Intelligent pure electric vehicle braking energy recovery control method

Abstract

The invention belongs to the technical field of fingerprint identification, and provides an intelligent pure electric vehicle braking energy recovery control method, which comprises the following steps of: acquiring data such as front images, vehicle distance speed, traffic flow and road congestion by virtue of a camera, a millimeter wave radar and an Internet of Vehicles module; identifying the road condition type by using a convolutional neural network and a logic judgment rule through a vehicle-mounted processor; driver operation and vehicle state data are recorded through an ECU, and aggressive, mild, economical and other driving habits are identified by using a clustering analysis algorithm and a K-Means algorithm; the method comprises the following steps: constructing an MLP neural network prediction model by taking vehicle speed, road condition type and driving habit type as an input layer and taking future vehicle speed change and braking demand as an output layer, and gradually reducing kinetic energy recovery strength in advance according to a specific time interval and amplitude when it is predicted that a vehicle is about to enter a congested road section and a driver is aggressive habit; and adjusting the recovery force according to a preset strategy curve when the road condition changes suddenly during driving.

Description

Technical Field

[0001] The present invention belongs to the technical field of kinetic energy recovery, and specifically relates to an intelligent braking energy recovery control method for pure electric vehicles. Background Art

[0002] With the enhancement of environmental awareness and the pursuit of energy sustainability, pure electric vehicles have received extensive attention and development; as one of the key technologies for pure electric vehicles to improve energy utilization efficiency, the braking energy recovery system can convert part of the kinetic energy into electrical energy and store it during vehicle braking, thereby increasing the vehicle's cruising range; However, there are some problems in the actual application of existing braking energy recovery systems. Among them, the more prominent one is the sense of jerk generated during kinetic energy recovery, which not only affects driving comfort but also may have a certain negative impact on the vehicle's handling performance; In addition, current braking energy recovery control methods often fail to fully consider factors such as road conditions and driving habits, resulting in difficulty in achieving the optimal energy recovery efficiency and driving experience; Therefore, the present invention provides an intelligent braking energy recovery control method for pure electric vehicles. Summary of the Invention

[0003] In order to make up for the deficiencies of the prior art and solve at least one of the technical problems proposed in the background art.

[0004] The technical solution adopted by the present invention to solve its technical problems is as follows: In the first aspect, the present invention provides an intelligent braking energy recovery control method for pure electric vehicles, including the following steps: Step 1: Collect the driving data, traffic flow, and road condition congestion data of the vehicle during driving, and use the road condition recognition algorithm to judge the current road condition type; Step 2: Continuously record the driver operation data and vehicle state data by using the ECU, and obtain the driver's driving habits through the data analysis module; Step 3: Based on the judged road condition type and the analyzed driving habit data, use the neural network algorithm to construct a prediction model, and use data training to obtain the kinetic energy recovery prediction model; Step 4: In order to achieve the balance between driving comfort and energy recovery efficiency under different road condition types and driving habits, it is necessary to preset the strategy curve of the kinetic energy recovery intensity changing with the road condition type and driving habit, and use the kinetic energy recovery prediction model to dynamically adjust the kinetic energy recovery intensity; The specific process of dynamically adjusting the kinetic energy recovery intensity is as follows: Taking the average driving speed of the vehicle as an index to measure the road condition congestion degree, and at the same time considering the driver's aggressiveness score s, the kinetic energy recovery intensity F is expressed as a percentage; A strategy curve function is obtained by fitting experimental data. ; When the prediction model predicts that the vehicle is about to enter a congested section, and the average vehicle speed v within the next 1 minute is predicted to be lower than 30 km / h and the driver has an aggressive driving habit, assuming the aggressiveness score s = 8, the system starts to gradually reduce the kinetic energy recovery intensity in advance.

[0005] In a second aspect, the present invention provides an intelligent pure electric vehicle braking energy recovery control system, including the following modules: Data acquisition module: Collect the driving data, traffic flow, and road condition congestion data of the vehicle during driving, and use a road condition recognition algorithm to judge the current road condition type; Data analysis module: Continuously record the driver operation data and vehicle state data by using the ECU, and obtain the driver's driving habit through the data analysis module; Kinetic energy recovery prediction module: Based on the judged road condition type and the analyzed driving habit data, use a neural network algorithm to construct a prediction model, and use data training to obtain a kinetic energy recovery prediction model; Recovery intensity adjustment module: In order to achieve a balance between driving comfort and energy recovery efficiency under different road condition types and driving habits, it is necessary to preset a strategy curve for the kinetic energy recovery intensity varying with the road condition type and driving habit, and use the kinetic energy recovery prediction model to dynamically adjust the kinetic energy recovery intensity; The specific process of dynamically adjusting the kinetic energy recovery intensity is as follows: Taking the average driving speed of the vehicle as an index to measure the road condition congestion degree, and at the same time considering the driver's aggressiveness score s, the kinetic energy recovery intensity F is expressed as a percentage; A strategy curve function is obtained by fitting experimental data. ; When the prediction model predicts that the vehicle is about to enter a congested section, and the average vehicle speed v within the next 1 minute is predicted to be lower than 30 km / h and the driver has an aggressive driving habit, assuming the aggressiveness score s = 8, the system starts to gradually reduce the kinetic energy recovery intensity in advance.

[0006] The beneficial effects of the present invention are as follows: 1. It can predict in advance the situation where the vehicle is about to enter a congested section and the driver has an aggressive driving habit, and gradually reduce the kinetic energy recovery intensity in advance, avoiding the jerks caused by excessive recovery force during frequent starts and stops, and greatly improving the riding experience during driving.

[0007] 2. By deeply analyzing the driver's operation data and vehicle status data, accurately identifying the types of driving habits, and establishing a prediction model in combination with road condition information, the kinetic energy recovery strategy can better adapt to the driving styles of different drivers, maximize the recovery of braking energy while ensuring comfort, and increase the vehicle's cruising range. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The present invention will be further described below with reference to the accompanying drawings.

[0009] Figure 1 is a flowchart of the steps of an intelligent pure electric vehicle braking energy recovery control method of the present invention; Figure 2 is a schematic diagram of the modules of an intelligent pure electric vehicle braking energy recovery control system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0010] In order to make the technical means, creative features, achieved purposes and effects of the present invention easy to understand, the present invention will be further described below in conjunction with specific embodiments.

[0011] Embodiment 1 As Figure 1 shown, an intelligent pure electric vehicle braking energy recovery control method according to an embodiment of the present invention includes: Step 1: Collect the driving data, traffic flow and road condition congestion data of the vehicle during driving, and use the road condition recognition algorithm to judge the current road condition type; During the driving of the vehicle, the camera captures the images in front of the vehicle in real time, including: traffic signs, lane lines; the millimeter wave radar monitors the distance and speed information of the vehicle in front in real time; the vehicle networking module obtains the latest traffic flow and road condition congestion data from the cloud server; The collected information is transmitted to the in-vehicle processor, and the current road condition type is analyzed and judged through the road condition recognition algorithm; Exemplarily, for example, if it is detected that the distance between the vehicles in front is small and the vehicle speed changes frequently, it is judged as a congested section in combination with the traffic flow data; Specifically, the millimeter wave radar can monitor the speed information of the vehicle in front in real time; it calculates the speed of the vehicle in front through the Doppler frequency shift method, and the formula is , where is the Doppler frequency shift, that is, the frequency difference between the transmitted wave and the reflected wave, is the radial speed of the target object relative to the radar, is the wavelength of the millimeter wave, is the frequency of the radar transmitted signal; By measuring the obtained , knowing the transmission frequency of the millimeter wave radar and the wavelength , the relative speed of the target object can be calculated ; The millimeter-wave radar can monitor the distance information of the vehicle ahead in real time. By using the principle of time delay, the distance of the vehicle ahead is calculated. The formula is , where R is the distance between the target object and the radar, and c is the speed of light. is the time delay between the transmitted signal and the received signal; Since the time delay between the transmitted and received signals of the millimeter-wave radar is proportional to the target distance, by measuring the precise , the distance R between the target object and the vehicle can be calculated; Based on the real-time front image captured by the camera, the millimeter-wave radar monitors the distance and speed of the vehicle ahead, and the vehicle networking module obtains traffic flow and road congestion data from the cloud; The driving data, traffic flow, and road congestion data information collected during driving are transmitted to the in-vehicle processor, and the road condition type is judged using the road condition recognition algorithm; Specifically, the convolutional neural network (CNN) algorithm is used to process the camera image to identify traffic signs, lane lines, and the distance and speed of the vehicle ahead, and a road condition recognition model is trained; Based on the road condition recognition model for feature extraction and classification, when judging the road condition according to the vehicle spacing and vehicle speed changes, simple logical judgment rules are used. Exemplarily, if the vehicle spacing is set to be less than a certain threshold and the standard deviation of the vehicle speed is greater than a certain value, and at the same time combined with traffic flow data, such as the number of vehicles passing through a certain section per unit time exceeds the set value, it is determined as a congested section or a non-congested section; Step 2: Use the ECU to continuously record the driver operation data and vehicle state data, and obtain the driver's driving habits through the data analysis module; The driving habit recognition and analysis step: Use the ECU to continuously record the driver operation data and vehicle state data, and regularly transmit the data to the data analysis module; The data analysis module uses machine learning algorithms to process and analyze the data to obtain the driver's driving habits; Exemplarily, for example, by analyzing the change frequency and amplitude of the accelerator pedal and brake pedal openings, it is judged whether the driver belongs to an aggressive, mild, or economical driving habit; Specifically, the ECU continuously records the driver operation data, including the accelerator pedal opening, brake pedal opening, gear shift operation, steering angle, and vehicle state data, including vehicle speed, acceleration, and deceleration, and regularly transmits them to the data analysis module; The clustering analysis algorithm is used to standardize the operation data and vehicle state data. The formula is , where x is the original data, μ is the data mean, and σ is the standard deviation. The data after normalization; map the operation data of different drivers to the same scale; then, use the K-Means algorithm to cluster the data into different categories, corresponding to aggressive, mild, and economical driving habits; Specifically, for example, for driver A, during a certain period of driving, the accelerator pedal opening data are successively [10%, 12%, 15%, 13%, 11%, 10%, 14%, 16%, 12%, 13%], the brake pedal opening data are [0%, 0%, 0%, 5%, 0%, 0%, 0%, 0%, 0%, 3%], and the vehicle speed data are [30 km / h, 32 km / h, 35 km / h, 33 km / h, 31 km / h, 30 km / h, 34 km / h, 36 km / h, 32 km / h, 33 km / h]; At the same time, record the gear shifting operation and the steering angle. Exemplarily, for example, during a turning process, the steering angle gradually increases from 0° to 30°, and then gradually returns to the original data; for acceleration and deceleration, they are calculated from the vehicle speed data; The ECU packs all the data recorded during this period and transmits it to the data analysis module through the in-vehicle network; the analysis module conducts data analysis. Taking the accelerator pedal opening data as an example, suppose the ECU records the mean μ and standard deviation σ of all the accelerator pedal opening data of 10 drivers in a week; through calculation, it is assumed that the mean μ = 15% and the standard deviation σ = 3% are calculated; For the first accelerator pedal opening data of driver A, which is 10%, the value after standardization is Perform such standardization processing on all the accelerator pedal opening data of driver A and the corresponding data of the other 9 drivers, so that the data of all drivers are on the same scale for subsequent analysis; Determine the number of clusters, divide the driving habits into 3 categories, namely aggressive, mild, and economical, so K = 3; initialize the cluster centers: randomly select 3 points from all the standardized accelerator pedal opening and brake pedal opening data points as the initial cluster centers c1, c2, c3; assume that the standardized value of the accelerator pedal opening corresponding to the initial cluster center c1 is -0.5, and the standardized value of the brake pedal opening is 0.2; the standardized value of the accelerator pedal opening corresponding to c2 is 1.2, and the standardized value of the brake pedal opening is -0.8; the standardized value of the accelerator pedal opening corresponding to c3 is 0.3, and the standardized value of the brake pedal opening is 0.1; Calculate the distance from the data point to the cluster center. For the first data point of driver A, that is, the standardized value of the accelerator pedal opening is -1.67 and the standardized value of the brake pedal opening is 0, calculate its distance to the three cluster centers; The distance to c1 ; The distance to c2 ; Distance to c3 ; Since the distance to c1 is the shortest, this data point is classified into the category where c1 is located; After traversing all the data points of all drivers and completing the classification, recalculate the cluster centers of each category; for example, for the category where c1 is located, assuming there are 2000 data points in this category, recalculate the average value of the normalized accelerator pedal opening and the average value of the normalized brake pedal opening to obtain the new cluster center; Repeat the above steps of calculating distances and updating cluster centers until the cluster centers no longer change significantly; If the normalized accelerator pedal opening values of the data points in a certain cluster are generally large and fluctuate frequently, and the normalized brake pedal opening values are also large and appear frequently, then the driving habit corresponding to this cluster is aggressive; If the normalized accelerator pedal opening values of the data points in a certain cluster are relatively stable and small, and the normalized brake pedal opening values are also small and appear less frequently, it corresponds to a gentle driving habit; For an economical driving habit, it is manifested as a moderate and stable accelerator pedal opening, and less use of the brake pedal, corresponding to the characteristics of the cluster where the cluster center c3 is located; Step 3: Based on the judged road condition type and the analyzed driving habit data, use the neural network algorithm to construct a prediction model, and use the data to train to obtain a kinetic energy recovery prediction model; Based on the collected road condition information and driving habit data, establish a prediction model; use the neural network algorithm to construct the model, with vehicle speed, road condition type, and driving habit type as the input layer, and vehicle speed change and braking demand in the future period of time as the output layer, and train the data to obtain a kinetic energy recovery prediction model. During the vehicle driving process, continuously input new actual driving data into the kinetic energy recovery prediction model, and adjust the model parameters through optimization algorithms such as the backpropagation algorithm to make the kinetic energy recovery prediction model more accurate; Specifically, use vehicle speed, road condition type, driving habit type, etc. as the input layer, and vehicle speed change and braking demand as the output, where the expected opening percentage of the brake pedal is represented; construct an MLP neural network with one hidden layer, and the hidden layer contains 5 neurons, and calculate the weighted sum of the hidden layer; Exemplarily, for the first neuron j = 1 in the hidden layer, there are three inputs in the input layer, which are x1 corresponding to the vehicle speed, x2 corresponding to the road condition type, that is, congested section or non-congested section, and x3 corresponding to the driving habit type; the corresponding output weights are w11, w12, w13; and the bias term b1; It should be noted that for the convenience of calculation and training, congested sections in the road condition types are marked as 1, and non-congested sections are marked as 0; in the driving habit types, the aggressive type is set to 0, the mild type is set to 1, and the economic type is set to 2; Exemplarily, assume that w11 = 0.3, w12 = -0.1, w13 = 0.2, b1 = 0.05. For a certain set of input data x1 = 50, that is, the vehicle speed is 50 km / h, x2 = 1, that is, a congested section, x3 = 1, that is, a mild driving habit, then the input z1 of this neuron is: ; where i = 1, 2, 3; Similarly, for other neurons in the hidden layer, weighted sum calculations are also performed according to this formula; The ReLU activation function f(z) = max(0, z) is adopted; for z1 = 15.15, after being processed by the ReLU activation function, f(z1) = max(0, 15.15) = 15.15; if the value of z is negative, such as z = -2, then f(z) = max(0, -2) = 0; the output of each neuron in the hidden layer after being processed by the ReLU activation function is used as the input of the next layer (output layer); Assume that the output layer has two neurons, corresponding to the vehicle speed change and braking demand within the next 10 seconds respectively; there are also weights and biases between the neurons in the output layer and the hidden layer; Exemplarily, for the first neuron in the output layer, that is, corresponding to the vehicle speed change prediction, the weights with the 5 neurons in the hidden layer are v11, v12, v13, v14, v15 respectively, and the bias term is e1. Assume that v11 = 0.2, v12 = 0.1, v13 = -0.05, v14 = 0.15, v15 = 0.08, e1 = -0.1, and the outputs of the 5 neurons in the hidden layer after being activated by ReLU are h1 = 15.15, h2 = 10.2, h3 = 0, h4 = 8.5, h5 = 12.3 respectively. Then the output of the first neuron in the output layer is: ; where j = 1, 2, 3, 4, 5; similarly, the output of the second neuron in the output layer can be calculated; The mean squared error loss function is used to optimize the model, and the formula is: , where m is the number of groups of training data. Assume that for a certain set of data, the true vehicle speed change within the next 10 seconds , the predicted value , the true braking demand , the predicted value , then the loss of this set of data is 0.0312625: Summing up the losses of all m groups of data and taking the average, the loss L of the entire training set is obtained; Next, use the backpropagation algorithm to update the parameters. Based on the principle of gradient descent, calculate the gradients of the loss function with respect to the weights and biases. After obtaining the gradients, update the weights and biases according to the gradient descent formula. Repeat the process of calculating the loss function and updating the parameters. After multiple rounds of iteration, the loss function gradually decreases, and the prediction accuracy of the kinetic energy recovery prediction model continuously improves; Step 4: Use the kinetic energy recovery prediction model to dynamically adjust the kinetic energy recovery intensity according to the pre-set strategy curve of the kinetic energy recovery intensity varying with road conditions and driving habits; Based on the kinetic energy recovery prediction model, when the model predicts that the vehicle is about to enter a congested section and the driver has an aggressive driving habit, gradually reduce the kinetic energy recovery intensity in advance to avoid jerks caused by excessive recovery force during frequent starts and stops of the vehicle; During the driving process of the vehicle, if the road conditions suddenly change from smooth to congested, the system adjusts the kinetic energy recovery strategy in real time and reduces the recovery intensity according to the pre-set strategy curve to ensure the balance between driving comfort and energy recovery efficiency; Specifically, in order to achieve the balance between driving comfort and energy recovery efficiency under different road conditions and driving habits, it is necessary to pre-set the strategy curve of the kinetic energy recovery intensity varying with road conditions and driving habits; Exemplarily, assume that we use the average driving speed of the vehicle as an indicator to measure the congestion degree of the road conditions, and at the same time consider the aggressive degree score s of the driver (the higher the score, the more aggressive the driving habit, with a range of 0 - 10); the kinetic energy recovery intensity F is expressed as a percentage (with a range of 0 - 100%); A strategy curve function is obtained by fitting experimental data, for example ; When the prediction model predicts that the vehicle is about to enter a congested section, for example, it is predicted that the average vehicle speed v will be lower than 30 km / h within the next 1 minute and the driver has an aggressive driving habit. Assuming the aggressive degree score s = 8, the system starts to gradually reduce the kinetic energy recovery intensity in advance; Assume that the current kinetic energy recovery intensity of the vehicle is , and the current vehicle speed ; According to the predicted congested road conditions and driving habits, we can calculate the target kinetic energy recovery intensity ; To avoid jerks caused by sudden changes in the recovery force, the system gradually reduces the kinetic energy recovery intensity at a certain time interval ; The reduction amplitude each time can be based on the remaining time, such as the predicted time to enter the congested section and the target reduction amplitude; During the vehicle's driving process, if the road condition suddenly changes from smooth to congested, the system needs to adjust the kinetic energy recovery strategy in real time. The system immediately calculates the adjustment amplitude according to the current kinetic energy recovery intensity and the new target kinetic energy recovery intensity. Similarly, to avoid jerks, the system reduces the kinetic energy recovery intensity at a certain adjustment rate; The technical solution of this embodiment is as follows: By means of a camera, a millimeter-wave radar and a vehicle networking module, data such as the front image, vehicle distance speed, traffic flow and road congestion are collected, and the vehicle-mounted processor uses a convolutional neural network and logical judgment rules to identify the road condition type; The ECU records the driver's operation and vehicle state data, and uses clustering analysis algorithms and K-Means algorithms to identify driving habits such as aggressive, mild, and economical; Taking vehicle speed, road condition type, and driving habit type as the input layer, and future vehicle speed changes and braking requirements as the output layer, an MLP neural network prediction model is constructed and optimized by the backpropagation algorithm to make the prediction more accurate; When it is predicted that the vehicle will enter a congested section and the driver has an aggressive driving habit, the kinetic energy recovery intensity is gradually reduced in advance at a specific time interval and amplitude; When the road condition suddenly changes during driving, the recovery intensity is adjusted in real time according to the preset strategy curve to ensure the balance between driving comfort and energy recovery efficiency, and improve the comprehensive performance of pure electric vehicles.

[0012] Embodiment 2 As Figure 2 described, based on Embodiment 1, the present invention also provides an intelligent pure electric vehicle braking energy recovery control system, including the following modules: Data acquisition module: Collect driving data, traffic flow and road congestion data during the vehicle's driving process, and use a road condition recognition algorithm to judge the current road condition type; During the vehicle's driving process, the camera takes real-time pictures of the front of the vehicle, including: traffic signs, lane lines; The millimeter-wave radar monitors the distance and speed information of the vehicle ahead in real time; The vehicle networking module obtains the latest traffic flow and road congestion data from the cloud server; The collected information is transmitted to the vehicle-mounted processor, and the current road condition type is analyzed and judged through a road condition recognition algorithm; Based on the real-time front image taken by the camera, the millimeter-wave radar monitors the distance and speed of the vehicle ahead, and the vehicle networking module obtains traffic flow and road congestion data from the cloud; The driving data, traffic flow and road congestion data information collected during the driving process are transmitted to the vehicle-mounted processor, and a road condition recognition algorithm is used to judge the road condition type; Specifically, the convolutional neural network (CNN) algorithm is used to process the camera image to identify traffic signs, lane lines and the distance and speed of the vehicle ahead, and a road condition recognition model is trained; Based on the road condition recognition model for feature extraction and classification, when judging the road condition according to the vehicle spacing and vehicle speed changes, simple logical judgment rules are used. Exemplarily, for example, it is set that the vehicle spacing is less than a certain threshold and the standard deviation of the vehicle speed is greater than a certain value. At the same time, combined with traffic flow data, such as the number of vehicles passing through a certain section per unit time exceeding the set value, it is determined as a congested section or a non-congested section; Data analysis module: The ECU continuously records the driver operation data and vehicle state data, and obtains the driver's driving habits by using the data analysis module; Steps for driving habit recognition and analysis: The ECU continuously records the driver operation data and vehicle state data, and regularly transmits the data to the data analysis module; The data analysis module uses machine learning algorithms to process and analyze the data to obtain the driver's driving habits; For example, by analyzing the change frequency and amplitude of the accelerator pedal and brake pedal opening degrees, it is judged whether the driver belongs to an aggressive, mild or economical driving habit; Specifically, the ECU continuously records the driver operation data, including the accelerator pedal opening degree, brake pedal opening degree, gear shift operation, steering angle, and vehicle state data, including vehicle speed, acceleration, deceleration, and regularly transmits them to the data analysis module; The clustering analysis algorithm is used to standardize the operation data and vehicle state data. The formula is , where x is the original data, μ is the data mean, σ is the standard deviation, is the normalized data; Map the operation data of different drivers to the same scale; Then, use the K-Means algorithm to cluster the data into different categories, corresponding to aggressive, mild, and economical driving habits; Specifically, for example, for driver A, during the driving process in a certain period of time, the accelerator pedal opening degree data are successively [10%, 12%, 15%, 13%, 11%, 10%, 14%, 16%, 12%, 13%], the brake pedal opening degree data are [0%, 0%, 0%, 5%, 0%, 0%, 0%, 0%, 0%, 3%], and the vehicle speed data are [30 km / h, 32 km / h, 35 km / h, 33 km / h, 31 km / h, 30 km / h, 34 km / h, 36 km / h, 32 km / h, 33 km / h]; At the same time, record the gear shift operation and steering angle. Exemplarily, for example, during a turning process, the steering angle gradually increases from 0° to 30°, and then gradually returns to the original data; For acceleration and deceleration, they are calculated from the vehicle speed data; The ECU packs all the data recorded during this period and transmits it to the data analysis module through the in-vehicle network; the analysis module conducts data analysis. Taking the accelerator pedal opening data as an example, suppose the ECU records the mean μ and standard deviation σ of all accelerator pedal opening data of 10 drivers in a week; through calculation, it is assumed that the calculated mean μ = 15% and standard deviation σ = 3%. For the first accelerator pedal opening data of 10% of driver A, the value after standardization is Perform such standardization on all the accelerator pedal opening data of driver A and the corresponding data of the other 9 drivers, so that the data of all drivers is on the same scale, facilitating subsequent analysis; Determine the number of clusters. Divide driving habits into 3 categories, namely aggressive, mild, and economical, so K = 3; Initialize the cluster centers: randomly select 3 points from all the standardized accelerator pedal opening and brake pedal opening data points as the initial cluster centers c1, c2, c3; Suppose the standardized value of the accelerator pedal opening corresponding to the initial cluster center c1 is -0.5, and the standardized value of the brake pedal opening is 0.2; The standardized value of the accelerator pedal opening corresponding to c2 is 1.2, and the standardized value of the brake pedal opening is -0.8; The standardized value of the accelerator pedal opening corresponding to c3 is 0.3, and the standardized value of the brake pedal opening is 0.1; Calculate the distance from the data point to the cluster center. For the first data point of driver A, that is, the standardized value of the accelerator pedal opening is -1.67 and the standardized value of the brake pedal opening is 0, calculate its distances to the three cluster centers; The distance to c1 ; The distance to c2 ; The distance to c3 ; Since the distance to c1 is the shortest, this data point is classified into the category where c1 is located; After traversing all the data points of all drivers and completing the classification, recalculate the cluster centers of each category; For example, for the category where c1 is located, suppose there are 2000 data points in this category, and recalculate the average value of the standardized accelerator pedal opening and the average value of the standardized brake pedal opening to obtain the new cluster center; Repeat the above steps of calculating distances and updating cluster centers until the cluster centers no longer change significantly; If the standardized values of the accelerator pedal opening in a certain cluster are generally large and fluctuate frequently, and the standardized values of the brake pedal opening are also large and appear frequently, then the driving habit corresponding to this cluster is aggressive; If the normalized value of the accelerator pedal opening of the data points in a certain cluster is relatively stable and small, and the normalized value of the brake pedal opening is also small and appears less frequently, it corresponds to a mild driving habit; For an economical driving habit, it is manifested that the accelerator pedal opening is moderate and changes smoothly, and the brake pedal opening is used less, corresponding to the characteristics of the cluster where the cluster center c3 is located; Kinetic energy recovery prediction module: Based on the judged road condition type and the analyzed driving habit data, use the neural network algorithm to build a prediction model, and use the data to train the kinetic energy recovery prediction model; Based on the collected road condition information and driving habit data, establish a prediction model; use the neural network algorithm to build the model, with vehicle speed, road condition type, and driving habit type as the input layer, and vehicle speed change and braking demand in the next period of time as the output layer. Through data training, obtain the kinetic energy recovery prediction model. During the vehicle driving process, continuously input new actual driving data into the kinetic energy recovery prediction model, and adjust the model parameters through optimization algorithms such as the backpropagation algorithm to make the kinetic energy recovery prediction model more accurate; Recovery force adjustment module: Use the kinetic energy recovery prediction model to dynamically adjust the kinetic energy recovery force according to the pre-set strategy curve of the kinetic energy recovery force changing with the road condition type and driving habit; Based on the kinetic energy recovery prediction model, when the model predicts that the vehicle is about to enter a congested section and the driver has an aggressive driving habit, gradually reduce the kinetic energy recovery force in advance to avoid jerks caused by excessive recovery force during frequent starts and stops of the vehicle; During the vehicle driving process, if the road condition suddenly changes from unobstructed to congested, the system adjusts the kinetic energy recovery strategy in real time, and reduces the recovery force according to the pre-set strategy curve to ensure the balance between driving comfort and energy recovery efficiency; Specifically, in order to achieve the balance between driving comfort and energy recovery efficiency under different road condition types and driving habits, it is necessary to pre-set the strategy curve of the kinetic energy recovery force changing with the road condition type and driving habit; Exemplarily, assume that we use the average driving speed of the vehicle as an index to measure the congestion degree of the road condition, and at the same time consider the aggressiveness score s of the driver (the higher the score, the more aggressive the driving habit, and the range is 0-10); the kinetic energy recovery force F is expressed as a percentage (the range is 0-100%); A strategy curve function is obtained by fitting experimental data, for example ; When the prediction model predicts that the vehicle is about to enter a congested section, for example, it is predicted that the average vehicle speed v will be lower than 30 km / h in the next 1 minute and the driver has an aggressive driving habit. Assuming the aggressiveness score s = 8, the system starts to gradually reduce the kinetic energy recovery force in advance; Assume that the current kinetic energy recovery force of the vehicle is , the current vehicle speed ; Based on the predicted traffic congestion and driving habits, we can calculate the target kinetic energy recovery intensity ; To avoid jerks caused by sudden changes in the recovery force, the system gradually reduces the kinetic energy recovery intensity at regular time intervals ; The reduction amplitude each time can be based on the remaining time, such as the predicted time to enter the congested section and the target reduction amplitude; During the vehicle driving process, if the road condition suddenly changes from unobstructed to congested, the system needs to adjust the kinetic energy recovery strategy in real time. The system immediately calculates the adjustment amplitude according to the current kinetic energy recovery intensity and the new target kinetic energy recovery intensity. Similarly, to avoid jerks, the system reduces the kinetic energy recovery intensity at a certain adjustment rate.

[0013] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent pure electric vehicle braking energy recovery control method, characterized in that: include: Step 1: Collect the driving data, traffic flow and traffic congestion data of the vehicle during driving, and use the road condition recognition algorithm to determine the current road condition type; Step 2: Use the ECU to continuously record the driver's operation data and vehicle status data, and obtain the driver's driving habits by using the data analysis module; Step 3: Based on the road condition type determined and the driving habit data analyzed, a prediction model is constructed using a neural network algorithm, and a kinetic energy recovery prediction model is obtained through data training; Step 4: In order to achieve a balance between driving comfort and energy recovery efficiency under different road conditions and driving habits, it is necessary to pre-set a strategy curve for how the kinetic energy recovery strength changes with road conditions and driving habits, and use a kinetic energy recovery prediction model to dynamically adjust the kinetic energy recovery strength; The specific process of dynamically adjusting the kinetic energy recovery intensity is as follows: The average speed of the vehicle is used as an indicator to measure the degree of traffic congestion, while the driver's aggressiveness score s is taken into account. The kinetic energy recovery strength F is expressed as a percentage; A strategy curve function is obtained by fitting the experimental data. ； When the prediction model predicts that the vehicle is about to enter a congested road section, and predicts that the average vehicle speed v in the next 1 minute will be lower than 30km / h and the driver has aggressive driving habits, assuming that the aggressiveness score s=8, the system begins to gradually reduce the kinetic energy recovery intensity in advance.

2. The intelligent pure electric vehicle braking energy recovery control method according to claim 1 is characterized in that: The specific process of using the road condition recognition algorithm to determine the current road condition type is as follows: Use the convolutional neural network algorithm to process camera images to identify traffic signs, lane lines, and the distance and speed of the vehicle ahead, and train a road condition recognition algorithm model; Feature extraction and classification are performed based on the road condition recognition algorithm model to determine whether the road is congested or not.

3. The intelligent pure electric vehicle braking energy recovery control method according to claim 1 is characterized in that: The recording of driver operation data specifically includes: Accelerator pedal opening, brake pedal opening, gear shifting operation, steering angle.

4. The intelligent pure electric vehicle braking energy recovery control method according to claim 1 is characterized in that: The recorded vehicle status data includes: Vehicle speed, acceleration, deceleration.

5. The intelligent pure electric vehicle braking energy recovery control method according to claim 1 is characterized in that: The specific process of obtaining the driver's driving habits by using the data analysis module is as follows: The data analysis module uses machine learning algorithms to analyze the frequency and amplitude of changes in the opening of the accelerator pedal and brake pedal to determine whether the driver has aggressive, gentle and economical driving habits.

6. The intelligent pure electric vehicle braking energy recovery control method according to claim 1 is characterized in that: The specific process of using the neural network algorithm to build a prediction model is as follows: A neural network algorithm is used to build a model, with vehicle speed, road condition type, and driving habit type as the input layer, and vehicle speed changes, braking requirements, etc. as the output layer. A kinetic energy recovery prediction model is obtained through data training.

7. The intelligent pure electric vehicle braking energy recovery control method according to claim 1 is characterized in that: The specific process of dynamically adjusting the kinetic energy recovery strength is as follows: During vehicle driving, if the road condition suddenly changes from smooth to congested, the system will adjust the kinetic energy recovery strategy in real time and reduce the recovery intensity according to the preset strategy curve.

8. The intelligent pure electric vehicle braking energy recovery control method according to claim 1 is characterized by: The specific process of dynamically adjusting the kinetic energy recovery strength is as follows: In order to avoid setbacks caused by changes in recovery force, the system gradually reduces the kinetic energy recovery force according to the remaining time.

Images