An intelligent control method for energy recovery intensity in coasting condition of new energy vehicles

Abstract

The present invention relates to an intelligent control method for the energy recovery intensity of a new energy vehicle in a coasting condition, comprising constructing a basic training set, establishing a self-learning model for real-time judgment of the recovery intensity, training the model, adjusting the deceleration as a dependent variable after the training set is updated, continuously iterating and optimizing to obtain a trained model, inputting the trained model with a working scene signal set, obtaining a coasting recovery intensity judgment condition, monitoring whether the driver additionally steps on the accelerator pedal and the brake pedal during the coasting cycle, and controlling the corresponding pedal opening, iterating and dynamically optimizing the model, and placing the vehicle in an optimal deceleration state during coasting. The control method of the present invention can dynamically calculate in real time the optimal energy recovery intensity if the driver releases the accelerator and before stepping on the brake; reduce driving fatigue and improve user experience by reducing the number of times the user operates the pedals; improve the energy recovery economy level, and avoid negative impacts on the user's driving experience.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy vehicles, and specifically relates to an intelligent control method for energy recovery intensity in a coasting condition of a new energy vehicle. Background Art

[0002] New energy vehicles can activate energy regeneration during braking and coasting, using the motor to generate electricity, converting some of the vehicle's kinetic energy into electricity for continued use and improving efficiency. During braking, the regeneration intensity can be adjusted in real time based on information such as the master cylinder pressure. However, during coasting, after the accelerator is released, the energy regeneration intensity is determined solely by the system's preset settings.

[0003] The drawback of this setting is that if the preset level is set too low, the energy recovered during coasting is low, resulting in poor fuel economy. If the preset level is set too high, it is more likely to cause deceleration beyond the driver's expectations, forcing the driver to hit the accelerator again to increase the speed. This not only poorly experiences a poor user experience, but also increases the number of energy conversions, which in turn reduces fuel economy. Some vehicles offer users multiple preset levels, but they lack real-time intelligent settings that adapt to road conditions and driver habits, ultimately requiring personal adaptation to the vehicle.

[0004] The prior art discloses a method for controlling the coasting energy recovery intensity of an electric vehicle, comprising: monitoring the accelerator pedal opening and vehicle speed information in real time while the vehicle is moving; if the accelerator pedal opening is lower than the accelerator pedal opening threshold and the vehicle speed is greater than or equal to the minimum coasting energy recovery speed, entering the coasting energy recovery stage; detecting the distance between the current vehicle and the vehicle in front, and obtaining the deceleration required for braking the current vehicle based on the distance; if the deceleration is greater than a preset deceleration threshold and less than or equal to the maximum deceleration allowed for coasting energy recovery, adjusting the coasting energy recovery required torque so that the deceleration provided by the adjusted coasting energy recovery required torque is the deceleration required for braking the current vehicle.

[0005] However, the document only discloses a method for controlling coasting energy recovery intensity. This method uses environmental perception data and real-time vehicle information to calculate the optimal coasting energy recovery according to a fixed formula. This control method does not integrate big data or artificial intelligence, nor does it incorporate real-time dynamic adjustments based on actual user driving habits. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent control method for the energy recovery intensity of new energy vehicles in the coasting condition, so as to solve the problem of relying on big data and artificial intelligence to empower the original vehicle control and improve the intelligence level of control.

[0007] The purpose of the present invention is achieved through the following technical solutions:

[0008] A method for intelligently controlling energy recovery intensity in a coasting condition of a new energy vehicle comprises the following steps:

[0009] A. Construct a basic training set on the cloud, which consists of a navigation map signal group and a vehicle signal group;

[0010] B. Establish a self-learning model for real-time judgment of recycling intensity

[0011] Based on the basic training set, a self-learning model for real-time judgment of recovery intensity is constructed, and the random forest regression algorithm is used to predict the deceleration a during sliding. c , applied to both the vehicle side and the cloud side;

[0012] C. Training the self-learning model for real-time judgment of recycling intensity

[0013] C1. After the self-learning model for obtaining real-time judgment of the initial recovery intensity is applied to the vehicle, the deceleration is adjusted based on the driver's feedback:

[0014] C2. Based on the real-time feedback from the driver during the current taxiing cycle, adjust the deceleration as the dependent variable after the training set is updated;

[0015] D. Apply the new training set generated in the previous step to the vehicle side and continuously iterate and optimize;

[0016] E. Repeat steps C and D. When the driver makes no or very few adjustments to the given coasting deceleration, the drive cycle will no longer be optimized. The new training set generated on the vehicle side will be uploaded to the cloud after user authorization to update the cloud-based basic training set.

[0017] F. The next driving cycle will continue to iterate based on the final optimization results of the previous driving cycle, obtaining a self-learning model that can determine the recovery intensity in real time after training.

[0018] G. Input the real-time working scene signal set into the trained model to calculate and generate the current taxiing recovery strength, wherein the working scene signal set is composed of a navigation map signal group and a vehicle signal group.

[0019] H. Obtain and implement coasting recovery intensity decisions, monitor whether the driver additionally presses the accelerator and brake pedals during the coasting cycle, and the corresponding pedal opening control results. Iterate the dynamic optimization model to ensure that the vehicle is in the optimal deceleration state in real time during the coasting process. The new training set generated on the vehicle side is uploaded to the cloud after user authorization to update the basic cloud training set.

[0020] Furthermore, in step A, the navigation map signal combination includes the intersection distance in the forward direction and the average slope 100m ahead.

[0021] Furthermore, in step A, the vehicle signal combination includes vehicle speed, accelerator pedal opening signal, brake pedal, front traffic light status recognition, front speed limit sign recognition, front traffic light distance recognition, front speed limit sign distance recognition, front vehicle distance recognition, and front vehicle distance change speed recognition.

[0022] Furthermore, in step A, the basic training set is sourced from two parts. One part is a preset training set developed by the OEM, which is designed by development engineers and collected during the testing process under ideal scenarios, comprehensively considering the control data of recycling efficiency and driving comfort; the other part is actual usage data authorized to be transmitted back by users.

[0023] Furthermore, in step B, the random forest is composed of multiple decision trees, and its prediction results depend on the results of each regression decision tree:

[0024]

[0025] Where: is the model prediction result; h(x,θ t ) is based on x and θ t The output of; x is the independent variable; θ t are independent and identically distributed random variables; T is the number of decision trees.

[0026] Furthermore, in step C1, the deceleration adjustment is performed according to the following formula:

[0027] a c =a0+c0·arctan(h(x acc ,x brk ))

[0028] Where a0 is the original deceleration; h(x acc ,x brk ) is the output based on braking / acceleration behavior; c0, h(x acc ,x brk ) is determined by road load measurement tests.

[0029] Furthermore, in step C2, the coasting period is 5 seconds after the start of coasting.

[0030] Furthermore, in step C2, the feedback behavior includes braking or acceleration behavior.

[0031] Furthermore, in step G, the navigation map signal combination includes the intersection distance in the forward direction and the average slope 100 meters ahead.

[0032] Furthermore, in step G, the vehicle signal combination includes vehicle speed, accelerator pedal opening signal, brake pedal, front traffic light status recognition, front speed limit sign recognition, front traffic light distance recognition, front speed limit sign distance recognition, front vehicle distance recognition, and front vehicle distance change speed recognition.

[0033] Compared with the prior art, the present invention has the following beneficial effects:

[0034] 1. The control method of the present invention uses artificial intelligence, combined with a large database generated by driving data sent back by users, to train an artificial intelligence model, dynamically calculating in real time the optimal energy recovery intensity when the driver releases the accelerator or brake pedal;

[0035] 2. The control method distributes vehicle kinetic energy during coasting conditions, converting it into the optimal choice between driving distance and electrical energy in real time. This improves economy while reducing driver fatigue and enhancing the user experience by reducing the number of pedal strokes.

[0036] 3. The control method includes the coasting energy recovery intensity control function and dynamic self-learning model training for real-time judgment of recovery intensity. Relying on big data and artificial intelligence, it empowers the original vehicle control, improves the intelligence level of control, and makes the coasting energy recovery function user experience better. It not only improves the energy recovery economy level, but also avoids negative impacts on the user's driving experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those skilled in the art, it is easy to understand without inventive means.

[0038] Figure 1 Flowchart of the intelligent control method for energy recovery intensity in coasting conditions of new energy vehicles;

[0039] Figure 2 Example coasting deceleration dynamic curve. DETAILED DESCRIPTION

[0040] The present invention will be further described below in conjunction with embodiment:

[0041] like Figure 1 As shown, the intelligent control method for energy recovery intensity in coasting conditions of new energy vehicles of the present invention iteratively optimizes the coasting recovery deceleration in coasting conditions for different vehicles based on the vehicle's driving status in coasting conditions, such as road conditions, vehicle speed, deceleration, and the braking / acceleration behavior within the next 5 seconds, so that the vehicle is in the optimal deceleration state when coasting and conforms to the driving habits of the vehicle driver.

[0042] A. Construct a cloud-based basic training set, which consists of a navigation map signal group and a vehicle signal group.

[0043] The navigation map signal combination includes the distance to the intersection in the forward direction and the average slope 100 meters ahead. The vehicle signal combination includes vehicle speed, accelerator pedal position signal, brake pedal, traffic light status recognition ahead, speed limit sign recognition ahead, traffic light distance recognition ahead, speed limit sign distance recognition ahead, vehicle distance recognition ahead, and vehicle distance change speed recognition ahead.

[0044] The basic training set is derived from two sources: one is a preset training set developed by the OEM, which is designed by development engineers and collected during the testing process under ideal scenarios, taking into account control data for recycling efficiency and driving comfort; the other is actual usage data authorized by users, as detailed in steps E and H.

[0045] B. Establish a self-learning model for real-time judgment of recycling intensity

[0046] Based on the basic training set, a self-learning model for real-time judgment of recovery intensity is constructed, and the random forest regression algorithm is used to predict the deceleration a during sliding. c , applied to both the vehicle side and the cloud side;

[0047] A random forest is composed of multiple decision trees, and its prediction results depend on the results of each regression decision tree:

[0048]

[0049] Where: is the model prediction result; h(x,θ t ) is based on x and θ t The output of; x is the independent variable; θ t are independent and identically distributed random variables; T is the number of decision trees.

[0050] C. Training the self-learning model for real-time judgment of recycling intensity

[0051] C1. After the self-learning model for obtaining real-time judgment of the initial recovery intensity is applied to the vehicle, the deceleration is adjusted according to the following formula based on the driver's feedback:

[0052] a c =a0+c0·arctan(h(x acc ,x brk ))

[0053] Where a0 is the original deceleration; h(x acc ,x brk ) is the output based on braking / acceleration behavior; c0, h(xacc ,x brk ) is determined by conducting road load measurement tests in accordance with Appendix CC of GB18352.6-2016.

[0054] C2. Based on the real-time feedback from the driver during the current taxiing cycle, such as braking / acceleration behavior, adjust the deceleration as the dependent variable after the training set is updated.

[0055] D. Apply the new training set generated in the previous step to the vehicle side and continuously iterate and optimize;

[0056] E. Repeat steps C and D. If the driver makes no or very few adjustments to the given coasting deceleration, the drive cycle will no longer be optimized. The new training set generated on the vehicle side is uploaded to the cloud after user authorization to update the cloud-based basic training set.

[0057] F. The next driving cycle will continue to iterate based on the final optimization results of the previous driving cycle, obtaining a self-learning model that can judge the recovery intensity in real time after training.

[0058] G. Input the real-time working scene signal set into the trained model to calculate and generate the current sliding recovery strength, wherein the working scene signal set is composed of a navigation map signal group and a vehicle signal group.

[0059] H. Decisions on coasting recovery intensity are implemented, monitoring whether the driver presses the accelerator and brake pedals within 5 seconds of coasting, as well as the corresponding pedal opening control results. The dynamic optimization model is iterated to ensure that the vehicle is in the optimal deceleration state in real time during the coasting process. The new training set generated on the vehicle side is uploaded to the cloud after user authorization to update the basic cloud training set.

[0060] From step A, step E, and step H, reducing user driving fatigue is reflected in the optimization goal of whether the user re-operates the accelerator / brake pedal within a short period of time after coasting. The optimal energy conversion strategy is reflected in the fact that, on the one hand, the basic training set has integrated the design engineers' consideration of vehicle recovery economy, and on the other hand, the user's reduced operation of the accelerator / brake pedal itself avoids the energy loss caused by frequent acceleration and deceleration. At the same time, the model training on the vehicle side can independently learn from each car owner's habits, avoiding the problem of inability to provide differentiated service to users due to unified factory settings. At the same time, the collection of user-authorized data can be fed back to update the cloud training set, so that the factory-set training model can adapt to as many users as possible, gradually improving the user's initial car purchase experience.

[0061] Example 1

[0062] A. Sample set acquisition

[0063] Based on the historical data of each vehicle within one year, we selected a dataset within 5 seconds after the start of each coasting as a sample data source. We then extracted the following signals within this sample time range for modeling:

[0064] 1. Map signal value, vehicle signal value, and deceleration value at the start of coasting;

[0065] 2. The following values ​​at the first braking / acceleration moment within 5 seconds: time from the start of coasting, distance traveled from the start of coasting;

[0066] 3. Pedal opening for the first braking / acceleration within 5 seconds;

[0067] 4. Duration of first braking / accelerating pedal depression within 5 seconds;

[0068] Based on each sample data source, sample data that can be used for modeling is calculated. Each vehicle has a personalized sample set to match the driving habits of different drivers.

[0069] B. Initial prediction: Build a model based on the original sample set and apply it to the vehicle side;

[0070] C. Dependent variable adjustment: Obtain the signal value under the coasting condition after the initial model is applied to the vehicle, and adjust the deceleration as the new dependent variable;

[0071] D. Model iteration: Apply the new training set generated in the previous step to the vehicle side and repeat steps C and D;

[0072] E. This drive cycle model iteration is complete: When the driver makes no or very few adjustments to the given coasting deceleration, this drive cycle will no longer be optimized;

[0073] F. Next driving cycle: Iteration will continue based on the final optimization results of the previous driving cycle.

[0074] Example 2

[0075] For example, when the red light is 300 meters ahead, the driver releases the accelerator pedal and does not step on the brake pedal. The starting speed is 54 km / h:

[0076] Scene 1: There is no car ahead. Press Figure 2 The medium deceleration curve 1 dynamically performs coasting energy recovery to ensure that the vehicle does not need to step on the accelerator to consume additional energy before reaching the intersection and stopping quickly, while at the same time recovering as much energy as possible into the battery.

[0077] In the second scenario, there is a car 50m ahead, and the speed is 36km / h. The distance between the two cars is gradually decreasing. Figure 2Medium deceleration curve 2 is executed, initially using a higher regenerative deceleration rate to reduce speed, ensuring the distance to the vehicle ahead is no longer reduced within a safe distance. Then, a lower regenerative deceleration rate is used to stop at the intersection. This process takes a long time to avoid the vehicle ahead, but appropriate coasting regenerative braking reduces mechanical braking intervention and maximizes energy recovery.

[0078] If the driver steps on the brake or accelerator pedal during the first scenario, the vehicle will automatically record this. This intelligent control method for energy recovery intensity during coasting conditions in new energy vehicles is applicable to controlling energy recovery intensity during coasting conditions in electric or hybrid vehicles. This method combines big data and artificial intelligence with actual user driving habits to dynamically adjust model training in real time.

[0079] Note that the above are only preferred embodiments of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and that various obvious changes, readjustments, and substitutions can be made by those skilled in the art without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments and may include many other equivalent embodiments without departing from the concept of the present invention. The scope of the present invention is determined by the scope of the appended claims.

Claims

1. An intelligent control method for energy recovery intensity in coasting condition of new energy vehicles, characterized in that: The following steps are involved: A. Construct a basic training set on the cloud, which consists of a navigation map signal group and a vehicle signal group; B. Establish a self-learning model for real-time judgment of recycling intensity Based on the basic training set, a self-learning model for real-time judgment of recovery intensity is constructed, and the random forest regression algorithm is used to predict the deceleration a during sliding. c , applied to both the vehicle side and the cloud side; C. Training the self-learning model for real-time judgment of recycling intensity C1. After the self-learning model for obtaining real-time judgment of the initial recovery intensity is applied to the vehicle, the deceleration is adjusted based on the driver's feedback: C2. Based on the real-time feedback from the driver during the current taxiing cycle, adjust the deceleration as the dependent variable after the training set is updated; D. Apply the new training set generated in the previous step to the vehicle side and continuously iterate and optimize; E. Repeat steps C and D. When the driver no longer makes adjustments to the given coasting deceleration, the driving cycle will no longer be optimized. The new training set generated on the vehicle side will be uploaded to the cloud after user authorization to update the cloud-based basic training set. F. The next driving cycle will continue to iterate based on the final optimization results of the previous driving cycle, obtaining a self-learning model that can determine the recovery intensity in real time after training. G. Inputting a real-time working scene signal set into the trained model to calculate and generate the current taxiing recovery strength, wherein the working scene signal set is composed of a navigation map signal group and a vehicle signal group; H. Obtain and implement coasting recovery intensity decisions, monitor whether the driver additionally presses the accelerator and brake pedals during the coasting cycle, and the corresponding pedal opening control results. Iterate the dynamic optimization model to ensure that the vehicle is in the optimal deceleration state in real time during the coasting process. The new training set generated on the vehicle side is uploaded to the cloud after user authorization to update the basic cloud training set.

2. The intelligent control method for energy recovery intensity in coasting condition of a new energy vehicle according to claim 1, characterized in that: Step A: The navigation map signal combination includes the intersection distance in the forward direction and the average slope of the next 100 meters.

3. The intelligent control method for energy recovery intensity in coasting condition of a new energy vehicle according to claim 1, characterized in that: In step A, the vehicle signal combination includes vehicle speed, accelerator pedal opening signal, brake pedal opening signal, traffic light status recognition ahead, speed limit sign recognition ahead, traffic light distance recognition ahead, speed limit sign distance recognition ahead, vehicle distance recognition ahead, and vehicle distance change speed recognition ahead.

4. The intelligent control method for energy recovery intensity in coasting condition of a new energy vehicle according to claim 1 is characterized by: In step A, the basic training set is derived from two sources: one is a preset training set developed by the OEM, which is designed by development engineers and collected during the testing process under ideal scenarios, taking into account control data for recycling efficiency and driving comfort; the other is actual usage data authorized by users.

5. The intelligent control method for energy recovery intensity in coasting condition of a new energy vehicle according to claim 1, characterized in that: Step B: Random Forest consists of multiple decision trees, and its prediction results depend on the results of each regression decision tree: Where: is the model prediction result; h(x,θ t ) is based on x and θ t The output of; x is the independent variable; θ t are independent and identically distributed random variables; T is the number of decision trees.

6. The intelligent control method for energy recovery intensity in coasting condition of a new energy vehicle according to claim 1, characterized in that: In step C1, the deceleration is adjusted according to the following formula: a c =a0+c0·arctan(h(x acc ,x brk )) Where a0 is the original deceleration; h(x acc ,x brk ) is the output based on braking / acceleration behavior; c0, h(x acc ,x brk ) is determined by road load measurement tests.

7. The intelligent control method for energy recovery intensity in coasting condition of a new energy vehicle according to claim 1 is characterized by: Step C2: The coasting period is 5 seconds after the start of coasting.

8. The intelligent control method for energy recovery intensity in coasting condition of a new energy vehicle according to claim 1 is characterized by: In step C2, the feedback behavior includes braking or accelerating behavior.

9. The intelligent control method for energy recovery intensity in coasting condition of a new energy vehicle according to claim 1, characterized in that: Step G: The navigation map signal combination includes the intersection distance in the forward direction and the average slope of the next 100 meters.

10. The intelligent control method for energy recovery intensity in coasting condition of a new energy vehicle according to claim 1, characterized in that: Step G, the vehicle signal combination includes vehicle speed, accelerator pedal opening signal, brake pedal, front traffic light status recognition, front speed limit sign recognition, front traffic light distance recognition, front speed limit sign distance recognition, front vehicle distance recognition and front vehicle distance change speed recognition.

Images