New energy vehicle battery thermal management intelligent prediction control method, system, device and storage medium

Abstract

The present application relates to the technical field of new energy vehicle battery thermal management, in particular to a new energy vehicle battery thermal management intelligent prediction control method, system, device and storage medium, through collecting vehicle operation parameters, battery electrical parameters and battery pack temperature distribution multi-source data, constructing a standard base matrix to orthogonal projection dimension reduction of temperature field, decoupling extraction of core eigenvectors and correction of battery nonlinear internal resistance, forming a multi-dimensional transient thermodynamic characteristic package; based on the characteristic package, predicting the future driving power trajectory, combining the internal resistance thermal effect and the electrochemical entropy change thermal effect to calculate the dynamic heat generation power, and generating the predicted thermal load through thermal inertia propagation correction; constructing the battery transient thermal state evolution surface, generating the dynamic thermal safety boundary through local curvature mutation inversion, and finally comparing the temperature evolution curve with the safety boundary to generate the grading active temperature control instruction. The problems of traditional thermal management response lag and safety boundary conservation are solved, and the battery thermal safety protection capability and operation efficiency are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of thermal management technology for new energy vehicle batteries, specifically to intelligent predictive control methods, systems, devices, and storage media for thermal management of new energy vehicle batteries. Background Technology

[0002] With the large-scale development and technological iteration of the new energy vehicle industry, lithium-ion power batteries, as the core power source of vehicles, directly determine charging and discharging efficiency, cycle life, and overall vehicle driving safety through their operating temperature. Currently, the industry has formed a battery thermal management technology system based primarily on liquid cooling, supplemented by air cooling and direct cooling. Mainstream control strategies are mostly based on real-time data collected by the battery management system to build a feedback closed loop. By monitoring parameters such as the average temperature of the battery pack, the maximum temperature of individual cells, and charging and discharging current, the operating power of the cooling water pump and radiator fan is adjusted. Some advanced solutions combine the state of charge and ambient temperature for simple feedforward compensation, which can maintain the battery temperature within a suitable operating range under steady-state conditions such as normal constant speed driving and slow charging, thus meeting basic thermal management requirements.

[0003] However, in practical use, it still has some drawbacks. The feedback control architecture inherently has a response lag. The cooling system needs a certain time delay from receiving the command to establishing effective heat dissipation capacity. Under transient high-current conditions such as rapid acceleration and high-power fast charging, the battery heat generation rate increases sharply, and the system cannot intervene in advance, which can easily lead to the rapid accumulation of local hot spots. Existing heat generation calculation models generally use the lumped parameter method, which simplifies the battery pack as a uniform heat-generating body and relies only on the average temperature and basic internal resistance to estimate the heat generation power. This completely ignores the influence of the internal temperature gradient distribution of the battery pack on the nonlinear internal resistance, and does not consider the anisotropic characteristics of heat diffusion in different regions. This results in a large deviation in the heat generation power estimation and makes it impossible to accurately capture local thermal risks. Meanwhile, the industry-standard thermal safety protection mechanisms all use fixed temperature thresholds, which cannot adapt to the decline in thermal stability caused by battery aging, nor can they distinguish the critical conditions for thermal runaway under different operating conditions. This presents a dual contradiction: excessive conservatism increases the energy consumption of the entire vehicle, while insufficient thresholds cause safety hazards. Moreover, most systems lack the ability to adapt to the evolution of thermal characteristics throughout the battery's entire life cycle. As the battery's internal resistance increases and its thermal conductivity deteriorates, its thermal management efficiency will decrease year by year. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, this invention provides an intelligent predictive control method, system, device, and medium for thermal management of new energy vehicle batteries. Through the following solutions, the problems of delayed response and conservative safety boundaries in traditional thermal management are solved, significantly improving battery thermal safety protection capabilities and operating efficiency.

[0005] The intelligent predictive control method for thermal management of new energy vehicle batteries according to the present invention includes:

[0006] S1: During vehicle operation, acquire vehicle speed, vehicle acceleration, and drive motor output torque as the first set of data; acquire total battery current and state of charge as the second set of data; acquire the real-time temperature distribution column vector of the temperature measurement nodes in the battery pack as the third set of data.

[0007] S2: Construct a standard basis matrix, orthogonally project the third set of data into the standard basis matrix space for dimensionality reduction, perform singular value decomposition on the temperature snapshot matrix formed by the third set of data, extract the first k left singular vectors whose cumulative singular value energy ratio reaches a preset threshold, and use them as core eigenvectors; use the core eigenvectors to characterize the main direction of thermal diffusion and the concentration of temperature gradient inside the battery pack, and correct the nonlinear internal resistance of the battery based on the concentration of temperature gradient; finally, a multidimensional transient thermodynamic feature package is formed.

[0008] S3: Based on the multidimensional transient thermodynamic feature package, analyze the expected driving power trajectory of the vehicle in the future prediction time domain; calculate the dynamic heat generation power trajectory through the heat generation power equation including the internal resistance heat effect and the electrochemical entropy change heat effect; calculate the heat propagation delay coefficient according to the rate of change of the dynamic heat generation power trajectory in the continuous time domain, and use the heat propagation delay coefficient to perform thermal inertial propagation correction on the dynamic heat generation power trajectory to generate predicted thermal load time series data.

[0009] S4: Based on the predicted thermal load time series data, core eigenvectors and the rate of change of the total battery current, construct the transient thermal state evolution surface of the battery; calculate the local curvature of the thermal state evolution surface, and generate a dynamic thermal safety boundary including the dynamic maximum temperature threshold based on the temperature value corresponding to the local curvature abrupt change region.

[0010] S5: Compare the temperature evolution curve corresponding to the predicted thermal load time series data with the dynamic thermal safety boundary in the time domain to generate an active temperature control command.

[0011] Preferably, in step S2, the temperature gradient concentration is obtained by obtaining a low-dimensional coefficient vector after orthogonally projecting the temperature distribution column vector onto the standard basis matrix space, calculating the mean and standard deviation of all components of the low-dimensional coefficient vector, and using the ratio of the standard deviation to the mean of the low-dimensional coefficient vector as the temperature gradient concentration at the corresponding time.

[0012] Preferably, in step S2, the correction process of the battery nonlinear internal resistance is as follows: the basic nonlinear internal resistance is calibrated by battery bench test, and the correction coefficient function corresponding to the temperature gradient concentration is calibrated at the same time; the corresponding correction coefficient is calculated according to the temperature gradient concentration at the current time, and the basic nonlinear internal resistance is multiplied by the sum of the correction coefficients to obtain the corrected battery nonlinear internal resistance.

[0013] Preferably, in step S3, the heat propagation delay coefficient is obtained by using the central difference method to calculate the rate of change of heat power at each predicted time step in the continuous time domain for the dynamic heat power trajectory, and using the forward difference method and the backward difference method for the first and last time steps in the predicted time domain, respectively; calculating the basic heat propagation delay coefficient based on the absolute value of the rate of change of heat power; extracting the principal direction vector of heat diffusion, and calculating the ratio of the maximum absolute value of the vector elements to the average absolute value of all elements as the anisotropy correction coefficient; multiplying the basic heat propagation delay coefficient by the anisotropy correction coefficient to obtain the final heat propagation delay coefficient.

[0014] Preferably, in step S3, the method for obtaining the predicted heat load time series data is to use the dynamic heat generation power at the current moment as the initial value, adopt the discrete recursive form of the first-order inertial element, and use the thermal propagation delay coefficient corresponding to each predicted time step to perform time-step thermal inertial propagation correction on the dynamic heat generation power trajectory; and arrange the corrected heat generation power of all time steps in the predicted time domain in chronological order to generate the predicted heat load time series data.

[0015] Preferably, in step S4, the construction method of the battery transient thermal state evolution surface is as follows: a dimension-reduced heat dissipation and conduction model is constructed based on the standard basis matrix. The dimension-reduced heat dissipation and conduction model includes a system state transition matrix and a thermal load input matrix. The low-dimensional coefficient vector at the current moment is used as the initial value, and the predicted low-dimensional coefficient vector for each predicted time step is recursively calculated in combination with the predicted thermal load time series data. The total current change rate of the battery at each predicted time step is calculated, and a current change rate correction term is introduced to correct the predicted low-dimensional coefficient vector. The physical position index of the temperature measurement node in the battery pack is used as the spatial dimension axis, and the predicted time offset is used as the time dimension axis. Based on the corrected predicted low-dimensional coefficient vector and the core eigenvector, the predicted temperature of each temperature measurement node is reconstructed. The reconstructed predicted temperature is used as the temperature dimension axis to construct a three-dimensional physical thermal state evolution surface.

[0016] Preferably, in step S4, the dynamic thermal safety boundary is obtained by calibrating the thermal state abrupt change threshold through a battery bench thermal runaway critical experiment; extracting the time series data of the predicted maximum temperature corresponding to the thermal state evolution surface, calculating the second derivative of the predicted maximum temperature with respect to time as the thermal state change rate, and performing moving average filtering on the thermal state change rate; identifying the moment when the filtered thermal state change rate reaches or exceeds the thermal state abrupt change threshold as the thermal state abrupt change moment, taking the predicted maximum temperature corresponding to the abrupt change moment as the critical temperature at that moment, and using the battery's nominal maximum operating temperature as the basic critical temperature at non-abrupt change moments; and sequentially introducing safety margin correction and battery aging correction into the critical temperature of each predicted time step to generate a dynamic thermal safety boundary that changes with the predicted time.

[0017] The present invention also provides a system for implementing the above-mentioned intelligent predictive control method for thermal management of new energy vehicle batteries, specifically an intelligent predictive control system for thermal management of new energy vehicle batteries, including a data acquisition module, a feature construction module, a thermal load prediction module, a safety boundary generation module, and a temperature control command generation module.

[0018] Data acquisition module: During vehicle operation, it acquires vehicle speed, vehicle acceleration, and drive motor output torque as the first set of data; acquires total battery current and state of charge as the second set of data; and acquires the real-time temperature distribution column vector of the temperature measurement nodes inside the battery pack as the third set of data.

[0019] Feature construction module: Construct a standard basis matrix, orthogonally project the third set of data into the standard basis matrix space for dimensionality reduction, perform singular value decomposition on the temperature snapshot matrix formed by the third set of data, extract the k left singular vectors whose cumulative singular value energy ratio reaches a preset threshold, and use them as core eigenvectors; use the core eigenvectors to characterize the main direction of thermal diffusion and the concentration of temperature gradient inside the battery pack, and correct the nonlinear internal resistance of the battery based on the concentration of temperature gradient; finally, a multidimensional transient thermodynamic feature package is constructed.

[0020] Thermal load prediction module: Based on the multidimensional transient thermodynamic feature package, it analyzes the expected driving power trajectory of the vehicle in the future prediction time domain; and calculates the dynamic heat generation power trajectory through the heat generation power equation that includes the internal resistance heat effect and the electrochemical entropy change heat effect; calculates the heat propagation delay coefficient according to the rate of change of the dynamic heat generation power trajectory in the continuous time domain, and uses the heat propagation delay coefficient to perform thermal inertia propagation correction on the dynamic heat generation power trajectory to generate predicted thermal load time series data.

[0021] Safety boundary generation module: Based on the predicted thermal load time series data, core eigenvectors and the rate of change of the total battery current, construct the transient thermal state evolution surface of the battery; calculate the local curvature of the thermal state evolution surface, and generate a dynamic thermal safety boundary including the dynamic maximum temperature threshold according to the temperature value corresponding to the local curvature abrupt change region.

[0022] The temperature control command generation module compares the temperature evolution curve corresponding to the predicted thermal load time series data with the dynamic thermal safety boundary in the time domain to generate an active temperature control command.

[0023] The present invention also provides a device for implementing the above-mentioned intelligent predictive control method for thermal management of new energy vehicle batteries. Specifically, it is an intelligent predictive control device for thermal management of new energy vehicle batteries, which is an electronic device, including a processor and a memory. The memory is electrically connected to the processor, and the memory stores a computer-executable program. When the processor executes the computer-executable program, it implements the intelligent predictive control method for thermal management of new energy vehicle batteries.

[0024] The present invention also provides a storage medium for implementing the above-mentioned intelligent predictive control method for thermal management of new energy vehicle batteries, specifically an intelligent predictive control medium for thermal management of new energy vehicle batteries, specifically a computer storage medium, including: a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program implements the intelligent predictive control method for thermal management of new energy vehicle batteries when executed by a processor.

[0025] The technical effects and advantages of this invention are as follows:

[0026] 1. This invention adopts a predictive active control architecture to predict the evolution trend of battery thermal state in advance and execute temperature control commands in advance, which completely solves the problem of lag in response of traditional feedback control. It can effectively cope with sudden thermal risks under transient high current conditions such as rapid acceleration and high-power fast charging, and significantly improve the battery thermal safety protection capability.

[0027] 2. This invention decouples high-dimensional temperature field data into low-dimensional core eigenvectors through orthogonal projection dimensionality reduction technology. While retaining most of the thermal energy information, it significantly reduces the amount of computation. At the same time, it introduces temperature gradient concentration to correct the nonlinear internal resistance of the battery, effectively solving the problem of heat generation estimation deviation caused by the traditional model relying only on average temperature, and significantly improving the accuracy of heat generation power calculation.

[0028] 3. By constructing a dynamic thermal safety boundary, this invention can adapt to the thermal limits of batteries under different operating conditions and aging states. By inverting the critical thermal runaway temperature through the thermal curvature abrupt change, and combining the safety margin and aging correction mechanism, it avoids the increase in energy consumption caused by excessively conservative fixed thresholds, and eliminates the risk of thermal runaway caused by excessively high thresholds.

[0029] 4. This invention adopts a graded active temperature control strategy, matching the corresponding cooling power according to the thermal risk level, effectively reducing the average energy consumption of the cooling system while ensuring thermal safety. At the same time, by regularly updating the global standard base matrix in the cloud, the system can continuously adapt to the changes in thermal characteristics throughout the battery's entire life cycle, improving the system's versatility and service life. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the overall method flow of the intelligent predictive control method for thermal management of new energy vehicle batteries according to the present invention.

[0031] Figure 2 This is a schematic diagram of the process for obtaining multi-dimensional transient thermodynamic feature packages in the intelligent predictive control method for thermal management of new energy vehicle batteries according to the present invention.

[0032] Figure 3 This is a flowchart illustrating the process of acquiring predicted thermal load time-series data in the intelligent predictive control method for thermal management of new energy vehicle batteries according to the present invention.

[0033] Figure 4 This is a schematic diagram illustrating the process of obtaining the dynamic thermal safety boundary in the intelligent predictive control method for thermal management of new energy vehicle batteries according to the present invention.

[0034] Figure 5 This is a flowchart illustrating the process of obtaining active temperature control commands in the intelligent predictive control method for thermal management of new energy vehicle batteries according to the present invention.

[0035] Figure 6 This is a schematic diagram of the system described in this invention. Detailed Implementation

[0036] 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.

[0037] Example 1:

[0038] As attached Figure 1 to Figure 5 As shown, this embodiment is an intelligent predictive control method for battery thermal management of new energy vehicles, including:

[0039] S1: During vehicle operation, acquire vehicle speed, vehicle acceleration, and drive motor output torque as the first set of data; acquire total battery current and state of charge as the second set of data; acquire the real-time temperature distribution column vector of the temperature measurement nodes in the battery pack as the third set of data.

[0040] It should be noted that the vehicle speed, overall vehicle acceleration, and drive motor output torque are acquired as the first set of data during vehicle operation. The specific process is as follows:

[0041] Vehicle speed: Pulse signals are collected by magnetoelectric wheel speed sensors installed on the four drive wheels of the vehicle, and the real-time vehicle speed is calculated by the vehicle control unit (VCU). The measurement accuracy is [not specified]. ±0.5km / h Measurement range 0~250km / h Response time ≤ 10ms The wheel speed sensor uses a non-contact mounting method, and the clearance between it and the gear ring is controlled within a certain range. 0.5~1.5mm Within the range, avoid mechanical wear and signal interference; sampling rate 10Hz It maintains consistency with the vehicle's CAN bus communication cycle to ensure the real-time and synchronous nature of vehicle speed data.

[0042] Vehicle acceleration: Selected MEMS Triaxial accelerometer, measurement range -5g~+5gMeasurement accuracy ±0.01g Frequency response range 0~200Hz Installed at the center of gravity of the vehicle chassis, rigidly fixed with bolts, with a flatness error of the mounting surface. ≤0.1mm To avoid measurement deviations caused by loose installation; the sampling rate is 10Hz, and it is collected synchronously with the vehicle speed signal.

[0043] Drive motor output torque: via drive motor controller MCU of CAN Real-time reading from the bus interface; measurement range covers the motor's rated torque. -150%～+150% Measurement accuracy ±1N·m Update cycle 100ms The torque signal is from MCU The data is calculated in real time based on the motor stator current, rotor position, and field-oriented control algorithm, requiring no additional sensors; sampling rate... 10Hz It keeps synchronized with the vehicle speed and acceleration signals.

[0044] It should be further explained that the total battery current and state of charge were obtained as the second set of data, and the specific process is as follows:

[0045] Total battery current: A closed-loop Hall current sensor is selected, with a measurement range of... -300A~+300A Measurement accuracy ± 0.5%FS Response time ≤1μs It is connected in series at the positive terminal busbar of the battery pack, using a rigid copper busbar connection, with a contact resistance of... ≤ 0.1mΩ To avoid heat generation and voltage drop under high current; sampling rate 20Hz The sampling rate is higher than other signals to capture transient current fluctuations during charging and discharging. The sampled data is processed by a first-order low-pass filter, with a cutoff frequency of [missing information]. 5Hz To suppress high-frequency electromagnetic interference.

[0046] Battery state of charge: determined by the battery management system BMS The calculation is performed in real time using the ampere-hour integration method combined with the open-circuit voltage calibration method, and the calculation accuracy is [not specified]. ±3% Within; the time step of ampere-hour integration is consistent with the current sampling step, and the initial integration value is determined by the open-circuit voltage calibration value when the vehicle was stationary for the last time; the open-circuit voltage calibration is performed when the vehicle has been stationary for more than 30 minutes and the average battery temperature is within; 10℃～45℃ The calibration process is executed automatically within the specified interval, disconnecting all high-voltage loads during the calibration process to ensure accurate open-circuit voltage measurement; sampling rate 20Hz It is updated synchronously with the total battery current signal.

[0047] It should be further explained that the real-time temperature distribution column vector of the temperature measurement nodes inside the battery pack is obtained as the third set of data. The specific process is as follows:

[0048] Temperature measurement node arrangement: pre-arranged inside the battery pack N indivual NTC The thermistor temperature measurement nodes include one temperature measurement point at the positive electrode tab and one at the middle of the casing of each individual cell, and one temperature measurement point at the coolant inlet, coolant outlet, high voltage junction box and main control board of the battery management system of the battery pack. NTC Thermistor measurement range - 40℃～+85℃ Measurement accuracy ±0.5℃ nominal B 0 value 3435K±1% Response time ≤5s All temperature sensing points are fixed with thermally conductive adhesive, and the thermal conductivity of the adhesive is [missing value]. ≥1.5W / (m·K) This ensures the accuracy and speed of temperature measurement.

[0049] Temperature distribution column vector construction: At each sampling time, the BMS reads the temperature values ​​of all temperature measurement nodes in a preset fixed physical order and arranges them into a real-time temperature distribution column vector of length N. T(t)=[T 1 (t),T 2 (t),...,T N （t）] T in T i （t） Indicates the first i Each temperature measurement node is located at t The temperature value at any given time; the sampling rate of 5Hz can accurately track the slow changing trend of the temperature field inside the battery pack, while controlling the amount of data storage and transmission; a system timestamp from the same source as the first and second groups of data is added to each temperature distribution column vector to ensure the temporal consistency of multi-source data.

[0050] The final output consists of three sets of time-series data with unified high-precision system timestamps. The first set of data consists of vehicle speed, vehicle acceleration, and drive motor output torque. The second set of data consists of total battery current and state of charge. The third set of temperature distribution column vector data consists of real-time temperature values ​​of temperature measurement nodes within the battery pack.

[0051] S2: Construct a standard basis matrix, orthogonally project the third set of data into the standard basis matrix space for dimensionality reduction, and perform singular value decomposition on the temperature snapshot matrix formed by the third set of data to extract the singular value cumulative energy ratios that reach a preset threshold. kA left singular vector is used as the core eigenvector; the core eigenvector is used to characterize the main direction of thermal diffusion and the concentration of temperature gradient inside the battery pack, and the nonlinear internal resistance of the battery is corrected based on the concentration of temperature gradient; finally, a multidimensional transient thermodynamic feature package is formed.

[0052] It should be specifically explained that a standard basis matrix is ​​constructed, and the third set of data is orthogonally projected into the standard basis matrix space for dimensionality reduction. Then, singular value decomposition is performed on the temperature snapshot matrix formed by the third set of data, and the top k left singular vectors whose cumulative singular value energy ratio reaches a preset threshold are extracted as core eigenvectors. The specific process is as follows:

[0053] Initial construction of the standard basis matrix: The global initial standard basis matrix obtained through bench testing; the bench testing covers the entire operating range of the target battery pack ( -20℃～60℃ Ambient temperature, 0～1C Charge / discharge rate, 0~120km / h (Equivalent vehicle speed), collect more than 10,000 sets of temperature distribution data to construct a temperature snapshot matrix. Where N is the total number of temperature measurement nodes, M is the total number of historical sampling times, and each column of the matrix corresponds to a temperature distribution column vector T(t) for a sampling time; performing singular value decomposition on the snapshot matrix yields: ,in It is a left singular vector matrix, and its column vectors are orthogonal eigenvectors; For a singular value diagonal matrix, the diagonal elements These are the singular values ​​sorted by energy from highest to lowest. It is a right singular vector matrix.

[0054] During vehicle operation, the accumulated temperature distribution data is automatically uploaded to the cloud big data platform every 3 months. Based on the full operating data of all vehicles of the same model, the cloud performs singular value decomposition again and updates the global standard basis matrix, and then pushes the updated matrix to all online vehicles.

[0055] Core eigenvector selection: before calculation k Cumulative energy percentage of each eigenvector E k :

[0056] ,in For the first i A singular value, k The number of core eigenvectors to be selected. N This represents the total number of temperature measurement nodes. Select those that meet the following criteria. E k ≥ 95% The smallest integer k The number of core eigenvectors (typically valued as follows)3～8） Extract the front of the left singular vector matrix k Constructing orthogonal basis matrices This matrix is ​​the standard basis matrix.

[0057] Online orthogonal projection dimensionality reduction: During the real-time operation of the vehicle, the temperature distribution column vector acquired at each sampling time is transformed... T(t) Orthogonal projection onto the standard basis matrix U k The resulting low-dimensional subspace yields a low-dimensional coefficient vector: ,in It is the transpose of the standard basis matrix. T(t) for t The column vector of temperature distribution at time points. k The number of core eigenvectors; c(t)=[c 1 (t),c 2 (t),...,c k （t）] T ，c i （t） For the first i The projection coefficients corresponding to the core eigenvectors; this projection process reduces the dimensionality of the N-dimensional high-dimensional temperature distribution data to... k While maintaining over 95% of the energy information of the internal thermal field of the battery pack, it achieves decoupling and compression of the thermal field.

[0058] It should be further explained that the core eigenvector is used to characterize the principal direction of thermal diffusion and the concentration of temperature gradient inside the battery pack, and the nonlinear internal resistance of the battery is corrected based on the concentration of temperature gradient. The specific process is as follows:

[0059] Characterization of principal directions of thermal diffusion: standard basis matrix U k The first column vector u 1 (corresponding to the maximum singular value) σ 1 The direction of heat diffusion inside the battery pack is represented by the sign and magnitude of the element, which reflects the temperature contribution of the corresponding temperature measurement node in the direction of heat diffusion.

[0060] In step S2, the temperature gradient concentration is calculated: the temperature gradient concentration is used to quantify the non-uniformity of the temperature distribution inside the battery pack, and is calculated using the dimensionless coefficient of variation. ,in The mean of the low-dimensional coefficient vector. The standard deviation of the low-dimensional coefficient vector. k The number of core eigenvectors,c i （t） The projection coefficients corresponding to the i-th core eigenvector; G(t) It is a dimensionless quantity under normal operating conditions. G （t） It is usually between 0 and 1. The larger the value, the more uneven the temperature distribution and the greater the local temperature gradient. G(t) =0 indicates that the temperature is completely uniformly distributed.

[0061] In step S2, the battery nonlinear internal resistance is corrected: the basic nonlinear internal resistance and correction coefficient function are calibrated through battery bench experiments; the basic nonlinear internal resistance... R 0 (SOC,T avg （t）） For different states of charge SOC Different average temperatures T avg （t） The internal resistance reference value is given below, where the average temperature is... , N This represents the total number of temperature measurement nodes. T i （t） For the first i Each temperature measurement node is located at t Temperature value at any given time; In this invention, the basic nonlinear internal resistance R 0 (SOC,T avg （t）） Different states of charge (SOC) The calibration method for the battery internal resistance benchmark value at different average battery temperatures is based on well-known common sense and conventional technical means in the field of lithium-ion power batteries, and combines... HPPC Test method or DCR The parameters obtained by the test method are the basic nonlinear internal resistance data required through conventional experiments; correction coefficients. α(G(t)) Using the form of a quadratic function, satisfying α（0）=0 And it is monotonically increasing:

[0062] ,in k 1 、k 2 These are constants calibrated through bench tests, with a typical range of values. , , G(t) for t The concentration of the temperature gradient at different times; the calibration experimental method is as follows: control different concentrations of the temperature gradient on the battery platform. G Measure the actual internal resistance under the corresponding operating conditions. RThe least squares method was used to fit the data. k 1 and k 2 The value of .

[0063] The corrected nonlinear internal resistance of the battery is: ,in R 0 (SOC,T avg （t）） Based on the nonlinear internal resistance, α(G(t)) This is the temperature gradient concentration correction coefficient; this correction mechanism takes into account the impact of local temperature unevenness on the battery's internal resistance, and solves the problem of heat generation calculation error caused by the traditional internal resistance model relying only on average temperature.

[0064] It should be further explained that the final formation of the multidimensional transient thermodynamic feature package is as follows:

[0065] Because the three sets of data have different sampling rates (the first set) 10Hz Group 2 20Hz Group 3 5Hz ),by 10Hz To standardize the baseline sampling rate; for the second set of data (total battery current, state of charge), a cutoff frequency of [missing value] was used. 5Hz After anti-aliasing low-pass filtering, the data is downsampled to 10Hz; the low-dimensional coefficient vector derived from the third set of data... c(t) Temperature gradient concentration G(t) Corrected internal resistance R(t) The frequency was upsampled to 10Hz using linear interpolation.

[0066] Based on the system timestamp, the first set of data (vehicle speed) at the same time is... v(t) longitudinal acceleration of the whole vehicle A ccx （t） , drive motor output torque T m （t） The second set of data (total battery current) I(t) State of charge SOC(t) Corrected nonlinear internal resistance of the battery R(t) and low-dimensional coefficient vectors c(t) Perform strict alignment.

[0067] All aligned features are concatenated sequentially to form a 6+k dimensional original feature vector:

[0068] ,in v(t) for t Vehicle speed at any given moment A ccx（t） for t The longitudinal acceleration of the entire vehicle at any given moment. T m （t） Let I(t) be the output torque of the drive motor at time t. t The total battery current at any given moment. SOC(t) )for t The state of charge of the battery at any given time. R(t) for t The nonlinear internal resistance of the battery after constant correction. c i （t） For the first i The projection coefficients corresponding to the core eigenvectors. k The number of core eigenvectors.

[0069] To eliminate the influence of differences in the dimensions of each component on subsequent calculations, the original eigenvector is standardized by zero mean and unit variance for each component: ,in F raw,i （t） The first feature vector is the first feature vector. i One portion, μ i This is the global mean of the component obtained during the offline training phase. s i This is the global standard deviation of the component obtained during the offline training phase; after standardization, the final multidimensional transient thermodynamic feature package is obtained. F(t) All components are dimensionless.

[0070] S3: Based on the multidimensional transient thermodynamic feature package, analyze the expected driving power trajectory of the vehicle in the future prediction time domain; calculate the dynamic heat generation power trajectory through the heat generation power equation including the internal resistance heat effect and the electrochemical entropy change heat effect; calculate the heat propagation delay coefficient according to the rate of change of the dynamic heat generation power trajectory in the continuous time domain, and use the heat propagation delay coefficient to perform thermal inertial propagation correction on the dynamic heat generation power trajectory to generate predicted thermal load time series data.

[0071] It should be specifically noted that, based on the aforementioned multidimensional transient thermodynamic feature package, the expected driving power trajectory of the vehicle in the future prediction time domain is analyzed, and the specific process is as follows:

[0072] Prediction Time Domain and Input Window Settings: Set the future prediction time domain to... T horizon =10s And time step Δt= 0.1s (Consistent with the update frequency of the multidimensional transient thermodynamic feature package), outputting driving power values ​​for 100 prediction time steps; input historical window length is...T hist =1s That is, the multidimensional transient thermodynamic feature packet sequence of the most recent 10 time moments is used as the input of the prediction model.

[0073] Driving power prediction model construction: A two-layer bidirectional long short-term memory network is used. (BiLSTM) Construct a driving power prediction model with the following input dimensions: 6+k (Consistent with the feature package dimension), the hidden layer dimension is 64, and the output dimension is 100 (corresponding to 100 prediction time steps); the model is pre-trained offline using real vehicle operation data, and the average absolute error of prediction is controlled within 5%.

[0074] During the real-time operation of the vehicle, a sequence of multidimensional transient thermodynamic feature packets at 10 consecutive time points will be generated. [F(t-9Δ t),F(t-8Δt),...,F(t) Input the prediction model to obtain the initial expected driving power trajectory. P drive,raw (t+δ) ,in δ ∈[0,T horizon ] This is for predicting the time offset.

[0075] The final output is the expected drive power trajectory over the next 10 seconds. P drive (t+δ ), δ=0,0.1,0.2,...,10s .

[0076] It should be further explained that the dynamic heat generation power trajectory is calculated by combining the desired driving power trajectory with the nonlinear internal resistance of the battery. The specific process is as follows:

[0077] Battery output current prediction: Based on the expected drive power trajectory, combined with motor efficiency. η m Inverter efficiency η inv and dynamic battery terminal voltage U batt (t+δ) Calculate the total battery current at each predicted time step:

[0078] ,in η m and η inv The efficiency values ​​are pre-calibrated and take the following values: 0.92 and 0.95 ; U batt (t+δ) Battery terminal voltage, based on dynamic SOCThe predicted values ​​are obtained by looking up the table:

[0079] ,in C nom Battery nominal capacity (unit) Ah ), U batt (t+δ) Based on the battery management system SOC(t+δ) and T avg （t） According to the table, the solution is obtained by iterative calculation, and one iteration is sufficient to meet the engineering accuracy requirements.

[0080] Dynamic heat generation power calculation: The battery's heat generation power consists of two parts: the internal resistance heat effect (Joule heating) and the electrochemical entropy change heat effect. The calculation formula is as follows:

[0081] ,in R(t) To correct the nonlinear internal resistance of the battery; T avg （t） This represents the current average battery temperature. S coeff (SOC) This is the battery entropy change coefficient, in units of... V / K , is a function of SOC that has been calibrated in advance through battery bench experiments, and its value range is . -0.5mV / K~+0.5mV / K During discharge, the value is usually negative, indicating heat absorption; during charging, the value is usually positive, indicating heat release.

[0082] The final output is a dynamic heat generation power trajectory synchronized with the desired drive power trajectory. P heat (t+δ) , δ=0,0.1, 0.2,...,10s .

[0083] It should be further explained that in step S3, the heat propagation delay coefficient is calculated based on the rate of change of the dynamic heat generation power trajectory in the continuous time domain. The specific process is as follows:

[0084] Calculation of the rate of change of heat production power: The central difference method is used to calculate the time-domain rate of change of heat production power at each prediction time step.

[0085] ,in Δt=0.1s The time step is denoted as ; for the first and last time steps in the prediction time domain, forward differencing and backward differencing are used for calculation, respectively.

[0086] Calculation of the basic heat propagation delay coefficient: The basic heat propagation delay coefficient is inversely proportional to the absolute value of the rate of change of heat production power. The faster the heat production power changes, the smaller the heat propagation delay. The calculation formula is as follows:

[0087] ,in k 0 The thermal inertia constant of the battery pack is determined through bench testing, and its value range is [value range missing]. 0.5～2W ; ε Minimum value (take) 10 -6 ), used to avoid the denominator being zero.

[0088] Anisotropy correction: Anisotropy correction is performed by introducing the principal direction of thermal diffusion u1, with the correction factor being:

[0089] ,in u 1,i The principal direction vector of thermal diffusion u 1 The i One element, N This represents the total number of temperature measurement nodes. β This reflects the degree of concentration of heat diffusion. β The larger the value, the more concentrated the heat diffusion is in a local area, and the longer the heat propagation delay.

[0090] The final heat propagation delay factor is:

[0091] ,in for The thermal propagation delay coefficient corresponding to time , in units of s This coefficient takes into account both the dynamic changes in heat generation power and the influence of thermal diffusion anisotropy, and can more accurately characterize the thermal inertial characteristics inside the battery pack.

[0092] It should be further explained that in step S3, the dynamic heat generation power trajectory is corrected by thermal inertia propagation using the aforementioned thermal propagation delay coefficient to generate predicted thermal load time series data. The specific process is as follows:

[0093] The thermal inertial propagation correction is implemented using a discrete recursive form of a first-order inertial element, which is suitable for embedded computing in battery management systems. The recursive formula is as follows:

[0094] ,in P corr (t+δ) The corrected heat production power; initial conditions are: P corr (t)=P heat （t） That is, the corrected heat production power at the current moment is equal to the actual heat production power.

[0095] After correction, the corrected heat generation power of all predicted time steps is arranged in chronological order to generate predicted heat load time series data: Q pred =[P corr (t),P corr (t+0.1s}),...,P corr (t+10s)] .

[0096] Final output: predicted thermal load time series data Q pred Its time resolution is 0.1s The predicted duration is 10s .

[0097] S4: Based on the predicted thermal load time series data, core eigenvectors and the rate of change of the total battery current, construct the transient thermal state evolution surface of the battery; calculate the local curvature of the thermal state evolution surface, and generate a dynamic thermal safety boundary including the dynamic maximum temperature threshold based on the temperature value corresponding to the local curvature abrupt change region.

[0098] It should be specifically noted that in step S4, based on the predicted thermal load time series data, the core eigenvector, and the rate of change of the total battery current, a transient thermal state evolution surface of the battery is constructed. The specific process is as follows:

[0099] Construction of a dimension-reduced heat conduction model: based on orthogonal basis matrices U_k Projecting the three-dimensional heat conduction equation onto k By using the eigenspace of dimensionless eigenvalues, we obtain the dimensionless heat dissipation conduction recurrence equation:

[0100] ,in for t+δ The predicted low-dimensional coefficient vector at time step; The system state transition matrix is ​​dimensionless. The thermal load input matrix has the following dimensions: ℃ / W ; Q pred (t+δ) for t+δ Predicted heat load at time t, in units of W ; Δt=0.1s The time step is [time step size]; the initial conditions are [initial conditions]. c pred (t)=c(t) That is, the actual low-dimensional coefficient vector at the current moment.

[0101] System matrix and input matrix calibration: The A and B matrices were calibrated through battery bench thermal experiments. During the experiment, step thermal loads of different powers were applied to the battery pack, and temperature response data of each temperature measurement node were collected to construct the input and output datasets. The least squares method was used to identify the parameters of the dimensionless heat conduction model to obtain the optimal values ​​of the A and B matrices. After calibration, the matrix parameters were stored offline in the BMS and updated every 6 months with cloud data.

[0102] Correction for total battery current change rate: The total battery current change rate at each prediction time step is calculated using the central difference method.

[0103] ,in I(t+δ) is t+δ Predicted total battery current at time, in units of A .

[0104] For the first and last time steps in the prediction time domain, forward differencing and backward differencing are used for calculation, respectively. A current change rate correction term is introduced to correct the coefficients corresponding to higher-order eigenvectors. The correction formula is as follows:

[0105] ,in c pred,i (t+δ) To correct the previous one i Prediction coefficients corresponding to each eigenvector; For the revised version i Prediction coefficients corresponding to each eigenvector; γ i Let be the current correction coefficient for the i-th eigenvector, which is pre-calibrated experimentally and satisfies . γ 1 <γ 2 <...<γ k That is, higher-order eigenvectors are more sensitive to changes in current, and their typical range is... γ i ∈[0,0.05s / A] .

[0106] Battery transient thermal state evolution surface construction: based on the physical location sequence of temperature measurement nodes within the battery pack. x The spatial dimension axis is δ, the predicted time offset is δ, and the reconstructed predicted temperature is δ. T pred (x,t+δ) Construct a three-dimensional physical thermal state evolution surface with temperature as the axis:

[0107] ,in x=1,2,...,N This refers to the physical location number of the temperature measurement node; δ=0,0.1,0.2,...,10s To predict the time offset; ui,x For the first i The core eigenvector at the th th x The element values ​​at each temperature measurement node; this surface fully characterizes the spatiotemporal evolution of temperature at various locations within the battery pack over the next 10 seconds, with all dimensions having clear physical meaning. It should be further noted that the calculation of the local curvature of the aforementioned thermal state evolution surface is as follows:

[0108] Maximum temperature time series extraction: based on the main direction of thermal diffusion u 1 Pre-determine the location of the hottest area within the battery pack (i.e. u 1 (The location of the temperature measurement node corresponding to the maximum value of the element); at each prediction time step δ The predicted temperature of the hottest region is directly extracted as the predicted maximum temperature at that moment. T max,pred (δ) .

[0109] Calculation of the rate of change of thermal state: The second derivative of the predicted maximum temperature with respect to time is used as a quantitative indicator of the severity of the change in thermal state, replacing the original mixed-dimensional mean curvature. The calculation formula is as follows:

[0110] ,in H(δ) for t+δ The rate of change of thermal state at time t, with dimensions of ℃ / s 2 Physically, it refers to the acceleration at the highest temperature, which can accurately reflect the characteristics of a rapid temperature rise in the early stages of thermal runaway.

[0111] Discrete calculations are performed using the second-order central difference method:

[0112] For the first and last two time steps in the prediction time domain, forward difference and backward difference are used for calculation, respectively.

[0113] To suppress spurious abrupt changes caused by measurement noise and calculation errors, a 5-point moving average filter was applied to the original thermal state change rate data to obtain a smoothed thermal state change rate:

[0114] This filtering process effectively removes high-frequency noise interference while preserving the true characteristics of thermal runaway precursors, thus improving the accuracy of subsequent abrupt change region identification.

[0115] It should be further explained that in step 4 of this embodiment, a dynamic thermal safety boundary, including a dynamic maximum temperature threshold, is generated based on the temperature value corresponding to the local curvature abrupt change region. The specific process is as follows:

[0116] Thermal state change threshold calibration: Through battery bench thermal runaway critical experiments, different data were collected. SOC Complete temperature evolution data of the battery from normal state to thermal runaway critical state under different initial temperatures and different charge / discharge rates, and calculate the corresponding thermal state change rate sequence.

[0117] Determine the threshold for thermal state abrupt change H th ,when H smooth (δ)≥H th At that time, it was determined that the moment was a sudden change in thermal state, and there was a risk of thermal runaway; H th The range of values ​​is 0.5～2.0℃ / s 2 The specific value was determined through experiments.

[0118] Time-by-time critical temperature inversion: Traversing all prediction time steps δ Identify all that meet the requirements H smooth (δ)≥H th The moment of thermal state abrupt change. For each abrupt change moment... δ The predicted highest temperature T at that moment max,pred (δ) represents the critical thermal runaway temperature at that moment. T crit (δ) For non-abrupt periods, the battery's nominal maximum operating temperature is used. T nom =55℃ As the fundamental critical temperature.

[0119] Dynamic thermal safety boundary generation: For each predicted time step δ, based on the critical temperature at the corresponding time, safety margin correction and battery aging correction are introduced to generate the dynamic thermal safety boundary at that time.

[0120] ,in ΔT safe =2℃ A fixed safety margin is used to compensate for prediction errors; The battery aging correction is expressed as a linear function:

[0121] ΔT aging (SOH)=k aging ·(1-SOH) ,in k aging This is an aging correction factor, with a value of 15℃; SOHThis represents the battery's health status, with a value ranging from 0 to 1. This correction mechanism takes into account the decrease in thermal stability caused by battery aging. SOH The lower the value, the stricter the safety boundary; the final output is a dynamic thermal safety boundary curve that varies over time. T bound (δ) , δ=0,0.1,0.2,..., 10s This curve can adapt to the battery thermal safety limit under different operating conditions and aging states, avoiding the overly conservative nature of a fixed threshold.

[0122] S5: Compare the temperature evolution curve corresponding to the predicted thermal load time series data with the dynamic thermal safety boundary in the time domain to generate an active temperature control command.

[0123] It should be noted that the specific process for extracting and comparing the predicted temperature evolution curve in the time domain is as follows:

[0124] Extraction of the predicted maximum temperature evolution curve: Directly obtain the predicted maximum temperature evolution curve. T max,pred (δ) The curve is based on the main direction of thermal diffusion. u_1 The location of the pre-determined hottest region is extracted; where δ=0,0.1,0.2,...,10s To predict the time offset, the evolution trend of the highest temperature inside the battery pack within the next 10 seconds was fully characterized.

[0125] Time-domain point-by-point comparison: comparing the predicted maximum temperature evolution curve with the dynamic thermal safety boundary curve T bound (δ) Perform time-by-time comparisons and calculate the temperature excess at each predicted time step: ΔT(δ)=T max,pred (δ)-T bound (δ) in ΔT(δ) is t+δ The predicted temperature exceeds the limit at any given time, in units of ℃ ; If ΔT(δ)>0 This indicates that the predicted maximum temperature at that moment will exceed the dynamic thermal safety boundary, posing a thermal safety risk.

[0126] Continuous trigger determination: To avoid false triggers caused by prediction errors and instantaneous fluctuations, a continuous trigger mechanism is introduced; triggers are only applied when there is continuous triggering. n One prediction time step ( n=3 (corresponding to 0.3 seconds) all meet the requirements. ΔT(δ)>0 Only when the temperature exceeds a certain threshold is the event considered a valid thermal risk event, triggering the generation of subsequent control commands; simultaneously, the maximum temperature exceedance is recorded. ΔT max =max(ΔT(δ)) and duration of riskt risk ,in t risk To continuously satisfy ΔT(δ)>0 The prediction time step multiplied by the time step size Δt The unit is s This is used to determine the intensity of control interventions.

[0127] It should be further explained that the generation of graded active temperature control commands based on thermal risk levels follows this process: Thermal risk level classification: based on the maximum temperature exceeding the limit... ΔT max and duration of risk t risk Thermal risks are divided into three levels:

[0128] Mild heat risk: 0<ΔT max ≤3℃ and t risk <1s；

[0129] Moderate heat risk: 3℃<ΔT max ≤6℃ or t risk ≥1s；

[0130] Severe heat risk: ΔT max >6℃ .

[0131] Generation of graded active temperature control commands: Active temperature control commands of corresponding strength are generated for different thermal risk levels and sent to the battery thermal management system controller.

[0132] Mild thermal risk: Generate a mild cooling command, controlling the cooling water pump speed to increase to 50% of the rated speed and the cooling fan speed to increase to 30% of the rated speed, thereby reducing the base temperature of the battery pack in advance.

[0133] Moderate heat risk: Generates a moderate cooling command, controlling the cooling water pump speed to increase to 80% of the rated speed and the cooling fan speed to increase to 70% of the rated speed, enhancing heat dissipation capacity to suppress the temperature rise trend.

[0134] Severe thermal risk: Generates a severe cooling command, controlling the cooling water pump and cooling fan to run at 100% of their rated speed to maximize heat dissipation power, while simultaneously opening the emergency heat dissipation channel inside the battery pack (if equipped).

[0135] All active temperature control commands are issued Δt in advance of the predicted risk occurrence time. lead=Executes in 2 seconds, taking advantage of the response delay characteristics of the thermal management system to ensure that the heat dissipation capacity improvement is completed before the temperature reaches the safe boundary.

[0136] Example 2:

[0137] As attached Figure 6 As shown, this embodiment is an intelligent predictive control system for thermal management of new energy vehicle batteries, used in the intelligent predictive control method for thermal management of new energy vehicle batteries described in Embodiment 1. It includes a data acquisition module, a feature construction module, a thermal load prediction module, a safety boundary generation module, and a temperature control command generation module.

[0138] Data acquisition module: During vehicle operation, it acquires vehicle speed, vehicle acceleration, and drive motor output torque as the first set of data; acquires total battery current and state of charge as the second set of data; and acquires the real-time temperature distribution column vector of the temperature measurement nodes inside the battery pack as the third set of data.

[0139] Feature construction module: Construct a standard basis matrix, orthogonally project the third set of data into the standard basis matrix space for dimensionality reduction, perform singular value decomposition on the temperature snapshot matrix formed by the third set of data, extract the k left singular vectors whose cumulative singular value energy ratio reaches a preset threshold, and use them as core eigenvectors; use the core eigenvectors to characterize the main direction of thermal diffusion and the concentration of temperature gradient inside the battery pack, and correct the nonlinear internal resistance of the battery based on the concentration of temperature gradient; finally, a multidimensional transient thermodynamic feature package is constructed.

[0140] Thermal load prediction module: Based on the multidimensional transient thermodynamic feature package, it analyzes the expected driving power trajectory of the vehicle in the future prediction time domain; and calculates the dynamic heat generation power trajectory through the heat generation power equation that includes the internal resistance heat effect and the electrochemical entropy change heat effect; calculates the heat propagation delay coefficient according to the rate of change of the dynamic heat generation power trajectory in the continuous time domain, and uses the heat propagation delay coefficient to perform thermal inertia propagation correction on the dynamic heat generation power trajectory to generate predicted thermal load time series data.

[0141] Safety boundary generation module: Based on the predicted thermal load time series data, core eigenvectors and the rate of change of the total battery current, construct the transient thermal state evolution surface of the battery; calculate the local curvature of the thermal state evolution surface, and generate a dynamic thermal safety boundary including the dynamic maximum temperature threshold according to the temperature value corresponding to the local curvature abrupt change region.

[0142] The temperature control command generation module compares the temperature evolution curve corresponding to the predicted thermal load time series data with the dynamic thermal safety boundary in the time domain to generate an active temperature control command.

[0143] Example 3:

[0144] This embodiment is a device for implementing Embodiment 1 above, specifically an electronic device, including a processor, a memory, and a computer program stored in the memory and capable of running on the processor.

[0145] It should be specifically noted that the device includes at least one processor, at least one memory communicatively connected to the processor, and a communication bus; the processor and the memory communicate and interact with each other through the communication bus; when the computer program stored in the memory is called by the processor, it executes all the steps of the aforementioned intelligent predictive control method for thermal management of new energy vehicle batteries.

[0146] It should be further noted that the device also includes a CAN bus communication interface, a power module, and a wireless communication module, which are used for vehicle system data interaction, power supply, and cloud data transmission, respectively. The device may be a functional module integrated into the BMS, an independent thermal management controller, a power domain controller, or a cloud server. This embodiment does not limit its specific form.

[0147] Example 4:

[0148] This embodiment uses a storage medium to implement the storage medium of Embodiment 1 above, specifically a computer storage medium. A computer storage medium stores a computer program on the computer-readable storage medium. When the computer program is executed by a processor, it implements all the steps of the aforementioned intelligent predictive control method for thermal management of new energy vehicle batteries.

[0149] It should be specifically noted that the storage medium is a non-transitory storage medium with a physical carrier; when the computer program is executed, it fully implements the aforementioned thermal management predictive control process and outputs graded temperature control and power intervention commands; the program supports online incremental updates and can be upgraded via cloud download without modifying the vehicle hardware.

[0150] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments disclosed in this invention. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other.

[0151] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent predictive control of battery thermal management in new energy vehicles, characterized in that, include: S1: During vehicle operation, acquire vehicle speed, vehicle acceleration, and drive motor output torque as the first set of data; acquire total battery current and state of charge as the second set of data; Obtain the real-time temperature distribution column vector of the temperature measurement nodes inside the battery pack as the third set of data; S2: Construct a standard basis matrix, orthogonally project the third set of data onto the standard basis matrix space for dimensionality reduction, and perform singular value decomposition on the temperature snapshot matrix formed by the third set of data to extract the first k left singular vectors whose cumulative singular value energy ratio reaches a preset threshold as core eigenvectors. The core eigenvectors are used to characterize the main direction of thermal diffusion and the concentration of temperature gradient inside the battery pack, and the nonlinear internal resistance of the battery is corrected based on the concentration of temperature gradient; finally, a multidimensional transient thermodynamic feature package is formed. S3: Based on the multidimensional transient thermodynamic feature package, analyze the expected driving power trajectory of the vehicle in the future prediction time domain; calculate the dynamic heat generation power trajectory through the heat generation power equation including the internal resistance heat effect and the electrochemical entropy change heat effect; calculate the heat propagation delay coefficient according to the rate of change of the dynamic heat generation power trajectory in the continuous time domain, and use the heat propagation delay coefficient to perform thermal inertial propagation correction on the dynamic heat generation power trajectory to generate predicted thermal load time series data; S4: Based on the predicted thermal load time series data, core eigenvectors and the rate of change of the total battery current, construct the transient thermal state evolution surface of the battery; calculate the local curvature of the thermal state evolution surface, and generate a dynamic thermal safety boundary including the dynamic maximum temperature threshold based on the temperature value corresponding to the local curvature abrupt change region. S5: Compare the temperature evolution curve corresponding to the predicted thermal load time series data with the dynamic thermal safety boundary in the time domain to generate an active temperature control command.

2. The intelligent predictive control method for thermal management of new energy vehicle batteries according to claim 1, characterized in that: In step S2, the temperature gradient concentration is obtained by obtaining a low-dimensional coefficient vector after orthogonally projecting the temperature distribution column vector onto the standard basis matrix space, calculating the mean and standard deviation of all components of the low-dimensional coefficient vector, and taking the ratio of the standard deviation to the mean of the low-dimensional coefficient vector as the temperature gradient concentration at the corresponding time.

3. The intelligent predictive control method for thermal management of new energy vehicle batteries according to claim 1, characterized in that: In step S2, the correction process for the nonlinear internal resistance of the battery involves calibrating the basic nonlinear internal resistance through a battery bench test, and simultaneously calibrating the correction coefficient function corresponding to the temperature gradient concentration; calculating the corresponding correction coefficient based on the temperature gradient concentration at the current moment, and multiplying the basic nonlinear internal resistance by the sum of the correction coefficients to obtain the corrected nonlinear internal resistance of the battery.

4. The intelligent predictive control method for thermal management of new energy vehicle batteries according to claim 1, characterized in that: In step S3, the heat propagation delay coefficient is obtained by using the central difference method to calculate the rate of change of heat power at each predicted time step in the continuous time domain for the dynamic heat power trajectory. The forward difference method and the backward difference method are used for the first and last time steps in the predicted time domain, respectively. The basic heat propagation delay coefficient is calculated based on the absolute value of the rate of change of heat power. The principal direction vector of heat diffusion is extracted, and the ratio of the maximum absolute value of the vector elements to the average absolute value of all elements is used as the anisotropy correction coefficient. The basic heat propagation delay coefficient is multiplied by the anisotropy correction coefficient to obtain the final heat propagation delay coefficient.

5. The intelligent predictive control method for thermal management of new energy vehicle batteries according to claim 1, characterized in that: In step S3, the method for obtaining the predicted heat load time series data is to use the dynamic heat generation power at the current moment as the initial value, adopt the discrete recursive form of the first-order inertial element, and use the thermal propagation delay coefficient corresponding to each predicted time step to perform time-step thermal inertial propagation correction on the dynamic heat generation power trajectory; and arrange the corrected heat generation power of all time steps in the predicted time domain in chronological order to generate the predicted heat load time series data.

6. The intelligent predictive control method for thermal management of new energy vehicle batteries according to claim 1, characterized in that: In step S4, the battery transient thermal state evolution surface is constructed by building a dimension-reduced heat dissipation and conduction model based on the standard basis matrix. The dimension-reduced heat dissipation and conduction model includes a system state transition matrix and a thermal load input matrix. The low-dimensional coefficient vector at the current moment is used as the initial value, and the predicted low-dimensional coefficient vector for each predicted time step is recursively calculated in combination with the predicted thermal load time series data. The total current change rate of the battery at each predicted time step is calculated, and a current change rate correction term is introduced to correct the predicted low-dimensional coefficient vector. The physical position index of the temperature measurement node in the battery pack is used as the spatial dimension axis, and the predicted time offset is used as the time dimension axis. Based on the corrected predicted low-dimensional coefficient vector and the core eigenvector, the predicted temperature of each temperature measurement node is reconstructed. The reconstructed predicted temperature is used as the temperature dimension axis to construct a three-dimensional physical thermal state evolution surface.

7. The intelligent predictive control method for thermal management of new energy vehicle batteries according to claim 1, characterized in that: In step S4, the dynamic thermal safety boundary is obtained by calibrating the thermal state change threshold through a battery bench thermal runaway critical experiment; extracting the time series data of the predicted maximum temperature corresponding to the thermal state evolution surface, calculating the second derivative of the predicted maximum temperature with respect to time as the thermal state change rate, and performing moving average filtering on the thermal state change rate. The moment when the rate of change of thermal state after filtering reaches or exceeds the threshold of thermal state abrupt change is identified as the moment of thermal state abrupt change. The predicted highest temperature corresponding to the moment of abrupt change is taken as the critical temperature at that moment. The nominal highest operating temperature of the battery is used as the basic critical temperature for non-abrupt change moments. Safety margin correction and battery aging correction are introduced sequentially for the critical temperature of each predicted time step to generate a dynamic thermal safety boundary that changes with the predicted time.

8. A system for implementing the intelligent predictive control method for thermal management of new energy vehicle batteries according to any one of claims 1 to 7, characterized in that: It includes a data acquisition module, a feature construction module, a thermal load prediction module, a safety boundary generation module, and a temperature control command generation module; Data acquisition module: During vehicle operation, it acquires vehicle speed, vehicle acceleration, and drive motor output torque as the first set of data; and acquires total battery current and state of charge as the second set of data. Obtain the real-time temperature distribution column vector of the temperature measurement nodes inside the battery pack as the third set of data; Feature construction module: Construct a standard basis matrix, orthogonally project the third set of data into the standard basis matrix space for dimensionality reduction, and perform singular value decomposition on the temperature snapshot matrix formed by the third set of data to extract the first k left singular vectors whose cumulative singular value energy ratio reaches a preset threshold as core eigenvectors; The core eigenvectors are used to characterize the main direction of thermal diffusion and the concentration of temperature gradient inside the battery pack, and the nonlinear internal resistance of the battery is corrected based on the concentration of temperature gradient; finally, a multidimensional transient thermodynamic feature package is formed. Thermal load prediction module: Based on the multidimensional transient thermodynamic feature package, it analyzes the expected driving power trajectory of the vehicle in the future prediction time domain; and calculates the dynamic heat generation power trajectory through the heat generation power equation that includes the internal resistance heat effect and the electrochemical entropy change heat effect; calculates the heat propagation delay coefficient according to the rate of change of the dynamic heat generation power trajectory in the continuous time domain, and uses the heat propagation delay coefficient to perform thermal inertia propagation correction on the dynamic heat generation power trajectory to generate predicted thermal load time series data; Safety boundary generation module: Based on the predicted thermal load time series data, core eigenvectors and the rate of change of the total battery current, construct the transient thermal state evolution surface of the battery; calculate the local curvature of the thermal state evolution surface, and generate a dynamic thermal safety boundary including the dynamic maximum temperature threshold according to the temperature value corresponding to the local curvature change region. The temperature control command generation module compares the temperature evolution curve corresponding to the predicted thermal load time series data with the dynamic thermal safety boundary in the time domain to generate an active temperature control command.

9. A device, characterized in that: The device includes a processor and a memory, the memory being electrically connected to the processor, the memory storing a computer-executable program, and the processor executing the computer-executable program to implement the intelligent predictive control method for thermal management of new energy vehicle batteries as described in any one of claims 1 to 7.

10. A storage medium, characterized in that, include: A computer-readable storage medium storing a computer program that, when executed by a processor, implements the intelligent predictive control method for thermal management of new energy vehicle batteries as described in any one of claims 1 to 7.

Images