Artificial intelligence based energy storage battery thermal management system

Abstract

This invention discloses an artificial intelligence-based thermal management system for energy storage batteries, belonging to the field of battery thermal management technology. The system includes a data acquisition module, a feature construction module, a condition judgment module, and a management execution module. The system acquires the temperature of the battery casing's neighborhood, the representative temperature of the battery cell, and electrical stress parameters that quantify the cost of wireless charging transmission to construct thermal and electrical characteristic indices. Using a self-referenced historical database indexed by an identifier, it determines the degree of deviation of the current characteristic index from the historical baseline set, thereby determining whether the system is in a condition where the casing heats up first and then reverses. When this condition is determined, thermal management reconfiguration is performed, switching the main feedback target of the control loop from the representative temperature of the battery cell to the temperature of the casing's neighborhood. This invention effectively solves the problem of thermal response hysteresis caused by casing eddy current heating under wireless charging compensation current rise conditions.

Description

Technical Field

[0001] This invention relates to the field of battery thermal management technology, and more specifically to an artificial intelligence-based energy storage battery thermal management system. Background Technology

[0002] High-power wireless power transfer technology is gradually becoming an important energy replenishment method for electric vehicles. In on-board energy storage systems, the power battery pack typically integrates a liquid-cooled thermal management system to maintain the appropriate operating temperature of the battery cells, while its bottom is equipped with a large-area metal protective plate to ensure mechanical strength. This metal plate is located directly under the vehicle chassis and, during charging, is only separated from the ground transmitter by an air gap. It is situated within a strongly coupled region of a high-frequency alternating magnetic field and serves as a key receiver in the wireless power transfer link.

[0003] However, existing battery thermal management control logic generally has a response blind spot for the special operating conditions of wireless charging. When there is a misalignment in the vehicle parking, the wireless charging system often adopts a current compensation strategy to maintain a constant power output. The surge in coil current will induce strong eddy current losses and hysteresis losses in the metal protective plate at the bottom of the battery pack, causing the base plate to accumulate heat and rise rapidly. Traditional thermal management systems only use the internal cell temperature as the single feedback source. Due to the huge heat capacity and thermal resistance of the battery pack structure, there is a significant thermal hysteresis. When the external base plate has reached a dangerous temperature, the internal cell temperature has not yet fluctuated significantly due to thermal inertia. This causes the system to misjudge the situation as safe and fail to start cooling in time, missing the best opportunity to block heat from entering the interior, causing structural adhesive aging and failure, or even inducing the risk of thermal runaway. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention proposes an artificial intelligence-based thermal management system for energy storage batteries. This system solves the problem of overheating of the battery pack's bottom casing due to eddy current caused by compensation current rise under wireless charging alignment deviation conditions. Traditional control strategies based on cell temperature feedback suffer from severe thermal hysteresis, leading to untimely thermal response and thermal interception failure.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an artificial intelligence-based energy storage battery thermal management system, comprising:

[0006] The data acquisition module acquires the ambient temperature of the battery pack's outer casing, the representative temperature of the battery cells inside the battery, electrical stress parameters that quantify the cost of wireless charging transmission, DC charging power, and an identification identifier containing charging location information.

[0007] The feature construction module determines the thermal characteristic index based on the dynamic change relationship between the temperature of the outer casing neighborhood and the representative temperature of the cell within a time window, and determines the electrical characteristic index based on the correspondence between electrical stress parameters and DC charging power.

[0008] The operating condition discrimination module retrieves the historical baseline set based on the identity identifier, determines the relative deviation of the current thermal characteristic index and electrical characteristic index from the historical baseline set, and determines whether the system is in the shell-heating-reversal operating condition based on the relative deviation.

[0009] The management execution module is used to perform thermal management reconfiguration when it is determined that the casing is in a reverse heating condition. It switches the main feedback target of the thermal management control loop from the cell representative temperature to the casing neighborhood temperature, and adjusts the cooling strategy for the battery pack based on the switched main feedback target.

[0010] Preferably, the self-referenced historical database indexed by the identity identifier is updated, as follows:

[0011] The identity identifier is used as the index key of the self-referenced history database, which is a non-volatile storage structure.

[0012] The current thermal characteristic index, electrical characteristic index, and DC charging power are combined into a record tuple, and the following dynamic grouping strategy is executed using a preset power determination threshold to store the record tuple in the self-referenced historical database:

[0013] Calculate the difference between the current DC charging power and the power value of the existing record tuple under this index key in the self-referenced historical database;

[0014] If there is a record tuple in the reference historical database whose absolute value of the difference is less than the power determination threshold, then the current record tuple is assigned to the power group to which the existing record tuple belongs.

[0015] If no record tuple that satisfies the difference condition exists in the reference historical database, a new power group is established based on the current DC charging power, and the current record tuple is stored in the new power group.

[0016] Compared with existing technologies, it has the following advantages:

[0017] This proposed AI-based thermal management system for energy storage batteries effectively solves the technical challenges of overheating of the battery pack's bottom casing due to eddy currents caused by compensation current rise during high-power wireless charging under alignment deviation conditions. Traditional cell-based temperature feedback control suffers from severe thermal hysteresis and response blind spots. This system constructs a self-referenced historical database indexed by identity identifiers. By accurately identifying the relative deviation of multi-dimensional feature indices, it pinpoints the casing overheating reversal condition and performs thermal management reconfiguration when an anomaly is detected. The main feedback target of the control loop is instantly switched from the sluggish cell-represented temperature to the sensitive casing neighborhood temperature, thereby building a thermal interception barrier at the base plate before heat can penetrate into the cell. This proactive intervention strategy not only eliminates the thermal control dead zone caused by external electromagnetic thermal effects, achieving zero-delay response and precise suppression of base plate overheating, but also effectively prevents the aging and failure of thermally conductive adhesives and the risk of thermal runaway. It also avoids the drawbacks of frequent power outages caused by simple trigger thresholds in traditional solutions, significantly improving the charging success rate of the wireless charging system while ensuring the safety of the battery throughout its entire life cycle. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the system framework of the present invention;

[0019] Figure 2 This is a schematic diagram of the operation process of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Please see Figures 1 to 2 This application provides an artificial intelligence-based energy storage battery thermal management system, including a data acquisition module, a feature construction module, an operating condition discrimination module, and a management execution module;

[0022] The data acquisition module obtains the cell's representative temperature through the vehicle-mounted sensor network and the vehicle-to-ground communication protocol. The cell temperature can be selected as either the highest or average temperature point within the module. The DC charging power is calculated using the current and voltage data reported by the on-board charger. The vehicle's static configuration and wireless charging handshake protocol are used to read the vehicle's identification ID.

[0023] In addition, a temperature sensor is placed in the electromagnetic thermal effect sensitive area at the bottom of the electric vehicle's battery pack to capture the heat accumulation caused by eddy current losses and collect the temperature of the surrounding area. Obtain a single scalar value characterizing the electrical stress intensity exerted by the wireless charging system to maintain power transmission, and label it as the electrical stress parameter. Specifically, electrical stress parameters According to Joule's law, the higher the current (RMS value) flowing through the wireless charging receiving or transmitting coil, the greater the risk of overheating in the coil and surrounding metal components. For example, under good alignment, to transmit 10kW of power, the system only needs a coil current of 20A. =20, under the condition of alignment deviation, in order to transmit 10kW of power, due to the decrease in efficiency, the system has to increase the coil current to 50A to maintain power output. =50, at the same output power The change from 20 to 50 means a sharp increase in transmission costs, which indicates that a huge amount of additional heat loss is occurring.

[0024] In detail, all parameters that characterize the "electrical cost of maintaining power" are included in the electrical stress parameters. The data includes various raw data that the wireless charging system may provide, such as the transmitter coil current, the phase difference between the primary and secondary currents or the reactive current component, and the resonant cavity voltage. However, in this example, the effective value of the transmitter coil current obtained by the system is typically used as the reference value. In another embodiment, if the system obtains combined data including current and phase difference, a single value needs to be calculated using a preset weighted formula, such as formula... ,in For coil current, For phase difference, and The weighting coefficients are pre-calibrated based on the electromagnetic coupling characteristics of different vehicle models, and are used to balance the contribution of current amplitude and phase deviation to heat loss.

[0025] It's important to note that the identification ID is a composite data tag used not only to distinguish vehicles but also to differentiate charging environments. Therefore, it must contain information uniquely pointing to the current ground-based transmitter. When the ground coil ID cannot be directly obtained via communication protocols, the location coordinate hash value of the ground facility or the ground QR code information identified by the vehicle's visual sensors can be used as a substitute. For example, if the vehicle code is VIN_LS12345 (representing a specific car) and the location code is LOC_SH_001 (representing transmitter board number 1 of a certain charging station), then the identification ID would be VIN_LS12345_LOC_SH_001. When the system detects this vehicle again at the same location, it will directly retrieve historical data under this ID (such as historical electrical stress parameters). The average value is 20), if the electrical stress parameters measured this time... If the value is 50, the system can immediately detect the anomaly.

[0026] It should be noted that regarding the temperature near the outer casing... The sensor placement should be within the electromagnetic thermal effect sensitive area. This sensitive area can be determined through thermal simulation during vehicle development or calibration using an infrared thermal imager under alignment deviation conditions. The area with the highest eddy current density or the fastest temperature rise rate at the bottom of the battery pack should be selected as the sensitive area. The sensor can be attached to the inner surface of the metal base plate or integrated into the sandwich layer of the anti-collision shell, as long as it can be thermally coupled to structural components that are heated by external magnetic fields. Assuming the vehicle is undergoing 11kW wireless charging at an ambient temperature of 25℃, under normal operating conditions, the base plate temperature may slowly rise to 30℃, with little difference from the internal battery cell temperature. Under special conditions (compensation current boost), due to the eddy current effect, the base plate temperature rapidly rises from 25℃ to 55℃ within 5 minutes of charging starting, while the internal battery cell temperature may only have just reached 26℃. The value "55" collected by the sensor at this time represents the temperature of the surrounding area. ;

[0027] Since the above signals originate from different subsystems such as the battery management system, wireless charging controller, and on-board charger, their sampling frequencies and transmission delays differ. The system adopts resampling technology, sets a unified sampling period, establishes a unified discrete time series index, uses downsampling processing for high-frequency data, and uses linear interpolation or zero-order hold processing for low-frequency data to align the data of all channels to the same time segment.

[0028] To suppress measurement noise in high-power electromagnetic environments, the system employs a moving average filtering algorithm to smooth the synchronized data. This is specifically designed for the temperature in the vicinity of the outer casing. With electrical stress parameters By setting a specific window length, the system performs an arithmetic average of the data within the window.

[0029] Specifically, by using the low-pass filter characteristics to filter out random spike noise caused by electromagnetic interference, due to the large thermal inertia and the fast electromagnetic fluctuations, an appropriate moving average can retain the trend information of temperature change, while smoothing out meaningless electromagnetic jitter, thus ensuring the numerical stability when calculating the temperature rise rate in the subsequent calculation.

[0030] The parameters, after synchronization and filtering, are packaged to construct a standardized synchronization state dataset. The system then processes the smoothed shell neighborhood temperature according to a predefined data protocol. Cell temperature Electrical stress parameters DC charging power The data frame is composed of the identity identifier (ID).

[0031] In another embodiment, for situations where ground transmitter data cannot be directly obtained through communication protocols, the present invention provides an alternative implementation method based on receiver data estimation. In this embodiment, electrical stress parameters... The voltage is not directly derived from the ground transmitter, but rather calculated from the electrical parameters of the vehicle-mounted receiver. The system collects the induced voltage, receiving current, and operating frequency of the vehicle-mounted receiving coil. Based on the mutual inductance coupling circuit model of the wireless charging system, and using Kirchhoff's voltage law, combined with the preset coil mutual inductance coefficient and receiver impedance parameters, it reverse-calculates the equivalent transmitter current or equivalent magnetic field strength required to maintain the current receiver voltage and current, and uses this calculated value as... Therefore, even when the communication protocol between the vehicle and the ground is not open or communication is restricted, the vehicle can still independently sense the electromagnetic transmission cost through its own data, thereby identifying special thermal conditions and enhancing the versatility and anti-interference capability of this solution.

[0032] In a preferred embodiment of the invention, the feature construction module receives the synchronization status dataset output by the data acquisition module, converts the original signal into feature indices, and updates and maintains a self-referencing historical database with self-learning capabilities. First, it receives data frames and performs differential processing on the temperature data in the data frames to obtain the temperature rise rate. Then, it constructs thermal feature indices. The specific process is as follows:

[0033] Set a computation time window (e.g., a computation time window contains m sampling periods, corresponding to 30 seconds), and read the shell neighborhood temperature at the current time k and the past time km. and the cell's representative temperature Then, calculate using the following formula. :

[0034]

[0035] In the formula, the numerator represents the temperature rise of the casing side within the time window, and the denominator represents the temperature rise of the cell side within the same time window. The thermal characteristic index, also known as the shell-heat reversal strength, is a dimensionless scalar that quantifies the dominance of external heat flux caused by external electromagnetic eddy current losses relative to internal heat sources caused by cell internal resistance during battery pack temperature rise. Specifically, in actual calculations, if the denominator of the above formula is zero or close to zero (e.g., before cell temperature rise occurs at the beginning of charging), then... To ensure the executability of the operation, the system will execute division-by-zero protection logic. Methods that can be used include: replacing the denominator with a preset, extremely small positive number (such as...). );

[0036] It should be noted that the above formula for calculating the heat reversal strength of the outer casing is only a preferred implementation method for calculating the dynamic change relationship. Any calculation method that can quantify the difference between the temperature rise trend of the outer casing and the temperature rise trend of the battery cell, such as calculating the difference in the temperature rise rate, the ratio of the slope fitting coefficient, or the dynamic cross-correlation coefficient, is covered within the scope of the dynamic change relationship.

[0037] Specifically, the difference in heat source location is amplified by utilizing a ratio relationship. Under normal charging conditions, heat is conducted from the inside out, and the temperature rise of the battery cell is usually greater than or close to the temperature rise of the casing. The value is small and stable. However, when the casing heats up and reverses due to compensation current rise, the external electromagnetic loss directly heats the casing, causing the numerator to increase sharply while the denominator changes little due to thermal hysteresis. The value of the absolute temperature shows a significant leap. By constructing a dimensionless ratio, the influence of the ambient base temperature (such as the difference between winter and summer temperatures) on the absolute temperature value is eliminated, enabling the algorithm to keenly capture the dynamic trend of heat flow in the opposite direction (from the outside to the inside).

[0038] It should be noted that the casing reversal caused by compensation current boost refers to the situation during high-power wireless charging of electric vehicles. When the coil coupling efficiency decreases due to reasons such as vehicle parking misalignment, but the system still tries to maintain the target power output, the wireless charging system forces the coil current to increase abnormally to compensate for transmission losses (i.e., compensation current boost). This high-frequency, high-current induces strong eddy current losses and hysteresis losses in the metal protective plate or shielding layer at the bottom of the battery pack. As a result, the temperature rise rate of the battery pack casing or bottom plate area is significantly higher than that of the internal cells. This exhibits the phenomenon of heat being conducted from the outside to the inside and the electrical transmission cost (such as the ratio of the square of the coil current to the received power) increasing abnormally.

[0039] Simultaneously, the system utilizes electrical stress parameters With DC charging power Construction of electrical characteristic index The specific process is as follows:

[0040] The system reads the data at the current time k and calculates using the following formula. :

[0041]

[0042] In the formula, To compensate for the rise current intensity, also known as the electrical characteristic index, the rise current intensity is a dimensionless scalar that quantifies the electrical stress cost borne by the wireless charging system in order to deliver a unit of effective power to the battery during the wireless charging process. Specifically, under ideal alignment coupling conditions, this index is usually kept at a low and stable level. However, under the condition of the casing heating up and reversing due to the rise current compensation, the coupling efficiency decreases, and the system forces the coil current to increase significantly in order to maintain the target power output. This causes the ratio to rise sharply and nonlinearly, reflecting that the current charging process is in a state of high electromagnetic loss and low transmission efficiency, indicating that the external structural components will face a serious risk of eddy current heating.

[0043] In detail, due to Characterizing current or electromagnetic stress, its square term is proportional to Joule heat loss. Therefore, This reflects the risk of heat loss at the transmission end of the wireless charging system in order to obtain a unit of effective charging power. Under the compensated current boost condition caused by alignment misalignment, the power obtained by the battery... The power remains unchanged or decreases slightly, but in order to maintain this power, It will rise significantly, leading to The numerical nonlinearity increases sharply. The compensated boost current intensity successfully decomposes the difference between "high-power charging" and "inefficient charging". During normal high-power charging, although the current is large, the received power is also large, and the ratio is relatively stable. Only under inefficient (compensated boost current) conditions will the ratio become abnormally prominent, thus providing high signal-to-noise ratio data for fault identification.

[0044] It should be noted that the construction of electrical characteristic indices is not limited to ratios; it can also be a weighted difference between electrical stress parameters and power, a nonlinear function mapping, etc.

[0045] The system uses identity identifiers (IDs) to archive and retrieve the two calculated feature indices, realizing the memory and adaptive functions of artificial intelligence, as detailed below:

[0046] The system builds, updates, and maintains a non-volatile database, namely a self-referenced historical database. This database uses the identity ID (i.e., "vehicle ID + ground loop ID") as the unique index key. The system performs the following bidirectional operations:

[0047] Write operation: Pair the currently calculated feature indices with... and the current power status As a record tuple, it is stored in the historical data bucket corresponding to the index key.

[0048] Search operation: Based on the current DC charging power In the historical data bucket with the same index key, all historical records with power values ​​within the power tolerance range are filtered out to form the historical baseline set H;

[0049] It should be noted that the power tolerance range refers to the DC charging power detected at the current moment. Using the central reference, extend the preset amplitude to the upper and lower sides of the numerical axis respectively. The constructed closed interval is specifically as follows: In the formula, This represents the measured value of DC charging power collected by the system at the current time step. This represents the preset power matching tolerance or one-sided screening bandwidth. Because the thermal loss characteristics of wireless charging systems (especially the heat generated by eddy currents on the base plate) are highly correlated with and have a non-linear relationship with the transmission power, mixing current high-power conditions (e.g., 11kW full load) with historical low-power conditions (e.g., 3kW trickle charging) during historical comparisons will lead to baseline inaccuracies. Therefore, it is necessary to set... To define a neighborhood of similar power, only historical samples with similar power are allowed to participate in the comparison. The numerical setting usually takes into account the measurement error range of the vehicle power sensor and the normal power fluctuation range during the charging process. It is preferably set to 0.5kW to 1.0kW, which can effectively eliminate heterogeneous power samples, ensure the scientific nature of the comparison under the same working conditions, and avoid insufficient number of historical samples due to the narrow range.

[0050] This module establishes a statistical baseline for vehicles, charging locations, and charging power. Because different vehicle models have varying chassis materials and the aging levels of ground coils at different charging locations differ, using a global threshold is prone to misjudgment. By locking the vehicle's identification ID, the system essentially compares today's vehicle with those that have been charged at the same location in the past, achieving adaptive learning without manual calibration. As the number of charging cycles increases, the richer the sample data in the self-referenced historical database, the more accurate the system's understanding of normal thermal characteristics under specific conditions becomes, effectively suppressing false alarms caused by vehicle aging or differences in facilities.

[0051] It should be noted that regarding the selection of the calculation time window, the differential time window used to calculate the temperature rise should not be too short. This is because heat transfer has inertia, and an excessively short time window (e.g., 1 second) will cause the temperature rise value to be overwhelmed by sensor quantization noise. Preferably, the calculation time window should range from 30 to 120 seconds, or a sliding window slope fitting method can be used instead of a simple two-point difference to obtain a more stable temperature rise rate ratio. Regarding the data-driven grouping strategy of the self-referenced historical database, power grouping is necessary to ensure the fairness of the comparison. This is because the heating characteristics of the coil are significantly different at different power levels (e.g., 3kW trickle charging and 100kW fast charging). The system does not need to preset fixed power levels, but instead adopts a data-driven dynamic bucketing strategy: when new data enters, if a record with a power difference less than a set threshold (e.g., 5kW) already exists in the database, it is assigned to that group; otherwise, a new power group is created. This ensures that the samples in the historical baseline set H are always comparable to the current operating conditions.

[0052] In another embodiment, to address the "cold start" problem—that is, when a vehicle first enters a charging station and there is no historical data corresponding to that identity ID in the historical reference database—this invention provides an alternative initialization method based on a universal model. In this embodiment, if the historical baseline set H is found to be empty, the system will call a pre-set universal vehicle benchmark library. This benchmark library is based on data obtained from similar vehicles on a standard test bench and pre-stores characteristic index distribution data of the same battery pack model measured on a laboratory calibration bench at different offset distances (e.g., X-axis offset 0 to 10 cm). The system will then match the current characteristic index with... The data is compared with normal alignment group data in a universal benchmark library. Once the vehicle completes its first full charge at this location, the system immediately generates a unique record for that vehicle and automatically switches back to a mode that prioritizes the use of the self-referenced historical database during the next charge. This implementation ensures that the system has basic thermal management capabilities at any stage of its lifecycle (including initial use), while retaining the ability to evolve personalized over long-term use.

[0053] In a preferred embodiment of the present invention, the operating condition discrimination module receives the current feature index and historical baseline set output by the feature construction module, and addresses the discrimination failure caused by individual vehicle differences, seasonal environmental changes, and facility aging. By identifying the relative position of the current operating condition in the historical statistical distribution, it achieves highly robust operating condition determination. The specific process is as follows:

[0054] Receive current characteristic indices, including shell preheating reversal strength. and compensation upflow intensity The system also includes the corresponding historical baseline set H. This set H contains n sets of historical records accumulated by the vehicle at the same charging location and with similar power levels. The system sorts the thermal characteristic indices and electrical characteristic indices in descending order within the database to obtain their relative rankings.

[0055] Thermal sorting: This will sort the current... Sort all historical thermal characteristic index values ​​in set H in descending order (i.e., the larger the value, the more severe the crustal inversion, and the higher the ranking). Record the current... The position number in this sequence is marked as the thermal reversal ranking. For example, if the current value is greater than all values ​​in the historical database, then =1.

[0056] Electrical sorting: Similarly, sort the current... Sort all historical electrical characteristic index values ​​in set H in descending order (i.e., the larger the value, the higher the transmission cost and the higher the ranking). Record the current... The position number in this sequence is marked as the electrical cost ranking. ;

[0057] It should be noted that relative ranking is used to eliminate environmental baseline drift: for example, during the high temperatures of summer, the shell temperature generally rises faster, and the overall historical database... The levels will all rise. If fixed judgment conditions are used (such as...) >2.0), which may lead to frequent false alarms during the summer. However, using a sorting method, as long as the current... No significant improvement over historical summer rankings (i.e., rank) If a vehicle is not ranked high (e.g., 50th), the system will not make a misjudgment, thus achieving the normalization of the judgment logic. This eliminates the need for manual calibration of vehicles in different seasons or with different degrees of aging, significantly reducing the calibration cost of algorithm deployment and demonstrating the system's adaptability to environmental changes.

[0058] It should be noted that thermal sorting and electrical sorting are one way to determine the relative deviation, but are not limited to sorting methods. They can also be methods such as calculating the Euclidean distance, Mahalanobis distance, etc., between the current feature index and the historical feature distribution center.

[0059] After obtaining the rankings in two independent dimensions, the system calculates a joint score J to quantify the degree to which the two operating conditions of "shell heating first" and "compensation upflow" occur simultaneously.

[0060] The system calculates J using the following formula:

[0061]

[0062] In the formula, the joint score J equals the sum of the thermal reversal ranking and the electrical cost ranking;

[0063] Specifically, by utilizing addition operations to construct a soft-decision mechanism based on "AND" logic, the system accurately captures the operating condition where the casing heats up first and then reverses due to compensation for the rising current. For example, in scenario one (ambient heat only): if the casing heats up only due to sun exposure, then... It may be extremely small (e.g., 1), but at this point the coil alignment is good and the transmission cost is low. The current will be very high (e.g., 50), resulting in J=51, a relatively large value, indicating a non-outer casing overheating condition. Scenario 2 (Coil aging only): If the slightly higher current is only due to coil aging, then... The value may be small (e.g., 5), but it did not cause the shell to invert. The value will be very large (e.g., 60), resulting in J=65, which is a relatively large value, and is judged as a non-outer shell heating condition. Scenario 3 (outer shell heating condition): Only when "compensation rise current" directly causes "outer shell heating" will both be simultaneously at the top of the historical rankings (e.g. =1, =2), thus obtaining a very small J value (such as 3), accurately locking the shell heating condition. This scoring mechanism effectively eliminates the interference of single-factor anomalies (such as simple environmental thermal interference or simple electrical fluctuations). Only when the cause and effect show a strong statistical correlation can the judgment threshold be passed, thereby greatly reducing the false alarm rate and improving the system's ability to identify true positive samples.

[0064] The system dynamically sets the discrimination boundary based on the data size of the historical baseline set and outputs the final discrimination result, as follows:

[0065] The system counts the total number of valid samples, n, in the historical baseline set H. Then, it calculates the adaptive frontier size. This is used as an adaptive threshold for judging abnormal operating conditions. Its calculation process involves rounding up the arithmetic square root of the total number of samples, i.e. Next, the system will combine the current joint score J with the adaptive frontier size. Comparison:

[0066] like If the system enters an abnormal operating condition dominant state, the abnormal operating condition flag will be set. Set to 1 (true). Specifically, the abnormal operating condition is the case where the casing heats up and reverses direction due to the compensation current rise.

[0067] like If the system is in normal operating condition, the abnormal operating condition flag will be set. Set to 0 (false).

[0068] Finally, the working condition discrimination module will J is output to the execution module.

[0069] Specifically, the stringency of the judgment is dynamically adjusted as data accumulates: in the early stages of system operation (e.g., n=9). =3, the discrimination is relatively lenient, allowing for quick capture of anomalies; as the runtime increases (e.g., n=100), the exceptions will become more complex. =10. Although the absolute value has increased, the statistical probability of simultaneously ranking in the top 10 out of 100 samples is much lower than ranking in the top 3 out of 9 samples. This means that by adopting a data-driven adaptive boundary, the consistency problem in the process from small sample startup to large sample stable operation has been solved, ensuring the reliability of the discrimination throughout the entire life cycle, avoiding the jitter of the discrimination criteria caused by changes in the amount of data, and as historical data becomes richer, the system's definition of abnormal operating conditions will become more and more convergent and accurate.

[0070] It should be noted that regarding the lower bound constraint on the total number of samples n and the cold start strategy, in order to prevent invalid judgments when historical data is extremely scarce (e.g., n=1), the system sets a minimum number of samples for starting. When the total number of samples in the historical baseline set does not reach this minimum starting number, the system will not trigger a judgment by default, or will use a preset conservative safety threshold as a temporary fallback, while only recording and accumulating data will be performed until the historical database meets the statistically significant minimum size.

[0071] It should be noted that, in order to avoid The signal frequently jumps at the critical point. In practical engineering, hysteresis comparison logic or time filtering logic can be introduced for the J value. For example, the J value must meet the judgment condition for several consecutive sampling periods before the signal is changed. Set it to the valid state; or add a buffer margin to the exit condition, and reset the flag only when the J value is significantly greater than the boundary value.

[0072] It should be noted that the joint score J value essentially characterizes the degree of deviation of the current operating point from the normal historical cluster center in the thermoelectric joint feature space. Although this embodiment uses a simplified rank summation method, in another embodiment where computational resources permit, the Mahalanobis distance from the current vector to the historical point cloud can be used instead of the J value, which can also achieve the identification of anomalous outliers.

[0073] Among them, the management execution module dynamically reconstructs the path and control logic of the thermal management system according to the current operating conditions to solve the limitation of the traditional battery thermal management system (BTMS) which only uses the cell internal resistance heating model for temperature control. Through the active intervention of the actuator, the heat flow caused by external electromagnetic induction is intercepted without interrupting charging.

[0074] First, read the operating condition flags. ,like =0, indicating that the system is in normal operating condition, and the execution module maintains or enters the cell temperature control priority mode;

[0075] like =1, indicating that the system is in an abnormal operating condition (in this example, the abnormal operating condition is the case heat reversal caused by compensation upflow), and the module immediately switches to the case heat flow interception mode.

[0076] When the system determines that the current operating condition is normal (including normal conductive charging, low-power wireless charging, or temperature rise caused only by the environment), the execution module maintains the standard control strategy based on the cell internal resistance heating model.

[0077] Specifically, the thermal management controller will represent the temperature of the battery cell. As the main feedback variable in the PID (Proportional-Integral-Derivative) control loop, the temperature uniformity within the battery pack is used as the optimization objective. The control algorithm calculates... The deviation from the preset optimal operating temperature of the battery cell (e.g., 25 to 35 degrees Celsius) is considered, and the speed of the electric water pump and the power of the compressor are adjusted accordingly. At this time, the coolant flow channel distribution valve is at a balanced opening, ensuring that the side and bottom flow channels of the battery module receive uniform heat dissipation flow. Thus, in the absence of significant external electromagnetic heat source interference, the uniformity of the battery cell temperature is prioritized, preventing local overheating or undercooling of the battery cells, which is beneficial to extending the battery's electrochemical life.

[0078] once When set to 1, the execution module immediately sends a path rearrangement command to the multi-way proportional valve and electric water pump in the coolant circulation loop, thereby changing the flow direction distribution of the heat exchange medium.

[0079] Specifically, the path reordering instruction drives the actuator to perform the following actions:

[0080] Flow channel weight tilt: Control the multi-way proportional valve to adjust the valve opening of the branch flowing through the bottom liquid cooling plate of the battery pack to the maximum (100%), while proportionally reducing the valve opening of the branch flowing through the side of the module or non-critical areas (e.g., reducing it to 30%).

[0081] Active flow surge: The electronic water pump is instructed to break through the conventional energy-saving speed curve and operate at rated maximum power or overclocking mode, forcibly increasing the flow rate of coolant through the base plate area to improve the convective heat transfer coefficient.

[0082] In detail, under abnormal operating conditions of compensated upflow, the heat mainly originates from the eddy current loss of the base plate metal. By prioritizing the increase of the base plate flow rate, the high-flow-rate coolant carries away the Joule heat accumulated in the base plate, effectively building a heat-absorbing wall between the external high-temperature base plate and the internal low-temperature battery cell. This cuts off the path of heat conduction upward to the battery cell, thus achieving precise, targeted heat dissipation. The limited cooling capacity is concentrated and delivered to the base plate area with the highest heat load, avoiding the overall low heat dissipation efficiency caused by the coolant being wasted in non-hot areas (such as the sides). This also effectively prevents the thermally conductive structural adhesive on the base plate from aging and failing due to long-term high temperatures.

[0083] Simultaneously, the execution module modifies the closed-loop feedback logic of the thermal management control algorithm and performs feedback rearrangement.

[0084] Specifically, the PID controller of the thermal management control system has undergone the following changes:

[0085] Main control variable switching: The input feedback source of the PID controller is represented by the temperature of the battery cell. Instantly switch to the temperature of the surrounding area of ​​the outer shell. .

[0086] Setpoint change: The target setpoint of the PID controller is switched from the cell's optimal temperature (e.g., 35°C) to the casing safety threshold (e.g., 55°C).

[0087] Control logic execution: Once It shows an upward trend and is approaching the enclosure safety threshold, regardless of the internal... Regardless of whether the temperature rises, the system determines that there is a need for heat dissipation and immediately starts the compressor for powerful cooling.

[0088] Safety constraint downgrade: original master variable It is transformed into a secondary safety constraint. That is, in control... During the process, the system continuously monitors in the background. Only when Safety protection strategies are only triggered when the limit safety value (such as 60 degrees Celsius) is exceeded. These strategies include shutdown, power outage, and alarm.

[0089] In detail, because the casing is the heat source and the battery cell is the acceptor, there is a significant thermal capacity delay between the two. If we continue to use... Control is crucial because when the cell temperature is detected to be rising, the base plate is often already at an extremely high temperature. This can be addressed through direct closed-loop control. The system eliminates thermal conduction hysteresis, achieving a zero-delay response to external heat intrusion, thus solving the hysteresis problem of traditional BMS in wireless charging overheat protection. By monitoring the properties of the casing, it ensures that the thermal management system removes heat from the base plate before it penetrates the battery cell;

[0090] It should be noted that the core of this invention lies in, when When the value is 1, the system performs active thermal intervention rather than passive power cut-off. Existing foreign object detection (FOD) technologies typically cut off charging directly when they detect metallic foreign objects or abnormal current, which severely impacts the user experience. This solution identifies the non-faulty but inefficient "compensation current boost" condition. The system is capable of keeping the floor temperature within a safe range, allowing the vehicle to continue the wireless charging process while consuming some additional energy. This "operating with heat" capability greatly improves the availability and charging success rate of the wireless charging system. Additionally, attention must be paid to changes in the total flow resistance of the hydraulic system. Excessive closure of side branches may cause a sudden increase in pressure in the main pipeline, affecting the lifespan of the water pump. Therefore, the preferred implementation is to use an intelligent thermal management module with an electronic bypass valve. While increasing the bottom flow rate, it maintains a constant system back pressure by dynamically adjusting the opening of the bypass branch, ensuring that the flow distribution meets the floor cooling requirements without damaging the hydraulic hardware.

[0091] In another embodiment, to further improve the system's response speed to rapidly changing operating conditions, the present invention provides a pre-cooling implementation method (i.e., trend predictive control) based on the joint score J change rate. In this embodiment, the execution module not only responds to the binarized... The signal also continuously monitors the dynamic trend of the continuous value J:

[0092] Specifically, the system calculates the rate of change of the joint score over time (i.e., the first derivative) in real time. As defined in this invention, the higher the degree of abnormality, the smaller the nominal value ( and The closer the value of J is to 1, the smaller the joint score J becomes. Therefore, when the system condition is deteriorating, the J value will show a decreasing trend, i.e., the rate of change... The value is negative. To accurately identify the precursory stage of the shell thermal reversal that is rapidly approaching an anomaly, the system presets a trend judgment threshold. (For example This threshold is used to distinguish between normal parameter fluctuations and deteriorating trends. When the joint score J is still greater than the adaptive threshold... (Right now =0, the system has not yet met the final judgment condition for abnormal operating conditions, but when the following two trend conditions are met simultaneously: rate of change (This indicates that the score is declining and the operating condition is developing in an abnormal direction); the absolute value of the rate of change. (This indicates an extremely rapid rate of descent, suggesting that the safety threshold will be breached within a short time); the system determines that it is currently in the precursor stage of a thermal reversal of the casing. Once the precursor stage is determined, the management execution module does not wait. Instead of flipping over, pre-cooling operations are performed immediately, including increasing the base speed of the electric water pump in advance (e.g., from 2000 rpm to 3500 rpm) or opening the active air intake grille (AGS) in advance.

[0093] This derivative-based control strategy utilizes the continuously changing nature of the J value to fill the time dead zone caused by binarization, thus smoothing the intervention process of the thermal management system and avoiding [further issues]. To prevent excessive hydraulic shock or noise during sudden transitions, a smoother, more seamless intervention is achieved.

[0094] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An artificial intelligence-based energy storage battery thermal management system, characterized in that, include: The data acquisition module acquires the ambient temperature of the battery pack's outer casing, the representative temperature of the battery cells inside the battery, electrical stress parameters that quantify the cost of wireless charging transmission, DC charging power, and an identification identifier containing charging location information. The feature construction module determines the thermal characteristic index based on the dynamic change relationship between the temperature of the outer casing neighborhood and the representative temperature of the cell within a time window, and determines the electrical characteristic index based on the correspondence between electrical stress parameters and DC charging power. The feature construction module is also used for: Update the self-referenced historical database indexed by identity identifier, as follows: The identity identifier is used as the index key of the self-referenced history database, which is a non-volatile storage structure. The current thermal characteristic index, electrical characteristic index, and DC charging power are combined into a record tuple, and the following dynamic grouping strategy is executed using a preset power determination threshold to store the record tuple in the self-referenced historical database: Calculate the difference between the current DC charging power and the power value of the existing record tuple under this index key in the self-referenced historical database; If there is a record tuple in the reference historical database whose absolute value of the difference is less than the power determination threshold, then the current record tuple is assigned to the power group to which the existing record tuple belongs. If no record tuple that satisfies the difference condition exists in the reference historical database, a new power group is established based on the current DC charging power, and the current record tuple is stored in the new power group. The operating condition discrimination module retrieves the historical baseline set based on the identity identifier, determines the relative deviation of the current thermal characteristic index and electrical characteristic index from the historical baseline set, and determines whether the system is in the shell-heating-reversal operating condition based on the relative deviation. The operating condition discrimination module is specifically used for: Based on the identity identifier, the corresponding storage bucket in the self-referenced historical database is entered. A preset power tolerance range is set with the current DC charging power as the center. All historical data tuples in the storage bucket whose power values ​​fall within the power tolerance range are retrieved and formed into a historical baseline set. Thermal characteristic indices are extracted from each record tuple in the historical baseline set to construct a historical thermal sequence; Add the current thermal characteristic index to the historical thermal sequence and sort it in descending order to determine the position number of the current thermal characteristic index in the sequence, which is used as the thermal reversal rank. Electrical characteristic indices are extracted from each record tuple in the historical baseline set to construct a historical electrical sequence; Add the current electrical characteristic index to the historical electrical sequence and sort it in descending order to determine the current position number of the electrical characteristic index in the sequence, which is used as the electrical cost ranking. The descending order arrangement rule is such that the smaller the position number, the larger the corresponding feature index value, thereby reflecting the positive deviation of the feature index from the historical baseline data. The management execution module is used to perform thermal management reconfiguration when it is determined that the casing is in a reverse heating condition. It switches the main feedback target of the thermal management control loop from the cell representative temperature to the casing neighborhood temperature, and adjusts the cooling strategy for the battery pack based on the switched main feedback target.

2. The artificial intelligence-based energy storage battery thermal management system according to claim 1, characterized in that, include: Set a calculation time window that includes multiple sampling periods, and calculate the difference between the current temperature of the shell neighborhood and the temperature of the shell neighborhood at the beginning of the calculation time window, as the shell temperature rise rate. The difference between the current cell temperature and the cell temperature at the start of the calculation time window is used as the cell temperature rise rate. The ratio of the temperature rise of the outer casing to the temperature rise of the battery cell is denoted as the thermal characteristic index; Calculate the square of the electrical stress parameter at the current moment, and label the ratio of the square to the DC charging power as the electrical characteristic index.

3. The artificial intelligence-based energy storage battery thermal management system according to claim 1, characterized in that, include: The algebraic sum of the thermal reversal ranking and the electrical cost ranking is used as the joint score; The total number of record tuples in the historical baseline set is counted, the arithmetic square root of the total number of record tuples is calculated, and the result is rounded up to obtain an integer value, which is used as the adaptive threshold. The joint score is compared with the adaptive threshold. When the joint score is less than or equal to the adaptive threshold, the system is determined to be in the shell-heating-reverse condition.

4. The artificial intelligence-based energy storage battery thermal management system according to claim 3, characterized in that, include: Calculate the rate of change of the joint score over time; When the value of the joint score is greater than the adaptive threshold, but the rate of change is negative and the absolute value of the rate of change is greater than the preset trend judgment threshold, the system is in the precursor stage of the shell heating reversal, and the base speed of the electronic water pump is increased in advance or the active air intake grille is opened.

5. The artificial intelligence-based energy storage battery thermal management system according to claim 3, characterized in that, include: When the historical baseline set is found to be empty, a pre-set universal vehicle benchmark library is invoked. The universal vehicle benchmark library contains characteristic index distribution data of the same type of vehicles measured on a standard test bench at different alignment offset distances. The current thermal and electrical characteristic indices are compared with normal alignment group data in the general vehicle benchmark library to determine whether the system is in the condition of the outer shell heating up and then reversing.

6. The artificial intelligence-based energy storage battery thermal management system according to claim 1, characterized in that, include: The electrical stress parameter is a scalar that quantitatively characterizes the electrical cost that a wireless charging system incurs to maintain the target power transmission. The electrical stress parameters are selected as the effective value of the wireless charging coil current, or as a weighted sum calculated based on the coil current intensity and the phase difference between the primary and secondary currents; Specifically, when the weighted sum is used, the coil current intensity and the phase difference are multiplied by their respective weighting coefficients and then summed to obtain the result. The weighting coefficients are pre-calibrated based on the electromagnetic coupling characteristics of different vehicle models and are used to balance the contribution of current amplitude and phase deviation to heat loss.

7. The artificial intelligence-based energy storage battery thermal management system according to claim 1, characterized in that, include: The casing neighborhood temperature refers to the temperature value of the electromagnetic thermal effect sensitive area located at the bottom of the battery pack; The electromagnetic thermal effect sensitive area is a pre-calibrated area at the bottom of the battery pack where the eddy current density is the highest or the temperature rise rate is the fastest. The temperature sensor is attached to the inner surface of the metal base plate in the area or integrated into the interlayer of the anti-collision shell to thermally couple to the structural component that is heated by the external magnetic field. The identification tag is a combined data tag containing vehicle information and charging environment information, which is composed of the vehicle's unique identification code and the unique identification code of the ground transmission facility; When the unique identification code of the ground launch facility cannot be obtained directly through the communication protocol, the location coordinate hash value of the ground facility or the ground identification information identified by the vehicle-mounted vision sensor is used as a substitute for the ground launch facility identification code.

8. The artificial intelligence-based energy storage battery thermal management system according to claim 1, characterized in that, The specific methods for implementing thermal management reconfiguration include: The controller's input feedback source was switched from the cell-represented temperature to the temperature of the surrounding casing. Switch the controller's target setpoint from the cell's optimal operating temperature to a preset casing safety threshold. The cell-represented temperature is transformed into a secondary safety constraint. During the control process with the temperature of the surrounding area of ​​the casing as the main feedback target, the cell-represented temperature is continuously monitored. When the cell-represented temperature exceeds the preset limit safety value, a safety protection strategy is triggered.

Images