System and application method for cooperative optimization of vehicle speed planning and energy management strategy

Abstract

The present application relates to the field of vehicle control, more particularly, to a system and application method based on cooperative optimization of vehicle speed planning and energy management strategy, wherein the system comprises: an offline optimization subsystem and an online application subsystem; wherein the offline optimization subsystem comprises: a scene pre-analysis module and a multi-dimensional dynamic programming algorithm module; the online application subsystem comprises: a double-LSTM model module and a vehicle-mounted control module; the scene pre-analysis module is used to acquire and analyze the scene of urban continuous intersections; then the multi-dimensional dynamic programming algorithm module is used to solve the optimal vehicle speed sequence and fuel cell control sequence with the best comprehensive performance as the training data set; the double-LSTM model module is used to obtain the economic vehicle speed and fuel cell output power; and the vehicle-mounted control module is used to obtain the power battery output power. Thus, the external interference factors from the driving scenes such as traffic lights and preceding vehicles are considered, and the economic vehicle speed determines the optimal output power of the fuel cell and the power battery at each moment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of vehicle control, and more particularly, to a system and application method for cooperative optimization of speed planning and energy management strategy. BACKGROUND

[0002] Fuel cell hybrid electric vehicles have diversity of energy sources and working modes. Therefore, it is crucial to achieve efficient energy management of fuel cell hybrid electric vehicles. However, both traditional rule-based energy management strategies and optimization-based energy management strategies are greatly affected by the actual driving environment. For example, the degree of traffic congestion, the location and period of traffic lights, pedestrians, rainy and snowy weather, and traffic random events. Such traffic information has the characteristics of wide sources, complex data, and even incomplete traffic information. Therefore, most current technologies focus on energy management strategies that integrate traffic information, that is, cooperative optimization of speed planning and energy management strategy. However, the diversification of the power system of fuel cell vehicles brings great challenges to the cooperative optimization of speed planning and energy management strategy, because there are two power sources in the power system, and the energy management strategy needs to determine the power distribution between the fuel cell and the power battery at each time according to the planned speed. In the joint optimization problem, the increase of state variables and control variables leads to the exponential increase of the difficulty and computational burden of solving. In addition, due to the simplification of the model, external disturbances from traffic signals and preceding vehicles and other driving scenarios are often ignored in the cooperative optimization process. Therefore, most cooperative optimization strategies are suitable for free-flow driving, but not practical for online implementation in urban driving scenarios.

[0003] The prior art discloses a power distribution method and device for a hybrid vehicle, and a vehicle controller. The power distribution method comprises the following steps: acquiring the current speed and the current demand torque of the hybrid vehicle; determining the current SOC value of the power battery when the driving mode of the hybrid vehicle is in parallel mode, and determining the current demand power of the hybrid vehicle; and distributing the output power of the engine and the working power of the motor generator according to the current SOC value of the power battery, the current demand power, and the engine economy line power of the hybrid vehicle. However, this method still ignores external disturbances from traffic signals and preceding vehicles and other driving scenarios, thereby affecting the power distribution performance. SUMMARY

[0004] The present application aims to disclose a system and application method for cooperative optimization of speed planning and energy management strategy with better performance and considering external factors.

[0005] In order to achieve the above-mentioned purpose, the present application provides a system and application method for cooperative optimization of speed planning and energy management strategy, comprising:

[0006] The off-line optimization subsystem and the on-line application subsystem; wherein the off-line optimization subsystem comprises: a scene pre-analysis module and a multi-dimensional dynamic programming algorithm module; the on-line application subsystem comprises: a double-LSTM model module and a vehicle-mounted control module;

[0007] The scene pre-analysis module: obtains the scene of the urban continuous intersection, and pre-analyzes the scene of the urban continuous intersection to obtain reasonable green wave passing time window and vehicle data, and obtains the driving speed interval of the green wave passing time window;

[0008] The multi-dimensional dynamic programming algorithm module: constructs a speed planning and energy management collaborative optimal control problem according to the driving speed interval and the vehicle data, and solves the optimal control problem by using a multi-dimensional dynamic programming algorithm to obtain a comprehensive performance optimal speed sequence and fuel cell control sequence as a training data set;

[0009] The double-LSTM model module: obtains the economic speed and fuel cell output power according to the training data set and the vehicle data;

[0010] The vehicle-mounted control module: obtains the power battery output power according to the economic speed and the fuel cell output power, and controls the vehicle according to the fuel cell output power and the power battery output power.

[0011] Further, the scene pre-analysis module comprises: a traffic intersection queue prediction submodule and a time and speed pre-analysis submodule;

[0012] The traffic intersection queue prediction submodule: obtains the length of the queued vehicles and the dissipation time of the queued vehicles according to the scene of the urban continuous intersection;

[0013] The time and speed pre-analysis submodule: obtains the green wave passing time window and calculates the driving speed interval according to the length of the queued vehicles and the dissipation time of the queued vehicles.

[0014] Further, the traffic intersection queue prediction submodule comprises: in the scene of the urban continuous intersection, when the traffic flow encounters a red light at the intersection, a queue will be formed successively behind the stop line, generating a stop wave When the signal light switches to green, a start wave is generated The meeting time of the start wave and the stop wave marks the end of the queue caused by the red light phase, and the distance from the stop line to the farthest waiting vehicle at this time is the farthest distance of the queue The time taken from the green light turning on to the last queued vehicle driving off the stop line is the queue dissipation time The formula is as follows:

[0015]

[0016]

[0017]

[0018]

[0019] wherein, , denote the flow and density of free-flow traffic flow, , denote the flow and density of saturated traffic flow, denote the density of congested traffic flow, are the times when the signal light switches to green and red, respectively.

[0020] Further, the time and speed pre-analysis submodule comprises:

[0021] The calculation formula of the travel time interval of the vehicle arriving at the intersection is as follows:

[0022]

[0023]

[0024] wherein, , are the minimum and maximum travel times of the vehicle on the first road segment; is the length of the first road segment; , are the upper and lower limit values of the vehicle travel speed, the upper limit value is set according to the road speed limit, and the lower limit value is set to prevent the vehicle from driving at an excessively low speed and affecting the driving of the following vehicle; , are the upper and lower limit values of the vehicle acceleration;

[0025] According to the phase and timing information of the red and green lights of each intersection, the estimated road travel time interval and the green light window are intersected to obtain a passable time interval;

[0026] The fastest and slowest times for the vehicle to arrive at the first intersection are ; after selecting the target travel time interval, the upper and lower limit values of the speed that should be driven on each road segment are determined:

[0027]

[0028] .

[0029] Further, the multi-dimensional dynamic programming algorithm module comprises an optimal control problem submodule and a dynamic programming solution algorithm submodule;

[0030] Optimal control problem submodule: used to build the optimal control problem of the fuel cell vehicle speed planning and energy management collaborative control;

[0031] Dynamic programming solution algorithm submodule: solve the optimal control problem, get the speed control sequence and fuel cell output power sequence with optimal cost function performance index value, as training data set.

[0032] Further, the optimal control problem submodule includes:

[0033] Establish the vehicle longitudinal dynamics model:

[0034] =

[0035]

[0036] In the formula is the wheel end drive power, represents the battery output power, represents the fuel cell output power, represents the motor efficiency, represents the vehicle mass (kg); represents the gravitational acceleration ( ); represents the rolling resistance coefficient; represents the road slope angle; represents the air resistance coefficient; represents the wind area ( ); represents the air density ( ); represents the driving speed ( ), represents the rotational mass conversion coefficient, represents the transmission efficiency;

[0037] The Thevenin model is used for mathematical modeling of the battery, and the state of charge of the power battery is calculated according to the ampere-hour integral method SOC :

[0038]

[0039] In the formula, is the initial charge state value of the battery, is the battery capacity, is the battery current;

[0040] The output voltage of the PEMFC single cell is:

[0041]

[0042] wherein is the activation polarization voltage loss, is the ohmic polarization voltage loss, is the concentration polarization voltage loss;

[0043] The single cell voltage of a PEMFC is about 0.7 V;

[0044] The total voltage of a fuel cell stack is:

[0045]

[0046] wherein is the number of monomers in the stack;

[0047] The hydrogen consumption rate of a fuel cell is expressed as:

[0048]

[0049]

[0050] wherein is the molar mass of hydrogen, is the superimposed current, and F is the Faraday constant;

[0051] The fuel cell performance degradation rate formula is:

[0052]

[0053] wherein is the fuel cell performance degradation rate, is the correction factor, , , and represent the duration of low-load operation, high-load operation and fast load change, start-stop cycle, respectively, , , and are the corresponding degradation coefficients;

[0054] The state variables have two dimensions, namely the speed and the state of charge of the power battery ; while the control variables are the motor demand power and the fuel cell output power , and the state transition equation is:

[0055]

[0056] The hydrogen consumption, fuel cell loss, driving time and speed fluctuation are taken as the optimization objectives to build the optimal control problem of the coordinated control of the speed planning and energy management of the fuel cell vehicle:

[0057]

[0058]

[0059] wherein, , and are the hydrogen consumption, fuel cell life and time consumed by the vehicle in the i-th stage, is the absolute value of the speed difference between the current stage and the previous stage, , , , are the weight factors of the corresponding cost items.

[0060] Further, the dynamic programming solving algorithm submodule includes:

[0061] The optimal control problem is converted into a distance-based discrete system, the road is divided into equal parts by dividing the total road length according to the distance interval of , and the number of stages is ; the multi-dimensional dynamic programming algorithm is used for solving: after the state variables are discretized, the state variable matrix is obtained, i.e. , i.e. , and , which are respectively denoted as , and , and the control variable is denoted as and ;

[0062] The dynamic programming is calculated from the final state to the initial state, and the lowest cost of each stage to reach the final state is calculated step by step, and the best decision path is determined based on the lowest cost. The cost function expression of each stage dynamic programming solving is as follows:

[0063]

[0064] is the cumulative optimal cost of the rear sub-route, is the cost function value of the state variable and the control variable in the current stage;

[0065] ​The multi-dimensional dynamic programming algorithm starts from the last stage N and solves backward, first calculates the transition cost of the feasible state point to the N-1 stage, and the feasible state point is determined by the maximum charging and discharging power of the power battery and the passing speed interval obtained by pre-analysis; records the decision of each state point in the N-1 stage to the optimal cost in the N stage, and so on During the whole planning process, the optimal vehicle speed and fuel cell power sequence are generated by the backtracking method; finally Starting from the initial value, the optimal control variable in the current state is indexed forward, and the state value of the next stage is solved through the state transition equation; in this way, the speed sequence and the fuel cell output power sequence that optimize the cost function are finally obtained as the training data set.

[0066] Further, the double-LSTM model module comprises an LSTM1 submodule and an LSTM2 submodule.

[0067] The LSTM1 submodule generates an economic vehicle speed according to the solving rule of the multi-dimensional dynamic programming algorithm for the vehicle speed planning problem obtained from the training data set.

[0068] The network structure of the LSTM1 comprises an input layer, a hidden layer and an output layer; the input layer has four variables, namely the vehicle speed V, the distance L from the vehicle to the next intersection, the time T from the current time to the next intersection signal light switching to green, and the length of the next intersection green light ; the four variables are input in the form of a sequence; the hidden layer comprises an LSTM layer and a fully connected layer, and the LSTM layer can capture the long-term dependence relationship in the time sequence data; the output layer is the vehicle speed at the next time ;

[0069] The LSTM2 submodule is used to generate the fuel cell output power.

[0070] The network structure of the LSTM2 comprises an input layer, a hidden layer and an output layer; the input of the LSTM2 network is the vehicle speed V, the state of charge S of the power battery and the acceleration A, wherein the acceleration is obtained according to the economic vehicle speed generated by the LSTM1, and the final output is the fuel cell power at the current time .

[0071] Further, the solving rule of the multi-dimensional dynamic programming algorithm for the vehicle speed planning problem obtained from the training data set comprises: interpolating and resampling the time sequence information in all training data sets according to a time interval , and reorganizing the resampled time sequence information according to each group of n time sequences in order, and taking the reorganized training data set as the input of the LSTM1 submodule.

[0072] The application also provides an application method of a system based on cooperative optimization of vehicle speed planning and energy management strategies, comprising:

[0073] S1: obtaining the scene of the urban continuous intersection through the scene pre-analysis module, pre-analyzing the scene of the urban continuous intersection, obtaining reasonable green wave passing time windows and vehicle data, and obtaining the driving speed interval of the green wave passing time windows;

[0074] S2: constructing a cooperative optimal control problem of vehicle speed planning and energy management according to the driving speed interval and the vehicle data through the multi-dimensional dynamic programming algorithm module, and solving the vehicle speed sequence and the fuel cell control sequence with optimal comprehensive performance by using the multi-dimensional dynamic programming algorithm as the training data set;

[0075] S3: obtaining the economic vehicle speed and the fuel cell output power according to the training data set and the vehicle data through the double-LSTM model module;

[0076] S4: obtaining the power battery output power according to the economic vehicle speed and the fuel cell output power through the vehicle-mounted control module, and controlling the vehicle according to the fuel cell output power and the power battery output power.

[0077] Compared with the prior art, the beneficial effects of the technical scheme of the application are:

[0078] The scene pre-analysis module is used to obtain and analyze the scene of the urban continuous intersection, so that external interference factors from driving scenes such as traffic lights and preceding vehicles are considered; then the multi-dimensional dynamic programming algorithm module is used to solve the vehicle speed sequence and the fuel cell control sequence with optimal comprehensive performance as the training data set; the double-LSTM model module is used to obtain the economic vehicle speed and the fuel cell output power; and the vehicle-mounted control module is used to obtain the power battery output power according to the economic vehicle speed and the fuel cell output power. Thus, the optimal output power of the fuel cell and the power battery at each moment is determined in combination with the driving scene and the economic vehicle speed. BRIEF DESCRIPTION OF DRAWINGS

[0079] Figure 1 It is a system block diagram based on cooperative optimization of vehicle speed planning and energy management strategies according to example one;

[0080] Figure 2 It is a vehicle cloud layered architecture diagram according to example two;

[0081] Figure 3 It is a schematic diagram of scene pre-analysis according to example two;

[0082] Figure 4 It is a fuel cell vehicle configuration diagram according to example two;

[0083] Figure 5is a schematic diagram of the solving process of the multi-dimensional dynamic programming algorithm described in embodiment two;

[0084] Figure 6 is an LSTM1 model structure described in embodiment two;

[0085] Figure 7 is an LSTM2 model structure described in embodiment two;

[0086] Figure 8 is an application method flowchart of the system based on speed planning and energy management strategy collaborative optimization described in embodiment one; DETAILED DESCRIPTION

[0087] The accompanying drawings are only for illustrative purposes and cannot be understood as limiting the patent;

[0088] The technical solutions of the present application will be further described below in combination with the drawings and embodiments.

[0089] Embodiment one:

[0090] The present embodiment provides a system and application method based on speed planning and energy management strategy collaborative optimization as shown in Figure 1 The system and application method based on speed planning and energy management strategy collaborative optimization comprises:

[0091] In order to achieve the above purpose, the present application provides a system and application method based on speed planning and energy management strategy collaborative optimization, comprising:

[0092] An offline optimization subsystem and an online application subsystem; wherein the offline optimization subsystem comprises a scene pre-analysis module and a multi-dimensional dynamic programming algorithm module; the online application subsystem comprises a double-LSTM model module and a vehicle-mounted control module;

[0093] The scene pre-analysis module: obtains the scene of the city continuous intersection, and pre-analyzes the scene of the city continuous intersection to obtain reasonable green wave passing time window and vehicle data, and obtains the driving speed interval of the green wave passing time window;

[0094] The multi-dimensional dynamic programming algorithm module: constructs a speed planning and energy management collaborative optimal control problem according to the driving speed interval and the vehicle data, and solves the optimal speed sequence and the fuel cell control sequence with the best comprehensive performance by using the multi-dimensional dynamic programming algorithm as the training data set;

[0095] The double-LSTM model module: obtains the economic speed and the fuel cell output power according to the training data set and the vehicle data;

[0096] The vehicle-mounted control module: obtains the power battery output power according to the economic speed and the fuel cell output power; and controls the vehicle according to the fuel cell output power and the power battery output power.

[0097] The embodiment obtains and analyzes the scene of the urban continuous intersection through the scene pre-analysis module, so as to consider the external interference factors from the driving scene of the traffic signal and the preceding vehicle; then the multi-dimensional dynamic programming algorithm module is used to solve the vehicle speed sequence and the fuel cell control sequence with the optimal comprehensive performance as the training data set; the economic vehicle speed and the fuel cell output power are obtained through the double-LSTM model module; the power battery output power is obtained through the vehicle-mounted control module according to the economic vehicle speed and the fuel cell output power. Thus, the optimal output power of the fuel cell and the power battery at each moment is determined in combination with the driving scene and the economic vehicle speed.

[0098] Embodiment two:

[0099] The embodiment is further disclosed on the basis of embodiment one:

[0100] The system cloud hierarchical architecture diagram based on the speed planning and energy management strategy cooperative optimization in the embodiment is shown in Figure 2 .

[0101] The scene pre-analysis module includes a traffic intersection queuing prediction submodule and a time and speed pre-analysis submodule.

[0102] The traffic intersection queuing prediction submodule: according to the scene of the urban continuous intersection, the length of the queuing vehicles at the intersection and the dissipation time of the queuing vehicles are obtained.

[0103] The time and speed pre-analysis submodule: according to the length of the queuing vehicles at the intersection and the dissipation time of the queuing vehicles, the green wave passing time window is obtained and the driving speed interval is calculated.

[0104] The traffic intersection queuing prediction submodule includes: in the scene of the urban continuous intersection, when the traffic flow encounters a red light at the intersection, a queuing will be formed successively behind the stop line, generating a stop wave ; when the signal light is switched to green, a start wave is generated ; the meeting time of the start wave and the stop wave marks the end of the queuing caused by the red light phase, and the distance from the stop line to the farthest waiting vehicle at this time is the farthest distance of the queuing ; the time spent from the green light on to the moment when the last queuing vehicle drives away from the stop line is the queuing dissipation time ; the formula is as follows:

[0105]

[0106]

[0107]

[0108]

[0109] wherein, , denote the flow and density of free-flow traffic, , denote the flow and density of saturated traffic, denote the density of congested traffic, are the times when the signal light switches to green and red, respectively.

[0110] The time and speed pre-analysis submodule includes:

[0111] The calculation formula of the travel time interval of the vehicle arriving at the intersection is as follows:

[0112]

[0113]

[0114] wherein, , are the minimum and maximum travel times of the vehicle on the first road segment; is the length of the first road segment; , are the upper and lower limit values of the vehicle travel speed, the upper limit value is set according to the road speed limit, and the lower limit value is set to prevent the vehicle from driving at an excessively low speed and affecting the driving of the following vehicle; , are the upper and lower limit values of the vehicle acceleration;

[0115] As shown in FIG. 4, according to the phase and timing information of the red and green lights of each intersection, the estimated road travel time interval is intersected with the green light window to obtain a passable time interval; Figure 3 The fastest and slowest times of the vehicle arriving at the first

[0116] intersection are ; after selecting the target travel time interval, the upper and lower limit values of the speed that should be driven on each road segment are determined:

[0117]

[0118] .

[0119] The multi-dimensional dynamic programming algorithm module includes an optimal control problem submodule and a dynamic programming solution algorithm submodule;

[0120] The optimal control problem submodule is used to construct the optimal control problem of the fuel cell vehicle speed planning and energy management cooperative control;

[0121] ​Dynamic programming solving algorithm submodule: solve the optimal control problem, get the speed control sequence and fuel cell output power sequence with the optimal cost function performance index value, as the training data set.

[0122] The optimal control problem submodule includes:

[0123] The research object of the embodiment is a fuel cell hybrid electric vehicle with a "fuel cell + power battery" configuration, as shown in FIG. 1. In this configuration, the fuel cell mainly meets the continuous power demand of the vehicle, while the battery supplements the peak and transient power demand, and recovers regenerative energy during braking. This configuration can weaken the power fluctuation of the fuel cell and reduce hydrogen consumption, thereby improving the service life of the fuel cell and the economy of the whole vehicle. Figure 4

[0124] A vehicle longitudinal dynamics model is established:

[0125]

[0126]

[0127] In the formula, is the wheel-end driving power, represents the battery output power, represents the fuel cell output power, represents the motor efficiency, represents the mass of the vehicle (kg); represents the gravitational acceleration ( ); represents the rolling resistance coefficient; represents the road slope angle; represents the air resistance coefficient; represents the wind area ( ); represents the air density ( ); represents the driving speed ( ), represents the rotational mass conversion coefficient, represents the transmission efficiency;

[0128] The Thevenin model is used for mathematical modeling of the battery, and the state of charge of the power battery is calculated according to the ampere-hour integral method: SOC

[0129]

[0130] In the formula, is the initial charge state value of the battery, is the battery capacity, is the battery current;​​​

[0131] The single cell output voltage of PEMFC is:

[0132]

[0133] wherein is the activation polarization voltage loss, is the ohmic polarization voltage loss, is the concentration polarization voltage loss;

[0134] The single cell voltage of PEMFC is about 0.7 V;

[0135] The total voltage of fuel cell stack is:

[0136]

[0137] wherein is the number of single cells in the stack;

[0138] The hydrogen consumption rate of fuel cell is expressed as:

[0139]

[0140]

[0141] wherein is the molar mass of hydrogen, is the superimposed current, and F is the Faraday constant;

[0142] The fuel cell performance attenuation rate formula is:

[0143]

[0144] wherein is the fuel cell performance attenuation rate, is the correction factor, , , and represent the duration of low load operation, high load operation and fast load change, start-stop cycle, respectively, , , and are the corresponding attenuation coefficients;

[0145] The state variables have two dimensions, namely the speed and the state of charge of the power battery ; and the control variables are the motor demand power and the fuel cell output power , the state transition equation is:

[0146]

[0147] The optimal control problem of the coordinated control of the speed planning and energy management of the fuel cell vehicle is constructed by taking the comprehensive performance of the hydrogen consumption, fuel cell loss, driving time and speed fluctuation as the optimization objective:

[0148]

[0149]

[0150] wherein, , and are the hydrogen consumed, fuel cell life and time consumed by the vehicle in the i-th stage, is the absolute value of the speed difference between the current stage and the previous stage, , , , , are the weight factors of the corresponding cost items.

[0151] The dynamic programming solving algorithm submodule includes:

[0152] The optimal control problem is converted into a discrete system based on distance, the total road length is divided according to the distance interval of , the road is divided into equal parts, and the number of stages is ; as shown in Figure 5 , a multi-dimensional dynamic programming algorithm is used for solving: after discretization of the state variables, a state variable matrix is obtained, i.e. , i.e. , and , respectively denoted as and , and the control variable is denoted as and ; The dynamic programming is calculated from the final state to the initial state, and the lowest cost of each stage to reach the final state is calculated step by step, and the best decision path is determined based on the lowest cost. The cost function expression of each stage dynamic programming solving is as follows:

[0153]

[0154]

[0155] is the cumulative optimal cost of the rear sub-route, is the cost function value of the state variable and control variable in the current stage;​​

[0156] The multi-dimensional dynamic programming algorithm is solved from the last stage N in reverse, first calculates the transition cost of the feasible state point to the N-1 stage, and the feasible state point is determined by the maximum charging and discharging power of the power battery and the passing speed interval obtained by pre-analysis; records the decision of each state point of the N-1 stage to the optimal cost of the N stage, and so on During the whole planning process, the optimal vehicle speed and fuel cell power sequence are generated by the backtracking method; finally The initial value is used to index the optimal control variable in the current state, and the state value of the next stage is solved through the state transition equation; in this way, the speed sequence and the fuel cell output power sequence with the optimal cost function are finally obtained as the training data set.

[0157] The double-LSTM model module comprises an LSTM1 submodule and an LSTM2 submodule.

[0158] The LSTM1 submodule generates an economic vehicle speed according to the solution rule of the vehicle speed planning problem based on the multi-dimensional dynamic programming algorithm obtained from the training data set.

[0159] The network structure of the LSTM1 is shown in FIG. 1 and comprises an input layer, a hidden layer and an output layer. Figure 6 The input layer has four variables, namely, the vehicle speed V, the distance L from the vehicle to the next intersection, the time T from the current time to the next time when the signal light of the next intersection changes to green, and the phase length of the green light of the next intersection. The four variables are input in the form of a sequence; the hidden layer comprises an LSTM layer and a fully connected layer, and the LSTM layer can capture the long-term dependence relationship in the time sequence data; and the output layer is the vehicle speed at the next time.

[0160] The LSTM2 submodule is used to generate the fuel cell output power.

[0161] The network structure of the LSTM2 is shown in FIG. 2 and comprises an input layer, a hidden layer and an output layer. Figure 7 The input of the LSTM2 network is the vehicle speed V, the state of charge S of the power battery and the acceleration A, wherein the acceleration is obtained according to the economic vehicle speed generated by the LSTM1, and the final output is the fuel cell power at the current time. .

[0162] The solution rule of the vehicle speed planning problem based on the multi-dimensional dynamic programming algorithm obtained from the training data set comprises: interpolating and resampling the time sequence information in all the training data sets according to a time interval , and recombining the resampled time sequence information according to each group of n time sequences in order, and taking the recombined training data set as the input of the LSTM1 submodule.​

[0163] The embodiment obtains and analyzes the scene of the urban continuous intersection through the scene pre-analysis module, so as to consider the external interference factors from the driving scene of the traffic signal and the preceding vehicle; then the multi-dimensional dynamic programming algorithm module is used to solve the vehicle speed sequence and the fuel cell control sequence with the optimal comprehensive performance as the training data set; the economic vehicle speed and the fuel cell output power are obtained through the double LSTM model module; the power battery output power is obtained through the vehicle-mounted control module according to the economic vehicle speed and the fuel cell output power. Thus, the optimal output power of the fuel cell and the power battery at each moment is determined in combination with the driving scene and the economic vehicle speed.

[0164] Embodiment three

[0165] The embodiment provides an application method of a system based on speed planning and energy management strategy cooperative optimization as shown in Figure 8 The application method comprises the following steps:

[0166] S1: obtaining the scene of the urban continuous intersection through the scene pre-analysis module, and pre-analyzing the scene of the urban continuous intersection to obtain reasonable green wave passing time windows and vehicle data, and obtain the driving speed interval of the green wave passing time windows;

[0167] S2: constructing a speed planning and energy management cooperative optimal control problem according to the driving speed interval and the vehicle data through the multi-dimensional dynamic programming algorithm module, and solving the vehicle speed sequence and the fuel cell control sequence with the optimal comprehensive performance by using the multi-dimensional dynamic programming algorithm as the training data set;

[0168] S3: obtaining the economic vehicle speed and the fuel cell output power according to the training data set and the vehicle data through the double LSTM model module;

[0169] S4: obtaining the power battery output power according to the economic vehicle speed and the fuel cell output power through the vehicle-mounted control module; and controlling the vehicle according to the fuel cell output power and the power battery output power.

[0170] The embodiment obtains and analyzes the scene of the urban continuous intersection through the scene pre-analysis module, so as to consider the external interference factors from the driving scene of the traffic signal and the preceding vehicle; then the multi-dimensional dynamic programming algorithm module is used to solve the vehicle speed sequence and the fuel cell control sequence with the optimal comprehensive performance as the training data set; the economic vehicle speed and the fuel cell output power are obtained through the double LSTM model module; the power battery output power is obtained through the vehicle-mounted control module according to the economic vehicle speed and the fuel cell output power. Thus, the optimal output power of the fuel cell and the power battery at each moment is determined in combination with the driving scene and the economic vehicle speed.

[0171] Obviously, the above embodiments of the present application are merely exemplary but not intended to limit the embodiments of the present application. Based on the above description, any other variations or changes can be made by those skilled in the art without departing from the spirit and principles of the present application. It is not necessary to list all the embodiments here. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present application shall fall within the scope of the claims of the present application.

Claims

1. A system based on the coordinated optimization of vehicle speed planning and energy management strategies, characterized in that, include: Offline optimization subsystem and online application subsystem; The offline optimization subsystem includes a scene pre-analysis module and a multi-dimensional dynamic programming algorithm module; the online application subsystem includes a dual LSTM model module and an on-board control module. Scene pre-analysis module: acquires the scene of continuous intersections in the city, performs pre-analysis on the scene of continuous intersections in the city, obtains a reasonable green wave passage time window and vehicle data, and obtains the driving speed range of the green wave passage time window; Multidimensional dynamic programming algorithm module: Based on the driving speed range and vehicle data, construct the optimal control problem of coordinated vehicle speed planning and energy management, and use the multidimensional dynamic programming algorithm to solve the vehicle speed sequence and fuel cell control sequence with the best comprehensive performance, which are used as training datasets; Dual LSTM model module: Obtains economic vehicle speed and fuel cell output power based on training dataset and vehicle data; Onboard control module: Calculates the power battery output power based on the economic vehicle speed and fuel cell output power; controls the vehicle based on the fuel cell output power and power battery output power. The scenario pre-analysis module includes: a traffic intersection queue prediction sub-module and a time and speed pre-analysis sub-module; Traffic intersection queue prediction submodule: Based on the scenario of continuous intersections in the city, it obtains the length of queued vehicles and the time for queued vehicles to dissipate. Time and speed pre-analysis submodule: Based on the length of vehicles queuing at the intersection and the time it takes for the vehicles to dissipate, the green wave passage time window is obtained and the driving speed range is calculated.

2. The system based on the collaborative optimization of vehicle speed planning and energy management strategies according to claim 1, characterized in that, The traffic intersection queue prediction submodule includes: In the scenario of continuous intersections in the city, when the traffic flow encounters a red light at the intersection, it will gradually form a queue after the stop line, generating a stop wave v. stop When the traffic light turns green, a starting wave v is generated. start The moment the starting wave and the stopping wave meet marks the end of the queue caused by the red light phase. At this point, the distance from the stop line to the farthest waiting vehicle is the farthest queue distance D. end The time from when the green light turns on until the last vehicle in the queue leaves the stop line is called the queue dissipation time T. dissipate The formula is as follows: In the formula, q0 and k0 represent the flow rate and density of traffic flow in the free-moving state, and q m k m k represents the flow rate and density at saturation. j t represents the density under congested traffic conditions. g t r These are the times when the traffic lights switch to green and red, respectively.

3. The system based on the collaborative optimization of vehicle speed planning and energy management strategies according to claim 1, characterized in that, The time and velocity pre-analysis submodules include: The formula for calculating the travel time interval to the intersection is as follows: In the formula, t low t high Let L be the minimum and maximum travel time for a vehicle to travel on the i-th road segment; i v is the length of the i-th road segment; max v min These are the upper and lower limits for vehicle speed. The upper limit is set according to the road speed limit, and the lower limit is set to prevent vehicles from driving at excessively low speeds and affecting the driving of vehicles behind them; a max a min These are the upper and lower limits of vehicle acceleration. Based on the traffic light phase and timing information at each intersection, the estimated travel time interval of the road segment is intersected with the green light window to obtain the passable time interval. The fastest and slowest times to reach the i-th intersection are: After selecting the target travel time range, determine the upper and lower speed limits for each road segment:

4. The system based on the collaborative optimization of vehicle speed planning and energy management strategies according to claim 1, characterized in that, The multidimensional dynamic programming algorithm module includes: an optimal control problem submodule and a dynamic programming solution algorithm submodule; Optimal Control Problem Submodule: Used to construct the optimal control problem for the coordinated control of vehicle speed planning and energy management in fuel cell vehicles; The dynamic programming algorithm submodule solves the optimal control problem, obtaining the speed control sequence and fuel cell output power sequence with the optimal cost function performance index, which serve as the training dataset.

5. The system based on the collaborative optimization of vehicle speed planning and energy management strategies according to claim 4, characterized in that, The optimal control problem submodule includes: Establish a longitudinal dynamics model for the vehicle: In the formula P drive It is the wheel-end drive power, P b P represents the battery output power. fc Indicates the output power of the fuel cell, η m The motor efficiency is represented by m, the car mass (kg), and g is the acceleration due to gravity (m / s²). 2 ); f represents the rolling resistance coefficient; α represents the road slope angle; C d A represents the air drag coefficient; A represents the frontal area (m²). 2 ); ρ represents air density (kg / m³) 3 v represents the travel speed (m / s), δ represents the rotational mass conversion factor, and η represents the rotational mass conversion factor. t Indicates transmission efficiency; The Thevenin model is used for mathematical modeling of the battery, and the state of charge (SOC) of the power battery is calculated using the ampere-hour integral method: In the formula, SOC(t0) is the initial state of charge of the battery, and Q... bat For battery capacity, I b This refers to the battery current. The PEMFC cell output voltage is: E cell =E Nernst -IN act -IN ohm -IN co In the formula U act It is the activation polarization voltage loss, U ohm It is the ohmic polarization voltage loss, U co It is the concentration polarization voltage loss; The total voltage of the fuel cell stack is: The fc =n cel And cel Where n cell It is the number of individual cells in the battery stack; Hydrogen consumption rate of fuel cells Represented as: in I is the molar mass of hydrogen. fc For superimposed currents, F is the Faraday coefficient; The formula for fuel cell performance degradation rate is: D fc =k p (k1t1-k2t2-k3t3-k4n) Where D fc It is the fuel cell performance degradation rate, k p These are correction factors. t1, t2, t3, and n represent the durations of low-load operation, high-load operation, and rapid load change and start-stop cycle, respectively. k1, k2, k3, and k4 are the corresponding attenuation coefficients. The state variables have two dimensions: speed v(k) and battery state of charge (SOC)(k); while the control variable is the motor power demand P. d and fuel cell output power P fc The state transition equation is: With the comprehensive performance of hydrogen consumption, fuel cell losses, driving time, and speed fluctuations as the optimization objective, an optimal control problem for the coordinated control of vehicle speed planning and energy management in fuel cell vehicles is constructed: In the formula, D fc (k) and t trip Let be the hydrogen consumed by the vehicle in stage k, the fuel cell lifespan, and the time; |v(k+1)-v(k)|] be the absolute value of the speed difference between the current stage and the previous stage; and λ1, λ2, λ3, and λ4 be the weighting factors for the corresponding cost items.

6. The system based on the coordinated optimization of vehicle speed planning and energy management strategies according to claim 4, characterized in that, The dynamic programming solution algorithm submodule includes: The optimal control problem is transformed into a distance-based discrete system. The total road length is divided into N-1 equal parts according to distance intervals Δs, with the number of stages k = 1, 2, ..., N. A multidimensional dynamic programming algorithm is used to solve the problem. After discretizing the state variables, an m1×m2 state variable matrix is ​​obtained, consisting of m1 V values ​​and m2 SOC values, denoted as and respectively. i = 1, 2, ..., m1 and j = 1, 2, ..., m2, the control variable is denoted as as well as Dynamic programming works backward from the final state to the initial state, calculating the minimum cost to reach the final state at each stage. Based on the minimum cost, the optimal decision path is determined. The cost function expressions for each stage of the dynamic programming solution are as follows: The cumulative optimal cost of the subsequent sub-routes. It is the cost function value of the state variables and control variables at the current stage; The multidimensional dynamic programming algorithm solves the problem backwards from the final stage N. First, it calculates the transition cost to the feasible state point in stage N-1, which is determined by the maximum charge / discharge power of the power battery and the speed range obtained from the pre-analysis. It records the decision of each state point in stage N-1 to the optimal cost in stage N, and so on, solving the entire planning process k = N-1…2,1. The optimal vehicle speed and fuel cell power sequence are generated by backtracking. Finally, starting from the initial value of k=1, the optimal control variable in the current state is indexed forward, and the state value of the next stage is solved by the state transition equation. In this way, the speed sequence and fuel cell output power sequence with the optimal cost function are finally obtained as the training dataset.

7. The system based on the collaborative optimization of vehicle speed planning and energy management strategies according to claim 1, characterized in that, The dual LSTM model module includes: an LSTM1 submodule and an LSTM2 submodule; The LSTM1 submodule: Based on the training dataset, it obtains the solution rules for the vehicle speed planning problem using a multidimensional dynamic programming algorithm, and generates an economical vehicle speed. The LSTM1 network structure includes an input layer, hidden layers, and an output layer. The input layer has four variables: vehicle speed V, distance L from the vehicle to the next intersection, time T from the current moment to the next intersection's traffic light turning green, and phase length T of the green light at the next intersection. g All four variables are input as sequences; the hidden layers include LSTM layers and fully connected layers, with the LSTM layers capturing long-term dependencies in the time series data; the output layer is the vehicle speed V at the next time step. t+1 ; LSTM2 submodule: Used to generate fuel cell output power; The LSTM2 network structure includes an input layer, hidden layers, and an output layer. The inputs to the LSTM2 network are vehicle speed V, battery state of charge S, and acceleration A, where the acceleration is obtained from the economic vehicle speed generated by LSTM1. The final output is the fuel cell power P at the current moment. fc .

8. The system based on the coordinated optimization of vehicle speed planning and energy management strategies according to claim 7, characterized in that, Based on the training dataset, the solution rules for the vehicle speed planning problem based on the multidimensional dynamic programming algorithm are as follows: interpolate and resample the temporal information in all training datasets according to the time interval Δt, and reassemble the resampled temporal information into groups of n time series in order, and use the reassembled training dataset as the input of the LSTM1 submodule.

9. An application method of the system based on the collaborative optimization of vehicle speed planning and energy management strategies as described in claim 1, characterized in that, include: S1: Obtain the scene of continuous intersections in the city through the scene pre-analysis module, and perform pre-analysis on the scene of continuous intersections in the city to obtain a reasonable green wave passage time window and vehicle data, and obtain the driving speed range of the green wave passage time window. S2: The multidimensional dynamic programming algorithm module constructs a coordinated optimal control problem of vehicle speed planning and energy management based on the driving speed range and vehicle data. The multidimensional dynamic programming algorithm is used to solve the vehicle speed sequence and fuel cell control sequence with the best comprehensive performance, which are used as training datasets. S3: The economic vehicle speed and fuel cell output power are obtained by using the dual LSTM model module based on the training dataset and vehicle data. S4: The power battery output power is obtained by the on-board control module based on the economic vehicle speed and the fuel cell output power. The vehicle is controlled based on the output power of the fuel cell and the output power of the power battery; The scenario pre-analysis module includes: a traffic intersection queue prediction sub-module and a time and speed pre-analysis sub-module; Traffic intersection queue prediction submodule: Based on the scenario of continuous intersections in the city, it obtains the length of queued vehicles and the time for queued vehicles to dissipate. Time and speed pre-analysis submodule: Based on the length of vehicles queuing at the intersection and the time it takes for the vehicles to dissipate, the green wave passage time window is obtained and the driving speed range is calculated.

Images