Battery management system

Abstract

The invention relates to the technical field of battery equalization control, in particular to a battery management system, which comprises a state determination unit used for acquiring the charge state of each battery in a battery pack according to an ampere-hour integral method, and determining the average charge state of the battery pack according to the charge state of each battery in the battery pack; and the difference calculation unit is used for determining the charge state difference of the batteries in the same battery pack according to the charge states of the batteries, and determining the average charge state difference of the battery pack in the battery system according to the average charge state of the battery pack. According to the invention, the energy consumption of each battery in the charging and discharging process tends to be consistent through the functions of intra-pack equalization and inter-pack equalization, the energy of the battery pack can be more uniformly distributed by eliminating the charge state difference between the batteries, the efficiency loss caused by excessive charging and discharging of individual batteries is avoided, and the battery efficiency is improved. Therefore, the energy utilization efficiency of the battery pack is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of battery equalization control, and particularly to a battery management system. Background Art

[0002] Traditional systems cannot detect the state of charge difference of batteries in a timely manner, and thus cannot effectively perform equalization correction. Due to the inaccurate state of charge of the batteries, it may lead to over-discharge or over-charge of the batteries, affecting the battery life and performance. Moreover, traditional systems do not clearly distinguish between intra-group equalization and inter-group equalization functions, and may only rely on simple equalization mechanisms, such as voltage-based equalization, and cannot effectively solve the state of charge difference between batteries, thus unable to achieve uniform energy distribution between batteries, which may cause some batteries to be over-charged or over-discharged, resulting in reduced efficiency, shortening the service life of the battery pack, and reducing the energy utilization efficiency. In addition, traditional systems usually do not integrate an intelligent temperature control mechanism. Temperature control may only rely on a single temperature sensor, or there is no temperature regulation strategy for the change of the equalization state of the battery pack. When the battery pack is unbalanced, some batteries may generate heat due to over-charge or over-discharge, but traditional systems may not be able to identify these changes in a timely manner and cannot accurately adjust the temperature, resulting in overheating or over-cooling of the batteries, increasing the risk of battery damage and affecting the safety and performance of the system. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to overcome the above-mentioned disadvantages of the prior art and provide a battery management system.

[0004] The technical solution adopted to solve the above technical problem is: A battery management system, comprising: A state determination unit, which is used to obtain the state of charge of each battery in the battery pack according to the ampere-hour integration method, and determine the average state of charge of the battery pack according to the state of charge of each battery in the battery pack; A difference calculation unit, which is used to determine the state of charge difference of the batteries in the same battery pack according to the state of charge of the batteries, and determine the average state of charge difference of the battery packs in the battery system according to the average state of charge of the battery pack; An equalization judgment unit, which is used to judge whether the battery system and the battery pack are in an unbalanced state according to the state of charge difference of the batteries in the same battery pack and the average state of charge difference of the battery packs in the battery system; A state equalization unit, which is used to perform intra-group equalization on the battery pack when the battery pack is in an unbalanced state, and perform inter-group equalization on the battery system when the battery system is in an unbalanced state; A temperature control unit, which is used to control the temperature of the battery pack according to a pre-trained temperature control model when the battery pack performs internal balancing.

[0005] Preferably, the calculation formula for the average state of charge of the battery pack is as follows: ; Wherein, represents the average state of charge of the th battery pack in the battery system, represents the state of charge of the th battery in the th battery pack in the battery system, represents the total number of batteries in the th battery pack in the battery system; The calculation formula for the difference in the state of charge of batteries within the same battery pack is as follows: ; Wherein, represents the difference in the state of charge between the th and the th batteries in the th battery pack in the battery system; The calculation formula for the average difference in the state of charge of battery packs in the battery system is as follows: ; Wherein, represents the average difference in the state of charge between the th and the th battery packs in the battery system.

[0006] Preferably, judging whether the battery system and the battery pack are in an unbalanced state according to the difference in the state of charge of batteries within the same battery pack and the average difference in the state of charge of battery packs in the battery system includes: Judging whether the difference in the state of charge is greater than a preset first threshold. If it is greater than the preset first threshold, it is determined that the battery pack is in an unbalanced state; Judging whether the average difference in the state of charge of battery packs in the battery system is greater than a preset second threshold. If it is greater than the preset second threshold, it is determined that the battery system is in an unbalanced state.

[0007] Preferably, when the battery pack is in an unbalanced state, performing internal balancing on the battery pack includes: The battery with the highest state of charge in the same battery pack charges the battery with the lowest state of charge; When the difference between the state of charge of the battery with the highest state of charge and the state of charge of the batteries within the same battery pack that do not participate in the equalization is less than the third threshold, then the batteries within the same battery pack transfer the excess state of charge to the battery with a low state of charge; Determine whether the battery pack is in an unbalanced state. When the battery pack is in an unbalanced state, repeat the above operations; When the battery pack is in a balanced state, determine whether the battery pack is in a working state. When the battery pack is in a working state, perform equalized discharge on the battery pack.

[0008] Preferably, when the battery system is in an unbalanced state, perform inter-group equalization on the battery system, including: Determine whether the battery system is in a working state; When the battery system is in a working state, determine whether the battery system is in a charging state; When the battery system is in a charging state, perform equalized charging on the battery system according to a pre-trained charging control model.

[0009] Preferably, the charging control model includes a charging optimization model and a deep reinforcement learning model. The charging optimization model includes a first objective function, a second objective function, and constraint conditions. The deep reinforcement learning model is used to solve the charging optimization model to obtain an optimal charging strategy.

[0010] Preferably, the expression of the first objective function is as follows: ; Wherein, represents the first objective function, represents the corresponding sampling step number, represents the state of charge at the sampling step number, represents the target state of charge of the battery pack, the expression of the second objective function is as follows: ; Wherein, represents the second objective function, represents the time-of-use electricity price of electricity at the charging sampling period represents the charging current, represents the charging voltage; the expression of the constraint condition is as follows: ; Wherein, Represents the maximum allowable charging current, Represents the maximum allowable charging voltage, Represents the maximum value of the state of charge, Represents the charging sampling period The temperature of the battery system at Represents the maximum allowable temperature of the battery system, where, , Represents the charging sampling period The surface temperature of the battery system at Represents the charging sampling period The core temperature of the battery system at

[0011] Preferably, the deep reinforcement learning model is used to solve the charging optimization model to obtain an optimal charging strategy, including: Determine the reward function according to the constraint conditions of the charging optimization model, where the reward function includes a first reward function, a second reward function, a third reward function, and a fourth reward function; Construct a charging policy network and a policy evaluation network, where the charging policy network is used to control the charging current of the battery, the policy evaluation network is used to evaluate the selection of each charging policy during training, the hidden layers of the charging policy network and the policy evaluation network are both fully connected layers, the first layer of the charging policy network is a relu function, the second layer includes the expectation and variance of the battery charging policy distribution, the activation function of the expectation part is a tanh function, and the activation function of the variance part is a softplus function; the policy evaluation network consists of two layers of networks, the activation function of the first layer is a relu function; the activation function of the second layer is a tanh function; Update the charging policy according to the charging objective function of the charging policy network, where the charging objective function of the charging policy network is as follows: ; Where, Represents the charging objective function of the charging policy network, Represents taking the mean value of the interval , Represents the charging policy network parameters of the current round, Represents the charging policy network parameters of the previous update round, Represents the battery charging policy of the current round, Represents the battery charging policy of the previous round, Represents the battery charging current Under the battery state The reward function; After the training phase, a charging policy network that maximizes the total reward of the charging rounds is obtained. The policy evaluation network used to assist in the training of the policy network does not participate in the execution phase. The charging policy network receives the battery charging state and outputs a charging policy probability distribution.

[0012] Preferably, the first reward function is as follows: ; Where represents the charging sampling period and is the first reward function at time The second reward function is as follows: ; Where represents the second reward function, represents the highest time-of-use electricity price in the past day; The third reward function is as follows: ; Where represents the third reward function; The fourth reward function is as follows: ; Where represents the fourth reward function.

[0013] Preferably, the temperature control model uses a BP neural network and a PID control algorithm. Among them, the input layer neuron variables of the BP neural network are the actual output temperature value, the desired output temperature value, the system error, and the control amount. According to the input variables of the BP neural network, the input and output of the intermediate layer nodes of the BP neural network are obtained using the Sigmoid function. According to the input and output of the intermediate layer nodes, the input and output of the output layer are determined. Among them, the output of the output layer is the proportional gain, integral gain, and derivative gain of the PID control algorithm. The PID control algorithm calculates the PID control amount according to the output of the output layer, discretizes the PID control amount to obtain the PID control increment, and controls the fan according to the PID control increment. Among them, the expressions for the input and output of the intermediate layer nodes are as follows: ; Where represents the input of the intermediate layer node, represents the weighting coefficient between the input layer neuron and the intermediate layer neuron, represents the input layer neuron variable, represents the output of the intermediate layer node, represents the Sigmoid function, represents the number of neurons in the middle layer; The input and output expressions of the output layer are as follows: ; in, represents the input of the output layer, represents the weight coefficient between the intermediate neurons and the output layer neurons, represents the output of the output layer, Represents the non-negative Sigmoid function; The calculation formula of the PID control increment is as follows: ; in, Represents the PID control increment, that is Indicates the fan speed. , and represents the proportional gain, integral gain and differential gain, Represents the deviation between the actual output and the expected output.

[0014] The beneficial effects of the present invention are as follows: (1) The present invention obtains the state of charge of each battery in the battery pack by the ampere-hour integration method, and determines whether the battery pack is in a balanced state according to the difference in the battery state of charge. The system can timely discover the problem of unbalanced state of the battery pack and make corrections. The balance of the battery directly affects its charging and discharging efficiency and service life. The balanced battery pack can effectively avoid damage caused by excessive discharge or charging of some batteries, thereby improving the performance of the overall battery system and extending the service life; (2) The present invention ensures that the energy consumption of each battery during the charging and discharging process tends to be consistent through the functions of intra-group balancing and inter-group balancing. By eliminating the difference in the state of charge between the batteries, the energy of the battery pack can be more evenly distributed, and the efficiency loss of individual batteries caused by excessive charging and discharging can be avoided, thereby improving the energy utilization efficiency of the battery pack; (3) When the battery pack enters an unbalanced state, the temperature control unit of the present invention performs temperature control according to a pre-trained temperature control model. When the battery pack is unbalanced, some batteries may generate heat due to excessive charging and discharging. The system avoids overheating or overcooling of the battery through intelligent temperature control, prevents battery damage or overheating caused by excessive temperature, ensures that the battery works within the optimal temperature range, and further improves safety and performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 A schematic diagram of the system architecture of an overall system in an embodiment of the present invention.

[0016] Figure numerals: 1. state determination unit; 2. difference calculation unit; 3. balance judgment unit; 4. state balance unit; 5. temperature control unit. Detailed implementation mode

[0017] Example 1, as Figure 1 shown, a battery management system proposed by the present invention includes: State determination unit 1, which is used to obtain the state of charge of each battery in the battery pack according to the ampere-hour integration method, and determine the average state of charge of the battery pack according to the state of charge of each battery in the battery pack; Difference calculation unit 2, which is used to determine the state of charge difference of the batteries in the same battery pack according to the state of charge of the batteries, and determine the average state of charge difference of the battery packs in the battery system according to the average state of charge of the battery packs; Balancing judgment unit 3, which is used to judge whether the battery system and the battery pack are in an unbalanced state according to the state of charge difference of the batteries in the same battery pack and the average state of charge difference of the battery packs in the battery system; State balancing unit 4, which is used to perform intra-group balancing on the battery pack when the battery pack is in an unbalanced state, and perform inter-group balancing on the battery system when the battery system is in an unbalanced state; Temperature control unit 5, which is used to perform temperature control on the battery pack according to a pre-trained temperature control model when the battery pack performs intra-group balancing.

[0018] In the present invention, the state of charge (SOC) of each battery in the battery pack is obtained by the ampere-hour integration method, and the average SOC of the battery pack is calculated based on this data. The ampere-hour integration method is a commonly used method for estimating the SOC of a battery. By accumulating the charge and discharge current of the battery and combining it with the initial battery charge, the current SOC of the battery can be calculated. Identify the SOC difference between each battery in the battery pack and the overall SOC difference of the battery pack. If the SOC difference of some batteries in the battery pack is too large, it may affect the overall performance of the battery pack. By calculating these differences, the health state of the battery can be evaluated. Analyze the balance state inside the battery pack and the overall battery system. If the SOC difference between some batteries in the battery pack is large, or there is an obvious SOC difference between different battery packs, then the battery pack is in an unbalanced state. At this time, the system needs to take measures to balance. Inside the battery pack, the balancing method usually reduces the SOC difference between batteries by adjusting the charge and discharge state of the battery or using a bypass current (current flowing from a high-capacity battery to a low-capacity battery). Intra-pack balancing can be achieved by the battery management system adjusting the charging strategy of each battery. When there is an SOC difference between different battery packs in the entire battery system, it is necessary to balance between the battery packs, which usually involves distributing the power between different battery packs so that the SOC of each battery pack is as close as possible, thereby improving the efficiency of the overall battery system and avoiding over-discharge or over-charging of some battery packs. The performance and lifespan of the battery are greatly affected by temperature. Too high or too low temperature will have a negative impact on the battery, which may lead to capacity attenuation, overheating, or even danger. Through the temperature control unit, the system can monitor the temperature of the battery pack in real time and maintain the battery within the ideal operating temperature range through cooling or heating. This unit can intelligently adjust the temperature according to the state of the battery pack and environmental conditions using a pre-trained temperature control model to optimize the operating state of the battery.

[0019] Embodiment 2. A battery management system proposed by the present invention. Compared with Embodiment 1, this embodiment further includes: The calculation formula for the average state of charge of the battery pack is as follows: ; Wherein, represents the average state of charge of the th battery pack in the battery system, represents the state of charge of the th battery in the th battery pack in the battery system, represents the total number of batteries in the th battery pack in the battery system; The calculation formula for the difference in the state of charge of batteries in the same battery pack is as follows: ; Wherein, Indicates the state of charge difference between the th and th and th batteries within the The calculation formula for the average state of charge difference of battery packs within the battery system is as follows: ; Wherein, Indicates the average state of charge difference between the th and th battery packs within the battery system.

[0020] In an alternative embodiment, determining whether the battery system and the battery packs are in an unbalanced state based on the state of charge difference between the batteries within the same battery pack and the average state of charge difference of the battery packs within the battery system includes: Determining whether the state of charge difference is greater than a preset first threshold. If it is greater than the preset first threshold, it is determined that the battery pack is in an unbalanced state; Determining whether the average state of charge difference of the battery packs within the battery system is greater than a preset second threshold. If it is greater than the preset second threshold, it is determined that the battery system is in an unbalanced state.

[0021] It should be noted that the state of charge difference of the battery pack refers to the maximum difference in the SOC values between different batteries within the battery pack. Assuming there are multiple batteries within the battery pack, the SOC value of each battery is monitored in real time by the battery management system. When the SOC difference between these batteries is too large, it indicates that the batteries in the battery pack have not reached an equilibrium state; if the maximum SOC difference (state of charge difference) within the battery pack exceeds the preset first threshold, the system will determine that the battery pack is in an unbalanced state. This first threshold is usually set according to the battery characteristics and performance requirements during the battery design stage, such as 0.5%, 1%, etc. The selection of this threshold needs to consider the type of battery pack, operating conditions, and the strictness of the balance requirement; if the average SOC difference between different battery packs within the battery system exceeds the preset second threshold, the system will determine that the entire battery system is in an unbalanced state. The setting of the second threshold may be adjusted according to factors such as the characteristics of different battery packs and load requirements. If the SOC difference exceeds this threshold, the system needs to perform balancing between the battery packs to ensure that the SOC of different battery packs is as consistent as possible.

[0022] In an alternative embodiment, when the battery pack is in an unbalanced state, in-group balancing of the battery pack is performed, including: The battery with the highest state of charge within the same battery pack charges the battery with the lowest state of charge; When the difference between the state of charge of the battery with the highest state of charge and the state of charge of the batteries within the same battery pack that do not participate in the equalization is less than the third threshold, then multiple batteries within the same battery pack transfer the excess state of charge to the battery with a low state of charge; Determine whether the battery pack is in an unbalanced state. When the battery pack is in an unbalanced state, repeat the above operations; When the battery pack is in a balanced state, then determine whether the battery pack is in a working state. When the battery pack is in a working state, perform balanced discharging on the battery pack.

[0023] It should be noted that when the SOC of a certain battery is the highest within the group, and the SOC of another battery is the lowest within the group, the battery with the highest SOC will charge the battery with the lowest SOC to reduce the SOC difference between them. This method can effectively help the SOCs of the batteries within the battery pack tend to be consistent, thereby improving the performance of the battery pack and extending the battery life; if some batteries within the group do not participate in the equalization (that is, the difference in their SOCs is less than the preset third threshold), then the remaining batteries (the batteries with higher SOCs) will transfer the excess state of charge to the batteries with lower SOCs, which means that the state of charge within the battery pack is allocated among the batteries to ensure that all batteries operate within a reasonable SOC range; when the battery pack is in an unbalanced state, the system will repeatedly perform the above charging and energy transfer operations until the battery pack reaches a balanced state. This process of repeated adjustment can prevent excessive SOC differences from occurring between the batteries and ensure the stability and long-term operation of the battery pack; in the working state, the balanced discharging of the battery pack means that the battery pack will adjust the SOC differences of the batteries within the battery pack to ensure that during the discharging process, the load and power output between the batteries are balanced; the balanced discharging operation will ensure that the SOC of each battery within the battery pack remains as consistent as possible during the discharging process, thereby preventing individual batteries from being damaged due to over-discharging during the discharging process. In addition, balanced discharging also helps to improve the overall efficiency of the battery pack and ensure that the battery pack can stably output electrical energy during operation; in some cases, when the SOC differences of some batteries are small and they do not participate in the equalization, the system will use the transfer of excess power to transfer the excess power in the batteries with high SOCs to the batteries with low SOCs. This method can not only improve the equalization of the battery pack but also prevent the batteries from being damaged due to overcharging or over-discharging.

[0024] In an optional embodiment, when the battery system is in an unbalanced state, perform inter-group equalization on the battery system, including: Determine whether the battery system is in a working state; When the battery system is in a working state, then determine whether the battery system is in a charging state; When the battery system is in a charging state, perform balanced charging on the battery system according to the pre-trained charging control model.

[0025] It should be noted that for the charging state judgment: when the battery is in the working state, it is necessary to further determine whether it is in the charging state. This is usually achieved through the following methods: Current direction: Whether the battery is charging can be judged by a current sensor. If the current flows from the charging device to the battery, it indicates that the battery is in the charging state; Battery voltage change: If the battery voltage gradually rises and is within the normal charging voltage range, it indicates that the battery is charging; Charging device state: The working state of the battery charger can also reflect whether the battery is in the charging state. If the charger is in the on state and is supplying current to the battery, the battery is in the charging state; Working state judgment: The working state of the battery system can be evaluated through the following indicators: Battery voltage: The voltage range of the battery during normal operation should be within the specified range. If the voltage is too low or too high, it may mean that the battery is no longer in the normal working state; Battery temperature: Excessive or too low temperature will affect the working state of the battery. If the battery temperature exceeds or is lower than the set range, it may trigger over-temperature protection, resulting in the battery no longer working; Current sensor: Whether the battery system is in the discharging or charging state can be judged by the magnitude of the current. If the current is zero, it may indicate that the battery system is not in the working state; Fault monitoring: If there are any faults in the battery system, such as communication faults, battery module faults, etc., it will also cause it to fail to work properly.

[0026] In an alternative embodiment, the charging control model includes a charging optimization model and a deep reinforcement learning model. The charging optimization model includes a first objective function, a second objective function, and constraint conditions. The deep reinforcement learning model is used to solve the charging optimization model to obtain the optimal charging strategy.

[0027] It should be noted that the charging optimization model describes the objectives of the battery charging process through mathematical models and objective functions, and based on these objective functions, it searches for the best charging strategy; Deep Reinforcement Learning (DRL) is an algorithm that combines deep learning and reinforcement learning, and can gradually learn the best strategy through experience accumulation in environmental interactions. In battery charging control, the deep reinforcement learning model can be used to solve the charging optimization model, especially in complex charging scenarios and dynamic environments, which has significant advantages.

[0028] In an alternative embodiment, the expression of the first objective function is as follows: ; Wherein, represents the first objective function, represents the sampling step corresponding to represents the state of charge at the Represents the target state of charge of the battery pack, Represents the sampling period; The expression of the second objective function is as follows: ; Among them, Represents the second objective function, Represents the charging sampling period The time-of-use electricity price of electricity at Represents the charging current, Represents the charging voltage; The expression of the constraint condition is as follows: ; Among them, Represents the maximum allowable charging current, Represents the maximum allowable charging voltage, Represents the maximum value of the state of charge, Represents the charging sampling period The temperature of the battery system at Represents the maximum allowable temperature of the battery system, among which, , Represents the charging sampling period The surface temperature of the battery system at Represents the charging sampling period The core temperature of the battery system at

[0029] In an alternative embodiment, the deep reinforcement learning model is used to solve the charging optimization model to obtain an optimal charging strategy, including: Determine the reward function according to the constraint conditions of the charging optimization model, among which, the reward function includes the first reward function, the second reward function, the third reward function and the fourth reward function; Construct a charging policy network and a policy evaluation network. Among them, the charging policy network is used to control the charging current of the battery, and the policy evaluation network is used to evaluate the selection of each charging policy during training. The hidden layers of the charging policy network and the policy evaluation network are both fully connected layers. The first layer of the charging policy network is the relu function, and the second layer includes the expectation and variance of the battery charging policy distribution. The activation function of the expectation part is the tanh function, and the activation function of the variance part is the softplus function; the policy evaluation network consists of two layers of networks. The activation function of the first layer is the relu function; the activation function of the second layer is the tanh function; Update the charging policy according to the charging objective function of the charging policy network. Among them, the charging objective function of the charging policy network is as follows: ; Among them, represents the charging objective function of the charging policy network, represents the interval to calculate the mean value, represents the charging policy network parameters in the current round, represents the charging policy network parameters in the previous update round, represents the battery charging policy in the current round, represents the battery charging policy in the previous round, represents the battery charging current under the battery state of the reward function; After the training phase, a charging policy network that maximizes the total reward of the charging rounds is obtained. The policy evaluation network for assisting the policy network training does not participate in the execution phase. The charging policy network receives the battery charging state and outputs the charging policy probability distribution.

[0030] It should be noted that the charging policy network is responsible for generating the probability distribution of the charging current and adjusting the charging policy according to the current state; the policy evaluation network is responsible for evaluating the quality of each charging policy. The role of the evaluation network is to "score" the selection of the charging policy to help the charging policy network update; in the training phase, the charging policy network and the policy evaluation network are jointly trained. The charging policy network generates the distribution of the charging current, and the policy evaluation network scores the policy to guide the charging policy network to update the parameters; in the inference phase (i.e., the execution phase), the charging policy network outputs the probability distribution of the charging policy according to the battery charging state and performs the corresponding charging operation.

[0031] In an alternative embodiment, the first reward function is as follows: ; wherein, represents the charging sampling period of the first reward function; The second reward function is as follows: ; wherein, represents the second reward function, represents the highest time-of-use electricity price in the past day; The third reward function is as follows: ; wherein, represents the third reward function; The fourth reward function is as follows: ; wherein, represents the fourth reward function.

[0032] In an optional embodiment, the temperature control model adopts a BP neural network and a PID control algorithm. Among them, the input layer neuron variables of the BP neural network are the actual output temperature value, the desired output temperature value, the system error, and the control amount. According to the input variables of the BP neural network, the input and output of the intermediate layer nodes of the BP neural network are obtained by using the Sigmoid function. The input and output of the output layer are determined according to the input and output of the intermediate layer nodes. Among them, the output of the output layer is the proportional gain, integral gain, and derivative gain of the PID control algorithm. The PID control algorithm calculates the PID control amount according to the output of the output layer, discretizes the PID control amount to obtain the PID control increment, and controls the fan according to the PID control increment. The expressions for the input and output of the intermediate layer nodes are as follows: ; Among them, represents the input of the intermediate layer node, represents the weighting coefficient between the input layer neuron and the intermediate layer neuron, represents the input layer neuron variable, represents the output of the intermediate layer node, represents the Sigmoid function, represents the number of intermediate layer neurons; The expressions for the input and output of the output layer are as follows: ; Among them, represents the input of the output layer, represents the weighting coefficient between the intermediate neuron and the output layer neuron, represents the output of the output layer, represents the non - negative Sigmoid function; The calculation formula for the PID control increment is as follows: ; Among them, represents the PID control increment, that is, represents the fan speed, , and represent the proportional gain, integral gain, and derivative gain, represents the deviation between the actual output and the expected output.

[0033] It should be noted that the PID control increment is calculated based on the difference between the control quantity at the current moment and the control quantity at the previous moment; according to the calculated PID control increment, the control quantity of the fan (such as the fan speed) will be adjusted. The basic purpose of fan control is to help the system achieve more accurate temperature control by changing the fan speed. The change in the control quantity of the fan should be as smooth as possible to avoid excessive temperature fluctuations.

[0034] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made without departing from the spirit of the present invention within the scope of knowledge possessed by those skilled in the art to which the present invention pertains.

Claims

1. A battery management system, characterized in that: include: A state determination unit (1), the state determination unit (1) being used to obtain the state of charge of each battery in a battery pack according to an ampere-hour integration method, and to determine the average state of charge of the battery pack according to the state of charge of each battery in the battery pack; A difference calculation unit (2), the difference calculation unit (2) being used to determine the difference in state of charge of batteries in the same battery pack according to the state of charge of the battery, and to determine the difference in average state of charge of battery packs in the battery system according to the average state of charge of the battery pack; A balancing judgment unit (3), the balancing judgment unit (3) being used to judge whether the battery system and the battery group are in an unbalanced state according to the difference in state of charge of the batteries in the same battery group and the difference in average state of charge of the battery groups in the battery system; A state balancing unit (4), the state balancing unit (4) being used to perform intra-group balancing on the battery group when the battery group is in an unbalanced state, and to perform inter-group balancing on the battery system when the battery system is in an unbalanced state; A temperature control unit (5), the temperature control unit (5) being used to control the temperature of the battery pack according to a pre-trained temperature control model when the battery pack is performing internal balancing.

2. A battery management system according to claim 1, characterized in that: The calculation formula of the average state of charge of the battery pack is as follows: ； in, Indicates the battery system The average state of charge of the battery pack, Indicates the battery system The battery pack The state of charge of each battery, Indicates the battery system The total number of cells in a battery pack; The calculation formula for the difference in state of charge of the batteries in the same battery pack is as follows: ； in, Indicates the battery system The battery pack and The difference in state of charge between the batteries; The calculation formula of the average state of charge difference of the battery packs in the battery system is as follows: ； in, Indicates the battery system and The average state of charge difference between the battery groups.

3. A battery management system according to claim 2, characterized in that: Determining whether the battery system and the battery group are in an unbalanced state according to the state of charge difference of the batteries in the same battery group and the average state of charge difference of the battery groups in the battery system comprises: Determining whether the state of charge difference is greater than a preset first threshold, and if so, determining that the battery pack is in an unbalanced state; It is determined whether the average state of charge difference of the battery groups in the battery system is greater than a preset second threshold value. If it is greater than the preset second threshold value, it is determined that the battery system is in an unbalanced state.

4. A battery management system according to claim 3, characterized in that: When the battery group is in an unbalanced state, performing internal balancing on the battery group includes: The battery with the highest state of charge in the same battery pack charges the battery with the lowest state of charge; When the difference between the state of charge of the battery with the highest state of charge and the state of charge of the batteries in the same battery group that do not participate in balancing is less than a third threshold, the batteries in the same battery group transfer excess state of charge to the battery with the low state of charge; Determine whether the battery pack is in an unbalanced state, and if the battery pack is in an unbalanced state, repeat the above operation; When the battery pack is in a balanced state, it is determined whether the battery pack is in a working state. When the battery pack is in a working state, the battery pack is discharged in a balanced manner.

5. A battery management system according to claim 4, characterized in that: When the battery system is in an unbalanced state, the battery system is balanced among groups, including: Determining whether the battery system is in working state; When the battery system is in a working state, determining whether the battery system is in a charging state; When the battery system is in a charging state, the battery system is balanced charged according to a pre-trained charging control model.

6. A battery management system according to claim 5, characterized in that: The charging control model includes a charging optimization model and a deep reinforcement learning model. The charging optimization model includes a first objective function, a second objective function and constraints. The deep reinforcement learning model is used to solve the charging optimization model to obtain an optimal charging strategy.

7. A battery management system according to claim 6, characterized in that: The expression of the first objective function is as follows: ； in, represents the first objective function, express The corresponding number of sampling steps, Indicates The state of charge at the sampling step, Indicates the target state of charge of the battery pack. Indicates the sampling period; The expression of the second objective function is as follows: ； in, represents the second objective function, Indicates the charging sampling period Time-of-use electricity prices, Indicates the charging current, Indicates charging voltage; The constraint condition is expressed as follows: ； in, Indicates the maximum allowable charging current, Indicates the maximum allowable charging voltage, Indicates the maximum value of the state of charge, Indicates the charging sampling period The temperature of the battery system, represents the maximum allowable battery system temperature, where , Indicates the charging sampling period The surface temperature of the battery system is Indicates the charging sampling period The core temperature of the battery system.

8. A battery management system according to claim 7, characterized in that: The deep reinforcement learning model is used to solve the charging optimization model to obtain the optimal charging strategy, including: Determining a reward function according to the constraint conditions of the charging optimization model, wherein the reward function includes a first reward function, a second reward function, a third reward function and a fourth reward function; Construct a charging strategy network and a strategy evaluation network, wherein the charging strategy network is used to control the charging current of the battery, and the strategy evaluation network is used to evaluate the selection of each charging strategy during training. The hidden layers of the charging strategy network and the strategy evaluation network are both fully connected layers. The first layer of the charging strategy network is a relu function, and the second layer contains the expectation and variance of the battery charging strategy distribution. The activation function of the expectation part is a tanh function, and the activation function of the variance part is a softplus function. The strategy evaluation network contains two layers of networks, the activation function of the first layer is a relu function, and the activation function of the second layer is a tanh function. The charging strategy is updated according to the charging objective function of the charging strategy network, wherein the charging objective function of the charging strategy network is as follows: ； in, represents the charging objective function of the charging strategy network, Represents the interval Find the mean, Represents the network parameters of the current round charging strategy, Indicates the network parameters of the charging strategy in the last update round, Indicates the battery charging strategy for the current round, represents the battery charging strategy of the previous round, Indicates the battery charging current In battery status The reward function under After the training phase, the charging strategy network that maximizes the total reward of the charging round is obtained. The strategy evaluation network used to assist the strategy network training does not participate in the execution phase. The charging strategy network accepts the battery charging status and outputs the charging strategy probability distribution.

9. A battery management system according to claim 8, characterized in that: The first reward function is as follows: ； in, Indicates the charging sampling period The first reward function when ; The second reward function is as follows: ； in, represents the second reward function, Indicates the highest time-of-use electricity price in the past day; The third reward function is as follows: ； in, represents the third reward function; The fourth reward function is as follows: ； in, Denotes the fourth reward function.

10. A battery management system according to claim 9, characterized in that: The temperature control model adopts BP neural network and PID control algorithm, wherein the input layer neuron variables of the BP neural network are the actual output temperature value, the expected output temperature value, the system error and the control quantity. According to the BP neural network input variables, the input and output of the intermediate layer node of the BP neural network are obtained by using Sigmoid function, and the input and output of the output layer are determined according to the input and output of the intermediate layer node, wherein the output of the output layer is the proportional gain, integral gain and differential gain of the PID control algorithm, and the PID control algorithm calculates the PID control quantity according to the output of the output layer, discretizes the PID control quantity to obtain the PID control increment, and controls the fan according to the PID control increment, wherein the input and output of the intermediate layer node are expressed as follows: ； in, represents the input of the middle layer node, represents the weight coefficient between the input layer neurons and the intermediate layer neurons, represents the input layer neuron variable, represents the output of the middle layer node, represents the Sigmoid function, represents the number of neurons in the middle layer; The input and output expressions of the output layer are as follows: ； in, represents the input of the output layer, represents the weight coefficient between the intermediate neurons and the output layer neurons, represents the output of the output layer, Represents the non-negative Sigmoid function; The calculation formula of the PID control increment is as follows: ； in, Represents the PID control increment, that is Indicates the fan speed. , and represents the proportional gain, integral gain and differential gain, Represents the deviation between the actual output and the expected output.

Images