New energy automobile battery management optimization method and system based on MCU chip

Abstract

The invention relates to the technical field of battery management, and discloses a new energy automobile battery management optimization method and system based on an MCU chip, and the method comprises the steps: carrying out the digital sampling and abnormal value filtering of the voltage, temperature and current of a new energy automobile battery pack through the MCU chip, and obtaining a battery pack multi-source data set; performing infinitesimal body division and temperature field gradient calculation to obtain thermal characteristic parameters of the battery pack; inputting a multi-cohesion residual network model of the MCU chip to carry out convolution feature extraction and state identification to obtain a first battery pack working state parameter; adjusting a weight parameter of the multi-cohesion residual network model through a dual feedback channel to obtain a battery state prediction model; and performing collaborative optimization according to the second battery pack working state parameters output by the battery state prediction model to obtain a battery management control strategy. According to the method, the dynamic adjustment of the battery management strategy is realized, and the adaptability of the system to the working condition change is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery management, and in particular to a new energy vehicle battery management optimization method and system based on an MCU chip. Background Art

[0002] Traditional battery management methods mainly rely on simple threshold judgment and linear control strategies, which are difficult to cope with the dynamic characteristics of battery packs under complex working conditions. As battery capacity increases and charge and discharge rates increase, the uneven temperature distribution and cell voltage differences of battery packs are becoming increasingly prominent, which places higher demands on the control accuracy and response speed of battery management systems.

[0003] The current battery management system faces multiple technical difficulties in practical applications: the battery pack is a complex system coupled with multiple physical fields, and its internal temperature field distribution and electrochemical reaction process are difficult to model accurately; secondly, the precise estimation of battery state parameters (such as SOC, SOH) is affected by many factors, and the robustness of existing estimation methods needs to be improved; traditional single control strategies are difficult to take into account multiple goals such as battery balancing efficiency, temperature uniformity, and charging and discharging safety at the same time. Summary of the invention

[0004] The present invention provides a new energy vehicle battery management optimization method and system based on an MCU chip. The present invention realizes dynamic adjustment of the battery management strategy and effectively improves the system's adaptability to changes in operating conditions.

[0005] In a first aspect, the present invention provides a new energy vehicle battery management optimization method based on an MCU chip, and the new energy vehicle battery management optimization method based on an MCU chip comprises: The voltage, temperature, and current of the new energy vehicle battery pack are digitally sampled and outliers are filtered through the MCU chip to obtain a multi-source data set of the battery pack; Dividing the new energy vehicle battery pack into micro-element bodies to obtain a plurality of micro-element bodies, and calculating heat flux density distribution and temperature field gradient of the plurality of micro-element bodies according to temperature data in the multi-source data set of the battery pack to obtain thermal characteristic parameters of the battery pack; Inputting the battery pack multi-source data set and the battery pack thermal characteristic parameters into the multi-condensed residual network model of the MCU chip to perform convolution feature extraction and state recognition to obtain a first battery pack working state parameter; Performing error calculation on the working state parameter of the first battery pack to obtain an error value, and adjusting the weight parameter of the multi-condensed residual network model according to the error value and the dual feedback channel of the MCU chip to obtain a battery state prediction model; The second battery pack operating state parameter output by the battery state prediction model is input into the multi-objective control module of the MCU chip, and the battery pack voltage balance, temperature field distribution and charging and discharging power are collaboratively optimized to obtain a battery management control strategy.

[0006] In a second aspect, the present invention provides a new energy vehicle battery management optimization system based on an MCU chip, and the new energy vehicle battery management optimization system based on an MCU chip includes: The sampling module is used to digitally sample the voltage, temperature, and current of the battery pack of new energy vehicles and filter out outliers through the MCU chip to obtain a multi-source data set of the battery pack; A calculation module, used to divide the new energy vehicle battery pack into micro-units to obtain a plurality of micro-units, and calculate the heat flux density distribution and temperature field gradient of the plurality of micro-units according to the temperature data in the multi-source data set of the battery pack to obtain thermal characteristic parameters of the battery pack; An identification module, used for inputting the battery pack multi-source data set and the battery pack thermal characteristic parameters into the multi-condensed residual network model of the MCU chip to perform convolution feature extraction and state recognition, so as to obtain a first battery pack working state parameter; An adjustment module, configured to perform error calculation on the working state parameter of the first battery pack to obtain an error value, and adjust a weight parameter of the multi-condensed residual network model according to the error value and a dual feedback channel of the MCU chip to obtain a battery state prediction model; The optimization module is used to input the working state parameters of the second battery pack output by the battery state prediction model into the multi-objective control module of the MCU chip, and coordinately optimize the battery pack voltage balance, temperature field distribution and charge and discharge power to obtain a battery management control strategy.

[0007] The third aspect of the present invention provides a new energy vehicle battery management optimization device based on an MCU chip, comprising: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory so that the new energy vehicle battery management optimization device based on the MCU chip executes the above-mentioned new energy vehicle battery management optimization method based on the MCU chip.

[0008] The fourth aspect of the present invention provides a computer-readable storage medium, which stores instructions. When the computer-readable storage medium is run on a computer, it enables the computer to execute the above-mentioned MCU chip-based new energy vehicle battery management optimization method.

[0009] In the technical solution provided by the present invention, a multi-condensed residual network model is introduced, and a lightweight deep learning architecture is designed in combination with the characteristics of the MCU chip, which effectively reduces the computational complexity and enables complex algorithms to run stably on the resource-constrained MCU platform; a microelement thermodynamic modeling method is adopted, combined with multi-source data acquisition, to achieve an accurate description of the temperature field distribution of the battery pack, providing a reliable theoretical basis for temperature balance control; a dual feedback mechanism is designed, and the accuracy and reliability of state estimation are significantly improved through the combination of prediction error ratio and uncertainty loss function; a multi-objective collaborative optimization framework is constructed, and unified coordination of voltage balance, temperature control and power management is achieved through adaptive weight adjustment; a distributed control architecture based on the CAN bus is developed, and the real-time and reliability of each control instruction are guaranteed through precise timing control and task scheduling; dynamic adjustment of the battery management strategy is achieved through dynamic programming and rolling time domain optimization, which effectively improves the system's adaptability to changes in working conditions.

[0010] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention are realized and obtained by the structures particularly pointed out in the description, claims and drawings.

[0011] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A schematic diagram of an embodiment of a new energy vehicle battery management optimization method based on an MCU chip in an embodiment of the present invention; Figure 2 The figure is a schematic diagram of an embodiment of a new energy vehicle battery management optimization system based on an MCU chip in an embodiment of the present invention. DETAILED DESCRIPTION

[0013] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] The terms "including" and "having" and any variations thereof mentioned in the embodiments of the present invention are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device end including a series of steps or units is not limited to the listed steps or units, but may optionally include other steps or units that are not listed, or may optionally include other steps or units that are inherent to these processes, methods, products or device ends.

[0015] To facilitate understanding of this embodiment, a new energy vehicle battery management optimization method based on an MCU chip disclosed in an embodiment of the present invention is first described in detail. Figure 1 As shown, the method comprises the following steps: 101. Digitally sample the voltage, temperature and current of the battery pack of new energy vehicles and filter outliers through the MCU chip to obtain a multi-source data set of the battery pack; It is understandable that the execution subject of the present invention can be a new energy vehicle battery management optimization system based on an MCU chip, or a terminal or a server, which is not specifically limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0016] Specifically, the positive and negative ends of each battery cell of the new energy vehicle battery pack, the surface temperature sensor and the main circuit current sampling resistor are connected through the data acquisition interface of the MCU chip. Through these connections, the MCU chip realizes real-time acquisition of key physical quantities in the battery pack. In order to ensure the time synchronization of data acquisition, a sampling clock signal is applied to each port to form a periodic sampling task, and a multi-source analog signal is obtained to reflect the actual parameter changes of the battery under different working conditions. The multi-source analog signal is input into the 16-bit precision analog-to-digital converter inside the MCU chip for digital quantity conversion. The 16-bit precision analog-to-digital converter greatly reduces the quantization error of the signal, thereby ensuring that the digitized signal has high accuracy and resolution. Signal conditioning is performed during the analog-to-digital conversion process, including steps such as gain adjustment and offset correction, to adapt to input signals of different amplitudes. After analog-to-digital conversion and signal conditioning, the original digital quantity signal sequence is obtained. The original digital quantity signal sequence is subjected to signal frequency domain analysis and filtering processing. The signal is analyzed in the frequency domain by fast Fourier transform, the main frequency components in the signal are identified, and high-frequency noise and other undesirable frequency components are removed by bandpass filtering or low-pass filtering to obtain a filtered digital signal. The filtered digital signal is screened to remove some abnormal data points introduced by sudden interference to obtain a valid digital signal. The valid digital signal is subjected to data standardization to eliminate the magnitude difference caused by different measurement units and ranges to obtain a standardized data matrix. The standardized data matrix is ​​subjected to time series processing to obtain a data sequence with a timestamp. Time series processing adds a corresponding timestamp to each sampling point, so that the data is not only a static set of values, but can reflect the process of system state changing over time. At the same time, the charging state, discharging state and static state of the data sequence with a timestamp are marked, and the specific working mode is determined according to the direction and amplitude of the collected current signal to form a data sequence with a working state identification. The data sequence with the working state identification is stored in the Flash memory of the MCU chip. A dual index structure combining a time series index table and a data type index table is used. The time series index table is used to manage data in chronological order to ensure that the data type of any time period can be quickly retrieved, while the data type index table is classified and indexed according to the data source and type, so as to efficiently access specific types of data. Through the dual index structure, a multi-source data set of the battery pack is formed.

[0017] 102. Divide the new energy vehicle battery pack into micro-element bodies to obtain multiple micro-element bodies, and calculate the heat flux density distribution and temperature field gradient of the multiple micro-element bodies according to the temperature data in the multi-source data set of the battery pack to obtain the thermal characteristic parameters of the battery pack; Specifically, the battery pack of new energy vehicles is divided into multiple micro-units by dividing the micro-units at equal intervals. The geometric center coordinates and boundary coordinates of each micro-unit are calculated to form a spatial coordinate set of the micro-unit. The temperature data in the multi-source data set of the battery pack is mapped to the spatial coordinate set of these micro-units according to the corresponding relationship of the spatial position. Through spatial mapping, the macroscopic temperature data is refined into the local temperature value of each micro-unit to obtain the initial temperature distribution matrix. This matrix describes the initial temperature state of each micro-unit and reflects the temperature distribution characteristics of the battery pack during operation. Based on the initial temperature distribution matrix, the temperature difference between the micro-units is analyzed, and the temperature data of adjacent micro-units in the initial temperature distribution matrix are interpolated to obtain the temperature difference between the micro-units. According to the temperature difference, the Fourier heat conduction equation is used to calculate the heat flux density of each micro-unit interface. The Fourier heat conduction equation describes the basic law of heat transfer in a conductor. By using this equation, the heat flux density vector field is obtained, thereby quantitatively describing the heat transfer characteristics between each micro-unit. Based on the heat flux density vector field, the temperature gradient components of each microelement in the x, y, and z directions are calculated to obtain the temperature field gradient matrix, which describes the temperature change trend of the entire battery pack in three-dimensional space. The temperature field gradient matrix is ​​reconstructed according to the spatial distribution relationship of the microelement, and the thermal conductivity between each microelement and the six surrounding adjacent microelement is calculated to obtain the thermal conductivity distribution array, which describes the heat transfer capacity between the microelement inside the battery pack. Based on the thermal conductivity distribution array, a three-dimensional thermal network model containing the thermal resistance and thermal capacity parameters of each microelement is established. The three-dimensional thermal network model is a refined simulation of the thermal behavior of the battery pack, in which the thermal resistance and thermal capacity parameters of each microelement jointly determine the thermal characteristics of the entire battery pack. The three-dimensional thermal network model is solved by the node analysis method to obtain the steady-state temperature distribution characteristics of the battery pack. The node analysis method is a mathematical method commonly used in circuit and thermal network analysis. The steady-state temperature of each node is obtained by solving the energy balance equation of each node in the network. The current data in the multi-source data set of the battery pack is processed in sections. According to the different current intervals, the heat source intensity of Joule heat, polarization heat and reaction heat of each microelement in different current intervals is calculated to obtain the heat source distribution function. Joule heat is the heat generated by the current passing through the resistor, polarization heat is caused by electrochemical polarization, and reaction heat is related to the chemical reaction inside the battery. By calculating different types of heat sources, the heat source intensity distribution of each microelement inside the battery pack is obtained, reflecting the thermal behavior of the battery pack under different working conditions. Based on the steady-state temperature distribution characteristics and the heat source distribution function, the equivalent heat capacity and thermal conductivity of the entire battery pack are calculated. These parameters are the concentrated expression of the overall thermal characteristics of the battery pack. The equivalent heat capacity reflects the heat storage capacity of the battery pack, while the thermal conductivity describes the efficiency of heat transfer of the battery pack. By calculating these parameters, the thermal characteristic parameters of the battery pack are obtained.

[0018] 103. Input the battery pack multi-source data set and the battery pack thermal characteristic parameters into the multi-condensed residual network model of the MCU chip to perform convolution feature extraction and state recognition to obtain the first battery pack working state parameter; Specifically, a multimodal time series input matrix is ​​constructed based on the multi-source data set and thermal characteristic parameters of the battery pack, which includes various physical parameters of the battery pack, such as voltage, current, temperature, and parameters such as thermal conductivity and heat flux density calculated by the above thermal characteristics, and is presented in a time series form, so that the model can capture the dynamic changes of these features over time. The constructed multimodal time series input matrix is ​​input into the multi-residual network structure of the multi-condensed residual network model in the MCU chip. The multi-residual network structure contains three parallel residual blocks, each of which consists of two 3×3 convolutional layers, a BatchNorm layer and a ReLU activation function. The number of convolution kernels in these convolutional layers is 32, 64, and 128, respectively, and the step size is set to 1 to ensure that the convolution operation can fully extract local features in the input matrix. The convolution layer operates on the input data by sliding the convolution kernel, thereby capturing the spatial correlation and complex nonlinear relationships in the data. The output of each residual block is reduced in dimension by a 1×1 convolution and then added to the input signal to form a residual connection, thereby effectively alleviating the problem of gradient disappearance in the deep network and enhancing the expression ability of the model. The residual feature allows information to be directly transferred in the network, retaining the useful information of the original input features. The resulting residual feature map contains rich features of the multimodal input data. The residual feature map is input into the agglomerative network. The structure of the agglomerative network contains three agglomerative units, each of which consists of a 4×4 maximum pooling layer and a 1×1 convolutional layer, where the pooling step is 2 and the number of convolution kernels is 64. The role of the pooling layer is to reduce the size of the feature map by taking the maximum value of the local area while retaining the significant information of the feature, thereby reducing the computational complexity of the model. The features are reprocessed by the 1×1 convolutional layer, and the LeakyReLU activation function is used to enhance the sensitivity of the model to negative features and increase the robustness of the model. The use of the Dropout layer (dropout rate is 0.3) helps prevent the model from overfitting and improves generalization ability. The agglomerative network outputs a condensed feature vector. In order to enhance the expression and understanding of the input features, the channel attention weight and spatial attention weight are calculated for the condensed feature vector. The attention mechanism enables the model to pay more attention to important information in different channels or spatial positions, thereby improving the expressiveness of features and the decision quality of the model. When performing weighted operations, voltage, current, temperature, and thermal characteristics are given different weights, which reflect the importance of each feature in the current state. Through weighted fusion, a fused feature matrix is ​​obtained. The fused feature matrix is ​​input into the fully connected layer network for processing. The fully connected layer network contains three hidden layers, each of which contains 512, 256, and 128 neurons, respectively. Each hidden layer uses BatchNorm and ReLU activation functions, and dropout (rate of 0.5) is set between layers to prevent overfitting.These fully connected layers perform nonlinear transformation on the fused features, map high-dimensional features into low-dimensional space, extract deeper features, and obtain deep feature vectors. The deep feature vectors are input into the feature decoupling module to decouple the deep feature vectors into features of state of charge (SOC), state of health (SOH), and state of life. The feature decoupling module extracts these three state features through three parallel fully connected layers, each with an output dimension of 64. The tanh activation function is used for nonlinear mapping to obtain a three-dimensional state feature tensor. The three-dimensional state feature tensors are input into three prediction heads for state prediction. Each prediction head contains two fully connected layers, with the dimension of the middle layer being 32 and the dimension of the output layer being 1, using a linear activation function. The function of these prediction heads is to perform regression prediction on each state feature, and finally obtain the working state parameters of the first battery pack.

[0019] 104. Perform error calculation on the working state parameter of the first battery pack to obtain an error value, and adjust the weight parameter of the multi-condensed residual network model according to the error value and the dual feedback channel of the MCU chip to obtain a battery state prediction model; Specifically, a battery state spatiotemporal correlation diagram is constructed based on the working state parameters of the first battery pack to reveal the mutual relationship and dependency characteristics between the working states of the battery pack. The spatiotemporal dependency feature analysis is performed on the spatiotemporal correlation diagram, including modeling and quantifying the changes of the working state parameters on the time axis and the space axis to obtain a state correlation matrix to describe the correlation and dependency between different state parameters of the battery. Based on each state parameter in the state correlation matrix, their prediction error ratios are calculated respectively to obtain an error ratio sequence. The prediction error ratio is the relative difference between the predicted value and the true value, reflecting the prediction accuracy of the model for different state parameters. By calculating the error ratio, the prediction deviation size of each state parameter is quantified, so as to adjust the model in a targeted manner. The error ratio sequence is input into the first feedback channel of the MCU chip for back-propagation adjustment of the weights of the multi-residual network structure in the multi-condensed residual network model. The back-propagation algorithm adjusts the weights of the network by calculating the gradient so that the model minimizes the error ratio, thereby improving its prediction accuracy for different state parameters. The residual network correction parameters obtained in this process can effectively correct the parameters of the model, so that the residual network can better capture the dynamic characteristics of the battery pack. At the same time, based on the working state parameters of the first battery pack, an uncertainty loss function is constructed to evaluate the reliability of state prediction and the stability of the model. The uncertainty loss function is used to describe the fluctuation of the state prediction value. By analyzing the probability distribution characteristics of the state prediction value, the variance and skewness of the prediction result are calculated. Variance is an important indicator to measure the discreteness of the prediction result, while skewness reflects the symmetry of the prediction distribution. By calculating these uncertainty characteristics, the stability and potential deviation of the model prediction under different states can be identified. The calculated uncertainty feature vector is input into the second feedback channel of the MCU chip for gradient optimization of the weights of the condensed network in the multi-condensed residual network model. In the second feedback channel, the weights of the condensed network are gradient optimized to reduce the uncertainty caused by uneven data distribution or insufficient sensitivity of the model to certain states. Through gradient optimization, the model can better adapt to changes in input data, thereby improving its adaptability to complex battery pack states. The optimized condensed network correction parameters enable the model to more accurately reflect the relationship and importance between the features in the feature fusion stage. The residual network correction parameters and the condensed network correction parameters are applied to the multi-condensed residual network model for joint update to achieve comprehensive adjustment of the entire model. The joint update process ensures that each part of the model can work together to achieve the best overall performance and obtain a battery state prediction model.

[0020] 105. Input the working state parameters of the second battery pack output by the battery state prediction model into the multi-objective control module of the MCU chip, coordinately optimize the battery pack voltage balance, temperature field distribution and charge and discharge power, and obtain the battery management control strategy.

[0021] Specifically, the working state parameters of the second battery pack output by the battery state prediction model are obtained, including the real-time voltage, temperature, charge and discharge power and health state of the battery pack. These parameters are input into the multi-objective control module of the MCU chip, which can synchronously analyze and coordinately control multiple physical parameters of the battery. In the multi-objective control module, the voltage balance of the battery pack is optimized. The energy transfer path is planned through this module, and the energy transfer efficiency of the balance loop is calculated according to the topological structure of the DC-DC converter in the active balance circuit and the conduction sequence of the switch tube, so as to determine the voltage difference between the battery cells, and obtain the voltage balance control scheme accordingly. This scheme makes the overall voltage distribution of the battery pack more uniform through the transfer of energy between different battery cells, and avoids the life and performance of the battery pack being affected by overcharging or over-discharging of some battery cells. The optimization of the topology of the DC-DC converter and the control sequence of the switch tube can maximize the energy transfer efficiency and realize the accurate allocation and effective utilization of energy. The temperature field distribution characteristic data of the battery pack is input into the temperature field optimization controller for processing. The temperature field optimization controller comprehensively analyzes the temperature field in the battery pack by establishing a thermodynamic model. Based on these data, the power required by the cooling system in different areas is determined by calculating the distribution coefficient of the cooling power. According to the temperature distribution of the microelement of the battery pack, the temperature field optimization controller can identify local overheating areas, focus on these areas, and determine the target of temperature field regulation. Based on the temperature field regulation target, a heat dissipation power distribution function is constructed. Through this function, combined with the liquid cooling channel layout of the new energy vehicle cooling system and the specific distribution of the radiator position, the power regulation coefficient of each cooling unit is calculated to obtain the temperature balance control instruction. The heat dissipation power distribution function takes into account the physical structural characteristics of the cooling system to ensure that each cooling unit can reasonably allocate cooling power according to specific temperature requirements to achieve the global temperature balance control target. While optimizing the voltage and temperature, the health status data in the working state parameters of the second battery pack are analyzed, focusing on the remaining capacity (SOC) and internal resistance changes of the battery. Based on these data, the limit curve of the charge and discharge power is calculated to determine a reasonable power constraint range to ensure that the charge and discharge power is reasonably controlled under the healthy state of the battery to avoid irreversible damage to the battery caused by excessive charge and discharge. By limiting the power range, the duration of the battery's healthy state is effectively extended, and the overall stability and safety of the battery pack are improved. The power constraint interval is input into the dynamic power planner, and the rolling horizon optimization method is used to calculate the reference sequence of power control. The rolling horizon optimization method is a dynamic programming algorithm that predicts and optimizes a period of time in the future at each moment to calculate the optimal power control reference sequence. The voltage balance control scheme, temperature balance control instructions and power control reference sequence are multi-objective weighted calculations to balance voltage, temperature and power control, and finally the optimal battery management control strategy is obtained.

[0022] By analyzing the voltage balancing control scheme in the battery management control strategy, it is converted into a specific switch control timing table. Based on the voltage difference of different battery cells, the timing table groups the battery pack into three levels of high, medium and low voltage, and sets the corresponding PWM wave duty cycle according to the balancing priority of these groups. By setting the PWM wave duty cycle, the energy transfer rate is effectively controlled to ensure the balanced distribution of energy between different battery cells, and the balancing loop control signal is obtained. The balancing loop control signal is allocated in time. According to the topology of the DC-DC converter, the switch on delay is set to ensure the coordinated work between different switch tubes to prevent circuit short circuit or efficiency reduction caused by asynchronous signal transmission. In this process, the dead time compensation parameter is generated. The dead time is to prevent the short circuit caused by the simultaneous conduction of two switches during the switching process of the switch tube. Combined with the delay and dead time compensation, the driving timing of the balancing loop is generated to ensure that the switch tube in the DC-DC converter can work stably and efficiently, and realize the voltage balancing of the battery pack. At the same time, the temperature balancing control instruction is parsed according to the distribution of the temperature of the battery pack microelement and converted into a heat dissipation power allocation matrix. The matrix analyzes the temperature state of each microelement in the battery pack and sets local cooling power for each microelement accordingly. Through detailed heat dissipation power distribution, more cooling power is applied to the area with higher local temperature to minimize the problem of local overheating and obtain the cooling power control sequence. The cooling power control sequence is spatially mapped, and the coolant flow distribution ratio of each channel is calculated according to the parallel structure of the liquid cooling channel to ensure that the cooling power can be effectively transmitted to the area that needs cooling. Based on the calculation results of the flow distribution, the control instructions of the cooling system are obtained, so as to achieve precise control of the entire cooling system and ensure that the temperature of each microelement is kept within a reasonable range. In order to control the power during the charging and discharging process, the power control reference sequence is used as the basis to calculate the limit of the charging and discharging current. According to the rated capacity and rate characteristic curve of the battery pack, the upper and lower limits of the current are set to obtain the current threshold parameters. The current threshold parameters ensure that the battery pack will not be damaged or excessively heated due to excessive current during the charging and discharging process, thereby ensuring the safety and stability of the battery pack. The current threshold parameters are input into the power controller, and the PID control algorithm is used to adjust the charging and discharging process in real time, and the duty cycle of the PWM wave is calculated to obtain the current control signal. The PID control algorithm enables the system to respond quickly to changes in the external environment by adjusting the error in real time, while maintaining the current within a safe range to achieve the best charging and discharging efficiency. The control bus is encapsulated for the equalization loop drive timing, cooling system control instructions, and current control signals, so that each control signal is managed and scheduled uniformly, allowing the entire system to operate in a coordinated manner. During the encapsulation process, the real-time control signal of the battery pack is constructed in the format of the CAN (Controller Area Network) protocol.As a multi-master serial communication protocol, CAN protocol has high real-time performance and reliability, and is suitable for the scenario of multi-component coordinated control in new energy vehicles. The constructed real-time control signal of the battery pack is sent to each execution unit through the communication interface of the MCU chip. The execution unit includes the switch tube of the active balancing circuit, the pump body and valve of the cooling system, and the charge and discharge management module.

[0023] In the embodiment of the present invention, a multi-condensed residual network model is introduced and a lightweight deep learning architecture is designed in combination with the characteristics of the MCU chip, which effectively reduces the computational complexity and enables complex algorithms to run stably on the resource-constrained MCU platform; a microelement thermodynamic modeling method is adopted, combined with multi-source data acquisition, to achieve an accurate description of the temperature field distribution of the battery pack, providing a reliable theoretical basis for temperature balance control; a dual feedback mechanism is designed, and the accuracy and reliability of state estimation are significantly improved through the combination of prediction error ratio and uncertainty loss function; a multi-objective collaborative optimization framework is constructed, and unified coordination of voltage balance, temperature control and power management is achieved through adaptive weight adjustment; a distributed control architecture based on the CAN bus is developed, and the real-time and reliability of each control instruction are guaranteed through precise timing control and task scheduling; dynamic adjustment of the battery management strategy is achieved through dynamic programming and rolling time domain optimization, which effectively improves the system's adaptability to changes in working conditions.

[0024] In a specific embodiment, the process of executing step 101 may specifically include the following steps: The positive and negative ends of each battery cell of the new energy vehicle battery pack, the surface temperature sensor and the main circuit current sampling resistor are connected to the ports through the data acquisition interface of the MCU chip, and a sampling clock signal is applied to the port to obtain a multi-source analog signal; Input the multi-source analog signals into the 16-bit precision analog-to-digital converter of the MCU chip for digital quantity conversion and signal conditioning to obtain the original digital quantity signal sequence, and perform signal frequency domain analysis and filtering on the original digital quantity signal sequence to obtain the filtered digital signal; Performing data screening on the filtered digital signal to obtain a valid digital signal, and performing data standardization on the valid digital signal to obtain a standardized data matrix; Perform time series processing on the standardized data matrix to obtain a data sequence with a time stamp, and mark the charging state, discharging state, and static state of the data sequence with a time stamp. At the same time, determine the working mode according to the current direction and amplitude to obtain a data sequence with a working state mark; The data sequence with working status identification is stored in the Flash memory of the MCU chip, and a timing index table and a data type index table are established. A multi-source data set of the battery pack is formed through a dual index structure.

[0025] Specifically, the data acquisition interface of the MCU chip is used to connect the positive and negative ends of each battery cell of the new energy vehicle battery pack, the surface temperature sensor, and the main circuit current sampling resistor. The voltage measurement of the positive and negative ends is to monitor the voltage changes of the battery cells. These voltage values ​​are helpful to evaluate the charging and discharging status of the battery. The surface temperature sensor is used to measure the temperature changes of the battery to monitor the thermal characteristics of the battery, and the current sampling resistor is used to measure the current of the main circuit. In order to ensure that these measurements are performed synchronously, a unified sampling clock signal is applied to each port. The sampling clock signal provides a time reference to ensure that different physical quantities are collected at the same time point to obtain multi-source analog signals. The multi-source analog signals are input into the 16-bit precision analog-to-digital converter (ADC) in the MCU chip for digital quantity conversion. The ADC converts the analog signal into a digital signal, and through the 16-bit precision, the quantization error of the signal is extremely small, which can ensure that the obtained digital signal has a high resolution. In the process of analog-to-digital conversion, signal conditioning operations such as gain adjustment and filtering are performed to ensure the stability and accuracy of the signal. The obtained original digital quantity signal sequence is a high-precision discrete representation of voltage, temperature and current, containing data information at multiple moments. The original digital signal sequence is subjected to frequency domain analysis and filtering. Frequency domain analysis converts the signal from the time domain to the frequency domain through fast Fourier transform to identify the main frequency components. If the original signal sequence contains a lot of high-frequency noise, the noise frequency is determined through frequency domain analysis, and then the high-frequency components are removed by a low-pass filter. The filtered digital signal is obtained by convolution calculation: ; in, is the filtered signal, is the original digital signal, is the impulse response of the filter, is the length of the filter. Through the convolution operation of the filter, the unwanted high-frequency noise is effectively removed, and a smoother and more reliable filtered digital signal is obtained. The filtered digital signal is screened to remove the existing abnormal points. The data screening adopts the sliding window method, that is, the mean and standard deviation are calculated within a window of a given length, and the signal points that exceed a certain range are removed. For example, if the window length is set to , then The data filtering conditions at the moment are expressed as: ; in, and are the mean and standard deviation of the data in the sliding window, respectively. Through this step, valid digital signals are screened out to ensure the reliability and accuracy of the data. The screened valid digital signals are subjected to data standardization to eliminate the influence caused by different dimensions. Using the mean normalization method, each signal is scaled according to the mean and standard deviation, so that its mean is 0 and its standard deviation is 1: ; Standardization can make data from different sources have the same magnitude, which is convenient for subsequent feature fusion and model training. The standardized data matrix is ​​processed in time series to form a data sequence with a timestamp. Each timestamp corresponds to a specific sampling moment. In this way, the signals at different time points are recorded sequentially to reflect the law of battery pack status changes over time. For each data sequence with a timestamp, the state is marked, including the charging state, discharging state and static state. The current working mode of the battery is determined by judging the direction and amplitude of the current. For example, when the current is positive and exceeds a certain threshold When the current is negative, it is in the charging state; when the current is close to zero and fluctuates within a certain small range, it is in the static state. The data sequence with the working status identification is stored in the Flash memory of the MCU chip. In order to make the stored data convenient and efficient to retrieve and manage, a timing index table and a data type index table are established. The timing index table is used to manage the data according to the time sequence, so that the data information within any time period can be quickly retrieved; and the data type index table is classified and indexed according to the physical type of the data (such as voltage, temperature, current), so that specific types of data can be quickly accessed when needed. A multi-source data set for the battery pack is formed through a dual index structure.

[0026] In a specific embodiment, the process of executing step 102 may specifically include the following steps: Divide the new energy vehicle battery pack into micro-element bodies with equal spacing to obtain multiple micro-element bodies, and calculate the geometric center coordinates and boundary coordinates of each micro-element body to obtain a micro-element space coordinate set; The temperature data in the multi-source data set of the battery pack is mapped to the spatial coordinate set of the micro-element according to the spatial position correspondence, and the temperature value of each micro-element is calculated to obtain the initial temperature distribution matrix; Perform difference calculation on the temperature data of adjacent micro-element bodies in the initial temperature distribution matrix, and calculate the heat flux density of each micro-element interface according to the Fourier heat conduction equation to obtain the heat flux density vector field; According to the heat flux density vector field, the temperature gradient components of each microelement in the three directions of x, y, and z are calculated to obtain the temperature field gradient matrix, and the temperature field gradient matrix is ​​reconstructed according to the spatial distribution relationship of the microelement, and the thermal conductivity between each microelement and the six adjacent microelement bodies is calculated to obtain the thermal conductivity distribution array; Based on the thermal conductivity distribution array, a three-dimensional thermal network model including the thermal resistance and heat capacity parameters of each microelement is established, and the node analysis method is used to solve the steady-state temperature distribution characteristics. The current data in the multi-source data set of the battery pack are processed segmentally, and the heat source intensity of Joule heat, polarization heat and reaction heat of each microelement in different current intervals is calculated to obtain the heat source distribution function. Based on the steady-state temperature distribution characteristics and the heat source distribution function, the equivalent heat capacity and thermal conductivity of the battery pack are calculated to obtain the thermal characteristic parameters of the battery pack.

[0027] Specifically, the entire battery pack is geometrically modeled to simplify the complex battery system into multiple small, easy-to-analyze units to better describe its heat transfer characteristics. The battery pack is divided into equal intervals. microelement, the size of each microelement is , through division, the geometric center coordinates and boundary coordinates of each microelement are obtained, thus forming the spatial coordinate set of the microelement. The geometric center coordinates are expressed as ,in: ; in Respectively represent the microelement in The boundary coordinates are the coordinates of the boundary points of the microelement in three-dimensional space. The temperature data in the multi-source data set of the battery pack is mapped to the spatial coordinate set of the microelement according to the correspondence between the spatial positions, so as to refine the temperature distribution of the entire battery pack to each microelement. Assuming that the temperature data of the battery pack is collected by multiple temperature sensors, and the position of each sensor is known, the temperature data of the sensor is mapped to the geometric center of the microelement according to the nearest neighbor or bilinear interpolation method to obtain the temperature value of each microelement, and finally form the initial temperature distribution matrix. Represents microelement The temperature difference is obtained by performing a difference operation on the temperature data of adjacent micro-elements in the initial temperature distribution matrix. , and calculate the heat flux density of each microelement interface according to the Fourier heat conduction equation. The basic form of the Fourier heat conduction equation is: ; in, is the heat flux density, is the thermal conductivity, is the temperature gradient. The temperature gradient is calculated by difference operation, for example, The temperature gradient in the direction is expressed as: ; Similarly, calculation in and Based on these gradient values ​​and combined with Fourier's law, the heat flux density at each microelement interface is calculated to form a heat flux density vector field. Based on the heat flux vector field, calculate the The temperature gradient components in three directions, the temperature field gradient matrix is ​​obtained In order to analyze the heat transfer characteristics, the temperature field gradient matrix is ​​reconstructed according to the spatial distribution relationship of the microelement, and the thermal conductivity between each microelement and the six adjacent microelement is calculated. The calculation of thermal conductivity takes into account the thermal properties of the material and the spatial distance between adjacent microelements, which is usually expressed as: ; By calculating the thermal conductivity of all microelements, we can get the thermal conductivity distribution array. , which describes the heat transfer capacity between each microelement in the entire battery pack. Based on the thermal conductivity distribution array, a three-dimensional thermal network model containing the thermal resistance and thermal capacity parameters of each microelement is established. In this model, each microelement is regarded as a node, and these nodes are connected by thermal conductivity. For each microelement, its thermal resistance By back-calculating the thermal conductivity, we can get: ; in is the area of ​​the microelement interface. It represents the heat storage capacity of each microelement, expressed as: ; in, is the material density, is the specific heat capacity, is the volume of the microelement. Through these thermal resistance and heat capacity parameters, a complete three-dimensional thermal network model is established, and then the node analysis method is used to solve the model to obtain the steady-state temperature distribution characteristics. The core of the node analysis method is to list the energy conservation equations for each node, and by jointly solving these equations, the steady-state temperature distribution of the system is obtained. The current data in the multi-source data set of the battery pack is processed in segments to analyze the heat source characteristics of each microelement in different current ranges. The heat generated by the battery during the charging and discharging process includes Joule heat, polarization heat and reaction heat. Joule heat It is caused by the current passing through the resistor, and its calculation formula is: ; in is the current, is the internal resistance of the battery. It is related to the electrochemical reaction inside the battery, and its size is related to the degree of electrochemical polarization of the battery. The heat released or absorbed by the chemical reaction itself is obtained by combining the heat source intensity with the current data segmentation to obtain the heat source distribution function. . Calculate the equivalent heat capacity of the battery pack based on the steady-state temperature distribution characteristics and heat source distribution function and thermal conductivity The equivalent heat capacity describes the heat storage capacity of the entire battery pack and is calculated as the weighted average of the heat capacities of all micro-elements: ; The thermal conductivity coefficient describes the thermal conductivity of the battery pack as a whole, and is obtained by averaging the thermal conductivity of all microelements: ; Through this method, the heat capacity and thermal conductivity that describe the thermal characteristics of the battery pack are finally obtained. For example, during the high-speed charging process of the battery pack, the changes in the equivalent heat capacity and thermal conductivity reflect the thermal response characteristics of the battery, helping the management system decide whether to adjust the cooling strategy to ensure that the battery pack does not experience thermal runaway.

[0028] In a specific embodiment, the process of executing step 103 may specifically include the following steps: Constructing a multi-modal time series input matrix based on multi-source data sets of battery packs and thermal characteristic parameters of battery packs; The multi-modal time series input matrix is ​​input into the multi-residual network structure of the multi-condensed residual network model in the MCU chip. The multi-residual network structure contains three parallel residual blocks. Each residual block consists of two 3×3 convolutional layers, a BatchNorm layer and a ReLU activation function. The number of convolution kernels is 32, 64, and 128 respectively, and the step size is 1. The output of each residual block is reduced in dimension by 1×1 convolution and then added to the input signal to obtain a residual feature map. The residual feature map is input into the condensed network. The condensed network contains three condensed units. Each condensed unit consists of a 4×4 maximum pooling layer and a 1×1 convolution layer. The pooling step size is 2, the number of convolution kernels is 64, and the LeakyReLU activation function and Dropout layer are used. The dropout rate is 0.3 to obtain the condensed feature vector. Calculate the channel attention weight and spatial attention weight of the condensed feature vector, and weight the voltage, current, temperature and thermal characteristic features to obtain a fusion feature matrix; The fused feature matrix is ​​input into the fully connected layer network. The fully connected layer network contains three hidden layers with 512, 256, and 128 neurons respectively. Each hidden layer uses BatchNorm and ReLU activation functions, and the inter-layer dropout rate is set to 0.5 to obtain the deep feature vector. The deep feature vector is input into the feature decoupling module. The feature decoupling module extracts the state of charge, state of health and life state features through three parallel fully connected layers. The output dimension of each fully connected layer is 64. The tanh activation function is used to obtain a three-dimensional state feature tensor. The three-dimensional state feature tensor is input into three prediction heads respectively. Each prediction head contains two fully connected layers, the middle layer dimension is 32, and the output layer dimension is 1. A linear activation function is used to obtain the working state parameters of the first battery pack.

[0029] Specifically, multiple physical quantities such as voltage, temperature, current, and thermal characteristic parameters are obtained from the multi-source data set of the battery pack, and a multi-modal time series input matrix is ​​constructed based on these data. Assume that the multi-source data set contains time steps, each time step contains multiple sensor data. Suppose the total number of voltage, temperature, current, thermal characteristics and other data is , then the multimodal time series input matrix is ​​defined as This matrix contains multi-modal feature information and can fully describe the state changes of the battery pack on the time axis. Input the multi-modal time series into the matrix The input to the multi-residual network structure of the multi-condensed residual network model in the MCU chip is composed of three parallel residual blocks. The input of each residual block passes through two The number of convolution kernels in the convolution layer is 32, 64 and 128 respectively, the step size is set to 1, and the convolution operation is described as: ; in, represents the eigenvalue of the convolution output, is the convolution kernel weight, is the input eigenvalue, is the bias term. The convolution operation is performed by sliding the convolution kernel on the input matrix to capture the local spatial features of the input data. The convolution is followed by the Batch Normalization layer, which is used to normalize each batch of data so that the output of each layer remains within the same scale range, which helps to speed up the convergence of the model. Then, the ReLU (rectified linear unit) activation function is used, and the calculation process is: ; Through nonlinear transformation, the nonlinear characteristics of the model are introduced to enhance the model's expressiveness. The output of each residual block is The convolution is then performed to reduce the dimension and then added to the original input signal to obtain the residual feature map. The residual connection operation effectively alleviates the gradient vanishing problem in the deep network and ensures that information can be efficiently transmitted in the network. The residual feature map is input into the condensed network. The condensed network consists of three condensed units, each of which contains a The maximum pooling layer and a The step size of the maximum pooling layer is set to 2. The role of pooling is to reduce the size of the feature map by taking the maximum value of the local area while retaining the most significant features. The formula is: ; in, is the range of the pooling window. Through the pooling operation, the computational complexity is significantly reduced and the generalization ability of the model is improved. The convolution layer processes the features, the number of convolution kernels is set to 64, and the LeakyReLU activation function is used, which is defined as: ; in It is usually set to a constant less than 1 (for example, 0.01). The advantage of LeakyReLU is that it avoids the problem of ReLU being completely inactive in the negative area and can retain a certain amount of negative information. In order to avoid overfitting of the model, a Dropout layer is added, and the dropout rate is set to 0.3, which means that 30% of the neuron outputs are randomly set to zero during training to improve the robustness and generalization ability of the model. After being processed by the condensed network, the output is a condensed feature vector. In order to enhance the expression of features, the condensed feature vector is weighted for channel attention and spatial attention. Channel attention is used to measure the importance of different feature channels and is defined as: ; in, is the channel weight matrix, and is the weight matrix of the fully connected layer, Represents the sigmoid activation function. Spatial attention is used to measure the importance of different spatial positions, and its calculation is: ; in, For one Convolution operation. Combining the weights of channel attention and spatial attention, the final fusion feature matrix is ​​obtained, which effectively fuses multimodal data such as voltage, current, temperature and thermal characteristics. The fusion feature matrix is ​​input into the fully connected layer network for processing. The fully connected layer network contains three hidden layers, and the number of neurons in each hidden layer is 512, 256 and 128 respectively. Each layer is normalized by Batch Normalization and then nonlinearly transformed by ReLU activation function. Dropout is added between layers, and the dropout rate is set to 0.5 to reduce overfitting of the model. Through the calculation of the fully connected layer, the deep feature vector is obtained: ; in, is the weight matrix of the fully connected layer, is the bias term, is the fusion feature matrix. The deep feature vector is input into the feature decoupling module, which extracts the state of charge (SOC), state of health (SOH) and state of life (SOL) features through three parallel fully connected layers. The output dimension of each fully connected layer is 64, and the tanh activation function is used: ; The output range of the tanh activation function is [-1,1], which helps to limit the eigenvalues ​​to a smaller range, thereby stabilizing the model training process. Through this step, a three-dimensional state feature tensor is obtained, which describes the battery's state of charge, state of health, and state of life respectively. The three-dimensional state feature tensor is input into three prediction heads for state prediction. Each prediction head contains two fully connected layers, where the middle layer dimension is 32 and the output layer dimension is 1. The middle layer is activated by ReLU, and the output layer uses a linear activation function to obtain the final prediction value. The specific calculation process is: ; in, is the weight of the output layer, is the feature after tanh activation, In this way, the working state parameters of the first battery pack are obtained, including the state of charge, the state of health and the state of life.

[0030] In a specific embodiment, the process of executing step 104 may specifically include the following steps: Constructing a battery state spatiotemporal correlation diagram based on the working state parameters of the first battery pack, and performing spatiotemporal dependency feature analysis on the battery state spatiotemporal correlation diagram to obtain a state correlation matrix; Calculate the prediction error ratio for each state parameter in the state correlation matrix respectively to obtain an error ratio value sequence; The error ratio sequence is input into the first feedback channel of the MCU chip, and the weights of the multi-residual network structure in the multi-condensed residual network model are adjusted by back-propagation to obtain the residual network correction parameters; An uncertainty loss function is constructed based on the working state parameter of the first battery pack, and the variance and skewness of the prediction result are calculated based on the probability distribution characteristics of the state prediction value to obtain an uncertainty feature vector; The uncertainty feature vector is input into the second feedback channel of the MCU chip, and the weight of the condensed network in the multi-condensed residual network model is gradient optimized to obtain the condensed network correction parameters; The residual network correction parameters and the condensed network correction parameters are applied to the multi-condensed residual network model for joint update to obtain the battery state prediction model.

[0031] Specifically, a battery state spatiotemporal correlation diagram is constructed based on the first battery pack working state parameter. The battery pack working state parameters include the values ​​of the battery voltage, temperature, internal resistance, etc. at different time points. The spatiotemporal correlation diagram is constructed by these state parameters to quantify the battery timing changes and the spatial dependence between each battery cell. Assuming that the battery pack is composed of multiple single cells, the battery state parameters in a certain time period are set as the matrix ,in is the length of the time series, is the number of battery cells. In the spatiotemporal correlation graph, each node represents the state of a battery cell at a specific time point, and the edge represents the dependency in time or space. By calculating the correlation coefficient between each state, the weight between nodes is established: ; in, Representation Node and nodes The correlation between the battery status and the time-space dependency is obtained, thereby obtaining a graph structure that describes the time-space dependency of the battery pack status. The time-space dependency feature analysis is performed on the battery status time-space dependency graph, and the graph convolution network is used to extract features from the correlation graph to obtain the state correlation matrix. ,in Representation Node and nodes The degree of dependence between them. Error analysis is performed based on the state parameters in the state correlation matrix. In order to quantify the error of the model in predicting the battery pack state, the prediction error ratio is calculated for each state parameter in the state correlation matrix. For the The true value of a state, is the corresponding predicted value, then the prediction error ratio It is expressed as: ; Through this step, we get the error ratio sequence , where each element represents the relative error of the corresponding state parameter. Error ratio sequence Reflects the prediction accuracy of the model on different state parameters, and inputs the error ratio into the first feedback channel of the MCU chip to perform back-propagation adjustment on the weights of the multi-residual network structure in the multi-condensed residual network model. The back-propagation process is to adjust the weights of the network through the gradient descent method to minimize the prediction error. Assume that the loss function is the mean square error: ; By the loss function About weight Find the partial derivative, get the gradient, and then update the weights according to the gradient descent method: ; in, is the learning rate, which determines the step size of each weight update. Through back propagation and gradient adjustment, the correction parameters of the residual network are obtained, so that the model can better fit the input data and reduce the prediction error. The uncertainty of the model is optimized. On this basis, an uncertainty loss function is constructed based on the working state parameters of the first battery pack to quantify the uncertainty in the model prediction results. Uncertainty mainly reflects the confidence of the model when making predictions, and is quantified by calculating the probability distribution characteristics of the predicted values, such as variance and skewness. Assume that the probability distribution of the prediction result is , then the variance It is expressed as: ; in, is the expected value of the forecast, and the skewness It indicates the degree of asymmetry of the distribution: ; By calculating the variance and skewness of the prediction results, we can get the uncertainty eigenvector , and the feature vector is input into the second feedback channel of the MCU chip for gradient optimization of the condensed network weights in the multi-condensed residual network model. The goal of uncertainty optimization is to reduce the variance in the prediction results and increase the stability and robustness of the model to each input. Similar to error feedback, gradient optimization also updates the weights of the model based on the partial derivative of the loss function to reduce uncertainty loss. The residual network correction parameters and the condensed network correction parameters are applied to the multi-condensed residual network model and jointly updated to obtain the battery state prediction model. By combining the feedback information of the two parts, the weight distribution of the entire network is optimized globally to ensure that the model can reduce both errors and prediction uncertainties when predicting the battery pack state.

[0032] In this embodiment, the error ratio sequence is input into the first feedback channel of the MCU chip, and the weights of the multi-residual network structure in the multi-condensed residual network model are back-propagated and adjusted to obtain the residual network correction parameters, including: mapping the error ratio sequence to the nodes of each layer of the multi-condensed residual network model according to the hierarchical relationship from input to output, and obtaining the network structure parameter matrix according to the input weight and activation state of each node; performing gradient calculation on the network structure parameter matrix on the forward propagation path, and calculating the contribution of the convolution layer, BatchNorm layer and ReLU layer in each residual block to the final error, and obtaining the gradient vector table; the gradient value in the gradient vector table is transmitted back according to the topological relationship of the residual connection, and the gradient at each jump connection is summed to obtain the residual path gradient matrix; the update of the weights of each layer is calculated based on the residual path gradient matrix The Adam optimizer is used to set the initial learning rate to 0.001, the momentum parameters β1 to 0.9, and β2 to 0.999 to obtain the weight adjustment vector; the weight adjustment vector is amplitude-limited, the maximum adjustment step is set to 0.1, and the weight change is limited to a reasonable range through the clip function to obtain the correction constraint vector; the correction constraint vector is weightedly combined with the original weight, and the exponential decay strategy is used to dynamically adjust the weight update ratio, and the decay coefficient is set to 0.95 to obtain the updated weight parameters; the updated weight parameters are regularized, and the L2 norm constraint is used to prevent the weight from being too large. The regularization coefficient is set to 0.01 to obtain the normalized weight matrix; the normalized weight matrix is ​​applied to the multi-residual network structure of the multi-condensed residual network model, and the connection weights and bias parameters of each layer are updated to obtain the residual network correction parameters.

[0033] In this embodiment, the uncertainty feature vector is input into the second feedback channel of the MCU chip, and the weight of the condensed network in the multi-condensed residual network model is gradient optimized to obtain the condensed network correction parameters, including: feature decomposition of the uncertainty feature vector, extracting the main uncertainty components through principal component analysis, and reducing the dimension of the feature according to the contribution rate to obtain the core feature matrix; mapping the core feature matrix to each condensed unit of the condensed network, establishing a feature propagation path according to the topological structure of the pooling layer and the convolution layer, and obtaining a feature transfer graph; performing sensitivity analysis on the pooling operation of each condensed unit in the feature transfer graph, calculating the amount of information loss in the feature compression process, establishing a loss function based on information entropy, and obtaining the pooling layer correction target; substituting the pooling layer correction target into the gradient descent algorithm, setting the initial learning rate The learning rate is set to 0.005, and the cosine annealing strategy is used to dynamically adjust the learning rate. The minimum learning rate is set to 0.0001 to obtain the pooling parameter adjustment amount. The gradient of the weight of the 1×1 convolutional layer in the condensed network is calculated based on the pooling parameter adjustment amount, and the partial derivative of the weight with respect to the loss function is derived using the chain rule to obtain the convolutional layer gradient matrix. The convolutional layer gradient matrix is ​​input into the adaptive moment estimation optimizer, and the decay rate β1 is set to 0.9, β2 is set to 0.999, and ε is set to 1e-8. The gradient is biased and the weight update vector is obtained. The weight update vector is gradient clipped, the gradient norm threshold is set to 5.0, and the gradient explosion is limited by normalization to obtain the normalized weight increment. The normalized weight increment is applied to the parameters of each layer of the condensed network, the weight configuration of the pooling layer and the convolution layer is updated, and the condensed network correction parameters are obtained.

[0034] In a specific embodiment, the process of executing step 105 may specifically include the following steps: Obtaining a second battery pack operating state parameter output by the battery state prediction model, and inputting the second battery pack operating state parameter output by the battery state prediction model into a multi-objective control module of the MCU chip; The energy transfer path is planned through the multi-objective control module. According to the DC-DC converter topology and switch conduction sequence of the active balancing circuit, the energy transfer efficiency of the balancing loop is calculated to obtain the voltage balancing control scheme. The temperature field distribution characteristic data is input into the temperature field optimization controller, the cooling power distribution coefficient is calculated through the thermodynamic model, and the local overheating area is determined according to the temperature distribution of the microelement of the new energy vehicle battery pack to obtain the temperature field adjustment target; Based on the temperature field regulation target, a heat dissipation power allocation function is constructed. Combined with the liquid cooling channel layout and radiator position distribution of the cooling system, the power regulation coefficient of each cooling unit is calculated to obtain the temperature balance control instruction. Analyze the health status data in the working status parameters of the second battery pack, calculate the charge and discharge power limit curve based on the remaining capacity and internal resistance change of the battery, and obtain the power constraint range; The power constraint interval is input into the dynamic power planner, and the power control reference sequence is calculated using the rolling horizon optimization method. Multi-objective weight calculation is performed on the voltage balance control scheme, temperature balance control instructions and power control reference sequence to obtain the battery management control strategy.

[0035] Specifically, the operating state parameters of the second battery pack are obtained from the battery state prediction model, including the voltage, temperature, state of charge (SOC), state of health (SOH), etc. of the battery pack. Through the MCU chip, the operating state parameters are input into the multi-objective control module, and multiple objectives are optimized and controlled at the same time, such as voltage balancing, temperature management, and charge and discharge power regulation, to ensure efficient and safe operation of the battery. The energy transfer path is planned through the multi-objective control module, and the energy transfer efficiency of the balancing loop is calculated based on the DC-DC converter topology of the active balancing circuit and the conduction sequence of the switch tube. The goal of active balancing is to redistribute the power between battery cells through the DC-DC converter so that the voltages of all battery cells tend to be consistent. Assume and are the energy transmission input and output power respectively, which are expressed by the efficiency formula: ; in, is the energy transfer efficiency, and Calculated by the voltage and current measurements of the DC-DC converter: ; According to the efficiency of energy transfer and the topology of the DC-DC converter, combined with the conduction sequence of the switch tube, a voltage balancing control scheme is formulated. By controlling the conduction timing of each switch tube, energy is efficiently transferred from the high-voltage monomer to the low-voltage monomer to achieve voltage balance. In terms of temperature management, the temperature field distribution characteristic data is input into the temperature field optimization controller. Through the thermodynamic model, the cooling power distribution coefficient of each microelement is calculated. Assume that the temperature of each microelement is , according to the temperature data, determine the local overheating area. The thermodynamic model is based on Fourier's law of heat conduction, heat flux density It is expressed as: ; in, is the thermal conductivity, is the temperature gradient. By calculating the heat flux density, the cooling requirements of each area are determined. Combined with the temperature values ​​of each microelement, determine which areas are in an overheated state and form the temperature field regulation target. Based on the temperature field regulation target, a heat dissipation power allocation function is constructed. Combined with the liquid cooling channel layout and radiator position distribution of the new energy vehicle cooling system, the power regulation coefficient of each cooling unit is calculated. Assume that the cooling system contains multiple parallel liquid cooling channels, and the cooling power of each channel is , which is allocated by the following function: ; in, is the distribution coefficient of the cooling channel, satisfying , is the total cooling power of the system. Through this step, the temperature balance control instruction is obtained to ensure that appropriate cooling power is applied to each overheating area, thereby achieving temperature field balance. For charge and discharge power control, based on the health status data in the working status parameters of the second battery pack, the remaining capacity and internal resistance change of the battery are analyzed, and then the charge and discharge power limit curve is calculated. Assume that the remaining capacity of the battery is (indicates the current state of charge), the internal resistance is , then the charge and discharge power limit curve is expressed as: ; in, is the battery voltage, is the maximum discharge current of the battery. Through this formula, the power constraint interval is obtained to prevent damage to the battery caused by excessive charging and discharging, and ensure that the battery works within a safe range. The power constraint interval is input into the dynamic power planner, and the power control reference sequence is calculated using the rolling time domain optimization method. The core of the rolling time domain optimization is to predict and optimize a period of time in the future at a certain time step to achieve dynamic adjustment. The power control reference sequence is obtained by minimizing the following objective function: ; in, is the target power, is the actual power, represents the power change rate, is a smoothing factor used to control the severity of power changes. By minimizing the objective function, the optimal power control reference sequence is obtained. The voltage balance control scheme, temperature balance control instructions and power control reference sequence are multi-objective weighted to obtain the final battery management control strategy. The weights of the three objectives are , the battery management control strategy is expressed as: ; in, For voltage balance control, For temperature balance control, is the power control reference sequence. By weighted combination of various objectives, a control strategy that comprehensively considers voltage, temperature and power is obtained.

[0036] In a specific embodiment, the new energy vehicle battery management optimization method based on the MCU chip also includes the following steps: The voltage balancing control scheme in the battery management control strategy is parsed into a switch control timing table, and the duty cycle of the PWM wave is set according to the balancing priority of the three voltage groups of high, medium and low to obtain the balancing loop control signal; Perform timing distribution on the control signal of the balancing loop, set the switch on delay according to the DC-DC converter topology, generate dead time compensation parameters, and obtain the driving timing of the balancing loop; The temperature balance control instruction is parsed into a heat dissipation power allocation matrix according to the temperature distribution of the microelement, and the local cooling power is set corresponding to the microelement structure of the battery pack to obtain a cooling power control sequence; The cooling power control sequence is spatially mapped, and the flow distribution ratio of each channel is calculated according to the parallel structure of the liquid cooling channel to obtain the cooling system control instructions; The charge and discharge current limit is calculated based on the power control reference sequence, and the current upper and lower limits are set in combination with the rated capacity of the battery pack and the rate characteristic curve to obtain the current threshold parameter. The current threshold parameter is input into the power controller, and the duty cycle of the PWM wave is calculated through the PID control algorithm to obtain the current control signal; The control bus is encapsulated for the balancing loop drive timing, cooling system control instructions and current control signals, and the battery pack real-time control signal is constructed according to the CAN protocol format. The battery pack real-time control signal is sent to each execution unit through the communication interface of the MCU chip.

[0037] Specifically, by analyzing the voltage balancing control scheme in the battery management control strategy, a specific switch control timing table is generated. This control scheme groups the battery cells into three levels of high, medium and low voltage according to the voltage difference of each single cell in the battery pack to determine the balancing priority of each group. For battery cells with higher voltage, the active balancing circuit transfers the excess power to the battery cells with lower voltage to achieve voltage balancing. In this process, the duty cycle of the pulse width modulation (PWM) wave is set to control the energy transfer rate, and the PWM duty cycle is set to , and , corresponding to high, medium and low voltage groups respectively, usually satisfying: ; The size of the duty cycle determines the speed of energy transfer. For high voltage groups, a higher duty cycle is required to quickly transfer excess energy. After analyzing the voltage balancing control scheme, the control signal of the balancing loop is timed to ensure that the switch tubes in the DC-DC converter are correctly turned on according to a specific timing. The topology of the DC-DC converter determines the control logic of each switch tube. By setting the turn-on delay of the switch tube, the dead time compensation parameter is generated. Dead time means that when adjacent switch tubes are switched, it must be ensured that both will not be turned on at the same time to avoid the risk of short circuit. Set the turn-on delay to and dead time is , then it must satisfy: ; Through timing arrangement, the driving timing signal of the balancing loop is obtained to ensure that each switch tube is turned on and off at the appropriate time point to achieve safe and efficient energy transfer. The temperature balancing control instruction is parsed into a heat dissipation power allocation matrix according to the temperature distribution of the microelement to achieve refined temperature control. The temperature of each microelement in the new energy vehicle battery pack is measured by a temperature sensor and used to identify overheating areas. Assume that the temperature of each microelement is In order to avoid local overheating, different cooling powers are allocated to each microelement to form a heat dissipation power allocation matrix The cooling power is proportional to the temperature. Let the proportionality coefficient of cooling power and temperature be , then: ; in, is the ambient temperature. According to the temperature of each microelement, the local cooling power is determined, and the cooling power control sequence is obtained to guide the cooling system to dissipate heat for each microelement. When spatially mapping the cooling power control sequence, the parallel structure of the liquid cooling channel of the cooling system needs to be considered. Assume that the cooling system contains There are parallel liquid cooling channels, and the cooling liquid flow rate of each channel is , the total flow is , calculate the flow distribution ratio according to the cooling power distribution ratio required by each channel ,satisfy: ; In this way, the flow distribution ratio of each cooling channel is obtained, thereby forming the control instructions of the cooling system to ensure that the liquid cooling system applies appropriate cooling force in different areas. For charge and discharge control, the charge and discharge current limit is calculated based on the power control reference sequence. The safe range of charge and discharge depends on the remaining capacity, internal resistance and rate characteristic curve of the battery. Assume that the rated capacity of the battery is , the rate characteristic is Crate, then the maximum charge and discharge current limit is expressed as: ; At the same time, combined with the internal resistance of the battery In order to avoid excessive current causing internal resistance heating, the upper and lower limits of the current are set to obtain the current threshold parameters and The current threshold parameter is input into the power controller, and the corresponding PWM wave duty cycle is calculated through the PID (proportional-integral-differential) control algorithm to accurately control the charge and discharge current. The PID control formula is: ; in, is the control signal (i.e. PWM duty cycle), is the error between the expected value and the actual value, , and They are proportional, integral and differential coefficients respectively. Through PID control, the duty cycle of PWM is dynamically adjusted to ensure that the current during charging and discharging is within a safe range. After completing the calculation of the above control signals, the balancing loop drive timing, cooling system control instructions and current control signals are encapsulated in the control bus. In order to ensure that each execution unit can work in a coordinated manner, the CAN protocol is used to encapsulate the control signal. The CAN protocol is a multi-master communication protocol with high real-time and reliability, and is suitable for communication between various execution units in new energy vehicles. Construct a data frame, each data frame contains an identifier, a control field, a data field and a check field. Assume that the data field contains the balancing loop drive timing , cooling system control instructions and current control signal , expressed as follows: ; Among them, ID is the identifier of the data frame, which is used to distinguish different types of control signals, and CRC is a cyclic redundancy check, which is used to ensure the integrity of data transmission. The constructed real-time control signal of the battery pack is sent to each execution unit through the communication interface of the MCU chip. The execution unit includes an active balancing circuit, a pump body and valve of the liquid cooling system, and a charge and discharge management module. These execution units receive the control signal sent by the MCU and operate according to the specified timing and control instructions, thereby realizing the balancing, temperature regulation, and charge and discharge control of the battery pack.

[0038] The above describes the new energy vehicle battery management optimization method based on the MCU chip in the embodiment of the present invention. The following describes the new energy vehicle battery management optimization system based on the MCU chip in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, a new energy vehicle battery management optimization system based on an MCU chip includes: The sampling module 201 is used to digitally sample the voltage, temperature and current of the battery pack of the new energy vehicle and filter out abnormal values ​​through the MCU chip to obtain a multi-source data set of the battery pack; The calculation module 202 is used to divide the new energy vehicle battery pack into micro-element bodies to obtain multiple micro-element bodies, and calculate the heat flux density distribution and temperature field gradient of the multiple micro-element bodies according to the temperature data in the multi-source data set of the battery pack to obtain the thermal characteristic parameters of the battery pack; The identification module 203 is used to input the battery pack multi-source data set and the battery pack thermal characteristic parameters into the multi-condensed residual network model of the MCU chip to perform convolution feature extraction and state recognition to obtain the first battery pack working state parameter; The adjustment module 204 is used to perform error calculation on the working state parameter of the first battery pack to obtain an error value, and adjust the weight parameter of the multi-condensed residual network model according to the error value and the dual feedback channel of the MCU chip to obtain a battery state prediction model; The optimization module 205 is used to input the second battery pack working state parameter output by the battery state prediction model into the multi-objective control module of the MCU chip, and coordinately optimize the battery pack voltage balance, temperature field distribution and charge and discharge power to obtain a battery management control strategy.

[0039] Through the collaborative cooperation of the above-mentioned components, by introducing the multi-condensed residual network model and combining the characteristics of the MCU chip, a lightweight deep learning architecture was designed, which effectively reduced the computational complexity and enabled complex algorithms to run stably on the resource-constrained MCU platform; the microelement thermodynamic modeling method was adopted, combined with multi-source data acquisition, to achieve an accurate description of the temperature field distribution of the battery pack, providing a reliable theoretical basis for temperature balance control; a dual feedback mechanism was designed, which significantly improved the accuracy and reliability of state estimation through the combination of prediction error ratio and uncertainty loss function; a multi-objective collaborative optimization framework was constructed, and the unified coordination of voltage balance, temperature control and power management was achieved through adaptive weight adjustment; a distributed control architecture based on the CAN bus was developed, and the real-time and reliability of each control instruction was guaranteed through precise timing control and task scheduling; dynamic adjustment of the battery management strategy was achieved through dynamic programming and rolling time domain optimization, which effectively improved the system's adaptability to changes in working conditions.

[0040] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0041] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program code.

[0042] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A new energy vehicle battery management optimization method based on MCU chip, characterized in that: The method comprises: The voltage, temperature, and current of the new energy vehicle battery pack are digitally sampled and outliers are filtered through the MCU chip to obtain a multi-source data set of the battery pack; Dividing the new energy vehicle battery pack into micro-element bodies to obtain a plurality of micro-element bodies, and calculating heat flux density distribution and temperature field gradient of the plurality of micro-element bodies according to temperature data in the multi-source data set of the battery pack to obtain thermal characteristic parameters of the battery pack; Inputting the battery pack multi-source data set and the battery pack thermal characteristic parameters into the multi-condensed residual network model of the MCU chip to perform convolution feature extraction and state recognition to obtain a first battery pack working state parameter; Performing error calculation on the working state parameter of the first battery pack to obtain an error value, and adjusting the weight parameter of the multi-condensed residual network model according to the error value and the dual feedback channel of the MCU chip to obtain a battery state prediction model; The second battery pack operating state parameter output by the battery state prediction model is input into the multi-objective control module of the MCU chip, and the battery pack voltage balance, temperature field distribution and charging and discharging power are collaboratively optimized to obtain a battery management control strategy.

2. The method for optimizing battery management of new energy vehicles based on MCU chip according to claim 1, characterized in that: The voltage, temperature and current of the new energy vehicle battery pack are digitally sampled and outliers are filtered through the MCU chip to obtain a multi-source data set of the battery pack, including: Connect the positive and negative ends of each battery cell, the surface temperature sensor and the main circuit current sampling resistor of the new energy vehicle battery pack through the data acquisition interface of the MCU chip, and apply a sampling clock signal to the port to obtain a multi-source analog signal; Input the multi-source analog signals into the 16-bit precision analog-to-digital converter of the MCU chip for digital quantity conversion and signal conditioning to obtain an original digital quantity signal sequence, and perform signal frequency domain analysis and filtering processing on the original digital quantity signal sequence to obtain a filtered digital signal; Performing data screening on the filtered digital signal to obtain a valid digital signal, and performing data standardization on the valid digital signal to obtain a standardized data matrix; Performing time series processing on the standardized data matrix to obtain a data sequence with a time stamp, and marking the data sequence with a time stamp with a charging state, a discharging state, and a static state, and at the same time, judging the working mode according to the current direction and amplitude to obtain a data sequence with a working state mark; The data sequence with the working status identification is stored in the Flash memory of the MCU chip, a timing index table and a data type index table are established, and a battery pack multi-source data set is formed through a dual index structure.

3. The new energy vehicle battery management optimization method based on MCU chip according to claim 2 is characterized in that: The method of dividing the new energy vehicle battery pack into micro-element bodies to obtain a plurality of micro-element bodies, and calculating the heat flux density distribution and the temperature field gradient of the plurality of micro-element bodies according to the temperature data in the multi-source data set of the battery pack to obtain the thermal characteristic parameters of the battery pack includes: Dividing the new energy vehicle battery pack into micro-element bodies with equal spacing to obtain a plurality of micro-element bodies, and calculating the geometric center coordinates and boundary coordinates of each micro-element body to obtain a micro-element body space coordinate set; Mapping the temperature data in the multi-source data set of the battery pack to the spatial coordinate set of the micro-element according to the spatial position correspondence, and calculating the temperature value of each micro-element to obtain an initial temperature distribution matrix; Performing difference calculation on the temperature data of adjacent micro-element bodies in the initial temperature distribution matrix, and calculating the heat flux density of each micro-element interface according to the Fourier heat conduction equation to obtain a heat flux density vector field; According to the heat flux density vector field, the temperature gradient components of each microelement in the three directions of x, y and z are calculated to obtain a temperature field gradient matrix, and the temperature field gradient matrix is ​​reconstructed according to the spatial distribution relationship of the microelement, and the thermal conductivity between each microelement and the surrounding six adjacent microelement bodies is calculated to obtain a thermal conductivity distribution array; Based on the thermal conductivity distribution array, a three-dimensional thermal network model including thermal resistance and thermal capacity parameters of each microelement is established and a node analysis method is used to solve and obtain steady-state temperature distribution characteristics; The current data in the multi-source data set of the battery pack is processed in segments, the heat source intensity of Joule heat, polarization heat and reaction heat of each microelement in different current intervals is calculated, and the heat source distribution function is obtained. Based on the steady-state temperature distribution characteristics and the heat source distribution function, the equivalent heat capacity and thermal conductivity of the battery pack are calculated to obtain the thermal characteristic parameters of the battery pack.

4. The method for optimizing battery management of new energy vehicles based on MCU chip according to claim 3 is characterized in that: The step of inputting the battery pack multi-source data set and the battery pack thermal characteristic parameters into the multi-condensed residual network model of the MCU chip for convolution feature extraction and state recognition to obtain the first battery pack working state parameter includes: Constructing a multi-modal time series input matrix based on the battery pack multi-source data set and the battery pack thermal characteristic parameters; The multimodal time series input matrix is ​​input into the multi-residual network structure of the multi-condensed residual network model in the MCU chip, wherein the multi-residual network structure comprises three parallel residual blocks, each residual block is composed of two 3×3 convolutional layers, a BatchNorm layer and a ReLU activation function, the number of convolution kernels is 32, 64, and 128 respectively, and the step size is 1. The output of each residual block is reduced in dimension by 1×1 convolution and then added to the input signal to obtain a residual feature map; The residual feature map is input into a condensed network, which includes three condensed units, each of which is composed of a 4×4 maximum pooling layer and a 1×1 convolution layer, with a pooling step of 2 and a number of convolution kernels of 64. A LeakyReLU activation function and a Dropout layer are used, and the dropout rate is 0.3, to obtain a condensed feature vector; Calculating the channel attention weight and the spatial attention weight of the condensed feature vector, and weighting the voltage, current, temperature and thermal characteristic features to obtain a fusion feature matrix; The fused feature matrix is ​​input into a fully connected layer network, wherein the fully connected layer network comprises three hidden layers, and the number of neurons is 512, 256, and 128 respectively. Each hidden layer uses BatchNorm and ReLU activation functions, and the inter-layer dropout rate is set to 0.5, to obtain a deep feature vector; The deep feature vector is input into a feature decoupling module, and the feature decoupling module extracts the state of charge, state of health and life state features respectively through three parallel fully connected layers, and the output dimension of each fully connected layer is 64. A tanh activation function is used to obtain a three-dimensional state feature tensor; The three-dimensional state feature tensor is input into three prediction heads respectively, each prediction head contains two fully connected layers, the middle layer dimension is 32, the output layer dimension is 1, and a linear activation function is used to obtain the working state parameters of the first battery pack.

5. The method for optimizing battery management of new energy vehicles based on MCU chip according to claim 4 is characterized in that: The error calculation is performed on the working state parameter of the first battery pack to obtain an error value, and the weight parameter of the multi-condensed residual network model is adjusted according to the error value and the dual feedback channel of the MCU chip to obtain a battery state prediction model, including: constructing a battery state spatiotemporal correlation diagram based on the first battery pack working state parameter, and performing spatiotemporal dependency feature analysis on the battery state spatiotemporal correlation diagram to obtain a state correlation matrix; Calculating the prediction error ratio for each state parameter in the state association matrix respectively to obtain an error ratio value sequence; Inputting the error ratio sequence into the first feedback channel of the MCU chip, performing back-propagation adjustment on the weights of the multi-residual network structure in the multi-condensed residual network model, and obtaining residual network correction parameters; Constructing an uncertainty loss function based on the first battery pack working state parameter, calculating the variance and skewness of the prediction result based on the probability distribution characteristics of the state prediction value, and obtaining an uncertainty feature vector; Inputting the uncertainty feature vector into the second feedback channel of the MCU chip, performing gradient optimization on the weights of the condensed network in the multi-condensed residual network model, and obtaining condensed network correction parameters; The residual network correction parameters and the condensed network correction parameters are applied to the multi-condensed residual network model for joint updating to obtain a battery state prediction model.

6. The new energy vehicle battery management optimization method based on MCU chip according to claim 5 is characterized in that: The second battery pack working state parameter output by the battery state prediction model is input into the multi-objective control module of the MCU chip, and the battery pack voltage balance, temperature field distribution and charge and discharge power are collaboratively optimized to obtain a battery management control strategy, including: Acquire a second battery pack operating state parameter output by the battery state prediction model, and input the second battery pack operating state parameter output by the battery state prediction model into a multi-objective control module of the MCU chip; The energy transfer path is planned by the multi-objective control module, and the energy transfer efficiency of the balancing loop is calculated according to the DC-DC converter topology structure and the switch tube conduction sequence of the active balancing circuit to obtain a voltage balancing control scheme; Inputting the temperature field distribution characteristic data into the temperature field optimization controller, calculating the cooling power distribution coefficient through the thermodynamic model, and determining the local overheating area according to the microelement temperature distribution of the new energy vehicle battery pack to obtain the temperature field adjustment target; Based on the temperature field regulation target, a heat dissipation power allocation function is constructed, and the power regulation coefficient of each cooling unit is calculated in combination with the liquid cooling channel layout and radiator position distribution of the cooling system to obtain a temperature balance control instruction; Analyze the health status data in the working status parameters of the second battery pack, calculate the charge and discharge power limit curve based on the remaining capacity and internal resistance change of the battery, and obtain the power constraint range; The power constraint interval is input into a dynamic power planner, a power control reference sequence is calculated using a rolling time domain optimization method, and multi-objective weight calculation is performed on the voltage balance control scheme, the temperature balance control instruction and the power control reference sequence to obtain a battery management control strategy.

7. The method for optimizing battery management of new energy vehicles based on MCU chip according to claim 6 is characterized in that: The MCU chip-based new energy vehicle battery management optimization method also includes: The voltage balancing control scheme in the battery management control strategy is parsed into a switch control timing table, and the duty cycle of the PWM wave is set according to the balancing priority of the three voltage groups of high, medium and low to obtain a balancing loop control signal; Performing timing distribution on the control signal of the balancing loop, setting the switch on delay according to the DC-DC converter topology, generating a dead time compensation parameter, and obtaining the driving timing of the balancing loop; The temperature balance control instruction is parsed into a heat dissipation power allocation matrix according to the temperature distribution of the microelement, and the local cooling power is set corresponding to the microelement structure of the battery pack to obtain a cooling power control sequence; Performing spatial mapping on the cooling power control sequence, calculating the flow distribution ratio of each channel according to the parallel structure of the liquid cooling channel, and obtaining a cooling system control instruction; Calculate the charge and discharge current limit based on the power control reference sequence, set the current upper and lower limits in combination with the rated capacity of the battery pack and the rate characteristic curve, obtain the current threshold parameter, input the current threshold parameter into the power controller, calculate the duty cycle of the PWM wave through the PID control algorithm, and obtain the current control signal; The balancing loop drive timing, the cooling system control instructions and the current control signal are encapsulated by the control bus, a battery pack real-time control signal is constructed according to the CAN protocol format, and the battery pack real-time control signal is sent to each execution unit through the communication interface of the MCU chip.

8. A new energy vehicle battery management optimization system based on MCU chip, characterized in that: Used to execute the new energy vehicle battery management optimization method based on MCU chip according to any one of claims 1 to 7, the system comprises: The sampling module is used to digitally sample the voltage, temperature, and current of the battery pack of a new energy vehicle and filter out abnormal values ​​through the MCU chip to obtain a multi-source data set of the battery pack; A calculation module, used to divide the new energy vehicle battery pack into micro-units to obtain a plurality of micro-units, and calculate the heat flux density distribution and temperature field gradient of the plurality of micro-units according to the temperature data in the multi-source data set of the battery pack to obtain thermal characteristic parameters of the battery pack; An identification module, used for inputting the battery pack multi-source data set and the battery pack thermal characteristic parameters into the multi-condensed residual network model of the MCU chip to perform convolution feature extraction and state recognition, so as to obtain a first battery pack working state parameter; An adjustment module, configured to perform error calculation on the working state parameter of the first battery pack to obtain an error value, and adjust a weight parameter of the multi-condensed residual network model according to the error value and a dual feedback channel of the MCU chip to obtain a battery state prediction model; The optimization module is used to input the working state parameters of the second battery pack output by the battery state prediction model into the multi-objective control module of the MCU chip, and coordinately optimize the battery pack voltage balance, temperature field distribution and charge and discharge power to obtain a battery management control strategy.

Images