Energy management method and system for hybrid electric vehicle model predictive control

Through multi-feature deep neural network and real-time error adjustment mechanism, combined with networking technology, the battery SOC reference trajectory is designed, and the problem of insufficient adaptability of hybrid vehicle energy management systems in prediction errors and driving styles is achieved, and more efficient energy management and Fuel economy.

CN118182439BActive Publication Date: 2024-11-26WUHAN UNIV OF SCI & TECH
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410515320.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-04-26
Publication Date
2024-11-26
Estimated Expiration
2044-04-26

AI Technical Summary

Technical Problem

The energy management system of existing hybrid vehicles has insufficient prediction errors and driving style adaptability, resulting in unreasonable power distribution and low fuel economy.

Method used

The multi-feature deep neural network prediction speed model is adopted, and the battery SOC reference trajectory is designed in combination with real-time prediction error adjustment mechanism and networking technology to optimize the power distribution of power sources and improve the system's adaptability and accuracy.

Benefits of technology

It realizes more efficient energy management, improves fuel economy and system adaptability to multiple operating conditions, keeps the motor and engine in high-efficiency areas at all times, and reduces energy waste and environmental pollution.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118182439B_ABST
    Figure CN118182439B_ABST
Patent Text Reader

Abstract

The present invention belongs to the technical field of new energy vehicles, and in particular, relates to an energy management method and system for model predictive control of hybrid electric vehicles, including: S1, training and establishing a neural network speed prediction model; S2, constructing a prediction error adjustment mechanism; S3, designing a battery SOC reference trajectory. The present invention adopts a multi-feature deep neural network as a speed prediction model, and uses the root mean square error between the predicted speed and the actual speed as an exponential decay factor to construct the weight of the instantaneous cost, while updating the battery SOC reference trajectory to improve the adaptability and accuracy of power distribution, so that the motor and engine are always in a high efficiency area, and improve fuel economy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy vehicles, and in particular relates to an energy management method and system for model predictive control of a hybrid electric vehicle. Background Art

[0002] In recent years, as the contradiction between energy shortage and environmental pollution has become increasingly prominent, countries have introduced timetables for banning the sale of traditional fuel vehicles, which has prompted the automotive industry to accelerate the pace of developing new energy vehicle technology. As the most mature new energy vehicle at this stage, hybrid vehicles can reduce the use of oil and ensure long driving range compared to traditional fuel vehicles. Plug-in Hybrid Electric Vehicle (PHEV) is equipped with a large-capacity battery and can directly obtain electricity from the power grid through an external interface, which can have a longer pure electric driving range and less internal combustion engine working time.

[0003] Since plug-in hybrid electric vehicles have dual power sources, how to reasonably allocate the output power of each power source to improve energy efficiency and reduce fuel consumption has become a hot topic for many scholars. Among them, the energy management strategy based on model predictive control (MPC) has good development potential. Model predictive control requires predicting the driving conditions in the short time domain as the input of the power model, solving the optimal power distribution control sequence through the optimization algorithm, and then inputting the first control variable into the power model as the real control quantity. At present, the intelligent transportation system is in the development stage, and the prediction of future working conditions is often realized by using prediction models. The traditional prediction model only uses the historical speed sequence as the prediction model input, and the factors that interfere with the vehicle's driving speed during driving are often related to the vehicle's driving environment and driver style. Therefore, it is necessary to consider the driving style, the distance and speed of the vehicle's surrounding environment characteristics as the prediction model input to improve the prediction accuracy. Although conventional road conditions can achieve good prediction results through multivariate feature prediction models, the randomness of road condition changes, such as temporary road repairs and other occasional situations, will greatly reduce the prediction accuracy of the prediction model. In order to reduce the situation where the prediction model error is too large and leads to unreasonable power distribution, it is considered to correct the weight of the instantaneous cost of each prediction step in the cost function in real time according to the prediction speed error, reduce the impact of the prediction error on the MPC optimization strategy, and improve the optimization ability and adaptability of the strategy. In the MPC strategy, the state quantity needs to be reasonably constrained to achieve global fuel economy, so the battery state of charge (SOC) reference trajectory needs to be designed as a constraint. At present, the design of SOC reference trajectory has problems such as ignoring driving style and adaptability to different road conditions and low accuracy. Therefore, it is necessary to design a new battery SOC reference trajectory design method to further improve the optimization and adaptability of MPC.

[0004] Through the above analysis, the existing technology has problems and defects such as poor adaptability and poor optimization. The present invention uses methods such as multi-feature prediction model, real-time prediction correction and SOC reference trajectory planning considering driving style to achieve reasonable distribution of output power of each power source to improve energy efficiency while enhancing the adaptability of the optimization strategy to different working conditions. Summary of the invention

[0005] In response to the problems existing in the prior art, the present invention provides an energy management method and system for hybrid vehicle model predictive control, which adopts a multi-feature deep neural network as a speed prediction model, and uses the root mean square error between the predicted speed and the actual speed as an exponential decay factor to construct the weight of the instantaneous cost. At the same time, the battery SOC reference trajectory is updated to improve the adaptability and accuracy of power distribution, so that the motor and engine are always in a high efficiency area, thereby improving fuel economy.

[0006] The present invention is implemented as follows: a method for energy management of a hybrid electric vehicle model predictive control, the method comprising:

[0007] S1, training and establishing a neural network speed prediction model;

[0008] S2, construct a prediction error adjustment mechanism;

[0009] S3, design the battery SOC reference trajectory.

[0010] Furthermore, S1 specifically includes: using radar, global positioning system (GPS), and vehicle speed sensor to collect historical vehicle operating data and vehicle surrounding environment information, establish a multi-dimensional feature database and discretize the data; construct a deep neural network model, determine the number of hidden layers, the number of neurons and the activation function, take the distance and speed of the preceding vehicle, driving style, the current and previous status of the vehicle, road slope and climate conditions as input features, and the future short-term speed sequence as the target output:

[0011]

[0012]

[0013] In the above formula, D pre is the distance between the front vehicle and the vehicle, V pre is the speed of the vehicle in front, Sty is the driving style of the driver, and Wea is the weather and light conditions when the vehicle is driving, which are eight combinations of [rainy, sunny, foggy, snowy] and [day, night], This car went past kH p The driving speed at time k is, It is the future k+1 to k+H p The road slope of the driving section at any time can be obtained through the intelligent transportation system;

[0014] Train and verify the prediction accuracy of the deep neural network model, establish a deep neural network operating condition prediction model driven by multivariate data, and realize short-term accurate operating condition prediction of online model predictive control.

[0015] Further, S2 specifically includes: calculating the historical cumulative prediction error V of this trip by collecting the predicted speed and the actual driving speed of the trip up to the current time K RMSE (k), prediction error threshold V based on expert experience Err , by comparing the error with the threshold exceeding degree to design the exponential decay factor, that is, when V RMSE (k)>V Err hour:

[0016]

[0017] Thus, the weight of the cost function at different times in the time domain is predicted:

[0018]

[0019] Where γ is the coefficient. RMSE (k)≤V Err When , the cost function weights at different times are:

[0020] α(k+i)=1

[0021] Prediction time domain H p The total cost function is:

[0022]

[0023] The cost function coefficient is designed according to the prediction error to predict the instantaneous cost function J(k+i) at different times in the time domain, so as to adaptively reduce the impact of the prediction error on power allocation.

[0024] Furthermore, S3 specifically includes: collecting the driver's historical working conditions and road condition information through the PHEV on-board chip, extracting the impact degree data of the last route driving condition, using the impact degree root mean square to represent the driving style of the condition, using network technology to obtain real-time future driving section condition data, selecting the condition data closest to the driving style of the vehicle, splicing all the road section conditions of the route, and if there is an abnormal acceleration (greater than the maximum physical acceleration of the vehicle) at the splicing point, linear interpolation is used for processing. Complete the route condition prediction based on driving style, use the Pontryagin's Minimum Principle (PMP) algorithm to calculate the battery SOC trajectory based on the global prediction condition, so as to obtain the optimal reference trajectory of the battery SOC, and use the prediction time domain terminal state quantity constraint for real-time MPC optimization control to achieve a quasi-global optimal energy management strategy.

[0025] Furthermore, the control process of the energy management method of the hybrid vehicle model predictive control specifically includes:

[0026] Step 1: Before the trip begins, obtain the driving route according to the set vehicle navigation destination;

[0027] Step 2: Collect the driver's historical driving conditions and road condition information through the on-board chip, obtain the impact data of the driving conditions on the last route and determine his driving style (aggressive, normal, calm);

[0028] Step 3: Use network technology to obtain real-time future driving section working condition data and select the working condition data that is most similar to the driving style of the vehicle;

[0029] Step 4: Select the working condition data that is most similar to the driving style of the vehicle, and splice the working conditions of all sections of the route. If there is an abnormal acceleration (greater than the maximum physical acceleration of the vehicle) at the splicing point, use linear interpolation to process and obtain the route prediction working condition based on the driving style;

[0030] Step 5: Use the PMP algorithm to calculate the battery SOC trajectory based on the global prediction condition to obtain the optimal reference trajectory of the battery SOC;

[0031] Step 6: At the time k of operation, use the on-board sensors to collect data about the vehicle and the surrounding driving environment, such as the past kH p Data such as speed, acceleration, distance and speed of the vehicle ahead, driving style, road slope and climate up to the present moment;

[0032] Step 7: Use the above real-time data as the input of the deep neural network prediction model to predict the future k: k+H p The speed sequence of moments;

[0033] Step 8: By collecting the predicted speed and actual speed of the trip up to the current time K, calculate the historical cumulative prediction error V of this trip RMSE (k);

[0034] Step 9: Calculate the exponential decay factor by comparing the error with the threshold exceeding degree, threshold V Err It is usually 2.5km / h, thus obtaining the weights of the cost function at different times in the prediction time domain;

[0035] Step 10: Design the cost function coefficient according to the prediction error to predict the instantaneous cost function J(k+i) at different times in the time domain;

[0036] Step 11: K+H p The battery SOC reference trajectory corresponding to the moment is taken as The optimization objective constraint is used to control the battery SOC drop rate and complete the cost function configuration in the prediction time domain;

[0037] Step 12: Use the PMP algorithm to complete the calculation of the minimum Hamiltonian function and meet the exit condition;

[0038] Step 13: Complete the calculation of the cost function in the minimization prediction time domain and obtain the optimal control sequence;

[0039] Step 14: According to the model predictive control principle, the control quantity u(k) of the first step of the optimal control sequence is used as the actual output of the controller;

[0040] Step 15: Complete the vehicle state update at time k by executing control;

[0041] Step 16: When there is a large difference between the future road section operating condition data obtained at time k and at time k+1, execute steps 2 to 5 to complete the battery SOC reference trajectory update;

[0042] Step 17: Enter time k+1 and repeat steps 6 to 16 until the entire journey is completed, thus realizing the hybrid vehicle model predictive control energy management with real-time error correction.

[0043] Another object of the present invention is to provide a hybrid vehicle model predictive control energy management system that implements the hybrid vehicle model predictive control energy management method, comprising:

[0044] Prediction model building module, used to train and build a neural network speed prediction model;

[0045] An adjustment mechanism building module, used to build a forecast error adjustment mechanism;

[0046] The reference trajectory design module is used to design the battery SOC reference trajectory.

[0047] Another object of the present invention is to provide a computer device, the computer device includes a memory and a processor, the memory stores a computer program, when the computer program is executed by the processor, the processor executes the steps of the energy management method of the hybrid vehicle model predictive control.

[0048] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the energy management method of the hybrid vehicle model predictive control.

[0049] Another object of the present invention is to provide an information data processing terminal, which is used to implement the energy management system of the hybrid vehicle model predictive control.

[0050] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0051] First, the present invention adopts a multi-feature deep neural network as a speed prediction model, and uses multivariate data as the prediction model input to realize short-term accurate working condition prediction of online model predictive control, and sets up a prediction failure compensation mechanism. When the occasional prediction error is too large, the root mean square error between the predicted speed and the actual speed is used as an exponential decay factor to construct the weight of the instantaneous cost, thereby reducing the impact of the prediction error on the MPC optimization capability. At the same time, the network technology is used to obtain the characteristics of future driving conditions to realize the generation of the optimal reference trajectory of the battery SOC based on the network technology, and the battery SOC reference trajectory is regularly updated as a prediction time domain state quantity constraint to improve the adaptability and accuracy of power distribution, so that the motor and engine are always in a high efficiency area, thereby improving fuel economy.

[0052] The present invention is to develop an energy management method and system for hybrid electric vehicle model predictive control, which can solve the problem that the error of the speed prediction model in a real driving environment is difficult to correct in real time, reduce the influence of the prediction error on the MPC optimization ability, and improve the adaptability of its optimization method to various working conditions. A new method for generating the optimal reference trajectory of the battery SOC is designed using network technology, and the reference trajectory of the driving style is considered as a state quantity constraint, which can greatly improve the adaptability of MPC to the driver and achieve online global energy management optimization with better fuel economy and environmental adaptability.

[0053] The technical solution of the present invention overcomes technical prejudice: the present invention provides an energy management method and system for hybrid vehicle model predictive control, which can solve the problem that the MPC energy management strategy is difficult to correct prediction errors in real time, and complete energy distribution management in multiple environments in real time and effectively, so that hybrid vehicles can operate at optimal operating efficiency.

[0054] Second, the energy management method of hybrid vehicle model predictive control proposed in the present invention aims to solve several key problems in the prior art and achieve significant technical progress. The following are the main technical problems solved by the present invention and the technical progress it brings:

[0055] Technical issues solved:

[0056] 1. Inaccurate speed prediction: Existing energy management systems are usually based on static or simple dynamic models and cannot accurately predict vehicle speed changes in complex traffic environments. This invention improves the accuracy of speed prediction by establishing a deep neural network speed prediction model, combining multi-dimensional driving data (such as radar, GPS, vehicle speed sensor data) and environmental information (such as weather and road conditions).

[0057] 2. Low energy allocation efficiency: Traditional energy management strategies fail to fully consider real-time driving status and environmental changes, resulting in low efficiency of battery energy utilization. This invention optimizes the battery's charging and discharging behavior and improves energy efficiency by designing a battery SOC reference trajectory and a real-time adjustment mechanism.

[0058] 3. Weak ability to deal with prediction errors: In the past, systems often lacked effective adjustment mechanisms when faced with prediction errors. The present invention introduces a prediction error adjustment mechanism that dynamically adjusts the cost function weight by calculating historical cumulative prediction errors and designing exponential decay factors, effectively reducing the impact of prediction errors on system performance.

[0059] Significant technological advances achieved:

[0060] 1. Improve the accuracy and responsiveness of energy management: Through real-time and accurate speed prediction model and its close integration with energy management system, the present invention provides a more efficient energy allocation strategy, significantly improving fuel economy and battery life.

[0061] 2. Enhanced system adaptability: Hybrid vehicles can adapt their energy management strategies according to different driving styles and road conditions, so that the system can maintain optimal performance in changing environments.

[0062] 3. Optimized user experience and environmental impact: By reducing energy waste and improving operating efficiency, the present invention not only improves the user's driving experience, but also helps to reduce the impact of automobiles on the environment, such as reducing emissions.

[0063] 4. Technological innovation and foresight: The deep learning technology and model predictive control strategy adopted in this invention have demonstrated innovation in the field of energy management of hybrid vehicles, laying a technical foundation for the development of future smart cars.

[0064] The present invention not only effectively solves the problems existing in traditional energy management systems, but also achieves significant progress in energy efficiency, system adaptability and environmental sustainability through technological innovation. These technological advances will promote the competitiveness and popularity of hybrid vehicles in the market and promote wider technological applications and development in the future. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] 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 of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0066] Figure 1 It is a schematic diagram of the structure of a deep neural network operating condition prediction model based on multivariate data drive provided in an embodiment of the present invention;

[0067] Figure 2 It is a schematic diagram of a method for generating an optimal reference trajectory of battery SOC based on network connection technology provided by an embodiment of the present invention;

[0068] Figure 3 is a flow chart of a real-time error correction hybrid vehicle model predictive control method provided by an embodiment of the present invention;

[0069] Figure 4 It is a structural diagram of an energy management system for a hybrid vehicle model predictive control provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0070] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0071] In view of the problems existing in the prior art, the present invention provides an energy management method and system for hybrid electric vehicle model predictive control. The present invention is described in detail below with reference to the accompanying drawings.

[0072] The following are two examples of energy management methods for hybrid electric vehicle model predictive control:

[0073] Embodiment 1

[0074] An automobile manufacturing company has developed a hybrid vehicle energy management system based on model predictive control. The system first collects a large amount of real vehicle data and uses machine learning technology to establish a neural network speed prediction model. The model can predict the future speed of the vehicle in real time and combine the vehicle's current status information, such as battery power and engine status, to predict energy demand.

[0075] Based on the predictions, the system calculates the best energy management strategy through an optimization algorithm. This includes determining the best power distribution between the engine and the motor, as well as the battery charging and discharging strategy. Through these optimization measures, the system aims to reduce fuel consumption, reduce emissions, and improve driving comfort.

[0076] In actual application, the system is applied to a hybrid SUV model. Through verification in actual road tests, the system can significantly improve the vehicle's fuel economy, reduce emissions, and maintain stable power output under various driving conditions.

[0077] Embodiment 2

[0078] Another automotive technology company has developed a hybrid bus energy management system based on model predictive control. The system uses advanced predictive models and combines bus operation characteristics and route information to accurately predict the vehicle's energy demand.

[0079] The system collects historical bus operation data and combines it with real-time traffic information to establish a high-precision speed prediction model, which can predict the bus's driving speed in the future and thus predict its energy demand.

[0080] In terms of energy management strategy, the system adjusts the power distribution of the engine and motor in real time according to the prediction results, and optimizes the battery charging and discharging strategy. At the same time, the system also takes into account the energy requirements of buses under different driving conditions such as urban congestion and high-speed cruising to achieve global optimal energy management. In actual applications, the energy management system is applied to urban bus routes. Through actual application verification, the system can significantly reduce the fuel consumption of buses, reduce operating costs, and improve the riding experience of passengers.

[0081] These two examples demonstrate the application of energy management methods of hybrid vehicle model predictive control in different vehicle models and different application scenarios. Through accurate prediction and optimized energy management strategies, these methods can effectively improve vehicle fuel economy, reduce emissions, and enhance driving comfort and passenger riding experience.

[0082] The specific implementation scheme of the embodiment of the present invention is as follows:

[0083] 1. Training and establishing a neural network speed prediction model

[0084] The historical working condition data of the real vehicle and the surrounding environment information of the vehicle are collected by radar, GPS, speed sensor, etc., a multi-dimensional feature database is established and the data is discretized. A deep neural network model is constructed, and parameters such as the number of hidden layers, number of neurons and activation function are determined. The distance and speed of the preceding vehicle, driving style, current and previous status of the vehicle, road slope and climate conditions are used as input features, and the speed sequence in the short time domain in the future is used as the target output:

[0085]

[0086]

[0087] like Figure 1 The figure shows the structure of a deep neural network operating condition prediction model driven by multivariate data. The prediction accuracy of the deep neural network model is trained and verified, and a deep neural network operating condition prediction model driven by multivariate data is established to achieve short-term accurate operating condition prediction of online model predictive control.

[0088] 2. Construct a forecast error adjustment mechanism

[0089] By collecting the predicted speed and actual speed of the trip up to the current time K, the historical cumulative prediction error V of this trip is calculated. RMSE (k), prediction error threshold V based on expert experience Err , by comparing the error with the threshold exceeding degree to design the exponential decay factor, that is, when V RMSE (k)>V Err hour:

[0090]

[0091] Thus, the weight of the cost function at different times in the time domain is predicted:

[0092]

[0093] When V RMSE (k)≤V Err When , the cost function weights at different times are:

[0094] α(k+i)=1

[0095] Prediction time domain H p The total cost function is:

[0096]

[0097] The cost function coefficient is designed according to the prediction error to predict the instantaneous cost function J(k+i) at different times in the time domain, so as to adaptively reduce the impact of the prediction error on power allocation.

[0098] 3. Design battery SOC reference trajectory

[0099] The driver's historical working conditions and road condition information are collected through the PHEV on-board chip, and the impact degree data of the last route driving condition is extracted. The impact degree root mean square is used to represent the driving style of the condition. The real-time future driving section condition data is obtained using network technology. The condition data closest to the driving style of the vehicle is selected, and all the road section conditions of the route are spliced. If there is an abnormal acceleration (greater than the maximum physical acceleration of the vehicle) at the splicing point, linear interpolation is used for processing. The route condition prediction based on driving style is completed, and the battery SOC trajectory based on the global prediction condition is calculated using the Pontryagin minimum principle) algorithm to obtain the optimal reference trajectory of the battery SOC, which is used as a constraint on the prediction time domain terminal state quantity for real-time MPC optimization control to achieve a quasi-global optimal energy management strategy.

[0100] The present invention is to develop an energy management method for hybrid vehicle model predictive control with real-time error correction, which corrects the error of the speed prediction model in a real driving environment in real time to achieve optimal distribution of engine and motor torque. Figure 3 This is the control process of the present invention, and the specific steps are as follows:

[0101] Step 1: Before the trip begins, obtain the driving route according to the set vehicle navigation destination;

[0102] Step 2: Collect the driver's historical driving conditions and road condition information through the on-board chip, obtain the impact data of the driving conditions on the last route and determine his driving style (aggressive, normal, calm);

[0103] Step 3: Use network technology to obtain real-time future driving section working condition data and select the working condition data that is most similar to the driving style of the vehicle;

[0104] Step 4: Select the working condition data that is most similar to the driving style of the vehicle, and splice the working conditions of all sections of the route. If there is an abnormal acceleration (greater than the maximum physical acceleration of the vehicle) at the splicing point, use linear interpolation to process and obtain the route prediction working condition based on the driving style;

[0105] Step 5: Use the PMP algorithm to calculate the battery SOC trajectory based on the global prediction condition to obtain the optimal reference trajectory of the battery SOC;

[0106] Step 6: At the time k of operation, use the on-board sensors to collect data about the vehicle and the surrounding driving environment, such as the past kH pData such as speed, acceleration, distance and speed of the vehicle ahead, driving style, road slope and climate up to the present moment;

[0107] Step 7: Use the above real-time data as the input of the deep neural network prediction model to predict the future k: k+H p The speed sequence of moments;

[0108] Step 8: By collecting the predicted speed and actual speed of the trip up to the current time K, calculate the historical cumulative prediction error V of this trip RMSE (k);

[0109] Step 9: Calculate the exponential decay factor by comparing the error with the threshold exceeding degree, threshold V Err It is usually 2.5km / h, thus obtaining the weights of the cost function at different times in the prediction time domain;

[0110] Step 10: Design the cost function coefficient according to the prediction error to predict the instantaneous cost function J(k+i) at different times in the time domain;

[0111] Step 11: K+H p The battery SOC reference trajectory corresponding to the moment is taken as The optimization objective constraint is used to control the battery SOC drop rate and complete the cost function configuration in the prediction time domain;

[0112] Step 12: Use the PMP algorithm to complete the calculation of the minimum Hamiltonian function and meet the exit condition;

[0113] Step 13: Complete the calculation of the cost function in the minimization prediction time domain and obtain the optimal control sequence;

[0114] Step 14: According to the model predictive control principle, the control quantity u(k) of the first step of the optimal control sequence is used as the actual output of the controller;

[0115] Step 15: Complete the vehicle state update at time k by executing control;

[0116] Step 16: When there is a large difference between the future road section operating condition data obtained at time k and at time k+1, execute steps 2-5 to complete the battery SOC reference trajectory update;

[0117] Step 17: Enter time k+1 and repeat steps 6-16 until the entire journey is completed, thus realizing the HEV model predictive control energy management with real-time error correction.

[0118] This complex driving and energy management system is designed to optimize the driving efficiency and battery maintenance of hybrid vehicles by using advanced on-board technologies and algorithms. This system integrates vehicle navigation, driving style analysis, real-time data collection and analysis, predictive models and optimization control algorithms. The following is a detailed explanation of how each step works:

[0119] Step 1: Get driving directions

[0120] The system uses the vehicle's navigation system to determine the best route before the journey begins, which forms the basis for planning and forecasting.

[0121] Step 2: Driving style recognition

[0122] The on-board chip collects the driver's historical driving data, including speed, acceleration, etc., to assess the driver's style (such as aggressive, normal or calm). This helps the system predict the driver's behavior pattern on the future route.

[0123] Step 3: Obtain real-time operating data

[0124] The vehicle's connected technology is used to collect real-time traffic, weather and other data on the road section to be traveled, and match it with the driver's style to optimize the route and predict road conditions.

[0125] Step 4: Processing of working condition data

[0126] The collected data of all road sections are spliced ​​together to check the data continuity. If there are abnormal acceleration data, they are processed by linear interpolation to ensure the smoothness and reliability of the data.

[0127] Step 5: Battery SOC trajectory optimization

[0128] The PMP (Pontryagin's Minimum Principle) algorithm is used to calculate the optimal state of charge (SOC, i.e., remaining battery capacity) trajectory of the battery to maximize trip efficiency and battery life.

[0129] Steps 6 to 17: Real-time data acquisition and control loop

[0130] Real-time data collection (step 6): Real-time monitoring of vehicle and environmental data during driving, such as speed, acceleration, distance to the vehicle ahead, etc.

[0131] Speed ​​prediction and error calculation (steps 7 to 8): Based on the collected data and the deep neural network model, the future speed is predicted and the prediction error is calculated.

[0132] Cost function adjustment (steps 9 to 11): The cost function is adjusted according to the error to optimize the SOC control strategy.

[0133] Optimization algorithm execution (steps 12 to 14): Use the PMP algorithm to update the control strategy, determine the optimal control sequence based on the prediction model, and execute the first control action.

[0134] Control loop and data update (steps 15 to 17): Update vehicle status and adjust control strategy based on real-time data and new operating condition information.

[0135] The goal of this system is to maximize energy efficiency and optimize vehicle performance during driving, ensuring that hybrid vehicles can maintain optimal performance under different driving conditions through continuous data collection and real-time feedback adjustments. In this way, the system can not only adapt to different driver styles, but also respond to changes in road conditions in real time, thereby effectively managing energy consumption while ensuring driving safety.

[0136] like Figure 4 As shown, the energy management system of the hybrid vehicle model predictive control provided by the embodiment of the present invention is characterized by comprising:

[0137] Prediction model building module, used to train and build a neural network speed prediction model;

[0138] An adjustment mechanism building module, used to build a forecast error adjustment mechanism;

[0139] The reference trajectory design module is used to design the battery SOC reference trajectory.

[0140] An application embodiment of the present invention provides a computer device, which includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor executes the steps of an energy management method for hybrid electric vehicle model predictive control.

[0141] An application embodiment of the present invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the steps of an energy management method for model predictive control of a hybrid electric vehicle.

[0142] An application embodiment of the present invention provides an information data processing terminal, which is used to implement an energy management system for model predictive control of a hybrid electric vehicle.

[0143] The detailed working principle of the energy management system of the hybrid vehicle model predictive control provided by the embodiment of the present invention is as follows:

[0144] First, the prediction model building module uses a large amount of historical data and real-time data to train and build a neural network speed prediction model through deep learning or other machine learning algorithms. This model can predict the vehicle speed in the future based on the current vehicle status, road conditions, driving habits and other information. The predicted speed information is crucial for the energy management system because it is directly related to the vehicle's energy demand.

[0145] Secondly, the adjustment mechanism building module will build a prediction error adjustment mechanism based on the difference between the actual vehicle speed and the predicted vehicle speed. When there is a deviation between the predicted vehicle speed output by the prediction model and the actual vehicle speed, the adjustment mechanism will adjust the parameters or weights of the prediction model in real time to reduce the error and improve the accuracy of the prediction.

[0146] Then, the reference trajectory design module designs the reference trajectory of the battery SOC (state of charge) according to the actual needs of the vehicle and the characteristics of the battery. The SOC reference trajectory is an idealized battery charge change curve that takes into account the energy demand of the vehicle under different driving conditions and the charging and discharging efficiency of the battery.

[0147] In the actual operation of the energy management system, the prediction model predicts the future vehicle speed based on the current state information, thereby predicting the future energy demand. Then, the system calculates the optimal energy allocation strategy through the optimization algorithm based on the predicted energy demand and the reference trajectory of the battery SOC. This strategy takes into account the energy conversion efficiency between the engine and the battery, as well as the charging and discharging characteristics of the battery, to ensure that the energy utilization efficiency is maximized while meeting the vehicle's power requirements.

[0148] In addition, as one of the application embodiments of the present invention, the computer device executes the steps of the energy management method of the hybrid vehicle model predictive control through its memory and processor. When the computer program is executed by the processor, the prediction model is trained and adjusted according to the preset algorithm and logic, the energy distribution strategy is calculated, and the control instructions are sent to the corresponding actuator through the communication interface with other vehicle systems to achieve optimal energy management.

[0149] The computer-readable storage medium stores computer programs for implementing these functions. When these programs are executed by the processor, the energy management system of the hybrid vehicle model predictive control can be driven to operate normally.

[0150] Finally, the information data processing terminal is another application embodiment. It is a dedicated device on the vehicle and a functional module integrated into other vehicle systems. This terminal is responsible for receiving and processing data from various sensors, executing energy management algorithms, and outputting control instructions to achieve optimal management of hybrid vehicle energy.

[0151] The energy management system of the hybrid vehicle model predictive control provided by the embodiment of the present invention realizes efficient and optimized management of the energy of the hybrid vehicle through the design of the neural network speed prediction model, the prediction error adjustment mechanism and the battery SOC reference trajectory. This can not only improve the energy efficiency and performance of the vehicle, but also reduce energy consumption and pollutant emissions, which is of great significance to promoting the development of new energy vehicles.

[0152] The energy management system of the hybrid vehicle model predictive control of the present invention uses multivariate data as the input of the deep neural network speed prediction model, which is equivalent to a single speed sequence as a feature input, and the short-term working condition prediction is more accurate. By setting a prediction failure compensation mechanism, when there is an occasional excessive prediction error, the root mean square error between the predicted speed and the actual speed is used as an exponential decay factor to construct the weight of the instantaneous cost, compared with the traditional MPC strategy without a compensation mechanism, the impact of the prediction error on the MPC optimization ability is reduced. The network connection technology is used to obtain the characteristics of future driving road conditions and the influence of driving style is considered to realize the generation of the optimal reference trajectory of the battery SOC based on the network connection technology, and the battery SOC reference trajectory is regularly updated as the prediction time domain state quantity constraint. Compared with the linear formula reference trajectory method, the proposed method greatly improves the adaptability and accuracy of power distribution, so that the motor and engine are always in the high efficiency area, and efficient energy utilization is achieved.

[0153] Compared with the traditional MPC strategy, the energy management method and system of the hybrid vehicle model predictive control of the present invention enhances the multi-scenario prediction capability of the model through multi-sensory data and real-time error correction mechanism, and improves the MPC optimization capability by generating the battery SOC optimal reference trajectory based on network connection technology. Therefore, the fuel economy of the energy management method and system of the present invention can be achieved to be comparable to that of the theoretical optimal global dynamic programming method, and at the same time has better online control adaptability.

[0154] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0155] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. An energy management method for hybrid electric vehicle model predictive control, characterized in that: The following steps are involved: S1, training and establishing a deep neural network speed prediction model based on multi-dimensional features to predict the speed series in the short time domain in the future; S2, constructing a prediction error adjustment mechanism, including designing an exponential decay factor to adjust the cost function weight by comparing the cumulative error between the actual driving speed and the predicted speed; S3, design the battery SOC reference trajectory and optimize the battery charging and discharging behavior through model predictive control to improve energy efficiency; Step S1 further includes using radar, GPS, and vehicle speed sensors to collect historical operating data of the actual vehicle and information about the vehicle's surrounding environment, establish a multi-dimensional feature database, and discretize the data; The deep neural network model described herein uses the preceding vehicle distance, preceding vehicle speed, driving style, current and previous vehicle status, road slope and weather conditions as input features; Step S2 specifically includes calculating the historical cumulative prediction error of this trip by collecting the predicted speed and actual driving speed up to the current time K of this trip, and designing an exponential decay factor by comparing the error and the threshold exceeding degree based on the prediction error threshold obtained by expert experience, so as to predict the weight of the cost function at different times in the time domain; By collecting the predicted speed and actual speed of the trip up to the current time K, the historical cumulative prediction error V of this trip is calculated. RMSE (k), prediction error threshold V based on expert experience Err , by comparing the error with the threshold exceeding degree to design the exponential decay factor, that is, when V RMSE (k)>V Err hour: Thus, the weight of the cost function at different times in the time domain is predicted: Where γ is the coefficient, when V RMSE (k)≤V Err When , the cost function weights at different times are: α(k+i)= The total cost function for predicting Hp in the time domain is: Where J(k) is the instantaneous cost function at time k.

2. The energy management method according to claim 1, characterized in that: The S3 step specifically includes collecting the driver's historical operating conditions and road condition information through the PHEV on-board chip, extracting the impact degree data of the driving conditions of the last route, using the root mean square of the impact degree to represent the driving style of the condition, and using network technology to obtain real-time future driving section condition data, selecting the condition data that is most similar to the driving style of the vehicle, and splicing all the road conditions of the route.

3. The energy management method according to claim 2, characterized in that: When there is an abnormal acceleration at the splicing point, linear interpolation is used to complete the route condition prediction based on the driving style. The Pontryagin minimum principle algorithm is used to calculate the battery SOC trajectory based on the global prediction condition, thereby obtaining the optimal reference trajectory of the battery SOC and constraining the predicted time domain terminal state quantity for real-time MPC optimization control.

4. The energy management method according to claim 1, characterized in that: The control process further includes obtaining a driving route according to a set vehicle navigation destination before the start of the journey; collecting vehicle and surrounding driving environment data through on-board sensors, using the input of a deep neural network prediction model to predict speed sequences at future times; and calculating an exponential decay factor based on the error between the predicted speed and the actual driving speed to design a cost function coefficient, thereby completing the calculation and execution of the optimal control sequence.

5. A hybrid vehicle model predictive control energy management system that implements the hybrid vehicle model predictive control energy management method according to any one of claims 1 to 4, characterized in that: include: Prediction model building module, used to train and build a neural network speed prediction model; An adjustment mechanism building module, used to build a forecast error adjustment mechanism; The reference trajectory design module is used to design the battery SOC reference trajectory.

6. A computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the energy management method of hybrid vehicle model predictive control as claimed in any one of claims 1 to 4.

7. An information data processing terminal, used for implementing the energy management system of the hybrid electric vehicle model predictive control as claimed in claim 5.

Citation Information

Cited By

  • Plug-in hybrid electric vehicle model predictive control energy management method based on reinforcement learning

    CN120116917A