Construction method, control method and system for hierarchical control model of plug-in hybrid power truck

Abstract

The invention discloses a construction method, a control method and a system for a hierarchical control model of a plug-in hybrid power truck, and the construction method comprises the steps: constructing an energy control model of a plug-in hybrid power vehicle based on an improved DDPG; determining a state set S and an action set action based on an improved DDPG plug-in hybrid electric vehicle energy control model, and a reward function J and a constraint function based on adaptive change of truck driving data; step 2, collecting a standard working condition and a gradient, taking the standard working condition and the gradient as a sample set, inputting the sample set into the energy control model of the plug-in hybrid electric vehicle based on the improved DDPG constructed in the step 1 for training, and obtaining a trained energy control model of the plug-in hybrid electric vehicle based on the improved DDPG; coordination processing is carried out on a surveying and mapping map of a mining platform, a mixed integer nonlinear programming model Ttotal is established, a driving path of a main queue and a driving path of a single vehicle are obtained through solving, and the transportation planning requirement can be met. Therefore, the problems of mining area intersection congestion, electric shovel transportation waiting, low information island decision-making efficiency and the like are effectively solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy vehicle design, and relates to a hierarchical coordination optimization and energy management control method, and specifically to a construction method, a control method and a system for a hierarchical control model of a plug-in hybrid power truck. Background Art

[0003] For a long time, there have been three major pain points in mining transportation, which can be simply described as safety, efficiency, and cost. Autonomous truck platoons provide an opportunity to solve these pain points. In a convoy, the closer distance between vehicles helps improve the aerodynamic performance of the autonomous vehicles behind, thereby greatly reducing energy consumption. At the same time, the automated process of "mining-transporting-dispatching" can achieve 24-hour continuous operation, making the efficiency of automated operations reach or even exceed the manual level, thereby realizing the dispatch management from a single truck to a group and then to a systematic one. Summary of the invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a method for constructing a hierarchical control model for a plug-in hybrid truck, a control method and a system to solve the technical problem of high energy consumption in the mining area transportation process in the prior art.

[0005] In order to solve the above technical problems, the present invention adopts the following technical solutions to achieve the above problems:

[0006] A method for constructing a plug-in hybrid electric vehicle control model specifically comprises the following steps:

[0007] Step 1: construct an energy control model of a plug-in hybrid electric vehicle based on an improved DDPG, and determine a state set S, an action set action, a reward function J and a constraint function that are adaptively changed based on truck driving data.

[0008] S={v i (t),acc i ,SoC i (t),SoH i (t)},i=1,2...I

[0009]

[0010] in:

[0011] v i (t) Vehicle speed along the main queue’s driving path, km / h;

[0012] acc i is the acceleration of truck i, m / s 2 ;

[0013] SOC i (t) is the power of truck i at time t;

[0014] SOH i (t) is the battery health status value of truck i at time t;

[0015] is the engine output power of truck i, kW;

[0016] α is the weight of fuel consumption;

[0017] β is the weight of battery charge maintenance;

[0018] γ is the weight of battery degradation cost;

[0019] fuel(t) is the fuel consumption at time t;

[0020] elec(t) is the energy consumption at time t;

[0021] SOH ref It is the reference value of battery health status;

[0022] SoC ref (t) is the spatial domain index SoC when the driving distance is d(t);

[0023] λ is the SoC decrease rate in the spatial domain;

[0024] SoC ini For the initial SoC;

[0025] SoC tgt For the target SoC;

[0026] L is the expected driving range after the battery is fully charged;

[0027] eq ini is the initial value of the equivalence factor;

[0028] eq′ is the value of the equivalent factor when the slope is 0;

[0029] gr is the slope value;

[0030] gr min is the minimum slope value;

[0031] gr max is the maximum slope value;

[0032] The constraint function based on the improved DDPG plug-in hybrid electric vehicle energy control model is as follows:

[0033]

[0034] in:

[0035] SoC is the battery power;

[0036] SoC min is the minimum value of SoC;

[0037] SoC max is the maximum value of SoC;

[0038] x = eng, MG1, MG2, eng is the engine, MG1 is the motor, MG2 is the generator;

[0039] T x is the torque of the power unit x, N·m;

[0040] is the minimum torque of the power unit x, N·m;

[0041] is the maximum torque of the power unit x, N·m;

[0042] w x is the speed of the power unit x, rpm;

[0043] is the minimum speed of the power unit x, rpm;

[0044] is the maximum speed of the power unit x, rpm;

[0045] Step 2: Collect standard working conditions and slopes as sample sets, and input the sample sets into the improved DDPG plug-in hybrid electric vehicle energy control model constructed in step 1 for training, so as to obtain a trained improved DDPG plug-in hybrid electric vehicle energy control model.

[0046] A control method for a plug-in hybrid electric truck platoon comprises the following steps:

[0047] Get the slope of the mining platform and the speed v of truck i i (t), and input it into the improved DDPG plug-in hybrid vehicle energy control model according to claim 1 to obtain the engine output power of truck i The truck engine output power The engine speed and torque are calculated by inputting them into the engine's optimal fuel operating curve, and the engine speed and torque are respectively input into the motor transmission path and the generator transmission path to calculate the corresponding motor speed and torque and generator speed and torque.

[0048] The slope of the mining platform and the speed v of the truck i are obtained.i (t), specifically comprising the following steps:

[0049] Step 1, obtaining a survey map of the mining platform where the truck queue performs the task, coordinate processing the survey map of the mining platform, and collecting the slope of the mining platform;

[0050] Step 2: Establish a mixed integer nonlinear programming model T according to the coordinates of the survey map of the mining platform obtained in step 1 total , solve the mixed integer nonlinear programming model T total Get the driving path of the main queue and the driving path of truck i;

[0051] The driving path of the main queue is the coordinate l where the truck leaves the main queue i and the coordinates c where the truck merges into the main queue i The driving path of truck i is the coordinate l where truck i leaves the main queue. i , task point coordinates q n The coordinate c where it joins the main queue i The connection;

[0052]

[0053] in:

[0054] T1 is the main queue from the starting point of the mining platform q s The time it takes for the first truck to leave the main queue at coordinate l1, T1 = ||q s -l1|| / v platoon , s;

[0055] v platoon The maximum speed of the main queue, km / h;

[0056] T I The coordinate c of the last truck to perform the task and merge into the main queue I To the end point of the mining platform f The time taken, T I =||q f -c I || / v platoon , s;

[0057] G n The value of indicates whether there are trucks performing tasks at both task point n and task point (n+1). If so, G n Take 1, otherwise G n Take 0;

[0058] j nrepresents the sum of the time when all trucks leave the main queue and the time when no trucks leave the main queue when the main queue passes the task point n, s;

[0059] c i represents the coordinates of the i-th truck merging into the main queue;

[0060] l i represents the coordinates of the i-th truck leaving the main queue;

[0061] Step 3, establish the longitudinal dynamics model of the plug-in hybrid truck as shown below;

[0062]

[0063] in:

[0064] is the speed of truck i at time t, km / h;

[0065] is the acceleration of truck i at time t, m / s 2 ;

[0066] is the differential of the actual torque of truck i at time t;

[0067] v i (t) Vehicle speed along the main queue’s driving path, km / h;

[0068] g is the acceleration due to gravity;

[0069] A is the front area of ​​truck i, m 2 ;

[0070] f i is the rolling resistance coefficient of truck i;

[0071] τ i is the time lag constant of the longitudinal power system of truck i;

[0072] m i is the mass of truck i, kg;

[0073] r i is the wheel radius of truck i, m;

[0074] η i is the transmission system efficiency of truck i;

[0075] C i_air is the air resistance coefficient of truck i;

[0076] ρ is air density, kg / m 2 ;

[0077] u i (t) is the expected torque of truck i at time t, N·m;

[0078] i is the serial number of the truck, i is 0, 1, 2, ...;

[0079] Step 4: construct a DMPC-based longitudinal control model for the queue; determine the expected state variables of each truck in the DMPC-based longitudinal control model for the queue according to the driving path of the main queue obtained in step 2. And the objective function The state variable y of the DMPC-based platoon longitudinal control model is determined according to the plug-in hybrid truck longitudinal dynamics model constructed in step 3. i (t) and the control variable u i (t);

[0080] Step 5: Use the ADMM algorithm to solve the objective function of the DMPC-based queue longitudinal control model constructed in step 4 to obtain the speed v of truck i in the driving path of the main queue. i (t).

[0081] Step 2 specifically includes the following steps:

[0082] Step 2.1, construct the mixed integer nonlinear programming model T total :

[0083]

[0084] in:

[0085] T1 is the main queue from the starting point of the mining platform q s The time it takes for the first truck to leave the main queue at coordinate l1, T1 = ||q s -l1|| / v platoon ;

[0086] v platoon The maximum speed of the main queue;

[0087] T I The coordinate c of the last truck to perform the task and merge into the main queue I To the end point of the mining platform f The time taken, T I =||q f -c I || / v platoon ;

[0088] G n The value of indicates whether there are trucks performing tasks at both task point n and task point (n+1). If so, G n Take 1, otherwise Gn Take 0;

[0089] j n It represents the sum of the time when all trucks leave the main queue and the time when no trucks leave the main queue when the main queue passes through task point n;

[0090] c i represents the coordinates of the i-th truck merging into the main queue;

[0091] l i represents the coordinates of the i-th truck leaving the main queue;

[0092] Step 2.3, use the mixed integer nonlinear programming method to solve the optimal T total , get the coordinates l of the truck leaving the main queue i , the coordinate c where the truck meets the main queue i ;

[0093] Step 2.4: The coordinates l of the truck leaving the main queue obtained in step 2.3 are i and the coordinates c where the truck merges into the main queue i Connect them one by one to get the driving path of the main queue; the coordinate l of the truck leaving the main queue i , task point coordinates q n and the coordinates c where the truck merges into the main queue i Connect them one by one to get I truck driving paths.

[0094] Step 4 specifically includes the following steps:

[0095] Step 4.1: Based on the main queue's driving path obtained in step 2, the position s0(t) and speed v0(t) of the main queue are given, and the expected state variables of each truck are obtained according to the following formula:

[0096]

[0097] in:

[0098] is the expected displacement of the i-th truck, m;

[0099] is the expected speed of the i-th truck, km / h;

[0100] d des is the distance between two adjacent trucks, m;

[0101] Step 4.2: Define the state variable of the plug-in hybrid truck longitudinal dynamics model established in step 3 as the current state y of truck i i (t) = [s i (t),vi (t)] T , the control variable is the torque u of truck i i (t);

[0102] Step 4.3, establish the objective function

[0103]

[0104] in:

[0105] is the predicted state of the i-th truck at time t+k, which includes the position and speed of truck i at time t+k;

[0106] is the predicted torque of the i-th truck at time t+k;

[0107] is the hypothetical state of the i-th truck at time t+k, which includes the position and speed of truck i at time t+k;

[0108] is the assumed state of the j-th truck at time t+k, which includes the position and speed of truck i at time t+k;

[0109] is the assumed torque of the i-th truck at time t+k;

[0110] is the expected state of the i-th truck at time t+k, which includes the position and speed of truck i at time t+k;

[0111] C D is the truck's drag coefficient;

[0112] ρ is the air density, kg / m 2 ;

[0113] A is the frontal area of ​​the truck, m 2 ;

[0114] m i is the mass of the i-th truck, kg;

[0115] g is the acceleration due to gravity;

[0116] f is the rolling resistance coefficient of the truck;

[0117] r i is the wheel radius of truck i, m;

[0118] η is the transmission efficiency of the truck;

[0119] is the penalty weight matrix for the deviation between the actual state and the expected state of the i-th truck at time t;

[0120] is the penalty weight matrix of the error between the actual state and the expected torque of the i-th truck;

[0121] F i (t) is the penalty weight matrix of the deviation between the actual state and the assumed state of the i-th truck;

[0122] is the penalty weight matrix for the deviation between the expected output of the i-th truck and the expected output of its neighboring vehicles. If the vehicle is adjacent to the lead vehicle, then

[0123] d i,j is the distance between the i-th truck and the j-th truck, m.

[0124] Step 5 specifically includes the following steps:

[0125] Step 5.1, for the objective function in step 4.3 Introducing dummy variables and Get the improved objective function L ρ ;

[0126]

[0127] in:

[0128] is the Lagrangian function of the i-th vehicle at time t;

[0129] is the set of constraints for the i-th truck at time t;

[0130] θ i (t) and They are the dual variables corresponding to the i-th truck at time t;

[0131] u i (t) is the torque of the i-th truck at time t;

[0132] ρ1 and ρ2 are penalty coefficients;

[0133] Step 5.2: Improve the objective function L obtained in step 5.1 ρ The dual variable in and Update and obtain the improved objective function L after updating the variables ρ ', using ADMM algorithm to update the improved objective function L after the variables are updated ρ'Solve and get the original residual and the dual residual

[0134]

[0135] in:

[0136] τ is the update step size of the Lagrangian term;

[0137] is the penalty coefficient;

[0138] Step 5.3, determine whether the original residual and the dual residual both satisfy the following formula, if so, the control variable is obtained;

[0139] The control variables include the vehicle speed v under the driving path of the main queue i (t);

[0140]

[0141] in:

[0142] ε pri is the original residual termination condition;

[0143] ε dual is the dual residual termination condition.

[0144] A control system for the plug-in hybrid electric truck platoon, based on the plug-in hybrid electric vehicle control model construction method and the plug-in hybrid electric truck platoon control method, comprising a map processing unit, a model construction unit, a model training unit and an output unit;

[0145] The map processing unit is used to obtain the surveying map of the mining platform where the truck queue performs the task, and coordinate the surveying map of the mining platform to collect the slope of the mining platform;

[0146] The model building unit is used to establish a mixed integer nonlinear programming model T according to the coordinates of the survey map of the mining platform obtained total , solve the mixed integer nonlinear programming model T total Get the driving path of the main queue and the driving path of truck i;

[0147]

[0148] Establish a longitudinal dynamics model for a plug-in hybrid truck;

[0149]

[0150] Constructing a DMPC-based longitudinal control model for a platoon; determining the expected state variables of each truck in the DMPC-based longitudinal control model for a platoon according to the obtained driving path of the main platoon And the objective function The state variable y of the DMPC-based platoon longitudinal control model is determined according to the constructed plug-in hybrid truck longitudinal dynamics model. i (t) and the control variable u i (t);

[0151] The ADMM algorithm is used to solve the objective function of the DMPC-based queue longitudinal control model to obtain the vehicle speed v under the driving path of the main queue. i (t); The second model building unit builds an energy control model of a plug-in hybrid electric vehicle based on an improved DDPG, and determines a state set S, an action set action, a reward function J and a constraint function based on the energy control model of the plug-in hybrid electric vehicle based on the improved DDPG;

[0152] S={v i (t),acc i ,SoC i (t),SoH i (t)},i=1,2...I

[0153]

[0154]

[0155] The model training unit is used to collect standard working conditions and slopes as sample sets, and input the sample sets into the constructed energy control model of the plug-in hybrid electric vehicle based on the improved DDPG for training, so as to obtain the trained energy control model of the plug-in hybrid electric vehicle based on the improved DDPG;

[0156] The output unit is used to convert the vehicle speed v obtained in step 5 i (t) and the slope of the mining platform are input into the trained DDPG plug-in hybrid vehicle energy control model to obtain the engine output power of truck i The truck engine output power The engine speed and torque are calculated by inputting them into the engine's optimal fuel operating curve, and the engine speed and torque are respectively input into the motor transmission path and the generator transmission path to calculate the corresponding motor speed and torque and generator speed and torque.

[0157] Compared with the prior art, the present invention has the following beneficial technical effects:

[0158] (I) In the present invention, the mining platform survey map is processed by coordinate processing to establish a mixed integer nonlinear programming model T total , the driving path of the main queue and the driving path of a single vehicle are solved to meet the transportation planning requirements. This effectively solves the problems of congestion at the mining intersection, waiting time for electric shovel transportation, and inefficient decision-making in information islands.

[0159] (II) In the present invention, a DMPC-based platoon longitudinal control model is constructed and the ADMM algorithm is used to solve the model, thereby achieving good vehicle following of each truck in the platoon under various unidirectional communication topology structures.

[0160] (III) The invention uses the trained DDPG plug-in hybrid electric vehicle energy control model to obtain the power component control variables, improve fuel economy, and reduce the vehicle driving cost. It effectively improves the energy efficiency of PHET in coal mine transportation scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0161] Figure 1 It is a framework diagram of the hierarchical energy management method for plug-in hybrid truck platoons.

[0162] Figure 2 It is a schematic diagram of the variation law of equivalent factor and slope.

[0163] Figure 3 is a graph showing the path planning result of the truck platoon in this embodiment;

[0164] Figure 4 is the absolute position diagram of the truck queue in this embodiment;

[0165] Figure 5 is a diagram showing changes in queue speed under different unidirectional topologies in this embodiment;

[0166] Figure 6 is the operating point distribution diagram of the truck engine in this embodiment;

[0167] Figure 7 is a diagram showing the global SOC change of the lead vehicle PHET1 under various control strategies in this embodiment;

[0168] Figure 8 is a battery degradation variation trend diagram during a driving cycle in this embodiment;

[0169] Fig. 9 is a speed data diagram of a standard working condition in this embodiment;

[0170] Fig.10 It is the slope data diagram in this embodiment.

[0171] The specific contents of the present invention are further explained in detail below in conjunction with embodiments. DETAILED DESCRIPTION

[0172] It should be noted that, unless otherwise specified, all components in the present invention are components known in the art.

[0173] Specific embodiments of the present invention are given below. It should be noted that the present invention is not limited to the following specific embodiments, and all equivalent changes made on the basis of the technical solution of this application fall within the protection scope of the present invention.

[0174] Example:

[0175] This embodiment provides a method for constructing a plug-in hybrid electric vehicle control model, which specifically includes the following steps:

[0176] Step 1: construct an energy control model of plug-in hybrid electric vehicle based on improved DDPG, determine the state set S, action set action, reward function J and constraint function based on adaptive changes of truck driving data based on the improved DDPG plug-in hybrid electric vehicle energy control model;

[0177] S={v i (t),acc i ,SoC i (t),SoH i (t)},i=1,2...I

[0178]

[0179] in:

[0180] v i (t) Vehicle speed along the main queue’s driving path, km / h;

[0181] acc i is the acceleration of truck i, m / s 2 ;

[0182] SOC i (t) is the power of truck i at time t;

[0183] SOH i (t) is the battery health status value of truck i at time t;

[0184] is the engine output power of truck i, kW;

[0185] α is the weight of fuel consumption;

[0186] β is the weight of battery charge maintenance;

[0187] γ is the weight of battery degradation cost;

[0188] fuel(t) is the fuel consumption at time t;

[0189] elec(t) is the energy consumption at time t;

[0190] SOH ref It is the reference value of battery health status;

[0191] SoC ref (t) is the spatial domain index SoC when the driving distance is d(t);

[0192] λ is the SoC decrease rate in the spatial domain;

[0193] SoC ini For the initial SoC;

[0194] SoC tgt For the target SoC;

[0195] L is the expected driving range after the battery is fully charged;

[0196] eq ini is the initial value of the equivalence factor;

[0197] eq′ is the value of the equivalent factor when the slope is 0;

[0198] gr is the slope value;

[0199] gr min is the minimum slope value;

[0200] gr max is the maximum slope value;

[0201] The constraint function of the energy control model of plug-in hybrid electric vehicles based on the improved DDPG is as follows:

[0202]

[0203] in:

[0204] SoC is the battery power;

[0205] SoC min is the minimum value of SoC;

[0206] SoC max is the maximum value of SoC;

[0207] x = eng, MG1, MG2, eng is the engine, MG1 is the motor, MG2 is the generator;

[0208] T x is the torque of the power unit x, N·m;

[0209] is the minimum torque of the power unit x, N·m;

[0210] is the maximum torque of the power unit x, N·m;

[0211] w x is the speed of the power unit x, rpm;

[0212] is the minimum speed of the power unit x, rpm;

[0213] is the maximum speed of the power unit x, rpm;

[0214] Step 2: Collect standard working conditions and slopes as sample sets, and input the sample sets into the improved DDPG plug-in hybrid electric vehicle energy control model constructed in step 1 for training, so as to obtain a trained improved DDPG plug-in hybrid electric vehicle energy control model.

[0215] In the above scheme, the general training process is: starting from the initial state set S, optimizing the action set action, so that the action set action can get higher and higher scores in the reward function J, thereby realizing training. For details, please refer to the authorized Chinese patent CN202211627066.4.

[0216] In this embodiment:

[0217] acc i According to v i (t) Solved to obtain;

[0218] SOC i (t) is the quantity that changes with time during the solution process;

[0219] SOH i (t) is the quantity that changes with time during the solution process;

[0220] SOC ref (t) is Figure 7 In the SoC_ref curve, the SoC decrease rate λ in the spatial domain can be found in Figure 7 get;

[0221] L is the mileage value of the specific working condition;

[0222] eq ini The value of is related to the remaining mileage of the specific working conditions;

[0223] gr min is the minimum value of the slope under specific working conditions;

[0224] gr maxis the maximum value of the slope under specific working conditions;

[0225] The sample set is the standard working condition and slope. The standard working condition includes the curve that changes with time, and its data is as follows Fig. 9 As shown, the slope data is Fig.10 As shown;

[0226] The specific description of each parameter in the energy control model of plug-in hybrid electric vehicle based on the improved DDPG is shown in the table:

[0227]

[0228]

[0229] A control method for a plug-in hybrid electric truck platoon comprises the following steps:

[0230] Get the slope of the mining platform and the speed v of truck i i (t), and input it into the improved DDPG plug-in hybrid vehicle energy control model in claim 1 to obtain the engine output power of truck i The truck engine output power The engine speed and torque are calculated by inputting them into the engine's optimal fuel operating curve, and the engine speed and torque are respectively input into the motor transmission path and the generator transmission path to calculate the corresponding motor speed and torque and generator speed and torque.

[0231] Get the slope of the mining platform and the speed v of truck i i (t), specifically comprising the following steps:

[0232] Step 1, obtaining a survey map of the mining platform where the truck queue performs the task, coordinate processing the survey map of the mining platform, and collecting the slope of the mining platform;

[0233] Step 2: Establish a mixed integer nonlinear programming model T according to the coordinates of the survey map of the mining platform obtained in step 1 total , solve the mixed integer nonlinear programming model T total Get the driving path of the main queue and the driving path of truck i;

[0234] The driving path of the main queue is the coordinate l where the truck leaves the main queue i and the coordinates c where the truck merges into the main queue i The path of truck i is the coordinate l where truck i leaves the main queue. i , task point coordinates q n The coordinate c where it joins the main queue i The connection;

[0235]

[0236] in:

[0237] T1 is the main queue from the starting point of the mining platform q s The time it takes for the first truck to leave the main queue at coordinate l1, T1 = ||q s -l1|| / v platoon , s;

[0238] v platoon The maximum speed of the main queue, km / h;

[0239] T I The coordinate c of the last truck to perform the task and merge into the main queue I To the end point of the mining platform f The time taken, T I =||q f -c I || / v platoon , s;

[0240] G n The value of indicates whether there are trucks performing tasks at both task point n and task point (n+1). If so, G n Take 1, otherwise G n Take 0;

[0241] j n represents the sum of the time when all trucks leave the main queue and the time when no trucks leave the main queue when the main queue passes the task point n, s;

[0242] c i represents the coordinates of the i-th truck merging into the main queue;

[0243] l i represents the coordinates of the i-th truck leaving the main queue;

[0244] Step 3, establish the longitudinal dynamics model of the plug-in hybrid truck as shown below;

[0245]

[0246] in:

[0247] is the speed of truck i at time t, km / h;

[0248] is the acceleration of truck i at time t, m / s 2 ;

[0249] is the differential of the actual torque of truck i at time t;

[0250] v i (t) Vehicle speed along the main queue’s driving path, km / h;

[0251] g is the acceleration due to gravity;

[0252] A is the front area of ​​truck i, m 2 ;

[0253] f i is the rolling resistance coefficient of truck i;

[0254] τ i is the time lag constant of the longitudinal power system of truck i;

[0255] m i is the mass of truck i, kg;

[0256] r i is the wheel radius of truck i, m;

[0257] η i is the transmission system efficiency of truck i;

[0258] C i_air is the air resistance coefficient of truck i;

[0259] ρ is air density, kg / m 2 ;

[0260] u i (t) is the expected torque of truck i at time t, N·m;

[0261] i is the serial number of the truck, i is 0, 1, 2, ...;

[0262] Step 4: Construct a DMPC-based longitudinal control model for the queue; determine the expected state variables of each truck in the DMPC-based longitudinal control model for the queue according to the driving path of the main queue obtained in step 2. And the objective function According to the plug-in hybrid truck longitudinal dynamics model constructed in step 3, the state variable y of the DMPC-based platoon longitudinal control model is determined. i (t) and the control variable u i (t);

[0263] Step 5: Use the ADMM algorithm to solve the objective function of the DMPC-based queue longitudinal control model constructed in step 4 to obtain the speed v of truck i in the driving path of the main queue. i (t).

[0264] In this embodiment, v platoon is the vehicle speed vi The maximum value of (t);

[0265] In this embodiment, j n Collected under specific working conditions;

[0266] In this embodiment, the specific description of each parameter in the longitudinal dynamics model of the plug-in hybrid truck is shown in the table:

[0267] symbol Specific instructions A <![CDATA[In this embodiment, A = 5.2 m 2 > <![CDATA[f i ]]> <![CDATA[In this embodiment, f i = 0.01]]> <![CDATA[m i ]]> <![CDATA[In this embodiment, m i = 17000 kg]]> <![CDATA[r i ]]> <![CDATA[In this embodiment, r i = 0.45 m]]> <![CDATA[η i ]]> <![CDATA[In this embodiment, η i = 98%]]> <![CDATA[C i_air ]]> <![CDATA[In this embodiment, C i_air = 0.7]]> ρ <![CDATA[In this embodiment, ρ = 1.29 kg / m 3 >

[0268] In this embodiment, the starting point of the mining platform is q s =(0,0), the destination of the vehicle queue leaving the mining platform is q f =(50,0) and the coordinates of the task point are q n , the coordinates of the truck leaving the main queue are obtained i and the coordinates c where the truck meets the main queue i , as shown in Table 1; the driving path of the main queue and the driving path of a single vehicle are obtained as follows Figure 4 As shown; the position of the main queue obtained is s i (t) Figure 5 As shown; the speed v of the main queue is obtained i (t) Figure 6 The engine power obtained is shown as Figure 7 As shown;

[0269] Depend on Figure 7 It can be seen that the engine power distribution is relatively concentrated under the present invention, which effectively avoids the phenomenon that some engines are in extreme working conditions. The engine always works stably and evenly in the high fuel consumption rate range; the battery SOC and SOH changes under the control of the improved DDPG plug-in hybrid vehicle energy control model are as follows: Figure 8 and Fig. 9 As shown, Figure 8 The figure shows the global changes of SOC of the lead vehicle PHET1 under various control strategies in a real driving scenario. Figure 8 It can be seen that the present invention achieves a uniform decrease in SOC throughout the driving cycle, indicating that EMS effectively considers the dynamic reference trajectory. The present invention further adds the equivalent factor and battery degradation cost to the cost function, and the intelligent agent sacrifices the reward of the SOC deviation part to achieve the overall optimal driving cost, making the SOC decrease more slowly. Therefore, the result given by the present invention is a representative solution that can reflect the overall optimization performance. Fig. 9 The battery degradation trend during the driving cycle. Under the premise of good thermal management of the battery, assuming that the average temperature inside the battery is constant, it can be seen that the present invention can effectively suppress the health degradation of the battery compared with the most advanced existing strategies.

[0270] Table 1

[0271]

[0272]

[0273] By processing the coordinates of the mining platform survey map, a mixed integer nonlinear programming model T is established. total , the driving path of the main queue and the driving path of a single vehicle are solved to meet the transportation planning requirements. This effectively solves the problems of congestion at the intersection of the mine area, waiting for electric shovel transportation, and inefficient decision-making in information islands; by constructing a queue longitudinal control model based on DMPC and solving the model using the ADMM algorithm, good vehicle following is achieved for each truck in the queue under various one-way communication topology structures. Using the trained energy control model of the plug-in hybrid electric vehicle based on the improved DDPG, the control variables of the power components are obtained, which improves fuel economy, reduces the driving cost of the whole vehicle, and effectively improves the energy efficiency of PHET in coal mine transportation scenarios.

[0274] Step 2 specifically includes the following steps:

[0275] Step 2.1, construct the mixed integer nonlinear programming model T total :

[0276]

[0277] in:

[0278] T1 is the main queue from the starting point of the mining platform q s The time it takes for the first truck to leave the main queue at coordinate l1, T1 = ||q s -l1|| / v platoon ;

[0279] v platoon The maximum speed of the main queue;

[0280] T I The coordinate c of the last truck to perform the task and merge into the main queue I To the end point of the mining platform f The time taken, T I =||q f -c I || / v platoon ;

[0281] G n The value of indicates whether there are trucks performing tasks at both task point n and task point (n+1). If so, G n Take 1, otherwise G n Take 0;

[0282] j n It represents the sum of the time when all trucks leave the main queue and the time when no trucks leave the main queue when the main queue passes through task point n;

[0283] c i represents the coordinates of the i-th truck merging into the main queue;

[0284] l i represents the coordinates of the i-th truck leaving the main queue;

[0285] Step 2.3, use the mixed integer nonlinear programming method to solve the optimal T total , get the coordinates l of the truck leaving the main queue i , the coordinate c where the truck meets the main queue i ;

[0286] Step 2.4: The coordinates l of the truck leaving the main queue obtained in step 2.3 are i and the coordinates c where the truck merges into the main queue i Connect them one by one to get the driving path of the main queue; the coordinate l of the truck leaving the main queue i , task point coordinates q n and the coordinates c where the truck merges into the main queue i Connect them one by one to get I truck driving paths.

[0287] Step 4 specifically includes the following steps:

[0288] Step 4.1: Based on the main queue's driving path obtained in step 2, the position s0(t) and speed v0(t) of the main queue are given, and the expected state variables of each truck are obtained according to the following formula:

[0289]

[0290] in:

[0291] is the expected displacement of the i-th truck, m;

[0292] is the expected speed of the i-th truck, km / h;

[0293] d des is the distance between two adjacent trucks, m;

[0294] Step 4.2: Define the state variable of the plug-in hybrid truck longitudinal dynamics model established in step 3 as the current state y of truck i i (t) = [s i (t),v i (t)] T, the control variable is the torque u of truck i i (t);

[0295] Step 4.3, establish the objective function

[0296]

[0297] in:

[0298] is the predicted state of the i-th truck at time t+k, which includes the position and speed of truck i at time t+k;

[0299] is the predicted torque of the i-th truck at time t+k;

[0300] is the hypothetical state of the i-th truck at time t+k, which includes the position and speed of truck i at time t+k;

[0301] is the assumed state of the j-th truck at time t+k, which includes the position and speed of truck i at time t+k;

[0302] is the assumed torque of the i-th truck at time t+k;

[0303] is the expected state of the i-th truck at time t+k, which includes the position and speed of truck i at time t+k;

[0304] C D is the truck's drag coefficient;

[0305] ρ is the air density, kg / m 2 ;

[0306] A is the frontal area of ​​the truck, m 2 ;

[0307] m i is the mass of the i-th truck, kg;

[0308] g is the acceleration due to gravity;

[0309] f is the rolling resistance coefficient of the truck;

[0310] r i is the wheel radius of truck i, m;

[0311] η is the transmission efficiency of the truck;

[0312] is the penalty weight matrix for the deviation between the actual state and the expected state of the i-th truck at time t;

[0313] is the penalty weight matrix of the error between the actual state and the expected torque of the i-th truck;

[0314] F i (t) is the penalty weight matrix of the deviation between the actual state and the assumed state of the i-th truck;

[0315] is the penalty weight matrix for the deviation between the expected output of the i-th truck and the expected output of its neighboring vehicles. If the vehicle is adjacent to the lead vehicle, then

[0316] d i,j is the distance between the i-th truck and the j-th truck, m.

[0317] In the above technical solution, The value of depends on whether vehicle i can obtain the information of the leading vehicle; if the following vehicle i cannot obtain the information of the leading vehicle, then Q i =0, otherwise Q i >0;

[0318] In the above technical solution, F i (t) , The initial value satisfies Step 5 specifically includes the following steps:

[0319] Step 5.1, for the objective function in step 4.3 Introducing a dummy variable ξ i (t) and Get the improved objective function L ρ ;

[0320]

[0321] in:

[0322] is the Lagrangian function of the i-th vehicle at time t;

[0323] is the set of constraints for the i-th truck at time t;

[0324] θ i (t) and They are the dual variables corresponding to the i-th truck at time t;

[0325] u i(t) is the torque of the i-th truck at time t;

[0326] ρ1 and ρ2 are penalty coefficients;

[0327] Step 5.2: Improve the objective function L obtained in step 5.1 ρ The dual variable in and Update and obtain the improved objective function L after updating the variables ρ ', using ADMM algorithm to update the improved objective function L after the variables are updated ρ 'Solve and get the original residual and the dual residual

[0328]

[0329] in:

[0330] τ is the update step size of the Lagrangian term;

[0331] is the penalty coefficient;

[0332] Step 5.3, determine whether the original residual and the dual residual both satisfy the following formula, if so, the control variable is obtained;

[0333] The control variables include the vehicle speed v of the main queue under the driving path i (t);

[0334]

[0335] in:

[0336] ε pri is the original residual termination condition;

[0337] ε dual is the dual residual termination condition.

[0338] In this embodiment, The value of is updated over time during the solution process;

[0339] In this embodiment, The value of is updated over time during the solution process;

[0340] In this embodiment, θ i (t) and The value of is updated over time during the solution process;

[0341] In this embodiment, ρ1=0.1, ρ2=10 -3 ;

[0342] In this embodiment, τ = 0.1;

[0343] In this embodiment,

[0344] In this embodiment, the original residual termination condition ε pri =10 -3 ;

[0345] In this embodiment, the dual residual termination condition ε dual =10 -3 .

[0346] The present invention provides a control system for a plug-in hybrid electric truck platoon, based on a construction method of a plug-in hybrid electric vehicle control model and a control method for a plug-in hybrid electric truck platoon, comprising a map processing unit, a model construction unit, a model training unit and an output unit;

[0347] The map processing unit is used to obtain the survey map of the mining platform where the truck queue performs the task, coordinate the survey map of the mining platform, and collect the slope of the mining platform;

[0348] The model building unit is used to establish a mixed integer nonlinear programming model T according to the coordinates of the survey map of the mining platform obtained. total , solve the mixed integer nonlinear programming model T total Get the driving path of the main queue and the driving path of truck i;

[0349]

[0350] Establish a longitudinal dynamics model for a plug-in hybrid truck;

[0351]

[0352] Construct a DMPC-based longitudinal control model for the queue; determine the expected state variables of each truck in the DMPC-based longitudinal control model for the queue according to the obtained driving path of the main queue And the objective function According to the constructed plug-in hybrid truck longitudinal dynamics model, the state variable y of the DMPC-based platoon longitudinal control model is determined. i (t) and the control variable u i (t);

[0353] The ADMM algorithm is used to solve the objective function of the DMPC-based queue longitudinal control model to obtain the vehicle speed v under the driving path of the main queue. i(t); The second model building unit builds an energy control model of a plug-in hybrid electric vehicle based on an improved DDPG, determines a state set S, an action set action, a reward function J and a constraint function based on the adaptive change of the truck driving data based on the improved DDPG plug-in hybrid electric vehicle energy control model;

[0354] S={v i (t),acc i ,SoC i (t),SoH i (t)},i=1,2...I

[0355]

[0356] The model training unit is used to collect standard working conditions and slopes and use them as sample sets. The sample sets are input into the constructed energy control model of plug-in hybrid electric vehicle based on improved DDPG for training, so as to obtain the trained energy control model of plug-in hybrid electric vehicle based on improved DDPG.

[0357] The output unit is used to convert the vehicle speed v obtained in step 5 into i (t) and the slope of the mining platform are input into the trained DDPG plug-in hybrid vehicle energy control model to obtain the engine output power of truck i The truck engine output power The engine speed and torque are calculated by inputting them into the engine's optimal fuel operating curve, and the engine speed and torque are respectively input into the motor transmission path and the generator transmission path to calculate the corresponding motor speed and torque and generator speed and torque.

Claims

1. A method for constructing a plug-in hybrid electric vehicle control model, characterized in that: The specific steps include: Step 1: construct an energy control model of a plug-in hybrid electric vehicle based on an improved DDPG, and determine a state set S, an action set action, a reward function J and a constraint function that are adaptively changed based on truck driving data. S={v i (t),acc i ,SoC i (t),SoH i (t)},i=1,2...I in: v i (t) Vehicle speed along the main queue’s driving path, km / h; acc i is the acceleration of truck i, m / s 2 ; SOC i (t) is the power of truck i at time t; SOH i (t) is the battery health status value of truck i at time t; is the engine output power of truck i, kW; α is the weight of fuel consumption; β is the weight of battery charge maintenance; γ is the weight of battery degradation cost; fuel(t) is the fuel consumption at time t; elec(t) is the energy consumption at time t; SOH ref It is the reference value of battery health status; SoC ref (t) is the spatial domain index SoC when the driving distance is d(t); λ is the SoC decrease rate in the spatial domain; SoC ini For the initial SoC; SoC tgt For the target SoC; L is the expected driving range after the battery is fully charged; eq ini is the initial value of the equivalence factor; eq′ is the value of the equivalent factor when the slope is 0; gr is the slope value; gr min is the minimum slope value; gr max is the maximum slope value; The constraint function based on the improved DDPG plug-in hybrid electric vehicle energy control model is as follows: in: SoC is the battery power; SoC min is the minimum value of SoC; SoC max is the maximum value of SoC; x = eng, MG1, MG2, eng is the engine, MG1 is the motor, MG2 is the generator; T x is the torque of the power unit x, N·m; is the minimum torque of the power unit x, N·m; is the maximum torque of the power unit x, N·m; w x is the speed of the power unit x, rpm; is the minimum speed of the power unit x, rpm; is the maximum speed of the power unit x, rpm; Step 2: Collect standard working conditions and slopes as sample sets, and input the sample sets into the improved DDPG plug-in hybrid electric vehicle energy control model constructed in step 1 for training, so as to obtain a trained improved DDPG plug-in hybrid electric vehicle energy control model.

2. A control method for a plug-in hybrid truck platoon, characterized in that: The method for constructing the plug-in hybrid electric vehicle control model according to claim 1 specifically comprises the following steps: Get the slope of the mining platform and the speed v of truck i i (t), and input it into the improved DDPG plug-in hybrid vehicle energy control model according to claim 1 to obtain the engine output power of truck i The truck engine output power The engine speed and torque are calculated by inputting them into the engine's optimal fuel operating curve, and the engine speed and torque are respectively input into the motor transmission path and the generator transmission path to calculate the corresponding motor speed and torque and generator speed and torque.

3. The control method of the plug-in hybrid electric truck platoon as claimed in claim 1, characterized in that: The slope of the mining platform and the speed v of the truck i are obtained. i (t), specifically comprising the following steps: Step 1, obtaining a survey map of the mining platform where the truck queue performs the task, coordinate processing the survey map of the mining platform, and collecting the slope of the mining platform; Step 2: Establish a mixed integer nonlinear programming model T according to the coordinates of the survey map of the mining platform obtained in step 1 total , solve the mixed integer nonlinear programming model T total Get the driving path of the main queue and the driving path of truck i; The driving path of the main queue is the coordinate l where the truck leaves the main queue i and the coordinates c where the truck merges into the main queue i The driving path of truck i is the coordinate l where truck i leaves the main queue. i , task point coordinates q n The coordinate c where it joins the main queue i The connection; in: T1 is the main queue from the starting point of the mining platform q s The time it takes for the first truck to leave the main queue at coordinate l1, T1 = ||q s -l1|| / v platoon , s; v platoon The maximum speed of the main queue, km / h; T I The coordinate c of the last truck to perform the task and merge into the main queue I To the end point of the mining platform f The time taken, T I =||q f -c I || / v platoon , s; G n The value of indicates whether there are trucks performing tasks at both task point n and task point (n+1). If so, G n Take 1, otherwise G n Take 0; j n represents the sum of the time when all trucks leave the main queue and the time when no trucks leave the main queue when the main queue passes the task point n, s; c i represents the coordinates of the i-th truck merging into the main queue; l i represents the coordinates of the i-th truck leaving the main queue; Step 3, establish the longitudinal dynamics model of the plug-in hybrid truck as shown below; in: is the speed of truck i at time t, km / h; is the acceleration of truck i at time t, m / s 2 ; is the differential of the actual torque of truck i at time t; v i (t) Vehicle speed along the main queue’s driving path, km / h; g is the acceleration due to gravity; A is the front area of ​​truck i, m 2 ; f i is the rolling resistance coefficient of truck i; τ i is the time lag constant of the longitudinal power system of truck i; m i is the mass of truck i, kg; r i is the wheel radius of truck i, m; η i is the transmission system efficiency of truck i; C i_air is the air resistance coefficient of truck i; ρ is air density, kg / m 2 ; u i (t) is the expected torque of truck i at time t, N·m; i is the serial number of the truck, i is 0, 1, 2, ...; Step 4: construct a DMPC-based longitudinal control model for the queue; determine the expected state variables of each truck in the DMPC-based longitudinal control model for the queue according to the driving path of the main queue obtained in step 2. And the objective function The state variable y of the DMPC-based platoon longitudinal control model is determined according to the plug-in hybrid truck longitudinal dynamics model constructed in step 3. i (t) and the control variable u i (t); Step 5: Use the ADMM algorithm to solve the objective function of the DMPC-based queue longitudinal control model constructed in step 4 to obtain the speed v of truck i in the driving path of the main queue. i (t).

4. The control method of the plug-in hybrid electric truck platoon as claimed in claim 1, characterized in that: Step 2 specifically includes the following steps: Step 2.1, construct the mixed integer nonlinear programming model T total : in: T1 is the main queue from the starting point of the mining platform q s The time it takes for the first truck to leave the main queue at coordinate l1, T1 = ||q s -l1|| / v platoon ; v platoon The maximum speed of the main queue; T I The coordinate c of the last truck to perform the task and merge into the main queue I To the end point of the mining platform f The time taken, T I =||q f -c I || / v platoon ; G n The value of indicates whether there are trucks performing tasks at both task point n and task point (n+1). If so, G n Take 1, otherwise G n Take 0; j n It represents the sum of the time when all trucks leave the main queue and the time when no trucks leave the main queue when the main queue passes through task point n; c i represents the coordinates of the i-th truck merging into the main queue; l i represents the coordinates of the i-th truck leaving the main queue; Step 2.3, use the mixed integer nonlinear programming method to solve the optimal T total , get the coordinates l of the truck leaving the main queue i , the coordinate c where the truck meets the main queue i ; Step 2.4: The coordinates l of the truck leaving the main queue obtained in step 2.3 are i and the coordinates c where the truck merges into the main queue i Connect them one by one to get the driving path of the main queue; the coordinate l of the truck leaving the main queue i , task point coordinates q n and the coordinates c where the truck merges into the main queue i Connect them one by one to get I truck driving paths.

5. The control method of the plug-in hybrid electric truck platoon as claimed in claim 1, characterized in that: Step 4 specifically includes the following steps: Step 4.1: Based on the main queue's driving path obtained in step 2, the position s0(t) and speed v0(t) of the main queue are given, and the expected state variables of each truck are obtained according to the following formula: in: is the expected displacement of the i-th truck, m; is the expected speed of the i-th truck, km / h; d des is the distance between two adjacent trucks, m; Step 4.2: Define the state variable of the plug-in hybrid truck longitudinal dynamics model established in step 3 as the current state y of truck i i (t) = [s i (t),v i (t)] T , the control variable is the torque u of truck i i (t); Step 4.3, establish the objective function in: is the predicted state of the i-th truck at time t+k, which includes the position and speed of truck i at time t+k; is the predicted torque of the i-th truck at time t+k; is the hypothetical state of the i-th truck at time t+k, which includes the position and speed of truck i at time t+k; is the assumed state of the j-th truck at time t+k, which includes the position and speed of truck i at time t+k; is the assumed torque of the i-th truck at time t+k; is the expected state of the i-th truck at time t+k, which includes the position and speed of truck i at time t+k; C D is the truck's drag coefficient; ρ is the air density, kg / m 2 ; A is the frontal area of ​​the truck, m 2 ; m i is the mass of the i-th truck, kg; g is the acceleration due to gravity; f is the rolling resistance coefficient of the truck; r i is the wheel radius of truck i, m; η is the transmission efficiency of the truck; is the penalty weight matrix for the deviation between the actual state and the expected state of the i-th truck at time t; is the penalty weight matrix of the error between the actual state and the expected torque of the i-th truck; is the penalty weight matrix of the deviation between the actual state and the assumed state of the i-th truck; is the penalty weight matrix for the deviation between the expected output of the i-th truck and the expected output of its neighboring vehicles. If the vehicle is adjacent to the lead vehicle, then d i,j is the distance between the i-th truck and the j-th truck, m.

6. The control method of the plug-in hybrid electric truck platoon as claimed in claim 5, characterized in that: Step 5 specifically includes the following steps: Step 5.1, for the objective function in step 4.3 Introducing dummy variables and Get the improved objective function L ρ ; in: is the Lagrangian function of the i-th vehicle at time t; is the set of constraints for the i-th truck at time t; and They are the dual variables corresponding to the i-th truck at time t; u i (t) is the torque of the i-th truck at time t; ρ1 and ρ2 are penalty coefficients; Step 5.2: Improve the objective function L obtained in step 5.1 ρ The dual variable in and Update and obtain the improved objective function L after updating the variables ρ ', using ADMM algorithm to update the improved objective function L after the variables are updated ρ 'Solve and get the original residual and the dual residual in: τ is the update step size of the Lagrangian term; is the penalty coefficient; Step 5.3, determine whether the original residual and the dual residual both satisfy the following formula, if so, the control variable is obtained; The control variables include the vehicle speed v under the driving path of the main queue i (t); in: ε pri is the original residual termination condition; ε dual is the dual residual termination condition.

7. A control system for the plug-in hybrid truck platoon, characterized in that: A method for constructing a plug-in hybrid electric vehicle control model based on claim 1 and a method for controlling a plug-in hybrid electric truck platoon according to any one of claims 2 to 6, comprising a map processing unit, a model construction unit, a model training unit and an output unit; The map processing unit is used to obtain the surveying map of the mining platform where the truck queue performs the task, and coordinate the surveying map of the mining platform to collect the slope of the mining platform; The model building unit is used to establish a mixed integer nonlinear programming model T according to the coordinates of the survey map of the mining platform obtained total , solve the mixed integer nonlinear programming model T total Get the driving path of the main queue and the driving path of truck i; Establish a longitudinal dynamics model for a plug-in hybrid truck; Constructing a DMPC-based longitudinal control model for a platoon; determining the expected state variables of each truck in the DMPC-based longitudinal control model for a platoon according to the obtained driving path of the main platoon And the objective function The state variable y of the DMPC-based platoon longitudinal control model is determined according to the constructed plug-in hybrid truck longitudinal dynamics model. i (t) and the control variable u i (t); The ADMM algorithm is used to solve the objective function of the DMPC-based queue longitudinal control model to obtain the vehicle speed v under the driving path of the main queue. i (t); The second model building unit builds an energy control model of a plug-in hybrid electric vehicle based on an improved DDPG, and determines a state set S, an action set action, a reward function J and a constraint function based on the energy control model of the plug-in hybrid electric vehicle based on the improved DDPG; S={v i (t),acc i ,SoC i (t),SoH i (t)},i=1,2...I The model training unit is used to collect standard working conditions and slopes as sample sets, and input the sample sets into the constructed energy control model of the plug-in hybrid electric vehicle based on the improved DDPG for training, so as to obtain the trained energy control model of the plug-in hybrid electric vehicle based on the improved DDPG; The output unit is used to convert the vehicle speed v obtained in step 5 i (t) and the slope of the mining platform are input into the trained DDPG plug-in hybrid vehicle energy control model to obtain the engine output power of truck i The truck engine output power The engine speed and torque are calculated by inputting them into the engine's optimal fuel operating curve, and the engine speed and torque are respectively input into the motor transmission path and the generator transmission path to calculate the corresponding motor speed and torque and generator speed and torque.

Images