Electric vehicle battery charge state prediction method, device and system and storage medium

Abstract

The invention discloses an electric vehicle battery charge state prediction method, device and system and a storage medium. The electric vehicle battery charge state prediction method comprises the steps of S1, obtaining environment data and battery data; s2, calibrating the environment data and the battery data; and S3, inputting the calibrated environment data and the battery data into the two-channel composite information layered convolutional neural network to predict the state of charge of the battery. By adopting the technical scheme of the invention, the SOC of the battery can be accurately predicted in real time, and the cruising ability and the safety of medium and heavy electric vehicles can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of power batteries for medium and heavy-duty electric vehicles, and particularly relates to a method and device, a system, and a storage medium for predicting the state of charge of an electric vehicle battery. Background Art

[0002] With the increasingly serious global energy crisis and environmental problems, medium and heavy-duty electric vehicles have gradually become the focus of the development of the global automotive industry due to their advantages such as zero emissions and high energy efficiency. As the core energy storage device of electric vehicles, the performance of lithium power batteries directly affects the driving range and safety of electric vehicles. SOC, the full name is State of Charge, the charging state of the battery, also known as the remaining power, represents the ratio of the remaining dischargeable power to the fully charged power after the battery has been used for a period of time or maintained for a long time, usually expressed as a percentage. Represented by a byte of hexadecimal, that is, two digits (the value range is 0 to 100), indicating that the remaining power is 0% to 100%. When SOC = 0, the battery is fully discharged, and when SOC = 100%, the battery is fully charged. The SOC (charging state) of the battery reflects the actual available power of the battery and is a very important indicator in the operation of electric vehicles. Currently, there are various forms of electric vehicles, such as hybrid, plug-in hybrid, fuel cell, pure electric, etc. Limited by battery technology and charging technology levels, there are constraints such as energy density, driving range, charging speed, and charging station construction. However, pure electric vehicles are the ultimate direction. Therefore, the accurate prediction of battery SOC has always been a major challenge in the field of electric vehicles.

[0003] Currently, the commonly used methods for estimating battery SOC include the open-circuit voltage method, the ampere-hour integration method, the Kalman filter method, the neural network method, etc. However, these methods all have certain limitations in practical applications. For example, the open-circuit voltage method has poor real-time performance and depends on the rest time; the ampere-hour integration method requires high accuracy of the initial SOC and has an error accumulation problem; the Kalman filter method has high computational complexity and strong dependence on the battery model; the existing methods do not fully and comprehensively consider the internal and external factors affecting battery SOC. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and device, a system, and a storage medium for predicting the state of charge of an electric vehicle battery. Based on a data-driven method, it fully considers external factors such as road conditions, weather conditions, vehicle load, travel route, and usage time, as well as the influence of the performance parameters of the battery itself, and accurately predicts the SOC of the battery in real time, improving the driving range and safety of medium and heavy-duty electric vehicles.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for predicting the state of charge of an electric vehicle battery, comprising:

[0007] Step S1: Obtain environmental data and battery data;

[0008] Step S2: Calibrate the environmental data and battery data;

[0009] Step S3: Input the calibrated environmental data and battery data into a dual-channel composite information hierarchical convolutional neural network to predict the state of charge of the battery.

[0010] Preferably, the dual-channel composite information hierarchical convolutional neural network includes: two groups of channels, and each channel includes: 1 input layer, 3 convolutional layers, and 3 pooling layers; the two groups of channels share a fully connected layer, and the fully connected layer is placed at the end of the convolutional neural network, and a Softmax classifier is used to output the classification result together. In the Softmax classification layer, calculate the probability of each category occurring, and the expression is where, o j is the output probability value of the j-th neuron in the output layer, a j is the output value of the j-th neuron in the output layer, and m is the number of neurons in the output layer; the output is a softmax output layer, and the output layer includes m neurons, where m is the number of state classifications. The output layer includes 10 neurons, that is, the output matrix is 1×10, representing 10 levels of SOC, that is, the state of charge of the battery is 0-10%, 11-20%, 21-30%, 31-40%, 41-50%, 51-60%, 61-70%, 71-80%, 81-90%, 91-100%; in the convolutional layer, the ReLU function is used as the non-linear activation function.

[0011] Preferably, the environmental data includes road conditions, traffic conditions, climate seasons, weather temperature, weather wind force level, vehicle type, vehicle load, vehicle travel mileage, and vehicle travel speed; the battery data includes battery number, battery replacement time, battery replacement amount, SOC when replaced, SOC when removed, charging time, charging amount, discharging time, discharging amount, number of battery charge and discharge cycles, designed mileage of the battery, and designed life of the battery.

[0012] The present invention also provides an electric vehicle battery state of charge prediction device, including:

[0013] An acquisition module, configured to acquire environmental data and battery data;

[0014] A calibration module, configured to calibrate the environmental data and battery data;

[0015] A prediction module, configured to input the calibrated environmental data and battery data into a dual-channel composite information hierarchical convolutional neural network to predict the state of charge of the battery.

[0016] Preferably, the dual-channel composite information hierarchical convolutional neural network includes: two groups of channels, each channel containing: 1 input layer, 3 convolutional layers, and 3 pooling layers; the two groups of channels share a fully connected layer, and the fully connected layer is placed at the end of the convolutional neural network. A Softmax classifier is used to output the classification results together. In the Softmax classification layer, the probability of each category occurring is calculated, and the expression is where, o j is the output probability value of the j-th neuron in the output layer, and a j is the output value of the j-th neuron in the output layer, and m is the number of neurons in the output layer; the output is a softmax output layer, and the output layer includes m neurons, where m is the number of classifications of the states. The output layer includes 10 neurons, that is, the output matrix is 1×10, representing 10 levels of SOC, that is, the state of charge of the battery is 0-10%, 11-20%, 21-30%, 31-40%, 41-50%, 51-60%, 61-70%, 71-80%, 81-90%, 91-100%; in the convolutional layer, the ReLU function is used as the non-linear activation function.

[0017] Preferably, the environmental data includes road conditions, traffic conditions, climate seasons, weather temperature, weather wind force level, vehicle type, vehicle load, vehicle travel mileage, and vehicle travel speed; the battery data includes battery number, battery replacement time, battery replacement amount, SOC when replaced, SOC when removed, charging time, charging amount, discharging time, discharging amount, number of battery charge and discharge cycles, designed mileage of the battery, and designed life of the battery.

[0018] The present invention also provides an electric vehicle battery state of charge prediction system, including: a memory and a processor, and a computer program is stored on the memory and run by the processor. The computer program executes the electric vehicle battery state of charge prediction method when run by the processor.

[0019] The present invention also provides a storage medium, on which a computer program is stored, and the computer program executes the electric vehicle battery state of charge prediction method when running.

[0020] The present invention has the following technical effects:

[0021] 1. Improve prediction accuracy: By comprehensively considering various factors such as the environment (external factor) and the battery (internal factor), the present invention can more accurately predict the SOC of the battery.

[0022] 2. Enhance real-time performance: Utilize the vehicle positioning system and the charge and battery replacement record system to obtain real-time data, and combine with the deep learning algorithm for rapid prediction to meet the requirements of dynamic monitoring of various electric vehicle batteries.

[0023] 3. Optimize energy usage: Adjusting the driving strategy or charging / discharging plan according to the prediction results helps extend the driving range of electric vehicles and improve energy utilization efficiency.

[0024] 4. Strong adaptability: The method of the present invention can be applied to different types of medium and heavy-duty electric trucks and power batteries, with strong versatility and practicality. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to the provided drawings.

[0026] Figure 1 It is a flowchart of the method for predicting the state of charge of an electric vehicle battery in an embodiment of the present invention;

[0027] Figure 2 It is a schematic structural diagram of a dual-channel composite information hierarchical convolutional neural network. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0029] To make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0030] Embodiment 1:

[0031] As Figure 1 shown, an embodiment of the present invention provides a method for predicting the state of charge of an electric vehicle battery, including:

[0032] Step S1, obtain environmental data and battery data;

[0033] Step S2, calibrate the environmental data and battery data;

[0034] Step S3, input the calibrated environmental data and battery data into a dual-channel composite information hierarchical convolutional neural network to predict the state of charge of the battery.

[0035] As an implementation manner of an embodiment of the present invention, in step S1, the environmental data includes road conditions, traffic conditions, climate season, weather temperature, weather wind force level, vehicle type, vehicle load, vehicle travel mileage, and vehicle travel speed; the battery data includes battery number, battery replacement time, battery replacement amount, SOC when replaced, SOC when taken off, charging time, charging amount, discharging time, discharging amount, battery charge and discharge times, battery designed mileage, and battery designed life.

[0036] Furthermore, the process of obtaining environmental data is as follows:

[0037] The road conditions are obtained according to the geographic information system; the traffic conditions are obtained from the map API and traffic information, and the integrated data of the GPS and Beidou systems is converted into the vehicle driving state (such as urban congestion, highway cruising, parking, etc.); the climate season is defined by month; the weather temperature and weather wind force are obtained through the weather real-time forecasting system; the vehicle load is obtained according to the vehicle weighing system; the vehicle travel mileage is obtained through the vehicle mileage measurement system; the vehicle travel time is obtained by reading the in-vehicle clock.

[0038] Collect the system data corresponding to the environmental (external factor) parameters every 5 seconds for 2 consecutive minutes. The data includes road conditions (x 1 ), traffic conditions (x 2 ), climate season (x 3 ), weather temperature (x 4 ), weather wind force level (x 5 ), vehicle type (x 6 ), vehicle load (x 7 ), vehicle travel mileage (x 8 ), and vehicle travel speed (x 9 ).

[0039] Furthermore, the process of obtaining battery data is as follows:

[0040] The battery designed mileage and battery designed life are obtained from its archive information database; the battery replacement time, battery replacement amount, SOC when replaced, SOC when taken off, charging time, charging amount, discharging time, discharging amount, and battery charge and discharge times are obtained according to the charge and discharge records.

[0041] Obtain the latest 15 data records of this battery from the battery charge and discharge records, that is, the battery (internal factor). Each record includes the battery designed mileage (y 1 ) and the battery designed life (y 2 ) obtained from its archive information database; the battery replacement time (y 3 ), the battery replacement amount (y 4 ), the SOC when replaced (y 5 ), the SOC when taken off (y 6 ), the charging time (y 7)、 Charge amount (y 8 )、 Discharge time (y 9 )、 Discharge amount (y 10 ) and Number of battery charge and discharge cycles (y 11 ).

[0042] As an implementation manner of the embodiment of the present invention, in step S2, the environmental data and battery data are calibrated and converted into an input matrix suitable for the intelligent network. Table 1 shows the data calibration results of the embodiment of the present invention.

[0043] Table 1

[0044]

[0045]

[0046] As an implementation manner of the embodiment of the present invention, a dual-channel composite information hierarchical convolutional neural network is used as the intelligent network to predict the state of charge of the battery, as Figure 2 shown.

[0047] The dual-channel data consists of environmental data and battery data. The environmental data is one channel, and the input matrix is 9×24 quantities. They overlap head and tail at a certain overlap rate, and after being expanded and replicated 9 times and then padded with zeros, a two-dimensional 64×32 sample matrix is formed; the battery (internal cause) data is the other channel, and the input matrix is 15×11 quantities, which also overlap head and tail at a certain overlap rate, and after being expanded and replicated 6 times and then padded with zeros, a two-dimensional 64×32 sample matrix is formed. The data set is expressed as: X∈R h ×w×d×n , where X represents a sample, h represents the row of the sample, w represents the column of the sample, d represents the number of channels, that is, the number of original signals, and n represents the number of samples; in the implementation of the present invention, each running sample is 200, forming a 4D sample set of 64×32×2×2000.

[0048] The dual-channel composite information hierarchical convolutional neural network in the embodiment of the present invention includes: two groups of channels, and each channel includes: 1 input layer, 3 convolutional layers, and 3 pooling layers; the two groups of channels share a fully connected layer, and the fully connected layer is placed at the end of the convolutional neural network, and a Softmax classifier is used to output the classification result together. In the Softmax classification layer, the probability of each category occurring is calculated, and the expression is In the formula, o j is the output probability value of the jth neuron in the output layer, and a j$y_j$ is the output value of the $j$-th neuron in the output layer, and $m$ is the number of neurons in the output layer; the output is a softmax output layer, and the output layer includes $m$ neurons. Among them, $m$ is the number of state classifications. The output layer includes 10 neurons, that is, the output matrix is $1\times10$, representing 10 levels of SOC, that is, the battery charge state is 0-10%, 11-20%, 21-30%, 31-40%, 41-50%, 51-60%, 61-70%, 71-80%, 81-90%, 91-100%; the ReLU function is used as the non-linear activation function in the convolutional layer.

[0049] Max pooling is used for pooling; the convolution process is: S i = f(con(M i , C i ) + B i ), where in the formula, S i is the output feature map, $f$ is the activation function, $con$ is the convolution operation, $M i is the input feature map, $C i is the convolution kernel, and $B i is the bias; the pooling layer also has a kernel similar to the convolution kernel, called the sliding window, whose function is to reduce the dimension of the sequence features while maintaining the scale invariance of the features and reduce the computational complexity. The present invention adopts the max pooling method, that is, each time it slides, the maximum value in its area is used as the output. The definition of max pooling is as follows: In the formula, is the $h$-th value in the $i$-th input feature vector of the $l + 1$ layer, and $D$ is the pooling domain; is the $g$-th value in the $i$-th output feature vector of the $l + 1$ layer.

[0050] The cross-entropy error is used as the target loss function, and the cross-entropy error is where $n$ is the number of samples, $k$ is the number of output neurons, $y nk is the target value of the $k$-th output of the $n$-th sample, is the predicted value of the $k$-th output of the $n$-th sample.

[0051] Example 2:

[0052] The embodiment of the present invention also provides an electric vehicle battery charge state prediction device, including:

[0053] An acquisition module for acquiring environmental data and battery data;

[0054] A calibration module for calibrating the environmental data and battery data;

[0055] A prediction module for inputting the calibrated environmental data and battery data into a dual-channel composite information hierarchical convolutional neural network to predict the battery charge state.

[0056] As an implementation manner of an embodiment of the present invention, the dual-channel composite information hierarchical convolutional neural network includes: two groups of channels, and each channel includes: 1 input layer, 3 convolutional layers, and 3 pooling layers; the two groups of channels share a fully connected layer, and the fully connected layer is placed at the end of the convolutional neural network, and a Softmax classifier is used to output the classification result together. In the Softmax classification layer, the probability of each category occurring is calculated, and the expression is where o j is the output probability value of the j-th neuron in the output layer, and a j is the output value of the j-th neuron in the output layer, and m is the number of neurons in the output layer; the output is a softmax output layer, and the output layer includes m neurons, where m is the number of classifications of the states. The output layer includes 10 neurons, that is, the output matrix is 1×10, representing 10 levels of SOC, that is, the state of charge of the battery is 0-10%, 11-20%, 21-30%, 31-40%, 41-50%, 51-60%, 61-70%, 71-80%, 81-90%, 91-100%; in the convolutional layer, the ReLU function is used as the non-linear activation function.

[0057] As an implementation manner of an embodiment of the present invention, the environmental data includes road conditions, traffic conditions, climate seasons, weather temperature, weather wind force level, vehicle type, vehicle load, vehicle travel mileage, and vehicle travel speed; the battery data includes battery number, battery replacement time, battery replacement amount, SOC when replaced, SOC when removed, charging time, charging amount, discharging time, discharging amount, number of battery charge and discharge cycles, designed mileage of the battery, and designed life of the battery.

[0058] Example 3:

[0059] An embodiment of the present invention further provides an electric vehicle battery state of charge prediction system, including: a memory and a processor, and a computer program is stored on the memory and run by the processor, and the computer program executes the electric vehicle battery state of charge prediction method when run by the processor.

[0060] Example 4:

[0061] An embodiment of the present invention further provides a storage medium, and a computer program is stored on the storage medium, and the computer program executes the electric vehicle battery state of charge prediction method when running.

[0062] The above-described embodiments are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention should fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for predicting the state of charge of an electric vehicle battery, characterized in that: include: Step S1, obtaining environmental data and battery data; Step S2, calibrating environmental data and battery data; Step S3: input the calibrated environmental data and battery data into a dual-channel composite information hierarchical convolutional neural network to predict the battery state of charge.

2. The method for predicting the state of charge of an electric vehicle battery according to claim 1, characterized in that: The dual-channel composite information hierarchical convolutional neural network includes: two groups of channels, each channel contains: 1 input layer, 3 convolution layers, 3 pooling layers; the two groups of channels share a fully connected layer, the fully connected layer is placed at the end of the convolutional neural network, and the Softmax classifier is used to output the classification results together. In the Softmax classification layer, the probability of each category is calculated, and the expression is Among them, j is the output probability value of the jth neuron in the output layer, a j is the output value of the jth neuron in the output layer, and m is the number of neurons in the output layer; the output is a softmax output layer, which includes m neurons, where m is the number of state classifications and the output layer includes 10 neurons, that is, the output matrix is ​​1×10, representing 10 levels of SOC, that is, the battery state of charge is 0-10%, 11-20%, 21-30%, 31-40%, 41-50%, 51-60%, 61-70%, 71-80%, 81-90%, 91-100%; the ReLU function is used as the nonlinear activation function in the convolutional layer.

3. The method for predicting the state of charge of an electric vehicle battery according to claim 1, characterized in that: Environmental data include road conditions, traffic conditions, climate season, weather temperature, weather wind level, vehicle type, vehicle load, vehicle mileage, and vehicle travel speed; battery data include battery number, battery replacement time, battery replacement amount, SOC when replaced, SOC when replaced, charging time, charging amount, discharging time, discharging amount, battery charge and discharge times, battery design mileage, and battery design life.

4. A device for predicting the state of charge of an electric vehicle battery, characterized in that: include: An acquisition module, used to acquire environmental data and battery data; Calibration module, used to calibrate environmental data and battery data; The prediction module is used to input the calibrated environmental data and battery data into a dual-channel composite information hierarchical convolutional neural network to predict the battery state of charge.

5. The electric vehicle battery charge state prediction device according to claim 4, characterized in that: The dual-channel composite information hierarchical convolutional neural network includes: two groups of channels, each channel contains: 1 input layer, 3 convolution layers, 3 pooling layers; the two groups of channels share a fully connected layer, the fully connected layer is placed at the end of the convolutional neural network, and the Softmax classifier is used to output the classification results together. In the Softmax classification layer, the probability of each category is calculated, and the expression is Among them, j is the output probability value of the jth neuron in the output layer, a j is the output value of the jth neuron in the output layer, and m is the number of neurons in the output layer; the output is a softmax output layer, which includes m neurons, where m is the number of state classifications and the output layer includes 10 neurons, that is, the output matrix is ​​1×10, representing 10 levels of SOC, that is, the battery state of charge is 0-10%, 11-20%, 21-30%, 31-40%, 41-50%, 51-60%, 61-70%, 71-80%, 81-90%, 91-100%; the ReLU function is used as the nonlinear activation function in the convolutional layer.

6. The electric vehicle battery charge state prediction device according to claim 5, characterized in that: Environmental data include road conditions, traffic conditions, climate season, weather temperature, weather wind level, vehicle type, vehicle load, vehicle mileage, and vehicle travel speed; battery data include battery number, battery replacement time, battery replacement amount, SOC when replaced, SOC when replaced, charging time, charging amount, discharging time, discharging amount, battery charge and discharge times, battery design mileage, and battery design life.

7. A system for predicting the state of charge of an electric vehicle battery, characterized in that: include: A memory and a processor, wherein the memory stores a computer program executed by the processor, and when the computer program is executed by the processor, the method for predicting the state of charge of a battery of an electric vehicle as claimed in any one of claims 1 to 3 is executed.

8. A storage medium, characterized in that: The storage medium stores a computer program, which executes the electric vehicle battery state of charge prediction method according to any one of claims 1 to 3 when running.

Images