BMS temperature control method and system

Abstract

The invention relates to the technical field of temperature control, and discloses a BMS temperature control method and system, and the method comprises the steps: carrying out the multi-dimensional coupling of battery data, and obtaining a micro-region thermal state; predicting heat accumulation and generating a heat dissipation instruction based on load abrupt change superposition feedforward compensation; calculating heat distribution deviation evaluation stability according to the feedback signal; defining a lead compensation range by using a heat evolution model to generate parameter configuration; system response characteristics are obtained through calibration of the frequency spectrum and the state transition matrix; and fusing the temperature gradient distribution to generate a final advanced adjustment instruction. According to the method, accurate prediction and full-link advanced active intervention of micro-region thermal unbalance can be achieved, the problem of heat accumulation caused by load dramatic change is effectively solved through multi-dimensional feature coupling and a dynamic closed-loop calibration mechanism, and the system response speed and the temperature field balance are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of battery management technology, and in particular to a BMS temperature control method and system. Background Technology

[0002] With the rapid development of new energy vehicles and energy storage systems, battery management systems are facing increasingly stringent requirements for controlling battery thermal safety. Under the complex operating conditions of high-rate charging and discharging, the heat generation inside the battery exhibits significant nonlinearity, time-varying characteristics, and spatial variability.

[0003] Current technologies primarily rely on single-point temperature sensor data acquisition and PID feedback regulation. This approach often suffers from significant lag, initiating heat dissipation only after the temperature sensor detects a temperature rise. Under conditions of drastic load fluctuations, such as rapid acceleration or fast charging, this lag causes heat to accumulate rapidly in micro-regions within the secondary battery, forming hotspots before external cooling devices respond. Furthermore, existing control strategies are typically based on the overall average battery temperature, lacking the ability to perceive the microscopic heat distribution within the battery and thus failing to precisely target and dissipate heat to specific overheated areas. Simultaneously, traditional algorithms struggle to accurately capture the system's response delays and environmental disturbances, leading to control oscillations during steady-state and transient transitions, impacting battery life and safety.

[0004] Therefore, existing technologies suffer from the problem of being unable to proactively compensate for and dynamically calibrate thermal accumulation in micro-regions of the battery due to the lag in temperature control response. Summary of the Invention

[0005] The purpose of this invention is to provide a BMS temperature control method and system to solve the problem in the prior art that the delayed temperature control response makes it impossible to perform advanced compensation and dynamic calibration for thermal accumulation in the micro-regions of the battery.

[0006] In a first aspect, the present invention provides a BMS temperature control method, comprising: The battery's current and temperature data are acquired, and multidimensional feature coupling and grid projection are performed on the current and temperature data to obtain the thermal state distribution results of the battery's micro-regions. Based on the thermal state distribution results, combined with environmental disturbance factors and using dynamic thermal conductivity coefficient for trend prediction, the predicted value of heat accumulation is obtained. The difference between the predicted cumulative heat value and the preset thermal balance threshold is calculated to obtain the load change amplitude and system response delay. A feedforward compensation signal is generated based on the load change amplitude and the system response delay. A basic heat dissipation command is generated through a PID control algorithm. The feedforward compensation signal and the basic heat dissipation command are fused to obtain a heat dissipation control command. Drive the heat dissipation device according to the heat dissipation control command, collect the feedback signal intensity, calculate the heat distribution deviation according to the feedback signal intensity and compare it with the preset steady-state threshold to obtain the stability evaluation result; If the stability assessment result is a stable state, then the dynamic error correction amount is calculated based on the feedback signal strength, the advance compensation range is defined according to the preset predicted heat model, and the thermal management parameter configuration is obtained by combining the dynamic error correction amount and the advance compensation range. Based on the thermal management parameter configuration, a frequency spectrum distribution matrix is ​​constructed and a state transition matrix is ​​generated. The signal prediction offset is extracted, and the thermal management parameter configuration is calibrated according to the signal prediction offset to obtain the system response characteristics. If the system response characteristics meet the preset load drastic change conditions, a temperature gradient distribution matrix is ​​constructed based on the thermal state distribution results. The temperature gradient distribution matrix and the system response characteristics are then fused to obtain a regional heat flux matrix. The regional heat flux matrix is ​​then weighted and adjusted to obtain the final adjustment instruction.

[0007] Secondly, the present invention provides a BMS temperature control system, comprising: The thermal distribution module is used to acquire the current data and temperature data of the battery, and to perform multi-dimensional feature coupling and grid projection on the current data and temperature data to obtain the thermal state distribution results of the battery micro-region. The heat prediction module is used to predict the cumulative heat value by combining the thermal state distribution results with environmental interference factors and using the dynamic thermal conductivity coefficient to make trend prediction. The instruction adjustment module is used to calculate the difference between the predicted heat accumulation value and the preset heat balance threshold, obtain the load change amplitude and system response delay, generate a feedforward compensation signal based on the load change amplitude and the system response delay, generate a basic heat dissipation instruction through a PID control algorithm, and fuse the feedforward compensation signal and the basic heat dissipation instruction to obtain a heat dissipation control instruction. The steady-state evaluation module is used to drive the heat dissipation device according to the heat dissipation control command, collect the feedback signal intensity, calculate the heat distribution deviation according to the feedback signal intensity and compare it with the preset steady-state threshold to obtain the stability evaluation result. The parameter configuration module is used to calculate the dynamic error correction amount based on the feedback signal intensity if the stability assessment result is a stable state, define the advance compensation range according to the preset predicted heat model, and obtain the thermal management parameter configuration by combining the dynamic error correction amount and the advance compensation range. The signal calibration module is used to construct a frequency spectrum distribution matrix and generate a state transition matrix based on the thermal management parameter configuration, extract the signal prediction offset, calibrate the thermal management parameter configuration according to the signal prediction offset, and obtain the system response characteristics. The instruction generation module is used to construct a temperature gradient distribution matrix based on the thermal state distribution results if the system response characteristics meet the preset load drastic change conditions, fuse the temperature gradient distribution matrix with the system response characteristics to obtain a regional heat flow matrix, and adjust the regional heat flow matrix by weight to obtain the final adjustment instruction.

[0008] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the BMS temperature control method described in any one of the above.

[0009] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to perform the BMS temperature control method described in any one of the above.

[0010] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention transforms macroscopic current and temperature data into microscopic thermal state distribution in micro-regions through multidimensional feature coupling and grid projection technology, breaking through the spatial limitations of traditional single-point monitoring and realizing accurate positioning of hot spots inside the battery.

[0011] (2) This invention introduces a trend prediction mechanism based on environmental interference factors and dynamic thermal conductivity coefficient, and combines load mutation and system response delay to generate feedforward compensation signal, realizing the transformation from post-event remediation to pre-event prevention, effectively overcoming the physical lag of thermal management system.

[0012] (3) In the stable state of the system, the present invention uses the heat evolution model for advance compensation, and uses wavelet transform and temperature gradient matrix fusion for final adjustment under the preset load drastic conditions. Through this hierarchical dynamic calibration mechanism, the temperature field balance and response speed of the BMS under all working conditions are guaranteed. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the BMS temperature control method provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the BMS temperature control system provided in the second embodiment of the present invention. Detailed Implementation

[0014] 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.

[0015] Reference Figure 1 The first embodiment of the present invention provides a BMS temperature control method, including the following steps: S1, acquire the current data and temperature data of the battery, perform multidimensional feature coupling and grid projection on the current data and temperature data to obtain the thermal state distribution results of the battery micro-region; S2. Based on the thermal state distribution results, combined with environmental interference factors and using the dynamic thermal conductivity coefficient to predict trends, the predicted value of heat accumulation is obtained. S3, calculate the difference between the predicted cumulative heat value and the preset thermal balance threshold to obtain the load change amplitude and system response delay, generate a feedforward compensation signal based on the load change amplitude and the system response delay, generate a basic heat dissipation command through a PID control algorithm, and fuse the feedforward compensation signal and the basic heat dissipation command to obtain a heat dissipation control command. S4, drive the heat dissipation device according to the heat dissipation control command, collect the feedback signal intensity, calculate the heat distribution deviation according to the feedback signal intensity and compare it with the preset steady-state threshold to obtain the stability evaluation result; S5. If the stability assessment result is a stable state, then the dynamic error correction amount is calculated based on the feedback signal strength, the advance compensation range is defined according to the preset predicted heat model, and the thermal management parameter configuration is obtained by combining the dynamic error correction amount and the advance compensation range. S6. Based on the thermal management parameter configuration, construct a frequency spectrum distribution matrix and generate a state transition matrix, extract the signal prediction offset, calibrate the thermal management parameter configuration according to the signal prediction offset, and obtain the system response characteristics; S7. If the system response characteristics meet the preset load drastic change conditions, then a temperature gradient distribution matrix is ​​constructed based on the thermal state distribution results. The temperature gradient distribution matrix and the system response characteristics are fused to obtain a regional heat flow matrix. The regional heat flow matrix is ​​then weighted and adjusted to obtain the final adjustment instruction.

[0016] In step S1, multidimensional feature coupling and grid projection are performed on the current data and the temperature data to obtain the thermal state distribution results of the battery micro-region, including: S11, acquire the battery's current data and temperature data to obtain a current dataset and a temperature dataset; perform feature extraction on the current dataset and the temperature dataset to obtain current feature data and temperature feature data; S12, perform multi-dimensional feature coupling processing on the current feature data and the temperature feature data to obtain a multi-dimensional fused feature matrix; S13, calculate the heat accumulation rate based on the multidimensional fusion feature matrix, and project the heat accumulation rate onto the preset grid model to obtain the thermal state distribution result of the battery micro-region.

[0017] In step S11, the current data and temperature data of the battery are acquired to obtain a current dataset and a temperature dataset; feature extraction is performed on the current dataset and the temperature dataset to obtain current feature data and temperature feature data.

[0018] It should be noted that current data is acquired using a high-precision Hall current sensor with a range of 0-500A and an accuracy of ±0.5%FS, while temperature data is acquired using an NTC thermistor sensor with a range of -40℃ to 125℃ and an accuracy of ±0.3℃. Both types of sensors are acquired synchronously at a sampling frequency of 10Hz. The sampling frequency has been experimentally verified to balance dynamic capture and resource consumption. The current dataset is stored in a two-dimensional structure of timestamp-current value, and the temperature dataset is stored in a three-dimensional structure of timestamp-temperature value-sensor number. The sensor number corresponds one-to-one with the 8×6 micro-region grid of the battery. Current features are extracted as current fluctuation frequency and current change rate. The fluctuation frequency is obtained by performing FFT on 1024 sampling points to obtain the main peak frequency in the 0.1Hz-10Hz frequency band, and the change rate is calculated as the average of 10 consecutive differences ΔI / Δt (Δt=0.1 seconds). Temperature features are extracted as temperature gradient and temperature cumulative value. The gradient is calculated as the x / y axis vector ΔT / 5cm (5cm spacing between adjacent sensors), and the cumulative value is the sum of 60 consecutive sampling points. The parameters are set based on a 10Hz matching thermal response cycle, a 1024-point FFT to balance resolution and efficiency, a 5cm bonding module structure, and a 6-second window to filter instantaneous fluctuations.

[0019] For example, the current dataset has sampling values ​​of 120A, 125A, 130A, 128A, 122A, and 118A over a certain period. After FFT, the fluctuation frequency is 0.8Hz and the average rate of change is -0.2A / s. The temperature sensor (3,2) corresponds to a temperature of 32.5℃-33.5℃, with a temperature difference of 0.3℃ from the adjacent sensor, a gradient of 0.06℃ / cm, and a cumulative value of 199.0℃ over 6 seconds. Finally, the current characteristic data is 0.8Hz, -0.2A / s and the temperature characteristic data is 0.06℃ / cm, 199.0℃.

[0020] In step S12, the current feature data and the temperature feature data are subjected to multi-dimensional feature coupling processing to obtain a multi-dimensional fused feature matrix.

[0021] It should be noted that before coupling, the dimensions are eliminated through min-max normalization. Each 8×6 micro-region of the battery corresponds to a set of four-dimensional feature vectors. The normalized fluctuation frequency, rate of change, gradient, and cumulative value are stacked in the order of micro-region numbering into a 48-row, 4-column fusion matrix, and the matrix elements are the normalized feature values ​​of the corresponding regions. The parameter settings are based on the principle that normalization preserves the relative distribution, and the matrix dimensions fit the module structure and computational efficiency requirements.

[0022] For example, the normalized eigenvalues ​​of sensor number (3,2) in S11 are 0.071, 0.48, 0.12, and 0.495, and the eigenvector is [0.071, 0.48, 0.12, 0.495]. The eigenvectors of the 48 micro-regions are stacked sequentially to obtain a 48-row, 4-column multidimensional fusion feature matrix. The element in the 20th row of the matrix is ​​the vector value.

[0023] In step S13, the heat accumulation rate is calculated based on the multidimensional fusion feature matrix, and the heat accumulation rate is projected onto a preset grid model to obtain the thermal state distribution results of the battery micro-region.

[0024] It should be noted that the heat accumulation rate is derived based on the Joule heat generation principle combined with dynamic correction, and the formula is as follows:

[0025] The unit is joules per minute (J / min). Where I is the real-time average current (A), R0 is the reference internal resistance of a single battery cell (Ω), taken as 0.0005Ω to adapt to the typical internal resistance range of power lithium batteries; α is the temperature coefficient of resistance (°C). - ¹), taking 0.00393 as the characteristic of the current collector material of the battery; T_grad is the temperature gradient (°C / cm), reflecting the influence of local temperature distribution on resistance; k_cond is the current change rate correction coefficient (J). s / A 3The following parameters are used: 0.1 represents the additional heat generated by the experimentally calibrated current surge; dI / dt is the rate of change of current (A / s); η is the heat dissipation correction factor, η=1+0.01×(T_avg-T_ref), where T_avg is the average temperature (°C) corresponding to the accumulated temperature value, and T_ref is the reference temperature (°C), taken as 30°C; a coefficient of 60 is used to convert power per second (W, i.e., J / s) into heat per minute. The preset mesh model is an 8×6 battery micro-region mesh, corresponding one-to-one with the sensor number. The projection method is to directly map the calculated heat accumulation rate value of each micro-region to the corresponding coordinates of the mesh, forming an 8x6 thermal state distribution matrix. The larger the matrix element value, the faster the heat accumulation in that region. The reference internal resistance and temperature coefficient refer to lithium battery industry standards, and the correction coefficient has been calibrated through 100 sets of experiments under different operating conditions. The mesh dimension fits the actual structure of the battery module, ensuring that the calculation results are consistent with engineering practice.

[0026] For example, the original feature data corresponding to sensor number (3,2) in S11 is the real-time average current I = 123.83A, current change rate dI / dt = -0.2A / s, temperature gradient T_grad = 0.06℃ / cm, and cumulative temperature value of 199.0℃ (sum of 6 seconds of sampling), corresponding to an average temperature T_avg ≈ 33.17℃. Substituting into the formula, the heat accumulation rate is calculated to be ≈ 458.4J / min. Projecting this rate value onto the grid coordinates (3,2), the element at position (3,2) in the final thermal state distribution matrix is ​​458.4 / min, which intuitively reflects the heat accumulation situation in this micro-region.

[0027] In step S2, based on the thermal state distribution results, combined with environmental disturbance factors and using the dynamic thermal conductivity coefficient for trend prediction, a predicted value for heat accumulation is obtained, including: S21, Based on the thermal state distribution results, obtain the associated temperature flow sequence, apply preset data weights to the temperature flow sequence to obtain a weighted thermal feature sequence; S22, perform a convolution operation between the weighted thermal feature sequence and the environmental interference factor to obtain the corrected thermal feature vector; S23, based on the modified thermal feature vector, predict the dynamic thermal conductivity coefficient of the preset time window to obtain the dynamic thermal conductivity coefficient matrix, and deduce the continuous thermal change trend according to the dynamic thermal conductivity coefficient matrix to obtain the thermal change trend trajectory. S24, perform time-domain integration on the thermal change trend trajectory to obtain the predicted cumulative heat value.

[0028] In step S21, the associated temperature flow sequence is obtained based on the thermal state distribution result, and the temperature flow sequence is weighted by applying preset data weights to obtain a weighted thermal feature sequence.

[0029] It should be noted that the temperature flow sequence is historical data of the micro-region corresponding to the thermal state distribution results, including the current sequence (sampling frequency 10Hz) and temperature sequence (sampling frequency 10Hz) for the past 120 minutes, which are associated one-to-one with the thermal state distribution micro-region of S13 according to the timestamp. The processing flow is as follows: the current sequence and temperature sequence are normalized by min-max to eliminate dimensions, and the normalization range is mapped to [0,1] to ensure that the two types of features are on the same order of magnitude; the preset data weights are set according to the time decay rule, with a weight of 0.6 for the most recent 30 minutes, 0.3 for 30-60 minutes, and 0.1 for 60-120 minutes, and time weighting is applied to the normalized current sequence and temperature sequence respectively; the normalized current value and normalized temperature value are weighted and fused by using feature weights w_I=0.4 (current feature weight) and w_T=0.6 (temperature feature weight), and the normalized current value and normalized temperature value are multiplied by the corresponding weight and then added. After weighting, each micro-region corresponds to a set of one-dimensional thermal feature sequences, with a length of 120×60×10=72000 data points (consistent with the original sampling data volume). The parameter settings are based on normalization to eliminate dimensional differences, and the time decay weight is tailored to the characteristic that recent data has a more significant impact on the current thermal trend. The feature weights are experimentally calibrated to match the contribution ratio of current heating and temperature accumulation to the thermal state.

[0030] For example, if the thermal state distribution rate of the micro-region (3,2) is 482.3 J / min, after preprocessing, the normalized range of the current sequence is [0.35, 0.65], and the normalized range of the temperature sequence is [0.40, 0.70]. The normalized mean values ​​for each time period are the most recent 30 minutes (0.60, 0.65), 30-60 minutes (0.55, 0.60), and 60-120 minutes (0.50, 0.55). After time weighting and feature fusion calculation, the weighted feature value is 0.605. This value is used as the core element and arranged in the order of the original sampling timestamps to finally form a weighted thermal feature sequence of 72,000 data points.

[0031] In step S22, the weighted thermal feature sequence is convolved with the environmental interference factor to obtain the corrected thermal feature vector.

[0032] It should be noted that the environmental interference factors include ambient temperature (range -10℃ to 45℃, accuracy ±0.2℃) and air velocity (range 0 to 5m / s, accuracy ±0.1m / s). Both parameters are normalized using min-max normalization and mapped to the [0,1] interval, forming a two-dimensional vector [F1,F2]. F1 is the normalized ambient temperature, and F2 is the normalized air velocity. Subsequently, a one-dimensional discrete convolution operation is performed: V_corr=Conv(W_seq, [F1,F2]), where V_corr is the corrected thermal feature vector, and W_seq is the weighted thermal feature sequence. The weighted thermal feature sequence is convolved with the interference factor vector in one dimension, and the mean vector of the convolution result is used as the corrected thermal feature vector. The vector dimension is consistent with the number of micro-regions, which is 48 dimensions (8×6). The parameter settings are based on the fact that ambient temperature and airflow velocity directly affect heat dissipation efficiency. Convolution operation can integrate the influence of external interference on thermal characteristics. Normalization processing ensures that the interference factor and the weighted thermal characteristic sequence are on the same order of magnitude, avoiding fusion deviation caused by dimensional differences.

[0033] For example, the weighted thermal feature sequence core value of the micro-region (3,2) in S21 is 68.5. The current ambient temperature is 35℃ and the air velocity is 0.3m / s. After min-max normalization, the interference factor vector is obtained as [0.818, 0.06]. After convolution operation, the mean is taken, and the element corresponding to the micro-region (3,2) in the corrected thermal feature vector is 68.5×0.818×0.6+68.5×0.06×0.4=68.5×(0.4908+0.024)=68.5×0.5148≈35.26, finally obtaining the 48-corrected thermal feature vector.

[0034] In step S23, the dynamic thermal conductivity coefficient of a preset time window is predicted based on the modified thermal feature vector to obtain the dynamic thermal conductivity coefficient matrix. The continuous thermal change trend is then deduced based on the dynamic thermal conductivity coefficient matrix to obtain the thermal change trend trajectory.

[0035] It should be noted that the preset time window is 30 minutes, divided into 30 time nodes at 1-minute intervals. The dynamic heat transfer coefficient is calculated using the formula k = k0 × (1 - 0.005℃). - ¹×ΔT), where k0 is the basic thermal conductivity of the battery material (0.8 W / (m²)). K), ΔT is the difference (in °C) between the temperature corresponding to the corrected thermal characteristic vector and the reference temperature (25 °C), 0.005 °C. - ¹ represents the temperature decay coefficient of the battery material's thermal conductivity (experimentally calibrated to reflect the temperature-dependent thermal conductivity of lithium batteries). The dynamic thermal conductivity matrix is ​​an 8×6×30 three-dimensional matrix, corresponding to the coefficients at 30 time points for each micro-region. When extrapolating the thermal change trend, the temperature value at each time point is calculated through matrix iteration.

[0036] Where Q is the heat accumulation rate of S13, ρcV is the micro-region heat capacity of the battery (fixed at 120J / K), and Δt = 60 seconds.

[0037] For example, in S22, the corrected thermal eigenvector element 35.26 in the micro-region (3,2) corresponds to ΔT = 33.5 - 25 = 8.5℃, and the dynamic thermal conductivity k = 0.8 × (1 - 0.005 × 8.5) = 0.8 × 0.9575 = 0.766 W / (m²). K). Substituting into the iterative formula, the initial temperature is 33.5℃, the heat accumulation rate is 28.83J / min, and the temperature in the first minute is T(1)=33.5+(28.83×60) / (120)×0.766 / 60=33.5+(28.83×0.766) / 120≈33.5+0.181≈33.68℃. Iterating through 30 nodes in sequence, the thermal change trend trajectory of the (3,2) micro-region from 33.5℃ to 39.2℃ within 30 minutes is obtained, and finally an 8×6×30 thermal change trend trajectory set is formed.

[0038] In step S24, the thermal change trend trajectory is integrated over time to obtain the predicted cumulative heat value.

[0039] It should be noted that the predicted heat accumulation value is the net heat accumulation of the battery micro-region within the integration interval (the sum of the differences between heat generation and heat dissipation). The preset integration window is 30 minutes, and the integration formula is as follows:

[0040] Where Q_acc is the overall predicted heat accumulation value (in J), P is the heat accumulation rate of each micro-region in S13 (in J / min, which needs to be converted to J / s for integration, i.e., P / 60), and k(t) is the dynamic heat transfer coefficient in S23 (in W / (m²)). K), T(t) represents the real-time temperature (in K) in the thermal change trend trajectory, T_env represents the real-time ambient temperature (in K), S represents the heat dissipation area of ​​a single micro-region (25cm²=0.0025m²), δ represents the battery casing thickness (0.01m), T represents the integration time (30min=1800s), and Σ represents the summation of the integration results of all micro-regions. The integration is calculated using the trapezoidal integration method, balancing computational accuracy and efficiency. The heat dissipation-related parameters (S, δ) are aligned with the mainstream square battery module structure design, and the calculation of the difference between heat generation and heat dissipation conforms to the physical principle of thermal balance, ensuring that the predicted value reflects the true heat accumulation state.

[0041] For example, the heat accumulation rate P in the microregion (3,2) of S23 is 482.3 J / min (converted to 8.04 J / s), and the mean k(t) corresponding to the 30-minute thermal change trend trajectory is 0.766 W / (m²). The ambient temperature is 35℃ (308K), the average T(t) is 36.3℃ (309.3K), S=0.0025m², δ=0.01m. The net cumulative heat value for this region, calculated using trapezoidal integration, is 2860J. The sum of the integral values ​​for the remaining 47 micro-regions is 81540J. The overall predicted cumulative heat value is 2860 + 81540 = 84400J, or 84.4kJ.

[0042] In step S3, the difference between the predicted cumulative heat value and the preset thermal balance threshold is calculated to obtain the load mutation amplitude and system response delay. A feedforward compensation signal is generated based on the load mutation amplitude and the system response delay. A basic heat dissipation command is generated through a PID control algorithm. The feedforward compensation signal and the basic heat dissipation command are fused to obtain a heat dissipation control command, including: S31, calculate the difference between the predicted heat accumulation value and the preset heat balance threshold to obtain the heat overflow value; S32, obtain the load change amplitude corresponding to the heat overflow value, sort the channels according to the load change amplitude, and obtain the adjustment priority order for different heat dissipation channels; S33, the heat overflow value is weighted and allocated according to the adjustment priority order, and a signal is generated in combination with the system response delay to obtain a feedforward compensation signal; S34: Generate basic heat dissipation commands through a PID control algorithm, and combine the feedforward compensation signal and the basic heat dissipation commands to obtain heat dissipation control commands.

[0043] In step S31, the difference between the predicted heat accumulation value and the preset heat balance threshold is calculated to obtain the heat overflow value.

[0044] It should be noted that the preset thermal balance threshold is set based on the core parameters of the battery module. For a ternary lithium battery with a rated capacity of 100Ah and a nominal voltage of 3.7V, the thermal safety specifications require calculations based on 60% of the battery's maximum allowable heat accumulation, specifically 80 kJ. This setting provides sufficient safety margin to avoid approaching the thermal runaway threshold. The heat overflow value is calculated using the formula: the heat overflow value equals the predicted heat accumulation value minus the preset thermal balance threshold. If the calculation result is less than or equal to 0, it is directly set to 0, indicating that the battery currently has no risk of heat overflow. The parameter setting is based on the fact that the thermal runaway temperature of a ternary lithium battery is approximately 150℃, and 80 kJ corresponds to a temperature of approximately 35℃, which is at the upper limit of the normal operating temperature range and meets the thermal safety design requirements.

[0045] For example, the overall cumulative heat prediction value obtained in S24 is 84.26kJ. Substituting this into the formula, the heat overflow value is 84.26-80=4.26kJ. This positive value indicates that the battery has a slight heat overflow and heat dissipation adjustment needs to be initiated.

[0046] In step S32, the load mutation amplitude corresponding to the heat overflow value is obtained, and the channels are sorted according to the load mutation amplitude to obtain the adjustment priority order for different heat dissipation channels.

[0047] It should be noted that the magnitude of the load change and the amount of heat overflow are correlated through a linear mapping relationship. The mapping formula is as follows:

[0048] Overflow, where k is a proportionality coefficient of 0.05 kW / kJ. This coefficient was obtained through fitting thermal response experiments under 100 different load conditions to ensure mapping accuracy error ≤ ±3%. The heat dissipation channels include four types: liquid cooling circulation channel (heat dissipation power 2.0 kW), air-cooled main channel (heat dissipation power 1.5 kW), forced convection channel (heat dissipation power 1.0 kW), and auxiliary heat sink channel (heat dissipation power 0.5 kW). The channels are ordered according to heat dissipation power from high to low priority; higher-priority channels participate in heat dissipation adjustment first, quickly offsetting heat overflow caused by sudden load changes. Parameter settings are based on the thermal conductivity of the liquid cooling channel (398 W / (m²)). K)) is higher than air cooling and heat sinks, making it suitable for scenarios with instantaneous heat overflow and meeting the dynamic response requirements of thermal load.

[0049] For example, the heat overflow value obtained from S31 is 4.26 kJ. Substituting this into the mapping formula, the load change amplitude ΔP = 0.05 × 4.26 = 0.213 kW is calculated. After sorting by heat dissipation power, the priority order is adjusted to liquid cooling circulation channel > air cooling main channel > forced convection channel > auxiliary heat sink channel to ensure that high-power heat dissipation channels are prioritized.

[0050] In step S33, the heat overflow value is weighted according to the adjustment priority order, and a signal is generated in combination with the system response delay to obtain a feedforward compensation signal.

[0051] It should be noted that the weighted allocation is based on priority, with the liquid cooling circulation channel having a weight of 0.5, the main air cooling channel 0.3, the forced convection channel 0.15, and the auxiliary heatsink channel 0.05, for a total weight of 1, ensuring that all heat overflow values ​​are allocated. The system response latency is an inherent property of each channel: 10ms for the liquid cooling circulation channel, 8ms for the main air cooling channel, 12ms for the forced convection channel, and 15ms for the auxiliary heatsink channel. This value was obtained by averaging multiple response tests. The feedforward compensation signal is calculated using the formula:

[0052] Where U_feed is the single-channel feedforward compensation signal voltage (in V), Q_dist is the heat overflow value after single-channel distribution (in kJ), k_signal is the signal conversion coefficient, with a value of 10 V / kJ, calibrated to match the output magnitude with the PID command, and τ_norm is the normalized value of the channel response delay, which is dimensionless and obtained by dividing the response delay τ of each channel by the reference delay of 10ms, thus eliminating the influence of the delay unit on the dimension. The signal voltage range is limited to 0-30V to adapt to the input requirements of the heat dissipation channel drive.

[0053] For example, in S31, the heat overflow value is 4.26 kJ: the liquid cooling circulation channel allocation value Q_dist = 4.26 × 0.5 = 2.13 kJ, the response delay τ = 10 ms, τ_norm = 10 ms / 10 ms = 1, substituting into the formula to calculate U_feed = (2.13 × 0.1) / 1 = 21.3 V; the air cooling main channel allocation value Q_dist = 4.26 × 0.3 = 1.278 kJ, the response delay is 8 ms, then τ_norm = 0.8, and the calculated U_feed = (1.278 × 0.1 × 0.8 = 10.224 V; the final feedforward compensation signal is the set of voltages for each channel (21.3 V, 10.224 V, 5.325 V, 4.26 V).

[0054] In step S34, a basic heat dissipation command is generated by a PID control algorithm, and the feedforward compensation signal and the basic heat dissipation command are fused to obtain a heat dissipation control command.

[0055] It should be noted that the PID control algorithm parameters are obtained using the Ziegler-Nichols method, with a proportional gain Kp = 5.0, an integral gain Ki = 0.1, a derivative gain Kd = 0.5, a control period Δt = 1 second, and an output limit of 0-30V. The basic heat dissipation command is generated using a discretized formula.

[0056] Where U_base is the basic thermal command voltage (in V), e n e represents the current cycle error (i.e., the amount of heat overflow, in kJ). n-1 The error of the previous cycle is ΣeΔt (initial value is 0, unit is kJ), and the cumulative sum of the errors of the previous 5 cycles is ΣeΔt (unit is kJ). s), Kp, Ki, and Kd are the proportional, integral, and derivative coefficients, respectively, and Δt is the control period. The fusion method is to superimpose the voltages of the same channel: U_ctrl = U_base + U_feed, where U_ctrl is the final heat dissipation control command voltage and U_feed is the feedforward compensation signal voltage; if the superimposed value exceeds 30V, then 30V is used; if it is below 0V, then 0V is used, ensuring that the command is within the operating range of the drive module. The parameter settings are based on the first-order inertial characteristics of the heat dissipation system. These PID parameters can ensure that the system overshoot is ≤5% and the settling time is ≤10 seconds.

[0057] For example, the heat overflow value e in S31 n =4.26kJ, cumulative error of the first 5 cycles Substitute into the formula to calculate The feedforward signal U_feed for the liquid cooling circulation channel is 21.3V. After fusion, U_ctrl = 25.56 + 21.3 = 46.86V. Since it exceeds the limit, it is taken as 30V. The final heat dissipation control command for the liquid cooling channel is 30V. The other channels are calculated according to the same logic to form a complete set of heat dissipation control commands.

[0058] In step S4, the heat dissipation device is driven according to the heat dissipation control command, the feedback signal intensity is collected, the heat distribution deviation is calculated based on the feedback signal intensity and compared with a preset steady-state threshold to obtain the stability evaluation result, including: S41, the heat dissipation control command is parsed to obtain the driving voltage value and the operating frequency value; S42, drive the heat dissipation device according to the driving voltage value and the operating frequency value to obtain feedback waveform data; S43, Denoise the feedback waveform data to obtain the feedback signal strength, and calculate the real-time heat distribution based on the feedback signal strength to obtain the heat distribution deviation; S44. Compare the heat distribution deviation with the preset steady-state threshold to obtain the stability evaluation result.

[0059] In step S41, the heat dissipation control command is parsed to obtain the driving voltage value and the operating frequency value.

[0060] It should be noted that the heat dissipation control commands are encapsulated as key-value pairs of channel number and voltage value according to channel type. During parsing, a voltage-frequency mapping relationship is established based on channel characteristics. The liquid cooling circulation channel uses a linear mapping formula:

[0061] Where U is the driving voltage (unit V), with a frequency range of 50-60Hz; Air-cooled main channel mapping formula:

[0062] Frequency range 30-50Hz; forced convection channel:

[0063] The frequency range is 25-40Hz; the auxiliary heatsink channel has a fixed frequency of 45Hz, and only the drive voltage (range 8-15V) is analyzed. The parameter settings are based on the rated parameters of each heat dissipation device. The rated voltage of the liquid cooling pump is 30V and the frequency is 60Hz, and the rated voltage of the fan is 24V and the frequency is 50Hz, which complies with the GB / T31485-2015 thermal safety standard.

[0064] For example, the heat dissipation control command for the liquid cooling circulation channel in S34 is 30V. Substituting into the formula, we can calculate f_liquid_cooling = 50 + (30 - 24) × 10 / 6 = 60Hz. The command for the air-cooled main channel is 15.975V. We can calculate f_air_cooling = 30 + (15.975 - 12) × 20 / 12 = 36.625Hz. The final analysis results are four sets of driving parameters: liquid cooling channel (30V, 60Hz), air-cooled main channel (15.975V, 36.625Hz), etc.

[0065] In step S42, the heat dissipation device is driven according to the driving voltage value and the operating frequency value, and feedback waveform data is collected.

[0066] It should be noted that the drive adopts PWM pulse width modulation. The liquid cooling circulation channel outputs a set voltage through a DC-DC converter to drive the brushless liquid cooling pump; the other channels output corresponding voltage and frequency drive signals through the motor drive module. Feedback data acquisition uses a high-precision current sensor (range 0-5A, accuracy ±0.2%FS) and a speed sensor (range 0-100Hz, accuracy ±0.1Hz), synchronously acquiring the current and speed waveforms of each channel at a sampling frequency of 100Hz for 5 seconds. The data format is a three-dimensional array of timestamp-current value-speed value. Parameter settings are based on the high-frequency data acquisition requirements; the 100Hz sampling frequency can fully capture the dynamic response of the drive signal, and the 5-second duration balances data integrity and computational efficiency.

[0067] For example, when the liquid cooling channel is driven at 30V and 60Hz, the current waveform collected is a sine wave with a peak value of 1.5A and a valley value of 0.8A, and the speed waveform is stable at 60Hz±0.2Hz; when the air cooling main channel is driven at 15.975V and 36.625Hz, the current waveform has a peak value of 0.6A and a valley value of 0.3A, and the speed waveform fluctuates between 36.5-36.8Hz, ultimately forming a feedback waveform dataset of 500 sets of sampled data.

[0068] In step S43, the feedback waveform data is denoised to obtain the feedback signal strength, and the real-time heat distribution is calculated based on the feedback signal strength to obtain the heat distribution deviation.

[0069] It should be noted that the denoising process uses a 3-level decomposition soft thresholding based on the db4 wavelet basis. The threshold calculation formula is as follows:

[0070] Where σ is the noise standard deviation (calculated from the first 100 sampling points, 0.02A), N is the number of sampling points (500), and λ is calculated to be 0.02 × √(2ln500) ≈ 0.156. After wavelet decomposition of the current waveform data, high-frequency coefficients with absolute values ​​less than λ are set to zero, and then the denoised current signal is reconstructed through inverse wavelet transform. The feedback signal strength is the effective value of the denoised current signal, and the formula is:

[0071] Where T = 5 seconds. Real-time heat distribution is calculated using the formula Q_real-time = Σ(P_channel × Δt), where P_channel = U × I_rms × η (η is the channel efficiency: 0.85 for liquid cooling, 0.8 for main air cooling, 0.75 for forced convection, and 0.7 for auxiliary heat sinks). The heat distribution deviation is equal to Q_real-time plus Q_predicted_base minus Q_predicted_total, where Q_predicted_base is the sum of predicted heat dissipation contributions from all channels in S24, which is 84.26 kJ, and Q_real-time is the actual cumulative heat dissipation.

[0072] For example, after noise reduction of the liquid cooling channel, the effective value of the current I_rms = √(1 / 5∫0) 5 i²(t)dt)=1.2A, P liquid cooling is 30×1.2×0.85=30.6W; the total power of the remaining channels is 42.3W, Q real time within 5 seconds is (30.6+42.3)×5 / 1000=0.3645kJ, and the heat distribution deviation is 0.3645+84.26-84.26=0.3645kJ.

[0073] In step S44, the heat distribution deviation is compared with the preset steady-state threshold to obtain the stability evaluation result.

[0074] It should be noted that the preset steady-state threshold is based on the battery thermal management steady-state requirements in the QCT1206.1-2024 standard, calculated in conjunction with the module's thermal capacity, and is set at ±0.5 kJ. This threshold corresponds to a maximum allowable temperature difference of 2°C in the battery's micro-region, ensuring stable system operation within the thermal equilibrium range. The comparison rule is as follows: if the absolute value of the heat distribution deviation is ≤0.5 kJ, it is considered thermally stable; if the absolute value of the deviation is >0.5 kJ and <1.0 kJ, it is considered slightly fluctuating; if the absolute value of the deviation is ≥1.0 kJ, it is considered a risk of thermal runaway. The parameter settings are based on the thermal safety characteristics of ternary lithium batteries to avoid cycle life degradation due to excessive temperature differences.

[0075] For example, the heat distribution deviation obtained in S43 is 0.3645kJ, and its absolute value is less than 0.5kJ, which meets the preset steady-state threshold requirement. The final stability assessment result is that the thermal state is stable. If the deviation is 0.6kJ, it is judged as a slight fluctuation, and the current heat dissipation strategy should be maintained and continuous monitoring should be carried out. If the deviation is 1.2kJ, it is judged as a risk of thermal runaway, and an emergency heat dissipation plan should be triggered.

[0076] In step S5, if the stability assessment result is a stable state, then the dynamic error correction amount is calculated based on the feedback signal strength, the advance compensation range is defined according to the preset predicted heat model, and the thermal management parameter configuration is obtained by combining the dynamic error correction amount and the advance compensation range, including: S51, if the stability assessment result is a stable state, then extract the transient fluctuation amplitude in the feedback signal intensity and construct a spatial residual heat distribution matrix; S52, the residual heat distribution matrix is ​​analyzed to obtain a dynamic error correction vector used to characterize the degree of deviation from the thermal state; S53, input the dynamic error correction vector into the preset predicted heat model, delineate the advanced compensation range based on the extreme points of the heat evolution trajectory, and determine the advanced compensation range as the preset compensation range; S54, combine the preset compensation range with the dynamic error correction vector to obtain the compensation gain value, and use the compensation gain value to perform correction calculations to obtain the thermal management parameter configuration.

[0077] In step S51, if the stability assessment result is a stable state, the transient fluctuation amplitude in the feedback signal intensity is extracted to construct a spatial residual heat distribution matrix.

[0078] It should be noted that the transient fluctuation amplitude is calculated using the feedback signal strength sequence, taking the absolute value of the difference between each sampling point in the denoised current signal sequence (500 sampling points) in S43 and the sequence mean. Where I_i is the current value at the i-th sampling point, and I_mean is the sequence mean. The residual heat mapping formula is: The parameter k_map = 0.5 J / A is derived by linearly fitting 100 different current fluctuation amplitudes with the actual residual heat estimated from temperature, with a fitting error ≤ ±2%. Physically, it characterizes the residual heat energy corresponding to a unit average current fluctuation. The matrix dimension is 8×6, corresponding one-to-one with the battery micro-region grid. Matrix elements are the residual heat values ​​of the corresponding micro-regions, with row indices corresponding to the x-coordinates (1-8) and column indices corresponding to the y-coordinates (1-6). The parameter settings are based on the fact that the fluctuation amplitude directly reflects the dynamic change of residual heat after heat dissipation. The 8×6 matrix fits the module structure, facilitating spatial distribution analysis.

[0079] For example, in S43, the mean of the feedback signal intensity sequence of the liquid cooling channel is I_mean=1.1A, the mean of a certain sampling point is I_i=1.2A, A_fluc=0.1A, and Q_res=0.1×0.5=0.05J; the mean of the fluctuation amplitude of 500 sampling points in the micro-region (3,2) is 0.08A, corresponding to Q_res=0.04J, and the element at position (3,2) in the final 8×6 residual heat distribution matrix is ​​0.04J.

[0080] In step S52, the residual heat distribution matrix is ​​analyzed to obtain a dynamic error correction vector used to characterize the degree of deviation from the thermal state.

[0081] It should be noted that the analytical method employs eigenvalue decomposition, performing singular value decomposition on the 8×6 residual heat distribution matrix M. The formula is as follows:

[0082] Where U is the left singular vector matrix (8×8), Σ is the singular value diagonal matrix (8×6), and V is the right singular vector matrix (6×6). The largest singular value in Σ is taken, and the corresponding left singular vector U1 is used as the dynamic error correction vector. The vector dimension is 8, and the element values ​​represent the deviation weights of the corresponding micro-regions at the x-coordinate. The magnitude ||U1|| = The modulus ranges from 0 to 1, with larger values ​​indicating more severe overall deviation. The parameter setting is based on the fact that the vector corresponding to the largest singular value can centrally reflect the main distribution characteristics of the matrix and accurately locate the deviation from the dominant direction.

[0083] For example, after decomposing the residual heat distribution matrix of S51, the maximum singular value λ1=0.32 corresponds to the left singular vector U1=[0.05,0.08,0.25,0.12,0.09,0.07,0.06,0.04], and the modulus ||U1||=√(0.05²+0.08²+…+0.04²)≈0.31, indicating that the micro-region column corresponding to x=3 (the third element 0.25) deviates the most significantly, and the dynamic error correction vector is U1.

[0084] In step S53, the dynamic error correction vector is input into the preset predicted heat model, and the advanced compensation range is determined based on the extreme points of the heat evolution trajectory. The advanced compensation range is then determined as the preset compensation range.

[0085] It should be noted that the preset heat prediction model is a trained LSTM model. The LSTM model receives an 8-dimensional dynamic error correction vector as input, first expands it to the initial state of the time series through a fully connected layer, and then predicts the heat evolution trajectory of each micro-region within the next 5 seconds. The output dimension is 8×6×time step, ensuring that the trajectory is aligned with the spatial distribution of the micro-region. For example, the output trajectory shows that the heat of micro-region (3,2) reaches an extreme value of 0.06J in the next 3 seconds, and other micro-region trajectories are processed similarly. The 5-second interval closely matches the battery thermal response cycle, which can fully cover the dynamic adjustment process of the heat dissipation device, balancing prediction timeliness and practicality. The model training data includes 1000 sets of thermal evolution data under different residual heat distributions, with 1000 training iterations. The loss function MSE≤0.01, the activation function is ReLU, and the optimizer is Adam. ReLU is a commonly used activation function in neural networks, which can introduce nonlinear features to help the model capture the complex patterns of heat evolution trajectory. Adam is an adaptive learning rate optimizer that can quickly and stably train the model and adapt to the efficient learning needs of thermal evolution data. The extreme point is extracted as the time t_extreme corresponding to the maximum heat value in the trajectory. The advance compensation time Δt = t_extreme, and the compensation range is [Q_current residual, Q_extreme], where Q_current residual is the mean of the matrix elements, and Q_extreme is the maximum value of the trajectory. The parameter settings are based on the fact that the LSTM model is good at time series prediction, and the 5-second window fits the thermal response cycle to ensure the timeliness of compensation.

[0086] For example, if the dynamic error correction vector U1 of S52 is input into the model, the output trajectory shows that the heat will reach the extreme value Q extreme value = 0.06J in the next 3 seconds, and the current residual heat average value Q current residual = 0.03J. Therefore, the advance compensation range is to intervene 3 seconds in advance, with an amplitude of 0.03-0.06J. This range is the preset compensation range.

[0087] In step S54, the compensation gain value is obtained by combining the preset compensation range and the dynamic error correction vector, and the heat dissipation control command is corrected by using the retrieved compensation gain value to obtain the thermal management parameter configuration.

[0088] It should be noted that the preset gain table is designed with a two-dimensional index of correction vector magnitude and compensation time. The magnitude is divided into three levels: 0-0.2, 0.2-0.4, and 0.4-0.6, and the compensation time is divided into three levels: 1-2s, 2-3s, and 3-5s, corresponding to the gain values ​​of different channels (liquid cooling, main air cooling, forced convection, and auxiliary heatsinks). After retrieval, the thermal management parameter correction formula is P correction equal to P current multiplied by G (P is the voltage / frequency parameter, and G is the gain value). The corrected parameters must conform to the rated range of each channel (liquid cooling voltage 24-30V, frequency 50-60Hz, etc.). The parameter settings are based on the gain value being verified under multiple operating conditions, which can improve the residual heat convergence speed by more than 30%.

[0089] For example, in S52, the correction vector magnitude is 0.31 (0.2-0.4 range), and in S53, the compensation time is 3 seconds (2-3 seconds range). Looking up the gain table, we get G=1.3 for liquid cooling, G=1.1 for main air cooling, G=1.05 for forced convection, and G=1.0 for auxiliary heatsinks. In S41, the current parameters for liquid cooling are 30V and 60Hz. After correction, the voltage is 30×1.3=39V (30V for exceeding the rated upper limit), and the frequency is 60×1.3=78Hz (60Hz for exceeding the rated upper limit). The current parameters for main air cooling are 15.975V and 36.625Hz. After correction, the voltage is 15.975×1.1≈17.57V, and the frequency is 36.625×1.1≈40.29Hz. The final thermal management parameters are configured as a set of corrected voltages and frequencies for the four channels.

[0090] In step S6, based on the thermal management parameter configuration, a frequency spectrum distribution matrix is ​​constructed and a state transition matrix is ​​generated. Signal prediction offsets are extracted, and the thermal management parameter configuration is calibrated according to the signal prediction offsets to obtain system response characteristics, including: S61, extract the current fluctuation frequency sequence and the load change amplitude sequence from the thermal management parameter configuration, and perform Fourier transform on the current fluctuation frequency sequence to obtain the frequency spectrum distribution matrix; S62, calculate the mutation detection threshold based on the frequency spectrum distribution matrix and the load mutation amplitude sequence. If the amplitude in the load mutation amplitude sequence exceeds the mutation detection threshold, extract the convergence vector from the load mutation amplitude sequence to obtain the iterative convergence vector. S63, construct a state transition matrix based on the iterative convergence vector, and perform eigenvalue decomposition on the state transition matrix to obtain the matrix eigenvalue sequence and signal prediction offset; S64, the signal prediction offset is used to calibrate the heat dissipation control command for the next cycle to obtain the system response characteristics that characterize the smoothness of the system response characteristic curve.

[0091] In step S61, the current fluctuation frequency sequence and the load change amplitude sequence are extracted from the thermal management parameter configuration, and the Fourier transform of the current fluctuation frequency sequence is performed to obtain the frequency spectrum distribution matrix.

[0092] It should be noted that the current fluctuation frequency sequence is extracted from the optimized thermal management parameter configuration of S54, corresponding to the real-time current data of an 8×6 micro-region. Data is collected at a sampling frequency of 10Hz, generating a one-dimensional sequence of 100 sampling points for each micro-region. The 10s acquisition time fully covers the typical cycle of current fluctuation. To meet the FFT algorithm's requirement for the number of points (N=1024) and optimize the frequency resolution, the 100-point sequence needs to be expanded to 1024 points by zero-padding. The load mutation amplitude sequence is the absolute value of the difference between the current in each micro-region and the rated current, expressed as ΔI=|I_real-I_rated|, where ΔI is the load mutation amplitude, I_real is the real-time current of the micro-region, and I_rated is the rated current, with a value of 100A. The Fourier transform employs the Fast Fourier Transform (FFT) algorithm with 1024 sampling points and a frequency resolution of 10Hz / 1024≈0.0098Hz. The resulting frequency spectrum distribution matrix has dimensions of 8×6×513, with 8 rows corresponding to the x-coordinate, 6 columns to the y-coordinate, and 513 frequency points. The matrix elements represent the amplitude of the corresponding micro-region at each frequency point, in A / Hz. The parameter settings are based on the principle that FFT can efficiently decompose the frequency components of current fluctuations, balancing resolution and computational efficiency with 1024 sampling points, and the matrix dimensions closely reflecting the spatial distribution of the micro-region.

[0093] For example, the current fluctuation frequency sequence of the micro-region (3,2) in S54 contains 100 sampling points. After zero-padding and expansion to 1024 points, an FFT transformation is performed. The amplitude at the 0.8Hz frequency point (corresponding to the frequency index k=0.8 / Δf≈82) is 0.3A / Hz, and the amplitude at the 1.2Hz frequency point is 0.1A / Hz. The element at position (3,2,82) in the final frequency spectrum distribution matrix is ​​0.3A / Hz, which fully presents the frequency energy distribution of the micro-region.

[0094] In step S62, a mutation detection threshold is calculated based on the frequency spectrum distribution matrix and the load mutation amplitude sequence. If the amplitude in the load mutation amplitude sequence exceeds the mutation detection threshold, a convergence vector is extracted from the load mutation amplitude sequence to obtain an iterative convergence vector.

[0095] It should be noted that the mutation detection threshold is calculated using the formula:

[0096] Where Th is the mutation detection threshold (unit is the same as the load mutation amplitude, which is kW in this embodiment), σ_spec is the standard deviation of all elements of the frequency spectrum distribution matrix (empirical value 0.05A / Hz), and μ_amp is the mean of the load mutation amplitude sequence (unit kW). The scaling factor is set to 1.5, which is dimensionless and reduces the false positive rate. 40 sampling points can completely capture the convergence process after abrupt changes. To eliminate dimensional differences, the unit of σ_spec needs to be converted to kW by multiplying it by the current-to-power conversion factor C, in kW·Hz / A, i.e., σ_spec_kW = σ_spec × C, where C is calculated based on the system's rated voltage V_rated (in V), C = V_rated / 1000. In this example, V_rated is taken as 400V, so C = 0.4kW·Hz / A. The corrected formula is:

[0097] Select the moment when the load mutation amplitude sequence first exceeds the threshold Th, extract 20 sampling points before and after it, calculate the amplitude difference between adjacent sampling points, that is, subtract the load mutation amplitude value corresponding to the sampling point to form a 40-dimensional one-dimensional vector, that is, the iterative convergence vector, and the positive or negative of the vector elements represent the upward or downward trend of the amplitude.

[0098] For example, in S61, the mean of the load mutation amplitude sequence is μ_amp=0.15kW, σ_spec=0.05A / Hz, and after conversion, σ_spec_kW=0.05×0.4=0.02kW. Th is calculated to be 1.5×0.02+0.15=0.18kW. At a certain sampling point in the micro-region (3,2), ΔP=0.25kW (exceeding the threshold), a mutation is detected. The difference sequence of the 20 sampling points before and after the mutation is extracted, and the iterative convergence vector [0.02,0.015,...,-0.008] is obtained, which represents the trend of gradual convergence of the amplitude after the load mutation.

[0099] In step S63, a state transition matrix is ​​constructed based on the iterative convergence vector, and eigenvalue decomposition is performed on the state transition matrix to obtain the matrix eigenvalue sequence and signal prediction offset.

[0100] It should be noted that the state transition matrix is ​​a 4×4 square matrix. It is constructed by dividing the 40-dimensional iterative convergence vector into four groups of 10-point segments, forming four groups of 10-dimensional sub-vectors. The matrix element M_ij = E[V_i×V_j^T] (where V_i is the i-th sub-vector and E is the expectation value) is obtained through autocorrelation operations on these sub-vectors. The eigenvalue decomposition uses the QR algorithm, which efficiently decomposes the state transition matrix and quickly obtains reliable eigenvalues ​​and eigenvectors, providing support for subsequent signal prediction offset calculations. The decomposition formula is as follows: Where U is a 4×4 eigenvector matrix, Λ is a diagonal matrix, and the diagonal elements are eigenvalues, i.e., the sequence of matrix eigenvalues. The signal prediction offset is the dot product of the eigenvector corresponding to the largest eigenvalue and the iterative convergence vector, with units of kW. The parameter settings are based on the fact that the 4×4 matrix fits the multimodal characteristics of the convergence trend, the QR algorithm has high decomposition accuracy, and the largest eigenvalue can reflect the dominant convergence mode.

[0101] For example, the state transition matrix M constructed after segmenting the iterative convergence vector in S62 is decomposed by QR to obtain the eigenvalue sequence [0.98, 0.72, 0.35, 0.11]. The eigenvector corresponding to the largest eigenvalue 0.98 is U1=[0.85, 0.32, 0.21, 0.18]. The signal prediction offset is equal to U1 multiplied by the convergence vector, which equals 0.03kW, representing the prediction offset direction and magnitude of the load amplitude in the next cycle.

[0102] In step S64, the signal prediction offset is used to calibrate the heat dissipation control command for the next cycle to obtain the system response characteristics that characterize the smoothness of the system response characteristic curve.

[0103] It should be noted that the calibration formula is: Where U is the original instruction of S54 heat dissipation control (e.g., 30V for liquid cooling channel), k is the calibration coefficient (valued at 0.5V / kW), and Δ offset is the signal prediction offset. The system response characteristics (smoothness) are calculated using the formula:

[0104] Where ΔU_i is the difference between adjacent values ​​in the calibrated command sequence, N is the sequence length (100 sampling points), and the smoothness range is 0-1, with a smoother response curve indicating a more stable response. The parameter settings are based on the fact that the calibration coefficient has been verified under multiple operating conditions to offset the influence of offset, and the smoothness calculation quantifies the stability of the response, meeting the stability requirements of BMS real-time control.

[0105] For example, in S63, the signal prediction offset is 0.03kW, and the original command U for the liquid cooling channel is 30V. Substituting these values ​​into the calibration formula, we get the calibrated U as 30 + 0.5 × 0.03 = 30.015V. After calibration, the sum of squares of adjacent differences in the command sequence Σ(ΔU_i)^2 = 0.002. With N = 100, the smoothness S is calculated to be approximately 0.9955, indicating that the system response characteristic curve tends to be stable.

[0106] In step S7, if the system response characteristics satisfy a preset load drastic change condition, a temperature gradient distribution matrix is ​​constructed based on the thermal state distribution results. The temperature gradient distribution matrix is ​​then fused with the system response characteristics to obtain a regional heat flux matrix. The regional heat flux matrix is ​​then weighted and adjusted to obtain a final adjustment instruction, including: S71, if the system response characteristics meet the preset load drastic change condition, then extract the thermal change trend trajectory based on the thermal state distribution result, perform wavelet transform processing on the thermal change trend trajectory, and obtain the temperature gradient distribution matrix. S72, calculate the heat imbalance threshold based on the temperature gradient distribution matrix and the system response characteristics. If the amplitude of the temperature gradient distribution matrix exceeds the heat imbalance threshold, extract the time delay component from the system response characteristics to obtain the delay compensation vector. S73, perform element-wise operations using the delay compensation vector and the temperature gradient distribution matrix to obtain the regional heat flow matrix and trend prediction offset; S74, the regional heat flow matrix is ​​weighted and adjusted using the trend prediction offset to obtain the final adjustment command for the micro-region heat imbalance.

[0107] In step S71, if the system response characteristics meet the preset load drastic change condition, the thermal change trend trajectory is extracted based on the thermal state distribution result, and wavelet transform is performed on the thermal change trend trajectory to obtain the temperature gradient distribution matrix.

[0108] It should be noted that the thermal change trend trajectory is extracted from the micro-region thermal state distribution data in S1, including the temperature time series of each micro-region within the next 30 minutes (sampling frequency 1Hz, total 1800 data points). The wavelet transform employs a 4-level multi-scale decomposition of the db4 wavelet basis. After decomposition, temperature gradient information is extracted from the detail coefficients. The matrix dimension is 8×6 (corresponding one-to-one with the battery micro-region grid), and the matrix elements are the temperature gradient values ​​of the corresponding micro-regions, in units of ℃ / cm. The parameter settings are based on adapting the db4 wavelet basis to the nonlinear characteristics of the temperature time series signal. The 4-level decomposition can accurately capture local thermal gradient abrupt changes, and the 8×6 matrix fits the module spatial distribution, meeting the requirements for precise micro-region control.

[0109] For example, the thermal change trend trajectory of the micro-region (3,2) in S1 shows that the temperature rises from 33.5℃ to 39.2℃ within 30 minutes. After 4-level decomposition by db4 wavelet, the temperature gradient value of this region is 0.8℃ / cm. The element at position (3,2) in the final temperature gradient distribution matrix is ​​0.8℃ / cm. The matrix as a whole shows the distribution characteristic that the gradient in the core region is higher than that in the edge region.

[0110] In step S72, a heat imbalance threshold is calculated based on the temperature gradient distribution matrix and the system response characteristics. If the amplitude of the temperature gradient distribution matrix exceeds the heat imbalance threshold, a time delay component is extracted from the system response characteristics to obtain a delay compensation vector.

[0111] It should be noted that the heat imbalance threshold is calculated using the following formula:

[0112] Where Th_imbal is the heat imbalance threshold (unit: °C / cm), and σ_grad is the standard deviation of all elements in the temperature gradient distribution matrix (empirical value: 0.15 °C / cm). The historical maximum value of the temperature gradient distribution matrix (unit: °C / cm) is used. S_resp is normalized to convert it to the actual temperature gradient dimension. S_resp represents the system response smoothness of S64 (range: 0-1). The time delay component is extracted from the phase lag data of the system response characteristics, taking the average time difference between the heat conduction response and the control signal. The delay compensation vector is an 8-dimensional column vector (corresponding to 8 x-coordinate micro-regions), with each element representing the time delay duration of the corresponding region in seconds. Parameter settings are based on 2.0 times the standard deviation to reduce the false negative rate, and the smoothness weighting term is correlated with system stability. The 8-dimensional vector adapts to spatial dimension compensation requirements.

[0113] For example, in S71, σ_grad = 0.15℃ / cm, and in S64, the system response smoothness S_resp = 0.9955. Th_imbal = 0.15 × 2.0 + 0.9955 × 0.1 ≈ 0.3996℃ / cm is calculated. The temperature gradient in the micro-region (3,2) is 0.8℃ / cm (exceeding the threshold). The time delay component of this region is extracted as 0.8s, and the final delay compensation vector is [0.5, 0.6, 0.8, 0.4, 0.3, 0.7, 0.5, 0.4]^T.

[0114] In step S73, the delay compensation vector and the temperature gradient distribution matrix are used to perform element-wise operations to obtain the regional heat flow matrix and trend prediction offset.

[0115] It should be noted that the delay compensation vector is an 8×1 dimensional column vector (denoted as V_delay), and the temperature gradient distribution matrix is ​​an 8×6 dimensional matrix, denoted as M_grad, with units of 1°C / cm = 100 K / m. Element-by-element calculations are performed according to the following formula:

[0116] in This indicates element-wise multiplication, where 1 represents a row vector of all 1s with dimensions 1×6. Will Expanded to an 8×6 matrix, R_thermal is the thermal resistance matrix, with units of m²·K / W, representing the resistance to heat transfer between micro-regions of the battery. Its reciprocal is the thermal conductivity, which is set according to the battery module material properties and structural design. Typical values ​​are 0.005–0.02 m²·K / W (refer to GB 38031-2025), consistent with the heat dissipation calculation parameters in S24.

[0117] in, This is the delay effect coefficient, which has no unit and an empirical value of 0.1. It is calibrated experimentally and is consistent with the calibration logic in S5 and S23. This is the historical maximum delay value, in seconds, used for normalization and is compatible with the threshold calculation logic in S72. The thickness of the battery micro-region casing is in meters (m), and is set to 0.01m based on step S24.

[0118] The calculated regional heat flow matrix M_heatflow has dimensions of 8×6, and the element unit is W / m² (heat flux density). The trend prediction offset is calculated according to the formula. The calculation is performed, where tr(M_heatflow) is the trace of the regional heat flow matrix (the sum of the diagonal elements), 8×6 is the matrix dimension product, and ΔT_pred is the scalar offset of the heat flow trend, in units of W / m².

[0119] For example, the M_grad value of the micro-region (3,2) is 0.8°C / cm, which translates to K / m: 0.8 × 100 = 80 K / m. The region has R_thermal = 0.01 m²·K / W and δ = 0.01 m. Given R_thermal = 1.0 W / (m·K), the corresponding delay value in V_delay is 0.8 s, the historical maximum delay V_delay,max = 1.5 s, and the delay correction term is 1 - 0.1 × (0.8 / 1.5) ≈ 0.9467. Substituting into the formula, we get M_heatflow(3,2) = 1.0 × 80 × 0.9467 ≈ 75.74 W / m². After calculating the matrix trace, the example value of ΔT_pred is 70.5 W / m².

[0120] In step S74, the regional heat flow matrix is ​​weighted and adjusted using the trend prediction offset to obtain the final adjustment instruction for the micro-regional heat imbalance.

[0121] It should be noted that the weighted adjustment formula is as follows:

[0122] Where M_heatflow_adj is the weighted adjusted regional heat flow matrix element (unit: W / m²), M_heatflow is the unadjusted regional heat flow matrix element (unit: W / m²), ΔT_pred is the trend prediction offset (unit: °C), and k_adj=0.3 (weighting coefficient, fitted by multi-condition experiments, error ≤ ±3%). The final adjustment command is the heat dissipation power adjustment value of each micro-region, calculated according to the formula P_adj=M_heatflow_adj×S_cell×δ, where S_cell is the micro-region area 0.0025m², δ is the heat dissipation equivalent thickness 0.01m, and P_adj is in W. The parameter setting is based on the fact that weighted adjustment can enhance the compensation effect of trend prediction, and the heat dissipation power calculation is close to the power output range of the actual heat dissipation device, complying with GB / T31485-2015 thermal safety specifications.

[0123] For example, in S73, the element of the regional heat flow matrix M_heatflow(3,2) is 0.64W / m², and the trend prediction offset ΔT_pred is 0.6℃. Substituting these values ​​into the formula, we get M_heatflow_adj = 0.64 × (1 + 0.6 × 0.3) = 0.64 × 1.18 = 0.7552W / m²; P_adj = 0.7552 × 0.0025 × 0.01 ≈ 1.89 × 10^-5kW = 0.0189W. The final heat imbalance adjustment instruction for the micro-region (3,2) is to increase the heat dissipation power by 0.02W. The other micro-regions are calculated using the same logic, forming a complete set of final adjustment instructions.

[0124] In summary, this invention discloses a BMS temperature control method. Through a multi-level processing chain, including parsing heat dissipation control commands and closed-loop acquisition of feedback signals, denoising feedback waveform data and inverting real-time heat distribution deviations, steady-state evaluation of heat distribution deviations and adaptive configuration of thermal management parameters, feature analysis of the residual heat matrix and advance compensation of a preset predicted heat model, spectral density analysis of current fluctuation frequencies and iterative convergence vector extraction of load mutations, extraction of time-delay components of system response delays and precise weighted adjustment of micro-region heat imbalances, this invention achieves refined perception of the thermal state of the battery management system, advance quantitative compensation for thermal imbalance risks, and smooth calibration output of control commands. This effectively improves the real-time performance, stability, and precision of temperature control, providing crucial technical support for the safe operation and efficient heat dissipation of battery systems.

[0125] Reference Figure 2 The second embodiment of the present invention provides a BMS temperature control system, comprising: The thermal distribution module is used to acquire the current data and temperature data of the battery, and to perform multi-dimensional feature coupling and grid projection on the current data and temperature data to obtain the thermal state distribution results of the battery micro-region. The heat prediction module is used to predict the cumulative heat value by combining the thermal state distribution results with environmental interference factors and using the dynamic thermal conductivity coefficient to make trend prediction. The instruction adjustment module is used to calculate the difference between the predicted heat accumulation value and the preset heat balance threshold, obtain the load change amplitude and system response delay, generate a feedforward compensation signal based on the load change amplitude and the system response delay, generate a basic heat dissipation instruction through a PID control algorithm, and fuse the feedforward compensation signal and the basic heat dissipation instruction to obtain a heat dissipation control instruction. The steady-state evaluation module is used to drive the heat dissipation device according to the heat dissipation control command, collect the feedback signal intensity, calculate the heat distribution deviation according to the feedback signal intensity and compare it with the preset steady-state threshold to obtain the stability evaluation result. The parameter configuration module is used to calculate the dynamic error correction amount based on the feedback signal intensity if the stability assessment result is a stable state, define the advance compensation range according to the preset predicted heat model, and obtain the thermal management parameter configuration by combining the dynamic error correction amount and the advance compensation range. The signal calibration module is used to construct a frequency spectrum distribution matrix and generate a state transition matrix based on the thermal management parameter configuration, extract the signal prediction offset, calibrate the thermal management parameter configuration according to the signal prediction offset, and obtain the system response characteristics. The instruction generation module is used to construct a temperature gradient distribution matrix based on the thermal state distribution results if the system response characteristics meet the preset load drastic change conditions, fuse the temperature gradient distribution matrix with the system response characteristics to obtain a regional heat flow matrix, and adjust the regional heat flow matrix by weight to obtain the final adjustment instruction.

[0126] It should be noted that the BMS temperature control system provided in this embodiment of the invention is used to execute all the process steps of the BMS temperature control method in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0127] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a BMS temperature control program. When the processor executes the computer program, it implements the steps described in the various BMS temperature control method embodiments above, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the thermal distribution module.

[0128] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0129] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0130] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0131] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0132] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0133] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0134] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A BMS temperature control method, characterized in that, include: The battery's current and temperature data are acquired, and multidimensional feature coupling and grid projection are performed on the current and temperature data to obtain the thermal state distribution results of the battery's micro-regions. Based on the thermal state distribution results, combined with environmental disturbance factors and using dynamic thermal conductivity coefficient for trend prediction, the predicted value of heat accumulation is obtained. The difference between the predicted cumulative heat value and the preset thermal balance threshold is calculated to obtain the load change amplitude and system response delay. A feedforward compensation signal is generated based on the load change amplitude and the system response delay. A basic heat dissipation command is generated through a PID control algorithm. The feedforward compensation signal and the basic heat dissipation command are fused to obtain a heat dissipation control command. Drive the heat dissipation device according to the heat dissipation control command, collect the feedback signal intensity, calculate the heat distribution deviation according to the feedback signal intensity and compare it with the preset steady-state threshold to obtain the stability evaluation result; If the stability assessment result is a stable state, then the dynamic error correction amount is calculated based on the feedback signal strength, the advance compensation range is defined according to the preset predicted heat model, and the thermal management parameter configuration is obtained by combining the dynamic error correction amount and the advance compensation range. Based on the thermal management parameter configuration, a frequency spectrum distribution matrix is ​​constructed and a state transition matrix is ​​generated. The signal prediction offset is extracted, and the thermal management parameter configuration is calibrated according to the signal prediction offset to obtain the system response characteristics. If the system response characteristics meet the preset load drastic change conditions, a temperature gradient distribution matrix is ​​constructed based on the thermal state distribution results. The temperature gradient distribution matrix and the system response characteristics are then fused to obtain a regional heat flux matrix. The regional heat flux matrix is ​​then weighted and adjusted to obtain the final adjustment instruction.

2. The BMS temperature control method according to claim 1, characterized in that, The process of performing multidimensional feature coupling and grid projection on the current data and temperature data to obtain the thermal state distribution results of the battery micro-region includes: Acquire current and temperature data from the battery to obtain current datasets and temperature datasets; extract features from the current datasets and temperature datasets to obtain current feature data and temperature feature data. The current feature data and the temperature feature data are subjected to multi-dimensional feature coupling processing to obtain a multi-dimensional fused feature matrix; The heat accumulation rate is calculated based on the multidimensional fusion feature matrix, and the heat accumulation rate is projected onto a preset grid model to obtain the thermal state distribution results of the battery micro-region.

3. The BMS temperature control method according to claim 1, characterized in that, The step of obtaining the predicted heat accumulation value based on the thermal state distribution results, combined with environmental disturbance factors and using the dynamic thermal conductivity coefficient for trend prediction includes: Based on the thermal state distribution results, the associated temperature flow sequence is obtained, and the temperature flow sequence is weighted by applying preset data weights to obtain a weighted thermal feature sequence. The weighted thermal feature sequence is convolved with the environmental interference factor to obtain the corrected thermal feature vector; Based on the modified thermal feature vector, the dynamic thermal conductivity coefficient of the preset time window is predicted to obtain the dynamic thermal conductivity coefficient matrix. The continuous thermal change trend is then deduced based on the dynamic thermal conductivity coefficient matrix to obtain the thermal change trend trajectory. The heat change trend trajectory is integrated over the time domain to obtain the predicted cumulative heat value.

4. The BMS temperature control method according to claim 1, characterized in that, The difference between the predicted cumulative heat value and the preset thermal balance threshold is calculated to obtain the load mutation amplitude and system response delay. A feedforward compensation signal is generated based on the load mutation amplitude and the system response delay. A basic heat dissipation command is generated through a PID control algorithm. The feedforward compensation signal and the basic heat dissipation command are then fused to obtain a heat dissipation control command, including: The difference between the predicted cumulative heat value and the preset heat balance threshold is calculated to obtain the heat overflow value; Obtain the load fluctuation amplitude corresponding to the heat overflow value, sort the channels according to the load fluctuation amplitude, and obtain the priority sequence for different heat dissipation channels; The heat overflow value is weighted according to the priority sequence, and a signal is generated by combining the system response delay to obtain a feedforward compensation signal; A basic heat dissipation command is generated by a PID control algorithm, and the heat dissipation control command is obtained by fusing the feedforward compensation signal and the basic heat dissipation command.

5. The BMS temperature control method according to claim 1, characterized in that, The process of driving the heat dissipation device according to the heat dissipation control command, collecting feedback signal intensity, calculating the heat distribution deviation based on the feedback signal intensity and comparing it with a preset steady-state threshold to obtain a stability evaluation result includes: The heat dissipation control command is parsed to obtain the driving voltage value and the operating frequency value; The heat dissipation device is driven according to the driving voltage value and the operating frequency value, and feedback waveform data is collected. The feedback waveform data is denoised to obtain the feedback signal strength. The real-time heat distribution is calculated based on the feedback signal strength to obtain the heat distribution deviation. By comparing the heat distribution deviation with the preset steady-state threshold, the stability evaluation result is obtained.

6. The BMS temperature control method according to claim 1, characterized in that, If the stability assessment result is a stable state, then a dynamic error correction amount is calculated based on the feedback signal strength, an advance compensation range is defined according to a preset predicted heat model, and the thermal management parameter configuration is obtained by combining the dynamic error correction amount and the advance compensation range, including: If the stability assessment result is a stable state, then the transient fluctuation amplitude in the feedback signal intensity is extracted to construct a spatial residual heat distribution matrix; The residual heat distribution matrix is ​​analyzed to obtain a dynamic error correction vector that characterizes the degree of deviation from the thermal state; The dynamic error correction vector is input into the preset predicted heat model, and the range of advance compensation is obtained by defining the extreme points of the heat evolution trajectory. By combining the advance compensation range and the dynamic error correction vector, a retrieval compensation gain value is obtained. The heat dissipation control command is then corrected using the retrieval compensation gain value to obtain the thermal management parameter configuration.

7. The BMS temperature control method according to claim 1, characterized in that, Based on the thermal management parameter configuration, a frequency spectrum distribution matrix is ​​constructed and a state transition matrix is ​​generated. Signal prediction offsets are extracted, and the thermal management parameter configuration is calibrated according to the signal prediction offsets to obtain system response characteristics, including: Extract the current fluctuation frequency sequence and load change amplitude sequence from the thermal management parameter configuration, and perform a Fourier transform on the current fluctuation frequency sequence to obtain the frequency spectrum distribution matrix; The mutation detection threshold is calculated based on the frequency spectrum distribution matrix and the load mutation amplitude sequence. If the amplitude in the load mutation amplitude sequence exceeds the mutation detection threshold, the convergence vector is extracted from the load mutation amplitude sequence to obtain the iterative convergence vector. A state transition matrix is ​​constructed based on the iterative convergence vector, and eigenvalue decomposition is performed on the state transition matrix to obtain the signal prediction offset. The heat dissipation control command is calibrated based on the predicted offset of the signal to obtain the system response characteristics.

8. The BMS temperature control method according to claim 1, characterized in that, If the system response characteristics satisfy a preset load drastic change condition, a temperature gradient distribution matrix is ​​constructed based on the thermal state distribution results. The temperature gradient distribution matrix and the system response characteristics are then fused to obtain a regional heat flux matrix. The regional heat flux matrix is ​​then weighted and adjusted to obtain a final adjustment instruction, including: If the system response characteristics meet the preset load drastic change condition, then the thermal change trend trajectory is extracted based on the thermal state distribution result, and wavelet transform is performed on the thermal change trend trajectory to obtain the temperature gradient distribution matrix. The heat imbalance threshold is calculated based on the temperature gradient distribution matrix and the system response characteristics. If the amplitude of the temperature gradient distribution matrix exceeds the heat imbalance threshold, the time delay component is extracted from the system response characteristics to obtain the delay compensation vector. By performing element-wise operations on the delay compensation vector and the temperature gradient distribution matrix, the regional heat flux matrix and trend prediction offset are obtained. The heat flow matrix of the region is weighted and adjusted using the trend prediction offset to obtain the final adjustment command for the heat imbalance in the micro-region.

9. A BMS temperature control system, characterized in that, include: The thermal distribution module is used to acquire the current data and temperature data of the battery, and to perform multi-dimensional feature coupling and grid projection on the current data and temperature data to obtain the thermal state distribution results of the battery micro-region. The heat prediction module is used to predict the cumulative heat value by combining the thermal state distribution results with environmental interference factors and using the dynamic thermal conductivity coefficient to make trend prediction. The instruction adjustment module is used to calculate the difference between the predicted heat accumulation value and the preset heat balance threshold, obtain the load change amplitude and system response delay, generate a feedforward compensation signal based on the load change amplitude and the system response delay, generate a basic heat dissipation instruction through a PID control algorithm, and fuse the feedforward compensation signal and the basic heat dissipation instruction to obtain a heat dissipation control instruction. The steady-state evaluation module is used to drive the heat dissipation device according to the heat dissipation control command, collect the feedback signal intensity, calculate the heat distribution deviation according to the feedback signal intensity and compare it with the preset steady-state threshold to obtain the stability evaluation result. The parameter configuration module is used to calculate the dynamic error correction amount based on the feedback signal intensity if the stability assessment result is a stable state, define the advance compensation range according to the preset predicted heat model, and obtain the thermal management parameter configuration by combining the dynamic error correction amount and the advance compensation range. The signal calibration module is used to construct a frequency spectrum distribution matrix and generate a state transition matrix based on the thermal management parameter configuration, extract the signal prediction offset, calibrate the thermal management parameter configuration according to the signal prediction offset, and obtain the system response characteristics. The instruction generation module is used to construct a temperature gradient distribution matrix based on the thermal state distribution results if the system response characteristics meet the preset load drastic change conditions, fuse the temperature gradient distribution matrix with the system response characteristics to obtain a regional heat flux matrix, and adjust the regional heat flux matrix by weight to obtain the final adjustment instruction.

Images