Method and system for adaptive allocation of charging power based on soh

Abstract

The application provides a SOH-based charging power adaptive allocation method and system, and relates to the electric two-wheeled vehicle charging technical field.The method comprises collecting the full life cycle data of the electric two-wheeled vehicle battery, establishing a double-path charging power allocation model, quantitatively analyzing the difference degree of two charging power allocation schemes in charging efficiency, battery loss and safety margin, verifying the effectiveness of the difference evaluation result and the actual adaptation scene of the two charging power allocation schemes, dynamically selecting the optimal allocation path or the advantage parameters of the fusion of the two paths, real-time monitoring the battery state drift in the charging process, dynamically correcting the allocation scheme difference threshold, and realizing the double-path collaborative optimization of the charging power adaptive allocation.The application prolongs the battery life and improves the charging safety through the double allocation mechanism combined with data support, difference evaluation verification and dynamic correction.

Description

Technical Field

[0001] This invention relates to the field of electric two-wheeled vehicle charging technology, and in particular to a charging power adaptive allocation method and system based on SOH. Background Technology

[0002] As a core mode of transportation for short-distance urban travel, electric two-wheelers have widely adopted ultra-fast charging technology to enhance user experience, enabling the battery to be replenished with a large amount of power in a short time.

[0003] However, the fast charging process for electric two-wheelers lacks the ability to accurately adapt to the battery's state throughout its entire lifecycle. State of Health (SOH), a core indicator reflecting battery degradation, directly impacts battery capacity, internal resistance, and thermal stability. However, most existing technologies lack a dynamic correlation between SOH and charging power, making it difficult to balance charging efficiency and battery protection. In real-world charging scenarios, batteries of different ages exhibit significant differences in SOH. Using a conservative charging strategy for new batteries wastes charging efficiency, while using high-power charging for degraded batteries exacerbates degradation and may even pose safety risks. Furthermore, electric two-wheelers operate in complex environments, with varying charging power requirements under different conditions such as normal temperature, low temperature, and high temperature, as well as dynamic load states. Existing static charging solutions cannot dynamically adapt to these changes in operating conditions, failing to provide stable and reliable charging guarantees for electric two-wheeler batteries under different SOH states and operating conditions. Summary of the Invention

[0004] This invention provides a charging power adaptive allocation method and system based on SOH to address the deficiencies in the prior art.

[0005] On one hand, the present invention provides a charging power adaptive allocation method based on SOH, comprising:

[0006] Collect data on the entire lifecycle of electric two-wheeler batteries and build a database linking battery health status, operating conditions, and power.

[0007] A dual-path charging power allocation model is established, including path A and path B. Path A is a static allocation mechanism based on SOH hierarchical-rule matching, and path B is a dynamic allocation mechanism based on SOH dynamic perception-deep learning prediction.

[0008] Two charging power allocation schemes were generated through two separate paths. A multi-dimensional difference evaluation index system was constructed to quantitatively analyze the differences between the two charging power allocation schemes in terms of charging efficiency, battery loss, and safety margin.

[0009] Design cross-validation experiments, combining a battery hardware-in-the-loop test platform, to verify the validity of the difference evaluation results and the actual adaptability of the two charging power allocation schemes in various scenarios.

[0010] An adaptive decision-making model is built based on the verification results. The optimal allocation path or the advantages of combining the two paths are dynamically selected according to the real-time SOH level, operating condition complexity and charging demand priority.

[0011] Real-time monitoring of battery state drift during charging, dynamic correction of allocation scheme difference thresholds, and adaptive allocation of charging power through dual-path collaborative optimization.

[0012] Output power allocation strategy, verification report and dynamic switching threshold parameters to complete charging power allocation with dual-path verification mechanism.

[0013] According to the present invention, a charging power adaptive allocation method based on State of Health (SOH) is provided. The full lifecycle data includes battery production parameters, cycle aging data, and fault history data. Battery production parameters include electrode material ratios and cell packaging process parameters. Cycle aging data includes capacity decay curves and internal resistance variation trends at different depths of charge and discharge. Fault history data includes overcharge fault records, over-discharge fault records, and temperature runaway warning logs. The battery health status-operating condition-power correlation database is stored using a time-series database, supporting multi-dimensional index queries by SOH interval and operating condition type. The SOH interval includes a first interval, a second interval, a third interval, and a fourth interval. Operating condition types include normal temperature static condition, low temperature fast charging condition, high temperature slow charging condition, and dynamic load condition.

[0014] According to the SOH-based adaptive charging power allocation method provided by the present invention, the process of constructing the SOH hierarchical-rule matching static allocation mechanism includes:

[0015] A graded power reference table is established based on the SOH graded threshold. The graded power reference table includes the correspondence between SOH range, reference charging power coefficient, maximum allowable current multiple, and temperature compensation coefficient.

[0016] Build a rule matching engine, input real-time operating parameters, and call the preset rule library to calculate the power correction value.

[0017] According to the present invention, a charging power adaptive allocation method based on SOH (State of Charge) is provided. The implementation process of constructing the SOH dynamic sensing-deep learning prediction dynamic allocation mechanism includes:

[0018] A dynamic SOH perception model is constructed, based on an attention mechanism-LSTM network, to predict the SOH change trend in real time within a set future time period.

[0019] A power prediction network is established, which takes the SOH prediction sequence and operating condition feature vector as input, and uses a dual-output convolutional neural network to output the optimal power and safety boundary.

[0020] A reinforcement learning optimization layer is introduced, with the shortest charging time and the minimum battery life loss as the reward function, to dynamically adjust the output power of path B.

[0021] According to the SOH-based adaptive charging power allocation method provided by the present invention, the process of constructing a multi-dimensional difference evaluation index system includes:

[0022] Define core difference indicators, which include efficiency difference, loss difference, and safety margin difference.

[0023] An indicator weight model was established, and the analytic hierarchy process (AHP) was used to calculate the weights of efficiency difference, loss difference, and safety margin difference, and the overall difference was calculated.

[0024] According to the SOH-based adaptive charging power allocation method provided by the present invention, the process of cross-validation experiments and hardware-in-the-loop testing verification includes:

[0025] Three sets of comparative experiments were designed. The first set of experiments combined the first SOH range with normal temperature conditions, the second set of experiments combined the third SOH range with low temperature conditions, and the third set of experiments combined the fourth SOH range with high temperature conditions.

[0026] A battery simulation model was built in a hardware-in-the-loop test platform. Full life cycle data was input to simulate the characteristics of a real battery. The charging power allocation schemes of path A and path B were run respectively, and the charging efficiency, SOH loss and safety boundary data of each group of experiments were collected.

[0027] Paired t-tests were used to verify the significance of the differences, and the appropriate scenarios were divided according to the verification results, including:

[0028] When the overall difference is less than the first difference threshold and the charging efficiency of path A is greater than or equal to the charging efficiency of path B, path A is adapted. When the overall difference is greater than or equal to the second difference threshold and the battery loss of path B is less than the battery loss of path A, path B is adapted. When the overall difference is greater than or equal to the first difference threshold and less than the second difference threshold, and the safety margin of path A is greater than the safety margin of path B, path fusion mode is activated.

[0029] According to the SOH-based adaptive charging power allocation method provided by the present invention, the process of constructing an adaptive decision model includes:

[0030] Decision rules are constructed using fuzzy logic reasoning. The input variables are SOH level, operating condition complexity, and charging priority, and the output variable is the path selection coefficient.

[0031] Define path fusion rules. When the path selection coefficient is greater than the first selection coefficient threshold and less than the second selection coefficient threshold, calculate the fusion power.

[0032] A safety boundary fusion mechanism is introduced and a decision model update mechanism is established. After each preset number of charging cycles, the fuzzy logic inference parameters are optimized using the gradient descent method based on historical verification data.

[0033] According to the present invention, a charging power adaptive allocation method based on SOH (State of Charge) is provided, and the process of realizing dual-path collaborative optimization of charging power adaptive allocation includes:

[0034] Define state drift indices, including SOH drift and temperature drift. SOH drift is the absolute value of the predicted SOH value compared to the actual SOH value. Temperature drift is the absolute value of the predicted temperature value compared to the actual temperature value.

[0035] Set a drift threshold. When the SOH drift exceeds the first drift threshold or the temperature drift exceeds the second drift threshold, trigger the difference threshold correction.

[0036] The path parameters are updated synchronously, and the SOH classification benchmark coefficient of path A is corrected.

[0037] The weights of the LSTM network for path B are optimized, and online learning is used to update the model parameters.

[0038] According to the present invention, a charging power adaptive allocation method based on SOH (System-in-the-Loop) provides dynamic switching threshold parameters including an SOH switching threshold, a comprehensive difference switching threshold, and a working condition complexity switching threshold. The SOH switching threshold includes a first SOH switching threshold and a second SOH switching threshold. The comprehensive difference switching threshold includes a first difference switching threshold and a second difference switching threshold. The working condition complexity switching threshold is a single threshold. When real-time parameters cross the threshold, the decision model automatically triggers path switching or fusion mode switching and records the switching event to the log system. The output verification report includes difference index data for each hardware-in-the-loop experiment, t-test results, and the criteria for determining the adaptation scenario.

[0039] On the other hand, the present invention also provides a charging power adaptive allocation system based on SOH, comprising:

[0040] The data acquisition module is used to collect data on the entire life cycle of electric two-wheeler batteries and build a database relating battery health status, operating conditions, and power.

[0041] The dual-path charging power allocation module is used to establish a dual-path charging power allocation model, including path A and path B. Path A is to build a static allocation mechanism based on SOH hierarchical-rule matching, and path B is to build a dynamic allocation mechanism based on SOH dynamic perception-deep learning prediction.

[0042] The multi-dimensional difference assessment module is used to generate charging power allocation schemes through two separate paths, construct a multi-dimensional difference assessment index system, and quantitatively analyze the differences between the two charging power allocation schemes in terms of charging efficiency, battery loss, and safety margin.

[0043] The verification and judgment module is used to design cross-validation experiments, combined with the battery hardware-in-the-loop test platform, to verify the validity of the difference evaluation results and the actual adaptation scenarios of the two charging power allocation schemes.

[0044] The decision fusion module is used to build an adaptive decision model based on the verification results. It dynamically selects the optimal allocation path or combines the advantages of the two paths according to the real-time SOH level, operating condition complexity and charging demand priority.

[0045] The monitoring and optimization module is used to monitor the battery state drift during the charging process in real time, dynamically correct the difference threshold of the allocation scheme, realize the adaptive allocation of charging power through dual-path collaborative optimization, and output the power allocation strategy, verification report and dynamic switching threshold parameters.

[0046] This invention provides a charging power adaptive allocation method and system based on State of Health (SOH). By constructing a full lifecycle data association database, it integrates and analyzes multi-dimensional data such as battery production, aging, and faults, providing comprehensive data support for charging power allocation. This solves the problem of incomplete understanding of battery status in existing technologies, making power allocation more closely match the actual characteristics of the battery. Through a dual-path allocation model, the static path achieves regular and rapid allocation based on SOH levels, while the dynamic path uses deep learning to predict SOH change trends and combines reinforcement learning for optimization. The two paths complement each other, ensuring allocation efficiency under simple operating conditions and improving adaptation accuracy under complex operating conditions, overcoming the limitation of single-path adaptation scenarios. By introducing a multi-dimensional difference evaluation and cross-validation mechanism, it achieves quantitative comparison and effectiveness verification of the two path allocation schemes, clarifying the optimal solution for different scenarios and avoiding inefficiency or battery damage caused by blindly selecting allocation strategies. Through an adaptive decision model combining real-time SOH levels, operating condition complexity, and charging demand priority, it achieves dynamic path selection and fusion, ensuring a balance between efficiency, loss, and safety during charging, meeting the needs of different users. Through state drift monitoring and parameter correction mechanisms, the system can respond to changes in battery state in real time, dynamically adjust allocation thresholds and path parameters, maintain allocation accuracy during long-term use, extend battery life, and improve charging safety, providing reliable charging technology support for the large-scale promotion of electric two-wheelers. Attached Figure Description

[0047] 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.

[0048] Figure 1 This is a schematic flowchart of a charging power adaptive allocation method based on SOH provided in an embodiment of the present invention;

[0049] Figure 2 This is a schematic diagram of a charging power adaptive allocation system based on SOH provided in an embodiment of the present invention. Detailed Implementation

[0050] 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.

[0051] The following is combined with Figures 1-2 The present invention describes a charging power adaptive allocation method and system based on SOH.

[0052] Figure 1 This is a flowchart illustrating a charging power adaptive allocation method based on SOH provided in an embodiment of the present invention.

[0053] like Figure 1 As shown, the charging power adaptive allocation method and system based on SOH provided in this embodiment of the invention can be executed by a charging power adaptive allocation method based on SOH, the method including:

[0054] Collect data on the entire lifecycle of electric two-wheeler batteries and build a database linking battery health status, operating conditions, and power.

[0055] The full lifecycle data includes battery production parameters, cycle aging data, and failure history data. Battery production parameters include electrode material ratios and cell packaging process parameters. Electrode material ratios specifically refer to the mass or molar proportions of key materials such as positive electrode active material, negative electrode active material, electrolyte, and separator. For example, the ratio of nickel, cobalt, and manganese in the positive electrode material of a ternary lithium battery. Cell packaging process parameters include packaging pressure, packaging temperature, sealant type and coating thickness, and tab welding parameters. Cycle aging data includes capacity decay curves and internal resistance trends at different charge / discharge depths. The charge / discharge depth is divided into multiple gradients, covering the range from 10% to 100%. At least 100 complete charge / discharge cycles are performed at each gradient, recording the actual capacity and DC internal resistance data after each cycle to form a continuous decay curve and trend graph. Historical fault data includes overcharge fault records, over-discharge fault records, and temperature runaway warning logs. Overcharge fault records must detail the charging voltage, current, duration, initial battery state, and subsequent battery performance changes at the time of the overcharge. The same applies to over-discharge fault records. Temperature runaway warning logs must include the battery core temperature, surface temperature, ambient temperature, charging stage, and corresponding voltage and current data at the time of warning triggering. The battery health status-operating condition-power correlation database uses a time-series database for storage, supporting multi-dimensional indexing and querying by SOH interval and operating condition type. The time-series database uses a product that supports high-concurrency writes and fast time-series queries, with data storage granularity accurate to the second, ensuring traceability of battery status data at every time point. The SOH interval includes four intervals: first interval SOH ≥ 85%, second interval 70% ≤ SOH < 85%, third interval 50% ≤ SOH < 70%, and fourth interval SOH < 50%. Operating conditions include ambient temperature resting condition, low temperature fast charging condition, high temperature slow charging condition, and dynamic load condition. Ambient temperature resting condition is defined as an ambient temperature of 20℃-25℃, with the battery under no additional load and resting for 30 minutes. Low temperature fast charging condition is defined as an ambient temperature of -10℃-0℃ and a charging current of not less than 1C. High temperature slow charging condition is defined as an ambient temperature of 35℃-45℃ and a charging current of not more than 0.3C. Dynamic load condition is defined as an ambient temperature of 15℃-30℃, with the battery undergoing alternating charging and discharging simultaneously, and the discharging current dynamically changing between 0.2C and 1C.

[0056] A dual-path charging power allocation model is established, including path A and path B. Path A is a static allocation mechanism based on SOH hierarchical-rule matching, and path B is a dynamic allocation mechanism based on SOH dynamic perception-deep learning prediction.

[0057] The process of constructing a static allocation mechanism based on SOH hierarchical rule matching includes:

[0058] A graded power reference table is established based on the SOH (State of Harshness) grading threshold. The graded power reference table includes the corresponding relationships between SOH ranges, reference charging power coefficients, maximum allowable current multiples, and temperature compensation coefficients. The graded power reference table is as follows:

[0059]

[0060] A rule-matching engine is built, which takes real-time operating parameters as input and calls a preset rule base to calculate the power correction value. The power correction calculation formula is as follows:

[0061]

[0062] In the formula, Output power for path A. This is the SOH grading reference coefficient, and its value directly corresponds to the reference charging power coefficient in the grading power reference table. Rated charging power, This is a temperature correction factor. Its calculation logic is as follows: when the ambient temperature is between 20℃ and 25℃, it takes a value of 1.0; for every 5℃ deviation of the temperature from this range, the factor is adjusted by 0.05, with a minimum of 0.5. This is the remaining power correction factor. This is the voltage correction factor. It is set to 1.0 when the charging interface voltage is within ±5% of the rated voltage. For every 1% deviation outside this range, the factor is adjusted by 0.02, with a minimum of 0.8. Real-time operating parameters include ambient temperature, remaining battery capacity, and charging interface voltage.

[0063] The implementation process of constructing the SOH dynamic perception-deep learning prediction dynamic allocation mechanism includes:

[0064] A dynamic SOH (Solar OH) perception model is constructed based on an attention mechanism-LSTM network. This network consists of an input layer, an attention layer, an LSTM layer, a fully connected layer, and an output layer. The input layer has a dimension of 4, corresponding to four input parameters: SOH(t), R(t), T(t), and I(t). The attention layer employs an additive attention mechanism, assigning different weights to input features through learned parameters to highlight the influence of key features on SOH prediction. The LSTM layer contains two hidden layers, each with 64 neurons. The activation function is the tanh function, and the forget gate, input gate, and output gate all use the sigmoid activation function. The fully connected layer contains 32 neurons, and the output layer has 1 neuron, outputting the predicted SOH value. The model predicts the SOH trend in real time over a set time period, which can be configured to 3 minutes, 5 minutes, or 10 minutes depending on the application scenario. The SOH prediction calculation formula is as follows:

[0065]

[0066] In the formula, for The predicted SOH value at time 10:00. For attention mechanism - LSTM model, Let SOH be the actual value at time t. Let be the internal resistance of the battery at time t. Let be the battery temperature at time t. Let t be the charging current at time t.

[0067] A power prediction network is established, taking the SOH prediction sequence and the operating condition feature vector as inputs. A dual-output convolutional neural network is used to output the optimal power and safety boundary. The dual-output convolutional neural network includes an input layer, convolutional layers, pooling layers, fully connected layers, and dual output layers. The dimension of the input layer is the sum of the length of the SOH prediction sequence and the dimension of the operating condition feature vector, where the length of the SOH prediction sequence equals the number of seconds in the set time period. The convolutional layers contain three convolutional blocks, each consisting of a convolutional layer, a batch normalization layer, and a ReLU activation function. The kernel sizes are 3×3, 5×5, and 3×3, and the number of kernels are 32, 64, and 128, respectively. The pooling layer uses max pooling with a 2×2 kernel size. The fully connected layer contains two hidden layers with 128 and 64 neurons, respectively. The dual output layers output the optimal power and the power safety upper limit, respectively, using linear and ReLU activation functions. The power prediction calculation formula is as follows:

[0068]

[0069] In the formula, Output power for path B. For the upper limit of power safety, This is the predicted sequence for SOH, including t+1 seconds to... All predicted SOH values ​​per second, For the working condition feature vector, It is a dual-output convolutional neural network. The operating condition feature vector includes ambient temperature, remaining battery power, and charging interface voltage.

[0070] A reinforcement learning optimization layer is introduced, with the reward function being the shortest charging time and the minimum battery life loss. The reinforcement learning uses the DQN (DeepQ-Network) algorithm, with the agent acting as a power adjustment module. The action space is the step size for adjusting the charging power, set to 0.05C. The state space includes the current SOH, battery temperature, remaining charge, and current charging power. The target network and evaluation network have the same structure, updated every 100 training steps. The experience replay pool has a capacity of 10,000, and the batch sampling size is 32. The output power of path B is dynamically adjusted, and the reward function is calculated using the following formula:

[0071]

[0072] In the formula, As a reward value, , These are the weighting coefficients. This refers to the actual charging time. The rated charging time, To predict charging loss, This represents the maximum permissible loss.

[0073] Two charging power allocation schemes were generated through two separate paths. A multi-dimensional difference evaluation index system was constructed to quantitatively analyze the differences between the two charging power allocation schemes in terms of charging efficiency, battery loss, and safety margin.

[0074] The process of constructing a multi-dimensional difference assessment indicator system includes:

[0075] Define core difference indicators, which include efficiency difference, loss difference, and safety margin difference.

[0076] The formula for calculating efficiency variability is:

[0077]

[0078] In the formula, This represents the efficiency difference index, ranging from 0 to 1. A larger value indicates a greater difference in charging efficiency between the two methods. For the charging efficiency of path A, The charging efficiency of path B is given. The charging efficiency of path A is the ratio of the product of the output power of path A and the charging time of path A to the grid input energy of path A, i.e. ,in The actual time it takes for path A to complete charging. The total energy input to the grid during the charging process of path A. The charging efficiency of path B is the ratio of the product of the output power of path B and the charging time of path B to the grid input energy of path B, i.e. ,in The actual time required to complete charging for path B. The total energy input to the power grid during the charging process of path B.

[0079] The formula for calculating the loss difference is:

[0080]

[0081] In the formula, For the degree of loss difference, For battery loss in path A, This represents the battery loss along path B. The battery loss along path A is the product of the loss coefficient, the square of the charging current along path A, the battery's internal resistance, and the charging time along path A. Where k is the loss coefficient, with a value ranging from 0.01 to 0.05. Let R be the charging current for path A, and R be the average internal resistance of the battery. The battery loss for path B is the product of the loss coefficient, the square of the charging current for path B, the battery's internal resistance, and the charging time for path B. ,in This is the charging current for path B.

[0082] The formula for calculating the safety margin variability is:

[0083]

[0084] In the formula, The safety margin variability is defined, with a value ranging from 0 to 1. For the safety margin of path A, This represents the safety margin for path B. The safety margin for path A is the ratio of the difference between the upper limit of the safe power of path A and the output power of path A to the upper limit of the safe power of path A, i.e. ,in The upper limit of the safe power for path A is determined by combining the battery's maximum allowable charging power and the safe power threshold at the current temperature. The safety margin for path B is the ratio of the difference between the upper limit of the safe power for path B and the output power of path B to the upper limit of the safe power for path B. ,in The upper limit of the safe power for path B is directly output by a dual-output convolutional neural network.

[0085] A weighted indicator model was established, and the analytic hierarchy process (AHP) was used to calculate the weights for efficiency difference, loss difference, and safety margin difference. The AHP construction process was as follows: First, a hierarchical structure was established, consisting of a target layer (comprehensive difference assessment), a criterion layer (efficiency difference, loss difference, and safety margin difference), and a scheme layer (two charging power allocation schemes). Then, a judgment matrix was constructed from the criterion layer to the target layer. This judgment matrix was determined based on expert scoring using a 1-9 scale, where 1 indicates that both indicators are equally important, and 9 indicates that one indicator is extremely more important than the other. Next, the largest eigenvalue and corresponding eigenvector of the judgment matrix were calculated, and the eigenvectors were normalized to obtain the weight values. Finally, a consistency check was performed, and consistency indicators were determined. ,in The largest eigenvalue, Given the number of indicators, when CI < 0.1, the judgment matrix meets the consistency requirement, and the weight allocation is reasonable. The sum of the efficiency difference weight, loss difference weight, and safety margin weight is 1.

[0086] The overall difference degree is calculated using the following formula:

[0087]

[0088] In the formula, The overall difference score ranges from 0 to 1. A larger value indicates a more significant overall difference between the two options. As the weight for efficiency differences, As the weight for the difference in loss, The weights for safety margin variability are all normalized weights calculated using the analytic hierarchy process (AHP).

[0089] Design cross-validation experiments, combining a battery hardware-in-the-loop test platform, to verify the validity of the difference evaluation results and the actual adaptability of the two charging power allocation schemes in various scenarios.

[0090] The process of cross-validation experiments and hardware-in-the-loop testing and verification includes:

[0091] Three sets of comparative experiments were designed. The first set of experiments combined the first SOH range with room temperature conditions; the second set combined the third SOH range with low temperature conditions; and the third set combined the fourth SOH range with high temperature conditions. In this embodiment, the battery degradation in the second SOH range is moderate, and the second SOH range is used as a comparison with various operating condition combinations.

[0092] A battery simulation model was built in a hardware-in-the-loop test platform. This model, based on MATLAB / Simulink, includes an electrochemical model, a thermal model, an electrical interface model, and a fault injection model. The electrochemical model uses a Thevenin model with extended Kalman filter parameter identification, accurately simulating battery voltage and current responses under different state of equilibrium (SOH) and operating conditions. The thermal model employs a lumped-parameter thermal model, considering heat transfer processes such as internal battery heat generation, surface heat dissipation, and environmental heat exchange, achieving a temperature simulation accuracy of ±1℃. The electrical interface model supports hardware interface interfacing with actual charging equipment, enabling real-time power and signal interaction. The fault injection model supports simulating fault scenarios such as overcharging, over-discharging, and temperature runaway. Full lifecycle data was input to simulate real battery characteristics. Charging power allocation schemes for paths A and B were run, and charging efficiency, SOH loss, and safety boundary data were collected for each experimental group.

[0093] The paired t-test was used to verify the significance of the difference. The formula for calculating the test statistic is as follows:

[0094]

[0095] In the formula, t is the test statistic. The mean of the difference index. Let s be the mean of the variance, s be the standard deviation of the variance index, and n be the sample size. We assume the mean of the variance is 0.

[0096] Based on the verification results, adaptation scenarios are defined. When the overall difference is less than the first difference threshold and the charging efficiency of path A is greater than or equal to the charging efficiency of path B, path A is adapted. When the overall difference is greater than or equal to the second difference threshold and the battery loss of path B is less than that of path A, path B is adapted. When the overall difference is greater than or equal to the first difference threshold and less than the second difference threshold, and the safety margin of path A is greater than the safety margin of path B, path fusion mode is activated.

[0097] An adaptive decision-making model is built based on the verification results. The optimal allocation path or the advantages of combining the two paths are dynamically selected according to the real-time SOH level, operating condition complexity and charging demand priority.

[0098] The process of building an adaptive decision model includes:

[0099] Decision rules are constructed using fuzzy logic reasoning, which includes a fuzzification module, a rule base module, an inference engine module, and a defuzzification module. The fuzzification module converts the three input variables—SOH level, operating condition complexity, and charging demand priority—into fuzzy sets. The SOH level is divided into three fuzzy subsets: high, medium, and low, with triangular membership functions. Operating condition complexity is divided into three fuzzy subsets: simple, medium, and complex, with trapezoidal membership functions. Charging demand priority is also divided into three fuzzy subsets: high, medium, and low, with triangular membership functions. The inference engine module uses the Mamdani inference method, obtaining the fuzzy output set through fuzzy conjunction, fuzzy disjunction, and fuzzy implication operations. The defuzzification module uses the centroid method to convert the fuzzy output set into path selection coefficients. The input variables are SOH level, operating condition complexity, and charging priority; the output variable is the path selection coefficient, calculated using the following formula:

[0100]

[0101]

[0102] In the formula, The path selection coefficient is denoted by f, which ranges from 0 to 1. A value closer to 1 indicates a stronger preference for path B, while a value closer to 0 indicates a stronger preference for path A. f is the fuzzy logic inference function. The SOH (Sodium Hydrochloric Oxide) level is determined based on the real-time SOH value: SOH ≥ 85% is "High", 70% ≤ SOH < 85% is "Medium", and SOH < 70% is "Low". For the complexity of the working conditions, For charging priority, The complexity coefficient is... For temperature fluctuation, This refers to current fluctuation. Charging priorities include fast charging demand priority, regular charging demand priority, and slow charging demand priority, with fast charging demand priority corresponding to... Suitable for scenarios where users need rapid power replenishment; priority is given to regular charging needs. Suitable for everyday charging scenarios; slow charging needs are prioritized accordingly. It is suitable for scenarios where charging speed is not a requirement and the goal is to maximize battery life.

[0103] Define path fusion rules: when the path selection coefficient is greater than the first selection coefficient threshold and less than the second selection coefficient threshold, the first selection coefficient threshold is set to 0.3 and the second selection coefficient threshold is set to 0.7. Calculate the fusion power using the following formula:

[0104]

[0105] In the formula, To integrate power, This is the path selection coefficient. Output power for path B. Output power for path A.

[0106] Simultaneously, security boundary fusion is introduced, and the formula for calculating the fused security boundary is:

[0107]

[0108] In the formula, To integrate the safety boundaries, the minimum value of the safety power upper limit for path A and path B is taken. The upper limit of safe power for path A. The upper limit of safe power for path B. This is the path selection coefficient.

[0109] Establish a decision model update mechanism. After each set number of charging cycles, optimize the fuzzy logic inference parameters based on historical verification data using the gradient descent method to improve decision accuracy.

[0110] Real-time monitoring of battery state drift during charging, dynamic correction of allocation scheme difference thresholds, and adaptive allocation of charging power through dual-path collaborative optimization are achieved. The process includes:

[0111] State drift indices are defined, including SOH drift and temperature drift. SOH drift is the absolute value of the predicted SOH value and the actual SOH value, i.e. in This represents the SOH drift amount. This is the predicted value of SOH. for The actual measured SOH value at time [time]. The temperature drift is the absolute value of the predicted temperature value and the actual temperature value, i.e. in The temperature value predicted by the path B model. for The actual temperature measurement at that moment.

[0112] A drift threshold is set, with the first drift threshold set to 2% and the second drift threshold set to 3°C. When the SOH drift exceeds the first drift threshold or the temperature drift exceeds the second drift threshold, a difference threshold correction is triggered. The formula for calculating the overall difference threshold after correction is as follows:

[0113]

[0114] In the formula, To correct the overall difference threshold, To correct the overall difference threshold, This represents the SOH drift amount. This is the temperature drift. This is the SOH drift influence coefficient, with a value range of 0.08-0.12 and a default value of 0.1. This is the temperature drift influence coefficient, with a value range of 0.03-0.07 and a default value of 0.05.

[0115] The path parameters are updated synchronously, and the SOH classification benchmark coefficient of path A is corrected. The formula for calculating the corrected SOH classification benchmark coefficient is as follows:

[0116]

[0117] In the formula, This is the corrected SOH classification benchmark coefficient. To correct the previous SOH classification benchmark coefficient, This represents the SOH drift amount. This is the correction factor for the SOH classification standard, with a value range of 0.01-0.03 and a default value of 0.02.

[0118] The weights of the LSTM network for path B are optimized by using online learning to update the model parameters. The learning rate for online learning is set to 0.001. Each time the drift threshold correction is triggered, the latest actual and predicted values ​​of SOH and the corresponding input parameters are used to construct training samples, and the LSTM network is fine-tuned to update the network weights, ensuring the prediction accuracy of the model.

[0119] Output power allocation strategy, verification report and dynamic switching threshold parameters to complete charging power allocation with dual-path verification mechanism.

[0120] The dynamic switching threshold parameters include the SOH switching threshold, the comprehensive difference switching threshold, and the operating condition complexity switching threshold. The SOH switching threshold includes a first SOH switching threshold and a second SOH switching threshold. The first SOH switching threshold is set to 85%, and the second SOH switching threshold is set to 50%, used to determine the path adaptation tendency corresponding to the SOH level. The comprehensive difference switching threshold includes a first difference switching threshold and a second difference switching threshold. The first difference switching threshold is set to 0.1, and the second difference switching threshold is set to 0.3, consistent with the difference thresholds in the cross-validation experiment. The operating condition complexity switching threshold is a single threshold set to 0.6. When the operating condition complexity is greater than this threshold, path B or the fusion mode is prioritized; when it is less than this threshold, path A is prioritized. When real-time parameters cross the threshold, the decision model automatically triggers path switching or fusion mode switching and records the switching event to the log system. The log system records the switching time, parameters before switching, path / mode after switching, and switching reason. The output validation report includes the difference index data for each hardware-in-the-loop experiment, t-test results, and the criteria for determining the suitability of the scenario. The difference index data is presented in tabular form, including the results for each experiment. , , , Specific numerical values; t-test results include t-values ​​and significance assessments; the criteria for determining the suitability of a scenario clearly define the optimal path or mode and the judgment logic for each experimental group.

[0121] In summary, this embodiment provides a charging power adaptive allocation method based on State of Health (SOH). By constructing a full lifecycle data association database, it integrates and analyzes multi-dimensional data such as battery production, aging, and faults, providing comprehensive data support for charging power allocation. This solves the problem of incomplete understanding of battery status in existing technologies, making power allocation more closely match the actual characteristics of the battery. Through a dual-path allocation model, the static path achieves regular and rapid allocation based on SOH levels, while the dynamic path uses deep learning to predict SOH change trends and combines reinforcement learning for optimization. The two paths complement each other, ensuring allocation efficiency under simple operating conditions while improving adaptation accuracy under complex operating conditions, overcoming the limitation of single-path adaptation scenarios. By introducing a multi-dimensional difference evaluation and cross-validation mechanism, it achieves quantitative comparison and effectiveness verification of the two path allocation schemes, clarifying the optimal solution for different scenarios and avoiding inefficiency or battery damage caused by blindly selecting allocation strategies. An adaptive decision model, combined with real-time SOH levels, operating condition complexity, and charging demand priority, achieves dynamic path selection and fusion, ensuring a balance between efficiency, loss, and safety during charging, meeting the needs of different users. Through state drift monitoring and parameter correction mechanisms, the system can respond to changes in battery state in real time, dynamically adjust allocation thresholds and path parameters, maintain allocation accuracy during long-term use, extend battery life, and improve charging safety, providing reliable charging technology support for the large-scale promotion of electric two-wheelers.

[0122] Based on the same general inventive concept, this invention also protects a charging power adaptive allocation system based on SOH. The following describes a charging power adaptive allocation system based on SOH provided by this invention. The charging power adaptive allocation system based on SOH described below and the charging power adaptive allocation method based on SOH described above can be referred to in correspondence.

[0123] Figure 2 This is a schematic diagram of a charging power adaptive allocation system based on SOH provided in an embodiment of the present invention.

[0124] like Figure 2 As shown, a charging power adaptive allocation system based on SOH includes a memory, a processor, and a computer program stored in the memory and capable of running on the processor. The processor includes a data acquisition module, a dual-path charging power allocation module, a multi-dimensional difference assessment module, a verification and judgment module, a decision fusion module, and a monitoring and optimization module.

[0125] The data acquisition module is used to collect data on the entire life cycle of electric two-wheeler batteries and build a database relating battery health status, operating conditions, and power.

[0126] The dual-path charging power allocation module is used to establish a dual-path charging power allocation model, including path A and path B. Path A is to build a static allocation mechanism based on SOH hierarchical-rule matching, and path B is to build a dynamic allocation mechanism based on SOH dynamic perception-deep learning prediction.

[0127] The multi-dimensional difference assessment module is used to generate charging power allocation schemes through two paths respectively, construct a multi-dimensional difference assessment index system, and quantitatively analyze the differences between the two charging power allocation schemes in terms of charging efficiency, battery loss, and safety margin.

[0128] The verification and judgment module is used to design cross-validation experiments, combined with the battery hardware-in-the-loop test platform, to verify the validity of the difference evaluation results and the actual adaptation scenarios of the two charging power allocation schemes.

[0129] The decision fusion module is used to build an adaptive decision model based on the verification results. It dynamically selects the optimal allocation path or combines the advantages of the two paths according to the real-time SOH level, operating condition complexity and charging demand priority.

[0130] The monitoring and optimization module is used to monitor the battery state drift during the charging process in real time, dynamically correct the difference threshold of the allocation scheme, realize the adaptive allocation of charging power through dual-path collaborative optimization, and output the power allocation strategy, verification report and dynamic switching threshold parameters.

[0131] 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.

[0132] 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 charging power adaptive allocation method based on SOH, characterized in that, include: Collect data on the entire lifecycle of electric two-wheeler batteries and build a database relating battery health status, operating conditions, and power. A dual-path charging power allocation model is established, including path A and path B. Path A is a static allocation mechanism based on SOH hierarchical-rule matching, and path B is a dynamic allocation mechanism based on SOH dynamic sensing-deep learning prediction. The process of constructing a static allocation mechanism based on SOH hierarchical rule matching includes: A graded power reference table is established based on the SOH graded threshold. The graded power reference table includes the correspondence between SOH range, reference charging power coefficient, maximum allowable current multiple, and temperature compensation coefficient. Build a rule matching engine, input real-time operating parameters, and call the preset rule library to calculate the power correction value; Two charging power allocation schemes were generated through two separate paths. A multi-dimensional difference evaluation index system was constructed to quantitatively analyze the differences between the two charging power allocation schemes in terms of charging efficiency, battery loss, and safety margin. Design cross-validation experiments, combined with a battery hardware-in-the-loop test platform, to verify the validity of the difference evaluation results and the actual adaptability of the two charging power allocation schemes in various scenarios. An adaptive decision-making model is built based on the verification results. The optimal allocation path or the advantages of combining the two paths are dynamically selected according to the real-time SOH level, operating condition complexity and charging demand priority. Real-time monitoring of battery state drift during charging, dynamic correction of allocation scheme difference thresholds, and realization of adaptive allocation of charging power through dual-path collaborative optimization; Output power allocation strategy, verification report and dynamic switching threshold parameters to complete charging power allocation with dual-path verification mechanism.

2. The adaptive charging power allocation method based on SOH according to claim 1, characterized in that, The full lifecycle data includes battery production parameters, cycle aging data, and fault history data; the battery production parameters include electrode material ratios and cell packaging process parameters; the cycle aging data includes capacity decay curves and internal resistance variation trends at different depths of charge and discharge. The fault history data includes overcharge fault records, over-discharge fault records, and temperature runaway warning logs; the battery health status-operating condition-power correlation database is stored in a time-series database and supports multi-dimensional index queries by SOH interval and operating condition type; the SOH interval includes the first interval, the second interval, the third interval, and the fourth interval; the operating condition type includes normal temperature static condition, low temperature fast charging condition, high temperature slow charging condition, and dynamic load condition.

3. The adaptive charging power allocation method based on SOH according to claim 1, characterized in that, The implementation process of constructing the SOH dynamic perception-deep learning prediction dynamic allocation mechanism includes: A dynamic SOH perception model is constructed based on an attention mechanism-LSTM network to predict the SOH change trend in real time within a set future time period. A power prediction network is established, with the SOH prediction sequence and operating condition feature vector as inputs, and a dual-output convolutional neural network is used to output the optimal power and safety boundary. A reinforcement learning optimization layer is introduced, with the shortest charging time and the minimum battery life loss as the reward function, to dynamically adjust the output power of path B.

4. The adaptive charging power allocation method based on SOH according to claim 1, characterized in that, The process of constructing a multi-dimensional difference assessment indicator system includes: Define core difference indicators, including efficiency difference, loss difference, and safety margin difference. An indicator weight model was established, and the analytic hierarchy process (AHP) was used to calculate the weights of efficiency difference, loss difference, and safety margin difference, and the overall difference was calculated.

5. The adaptive charging power allocation method based on SOH according to claim 1, characterized in that, The process of cross-validation experiments and hardware-in-the-loop testing and verification includes: Three sets of comparative experiments were designed. The first set of experiments combined the first SOH range with normal temperature conditions, the second set of experiments combined the third SOH range with low temperature conditions, and the third set of experiments combined the fourth SOH range with high temperature conditions. A battery simulation model was built in the hardware-in-the-loop test platform. Full life cycle data was input to simulate the characteristics of a real battery. The charging power allocation schemes of path A and path B were run respectively, and the charging efficiency, SOH loss and safety boundary data of each group of experiments were collected. Paired t-tests were used to verify the significance of the differences, and the appropriate scenarios were divided according to the verification results, including: When the overall difference is less than the first difference threshold and the charging efficiency of path A is greater than or equal to the charging efficiency of path B, path A is adapted; when the overall difference is greater than or equal to the second difference threshold and the battery loss of path B is less than the battery loss of path A, path B is adapted; when the overall difference is greater than or equal to the first difference threshold and less than the second difference threshold and the safety margin of path A is greater than the safety margin of path B, the path fusion mode is activated.

6. The adaptive charging power allocation method based on SOH according to claim 1, characterized in that, The process of building an adaptive decision model includes: Decision rules are constructed using fuzzy logic reasoning, with input variables being SOH level, operating condition complexity, and charging priority, and output variable being path selection coefficient; Define path fusion rules, and calculate fusion power when the path selection coefficient is greater than the first selection coefficient threshold and less than the second selection coefficient threshold; A safety boundary fusion mechanism is introduced and a decision model update mechanism is established. After each preset number of charging cycles, the fuzzy logic inference parameters are optimized using the gradient descent method based on historical verification data.

7. The adaptive charging power allocation method based on SOH according to claim 1, characterized in that, The process of achieving adaptive charging power allocation through dual-path collaborative optimization includes: Define state drift indices, which include SOH drift and temperature drift; the SOH drift is the absolute value of the predicted SOH value and the actual SOH value; the temperature drift is the absolute value of the predicted temperature value and the actual temperature value. Set a drift threshold. When the SOH drift exceeds the first drift threshold or the temperature drift exceeds the second drift threshold, trigger the difference threshold correction. The path parameters are updated synchronously, and the SOH classification benchmark coefficient of path A is corrected. The weights of the LSTM network for path B are optimized, and online learning is used to update the model parameters.

8. The adaptive charging power allocation method based on SOH according to claim 1, characterized in that, The dynamic switching threshold parameters include the SOH switching threshold, the comprehensive difference switching threshold, and the operating condition complexity switching threshold; the SOH switching threshold includes a first SOH switching threshold and a second SOH switching threshold; the comprehensive difference switching threshold includes a first difference switching threshold and a second difference switching threshold; the operating condition complexity switching threshold is a single threshold; when the real-time parameters cross the threshold, the decision model automatically triggers path switching or fusion mode switching and records the switching event to the log system; the output verification report includes the difference index data of each hardware-in-the-loop experiment, the t-test results, and the criteria for determining the adaptation scenario.

9. A charging power adaptive allocation system based on SOH, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements a charging power adaptive allocation method based on SOH as described in any one of claims 1 to 8.

Images