A hybrid electric vehicle energy management control method based on stochastic dynamic programming
Through the energy management control method based on random dynamic programming, cluster analysis and linear combination using historical and actual driving data, the problem of poor adaptability of energy management strategies in the existing technology is solved, and specific working conditions and different vehicle routes are realized targeted energy management improves control accuracy and adaptability.
Patent Information
- Application Number
- CN202210895082.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-07-28
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2042-07-28
AI Technical Summary
The existing hybrid vehicle energy management strategies lack adaptability and cannot control the vehicle's driving conditions in a targeted manner, especially in application scenarios with relatively fixed routes, and relying on known operating conditions, it cannot adapt to different vehicles and routes. change.
The energy management control method based on random dynamic programming is adopted, and the transfer probability matrix is established by obtaining historical driving data for preprocessing and clustering analysis. The stochastic dynamic programming iterative algorithm is used to solve energy management strategies under different categories of working conditions, and combined with actual driving. The data are linearly combined to output the optimal energy management strategy.
It improves the targetedness and adaptability of energy management strategies, can control specific driving conditions more accurately, is suitable for different vehicles and routes, and improves the accuracy and adaptability of energy management control.
Smart Images

Figure CN115257694B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of hybrid electric vehicles, and in particular to a hybrid electric vehicle energy management control method based on random dynamic programming. Background Art
[0002] Hybrid electric vehicles have great potential in alleviating the increasingly serious global warming, oil consumption and environmental pollution. The hybrid system, equipped with a variety of energy storage systems and power sources, provides design freedom for the efficient power output of electric motors and internal combustion engines. The energy management strategy plays an important role in coordinating the energy distribution of different power sources and is of great significance to give full play to the structural advantages of the hybrid system. Since the energy management strategy is the link between the fuel power system and the electric drive system, it directly affects the power, economy, comfort and emissions of the whole vehicle, so it is the focus and difficulty of research in the field of hybrid vehicles.
[0003] At present, energy management strategies include rule-based energy management strategies and optimization-based energy management strategies. Among them, rule-based control strategies generally consider parameters such as the power requirements of the engine and motor, vehicle speed, and battery SOC, and set the switching rules of the working mode in advance based on manual experience. Generally, a flow chart or state diagram is used to realize the conversion from one mode to another. Rule-based energy management strategies can be divided into deterministic rules and fuzzy rules. Whether it is simple rules or fuzzy logic, rule-based control strategies can be implemented by looking up tables. The algorithm is simple and practical, which is the basis for realizing complex control strategies. Therefore, it is widely used in the energy management system of hybrid vehicles. However, this control method has an important defect, that is, the algorithm model relies on expert experience and does not have good adaptability to working conditions.
[0004] Energy management strategies based on optimization can be divided into global optimization and instantaneous optimization. In theory, the optimal energy management can be obtained under known operating conditions. However, the actual working environment of automobiles is complex and changeable, which limits its wide application. For example, for hybrid vehicles with relatively fixed routes (such as buses, logistics vehicles, etc.), the above energy management strategies cannot specifically control the vehicle driving conditions, nor can they adapt to application scenarios of different vehicles and different routes. Summary of the invention
[0005] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide a hybrid vehicle energy management control method based on random dynamic programming. For hybrid vehicles with relatively fixed routes, it can solve the defects of the current hybrid vehicle energy management strategy having poor adaptability and dependence on known operating conditions.
[0006] The object of the present invention can be achieved by the following technical solution: A hybrid electric vehicle energy management control method based on random dynamic programming, comprising the following steps:
[0007] S1. Acquire historical driving data of the hybrid vehicle and pre-process the historical driving data;
[0008] S2. Based on the pre-processed historical driving data, a Gaussian mixture model is used for cluster analysis. Through model training, a classification model and different types of operating condition data are obtained;
[0009] S3. According to the data of different types of working conditions, a transition probability matrix of different types of working conditions is established;
[0010] S4. Using the transfer probability matrix of different types of working conditions and adopting the random dynamic programming iterative algorithm, the energy management strategy under different types of working conditions is solved;
[0011] S5, collecting actual driving data of hybrid vehicles, combining with the classification model, and outputting the optimal energy management strategy by linearly combining energy management strategies under different types of working conditions;
[0012] S6. According to the optimal energy management strategy, the driving state of the hybrid vehicle is controlled accordingly.
[0013] Furthermore, the driving data includes vehicle speed and SOC information, and the driving data is specifically collected by sensors.
[0014] Furthermore, the specific process of preprocessing the historical driving data in step S1 is as follows:
[0015] Model vehicle power requirements;
[0016] The historical vehicle speed data is input into the vehicle power demand model, and the corresponding power demand data is output.
[0017] Furthermore, the vehicle power demand model is specifically:
[0018]
[0019] Among them, F req is the driving force required by the car, Cd is the drag coefficient, A is the frontal area, v is the vehicle speed, m is the vehicle mass, g is the acceleration of gravity, α is the road slope, f is the rolling resistance coefficient, and t is the time.
[0020] Furthermore, the specific process of step S2 is: taking the historical vehicle speed and the corresponding power demand data as driving characteristics, determining the number of clustering categories, using a Gaussian mixture model to perform clustering analysis, estimating the parameters of the Gaussian mixture model, and training to obtain a classification model and different categories of operating condition data.
[0021] Furthermore, the Gaussian mixture model is specifically:
[0022]
[0023] Among them, w is the weight coefficient, φ i (x|θ i ) is a Gaussian distribution, x is a feature variable, θ is a parameter of the Gaussian distribution, i = 1, 2, …, M, and M is the number of clustering categories.
[0024] Furthermore, the step S2 specifically adopts an EM (Expectation-Maximum) algorithm to estimate the parameters of the Gaussian mixture model.
[0025] Furthermore, the specific process of step S3 is: using the nearest neighbor method, the power demand is discretized into several states, the change process of the power demand over time is used as a Markov chain, and according to the clustered different categories of working condition data, a transition probability matrix under the same category of working condition is established, and the calculation formula is as follows:
[0026]
[0027] Among them, p ij is the probability of state i transitioning to state j, n ij is the total number of transitions from state i to state j, n i is the total number of occurrences of state i.
[0028] Furthermore, the specific process of step S4 is: using the calculated transfer probability matrix of different types of operating conditions, taking the minimum fuel consumption of the whole vehicle and maintaining the balance of battery SOC under low power state as the optimization goal, and taking SOC and vehicle power demand as the state of the energy management strategy, according to the random dynamic programming value iteration algorithm, solving the energy management strategy under different types of operating conditions, thereby obtaining multiple energy management strategies with SOC and power demand as table indexes and energy management target values as table values.
[0029] Furthermore, the specific process of step S5 is: using a sensor to obtain the current vehicle speed, using a vehicle power demand model to calculate the current power demand, and using this as a feature of the actual working condition;
[0030] The classification model is used to calculate the posterior probability corresponding to the actual working condition, which is used as the weight to linearly combine the energy management strategies of different types of working conditions, and the optimal energy management strategy corresponding to the hybrid vehicle is output.
[0031] Compared with the prior art, the present invention has the following advantages:
[0032] 1. The present invention takes into account that vehicles with relatively fixed routes will generate a large amount of driving data during driving. Therefore, the present invention uses these data to perform cluster analysis to establish a classification model and obtain different types of operating condition data, and then calculates the corresponding multiple transition probability matrices and uses random dynamic programming to iteratively solve the corresponding energy management strategy, thereby ensuring that the energy management strategy is more targeted to specific driving conditions and improving the accuracy of energy management control.
[0033] 2. The present invention takes into account the differences in driving routes between different vehicles, and uses a classification model to calculate the posterior probability corresponding to the operating conditions. This is used as a weight to linearly combine the energy management strategies of different categories of operating conditions, thereby obtaining the vehicle's optimal capacity management strategy. In this way, the corresponding energy management strategies can be calculated for different vehicles and different routes, effectively improving the adaptability of energy management control. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] Figure 1 It is a schematic diagram of the method flow of the present invention;
[0035] Figure 2 A schematic diagram of the process of determining energy management strategies under different types of working conditions in the present invention;
[0036] Figure 3 It is a schematic diagram of the operation control principle of the embodiment;
[0037] Figure 4 Schematic diagram of the application framework in the embodiment. DETAILED DESCRIPTION
[0038] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.
[0039] Example
[0040] like Figure 1 As shown, a hybrid electric vehicle energy management control method based on stochastic dynamic programming includes the following steps:
[0041] S1. Acquire historical driving data of the hybrid vehicle and pre-process the historical driving data;
[0042] S2. Based on the pre-processed historical driving data, a Gaussian mixture model is used for cluster analysis. Through model training, a classification model and different types of operating condition data are obtained;
[0043] S3. According to the data of different types of working conditions, a transition probability matrix of different types of working conditions is established;
[0044] S4. Using the transfer probability matrix of different types of working conditions and adopting the random dynamic programming iterative algorithm, the energy management strategy under different types of working conditions is solved;
[0045] S5, collecting actual driving data of hybrid vehicles, combining with the classification model, and outputting the optimal energy management strategy by linearly combining energy management strategies under different types of working conditions;
[0046] S6. According to the optimal energy management strategy, the driving state of the hybrid vehicle is controlled accordingly.
[0047] This embodiment applies the above technical solution, such as Figure 2 to Figure 4 As shown, it mainly includes:
[0048] (1) Using sensors to collect data from real hybrid vehicles during real-world driving, including information about vehicle speed over time and SOC information, and using it as input to the vehicle power demand model;
[0049] The vehicle power demand model is established in advance, taking into account the longitudinal dynamics of the vehicle, including rolling resistance, wind resistance, acceleration resistance, and inertial resistance. The mathematical expression of the vehicle power demand model is as follows:
[0050]
[0051] In the formula, F req is the driving force of automobile demand, Cd is the drag coefficient, A is the frontal area, v is the vehicle speed, m is the vehicle mass, g is the acceleration of gravity, α is the road slope, f is the rolling resistance coefficient, and t is the time;
[0052] According to the established vehicle power demand model, the information of vehicle power changing over time can be calculated;
[0053] (2) Taking vehicle speed and power demand as the features of driving data, a Gaussian mixture model is used for cluster analysis. The mathematical expression of the Gaussian mixture model is as follows:
[0054]
[0055] Where w is the weight coefficient, φ i (x|θ i ) is a Gaussian distribution, x is a characteristic variable, θ is a parameter of the Gaussian distribution, i = 1, 2, ..., M, M is the number of cluster types;
[0056] Using the collected vehicle speed data and the calculated power demand data, the parameters of the Gaussian mixture model are estimated to construct a classification model;
[0057] For a specific working condition, according to the characteristics of the working condition, the classification model calculates the posterior probability of belonging to each category. This working condition belongs to the category with the highest probability. The posterior probability calculation formula is as follows:
[0058]
[0059] (3) Based on the clustered operating condition data, the nearest neighbor method is used to establish the transition probability matrix under the same category of operating conditions. The calculation formula is as follows:
[0060]
[0061] In the formula, p ij is the probability of state i transitioning to state j, n ij is the total number of transitions from state i to state j, n i is the total number of occurrences of state i;
[0062] (4) using the transfer probability matrix of different types of working conditions calculated in step (3), and according to the relevant theory of stochastic dynamic programming, an iterative algorithm is used to solve the energy management strategy of different types of working conditions;
[0063] (5) In the actual operation of the energy management strategy, the current vehicle speed is first obtained using sensors, and its power demand is calculated accordingly, which is used as the characteristic of the working condition;
[0064] Then, the posterior probability corresponding to the working condition is calculated according to the trained Gaussian mixture model (i.e., the classification model), and this is used as the weight to linearly combine the energy management strategies of different types of working conditions as the final energy management strategy for the vehicle.
[0065] In this embodiment, taking a series hybrid electric vehicle as an example, compared with a pure electric vehicle, a series hybrid electric vehicle has an additional range extender that can charge the battery or power the drive motor, thereby increasing the vehicle's cruising range; the energy management strategy of a series hybrid electric vehicle is mainly to control the output power of the range extender.
[0066] The specific application process includes:
[0067] 1. Use GPS modules, wheel speed sensors, etc. to collect data from actual hybrid vehicles in the real world, including information about vehicle speed changes over time, SOC information, etc., and use it as the actual working condition of the vehicle power demand model;
[0068] Second, measure the actual vehicle parameters and establish the vehicle power demand model, taking into account the longitudinal dynamics of the vehicle, including rolling resistance, wind resistance, acceleration resistance, and inertial resistance, to simulate the actual power demand during vehicle operation. The mathematical expression of the model is as follows:
[0069]
[0070] In the formula, F req is the driving force of automobile demand, Cd is the drag coefficient, A is the frontal area, v is the vehicle speed, m is the vehicle mass, g is the acceleration of gravity, α is the road slope, f is the rolling resistance coefficient, and t is the time;
[0071] According to the established power demand model, calculate the information of vehicle power changing over time;
[0072] 3. Taking vehicle speed and power demand as the features of driving data, determine the number of cluster types M, and use Gaussian mixture model for cluster analysis. The mathematical expression of Gaussian mixture model is as follows:
[0073]
[0074] Where w is the weight coefficient, φ i (x|θ i ) is a Gaussian distribution, x is a characteristic variable, θ is a parameter of the Gaussian distribution, i = 1, 2, …, M;
[0075] Then, the vehicle speed data collected above and the corresponding power demand data calculated are used to estimate the parameters of the Gaussian mixture model using the EM (Expectation-Maximum) algorithm to train a classification model.
[0076] Fourth, for a specific working condition, the vehicle speed and power under the working condition are used as features to calculate the posterior probability of belonging to each category. This working condition belongs to the category with the highest probability. The posterior probability calculation formula is as follows:
[0077]
[0078] 5. Use the nearest neighbor method to discretize the power demand into several states, and regard the change of power demand over time as a Markov chain. According to the clustered working condition data, establish the transition probability matrix under the same category of working conditions. The calculation formula is as follows:
[0079]
[0080] In the formula, p ij is the probability of state i transitioning to state j, n ij is the total number of transitions from state i to state j, n iis the total number of occurrences of state i;
[0081] 6. Using the calculated transfer probability matrix of different types of working conditions, taking the minimum fuel consumption of the whole vehicle and maintaining the balance of battery SOC under low power state as the optimization goal, and taking SOC and vehicle power demand as the state of energy management strategy, according to the random dynamic programming value iteration algorithm, solve the power output value of the range extender under the energy management strategy state of different types of working conditions, thereby obtaining multiple energy management strategies with SOC and power demand as table index and range extender output power as table value;
[0082] 7. In the actual operation of the energy management strategy, the current vehicle speed is obtained by using sensors to calculate its power demand, which is used as the characteristic of the working condition. The posterior probability of the working condition is calculated based on the trained Gaussian mixture model. This is used as the weight to linearly combine the energy management strategies of different types of working conditions to obtain the final range extender output power of the vehicle.
[0083] In summary, the present technical solution takes into account that vehicles with relatively fixed routes will generate a large amount of driving data during driving, so these data are used for cluster analysis and the corresponding energy management strategy is calculated, so that the energy management strategy is more targeted to specific driving conditions; in addition, considering that the driving routes of different vehicles may be different, the present technical solution can calculate the corresponding energy management strategy for different vehicles and different routes, and is adaptable; in practical applications, the present technical solution can be combined with Internet of Vehicles technology: the actual running vehicles upload the data to the cloud, and in the cloud, the method mentioned in the present technical solution can be used to calculate the energy management strategy for the vehicle in a specific scenario.
Claims
1. A hybrid electric vehicle energy management control method based on stochastic dynamic programming, characterized in that: The following steps are involved: S1. Acquire historical driving data of the hybrid vehicle and pre-process the historical driving data; S2. Based on the pre-processed historical driving data, a Gaussian mixture model is used for cluster analysis. Through model training, a classification model and different types of operating condition data are obtained; S3. According to the data of different types of working conditions, a transition probability matrix of different types of working conditions is established; S4. Using the transfer probability matrix of different types of working conditions and adopting the random dynamic programming iterative algorithm, the energy management strategy under different types of working conditions is solved; S5, collecting actual driving data of hybrid vehicles, combining with the classification model, and outputting the optimal energy management strategy by linearly combining energy management strategies under different types of working conditions; S6. controlling the driving state of the hybrid vehicle accordingly according to the optimal energy management strategy; The driving data includes vehicle speed and SOC information, and the driving data is specifically collected by a sensor; The specific process of preprocessing the historical driving data in step S1 is as follows: Model vehicle power requirements; Input the historical vehicle speed data into the vehicle power demand model, and output the corresponding power demand data; The vehicle power demand model is specifically: Among them, F req is the driving force required by the car, Cd is the drag coefficient, A is the frontal area, v is the vehicle speed, m is the vehicle mass, g is the acceleration of gravity, α is the road slope, f is the rolling resistance coefficient, and t is the time; The specific process of step S2 is: taking the historical vehicle speed and the corresponding power demand data as driving characteristics, determining the number of clustering categories, using a Gaussian mixture model to perform clustering analysis, estimating the parameters of the Gaussian mixture model, and training to obtain a classification model and different categories of operating condition data; The Gaussian mixture model is specifically: Among them, w is the weight coefficient, φ i (x|θ i ) is Gaussian distribution, x is the feature variable, θ is the parameter of Gaussian distribution, i = 1, 2, ..., M, M is the number of clustering categories; The step S2 specifically uses the EM algorithm to estimate the parameters of the Gaussian mixture model; The specific process of step S3 is: using the nearest neighbor method, discretizing the power demand into several states, taking the change process of the power demand over time as the Markov chain, and establishing the transition probability matrix under the same category of working conditions according to the clustered working condition data of different categories. The calculation formula is as follows: Among them, p ij is the probability of state i transitioning to state j, n ij is the total number of transitions from state i to state j, n i is the total number of occurrences of state i; The specific process of step S4 is: using the calculated transfer probability matrix of different types of working conditions, taking the minimum fuel consumption of the whole vehicle and maintaining the balance of the battery SOC under the low power state as the optimization goal, and taking the SOC and the power demand of the whole vehicle as the state of the energy management strategy, according to the random dynamic programming value iteration algorithm, solving the energy management strategy under different types of working conditions, thereby obtaining multiple energy management strategies with SOC and power demand as table indexes and energy management target values as table values; The specific process of step S5 is: using the sensor to obtain the current vehicle speed, using the vehicle power demand model to calculate the current power demand, and using this as the characteristic of the actual working condition; The classification model is used to calculate the posterior probability corresponding to the actual working condition, which is used as the weight to linearly combine the energy management strategies of different types of working conditions, and the optimal energy management strategy corresponding to the hybrid vehicle is output.
Citation Information
Patent Citations
Energy management method and system of hybrid vehicle
CN102717797A
Prediction energy management method of networked hybrid electric vehicle
CN110696815A
