A method, storage medium, and device for estimating the remaining capacity of a vehicle-mounted lead-acid battery.

Abstract

This invention discloses a method, storage medium, and device for estimating the remaining capacity of a vehicle-mounted lead-acid battery, relating to the field of vehicle-mounted lead-acid battery technology. The method includes determining the battery state; if in use, using the ampere-hour integral method to estimate the state of charge (SOC) in real time; if in a static state, using the open-circuit voltage method to estimate the SOC, and recording the corresponding voltage, current, and temperature; constructing a neural network, using voltage, current, and temperature as input, and the estimated SOC as the desired output; the input is passed through neurons and added to a bias, then through an activation function to obtain the output; the output error is fed back into the input layer to calculate the error of the neuron and the gradient of the loss function with respect to the neuron's weights and biases; the weights and biases of the neurons are updated based on the gradient, and a new output is calculated; the trained neural network is used to estimate the remaining battery capacity. This method can reduce the error in estimating the remaining battery capacity and improve the estimation accuracy.

Description

Technical Field

[0001] This invention relates to the field of vehicle lead-acid battery technology, and in particular to a method, storage medium, and device for estimating the remaining power of a vehicle lead-acid battery. Background Technology

[0002] Lead-acid batteries are auxiliary power sources in gasoline-powered vehicles, typically used for engine starting and auxiliary power supply. SOC (State of Charge) describes the state of charge of a lead-acid battery, representing the ratio of its remaining capacity to its capacity at full charge, usually expressed as a percentage. Its value ranges from 0 to 1; SOC = 0 indicates the battery is fully discharged, and SOC = 1 indicates the battery is fully charged. SOC is a crucial input parameter for the power management strategy of a gasoline-powered vehicle's energy management system, and its estimation accuracy significantly impacts the effectiveness of overall vehicle energy management. Regularly monitoring SOC ensures the battery remains within a reasonable charge range during use, avoiding the negative effects of overcharging or discharging on battery performance and lifespan.

[0003] Currently, commonly used methods for estimating State of Charge (SOC) include: the open-circuit voltage method, which determines the SOC value based on the battery's open-circuit voltage (OCV). This involves measuring the battery's open-circuit voltage at different SOC values ​​and then obtaining a function of SOC-OCV through data fitting. The ampere-hour integration method integrates the current flowing into or out of the battery over time; based on the battery's initial SOC value, the remaining charge at a given moment can be obtained. Model-based methods, such as the Kalman filter, etc., are also used. However, various methods for estimating SOC struggle to achieve accurate estimates, exhibiting significant errors due to interference from factors such as temperature. Summary of the Invention

[0004] The purpose of this invention is to address the problem that existing technologies struggle to accurately estimate the State of Charge (SOC) and suffer from significant errors. This invention proposes a method for estimating the remaining capacity of a vehicle-mounted lead-acid battery, comprising the following steps:

[0005] S1. Conduct a SOC-OCV experiment on the battery, plot the curves of battery SOC and OCV, and fit the relationship between battery SOC and OCV.

[0006] S2. Measure the charging and discharging current of the battery to determine whether the battery is in a resting state or in use.

[0007] S3. If the battery is in use, the open-circuit voltage-ampere-hour integral estimation method is used to estimate the remaining battery capacity in real time; if the battery is in a static state, the battery SOC and OCV relationship fitted in S1 is used to estimate the remaining battery capacity, and the corresponding voltage, current and temperature are recorded.

[0008] S4. Construct a neural network, using the voltage, current, and temperature recorded in S3 as the neural network input, and the estimated remaining battery power as the actual output label.

[0009] S5. Associate the input signal with each neuron and sum it with the bias, then pass it through the activation function to obtain the output of the neural network;

[0010] S6. Input the difference between the output of the neural network obtained in S5 and the true output label back into the input layer of the neural network, and calculate the gradient of the loss function of the neural network with respect to the output layer and the hidden layer.

[0011] S7. Update the weights and biases of the hidden layer and output layer based on the gradient, and use the updated weights and biases to calculate the new output;

[0012] S8. Repeat S4-S7 until the output of the neural network reaches the preset value or the number of neural network training iterations reaches the preset value. Input the voltage, current and temperature measured under various battery conditions into the trained neural network to obtain the final estimated remaining battery power.

[0013] Furthermore, a SOC-OCV experiment was conducted on the battery, specifically as follows:

[0014] Fully charge the battery, let it stand for 1 hour, and measure the open circuit voltage when SOC=100%.

[0015] An electronic load discharges the battery at a constant current of 10A for 1 hour, then allows it to stand for 1.5 hours, and measures the open-circuit voltage when the SOC is 90%.

[0016] The open-circuit voltage was measured sequentially at each stage: SOC = 80%, SOC = 70%, ..., SOC = 10%, and SOC = 0%.

[0017] Record the open-circuit voltage, constant current discharge value, and settling time at each stage.

[0018] Furthermore, the relationship between the fitted battery state of charge and the battery open-circuit voltage is expressed as:

[0019] OCV=a+b×SOC+c×SOC2+d×SOC3

[0020] Where OCV represents open-circuit voltage, SOC represents state of charge, and a, b, c, and d are parameters that need to be identified.

[0021] Furthermore, the open-circuit voltage-ampere-hour integral estimation method is used to estimate the remaining battery capacity in real time, specifically as follows:

[0022]

[0023] Where SOC(t) represents the state of charge at time t, SOC(t0) represents the initial state of charge, and I(t) represents the current at time t.

[0024] Furthermore, the loss function of the neural network is expressed as:

[0025]

[0026]

[0027] z o =W T x+b

[0028] Where L represents the loss function, x represents the input of the neural network, and σ(z) o ) represents z o The output after the Sigmoid activation function, where W represents the weight matrix of the input neuron and b represents the bias of the hidden layer.

[0029] Furthermore, the loss function of the neural network with respect to the gradients of the output and hidden layers is expressed as follows:

[0030]

[0031] Where, δ o Let L represent the gradient of the loss function with respect to the output layer, and let z represent the loss function. o Let y represent the input to the Sigmoid activation function, and y represent the output of the neural network. true Represents the actual output label, σ′(z o )=y(1-y) is the derivative of the Sigmoid activation function of the output layer;

[0032]

[0033] Where, δ j z represents the gradient of the loss function with respect to the neuron in the j-th hidden layer. j V represents the weighted sum of neurons in the j-th hidden layer. j σ′(z) represents the weight from the j-th neuron in the hidden layer to the output layer. j ) = a j (1-a j ) represents the derivative of the Sigmoid activation function in the hidden layer, a jThis represents the output value of the j-th hidden layer neuron after passing through the activation function.

[0034] Furthermore, the weights and biases of the hidden and output layers updated based on gradients are expressed as follows:

[0035] The weights and biases of the output layer are updated based on the gradient of the output layer using the loss function:

[0036] v j =v j -α·δ o ·a j

[0037] b o =b o -α·δ o

[0038] Among them, v j The weights from the j-th neuron in the hidden layer to the output layer are represented by α, δ, and α represents the learning rate. o a represents the gradient of the loss function with respect to the output layer. j b represents the output value of the j-th hidden layer neuron after passing through the activation function. o Indicates the bias of the output layer;

[0039] The weights and biases of the hidden layer are updated based on the gradient of the hidden layer using the loss function:

[0040] w ji =w ji -α·δ j ·x i

[0041] b j =b j -α·δ j

[0042] Among them, w ji δ represents the weight connecting the i-th input to the j-th neuron in the hidden layer. j Let x represent the gradient of the loss function with respect to the neuron in the j-th hidden layer. i Let b represent the i-th input of the neural network. j This represents the bias of the j-th hidden layer.

[0043] The present invention also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for estimating the remaining power of a vehicle-mounted lead-acid battery.

[0044] The present invention also proposes an electronic device, including a processor and a memory, wherein the processor and the memory are interconnected, wherein the memory is used to store a computer program, the computer program including computer-readable instructions, and the processor is configured to invoke the computer-readable instructions to execute the above-described method for estimating the remaining power of a vehicle-mounted lead-acid battery.

[0045] The beneficial effects of the technical solution provided by this invention are:

[0046] This invention proposes a method for estimating the remaining capacity of a vehicle-mounted lead-acid battery. It estimates the remaining capacity of the battery in both its resting and active states using the open-circuit voltage method and the open-circuit voltage-ampere-hour integral estimation method, recording the corresponding voltage, current, and temperature. The voltage, current, and temperature are used as inputs to a neural network, and the estimated remaining capacity is used as the desired output. The weighted input neurons of the neural network are activated and summed to obtain the output. Gradients are used to update the weights of the neurons and the bias of the neural network, and the network is iteratively trained. The trained neural network is then used for estimating the remaining battery capacity. This method can reduce the error in estimating the remaining battery capacity and improve the estimation accuracy. Attached Figure Description

[0047] Figure 1 This is an embodiment of the vehicle-mounted lead-acid battery management system of the present invention;

[0048] Figure 2 This is a schematic diagram of the battery data acquisition board structure in an embodiment of the present invention;

[0049] Figure 3 This is a flowchart of the reference model for estimating remaining battery power built into the MCU in this embodiment of the invention;

[0050] Figure 4 This is a schematic diagram of a real-time model for estimating remaining battery power based on data-driven technology in an embodiment of the present invention.

[0051] Figure 5 This is a flowchart of the method for estimating the remaining power of a vehicle-mounted lead-acid battery according to an embodiment of the present invention;

[0052] Figure 6 This is a block diagram of an electronic device according to an exemplary embodiment of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0054] The vehicle-mounted lead-acid battery management system of this invention embodiment is as follows: Figure 1As shown, it includes: main control communication board, battery status measurement board, cloud platform, and user terminal.

[0055] The main control communication board includes: a 4G module for communicating with the cloud platform; a storage module for saving battery status data when power is off; a power supply module for powering the main control communication board; a BMS (Battery Management System) chip; a Bluetooth chip for communicating with the user terminal; an MCU main control chip for controlling Bluetooth and 4G communication in the battery management system; and a battery data acquisition board for real-time acquisition of vehicle battery operating status data.

[0056] The battery status measurement board includes: a power module for powering the battery status measurement board; and a BMS chip, which is connected to the BMS chip of the main control communication board via two TTL (transistor-to-transistor logic integrated circuit) signal lines to transmit the data measured by the battery status measurement board to the MCU main control chip.

[0057] The cloud platform is used to store the battery SOC estimation program, communicate with the main control communication board to obtain real-time battery operating status data, and perform battery SOC estimation.

[0058] refer to Figure 2 , Figure 2 This is a schematic diagram of the battery data acquisition board structure in this invention, specifically including voltage divider resistors, dual-channel operational amplifiers, filter capacitors, and BMS chips.

[0059] The voltage divider resistors reduce the input voltage to a range that the BMS chip can withstand. The dual-channel operational amplifier is designed as two voltage followers, whose output voltage closely follows the input voltage. It features high input impedance and low output impedance, isolating the signal source and load, separating the BMS chip's power ground from the battery signal acquisition ground, and providing greater drive capability to ensure the BMS chip's normal operation during acquisition. The filter capacitors remove high-frequency noise and interference signals from the circuit, improving the quality of the power supply and acquired voltage signals. The BMS chip can measure battery voltage and temperature and communicate with other BMS chips.

[0060] Please refer to Figure 3 , Figure 3 This is a flowchart of the reference model for estimating remaining battery power embedded in the MCU in this invention.

[0061] The open-circuit voltage method estimates the State of Charge (SOC) value by obtaining a battery pack voltage fitting curve. However, due to battery polarization effects, it cannot update the SOC value in real time and can only be performed when the battery pack is idle. The ampere-hour integration method calculates the SOC value in real time by collecting data from the battery pack's charging and discharging circuit. However, there is accumulated error during the calculation process, requiring calibration of the initial SOC value. Furthermore, it only calculates the external charge and cannot estimate internal battery influencing factors. Through research and analysis of the above two methods, it was found that the advantages and disadvantages of the two methods are complementary. Therefore, the battery management device of this invention uses two algorithms to calculate and obtain the SOC value in parallel: the open-circuit voltage method obtains the initial SOC value, and the current is obtained through a Hall sensor. When the battery is in use, the main controller integrates the collected current and adds it to the initial SOC value to obtain the actual SOC value.

[0062] Please refer to Figure 4 , Figure 4 This is a schematic diagram of a real-time battery remaining capacity estimation model based on data-driven technology. The real-time battery remaining capacity estimation model based on data-driven technology uses a reference model for battery remaining capacity estimation to train a three-input single-output dataset of voltage, current, and temperature-SOC. Voltage, current, and temperature are used as input signals, with randomly preset weights W1, W2, W3…W n The input signal is associated with each neuron and summed with the bias b. Then, it is passed through the activation function f(x) to obtain the output, which is the SOC predicted by the data model. The error is calculated based on the output. The gradient of the loss function with respect to the weights and biases of each neuron is calculated in reverse based on the error, and the weights and biases are updated. The output SOC is recalculated based on voltage, current, and temperature using the updated weights and biases until the error is lower than the preset threshold or the number of model iterations reaches the threshold.

[0063] The method of this invention combines a battery remaining capacity estimation reference model and a real-time battery remaining capacity estimation model based on data-driven technology. The flowchart of the vehicle-mounted lead-acid battery remaining capacity estimation method of this invention is as follows: Figure 5 Specifically, it includes the following steps:

[0064] S1. Conduct a SOC-OCV experiment on the battery, plot the curve of battery state of charge versus battery open circuit voltage, and fit the relationship between battery state of charge (SOC) and battery open circuit voltage (OCV).

[0065] Specifically:

[0066] 1. Fully charge the battery, let it stand for 1 hour, and measure its open circuit voltage. This is the open circuit voltage corresponding to SOC=100%.

[0067] 2. The electronic load discharges the battery at a constant current of 10A for 1 hour, and then lets it stand for 1.5 hours. The open circuit voltage measured at this time is the open circuit voltage when SOC = 90%.

[0068] 3. Repeat step 2 to obtain the open-circuit voltage at SOC = 80%, SOC = 70%, ..., SOC = 10%, and SOC = 0%. When SOC is approximately 0%, the measured voltage should be approximately equal to the discharge cutoff voltage. Similarly, the last test requires a 1.5-hour rest period before measuring the open-circuit voltage corresponding to SOC = 0%. Record the open-circuit voltage, constant current discharge value, and resting time for each test stage.

[0069] The fitted relationship between the battery's state of charge and its open-circuit voltage is expressed as:

[0070] OCV = a + b × SOC + c × SOC 2 +d×SOC 3

[0071] The parameters a, b, c, and d are obtained by using multiple sets of data.

[0072] S2. Measure the charging and discharging current of the battery to determine whether the battery is in a resting state or in use.

[0073] When the battery's charging and discharging current is 0, the battery is in a resting state; when the battery's charging and discharging current is not 0, the battery is in a state of active use.

[0074] S3. If the battery is in use, the remaining battery capacity is estimated in real time using the open-circuit voltage-ampere-hour integral estimation method.

[0075] The ampere-hour (AH) integral method is a fundamental method for battery capacity measurement. It uses the AH accumulation method to estimate the state of charge (SOC) of a dynamically changing battery in real time. The calculation formula for the ampere-hour integral method is as follows:

[0076]

[0077] Where SOC(t) represents the state of charge at time t, SOC(t0) represents the initial state of charge, and I(t) represents the current at time t.

[0078] If the battery is in a quiescent state, the remaining battery capacity is estimated using the relationship between the battery state of charge and the battery open-circuit voltage fitted by S1.

[0079] Based on the relationship between the battery state of charge and the battery open-circuit voltage obtained by S1 fitting, the relevant parameters of the relationship are calculated by measuring multiple sets of data. The battery open-circuit voltage when the battery is in a static state is measured, and the battery state of charge is calculated by the relationship.

[0080] And record the corresponding voltage, current, and temperature.

[0081] S4. Construct a neural network, using the voltage, current, and temperature recorded in S3 as inputs and the estimated remaining battery power as the actual output label. Preset the weights and biases of the neurons in the neural network.

[0082] S5. Associate the input signal with each neuron and sum it with the bias, then pass it through the activation function f(x) to obtain the output, which is the SOC predicted by the data model.

[0083] S6. Input the difference between the output of the neural network obtained in S5 and the true output label back into the input layer of the neural network, and calculate the gradient of the loss function of the neural network with respect to the output layer and the hidden layer.

[0084] The loss function of a neural network is expressed as:

[0085]

[0086]

[0087] z o =W T x+b

[0088] Where L represents the loss function, x represents the input of the neural network, and σ(z) o ) represents z o The output after the Sigmoid activation function, where W represents the weight matrix of the input neuron and b represents the bias of the hidden layer.

[0089] The gradient is represented as:

[0090]

[0091] Where, δ o Let L represent the gradient of the loss function with respect to the output layer, and let z represent the loss function. o Let y represent the input to the Sigmoid activation function, and y represent the output of the neural network. true Represents the actual output label, σ′(z o )=y(1-y) is the derivative of the Sigmoid activation function of the output layer.

[0092]

[0093] Where, δ j z represents the gradient of the loss function with respect to the neuron in the j-th hidden layer. j V represents the weighted sum of neurons in the j-th hidden layer. jσ′(z) represents the weight from the j-th neuron in the hidden layer to the output layer. j ) = a j (1-a j ) represents the derivative of the Sigmoid activation function in the hidden layer, a j This represents the output value of the j-th hidden layer neuron after passing through the activation function.

[0094] S7. Update the weights and biases of the hidden and output layers based on the gradient, and use the updated weights and biases to calculate the new output.

[0095] The weights and biases of the output layer are updated based on the gradient of the output layer using the loss function:

[0096] v j =v j -α·δ o ·a j

[0097] b o =b o -α·δ o

[0098] Among them, v j The weights from the j-th neuron in the hidden layer to the output layer are represented by α, δ, and α represents the learning rate. o a represents the gradient of the loss function with respect to the output layer. j b represents the output value of the j-th hidden layer neuron after passing through the activation function. o This indicates the bias of the output layer.

[0099] The weights and biases of the hidden layer are updated based on the gradient of the hidden layer using the loss function:

[0100] w ji =w ji -α·δ j ·x i

[0101] b j =b j -α·δ j

[0102] Among them, w ji δ represents the weight connecting the i-th input to the j-th neuron in the hidden layer. j Let x represent the gradient of the loss function with respect to the neuron in the j-th hidden layer. i Let b represent the i-th input of the neural network. j This represents the bias of the j-th hidden layer.

[0103] S8. Repeat S4-S7 until the output of the neural network reaches the preset value or the number of neural network training iterations reaches the preset value. Input the voltage, current and temperature measured under various battery conditions into the trained neural network to obtain the final estimated remaining battery power.

[0104] In one exemplary embodiment, a computer-readable storage medium is included, which stores a computer program that, when executed by a processor, implements the above-described method for estimating the remaining charge of a vehicle-mounted lead-acid battery.

[0105] Please see Figure 6 In one exemplary embodiment, the device further includes an electronic device including at least one processor, at least one memory, and at least one communication bus.

[0106] The memory stores a computer program, which includes computer-readable instructions. The processor calls the computer-readable instructions stored in the memory through the communication bus to execute the above-mentioned method for estimating the remaining power of the vehicle lead-acid battery.

[0107] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for estimating the remaining capacity of a vehicle-mounted lead-acid battery, characterized in that, Includes the following steps: S1. Conduct a SOC-OCV experiment on the battery, plot the curves of battery SOC and OCV, and fit the relationship between battery SOC and OCV. S2. Measure the charging and discharging current of the battery to determine whether the battery is in a resting state or in use. S3. If the battery is in use, the open-circuit voltage-ampere-hour integral estimation method is used to estimate the remaining battery capacity in real time; if the battery is in a static state, the battery SOC and OCV relationship fitted in S1 is used to estimate the remaining battery capacity, and the corresponding voltage, current and temperature are recorded. S4. Construct a neural network, using the voltage, current, and temperature recorded in S3 as the neural network input, and the estimated remaining battery power as the actual output label. S5. Associate the input signal with each neuron and sum it with the bias, then pass it through the activation function to obtain the output of the neural network; S6. Input the difference between the output of the neural network obtained in S5 and the true output label back into the input layer of the neural network, and calculate the gradient of the loss function of the neural network with respect to the output layer and the hidden layer. The loss function of a neural network is expressed as: Where L represents the loss function, and x represents the input of the neural network. express The output after the Sigmoid activation function represents the weight matrix of the input neuron, and b represents the bias of the hidden layer; S7. Update the weights and biases of the hidden layer and output layer based on the gradient, and use the updated weights and biases to calculate the new output; S8. Repeat S4-S7 until the output of the neural network reaches the preset value or the number of neural network training iterations reaches the preset value. Input the voltage, current and temperature measured under various battery conditions into the trained neural network to obtain the final estimated remaining battery power.

2. The method for estimating the remaining capacity of a vehicle-mounted lead-acid battery according to claim 1, characterized in that, The SOC-OCV experiment was conducted on the battery, specifically as follows: Fully charge the battery, let it stand for 1 hour, and measure the open circuit voltage when SOC=100%. An electronic load discharges the battery at a constant current of 10A for 1 hour, then allows it to stand for 1.5 hours, and measures the open-circuit voltage when the SOC is 90%. The open-circuit voltage was measured sequentially at each stage: SOC=80%, SOC=70%, ..., SOC=10%, SOC=0%. Record the open-circuit voltage, constant current discharge value, and settling time at each stage.

3. The method for estimating the remaining capacity of a vehicle-mounted lead-acid battery according to claim 1, characterized in that, The fitted relationship between the battery's state of charge and its open-circuit voltage is expressed as: OCV = a + b × SOC + c × SOC² + d × SOC³ Where OCV represents open-circuit voltage, SOC represents state of charge, and a, b, c, and d are parameters that need to be identified.

4. The method for estimating the remaining capacity of a vehicle-mounted lead-acid battery according to claim 1, characterized in that, The real-time estimation of the remaining battery capacity using the open-circuit voltage-ampere-hour integral estimation method is as follows: Where SOC(t) represents the state of charge at time t, and SOC(t0) represents the initial state of charge. This represents the current at time t.

5. The method for estimating the remaining capacity of a vehicle-mounted lead-acid battery according to claim 1, characterized in that, The loss function of a neural network with respect to the gradients of the output and hidden layers is expressed as follows: in, Let L represent the gradient of the loss function with respect to the output layer. Let y represent the input to the sigmoid activation function, and y represent the output of the neural network. Indicates the actual output label. It is the derivative of the Sigmoid activation function of the output layer; in, This represents the gradient of the loss function with respect to the neuron in the j-th hidden layer. This represents the weighted sum of neurons in the j-th hidden layer. This represents the weight from the j-th neuron in the hidden layer to the output layer. This represents the derivative of the Sigmoid activation function in the hidden layer. This represents the output value of the j-th hidden layer neuron after passing through the activation function.

6. The method for estimating the remaining capacity of a vehicle-mounted lead-acid battery according to claim 5, characterized in that, The gradient-based update of the weights and biases of the hidden and output layers is represented as follows: The weights and biases of the output layer are updated based on the gradient of the output layer using the loss function: in, This represents the weight from the j-th neuron in the hidden layer to the output layer. Indicates the learning rate. This represents the gradient of the loss function with respect to the output layer. This represents the output value of the j-th hidden layer neuron after passing through the activation function. Indicates the bias of the output layer; The weights and biases of the hidden layer are updated based on the gradient of the hidden layer using the loss function: in, This represents the weight connecting the i-th input to the j-th neuron in the hidden layer. This represents the gradient of the loss function with respect to the neuron in the j-th hidden layer. This represents the i-th input to the neural network. This represents the bias of the j-th hidden layer.

7. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.

8. An electronic device, characterized in that, The device includes a processor and a memory, the processor being interconnected with the memory, wherein the memory is used to store a computer program, the computer program including computer-readable instructions, and the processor is configured to invoke the computer-readable instructions to perform the method as described in any one of claims 1-6.

Images