技术领域Technical field
本发明涉及智能电动汽车动力电池技术领域，尤其是涉及一种基于机器学习的动力电池热管理系统。The present invention relates to the technical field of smart electric vehicle power batteries, and in particular to a power battery thermal management system based on machine learning.
背景技术Background technique
动力电池是智能电动汽车储能单元，其通过内部的化学反应释放电能，从而为电动汽车提供足够动力。动力电池系统一般由电池模组、电池管理系统BMS、热管理系统以及一些电器和机械系统等组成。智能电动汽车的安全性始终受到行业的重视。动力锂离子电池在过充电、针刺、碰撞情况下容易引起热失控而造成冒烟失火甚至爆炸等事故。同时高温将影响动力电池的性能，包括能量密度、使用寿命等参数，因此动力电池的热管理系统是车载电池的核心子系统之一。Power batteries are smart electric vehicle energy storage units that release electrical energy through internal chemical reactions to provide sufficient power for electric vehicles. Power battery systems generally consist of battery modules, battery management systems (BMS), thermal management systems, and some electrical and mechanical systems. The safety of smart electric vehicles has always been taken seriously by the industry. Power lithium-ion batteries can easily cause thermal runaway when overcharged, punctured, or collided, resulting in accidents such as smoke, fire, or even explosion. At the same time, high temperature will affect the performance of power batteries, including energy density, service life and other parameters. Therefore, the thermal management system of power batteries is one of the core subsystems of vehicle batteries.
电动汽车上的动力电池由多个动力电池单体电芯构成，而车辆在不同的行驶状况下，单体电芯由于其自身有一定的内阻，在输出电能的同时会产生一定的热量，使得自身温度变高，动力电池系统在工作过程中产生大量的热聚集在狭小的电池箱体内，热量如果不能够及时地快速散出，当自身温度超出其正常工作温度范围间时会影响电池的性能和寿命，且高温会影响动力电池寿命甚至出现热失控，导致起火爆炸等。The power battery on an electric vehicle is composed of multiple power battery cells. Under different driving conditions of the vehicle, the single cells will generate a certain amount of heat while outputting electric energy due to their own internal resistance. As a result, the power battery system generates a large amount of heat that accumulates in the small battery box during operation. If the heat cannot be quickly dissipated in a timely manner, it will affect the battery's performance when its temperature exceeds its normal operating temperature range. Performance and life, and high temperatures will affect the life of the power battery and even cause thermal runaway, leading to fire and explosion.
动力电池的热管理系统主要通过温度感知和控制装置，加强电池的加热和散热能力，保证电池工作在合适的温度范围和保持电池箱内合理的温度分布。针对动力电池温度控制的特点，目前的温度控制方法主要方式为：The thermal management system of the power battery mainly uses temperature sensing and control devices to enhance the heating and heat dissipation capabilities of the battery, ensuring that the battery operates in a suitable temperature range and maintains a reasonable temperature distribution in the battery box. Based on the characteristics of power battery temperature control, the current main temperature control methods are:
(1)优化动力电池内部结构和材料，正负极等特殊部位采用低电阻材料，减少动力电池工作过程中的温升，但改变电池材质，选用贵重材料零部件时将充分提到电池成本，导致动力电池价格比较昂贵；(1) Optimize the internal structure and materials of the power battery. Use low-resistance materials for special parts such as the positive and negative electrodes to reduce the temperature rise during the operation of the power battery. However, changing the battery material and selecting precious material components will fully consider the battery cost. As a result, power batteries are more expensive;
(2)在电池外部选用导热性比较好的材料，通过原始的风冷散热形式降低电池工作过程中的温升。自然风冷散热形式难以控制扩散的热量，从而无法将动力电池温度精确控制在较小的合适范围。另外自然风冷的形式无法在寒冷低温的环境下使电池温度上升到合适工作范围。(2) Use materials with relatively good thermal conductivity on the outside of the battery to reduce the temperature rise during battery operation through original air-cooling heat dissipation. Natural air cooling is difficult to control the heat dissipation, and thus cannot accurately control the temperature of the power battery within a small appropriate range. In addition, natural air cooling cannot raise the battery temperature to a suitable operating range in cold and low-temperature environments.
(3)在电池内部安装温度传感器，并通过传统PID方法控制载热液的液体温度方法，使电池稳定在合适范围。但传统PID方法存在大量的超参数，需要不断测试每一组参数效果，难以得到最优解。(3) Install a temperature sensor inside the battery and control the liquid temperature of the heat transfer fluid through the traditional PID method to stabilize the battery within an appropriate range. However, the traditional PID method has a large number of hyperparameters, and it is necessary to continuously test the effect of each set of parameters, making it difficult to obtain the optimal solution.
发明内容Contents of the invention
本发明的目的就是为了克服上述现有技术存在的缺陷而提供一种基于机器学习的动力电池热管理系统。The purpose of the present invention is to provide a power battery thermal management system based on machine learning in order to overcome the above-mentioned shortcomings of the prior art.
本发明的目的可以通过以下技术方案来实现：The object of the present invention can be achieved through the following technical solutions:
一种基于机器学习的动力电池热管理系统，包括依次连接且构成回路的电池包、电子水泵、电加热PTC和制冷换热器，电池包、电子水泵、电加热PTC和制冷换热器分别连接控制器，所述电池包内部包括电池模组、动力电池载热管和温度传感器，所述控制器内部集成有用以计算当前时刻所需载热剂的温度和流量的机器学习算法。A power battery thermal management system based on machine learning, including a battery pack, an electronic water pump, an electric heating PTC and a refrigeration heat exchanger that are connected in sequence and form a loop. The battery pack, electronic water pump, electric heating PTC and refrigeration heat exchanger are connected separately. A controller. The battery pack includes a battery module, a power battery heat transfer tube and a temperature sensor. The controller is integrated with a machine learning algorithm for calculating the temperature and flow rate of the heat transfer agent required at the current moment.
所述动力电池载热管包括载热管进口、载热管出口和载热支管，所述载热管进口与所述载热管出口位于一侧，整个流道设有两个分别用以冷却两侧的模组和电器件的一级支路，每条一级支路分设有若干载热支管，各所述载热支管内设有自适应流量调节阀，当动力电池的温度低于阈值下限或高于阈值上限时，所述自适应流量调节阀自动扩张。The power battery heat-carrying tube includes a heat-carrying tube inlet, a heat-carrying tube outlet, and a heat-carrying branch pipe. The heat-carrying tube inlet and the heat-carrying tube outlet are located on one side. The entire flow channel is provided with two modules for cooling both sides. and electrical devices. Each first-level branch is equipped with a number of heat-carrying branch pipes. Each heat-carrying branch pipe is equipped with an adaptive flow regulating valve. When the temperature of the power battery is lower than the lower threshold or higher than the threshold, When the upper limit is reached, the adaptive flow regulating valve automatically expands.
所述温度传感器设于电池包内部的每个电池模组的表面。The temperature sensor is provided on the surface of each battery module inside the battery pack.
进一步地，电池包内部的每个电池模组的上、下表面分别粘贴有五个温度传感器，各温度传感器分别设于每个电池模组的四个顶角及中心位置处，每个电池模组的温度通过温度传感器融合算法估算。Further, five temperature sensors are respectively pasted on the upper and lower surfaces of each battery module inside the battery pack. Each temperature sensor is respectively located at the four top corners and the center of each battery module. Each battery module The temperature of the group is estimated by a temperature sensor fusion algorithm.
每个电池模组的温度通过温度传感器融合算法估算的表达式为：The expression for the temperature of each battery module estimated by the temperature sensor fusion algorithm is:
式中：tf为融合后温度值；tk为第k个温度传感器的测量值，n为温度传感器的总数；ωk为权重；为标准正态分布密度函数；μ为测量温度平均值；σ为测量温度标准差。In the formula: tf is the temperature value after fusion; tk is the measured value of the kth temperature sensor, n is the total number of temperature sensors; ωk is the weight; is the standard normal distribution density function; μ is the average value of the measured temperature; σ is the standard deviation of the measured temperature.
每个电池模组的温度通过在电池包平面构造M×N温度分布格栅图表示，温度分布格栅图的各格栅中的温度通过线性插值方法获取。各所述格栅中，温度值与动力电池的状态构造成固定长度的特征向量，动力电池的状态包括每个电池模组的电压、总电流、电池SOC和工作模式。The temperature of each battery module is represented by constructing an M×N temperature distribution grid diagram on the battery pack plane, and the temperature in each grid of the temperature distribution grid diagram is obtained by linear interpolation method. In each of the grids, the temperature value and the status of the power battery are constructed into a fixed-length feature vector. The status of the power battery includes the voltage, total current, battery SOC and working mode of each battery module.
将各格栅的温度值与动力电池的状态作为输入，通过神经网络输出载热剂在载热管中所需的温度和流量，进而控制制冷换热器、电加热PTC和电子水泵。Taking the temperature value of each grid and the status of the power battery as input, the required temperature and flow rate of the heat transfer agent in the heat transfer tube are output through the neural network, and then the refrigeration heat exchanger, electric heating PTC and electronic water pump are controlled.
所述神经网络包括多层感知器网络和卷积网络，其中所述多层感知器网络有两层全连接层组成，深度提取各格栅中长度为C的高维特征，所述高维特征在M×N格栅中的分布构造成为特征图C×M×N；所述卷积网络包括卷积层和全连接层，所述卷积层将C×M×N的所述特征图进行高维特征提取，并将得到的特征图展开成特征向量后，使用全连接层回归温度和流量两个参数。The neural network includes a multi-layer perceptron network and a convolutional network, wherein the multi-layer perceptron network consists of two fully connected layers, and deeply extracts high-dimensional features with a length of C in each grid. The high-dimensional features The distribution in the M×N grid is constructed into a feature map C×M×N; the convolutional network includes a convolution layer and a fully connected layer, and the convolution layer performs the feature map of C×M×N After extracting high-dimensional features and expanding the obtained feature map into feature vectors, the fully connected layer is used to regress the two parameters of temperature and flow.
所述神经网络在使用前通过训练获取所述神经网络的参数，训练数据通过人工专家实际调节载热剂温度和流量后获取。The parameters of the neural network are obtained through training before use, and the training data is obtained after manual experts actually adjust the temperature and flow rate of the heat transfer agent.
本发明提供的基于机器学习的动力电池热管理系统，相较于现有技术至少包括如下有益效果：Compared with the existing technology, the power battery thermal management system based on machine learning provided by the present invention at least includes the following beneficial effects:
1)本发明无需改变动力电池系统的零部件和核心材料，且可根据实际情况选用低成本动力电池系统，从而降低整个动力电池系统的成本。1) The present invention does not need to change the components and core materials of the power battery system, and can select a low-cost power battery system according to the actual situation, thereby reducing the cost of the entire power battery system.
2)本发明基于机器学习方法自动学习人工专家调参经验，当动力电池热管理系统确定完毕后，只需要人工专家手动调节一段时间，系统自动可以学习专家经验，并完成动力电池的温度控制，方法比较简便实用。2) The present invention automatically learns the parameter adjustment experience of artificial experts based on the machine learning method. After the power battery thermal management system is determined, it only requires manual adjustment by artificial experts for a period of time. The system can automatically learn from the expert experience and complete the temperature control of the power battery. The method is relatively simple and practical.
3)本发明无需人工调整控制参数，系统所有参数均根据专家经验，通过反向传播自动修正机器学习模型参数，从而减少控制人员的负担，提高动力电池热管理系统开发效率，缩短热管理系统开发周期。3) This invention does not require manual adjustment of control parameters. All system parameters are automatically corrected through backpropagation of machine learning model parameters based on expert experience, thereby reducing the burden on controllers, improving the development efficiency of power battery thermal management systems, and shortening the development of thermal management systems. cycle.
附图说明Description of the drawings
图1为实施例中基于机器学习的动力电池热管理系统的结构示意图；Figure 1 is a schematic structural diagram of the power battery thermal management system based on machine learning in the embodiment;
图2为实施例中动力电池内部载热管的结构示意图；Figure 2 is a schematic structural diagram of the heat transfer tube inside the power battery in the embodiment;
图3为实施例中动力电池内部温度格栅图分布图；Figure 3 is a grid distribution diagram of the internal temperature of the power battery in the embodiment;
图4为实施例中基于机器学习的动力电池热管理系统的温度控制方法示意图；Figure 4 is a schematic diagram of the temperature control method of the power battery thermal management system based on machine learning in the embodiment;
图5为实施例中采用的神经网络模型结构示意图；Figure 5 is a schematic structural diagram of the neural network model used in the embodiment;
图6为实施例中神经网络训练集制作示意图。Figure 6 is a schematic diagram of the preparation of the neural network training set in the embodiment.
具体实施方式Detailed ways
下面结合附图和具体实施例对本发明进行详细说明。显然，所描述的实施例是本发明的一部分实施例，而不是全部实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动的前提下所获得的所有其他实施例，都应属于本发明保护的范围。The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts should fall within the scope of protection of the present invention.
实施例Example
本发明涉及一种基于机器学习的动力电池热管理系统，该系统在智能电动汽车上的合适位置处布置安装相应的零部件，并通过载热管连接各零部件形成回路。The invention relates to a power battery thermal management system based on machine learning. The system arranges and installs corresponding components at appropriate positions on a smart electric vehicle, and connects the components through heat-carrying pipes to form a loop.
如图1所示，该系统包括电池包、电子水泵、电加热PTC、制冷换热器和控制器。电池包、电子水泵、电加热PTC、制冷换热器依次连接，且制冷换热器的输出端连接电池包。电池包、电子水泵、电加热PTC、制冷换热器分别与控制器连接。其中电池包内部包括模组、动力电池载热管和温度传感器。控制器内部集成有机器学习算法，能够根据电池温度和电池状态，计算出当前时刻所需载热剂的温度和流量，从而调节电池温度保持在合适范围。机器学习算法采用神经网络，神经网络的训练数据是通过人工专家的实际调节数据制作而成。As shown in Figure 1, the system includes a battery pack, electronic water pump, electric heating PTC, refrigeration heat exchanger and controller. The battery pack, electronic water pump, electric heating PTC, and refrigeration heat exchanger are connected in sequence, and the output end of the refrigeration heat exchanger is connected to the battery pack. The battery pack, electronic water pump, electric heating PTC, and refrigeration heat exchanger are connected to the controller respectively. The interior of the battery pack includes modules, power battery heat pipes and temperature sensors. The controller is integrated with a machine learning algorithm, which can calculate the temperature and flow rate of the heat transfer agent required at the current moment based on the battery temperature and battery status, thereby adjusting the battery temperature to maintain an appropriate range. The machine learning algorithm uses a neural network, and the training data for the neural network is made from actual adjusted data by human experts.
在本发明动力电池热管理系统中，电加热PTC主要用于对电池包进行加热，提高电池包的温度；制冷换热器主要是利用空调冷却回路中的冷却液，将动力电池热管理回路中的热量带走，从而使载热液降温，进一步降低电池包的温度；电子水泵主要使电池包内部的载热管中的载热液流动起来，加快热量传递。In the power battery thermal management system of the present invention, the electric heating PTC is mainly used to heat the battery pack and increase the temperature of the battery pack; the refrigeration heat exchanger mainly uses the coolant in the air-conditioning cooling circuit to heat the power battery thermal management circuit. The heat is taken away, thereby cooling the heat transfer fluid and further reducing the temperature of the battery pack; the electronic water pump mainly causes the heat transfer fluid in the heat transfer tube inside the battery pack to flow to speed up heat transfer.
动力电池载热管包括载热管进口、载热管出口和载热支管。如图2所示，电池包内的载热管具体结构为对称式。进水口与出水口位于一侧，整个流道先分为两个支路(一级支路)，分别冷却两侧的模组和电器件。每条支路再分为若干载热支管(二级支路)。载热支管通过自适应流量调整阀调整每个模组底部流道的流阻，当温度高于上限阈值或者低于下限阈值，节流阀内部通道扩张，增大流量，从而能够快速改变电池局部温度。自适应流量调整阀可分为两段，一段由热膨胀材料组成，一段由冷膨胀材料组成，二者相互拼接，位置的前后关系无限制要求，使支管之间能够根据附近模组的温度自适应调节载热液的流量。载热管的结构设计主要考虑软包模组电芯温度由中间向两边依次递减，中间电芯温度最高，因此载热管设计载热液由中间流入，可以最大限度的减少模组内电芯的温差。The power battery heat-carrying pipe includes a heat-carrying pipe inlet, a heat-carrying pipe outlet and a heat-carrying branch pipe. As shown in Figure 2, the specific structure of the heat transfer tube in the battery pack is symmetrical. The water inlet and outlet are located on one side, and the entire flow channel is first divided into two branches (first-level branches) to cool the modules and electrical devices on both sides respectively. Each branch is further divided into several heat-carrying branch pipes (secondary branches). The heat-carrying branch pipe adjusts the flow resistance of the flow channel at the bottom of each module through an adaptive flow adjustment valve. When the temperature is higher than the upper threshold or lower than the lower threshold, the internal channel of the throttle valve expands to increase the flow rate, thereby quickly changing the local area of the battery. temperature. The adaptive flow adjustment valve can be divided into two sections. One section is composed of thermal expansion material and the other section is composed of cold expansion material. The two are spliced to each other. There are no restrictions on the relationship between the positions, so that the branch pipes can adapt according to the temperature of the nearby modules. Adjust the flow of hot fluid. The structural design of the heat transfer tube mainly considers that the temperature of the battery core of the soft package module decreases from the middle to both sides, and the temperature of the middle battery core is the highest. Therefore, the heat transfer tube is designed to flow in the heat transfer liquid from the middle, which can minimize the temperature difference of the battery cores in the module. .
如图3所示，为了正确探测电池包内的温度，实时控制电池包内的温度保持在固定的范围，各模组电池表面粘贴温度传感器，温度传感器分别位于模组的四个顶角和中心。作为优选方案，温度传感器紧贴于电池包的单个模组表面，即每个模组的上下面分别粘贴有5个温度传感器，分别位于四个顶角和中心。As shown in Figure 3, in order to correctly detect the temperature in the battery pack and control the temperature in the battery pack in real time to maintain a fixed range, temperature sensors are attached to the battery surface of each module. The temperature sensors are located at the four top corners and center of the module. . As a preferred solution, the temperature sensor is closely attached to the surface of a single module of the battery pack, that is, five temperature sensors are pasted on the upper and lower sides of each module, respectively located at the four top corners and the center.
考虑到温度传感器的失效和误差，每个模组的温度通过传感器融合算法估算，公式为：Taking into account the failure and error of the temperature sensor, the temperature of each module is estimated through the sensor fusion algorithm, and the formula is:
式中：tf为融合后温度值；tk为第k个温度传感器的测量值，n为温度传感器的总数；ωk为权重；为标准正态分布密度函数；μ为测量温度平均值；σ为测量温度标准差。In the formula: tf is the temperature value after fusion; tk is the measured value of the kth temperature sensor, n is the total number of temperature sensors; ωk is the weight; is the standard normal distribution density function; μ is the average value of the measured temperature; σ is the standard deviation of the measured temperature.
各模组的温度值通过线性插值方法，生成动力电池温度分布格栅图，即：将电池包的平面构造一个M×N温度分布格栅图，各格栅中的温度通过线性插值方法，由每个模组的温度计算得到。格栅中，温度值与动力电池的状态构造成固定长度的特征向量。动力电池的状态包括每个模组的电压、总电流、电池SOC和工作模式。为扩充每个格栅中的特征信息，通过历史帧信息，采用三次方程进行拟合后，可计算出当前温度变化速度、温度变化加速度，并进一步通过等加速度温度变化模型，预测未来时刻的电池包温度。每个格栅有电池温度和电池状态参数组成，从而构造成为神经网络算法的输入量。The temperature value of each module is used to generate a power battery temperature distribution grid diagram through linear interpolation method, that is, an M×N temperature distribution grid diagram is constructed on the plane of the battery pack, and the temperature in each grid is calculated by linear interpolation method. The temperature of each module is calculated. In the grid, the temperature value and the state of the power battery are constructed into a fixed-length feature vector. The status of the power battery includes the voltage, total current, battery SOC and working mode of each module. In order to expand the characteristic information in each grid, through historical frame information and fitting with cubic equations, the current temperature change speed and temperature change acceleration can be calculated, and further through the equal acceleration temperature change model, the battery in the future can be predicted. package temperature. Each grid consists of battery temperature and battery status parameters, which are constructed as inputs to the neural network algorithm.
插值算法公式为：The interpolation algorithm formula is:
式中：t为当前格栅的温度值；ti和tj为中间变量；ta、tb、tc和td分别为左上角、右上角、左下角和右下角相邻传感器的温度值；(x，y)为当前格栅坐标；(x1，y1)为左上角传感器坐标；(x2，y2)为右下角传感器坐标。In the formula: t is the temperature value of the current grid; ti and tj are intermediate variables; ta , tb , tc and td are the temperatures of adjacent sensors in the upper left corner, upper right corner, lower left corner and lower right corner respectively. value; (x, y) is the current grid coordinate; (x1 , y1 ) is the upper left corner sensor coordinate; (x2 , y2 ) is the lower right corner sensor coordinate.
如图4所示，本发明主要采用神经网络方法完成动力电池温度控制，神经网络输入为电池温度和电池状态参数，输出为载热剂的温度和载热剂的流量。通过构造输入和输出的训练样本，神经网络自动调节网络权重，使网络能够自动完成动力电池的温度控制。神经网络集成于控制器中，根据输入量得到载热剂在载热管中所需的温度和流量，并进一步控制制冷换热器、电加热PTC和电子水泵，从而完成动力电池包的温度控制。As shown in Figure 4, the present invention mainly uses a neural network method to control the temperature of the power battery. The input of the neural network is the battery temperature and battery status parameters, and the output is the temperature of the heat transfer agent and the flow rate of the heat transfer agent. By constructing input and output training samples, the neural network automatically adjusts the network weights so that the network can automatically complete the temperature control of the power battery. The neural network is integrated into the controller to obtain the required temperature and flow rate of the heat transfer agent in the heat transfer tube based on the input amount, and further controls the refrigeration heat exchanger, electric heating PTC and electronic water pump to complete the temperature control of the power battery pack.
神经网络模型的结构示意图如图5所示，神经网络主要包括多层感知器网络和卷积网络，其中多层感知器网络主要有2层全连接层组成，深度提取各格栅中长度为C的高维特征。高维特征在M×N格栅中的分布，可以构造成为特征图C×M×N。卷积网络主要包括卷积层和全连接层，卷积层主要将C×M×N的所述特征图进行高维特征提取，并将得到的特征图展开成特征向量，再使用全连接层回归温度和流量两个参数。The structural diagram of the neural network model is shown in Figure 5. The neural network mainly includes a multi-layer perceptron network and a convolutional network. The multi-layer perceptron network mainly consists of 2 fully connected layers. The length of each grid in the depth extraction grid is C. high-dimensional features. The distribution of high-dimensional features in the M×N grid can be constructed as a feature map C×M×N. The convolutional network mainly includes a convolutional layer and a fully connected layer. The convolutional layer mainly extracts high-dimensional features from the feature map of C×M×N, and expands the obtained feature map into a feature vector, and then uses the fully connected layer. Return the two parameters of temperature and flow.
作为优选方案，卷积网络可采用主流的图像分类网络，卷积部分可采用Resnet50、Resnet101、FPN等网络结构，并使用网络的预训练权重。所述卷积网络的全连接层可采用Kaiming Initialization方法初始化参数，从而有利于整个网络的快速收敛。As a preferred solution, the convolutional network can use mainstream image classification networks. The convolution part can use network structures such as Resnet50, Resnet101, and FPN, and use the pre-training weights of the network. The fully connected layer of the convolutional network can use the Kaiming Initialization method to initialize parameters, which is beneficial to the rapid convergence of the entire network.
基于上述神经网络模型，首先将构造各温度格栅图中的特征向量，特征向量包括温度相关特征(当前温度值、历史前几帧温度值、当前温度变化速度、温度变化加速度、未来几帧温度值)和电池状态特征(每个模组的电压、总电流、电池SOC、工作模式)，更多的特征信息更有利于神经网络学习。网络将输入量映射至高维空间，从而完成对应的温度控制任务。输入特征向量先通过多层感知器网络MLP分别对每个格栅中的特征进行特征提取，提取高维特征以后，构造成为多通道的伪图像，从而可以采用卷积神经网络CNN进一步回归得到温度和流量值。在训练过程中，采用SmoothL1损失函数，公式为：Based on the above neural network model, the feature vectors in each temperature grid diagram will first be constructed. The feature vectors include temperature-related features (current temperature value, temperature values in previous frames in history, current temperature change speed, temperature change acceleration, temperature in future frames value) and battery status characteristics (voltage, total current, battery SOC, working mode of each module), more characteristic information is more conducive to neural network learning. The network maps the input quantity to a high-dimensional space to complete the corresponding temperature control task. The input feature vector is first extracted from the features in each grid through the multi-layer perceptron network MLP. After extracting the high-dimensional features, it is constructed into a multi-channel pseudo image, so that the convolutional neural network (CNN) can be used to further regression to obtain the temperature. and flow value. During the training process, the SmoothL1 loss function is used, and the formula is:
ltotal＝SmoothL1(Tpred-Tgt)+SmoothL1(Qpred-Qgt)ltotal =SmoothL1(Tpred -Tgt )+SmoothL1(Qpred -Qgt )
式中：Tpred为网络输出的载热液温度；Tgt为实际载热液所需温度；Qpred为网络输出的载热液流量；Qgt为实际载热液所需流量。In the formula: Tpred is the temperature of the thermal fluid output by the network; Tgt is the actual required temperature of the thermal fluid; Qpred is the flow rate of the thermal fluid output by the network; Qgt is the actual required flow rate of the thermal fluid.
如图6所示，在制作神经网络训练集数据过程中，需将当前电池实际温度和电池状态参数实际显示并记录，人工专家通过观察动力电池实际情况，根据经验调节载热管中载热剂的流量和温度，载热剂的流量和温度数据也实时保存。从而通过人工专家经验数据制作训练所需数据集。As shown in Figure 6, in the process of making neural network training set data, the current actual battery temperature and battery status parameters need to be actually displayed and recorded. Manual experts observe the actual conditions of the power battery and adjust the temperature of the heat transfer agent in the heat transfer tube based on experience. Flow and temperature, the flow and temperature data of heat transfer agent are also saved in real time. Thus, the data set required for training is produced through artificial expert experience data.
本发明方法结合神经网络自学习的优势，通过简单的方式构造神经网络所需要的训练集，训练集数据采集过程中可以由经验丰富的人工专家完成，只需要人工调节若干组数据，神经网络即可得到相关经验的网络模型，具有简单实用等优点。The method of the present invention combines the advantages of neural network self-learning and constructs the training set required by the neural network in a simple way. The training set data collection process can be completed by experienced artificial experts. It only needs to manually adjust several sets of data, and the neural network is The network model can obtain relevant experience and has the advantages of simplicity and practicality.
以上所述，仅为本发明的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的工作人员在本发明揭露的技术范围内，可轻易想到各种等效的修改或替换，这些修改或替换都应涵盖在本发明的保护范围之内。因此，本发明的保护范围应以权利要求的保护范围为准。The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any worker familiar with the technical field can easily think of various equivalent methods within the technical scope disclosed in the present invention. Modifications or substitutions shall be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111391780.3ACN114156567B (en) | 2021-11-19 | 2021-11-19 | Power battery thermal management system based on machine learning |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111391780.3ACN114156567B (en) | 2021-11-19 | 2021-11-19 | Power battery thermal management system based on machine learning |
| Publication Number | Publication Date |
|---|---|
| CN114156567A CN114156567A (en) | 2022-03-08 |
| CN114156567Btrue CN114156567B (en) | 2024-03-26 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111391780.3AActiveCN114156567B (en) | 2021-11-19 | 2021-11-19 | Power battery thermal management system based on machine learning |
| Country | Link |
|---|---|
| CN (1) | CN114156567B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116027863A (en)* | 2022-12-23 | 2023-04-28 | 武汉三氟新材料科技有限公司 | Immersed liquid cooling heat dissipation regulation and control method and system, electronic equipment and storage medium |
| CN116387692B (en)* | 2023-05-31 | 2023-08-18 | 宁波齐云新材料技术有限公司 | Lithium battery pack intelligent heat dissipation control system and control method thereof |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109193075A (en)* | 2018-09-28 | 2019-01-11 | 合肥工业大学 | Power battery of pure electric automobile method for controlling cooling system based on intensified learning |
| CN110146823A (en)* | 2019-06-06 | 2019-08-20 | 重庆大学 | A SOC Estimation Method for Series Battery Packs Based on Information Fusion |
| CN111079905A (en)* | 2019-12-27 | 2020-04-28 | 北京迈格威科技有限公司 | Convolutional neural network processing method, device and electronic system |
| CN211265669U (en)* | 2019-12-25 | 2020-08-14 | 合肥国轩高科动力能源有限公司 | A fully parallel power battery cooling system |
| CN111572407A (en)* | 2020-04-29 | 2020-08-25 | 辽宁工业大学 | A hybrid vehicle thermal management system and its control method |
| CN113300027A (en)* | 2021-04-19 | 2021-08-24 | 江苏大学 | Power battery thermal management system based on predictive control and control method thereof |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11850970B2 (en)* | 2019-08-18 | 2023-12-26 | Board Of Regents, The University Of Texas System | J-type air-cooled battery thermal management system and method |
| US11709080B2 (en)* | 2019-12-26 | 2023-07-25 | Horiba Stec, Co., Ltd. | Flow rate estimation for mass flow controller |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109193075A (en)* | 2018-09-28 | 2019-01-11 | 合肥工业大学 | Power battery of pure electric automobile method for controlling cooling system based on intensified learning |
| CN110146823A (en)* | 2019-06-06 | 2019-08-20 | 重庆大学 | A SOC Estimation Method for Series Battery Packs Based on Information Fusion |
| CN211265669U (en)* | 2019-12-25 | 2020-08-14 | 合肥国轩高科动力能源有限公司 | A fully parallel power battery cooling system |
| CN111079905A (en)* | 2019-12-27 | 2020-04-28 | 北京迈格威科技有限公司 | Convolutional neural network processing method, device and electronic system |
| CN111572407A (en)* | 2020-04-29 | 2020-08-25 | 辽宁工业大学 | A hybrid vehicle thermal management system and its control method |
| CN113300027A (en)* | 2021-04-19 | 2021-08-24 | 江苏大学 | Power battery thermal management system based on predictive control and control method thereof |
| Title |
|---|
| 刘国良，杨成慧主编；白旭灿，庄淑君，邵麦顿副主编.《MATLAB程序设计基础教程》.西安：西安电子科技大学出版社,2012,235.* |
| 周浩敏，钱政编著.《智能传感技术与系统》.北京：北京航空航天大学出版社,2008,44-50.* |
| 基于实时出行需求和交通路况的电动汽车充电负荷预测;吴钉捷;李晓露;;电力建设;20200801(08);61-71* |
| Publication number | Publication date |
|---|---|
| CN114156567A (en) | 2022-03-08 |
| Publication | Publication Date | Title |
|---|---|---|
| CN114156567B (en) | Power battery thermal management system based on machine learning | |
| CN113300027B (en) | Power battery thermal management system based on predictive control and control method thereof | |
| CN106671785B (en) | Battery management system and method for electric vehicle | |
| CN104577254B (en) | Thermal management system and working method for electric vehicle battery pack | |
| CN102324590B (en) | Temperature control system and method during charging and discharging of lithium-ion power battery pack | |
| CN112864488A (en) | Energy storage liquid cooling system device and control method thereof | |
| CN112092676A (en) | A method for estimating and correcting battery pack temperature field using virtual temperature sensor | |
| CN109962318A (en) | Battery pack coolant flow control system and control method based on dual fuzzy control | |
| CN108417906A (en) | Battery management system with intelligent thermal management system control | |
| CN115732810A (en) | Control method of electric vehicle battery pack heating system | |
| CN119440142B (en) | Energy storage cabinet temperature control method based on cloud platform and temperature control energy storage cabinet | |
| CN116470187A (en) | Battery cell temperature equalizing system and method suitable for immersed liquid cooling energy storage system | |
| CN219575759U (en) | Immersed energy storage cooling system | |
| CN118350260A (en) | A thermal management method for electric vehicle fast charging module based on data-driven model | |
| Ma et al. | Multi-layer NMPC for battery thermal management optimization strategy of connected electric vehicle integrated with waste heat recovery | |
| CN115879370A (en) | Method for predicting and monitoring internal temperature field of lithium ion battery pack | |
| CN111572407B (en) | Hybrid electric vehicle thermal management system and control method thereof | |
| CN116632404A (en) | Immersed energy storage cooling system and temperature control method thereof | |
| CN120121993A (en) | Data-driven lithium-ion battery state prediction method integrating physical information | |
| CN110380152B (en) | A battery thermal management module and its control method | |
| CN106602172A (en) | Constant temperature system of power battery | |
| CN117219917B (en) | A lithium battery thermal balance system device and method based on improved PID intelligent control | |
| CN220984643U (en) | A battery thermal management system for an aircraft | |
| CN205283183U (en) | A battery pack active balancing control system based on battery energy state estimation | |
| CN117525480A (en) | Fuel cell automobile integrated thermal management system and method based on neural network |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |