimization control technology, and in particular to a method for dynamic mapping and protection against extreme charging risks in digital twin-driven electric two-wheeled vehicles. Background Technology

[0002] In recent years, electric two-wheelers have rapidly gained popularity due to their convenience and environmental friendliness. However, safety issues during fast charging have become increasingly prominent. Traditional fast charging technology lacks a deep understanding of the internal micro-state of the battery and its coordination with the entire vehicle system. This often leads to thermal runaway caused by the uncontrolled micro-characteristics of the battery cells, and can even result in fires and explosions. Furthermore, various data sources are scattered and isolated, resulting in low sampling accuracy and difficulty in time sequence alignment, making it difficult to accurately reflect the dynamic process of fast charging.

[0003] Existing risk control methods mainly rely on empirical thresholds or single models, which have significant drawbacks: First, they cannot capture the advance impact of micro-level characteristics on macro-level risks, and early signs of loss of control are easily overlooked; second, protection strategies rely on static rules and do not combine real-time data to dynamically assess risk levels, resulting in a lack of scientific rigor and adaptability in the selection of protection paths; third, after long-term model operation, errors accumulate, and the lack of real-time correction mechanisms leads to increased risk prediction deviations and the failure of protection measures. With the development of digital twin technology, multi-scale data fusion and real-time simulation can be achieved by constructing digital mapping models of physical entities. However, how to dynamically correct model errors through real-time data interaction remains an industry challenge. Furthermore, there is a lack of systematic solutions for how to scientifically evaluate graded protection strategies based on risk mapping data, dynamically select the optimal path, and ensure the effectiveness of implementation.

[0004] Against this backdrop, there is an urgent need for a digital twin-driven method for dynamic mapping and protection against extreme charging risks in electric two-wheelers, in order to address the core defects of existing extreme charging safety technologies. Summary of the Invention

[0005] This invention provides a method for dynamic mapping and protection of charging risks in electric two-wheeled vehicles driven by digital twins, in order to overcome the deficiencies in the prior art.

[0006] This invention provides a method for dynamic mapping and protection against charging risks in electric two-wheeled vehicles driven by digital twins, including:

[0007] Collect data related to the electric two-wheeler's charging system and the vehicle's overall system data, and construct a digital twin charging model by combining the microscopic characteristics of the battery cells.

[0008] A real-time correction mechanism is established through Lyapunov incremental learning to correct errors in the digital twin extreme-filled model, resulting in a corrected extreme-filled model. Early runaway features are captured through a multi-scale feature fusion method to generate risk mapping data.

[0009] Based on risk mapping data and using the rough set-grey relational coupling method, a graded protection strategy is generated, and the optimal protection path is dynamically selected.

[0010] The optimal protection path is parsed into an execution instruction set, and its execution is simulated in the modified extreme charging model to modify the hierarchical protection strategy.

[0011] This invention provides a method for dynamic mapping and protection against charging risks in electric two-wheeled vehicles driven by digital twins. The steps for obtaining charging-related data and vehicle system data include:

[0012] The characteristics of the extreme charging process, charging pile parameters, and environmental data are used as multi-mode extreme charging data, while the battery pack structural parameters, BMS status data, vehicle load characteristics, and motor controller status are used as vehicle data.

[0013] The sensor type, communication protocol, sampling frequency, and accuracy are determined based on multi-mode extreme charging data and vehicle data.

[0014] Based on the sampling frequency and accuracy triggered during the extreme charging stage, enhanced collection of risk characteristics is triggered according to the anomaly type, and vehicle data and multi-mode extreme charging data are aligned according to the timestamp.

[0015] This invention provides a method for dynamic mapping and protection against extreme charging risks in electric two-wheeled vehicles driven by digital twins. The steps for constructing a digital twin extreme charging model include:

[0016] Lithium-ion migration rate, SEI film impedance, and lithium dendrite length are collected as microscopic characteristics of the battery cell. Combined with extreme charging correlation data and vehicle system data, extreme charging data is obtained by standardization according to micro-meso-macro.

[0017] A correlation dictionary was constructed by extracting lithium dendrite growth rate, SEI film rupture voltage threshold, cell voltage difference, temperature field gradient, charging curve slope, and vehicle energy consumption feedback coefficient from the extreme charging data.

[0018] Based on the association dictionary, a microscopic model of the battery cell, a mesoscopic model of the battery pack, and a macroscopic model of the whole vehicle and charging pile are constructed.

[0019] High-dimensional micro-features are compressed into a low-dimensional space that matches meso-level features using an adaptive dimensionality reduction algorithm, and then a nonlinear mapping with macro-level features is established using a fully connected neural network.

[0020] By utilizing long short-term memory networks to learn the advanced impact of micro-feature changes on macro-risks, a temporal correlation mechanism is established. A digital twin extreme charging model is obtained by fusing cell micro-models, battery pack meso-models, and vehicle-charging pile macro-models using data interfaces and semantic layer architecture.

[0021] This invention provides a method for dynamic mapping and protection against charging risks in electric two-wheeled vehicles driven by digital twins. The steps for establishing a real-time correction mechanism include:

[0022] For the three-level structure in the digital twin extreme-filling model, a multi-scale error index is defined, a weighted fusion algorithm is used to generate a comprehensive error vector, and weights are dynamically allocated based on data credibility.

[0023] Using the comprehensive error vector as input, a positive definite function is constructed as a stability criterion. The first derivative of the positive definite function is calculated to determine whether it is less than 0. Otherwise, the error diverges and a correction mechanism is triggered.

[0024] This invention provides a method for dynamic mapping and protection against extreme charging risks in digital twin-driven electric two-wheeled vehicles. The steps for obtaining the corrected extreme charging model include:

[0025] Based on real-time data interaction between the digital twin extreme charging model and the physical system, predicted and measured values ​​are collected synchronously, and the contribution analysis method is used to locate the dominant error source.

[0026] Based on the analysis of the dominant error sources, the influence parameters of the current comprehensive error vector on the digital twin extreme-filling model are formed to create a set of parameters to be corrected.

[0027] With the goal of minimizing the positive definite function, the correction parameters are obtained by updating the set of parameters to be corrected using the incremental gradient descent algorithm.

[0028] The modified extreme-charge model is obtained by adjusting the digital twin extreme-charge model according to the correction parameters.

[0029] This invention provides a method for dynamic mapping and protection against charging risks in electric two-wheeled vehicles driven by digital twins. The steps for generating risk mapping data include:

[0030] Early runaway characteristics are defined according to the three-level scale of micro-meso-macro, and noise filtering, outlier processing, standardization, and spatiotemporal alignment are performed.

[0031] Based on the importance of early loss of control features and the dynamic adjustment of feature weights at risk stages, an attention mechanism and cross-scale correlation network are used to design a fusion model. The mutual information value between the fusion feature and the early loss of control feature is calculated, and those that reach a preset threshold are retained as global feature vectors.

[0032] Historical risk data is collected, and a gradient boosting tree model trained based on the historical risk data is constructed as a risk identification model. The global feature vector is input, and the multi-dimensional risk probability is output.

[0033] Based on multi-dimensional risk probabilities and spatial alignment information, a risk spatial distribution heatmap is generated, and risk mapping data is obtained by predicting risk evolution trends based on the temporal changes of fused features.

[0034] This invention provides a method for dynamic mapping and protection against charging risks of electric two-wheeled vehicles driven by digital twins. The steps for evaluating and generating a graded protection strategy include:

[0035] Risk mapping data is transformed into discrete attributes by equal-frequency discretization, and a decision information table is constructed by using historical risk samples as rows and conditional attributes and decision attributes as columns.

[0036] Calculate the dependence of conditional attributes on decision attributes, identify key influencing features, and generate decision rules.

[0037] The ideal feature vector of the decision rule is used as the reference sequence, and the current extreme risk mapping data is used as the comparison sequence.

[0038] Calculate the single-index similarity between the comparison sequence and the reference sequence of each decision rule.

[0039] The overall correlation degree is obtained by weighted summation based on dependence and single-indicator similarity.

[0040] Based on the safety requirements of the extreme charging scenario, a correlation threshold is set, and a graded protection strategy is obtained by matching the corresponding level protection strategy according to the comprehensive correlation.

[0041] This invention provides a digital twin-driven method for dynamic mapping and protection against charging risks in electric two-wheeled vehicles. The steps for obtaining the optimal protection path include:

[0042] Ranking indicators are set based on comprehensive relevance, compatibility of protection strategies, confidence of decision rules, and balance of multiple objectives.

[0043] The comprehensive ranking score of each candidate path in the hierarchical protection strategy is calculated based on the ranking index, and the candidate paths are arranged in descending order of the score to obtain the path ranking table.

[0044] Based on the current hardware status of the Extreme Charge system and the dual verification path ranking table of the current Extreme Charge scenario, the candidate path that has passed the verification is selected as the optimal protection path.

[0045] This invention provides a method for dynamic mapping and protection against charging risks in electric two-wheeled vehicles driven by digital twins. The steps for parsing and obtaining the execution instruction set include:

[0046] Determine the combination of measures based on the optimal protection path, break it down into basic actions, and label the implementing entities.

[0047] For each basic action, the execution details are determined by combining risk mapping data with the hardware parameters of the Extreme Charge system.

[0048] Develop priority rules for the charging scenario, sort out the dependencies between the execution details of different basic actions, and determine the start time, execution duration and completion flag of each basic action to generate a timing execution table.

[0049] The timing execution table is verified based on action conflicts, hardware boundaries, battery safety, and communication protocols. According to the unified instruction encoding rules of the Extreme Charge system, a unique instruction ID is assigned to each basic action to obtain the execution instruction set.

[0050] This invention provides a method for dynamic mapping and protection against charging risks in electric two-wheeled vehicles driven by digital twins. The steps for modifying the graded protection strategy include:

[0051] The timing logic and parameter details of the execution instruction set are analyzed, and the instruction format is converted according to the interface protocol of the modified extreme charge model.

[0052] The instructions are executed sequentially and simulated in the modified extreme charging model. Risk-related indicators, execution effect indicators, and battery status indicators are recorded as simulation results.

[0053] By comparing the expected results of the instructions with the simulation results, the deviations of risk links, parameter execution deviations, and timing execution deviations are calculated as deviation indicators.

[0054] Determine whether the deviation index is within the preset deviation threshold. If it is, no correction is needed. Otherwise, locate the root cause of the deviation and correct the graded protection strategy according to the type of deviation root cause until it meets the preset deviation threshold.

[0055] This invention provides a digital twin-driven method for dynamic mapping and protection of risks during the extreme charging of electric two-wheelers. By constructing a digital twin extreme charging model, it monitors and assesses risks in real time during the extreme charging process, enabling timely detection of early runaway characteristics, generating risk mapping data, and formulating and implementing graded protection strategies, effectively reducing the safety risks of electric two-wheelers during extreme charging. A real-time correction mechanism is established using Lyapunov incremental learning, continuously correcting errors in the digital twin extreme charging model and improving its accuracy and reliability. Simultaneously, this method can dynamically adjust model parameters and protection strategies according to different extreme charging scenarios and data changes, enhancing the model's adaptability to various situations. By capturing early runaway characteristics through multi-scale feature fusion and combining this with a rough set-grey relational coupling method to evaluate and generate graded protection strategies, and dynamically selecting the optimal protection path, it achieves accurate risk assessment and protection, improving the effectiveness of protection measures and reducing unnecessary waste of protection resources. Through simulated execution and correction of the graded protection strategies, and optimization of the execution instruction set, it ensures the effective execution of the protection strategies during actual extreme charging, optimizing the extreme charging process, improving charging efficiency, and extending battery life. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0057] Figure 1 This is one of the flowcharts illustrating the dynamic mapping and protection method for charging risks of electric two-wheeled vehicles driven by digital twins provided in this embodiment of the invention.

[0058] Figure 2 This is the second flowchart of the method for dynamic mapping and protection of charging risks of electric two-wheeled vehicles driven by digital twins provided in the embodiments of the present invention;

[0059] Figure 3 This is the third flowchart of the method for dynamic mapping and protection of charging risks of digital twin-driven electric two-wheeled vehicles provided in the embodiments of the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0061] The following is combined Figures 1-3 This invention describes a digital twin-driven method for dynamic mapping and protection against charging risks in electric two-wheeled vehicles.

[0062] like Figure 1 As shown in the embodiment of the present invention, the method for dynamic mapping and protection against charging risks of electric two-wheeled vehicles driven by digital twins includes:

[0063] Collect data related to the electric two-wheeler's charging system and the vehicle's overall system data, and construct a digital twin charging model by combining the microscopic characteristics of the battery cells.

[0064] The steps to obtain the data related to the charging station and the vehicle system include:

[0065] The characteristics of the extreme charging process, charging pile parameters, and environmental data are used as multi-mode extreme charging data, while the battery pack structural parameters, BMS status data, vehicle load characteristics, and motor controller status are used as vehicle data.

[0066] The characteristics of the extreme charging process may include micro-current fluctuations, CO / H2 gas concentration, and cell expansion. Charging pile parameters may include output current / voltage and power step changes. Environmental data may include temperature, humidity, and air pressure. In extreme environments, vibration and electromagnetic interference intensity monitoring will be added.

[0067] The sensor type, communication protocol, sampling frequency, and accuracy are determined based on multi-mode extreme charging data and vehicle data.

[0068] The battery pack is equipped with a miniature current sensor, a gas sensor array, and a distributed temperature sensor. The charging station is fitted with a high-frequency voltage / current sensor, and the vehicle integrates BeiDou positioning and attitude sensors. Temperature sensors are positioned close to the cell surface, avoiding electromagnetic interference sources, and gas sensors have pre-drilled ventilation holes, meeting electrical safety clearance requirements. Calibration is performed using a standard signal source before installation, and the system is re-inspected every 3 months under extreme charging conditions; immediate calibration is performed if the error exceeds ±5%.

[0069] Based on the sampling frequency and accuracy triggered during the extreme charging stage, enhanced collection of risk characteristics is triggered according to the anomaly type, and vehicle data and multi-mode extreme charging data are aligned according to the timestamp.

[0070] The extreme charging phase may include: pre-charging phase: collecting initial SOC of the battery, individual cell voltage consistency, and ambient temperature to establish baseline data.

[0071] Constant current charging stage: High-frequency acquisition of voltage dynamic distortion and temperature field gradient under high current, and simultaneous capture of micro-current pulses and gas concentration change rate.

[0072] Constant pressure / trickle flow stage: Reduce the frequency of conventional parameter acquisition, retain gas concentration and cell expansion monitoring, and capture data on "tail heat" phenomenon.

[0073] When a voltage fluctuation >20% or a temperature rise rate >2℃ / min is detected, the corresponding parameter acquisition frequency will be automatically increased by 3 to 5 times.

[0074] When thermal runaway precursors are triggered, high-frequency acquisition of all parameters and data caching are initiated.

[0075] The steps to construct a digital twin extreme charge model include:

[0076] Lithium-ion migration rate, SEI film impedance, and lithium dendrite length are collected as microscopic characteristics of the battery cell. Combined with extreme charging correlation data and vehicle system data, extreme charging data is obtained by standardization according to micro-meso-macro.

[0077] A correlation dictionary was constructed by extracting lithium dendrite growth rate, SEI film rupture voltage threshold, cell voltage difference, temperature field gradient, charging curve slope, and vehicle energy consumption feedback coefficient from the extreme charging data.

[0078] Based on the association dictionary, a microscopic model of the battery cell, a mesoscopic model of the battery pack, and a macroscopic model of the whole vehicle and charging pile are constructed.

[0079] The steps for constructing a microscopic model of a battery cell may include: constructing a coupled model of lithium-ion migration and lithium dendrite growth based on molecular dynamics and electrochemical theories.

[0080] Based on the Newman equation, a concentration polarization correction term under extremely high current is introduced to simulate the migration rate of lithium ions between the positive and negative electrodes.

[0081] By combining the phase-field model, the risk of SEI film rupture is dynamically calculated using microscopic data such as lithium dendrite length and growth rate.

[0082] Model inputs: electrode charging current density (mA / cm²), electrolyte concentration (mol / L), cell temperature (°C).

[0083] Output: Lithium ion concentration distribution, lithium dendrite growth state, and microscopic internal resistance change.

[0084] The steps for constructing a mesoscopic model of a battery pack may include: constructing an electro-thermal-mechanical multiphysics coupling model.

[0085] The electrical components are based on an equivalent circuit model, which integrates data such as individual cell voltage differences and loop impedance to simulate the voltage / current distribution during the charging and discharging process of the battery pack.

[0086] The thermal section combines Fourier's law and Joule's law, introducing a contact thermal resistance correction term to simulate dynamic changes in the temperature field. The mechanical section is based on Hooke's law, relating the nonlinear relationship between cell expansion and internal pressure.

[0087] Model inputs: Internal resistance increment, extreme charging power, and cooling system parameters from the microscopic model output. Outputs: Battery pack temperature field distribution, individual cell voltage consistency, and structural stress distribution.

[0088] The steps to construct a macroscopic model of the whole vehicle and charging pile may include: integrating the output of the battery pack mesoscopic model, the load characteristics of the whole vehicle, and the parameters of the charging pile, and simulating the energy flow and information flow of "charging pile-battery pack-whole vehicle circuit".

[0089] Characterize the interaction process of the Extreme Charge protocol: simulate the handshake signal and power negotiation logic under the GB / T36944 protocol, and output macroscopic state parameters such as charging time prediction and vehicle energy consumption feedback.

[0090] High-dimensional micro-features are compressed into a low-dimensional space that matches meso-level features using an adaptive dimensionality reduction algorithm, and then a nonlinear mapping with macro-level features is established using a fully connected neural network.

[0091] By utilizing long short-term memory networks to learn the advanced impact of micro-feature changes on macro-risks, a temporal correlation mechanism is established. A digital twin extreme charging model is obtained by fusing cell micro-models, battery pack meso-models, and vehicle-charging pile macro-models using data interfaces and semantic layer architecture.

[0092] The cell micro-model transmits parameters to the battery pack meso-model via a standardized data interface. The battery pack meso-model injects data such as temperature field and voltage distribution into the vehicle-charging pile macro-model. Parameters such as vehicle load and charging strategy fed back from the vehicle-charging pile macro-model inversely adjust the boundary conditions of the cell micro-model and the battery pack meso-model. For example, when the charging current is dynamically adjusted, the micro-model updates the lithium-ion migration rate in real time.

[0093] A real-time correction mechanism is established through Lyapunov incremental learning to correct errors in the digital twin extreme-filled model, resulting in a corrected extreme-filled model. Early runaway features are captured through a multi-scale feature fusion method to generate risk mapping data.

[0094] The steps to establish a real-time correction mechanism include:

[0095] For the three-level structure in the digital twin extreme-filling model, a multi-scale error index is defined, a weighted fusion algorithm is used to generate a comprehensive error vector, and weights are dynamically allocated based on data credibility.

[0096] Using the comprehensive error vector as input, a positive definite function is constructed as a stability criterion. The first derivative of the positive definite function is calculated, and it is determined whether it is less than 0; otherwise, error divergence triggers a correction mechanism. The formula for the positive definite function is as follows:

[0097]

[0098] In the formula, It is the comprehensive error vector. It is a positive definite matrix. It is a positive definite function. It is the transpose of the comprehensive error vector.

[0099] The steps to obtain the modified extreme-charge model include:

[0100] Based on real-time data interaction between the digital twin extreme charging model and the physical system, predicted and measured values ​​are collected synchronously, and the contribution analysis method is used to locate the dominant error source.

[0101] Microscopic layer: Compare the lithium dendrite growth length and lithium ion migration rate predicted by the model with the in-situ detection data to calculate the microscopic error.

[0102] Mesoscopic layer: Compare the battery pack temperature field distribution and individual cell voltage difference output by the model with the measured values ​​of the distributed sensors to calculate the mesoscopic error.

[0103] Macro level: Compare the charging current and power predicted by the model with the actual values ​​fed back by the charging pile / BMS to calculate the macro error.

[0104] The contribution rate of each error term to the overall error vector is calculated. .

[0105] If n1 > 50%, it is determined to be a mismatch of microscopic model parameters, such as a deviation in the lithium-ion diffusion coefficient. If n2 > 50%, it is determined to be an error in the mesoscopic multiphysics coupling model, such as a deviation in the thermal conductivity setting. If n3 > 50%, it is determined to be an error in the macroscopic system interaction model, such as a deviation in the charging protocol response delay parameter.

[0106] Based on the analysis of the dominant error sources, the influence parameters of the current comprehensive error vector on the digital twin extreme-filling model are formed to create a set of parameters to be corrected.

[0107] If microscopic errors are the dominant factor, select microscopic parameters such as lithium-ion diffusion coefficient, SEI film impedance, and lithium dendrite growth rate coefficient to form a set.

[0108] If the error is mainly at the mesoscopic level, select mesoscopic parameters such as thermal conductivity, contact thermal resistance, and voltage equalization coefficient to form a set.

[0109] If macroscopic errors dominate: select macroscopic parameters such as power loss coefficient, charging protocol handshake delay, and load feedback coefficient to form a set.

[0110] With the objective of minimizing the positive definite function, the correction parameters are obtained by updating the set of parameters to be corrected using the incremental gradient descent algorithm, as expressed by the following formula:

[0111]

[0112] In the formula, It is the number of iterations. It's the learning rate. It is the set of parameters to be corrected. It is the first The combined error vector of the steps, It is a parameter correction. It is the first The set of parameters to be corrected during each iteration.

[0113] The modified extreme-charge model is obtained by adjusting the digital twin extreme-charge model according to the correction parameters.

[0114] like Figure 2 As shown, the steps for generating risk mapping data include:

[0115] Early runaway characteristics are defined according to the three-level scale of micro-meso-macro, and noise filtering, outlier processing, standardization, and spatiotemporal alignment are performed.

[0116] Microscale: Reflects the earliest electrochemical and physical changes in thermal runaway, which may include: Microcurrent characteristics: Abnormal pulse current generated by lithium dendrites piercing the SEI film.

[0117] Gas release characteristics: characteristic gas concentrations and release rates in the early stages of thermal runaway.

[0118] Microstructural characteristics: cell expansion and electrode microcrack density.

[0119] Mesoscale: Reflects the process of local anomalies spreading to the module, which may include: temperature field characteristics: sudden increase in temperature difference of individual cells, local hot spot temperature.

[0120] Voltage consistency characteristics: sudden increase in individual unit voltage difference, abnormal drop in voltage plateau.

[0121] Impedance characteristics: The real part of the AC impedance of the battery pack increases abruptly.

[0122] Macro-scale: Reflects system-level abnormal behavior, which may include: charging curve characteristics: current fluctuation amplitude, voltage-time curve slope abrupt change.

[0123] Energy consumption characteristics: A sudden increase in energy consumption per unit of SOC.

[0124] Environmental interaction characteristics: communication timeout between charging pile and BMS, sudden increase in cooling system load.

[0125] Microscopic features: Wavelet threshold noise reduction is applied to micro-current pulses to filter out magnetic interference from electrode charging. Gas concentration data is filtered using a sliding window mean filter to eliminate sensor drift.

[0126] Mesoscopic features: Temperature field data is corrected using Kalman filtering to eliminate abnormal jumps in local temperature measurement points. Voltage difference data uses the Laida criterion to remove instantaneous interference values.

[0127] Macroscopic features: Charging curve data is smoothed using Savitzky-Golay filtering to preserve trend changes. Discrete features such as communication timeouts are imputed to remove missing values ​​using time series data.

[0128] Z-score standardization is employed to map features of different magnitudes to the same scale, avoiding the impact of feature magnitude differences on fusion weights. Interpolation is performed on meso- and macro-level features based on the highest sampling frequency of micro-features to achieve time alignment. Spatial alignment is achieved by associating micro-features with meso- / macro-level spatial locations through the mapping relationship between cell number, module location, and vehicle coordinate system.

[0129] Based on the importance of early loss of control features and the dynamic adjustment of feature weights at risk stages, an attention mechanism and cross-scale correlation network are used to design a fusion model. The mutual information value between the fusion feature and the early loss of control feature is calculated, and those that reach a preset threshold are retained as global feature vectors.

[0130] Microscopic features are encoded using a Long Short-Term Memory (LSTM) network to represent temporal features, mesoscopic features using a CNN to represent spatial features, and macroscopic features using a fully connected network to represent system features, outputting feature vectors at each scale. A multi-head attention mechanism is introduced to calculate the correlation weights between feature vectors at different scales, such as the correlation between microscopic gas features and mesoscopic temperature features. These weighted features are then fused into a global feature vector, highlighting the causal relationship between "microscopic precursors and mesoscopic responses," such as the strong correlation between rising CO concentration and local temperature increases.

[0131] By using a gating mechanism to filter out unrelated features, the consistency of the physical meaning of the fused features is ensured.

[0132] Historical risk data is collected, and a gradient boosting tree model trained based on the historical risk data is constructed as a risk identification model. The global feature vector is input, and the multi-dimensional risk probability is output.

[0133] Multidimensional risk probabilities can include: thermal runaway probability, overcharge risk probability, insulation failure probability, etc.

[0134] Based on multi-dimensional risk probabilities and spatial alignment information, a risk spatial distribution heatmap is generated, and risk mapping data is obtained by predicting risk evolution trends based on the temporal changes of fused features.

[0135] The risk spatial distribution heatmap can include: microscopic layer: marking the location of high-risk cells, such as cell A3 where the lithium dendrite growth rate exceeds the standard, with a risk value of 0.8.

[0136] Mesoscopic level: Draw a heat map of the temperature field risk of the battery pack, such as red areas representing hot spot risk ≥0.7.

[0137] Macro level: Mark the risk nodes of the whole vehicle charging system, such as the risk value of the charging pile power module is 0.6.

[0138] Based on risk mapping data and using the rough set-grey relational coupling method, a graded protection strategy is generated, and the optimal protection path is dynamically selected.

[0139] like Figure 3 As shown, the steps for evaluating and generating a tiered protection strategy include:

[0140] Risk mapping data is transformed into discrete attributes by equal-frequency discretization, and a decision information table is constructed by using historical risk samples as rows and conditional attributes and decision attributes as columns.

[0141] Calculate the dependency of conditional attributes on decision attributes, identify key influencing features, and generate decision rules. The formula for calculating dependency is expressed as:

[0142]

[0143] In the formula, It is a conditional attribute. It is a decision-making attribute. It is the positive domain of the conditional attribute with respect to the decision attribute. It's about dependence. It represents the number of historical risk samples.

[0144] The ideal feature vector of the decision rule is used as the reference sequence, and the current extreme risk mapping data is used as the comparison sequence.

[0145] The single-index similarity between the comparison sequence and the reference sequence of each decision rule is calculated using the following formula:

[0146]

[0147] In the formula, It is the first Conditional attributes It is the resolution coefficient. It is the minimum absolute difference of all rules and all indicators. It is the maximum absolute difference of all rules and all indicators.

[0148] The comprehensive correlation degree is obtained by weighted summation based on dependency and single-index similarity, expressed by the formula:

[0149]

[0150]

[0151] In the formula, It is the first The importance weight of each attribute It is the overall correlation. Remove the first from the set of conditional attributes After considering all attributes, the dependence of the remaining subset of conditional attributes on the set of decision attributes.

[0152] Based on the safety requirements of the extreme charging scenario, a correlation threshold is set, and a graded protection strategy is obtained by matching the corresponding level protection strategy according to the comprehensive correlation.

[0153] Yellow risk: The risk value range is 0.2≤A<0.3, the correlation threshold is ≥0.80, the protection strategy type is low intervention, and the recommended combination of measures is APP warning + charging pile voice reminder.

[0154] Orange risk: The risk value range is 0.3≤A<0.6, the correlation threshold is ≥0.85, the protection strategy type is medium intervention, and the recommended combination of measures is 30% flow reduction + forced start of cooling system.

[0155] Red Risk: Risk value range is A≥0.6, correlation threshold is≥0.90, protection strategy type is high intervention, recommended combination of measures is emergency power outage + individual battery isolation + fire linkage.

[0156] The steps to obtain the optimal protection path include:

[0157] Ranking indicators are set based on comprehensive relevance, compatibility of protection strategies, confidence of decision rules, and balance of multiple objectives.

[0158] The comprehensive ranking score of each candidate path in the hierarchical protection strategy is calculated based on the ranking index, and the candidate paths are arranged in descending order of the score to obtain the path ranking table.

[0159] Based on the current hardware status of the Extreme Charge system and the dual verification path ranking table of the current Extreme Charge scenario, the candidate path that has passed the verification is selected as the optimal protection path.

[0160] The optimal protection path is parsed into an execution instruction set, and its execution is simulated in the modified extreme charging model to modify the hierarchical protection strategy.

[0161] The steps to parse and obtain the execution instruction set include:

[0162] Determine the combination of measures based on the optimal protection path, break it down into basic actions, and label the implementing entities.

[0163] For each basic action, the execution details are determined by combining risk mapping data with the hardware parameters of the Extreme Charge system.

[0164] For current / power-related actions: clearly define the target value, adjustment range, execution time, and accuracy requirements.

[0165] Thermal management actions: Define the startup mode, target temperature, and operating parameters.

[0166] Warning actions: Clearly state the content, volume, and number of repetitions. Example: Voice broadcast: The current battery temperature rise is abnormal, and the charging power has been reduced. Please pay attention. Volume: 80dB. Repeat twice, with a 1-second interval.

[0167] Emergency actions: Clearly define triggering conditions, execution priority, and confirmation mechanism. Example: Emergency power outage: Immediately disconnect the main charging circuit. No confirmation is required before execution. After power outage, a power outage success signal is fed back. Response time ≤ 50ms.

[0168] Develop priority rules for the charging scenario, identify the dependencies between the execution details of different basic actions, and determine the start time, execution duration, and completion flag for each basic action to generate a sequence execution table. The priority rule can be: Emergency safety actions > Risk blocking actions > Warning and prompt actions.

[0169] Execution schedule:

[0170]

[0171] The timing execution table is verified based on action conflicts, hardware boundaries, battery safety, and communication protocols. According to the unified instruction encoding rules of the Extreme Charge system, a unique instruction ID is assigned to each basic action to obtain the execution instruction set.

[0172] Action conflict verification: Check if there are mutually exclusive actions or resource preemption actions in the timing table. If a conflict exists, remove the lower priority action according to priority or adjust the execution timing.

[0173] Hardware boundary verification: Confirm that the parameters of each action do not exceed the rated range of the hardware. For example, if the cooling start power is 1.2kW, it is necessary to verify that the maximum power supply of the cooling system is ≥1.2kW. If it does not meet the requirement, the parameters should be adjusted.

[0174] Battery safety verification: Ensure that the command parameters meet the battery safety boundaries, such as "after current reduction, the current of 21A must be verified to be no less than the minimum charging current of the battery to avoid battery depletion or damage."

[0175] Communication protocol verification: Confirms the communication protocol supported by the executing entity for the instruction; if incompatible, it is converted to an adapted protocol instruction.

[0176] Standardized instruction format: Adopting an instruction header + parameter field + checksum + instruction tail structure. Example: Instruction header: Instruction type identifier. Parameter field: Quantized execution parameters. Checksum: Calculated based on the CRC8 algorithm to prevent instruction transmission errors. Instruction tail: Instruction end marker.

[0177] Classified and encapsulated by execution subject: The instruction set is classified into "charging pile instruction, BMS instruction, cooling system instruction" to generate multiple sub-instruction sets, avoiding cross-module instruction confusion.

[0178] The steps to revise the graded protection strategy include:

[0179] The timing logic and parameter details of the execution instruction set are analyzed, and the instruction format is converted according to the interface protocol of the modified extreme charge model.

[0180] The instructions are executed sequentially and simulated in the modified extreme charging model. Risk-related indicators, execution effect indicators, and battery status indicators are recorded as simulation results.

[0181] Risk-related indicators may include risk values, simulated CO concentration, and rate of temperature rise. Performance indicators may include actual current / voltage values ​​and temperature change curves. Battery status indicators may include changes in internal resistance and cell expansion.

[0182] By comparing the expected results of the instructions with the simulation results, the deviations of risk links, parameter execution deviations, and timing execution deviations are calculated as deviation indicators.

[0183] Risk mitigation deviation: The difference between the expected risk mitigation rate and the actual simulated mitigation rate, with a threshold of ≤10%.

[0184] Parameter execution deviation: The absolute deviation between the target parameters of the instruction and the actual output parameters of the model, with current deviation ≤ ±0.5A and temperature deviation ≤ ±1℃.

[0185] Timing execution deviation: The difference between the execution time required by the instruction and the actual completion time in the model, with a threshold of ≤20ms.

[0186] Determine whether the deviation index is within the preset deviation threshold. If it is, no correction is needed. Otherwise, locate the root cause of the deviation and correct the graded protection strategy according to the type of deviation root cause until it meets the preset deviation threshold.

[0187] If parameter execution deviation exceeds the limit: Investigate whether the instruction parameters are unreasonable or whether the model and physical system hardware parameters are incompatible. Corrective steps may include: adjusting instruction quantization parameters and optimizing timing intervals. For example, if the temperature deviation after cooling is 5°C, increase the cooling power from 1.2kW to 1.5kW. If current fluctuation exceeds the limit after current reduction, extend the current reduction execution time. If cooling starts before the pre-current reduction action is stable, causing voltage fluctuations, adjust the cooling start lag time from 50ms to 100ms to ensure the pre-current reduction action is completed.

[0188] If the risk mitigation deviation exceeds the limit: Investigate whether the combination of measures is inappropriate or the decision-making rules are mismatched. Corrective steps may include: supplementing / replacing measures and adjusting the priority of measures. For example, if current reduction alone cannot effectively control the temperature, add measures to open the heat dissipation channel of a single battery in the instruction set. If there is no isolation measure after an emergency power outage, resulting in residual risk, add instructions to isolate a single battery.

[0189] If a situation arises where early warning actions and risk mitigation actions are carried out concurrently, leading to resource contention, the early warning actions will be delayed until the risk mitigation actions are completed.

[0190] If the timing deviation exceeds the limit: check whether it is a conflict in the order of instruction execution.

[0191] This embodiment provides a digital twin-driven method for dynamic mapping and protection against extreme charging risks in electric two-wheelers. By capturing early signals such as lithium dendrite growth through microscopic features, it identifies risks minutes or even earlier before thermal runaway occurs. Furthermore, the constructed modified extreme charging model possesses self-correcting capabilities, continuously optimizing with battery aging and environmental changes, significantly improving prediction accuracy and robustness. The generated risk mapping data and tiered protection strategies provide a scientific basis for extreme charging management of electric two-wheelers, helping manufacturers, operators, and users better understand extreme charging risks and take reasonable measures to ensure extreme charging safety. This method improves extreme charging safety, enhances model accuracy and adaptability, and achieves precise protection against extreme charging risks.

[0192] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0193] 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 the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A digital twin driven e-two-wheeler extreme charging risk dynamic mapping and protection method, characterized by, The method comprises the following steps: Collecting electric two-wheeled vehicle extreme charging correlation data and whole vehicle system data, and constructing a digital twin extreme charging model in combination with electric core micro characteristics; A real-time correction mechanism is established through Lyapunov increment learning to correct errors of the digital twin extreme charging model to obtain a corrected extreme charging model, early out-of-control features are captured through a multi-scale feature fusion method, and risk mapping data is generated; According to the risk mapping data and based on a rough set-gray correlation coupling method, a hierarchical protection strategy is generated and an optimal protection path is dynamically selected; The optimal protection path is parsed into an execution instruction set, and the execution instruction set is simulated in the corrected extreme charging model, and the hierarchical protection strategy is corrected.

2. The digital twin driven e-two-wheeler extreme charging risk dynamic mapping and guard method as claimed in claim 1 wherein, The steps of obtaining the extreme charging correlation data and the whole vehicle system data comprise: Taking extreme charging process features, charging pile parameters and environmental data as multi-mode extreme charging data, and taking battery pack structure parameters, BMS state data, whole vehicle load characteristics and motor controller state as whole vehicle data; According to the multi-mode extreme charging data and the whole vehicle data, determine the sensor type, communication protocol and sampling frequency and accuracy; According to the extreme charging stage, trigger the sampling frequency and accuracy, trigger the enhanced collection of risk features according to the abnormal type, and align the whole vehicle data and the multi-mode extreme charging data according to the time stamp.

3. The digital twin driven e-two-wheeler extreme charging risk dynamic mapping and guard method as claimed in claim 1 wherein, The steps of constructing the digital twin extreme charging model comprise: Collecting lithium ion migration rate, SEI film impedance and lithium dendrite length as the electric core micro characteristics, combining the extreme charging correlation data and the whole vehicle system data, and performing standardization according to micro-meso-macro to obtain extreme charging data; From the extreme charging data, extract lithium dendrite growth rate, SEI film rupture voltage threshold, single cell voltage difference, temperature field gradient, charging curve slope and whole vehicle energy consumption feedback coefficient to construct an association dictionary; Based on the association dictionary, construct an electric core micro model, a battery pack meso model and a whole vehicle-charging pile macro model; Through an adaptive dimension reduction algorithm, high-dimensional micro features are compressed to a low-dimensional space matching meso features, and then a full-connection neural network is used to establish a nonlinear mapping with macro features; Using a long short-term memory network to learn the advanced influence of micro feature changes on macro risks, a time sequence association mechanism is established, and a data interface and a semantic layer architecture are used to fuse the electric core micro model, the battery pack meso model and the whole vehicle-charging pile macro model to obtain the digital twin extreme charging model.

4. The digital twin driven e-two-wheeler extreme charging risk dynamic mapping and guard method as claimed in claim 1 wherein, The steps of establishing the real-time correction mechanism comprise: For the three-level structure in the digital twin extreme charging model, define a multi-scale error index, generate a comprehensive error vector using a weighted fusion algorithm, and dynamically allocate weights based on data reliability; Taking the comprehensive error vector as input, constructing a positive definite function as a stability criterion, calculating the first derivative of the positive definite function, and judging whether it is less than 0, otherwise the error diverges and triggers the correction mechanism.

5. The digital twin driven e-two-wheeler extreme charging risk dynamic mapping and guard method as claimed in claim 4 wherein, The steps of obtaining the corrected extreme charging model comprise: Based on real-time data interaction between the digital twin extreme charging model and the physical system, synchronously collecting predicted values and measured values, and using a contribution amount analysis method to locate a dominant error source; Based on the dominant error source, analyze the influence parameters of the current comprehensive error vector on the digital twin extreme charging model to form a set of parameters to be corrected; updating the to-be-corrected parameter set by an incremental gradient descent algorithm to obtain a corrected parameter, so as to minimize the positive definite function; adjusting the digital twin extreme charging model according to the corrected parameter to obtain the corrected extreme charging model.

6. The digital twin driven e-two-wheeler extreme charging risk dynamic mapping and guard method as claimed in claim 1 wherein, The step of generating the risk mapping data comprises: defining early out-of-control features according to micro-meso-macro three-level scales, performing noise filtering and abnormal value processing, and standardization and spatio-temporal alignment; dynamically regulating feature weights based on importance of the early out-of-control features and risk stages, designing a fusion model using an attention mechanism and a cross-scale correlation network, calculating mutual information values of fusion features and the early out-of-control features, and retaining those reaching a preset threshold as global feature vectors; collecting historical risk data, constructing a gradient boosting tree model trained based on the historical risk data as a risk identification model, inputting the global feature vectors, and outputting to obtain multi-dimensional risk probabilities; generating a risk space distribution heat map according to the multi-dimensional risk probabilities combined with spatial alignment information, and predicting risk evolution trends based on time sequence changes of the fusion features to obtain the risk mapping data.

7. The digital twin driven e-two wheeler extreme charging risk dynamic mapping and guard method as claimed in claim 1 wherein, The step of evaluating the hierarchical protection strategy comprises: discretizing the risk mapping data by equal frequency, and constructing a decision information table by taking historical risk samples as rows, and taking conditional attributes and decision attributes as columns; calculating the dependency of the conditional attributes on the decision attributes, identifying key influence features to generate decision rules; taking ideal feature vectors of the decision rules as reference sequences, and taking risk mapping data of the current extreme charging as comparison sequences; calculating single-index similarity of the comparison sequences and reference sequences of each decision rule; weighting and summing to obtain a comprehensive correlation degree according to the dependency and the single-index similarity; combining safety requirements of the extreme charging scene, setting a correlation degree threshold, and matching a corresponding hierarchical protection strategy according to the comprehensive correlation degree to obtain the hierarchical protection strategy.

8. The digital twin driven e-two-wheeler extreme charging risk dynamic mapping and guard method as claimed in claim 7, wherein, The step of obtaining the optimal protection path comprises: setting a sorting index according to the comprehensive correlation degree, measure compatibility of the protection strategy, confidence of the decision rule, and multi-objective balance; calculating comprehensive sorting scores of each candidate path in the hierarchical protection strategy according to the sorting index, ranking the candidate paths in descending order of scores to obtain a path sorting table; combining current hardware states of the extreme charging system and the current extreme charging scene to verify the path sorting table, and selecting a candidate path that passes the verification as the optimal protection path.

9. The digital twin driven e-two wheeler extreme charging risk dynamic mapping and guard method as claimed in claim 1 wherein, The step of analyzing to obtain the execution instruction set comprises: determining a measure combination according to the optimal protection path, disassembling it into basic actions and labeling execution subjects; determining execution details for each basic action in combination with the risk mapping data and hardware parameters of the extreme charging system; formulating an extreme charging scene priority rule, combing dependency relationships between execution details of different basic actions, determining execution start time, execution duration and completion flag of each basic action to generate a time sequence execution table; verifying the time sequence execution table from action conflicts, hardware boundaries, battery safety and communication protocols, and assigning a unique instruction ID to each basic action according to a unified instruction coding rule of the extreme charging system to obtain the execution instruction set.

10. The digital twin driven e-two wheeler extreme charging risk dynamic mapping and guard method as claimed in claim 1 wherein, The step of modifying the hierarchical protection strategy comprises: analyzing the timing logic and parameter details of the execution instruction set, and converting the instruction format according to the interface protocol of the modified extreme charging model; simulating execution in the modified extreme charging model according to the timing sequence of the execution instruction set, and recording risk-related indicators, execution effect indicators and battery state indicators as simulation results; comparing the expected results of the instructions with the simulation results, and calculating risk link deviations, parameter execution deviations and timing execution deviations as deviation indicators; judging whether the deviation indicators are within a preset deviation threshold, and if yes, no modification is needed, otherwise, the root cause of the deviation is located, the hierarchical protection strategy is modified according to the type of the root cause of the deviation, and the modification is repeated until the preset deviation threshold is met.

Images