Data analysis method, device and storage medium

Abstract

The present invention relates to the field of data analysis technology, and in particular to a data analysis method, device and storage medium. The method comprises the following steps: independently classifying the stop state and driving state of an electric vehicle, and separately training each historical data that can affect the energy consumption of the electric vehicle by respectively training the stop state and driving state of the electric vehicle using two independent neural network models. When it is necessary to predict the energy consumption per unit time of the electric vehicle, it is only necessary to input multiple groups of data collected from the two states of the electric vehicle into the trained neural network model, respectively obtain multiple SOC change values, and respectively calculate the mean SOC change values ​​of the two states. Then, by calculating the time ratio of the stop state and the driving state in the historical state of the electric vehicle as a weight, the energy consumption of the electric vehicle per unit time can be more accurately predicted.

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, and in particular to a data analysis method, device and storage medium. Background Art

[0002] In recent years, with the rapid development of new energy, it can replace traditional energy in many scenarios. Electric vehicles have gradually replaced traditional fuel vehicles as the mainstream civilian transportation tool. Electric vehicles generally use lithium batteries as energy. Exploring the energy consumption law of electric vehicles is a key factor in studying the range of electric vehicles. There are many factors affecting the energy consumption of electric vehicles, including environmental factors, such as ambient temperature, humidity and air pressure, as well as factors of vehicle motion state, including moving speed, acceleration, motor torque, and motor speed. Since electric vehicles contain a variety of uncertain operating states in actual work, such as stopping, accelerating, decelerating, and driving at a constant speed, the changes in these operating states are affected by the individual driving habits of users and the unique driving path rules. Therefore, it is difficult to use general common laws to accurately predict the energy consumption changes of electric vehicles. Therefore, how to analyze multiple data affecting the energy consumption of electric vehicles to accurately predict their energy consumption has become a technical problem that needs to be solved urgently. Summary of the invention

[0003] The technical problem to be solved by the present invention is: how to analyze multiple data affecting the energy consumption of electric vehicles to achieve accurate prediction of their energy consumption.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] The present invention relates to a data analysis method, comprising the following steps:

[0006] Get the unit time when the electric vehicle is stopped The average value of ambient temperature, humidity and pressure at t; get the average value of ambient temperature, humidity and pressure at the same time. The average current I1 at both ends of the lithium battery within t;

[0007] According to the formula Calculate the unit time Change value of the charge state of the lithium battery within t , the Q max is the rated capacity of the lithium battery;

[0008] The unit time The mean value of ambient temperature, ambient humidity and ambient pressure at t is input into the first relationship model to obtain the unit time The charge state change value within t The first relationship model is obtained by training a plurality of first sample data based on a neural network, and the first sample data includes a unit time of the corresponding relationship The average value of ambient temperature, humidity and pressure at t and the corresponding state of charge change value ;

[0009] Get the unit time of the electric vehicle in driving state The average ambient temperature, humidity, pressure, speed, acceleration, motor torque and speed of the vehicle at t, and the average value of the unit time The average current I2 at both ends of the lithium battery within t;

[0010] According to the formula Calculate the unit time Change value of the charge state of the lithium battery within t ;

[0011] The unit time The mean values ​​of ambient temperature, ambient humidity, ambient pressure, speed, acceleration, motor torque and motor speed at t are input into the second relationship model to obtain the unit time The charge state change value within t The second relationship model is obtained by training a plurality of second sample data based on a neural network, and the second sample data includes a unit time of the corresponding relationship tThe average value of ambient temperature, ambient humidity, ambient pressure, speed, acceleration, motor torque and motor speed and the corresponding charge state change value ;

[0012] Obtain the sum of the stop state time t1 in the electric vehicle historical data and the sum of the driving state time t2 in the vehicle historical data;

[0013] Get multiple unit times of stop state in historical data of electric vehicles The mean values ​​of ambient temperature, ambient humidity and ambient pressure within t are used as test samples, and multiple values ​​are obtained through the first relationship model. , and get multiple The mean ;

[0014] Get multiple time units of driving status in the electric vehicle historical data The mean values ​​of ambient temperature, ambient humidity, ambient pressure, speed, acceleration, motor torque and motor speed within t are used as test samples, and multiple , and get multiple The mean ;

[0015] According to the formula , get the electric vehicle unit time Predicted value of energy consumption at t .

[0016] Furthermore, in the above data analysis method, the second relationship model is established based on a generative adversarial network; wherein the generative adversarial network includes a generator and a classifier, and both the generator and the classifier include a convolutional neural network and a fully connected layer.

[0017] Furthermore, in the above data analysis method, the first relationship model is obtained by training a neural network according to a plurality of first sample data, and the training method of the first relationship model includes the following steps:

[0018] A plurality of first sample data are obtained, wherein the first sample data includes a unit time of a corresponding relationship The average value of ambient temperature, humidity and pressure at t and the corresponding state of charge change value ;

[0019] For any first sample data, input the first sample data into the neural network, and output the matching probability corresponding to the first sample data;

[0020] Calculating a loss value according to a preset loss function and a matching probability corresponding to the first sample data;

[0021] If the loss value is less than a preset threshold, the training of the first relationship model is completed.

[0022] Furthermore, in the above data analysis method, the second relationship model is obtained by training the second relationship model based on a neural network according to a plurality of second sample data, and the training method of the second relationship model comprises the following steps:

[0023] In the generative adversarial network structure, the original data is input into the generator; wherein the original data is a subset of the real data, and the real data includes the unit time of the corresponding relationship The average values ​​of ambient temperature, humidity, pressure, speed, acceleration, motor torque and speed of the vehicle at t and the corresponding state of charge change values ;

[0024] The generator receives input data and outputs an output obtained by predicting the input data through a convolutional neural network output;

[0025] The output of the generator is combined with the real data as the input of the classifier, so that the classifier can learn the real data and the generated data, and back propagation is used to improve the training accuracy;

[0026] The output of the classifier is returned to train the generator.

[0027] The present invention also relates to a data analysis device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the data analysis method is implemented when the processor executes the program.

[0028] The present invention also relates to a computer-readable storage medium storing a computer program, wherein the computer program comprises program instructions which, when executed by a processor, cause the processor to perform the above-mentioned data analysis method.

[0029] The beneficial effect of the present invention is that in the data analysis method of the present invention, the stopping state and the driving state of the electric vehicle are independently classified, and the historical data that can affect the energy consumption of the electric vehicle are obtained by respectively training the stopping state and the driving state of the electric vehicle using two independent neural network models for separate training. When it is necessary to predict the energy consumption of the electric vehicle per unit time, it is only necessary to input multiple groups of data collected from the two states of the electric vehicle into the trained neural network model, and obtain multiple SOC change values ​​respectively, and calculate the mean SOC change values ​​of the two states respectively, and then calculate the time ratio of the stopping state and the driving state in the historical state of the electric vehicle as a weight, so that the energy consumption of the electric vehicle per unit time can be predicted more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 A schematic diagram of a data analysis method according to a specific embodiment of the present invention; DETAILED DESCRIPTION

[0031] In order to explain the technical content, achieved objectives and effects of the present invention in detail, the following is an explanation in combination with the implementation modes and the accompanying drawings.

[0032] Please refer to Figure 1 , Figure 1 A flow chart of a data analysis method provided for a specific embodiment of the present invention; the method comprises the following steps:

[0033] S1: Get the unit time when the electric vehicle is stopped The average value of ambient temperature, humidity and pressure at t; get the average value of ambient temperature, humidity and pressure at the same time. The average current I1 at both ends of the lithium battery within t;

[0034] In the above steps, the ambient temperature, ambient humidity and ambient air pressure of the electric vehicle can be collected by the vehicle-mounted temperature sensor, vehicle-mounted humidity sensor and vehicle-mounted air pressure sensor, and can be collected according to the unit time. t collects multiple discrete data points and calculates the unit time through the processor The mean value of ambient temperature, ambient humidity and ambient air pressure within t;

[0035] S2: According to the formula Calculate the unit time Change value of the charge state of the lithium battery within t , the Q max is the rated capacity of the lithium battery;

[0036] In the above steps, by measuring the unit time The discrete data of multiple continuous current point values ​​within t is used to calculate the average current value I1 within the unit time through the processor, and the unit time is calculated according to the above formula Change value of the state of charge of the lithium battery within t ;

[0037] S3: Unit time The mean value of ambient temperature, ambient humidity and ambient pressure at t is input into the first relationship model to obtain the unit time The charge state change value within t The first relationship model is obtained by training a plurality of first sample data based on a neural network, and the first sample data includes a unit time of the corresponding relationship The average value of ambient temperature, humidity and pressure at t and the corresponding state of charge change value ;

[0038] S4: Obtain the unit time of the electric vehicle in the driving state The average ambient temperature, humidity, pressure, speed, acceleration, motor torque and speed of the vehicle at t, and the average value of the unit time The average current I2 at both ends of the lithium battery within t;

[0039] In the above steps, the ambient temperature, ambient humidity and ambient air pressure of the electric vehicle can be collected by the vehicle-mounted temperature sensor, the vehicle-mounted humidity sensor and the vehicle-mounted air pressure sensor, and the speed and acceleration can be realized by any device equipped with a GPS system, such as a mobile device. Specifically, the mobile device obtains the driving trajectory data of the target vehicle based on the GPS system. For example, the GPS data of the target vehicle is obtained through the GPS system, which mainly includes the speed, position and time of the target vehicle. Then, the acceleration corresponding to the target vehicle at the current moment can be calculated according to the speed, position and time, thereby obtaining the speed and acceleration corresponding to the target vehicle at the current moment. The torque value and motor speed of the motor can be obtained through the torque sensor and the speed sensor, which can be calculated according to the unit time. t collects multiple discrete data points and calculates the unit time through the processor The average ambient temperature, average ambient humidity, average ambient pressure, average speed, average acceleration, average motor torque and average motor speed within t.

[0040] S5: According to the formula Calculate the unit time Change value of the charge state of the lithium battery within t ;

[0041] In the above steps, by measuring the unit time The discrete data of multiple continuous current point values ​​within t, the average value I2 of the current within the unit time is calculated by the processor, and the unit time is calculated according to the above formula Change value of the state of charge of the lithium battery within t ;

[0042] S6: Unit time The mean values ​​of ambient temperature, ambient humidity, ambient pressure, speed, acceleration, motor torque and motor speed at t are input into the second relationship model to obtain the unit time The charge state change value within t The second relationship model is obtained by training a plurality of second sample data based on a neural network, and the second sample data includes a unit time of the corresponding relationship tThe average value of ambient temperature, ambient humidity, ambient pressure, speed, acceleration, motor torque and motor speed and the corresponding charge state change value ;

[0043] S7: Obtaining the sum of the stop state time t1 in the electric vehicle historical data and the sum of the driving state time t2 in the electric vehicle historical data;

[0044] In the above steps, when the user purchases the vehicle, the stopping time and driving time of the electric vehicle will be counted. This is the habitual data reflecting the user's vehicle usage frequency, which is a personal factor for predicting the energy consumption of the vehicle.

[0045] S8: Get multiple unit times of the stop state in the electric vehicle historical data The mean values ​​of ambient temperature, ambient humidity and ambient pressure within t are used as test samples, and multiple values ​​are obtained through the first relationship model. , and get multiple The mean ;

[0046] S9: Obtain multiple unit times of driving status in historical data of electric vehicles The mean values ​​of ambient temperature, ambient humidity, ambient pressure, speed, acceleration, motor torque and motor speed within t are used as test samples, and multiple , and get multiple The mean ;

[0047] S10: According to the formula , get the electric vehicle unit time Predicted value of energy consumption at t .

[0048] In the above steps, since the first relationship model and the second relationship model are close to the user's driving habit preferences after a lot of training, the corresponding SOC change mean of the stopping state and the driving state are obtained through multiple test samples, and then weighted by the time proportion of the two states in the historical data to obtain the prediction result.

[0049] As a preferred implementation, the second relationship model is established based on a generative adversarial network; wherein the generative adversarial network includes a generator and a classifier, and both the generator and the classifier include a convolutional neural network and a fully connected layer.

[0050] In the above method, the generator and classifier in the generative adversarial network (GANs) are a combination of convolutional neural networks (CNNs) and fully connected layers. The generator uses convolutional neural networks to extract features, combines the features through fully connected layers, and outputs the predicted results through deconvolution layers. Its input is multiple unit times of driving status. The average value of ambient temperature, humidity, pressure, speed, acceleration, motor torque and motor speed within t are processed and output as the corresponding SOC change value; the classifier uses convolutional neural network to extract features and classifies through the fully connected layer to determine whether the input is real data or predicted data, and its output is a Boolean value. The structure of the generative adversarial network can improve the accuracy of energy consumption prediction.

[0051] As a preferred embodiment, the first relationship model is obtained by training a neural network according to a plurality of first sample data, and the training method of the first relationship model comprises the following steps:

[0052] A plurality of first sample data are obtained, wherein the first sample data includes a unit time of a corresponding relationship The average value of ambient temperature, humidity and pressure at t and the corresponding state of charge change value ;

[0053] For any first sample data, input the first sample data into the neural network, and output the matching probability corresponding to the first sample data;

[0054] Calculating a loss value according to a preset loss function and a matching probability corresponding to the first sample data;

[0055] If the loss value is less than a preset threshold, the training of the first relationship model is completed.

[0056] In the above steps, in order to prevent overfitting in the training process, L2 regularization and dropout algorithms are used to limit the model parameters of the convolutional neural network, increase the generalization ability of the convolutional neural network, and improve the accuracy. Among them, the preset loss function of the convolutional neural network adds an L2 regularization term. In order to avoid the gradient disappearance or gradient explosion caused by the deepening of the number of layers of the convolutional neural network and the slow convergence speed, the batch normalization (BatchNormalization, referred to as BN) algorithm can be used to standardize the input of certain layers or all layers, thereby fixing the mean and variance of the input signal of each layer, so that the input of each layer of the convolutional neural network maintains the same distribution during the training process, which simplifies the construction of deep neural networks and speeds up the learning convergence speed.

[0057] As a preferred embodiment, the second relationship model is obtained by training a neural network according to a plurality of second sample data, and the training method of the second relationship model comprises the following steps:

[0058] In the generative adversarial network structure, the original data is input into the generator; wherein the original data is a subset of the real data, and the real data includes the unit time of the corresponding relationship The average values ​​of ambient temperature, humidity, pressure, speed, acceleration, motor torque and speed of the vehicle at t and the corresponding state of charge change values ;

[0059] The generator receives input data and outputs an output obtained by predicting the input data through a convolutional neural network output;

[0060] The output of the generator is combined with the real data as the input of the classifier, so that the classifier can learn the real data and the generated data, and back propagation is used to improve the training accuracy;

[0061] The output of the classifier is returned to train the generator.

[0062] In the second relational model, the generator and classifier are trained separately. The loss function of the classifier is the cross entropy of the classification result, while the loss function of the generator is [the sum of the squares of the error between the generator's output and the expected value] plus [the result of inputting the generator's output into the classifier]. The two are not added directly, but added according to certain weights. In this way, the purpose of using the results of the classifier to train the generator is achieved.

[0063] The present invention also relates to a data analysis device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned data analysis method when executing the program.

[0064] It should be noted that the computer program in the above-mentioned memory can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0065] The present invention also relates to a computer-readable storage medium storing a computer program, wherein the computer program comprises program instructions which, when executed by a processor, cause the processor to perform the above-mentioned data analysis method.

[0066] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's specification and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A data analysis method, characterized in that: The following steps are involved: Get the unit time when the electric vehicle is in a stopped state The average value of ambient temperature, humidity and air pressure of the unit time The average current value I1 at both ends of the lithium battery; According to the formula Calculate the unit time Change value of the charge state of the lithium battery , the Q max is the rated capacity of the lithium battery; The unit time The mean value of the ambient temperature, the mean value of the ambient humidity and the mean value of the ambient pressure are input into the first relationship model to obtain the unit time The charge state change value within The first relationship model is obtained by training a plurality of first sample data based on a neural network, and the first sample data includes a unit time of the corresponding relationship The average value of ambient temperature, humidity and pressure and the corresponding state of charge change value ; Get the unit time of the electric vehicle in driving state The average ambient temperature, humidity, pressure, speed, acceleration, motor torque and speed of the vehicle are obtained at the same time. The average current value I2 at both ends of the lithium battery; According to the formula Calculate the unit time Change value of the charge state of the lithium battery ; The unit time The average value of ambient temperature, the average value of ambient humidity, the average value of ambient pressure, the average value of speed, the average value of acceleration, the average value of motor torque and the average value of motor speed are input into the second relationship model to obtain the unit time The charge state change value within The second relationship model is obtained by training a plurality of second sample data based on a neural network, and the second sample data includes a unit time of the corresponding relationship The average values ​​of ambient temperature, ambient humidity, ambient pressure, speed, acceleration, motor torque and motor speed and the corresponding state of charge change values ; Obtain the sum of the stop state time t1 in the electric vehicle historical data and the sum of the driving state time t2 in the vehicle historical data; Get multiple unit times of stop state in historical data of electric vehicles The mean of ambient temperature, humidity and pressure in the environment are taken as test samples, and multiple , and get multiple The mean ; Get multiple time units of driving status in the electric vehicle historical data The mean values ​​of ambient temperature, ambient humidity, ambient pressure, speed, acceleration, motor torque and motor speed are taken as test samples, and multiple values ​​are obtained through the second relationship model. , and get multiple The mean ; According to the formula , get the electric vehicle unit time The predicted value of energy consumption .

2. The data analysis method according to claim 1, characterized in that: The second relationship model is established based on a generative adversarial network; wherein the generative adversarial network includes a generator and a classifier, and both the generator and the classifier include a convolutional neural network and a fully connected layer.

3. The data analysis method according to claim 1, characterized in that: The first relationship model is obtained by training a neural network according to a plurality of first sample data, and the training method of the first relationship model comprises the following steps: A plurality of first sample data are obtained, wherein the first sample data includes a unit time of a corresponding relationship The average value of ambient temperature, humidity and pressure at t and the corresponding state of charge change value ; For any first sample data, input the first sample data into the neural network, and output the matching probability corresponding to the first sample data; Calculating a loss value according to a preset loss function and a matching probability corresponding to the first sample data; If the loss value is less than a preset threshold, the training of the first relationship model is completed.

4. The data analysis method according to claim 2, characterized in that: The second relationship model is obtained by training a neural network according to a plurality of second sample data, and the training method of the second relationship model comprises the following steps: In the generative adversarial network structure, the original data is input into the generator; wherein the original data is a subset of the real data, and the real data includes the unit time of the corresponding relationship The average values ​​of ambient temperature, humidity, pressure, speed, acceleration, motor torque and speed of the vehicle at t and the corresponding state of charge change values ; The generator receives input data and outputs an output obtained by predicting the input data through a convolutional neural network output; The output of the generator is combined with the real data as the input of the classifier, so that the classifier can learn the real data and the generated data, and back propagation is used to improve the training accuracy; The output of the classifier is returned to train the generator.

5. Data analysis equipment, characterized in that The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the data analysis method according to any one of claims 1 to 4 when executing the program.

6. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes the data analysis method according to any one of claims 1 to 4.

Images