A-ecms energy management method based on predictive road traffic information

Abstract

The application belongs to the technical field of vehicle energy management, and discloses an A-ECMS energy management method based on predictive road traffic information, which comprises the following steps: obtaining historical driving data of a vehicle, and calculating characteristic parameters of the historical driving data; inputting the characteristic parameters into an offline driving data identification library to obtain current optimization power control parameters; obtaining predictive road data in front of the vehicle, inputting the predictive road data into the offline driving data identification library for scene identification to obtain a prediction scene and corresponding prediction optimization power control parameters; calculating a long-term energy trajectory based on an online dynamic programming algorithm; calculating final power control parameters based on a DRL algorithm in combination with the current optimization power control parameters, the prediction optimization power control parameters, the long-term energy trajectory and a current energy trajectory; and calculating an optimal control sequence of a vehicle power system in real time according to the final power control parameters. The technical scheme can realize globally optimal energy management.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of vehicle energy management, and particularly relates to an A-ECMS energy management method based on predictive road traffic information. BACKGROUND

[0002] With the rapid development of the global economy and the continuous growth of the population, the demand for energy is increasing, and traditional fuel vehicles rely on fossil fuels such as oil, which is limited and faces the crisis of exhaustion. The large-scale use of fuel vehicles has led to serious air pollution and greenhouse gas emissions, exacerbating environmental problems such as global warming. Therefore, it is urgent to find clean and efficient ways to use energy.

[0003] Before the technology of pure electric vehicles is fully mature, hybrid electric vehicles, as a transitional vehicle type, combine the advantages of traditional fuel vehicles and pure electric vehicles, which can reduce fuel consumption and emissions, and maintain longer cruising range and better power performance. The core of hybrid electric vehicles lies in the design and control of its energy management system. Through reasonable energy management strategies, the optimization and collaborative work of multiple power sources (such as engines, motors and batteries) can be realized, thereby improving the energy utilization efficiency of the whole vehicle.

[0004] Traditional energy management strategies, including rule-based and instantaneous optimization-based, mainly rely on instantaneous driving cycle information, which is determined by the immediate operation of the driver, and they need a lot of engineering experience to pre-calibrate the internal rules and control parameters of the strategy. In the control process, fixed rules and control parameters cannot adapt to complex and variable road environments and driving needs, which will lead to low energy utilization efficiency, high fuel consumption and increased emissions. Global optimization-based energy management strategies (such as dynamic programming, convex optimization, genetic algorithm, etc.) can provide theoretically optimal solutions, but their calculation amount is huge and difficult to apply in real time. Energy management strategies based on intelligent algorithms (such as model predictive control, reinforcement learning, etc.) optimize control strategies through learning and training, which have strong adaptability, but require high algorithm training and calculation, and are highly dependent on the accuracy of the prediction model. In summary, traditional energy management strategies that rely on instantaneous driving cycle information lack adaptability.

[0005] Although the energy management strategy based on working condition recognition can switch the preset control parameter scheme in real time according to the historical driving information in the past short period, it still cannot cope with the complex traffic environment in the front road section in advance due to the lack of the traffic flow vision of the front road section. The energy management strategy based on the forward driving cycle prediction introduces the forward-looking information of the front road in the energy management, and optimizes the energy distribution by predicting the driving cycle in the short term, but such algorithm usually needs stronger computing power and is difficult to realize in the real vehicle, and the forward-looking information cannot guarantee the global energy optimization. The energy management strategy based on the global driving cycle prediction further expands the vision of the energy management, considers the whole trip process, and combines the cloud computing and edge computing platform to provide decision support for the vehicle, but the traffic flow information of different road sections in the trip changes over time, which will lead to the instability of the global planning energy management strategy. SUMMARY

[0006] The purpose of the present application is to provide an A-ECMS energy management method based on predictive road traffic information to solve the problems existing in the prior art.

[0007] To achieve the above purpose, the present application provides an A-ECMS energy management method based on predictive road traffic information, comprising:

[0008] Obtaining the historical driving data of the vehicle, calculating the characteristic parameters of the historical driving data;

[0009] Inputting the characteristic parameters into the offline driving data recognition library to obtain the current optimization power control parameter; wherein the offline driving data recognition library is a multi-dimensional power parameter optimization library including different driving styles, driving scenes, initial SOC and target SOC;

[0010] Obtaining the predictive road data in front of the vehicle, inputting the predictive road data into the offline driving data recognition library for scene recognition to obtain the predicted scene and the corresponding predicted optimization power control parameter; using the online dynamic programming algorithm to calculate the long-term energy trajectory;

[0011] Based on the DRL algorithm, combining the current optimization power control parameter, the predicted optimization power control parameter, the long-term energy trajectory and the current energy trajectory to calculate the final power control parameter;

[0012] According to the final power control parameter, the optimal control sequence of the vehicle power system is calculated in real time.

[0013] Optionally, the construction process of the offline driving data recognition library comprises:

[0014] Obtaining the original data, the original data including driving related data, performing data cleaning and normalization processing on the original data to obtain the preprocessed original data;

[0015] performing principal component analysis on the pretreated original data to obtain dimension-reduced data;

[0016] selecting a plurality of data points in the dimension-reduced data as initial clustering centers to perform clustering analysis, and obtaining driving scene type data and driving style type data with different characteristic parameters;

[0017] constructing a nonlinear relationship between the characteristic parameters and the driving scene and the driving style based on a supervised learning recognition algorithm, and establishing a driving scene library and a driving style library; the characteristic parameters include driving scene characteristic parameters and driving style characteristic parameters;

[0018] the library inputs of the driving scene library and the driving style library are driving scene characteristic parameters and driving style characteristic parameters respectively, and the outputs are corresponding driving scenes and driving styles;

[0019] setting a plurality of initial energy trajectories and target energy trajectories, and based on an intelligent optimization algorithm, iteratively optimizing a typical driving condition with different driving styles, initial energy trajectories and target energy trajectories to obtain corresponding optimized power control parameters, and establishing a multi-dimensional power parameter optimization library of different driving styles, driving scenes, initial SOC and target SOC.

[0020] Optionally, the driving scene characteristic parameters include maximum positive acceleration, positive acceleration standard deviation, positive acceleration average value, acceleration time proportion, maximum negative acceleration, negative acceleration standard deviation, negative acceleration average value, negative acceleration time proportion, maximum vehicle speed, vehicle speed standard value, average vehicle speed, average acceleration, acceleration standard deviation, cruise section average vehicle speed and cruise time proportion.

[0021] Optionally, the driving style characteristic parameters include maximum positive acceleration, positive acceleration standard deviation, positive acceleration average value, positive jerk standard deviation, positive jerk average value, acceleration time proportion, maximum negative acceleration, negative acceleration standard deviation, negative acceleration average value, negative acceleration time proportion, negative jerk standard deviation, negative jerk average value, maximum vehicle speed, average vehicle speed, average acceleration, acceleration standard value, cruise section average vehicle speed, cruise time proportion, accelerator pedal standard deviation, accelerator pedal change rate standard deviation, brake pedal standard deviation, brake pedal change rate standard deviation, average jerk, jerk standard deviation.

[0022] Optionally, the characteristic parameters are input into an offline driving data recognition library to obtain the current optimized power control parameters, specifically including:

[0023] The feature parameters are input into an offline driving data recognition library for scene recognition and style recognition, to obtain a typical driving scene and a driving style corresponding to the feature parameters, and the offline driving data recognition library outputs a corresponding current energy trajectory, a target energy trajectory and a current optimization power control parameter based on the recognized typical driving scene and driving style.

[0024] Optionally, the predictive road data includes a green wave speed range, a remaining distance of the entire driving route, a remaining required time of the entire driving route, average speeds of each road section of the entire driving route, congestion conditions of each road section of the entire driving route, distances of each road section of the entire driving route at the average speeds, distances of the congestion road sections and passing times of the congestion road sections.

[0025] Optionally, an online dynamic programming algorithm is used to calculate a long-term energy trajectory based on the whole-process predictive road traffic information, and a specific calculation formula is as follows:

[0026]

[0027] The constraint condition is as follows:

[0028]

[0029] The cost function is as follows:

[0030] V L = V eng / L (T eng ,N eng )+ V batt / L (P batt )

[0031] In the formula, E m is an open-loop voltage of the battery, R int is an internal resistance of the battery, Q c is a total capacity of the battery, SOC k is an SOC of the kth stage, SOC k+1 is an SOC of the (k+1)th stage, is an average battery power of the kth stage, is an average engine power of the kth stage, is an average engine speed of the kth stage; SOC min and SOC max are lower and upper limits of the SOC, P batt / chg and P batt / dischg are battery charging power and battery discharging power, T min and T max are lower and upper limits of engine torque, N min and N max are lower and upper limits of engine speed, and Veng / L and V batt / L respectively are engine fuel consumption and battery equivalent fuel consumption, V L is the comprehensive fuel consumption.

[0032] Optionally, based on the DRL algorithm, the final power control parameter is calculated in combination with the current optimization power control parameter, the predicted optimization power control parameter, the long-term energy trajectory and the current energy trajectory, and the specific calculation formula is:

[0033] s final (t)=F(s t (t),s pre (t),SOC ref (t),SOC t (t))

[0034] Wherein, s final is the A-ECMS final power control parameter, s t is the power control parameter based on short-term working condition identification; s pre is the power control parameter corresponding to the front road section working condition based on predictive road information identification; SOC ref is the long-term energy trajectory obtained based on online dynamic programming; SOC t is the current SOC, and F is the power control parameter adjustment strategy trained using the deep reinforcement learning algorithm.

[0035] Optionally, the optimal control sequence of the vehicle power system is calculated in real time according to the final power control parameter, and specifically includes:

[0036] A discrete sequence set is constructed, and the discrete sequence set includes an engine discrete sequence, a motor discrete sequence and a battery discrete sequence;

[0037] The instantaneous optimization constraint condition of the ECMS is set based on the discrete sequence set;

[0038] Based on the instantaneous optimization constraint condition and the final power control parameter, an instantaneous optimization objective function is constructed with the minimum equivalent fuel consumption of the vehicle power system as the target;

[0039] The instantaneous optimization objective function is solved to obtain the optimal control sequence of the vehicle power system at the current time.

[0040] The technical effects of the present application are:

[0041] The application proposes an A-ECMS energy management method based on predicted road traffic information. At the current time, the A-ECMS energy management strategy based on working condition recognition is used to adjust the power control parameters to optimize power distribution; the power control parameters of the front road section are obtained by introducing the front road traffic flow information for working condition recognition; the optimal energy trajectory for the whole journey is obtained by using a simplified online dynamic programming strategy through the introduction of the global driving cycle information of the whole journey; finally, based on the deep reinforcement learning algorithm, the current power control parameters, the power control parameters of the front road section and the optimal energy trajectory for the whole journey are used as state quantities, and the final power control parameters are used as action quantities to train the deep network, and the final power control parameters are calculated in real time. The energy management strategy combines the advantages of various energy management strategies, can cope with complex and variable traffic environment, has low computational complexity, strong robustness, and has a forward-looking and global view, and can realize an energy management strategy close to global optimization.

[0042] The application combines the advantages of various energy management strategies, uses the energy management strategy based on working condition recognition to improve the working condition adaptability of the energy management strategy, and has more robustness than other prediction algorithms; in terms of driving field of view, instantaneous optimization based on the current driving state, short-term prediction optimization based on forward road information and global energy trajectory planning based on whole journey traffic flow information are considered respectively, and the determination of the final power control parameters is combined with the deep reinforcement learning (DRL) algorithm, and the combination of long-term and short-term strategies helps to realize an energy management close to global optimization. BRIEF DESCRIPTION OF DRAWINGS

[0043] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only constitute some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0044] The drawings that form a part of the present application are used to provide a further understanding of the present application, the illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation on the present application. In the drawings:

[0045] Figure 1 The flowchart is developed for the offline part in the embodiments of the present application;

[0046] Figure 2 The flowchart is developed for the principal component analysis in the embodiments of the present application;

[0047] Figure 3 Flowchart for k-means clustering algorithm in embodiments of the present application;

[0048] Figure 4 Flowchart for online energy trajectory planning based on predictive road information in embodiments of the present application;

[0049] Figure 5 Flowchart for reinforcement learning training and application in embodiments of the present application;

[0050] Figure 6 Architectural diagram for overall strategy in embodiments of the present application;

[0051] Figure 7 Flowchart for driving scene style recognition and power parameter control selection in embodiments of the present application;

[0052] Figure 8 Flowchart for discrete sequence calculation in embodiments of the present application. DETAILED DESCRIPTION

[0053] Various illustrative embodiments of the present application are described below in detail, which should not be considered limiting on the present application, but rather as being illustrative of certain aspects, features and embodiments of the present application.

[0054] It should be understood that the terms used in the present application merely describe particular embodiments and are not intended to limit the present application. In addition, for numerical ranges in the present application, it should be understood that each intermediate value between the upper limit and the lower limit of the range is specifically disclosed. Each smaller range between any stated value or stated range and any other stated value or stated range is also included in the present application. The upper and lower limits of these smaller ranges can be independently included or excluded from the range.

[0055] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. Although methods are described herein, any method similar or equivalent to those described herein can be used in the practice or testing of the present application. All documents mentioned in this specification are incorporated by reference to disclose and describe the methods associated with the documents. In the event of conflict between the content of this specification and any incorporated document, the content of this specification controls.

[0056] Many modifications and variations of this application specification can be made without departing from the scope or spirit of the application, which would be apparent to one of skill in the art. Other embodiments of the application will be apparent to those skilled in the art from consideration of the specification and practice of the application. The specification and examples of this application are exemplary only.

[0057] The terms “include,” “including,” “have,” “contain,” etc., used in this article are all open-ended terms, meaning that they include but are not limited to.

[0058] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0059] Example 1

[0060] like Figure 1 - Figure 8 As shown, this embodiment provides an A-ECMS energy management method (Adaptive-Equivalent Consumption Minimization Strategy) based on predictive road traffic information, including: acquiring historical driving data of the vehicle and calculating feature parameters of the historical driving data; inputting the feature parameters into an offline driving data recognition library to obtain current optimized power control parameters; wherein, the offline driving data recognition library is a multi-dimensional power parameter optimization library including different driving styles, driving scenarios, initial SOC and target SOC; acquiring predictive road data ahead of the vehicle and inputting the predictive road data into the offline driving data recognition library for scene recognition to obtain the predicted scene and the corresponding predicted optimized power control parameters; using an online dynamic programming algorithm to calculate the long-term energy trajectory; based on the DRL algorithm, combining the current optimized power control parameters, the predicted optimized power control parameters, the long-term energy trajectory and the current energy trajectory to calculate the final power control parameters; and calculating the optimal control sequence of the vehicle power system in real time based on the final power control parameters.

[0061] In order to effectively utilize the big data platform and real-time traffic flow information, improve the ability of energy management strategy to cope with complex working conditions and random traffic environment, and make it close to the global optimal effect, the patent proposes an A-ECMS energy management method based on predicted road traffic information. At the current time, the A-ECMS energy management strategy based on working condition recognition is used to adjust the power control parameters to optimize power distribution; by introducing the front road traffic flow information for working condition recognition, the power control parameters of the front road section are obtained; by introducing the global driving cycle information of the whole driving process, the simplified online dynamic planning strategy is used to obtain the optimal energy trajectory; finally, based on the deep reinforcement learning algorithm, the current power control parameters, the front road section power control parameters and the optimal energy trajectory are used as state quantities, and the final power control parameters are used as action quantities to train the deep network, and applied to real-time calculation of the final power control parameters. The energy management strategy combines the advantages of various energy management strategies, can cope with complex and variable traffic environment, has low computational complexity, strong robustness, and has forward-looking and global vision, and can realize the energy management strategy close to global optimal.

[0062] Figure 1 Develop the offline part process, including the development of driving style and driving scene library and classification recognition system based on offline big data learning, and the development of driving style and scene recognition power control parameter optimization library based on offline global optimization.

[0063] 1. Development of driving style and driving scene library and classification recognition system based on offline big data learning

[0064] The offline big data learning model aims to establish a typical driving scene library and driving style library to provide data support and data processing basis for online work. The learning model includes data analysis, unsupervised scene clustering and driving style clustering, clustering effect evaluation, classification recognition training and other parts. M language is used to develop in MATLAB environment, and mathematical tools such as PCA principal component analysis, k-means unsupervised clustering algorithm and distance discrimination method are used to establish online driving scene recognizer and driving style recognizer. The clustering effect is evaluated by contour coefficient, Calinski-Harabasz index and other clustering evaluation indexes; the accuracy of driving recognizer is evaluated by root mean square error and other mathematical evaluation indexes.

[0065] Offline big data learning analyzes the driving related data provided by the database, and obtains various driving styles and driving scene types through algorithm clustering to establish a typical driving scene library and driving style library. The original driving data obtained through testing is obtained from the database, which covers most of the feature parameters that can represent driving style and driving scene in the driving process, which is shown in the following table:

[0066] Table 1 Driving scene feature parameters

[0067] Serial number Driving scene characteristic parameter 1 Maximum positive acceleration 2 Positive acceleration standard deviation 3 Positive acceleration average value 4 Acceleration time proportion 5 Maximum negative acceleration 6 Negative acceleration standard deviation 7 Negative acceleration average value 8 Negative acceleration time proportion 9 Maximum vehicle speed 10 Vehicle speed standard value 11 Average vehicle speed 12 Average acceleration 13 Acceleration standard deviation 14 Cruise section average vehicle speed 15 Cruise time proportion

[0068] Table 2 Driving style characteristic parameters

[0069]

[0070]

[0071] First, the data is processed, the abnormal values in the data are completed, the noise is cleaned, and the data is normalized to eliminate the interference of data magnitude difference on clustering. PCA principal component analysis is used to reduce the dimension of characteristic parameters, and the cumulative contribution threshold is set. Select N principal components with cumulative contribution exceeding the threshold as clustering parameters to reduce the number of clustering parameters, make the clustering parameters easy to calculate and representative. The principal component analysis process is shown in Figure 2 ;

[0072] In the figure, the coefficient matrix C is:

[0073]

[0074] Subsequently, the CH index and the DB index are used to evaluate the clustering effect of the characteristic parameter matrix under different numbers of clustering centers. The calculation formula of the CH index is

[0075]

[0076] Where B k is the inter-class covariance matrix, W k is the intra-class data covariance matrix, and the detailed formula is

[0077]

[0078]

[0079] Where c q represents the center point of class q, c e represents the center point of the data set, n q represents the number of data in class q, and C q represents the data set of class q. The calculation formula of the DB index is

[0080]

[0081] Where s i represents the sample point dispersion in the class, and M ij is the distance dispersion of the two-class clustering centers.

[0082] After calculating the CH index and the DB index, the best number of cluster centers is selected comprehensively, and a k-means unsupervised clustering algorithm is used for clustering to obtain different types of driving scene (for example, congestion scene, high-speed scene, etc.) and driving style (for example, aggressive style, moderate style, etc.) in the database, and the cluster center coordinates of each type of driving scene and driving style are obtained. The process of the k-means clustering algorithm and the related calculation formula are shown in Figure 3

[0083] Finally, by evaluating the recognition accuracy and calculation real-time of various supervised learning recognition algorithms (for example, neural network method, distance discrimination method, etc.), a suitable recognition algorithm is selected to construct the nonlinear relationship between the feature parameters and the driving scene and the driving style, and to establish the driving scene library and the driving style library. The driving scene library and the driving style library obtained by clustering will be used for training of the supervised learning recognition algorithm (for example, neural network method, distance discrimination method, etc.), and an online driving scene recognizer and a driving style recognizer are established. Among them, the library inputs of the driving scene library and the driving style library are driving scene feature parameters and driving style feature parameters respectively, and the library outputs are corresponding driving scene and driving style, as shown in Table 3;

[0084] Table 3 Library type and library output

[0085]

[0086]

[0087] 2. Driving style and scene recognition dynamic control parameter optimization library based on offline global optimization

[0088] An offline global optimization algorithm is used to optimize the dynamic control parameters of different driving styles and driving scenes, so that the energy management strategy can be adjusted according to the current driving characteristics in real time. M language is used to develop in the Matlab environment, and intelligent optimization algorithm is used to globally optimize the equivalent factor for typical driving scenes, typical driving styles, initial SOC, and target SOC, to establish a dynamic control parameter optimization library.

[0089] Firstly, typical driving conditions of each driving scene and driving style type are constructed as the target of the global optimization algorithm. In the driving scene segment library, connect several scene segments closest to each type of scene cluster center to construct the typical scene of the type of driving scene. In the driving style segment library, select the feature parameters of each type of driving style cluster center as the driver feature followed by the driving condition. Between 30% and 80%, set multiple initial SOC and target SOC as the start and end SOC of the intelligent optimization algorithm.

[0090] ​Then, the intelligent optimization algorithm is used to iteratively optimize the typical driving conditions with different driving styles, initial SOC and target SOC, with the minimum comprehensive fuel consumption as the target, to obtain the corresponding optimized power control parameters, and to establish a multi-dimensional power parameter optimization library for different driving styles, driving scenarios, initial SOC and target SOC.

[0091] 3. Online energy trajectory planning based on predictive road information

[0092] During driving, the predictive road information provides the traffic flow information of the front road section for energy management. The intelligent energy management system receives the global traffic flow information, predicts the possible driving data changes when the vehicle enters the front road section, identifies and plans the driving scenarios and global node energy information of the front road section, as shown in the above figure. Among them, the front road section is the road section within a certain mileage window in front, and during driving, the front road section window constantly moves to the future. The identification method of the front road section driving scenario uses offline part.

[0093] The intelligent energy management system receives the traffic flow information of the front road section that will be entered, predicts the possible driving data changes when the vehicle enters the front road section, and adjusts the power control parameters in advance to optimize energy management. Based on the traffic flow information of the front road section, the online dynamic planning global optimization algorithm is used to plan the SOC trajectory of the vehicle reaching the front road section in advance. The update frequency of the energy trajectory planning is consistent with the update frequency of obtaining the predictive road information, and when the traffic flow of the front road section changes, the SOC trajectory is re-planned to adapt to the change of the traffic scenario. The required predictive road information is shown in Table 4.

[0094] Table 4 Predictive road information

[0095]

[0096] Based on the map navigation information, the traffic flow information (such as average speed) of the front navigation road section in the future period of time can be obtained, which is used for planning the long-term optimal SOC trajectory in online dynamic planning. The online dynamic planning plans the long-term SOC trajectory with the average speed of each stage as the demand. The discrete settings of online dynamic planning are shown in Table 5;

[0097] Table 5 Discrete settings of online dynamic planning

[0098]

[0099]

[0100] The control variables are:

[0101]

[0102] Among them, is the average battery power of the kth stage, is the average engine power of the kth stage, is the average engine speed of the kth stage.

[0103] The state transition equation is:

[0104]

[0105] where E m is the battery open-circuit voltage, R int is the battery internal resistance, Q c is the total capacity of the battery, SOC k is the SOC of the kth stage, SOC k+1 is the SOC of the (k+1)th stage.

[0106] The constraint condition is:

[0107]

[0108] where SOC min and SOC max are the lower and upper limits of SOC, P batt / chg and P batt / dischg are the battery charging power and the battery discharging power, T min and T max are the lower and upper limits of engine torque, N min and N max are the lower and upper limits of engine speed. The cost function is

[0109] V L = V eng / L (T eng ,N eng ) + V batt / L (P batt )

[0110] where V eng / L and V batt / L are the engine fuel consumption and the battery equivalent fuel consumption, V L is the comprehensive fuel consumption.

[0111] 4. Power control parameter regulation strategy based on deep reinforcement learning

[0112] The power control parameter adjustment strategy is trained using a deep reinforcement learning (DRL) algorithm to obtain the final power control parameters of A-ECMS, as shown in the following formula:

[0113] s final (t) = F(st (t), s pre (t), SOC ref (t), SOC t (t))

[0114] where s final is the A-ECMS final power control parameter, s t is the power control parameter based on short-term driving cycle recognition, which enables the optimal energy consumption at the current moment; s pre is the power control parameter corresponding to the front road section driving cycle based on predictive road information recognition, which enables the short-term optimal performance in the foreseeable future; SOC ref is the long-term energy trajectory obtained based on online dynamic programming, which provides global energy consumption optimal guidance for energy management strategy; SOC t is the current moment SOC, and F is the power control parameter adjustment strategy trained using deep reinforcement learning algorithm.

[0115] The state space of deep reinforcement learning is as follows:

[0116] x(t) = [s t (t), s pre (t), SOC ref (t), SOC t (t)] T

[0117] For each state combination x t of the state space, the corresponding action a t in the action space is defined as the possible A-ECMS final equivalent factor s final ;

[0118] A = [a1, a2, …, a m , …, a n-1 , a n ] T

[0119] For each action of the agent, the environment returns the corresponding reward r t+1 :

[0120] r t+1 = V eng / L (T eng , N eng ) + V batt / L (P batt )

[0121] where V eng / L and V batt / L are the engine fuel consumption and the battery equivalent fuel consumption, respectively, and V L is the comprehensive fuel consumption, T engand N eng are engine torque and speed, respectively, P batt is the battery power.

[0122] The environment for reinforcement learning includes four parts, i.e., a hybrid propulsion system model, an A-ECMS energy management strategy model, an energy trajectory prediction model based on predictive road information, and a driving condition recognition system model.

[0123] Figure 5 The flow of reinforcement learning training and application is described. Using historical driving speed curves, in each learning set, the environment randomly extracts a speed curve and the corresponding predictive road information (extracted from the map API) from the historical data with which the agent interacts. In the agent model, the upper and lower limits of the battery SOC are set to control the energy consumption level of the entire route. When the battery SOC exceeds the upper limit or is lower than the lower limit, the environment will perform an early termination event.

[0124] 5. Online A-ECMS control strategy based on predictive road information

[0125] The A-ECMS control strategy based on predictive road information combines offline big data learning, global optimization, power control parameter adjustment strategy based on deep reinforcement learning, and predictive energy trajectory planning with the A-ECMS strategy, so that the optimization effect of the hybrid electric vehicle energy management strategy breaks through the limitation of instantaneous optimization and approaches the global optimum.

[0126] During driving, the final power control parameter scheme combines the optimization results of driving condition recognition and predictive road information, and uses the power control parameter adjustment strategy based on deep reinforcement learning to output the final power control parameter, thereby improving the adaptability of the energy management system to the driving scenario and making the energy management strategy have a certain road predictability. The overall strategy architecture diagram is shown in Figure 6 .

[0127] First, the characteristic parameters are calculated by collecting historical short-term driving data, the characteristic parameters of the historical driving data are calculated, and input to the driving scene recognizer and driving style recognizer to identify the current driving scene and driving style. The optimized power control parameters s t corresponding to the current driving scene, driving style, current SOC, and target SOC are searched. The collected driving scene data is also uploaded to the cloud for scene extraction calculation to periodically update (OTA) the typical driving scene and driving style library.

[0128] Subsequently, the traffic flow information of the road section within a certain mileage window ahead is obtained based on the predictive road information, input to the driving scene recognizer to identify the driving scene of the road section within a certain mileage ahead, and the corresponding optimized power control parameters s preAt the same time, online dynamic programming is used to obtain the future long-term energy trajectory SOC ref .

[0129] Finally, the final power control parameter s is output by using the power control parameter adjustment strategy based on the deep reinforcement learning algorithm final to the A-ECMS energy control strategy, so as to optimize the power distribution of the hybrid power system. The calculation formula of the final power control parameter is as follows

[0130] s final (t)=F(s t (t),s pre (t),SOC ref (t),SOC t (t))

[0131] The A-ECMS control target calculation model calculates the energy optimization distribution of the hybrid system power architecture in multiple energy sources in real time, while meeting the driving demand and power system constraints, so as to maximize the optimization of energy utilization efficiency and reduce fuel consumption.

[0132] Based on the DRL (Deep Reinforcement Learning) algorithm, the current optimized power control parameter, the short-term predicted optimized power control parameter, the long-term energy trajectory and the current energy trajectory are used as the reinforcement learning state quantity, and the final power control parameter is used as the action. The deep network is trained offline, and the final power control parameter is calculated in real time by applying it in real time. According to the final power control parameter, the optimal control sequence of the vehicle power system is calculated in real time based on the ECMS control strategy.

[0133] Firstly, the vehicle demand torque is calculated by analyzing the driving intention, accelerator pedal position, vehicle speed, battery state and motor characteristics. The optimal working point is calculated by the ECMS algorithm, which has the characteristics of fast calculation speed and real-time control, and the total equivalent fuel consumption is used as the optimization target. The overall structure is as follows:

[0134] (1) Construct a discrete sequence:

[0135] Firstly, the current driving state of the vehicle such as vehicle speed, engine speed, etc. can be discretely obtained all engine working point sequence, corresponding to the formula as follows:

[0136] (T EngVec ,ω EngVec )=f(v,ω Eng )

[0137] In the formula, T EngVec , ω EngVecThe torque sequence and the corresponding speed sequence obtained from the engine discrete analysis together constitute the engine discrete sequence; v is the current vehicle speed; ω Eng This refers to the current engine speed of the vehicle.

[0138] Secondly, by obtaining the engine discrete sequence, the driver's required torque (wheel-end required torque), and the current vehicle speed, the discrete sequence point of the motor can be determined, and the corresponding formula is as follows:

[0139] (T MotVec ,ω MotVec )=f(T EngVec ,ω EngVec ,v,T req )

[0140] In the formula, T MotVec ,ω MotVec The torque sequence and the corresponding speed sequence obtained from the motor discretization together constitute the motor discrete sequence; T req This refers to the torque required by the driver (wheel-end torque).

[0141] Finally, by using the discrete engine sequence, motor sequence, vehicle speed, and driver-demanded torque (wheel-end torque) obtained through discretization, the possible discrete sequence points of the battery can be calculated, as shown in the following formula:

[0142] P BattVec =f(T) EngVec ,ω EngVec ,T MotVec ,ω MotVec ,v,T req )

[0143] In the formula, P BattVec These are discrete sequence points of battery power, i.e., the discrete sequence of the battery.

[0144] Overall schematic diagram as follows Figure 8 As shown:

[0145] (2) Instantaneous optimization constraints of ECMS:

[0146] The constraints are mainly divided into three parts: engine module constraints, motor constraints, and battery module constraints. The total engine torque must satisfy the external characteristic constraints; the motor torque must satisfy the external characteristic constraints; and the battery must satisfy the charge / discharge limit power constraints. The ECMS constraint conditions are shown in the following equation:

[0147]

[0148] In the formula, ω Mot ω represents the real-time rotational speed of the electric motor, expressed in rpm. Eng Engine command speed, rpm; TMax Nm; P is the maximum torque at the corresponding speed, Nm; BattMaxDch P is the maximum discharge power of the battery, W; P BattMaxch P is the maximum charge power of the battery;

[0149] (3) Constructing the instantaneous optimization objective function of ECMS

[0150] In order to enable the battery pack to be fully discharged in the entire cycle working condition, thereby reducing the pure engine fuel consumption, the equivalent factor can be adjusted, and the Hamilton function is as follows:

[0151]

[0152]

[0153] In the formula, P is the engine fuel consumption rate, kg / s; P is the battery equivalent engine fuel consumption rate, kg / s; s final (t) is the instantaneous fuel equivalent factor, Q lhv P is the low heat value of fuel, kj / kg; P batt (t) is the battery power, kw; p(SOC(t)) is the SOC penalty function.

[0154] The SOC of the battery is corrected by using the SOC penalty function, and the SOC is controlled in the interval with smaller internal resistance and higher efficiency, so as to eliminate the biased use of electric quantity in the control strategy. The expression of p(SOC(t)) is as follows:

[0155]

[0156] In the formula, SOC max SOC min SOC is the upper and lower limit of the battery SOC, and e is the penalty function index.

[0157] (4) Solving the optimal working point:

[0158] When the Hamilton function H reaches the minimum value, the system obtains the optimal solution:

[0159]

[0160] In the formula, P is the equivalent fuel consumption, kg; u * (t) is the optimal control sequence at t;

[0161] The above description is only the preferred embodiment of the present application, but the protection scope of the present application is not limited to this, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An A-ECMS energy management method based on predictive road traffic information, characterized in that, The method comprises the following steps: obtaining historical driving data of a vehicle, and calculating a characteristic parameter of the historical driving data; inputting the characteristic parameter into an offline driving data recognition library to obtain a current optimized power control parameter; wherein the offline driving data recognition library is a multi-dimensional power parameter optimization library including different driving styles, driving scenes, initial SOC and target SOC; inputting the characteristic parameter into the offline driving data recognition library to obtain the current optimized power control parameter, specifically comprising: inputting the characteristic parameter into the offline driving data recognition library for scene recognition and style recognition to obtain a typical driving scene and a driving style corresponding to the characteristic parameter, and the offline driving data recognition library outputs corresponding current energy trajectory, target energy trajectory and current optimized power control parameter based on the recognized typical driving scene and driving style; the construction process of the offline driving data recognition library comprises: obtaining original data, wherein the original data includes driving related data, and performing data cleaning and normalization processing on the original data to obtain preprocessed original data; performing principal component analysis on the preprocessed original data to obtain reduced dimension data; selecting a plurality of data points in the reduced dimension data as initial clustering centers for clustering analysis to obtain driving scene type data and driving style type data with different characteristic parameters; constructing a nonlinear relationship between the characteristic parameters and the driving scene and the driving style based on a supervised learning recognition algorithm, and establishing a driving scene library and a driving style library; the characteristic parameters include driving scene characteristic parameters and driving style characteristic parameters; the library inputs of the driving scene library and the driving style library are respectively the driving scene characteristic parameters and the driving style characteristic parameters, and the outputs are corresponding driving scenes and driving styles; setting a plurality of initial energy trajectories and target energy trajectories, and based on an intelligent optimization algorithm, iteratively optimizing typical driving conditions with different driving styles, initial energy trajectories and target energy trajectories to obtain corresponding optimized power control parameters, and establishing a multi-dimensional power parameter optimization library of different driving styles, driving scenes, initial SOC and target SOC; obtaining predictive road data in front of the vehicle, inputting the predictive road data into the offline driving data recognition library for scene recognition to obtain a prediction scene and corresponding prediction optimized power control parameter; using an online dynamic programming algorithm, calculating a long-term energy trajectory based on the whole predictive road traffic information, and the specific calculation formula is: ； the constraint condition is: ； the cost function is: ； wherein, V is the open circuit voltage of the battery, R is the internal resistance of the battery, C is the total capacity of the battery, SOk is the SOC of the kth stage, SOC(k+1) is the SOC of the (k+1)th stage, Pbat(k) is the average battery power of the kth stage, Peng(k) is the average engine power of the kth stage, Neng(k) is the average engine speed of the kth stage; and SOL and SOL are the lower and upper limits of the SOC, respectively, and Pbat and Pbat are the battery charging power and the battery discharging power, respectively, and Teng and Teng are the lower and upper limits of the engine torque, respectively, and Neng and Neng are the lower and upper limits of the engine speed, respectively, and Ceng and Cbat are the engine fuel consumption and the battery equivalent fuel consumption, respectively, C is the overall fuel consumption. based on the DRL algorithm, combining the current optimized power control parameter, the prediction optimized power control parameter, the long-term energy trajectory and the current energy trajectory to calculate a final power control parameter; according to the final power control parameter, real-time calculating an optimal control sequence of the vehicle power system. The predictive road data includes a green wave speed range, a remaining mileage of the entire current driving route, a remaining required time of the entire current driving route, average speeds of road segments of the entire driving route, congestion conditions of the road segments of the entire driving route, distances of the road segments at the average speeds of the road segments of the entire driving route, distances of congestion road segments, and passing times of the congestion road segments.

2. The A-ECMS energy management method based on predictive road traffic information according to claim 1, characterized in that, The driving scene feature parameters include a maximum positive acceleration, a positive acceleration standard deviation, a positive acceleration average, an acceleration time proportion, a maximum negative acceleration, a negative acceleration standard deviation, a negative acceleration average, a negative acceleration time proportion, a maximum vehicle speed, a vehicle speed standard value, an average vehicle speed, an average acceleration, an acceleration standard deviation, a cruise section average vehicle speed, and a cruise time proportion.

3. The A-ECMS energy management method based on predictive road traffic information according to claim 1, characterized in that, The driving style feature parameters include a maximum positive acceleration, a positive acceleration standard deviation, a positive acceleration average, a positive jerk standard deviation, a positive jerk average, an acceleration time proportion, a maximum negative acceleration, a negative acceleration standard deviation, a negative acceleration average, a negative acceleration time proportion, a negative jerk standard deviation, a negative jerk average, a maximum vehicle speed, an average vehicle speed, an average acceleration, an acceleration standard value, a cruise section average vehicle speed, a cruise time proportion, an accelerator pedal standard deviation, an accelerator pedal change rate standard deviation, a brake pedal standard deviation, a brake pedal change rate standard deviation, an average jerk, a jerk standard deviation.

4. The A-ECMS energy management method based on predictive road traffic information according to claim 1, characterized in that, Based on the DRL algorithm, the final power control parameter is calculated based on the current optimized power control parameter, the predicted optimized power control parameter, the long-term energy trajectory, and the current energy trajectory, and a specific calculation formula is as follows: ； wherein, is a A-ECMS final power control parameter, is a power control parameter based on short-term driving condition recognition; is a power control parameter corresponding to the front road section driving condition based on predictive road information recognition; is a long-term energy trajectory obtained based on online dynamic programming; is a current time SOC, is a power control parameter adjustment strategy trained using a deep reinforcement learning algorithm.

5. The A-ECMS energy management method based on predictive road traffic information according to claim 1, characterized in that, According to the final power control parameter, an optimal control sequence of the vehicle power system is calculated in real time, and the specific calculation includes: A discrete sequence set is constructed, and the discrete sequence set includes an engine discrete sequence, a motor discrete sequence, and a battery discrete sequence; An instantaneous optimization constraint condition of the ECMS is set based on the discrete sequence set; Based on the instantaneous optimization constraint condition and the final power control parameter, an instantaneous optimization objective function is constructed with the minimum equivalent fuel consumption of the vehicle power system as the target; The instantaneous optimization objective function is solved to obtain the optimal control sequence of the vehicle power system at the current time.

Images