Energy consumption control method and system for extended-range electric loader

Abstract

The present invention discloses an energy consumption control method and system for an extended-range electric loader, the method comprising: obtaining system state parameters of a vehicle; predicting the vehicle speed in a future time window according to the vehicle speed and vehicle acceleration at the current moment by using a speed prediction model based on a recursive neural network, and calculating the required power of a drive motor in the corresponding time window; optimizing and solving a pre-constructed energy consumption optimization objective function by using the rolling optimization idea of ​​model predictive control and a particle swarm algorithm based on the current battery pack charge state, the calculated required power of the drive motor in the time window, and a predetermined range extender power-optimal fuel consumption mapping model, obtaining the optimal equivalent factor of the current prediction window, and calculating the optimal output power combination sequence of the range extender and the battery pack; and controlling the power output of the range extender and the battery pack in the prediction window according to the optimal output power combination sequence of the range extender and the battery pack.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy machinery and relates to an energy consumption control method and system for an extended-range electric loader. Background Art

[0002] Traditional loaders have the disadvantages of high energy consumption and high emissions. At present, although pure electric loaders can achieve zero emissions and efficient energy utilization, the battery packs loaded will reduce the endurance of the vehicle due to sudden changes in ambient temperature and load, and cannot ensure the continuous and efficient operation of the vehicle. For this reason, extended-range electric loaders give full play to the advantages of zero emissions of electric energy and continuous operation of fuel power systems, becoming one of the effective ways to solve this problem. In this process, whether a reasonable energy management strategy can be adapted to the vehicle directly affects the fuel economy of the vehicle.

[0003] At present, the typical energy management strategies proposed for extended-range electric vehicles are mainly divided into two categories: rule-based energy management strategies and optimization-based energy management strategies. Different energy management strategies will bring different energy consumption optimization effects. Among them, the rule-based energy management strategy has simple control logic and good real-time performance, but poor adaptability to working conditions and cannot achieve the optimal control effect. The energy management strategy based on optimization mainly includes offline optimization strategy and real-time optimization strategy. Although the offline optimization strategy can achieve global energy optimization, it has a large amount of calculation and requires accurate knowledge of the vehicle's global operating status information in advance. It is difficult to directly implant it into the controller for engineering application. The most typical real-time optimization strategy is the equivalent fuel consumption minimum strategy (ECMS), but ECMS does not consider the future information of driving conditions during the energy optimization process, resulting in it only being able to achieve instantaneous optimization of energy consumption, and it is difficult to achieve overall optimization of energy consumption.

[0004] Therefore, it is necessary to design an improved ECMS energy management strategy for extended-range electric loaders. Summary of the invention

[0005] Purpose: In order to overcome the deficiencies in the prior art, the present invention provides an energy consumption control method and system for an extended-range electric loader.

[0006] Technical solution: To solve the above technical problems, the technical solution adopted by the present invention is:

[0007] In a first aspect, a method for controlling energy consumption of an extended-range electric loader is provided, comprising:

[0008] Acquire system state parameters of the vehicle, wherein the system state parameters include vehicle speed v, vehicle acceleration a, and battery pack state of charge SOC;

[0009] According to the current vehicle speed v and vehicle acceleration a, the pre-trained speed prediction model based on recurrent neural network RNN ​​is used to predict the vehicle speed in the future prediction window.

[0010] Vehicle speed in the future prediction window based on the prediction Calculate the required power P of the drive motor in the corresponding prediction window d ;

[0011] Based on the current battery pack state of charge SOC and the calculated drive motor demand power P in the prediction window d and the predetermined range extender power-optimal fuel consumption mapping model, and use the rolling optimization idea of ​​model predictive control MPC and particle swarm algorithm to optimize and solve the pre-built energy consumption optimization objective function to obtain the optimal equivalent factor s in the current prediction window. * ;

[0012] Based on the optimal equivalent factor s obtained in the rolling time domain * , using the equivalent energy consumption minimum strategy ECMS algorithm to calculate the optimal output power combination sequence of the range extender and the battery pack;

[0013] The power output of the range extender and the battery pack in the prediction window is controlled according to the optimal output power combination sequence of the range extender and the battery pack.

[0014] In some embodiments, a method for acquiring a speed prediction model based on a recurrent neural network (RNN) includes:

[0015] S21, acquiring historical information of operating condition data, wherein the historical information of operating condition data includes vehicle speed and acceleration;

[0016] S22, segmenting the historical operating condition data according to the preset prediction window τ to form N sets of training sample sets;

[0017] S23, at the current vehicle speed v t , acceleration a t is the input, the speed at the next moment As output, a speed prediction model is constructed using a recurrent neural network (RNN);

[0018]

[0019] Where [v t ,a t ] T and are the input and output of RNN respectively; s t and Respectively represent the input and output of the hidden layer nodes; o trepresents the input of the output layer node; g and σ represent the activation functions of the hidden layer node and the output layer node, respectively, and select tanh and purelin functions; U 52 , W 55 、V 15 They represent the weight matrix from the input layer nodes to the hidden layer nodes, the weight matrix between the hidden layer nodes, and the weight matrix from the hidden layer nodes to the output layer nodes respectively; 0 and t f Respectively represent the starting and ending time of the vehicle;

[0020] S24, optimizing the initial parameters of the speed prediction model based on a hybrid optimization algorithm of genetic algorithm GA / particle swarm algorithm PSOA to obtain the optimal initial parameters, wherein the initial parameters include: 52 , W 55 、V 15 ,

[0021] S25, using the training sample set to train the speed prediction model loaded with the optimal initial parameters, U 52 , W 55 、V 15 The gradient descent method is used for training and updating until the set conditions are met, and a trained speed prediction model based on the recurrent neural network (RNN) is obtained.

[0022] Further, in some embodiments, the S24 includes the following steps:

[0023] S241, the U in the speed prediction model 52 ,W 55 ,V 15 ,as well as The parameters of the kth individual C are encoded to form the initial parameter set to be optimized, and form individuals and populations; k And the population C consisting of K individuals is marked as:

[0024]

[0025] Where M is U, W, V and The total number of parameters;

[0026] S242, calculating the fitness values ​​Fit of 2·K individuals, and selecting K individuals with small fitness values ​​according to the sequence numbers to form an initial parent population; and selecting the individual with the smallest fitness value as the optimal individual of the parent population;

[0027]

[0028] Among them, v i and are the true value and predicted value of the speed of the RNN network, and ε is the set training accuracy condition;

[0029] S243, based on the same parent population, GA and PSOA update the population according to their respective population update mechanisms to form their own offspring populations;

[0030] The two offspring populations S244, GA, and PSOA use the speed prediction model to predict the vehicle speed at the next moment and calculate the fitness value. The fitness values ​​calculated for each individual in the two offspring populations are compared in turn according to the individual sequence number, and the individual with the smallest fitness value in each sequence number is selected to re-form a population consisting of K individuals. The fitness value calculated for each individual in the population is compared with the fitness value calculated for the individual in the parent population in turn, and the individuals with smaller fitness values ​​are selected to form a new population, which is used as the new population for the next iteration of GA and PSOA.

[0031] S245, select the individual with the smallest fitness value in the new population of step S244, compare it with the fitness value of the best individual in step S242, and select the individual with the smallest fitness value as the new parent of the individual for the next iteration of GA and PSOA;

[0032] S246. Determine whether the number of iterations reaches the upper limit or satisfies the fitness value training accuracy condition ε. If so, the new parent individual output in step S245 is output as the optimal individual; otherwise, GA and PSOA use the population obtained in step S244 as the parent population and return to execute step S243.

[0033] In some embodiments, the driving motor requires power P d The calculation methods include:

[0034]

[0035]

[0036] Among them, η t is the overall powertrain efficiency, F t For driving force;

[0037] m is the vehicle mass, g is the acceleration of gravity, f is the rolling resistance coefficient, α is the road slope, ρ is the air density, C d is the drag coefficient, A is the frontal area, and δ is the rotational mass conversion coefficient.

[0038] In some embodiments, the method for determining the range extender power-optimal fuel consumption mapping model includes:

[0039] According to the engine speed n in the range extendere -Power P e - Fuel consumption rate Corresponding map data, determine the range extender power P e -Optimal fuel consumption rate point gather:

[0040]

[0041] The back propagation neural network (BPNN) is used to construct a range extender power-optimal fuel consumption mapping model;

[0042] Using the range extender power P e -Optimal fuel consumption rate point The collection trains and fits the pre-built range extender power-optimal fuel consumption mapping model to determine the range extender power-optimal fuel consumption mapping model.

[0043] Furthermore, in some embodiments, the range extender power-optimal fuel consumption mapping model is:

[0044]

[0045] Among them, x and y represent the network input and output respectively, W 1 , W 2 Respectively represent the weight matrices of each node from the input layer to the hidden layer and from the hidden layer to the output layer, θ 1 ,θ 2 represents the threshold matrix of each node in the hidden layer and the output layer respectively; the hidden layer node activation function f(*) and the output layer node activation function h(*) are selected as tansig and purelin functions respectively; H in , H ou Respectively represent the input and output of the hidden layer;

[0046] The variables are represented as:

[0047]

[0048] In some embodiments, optimizing and solving a pre-built energy consumption optimization objective function includes:

[0049] The energy consumption optimization objective function is expressed as:

[0050]

[0051] in, is the equivalent fuel consumption rate, Q lhv is the fuel calorific value, Δt is the time step, s is the equivalent factor, and p is the penalty factor:

[0052]

[0053] Among them, SOC is the current state of charge of the battery pack. target is the target value of the battery pack state of charge, SOC max and SOC min are the maximum and minimum values ​​of the battery pack’s state of charge, respectively.

[0054] In a second aspect, an energy consumption control device for an extended-range electric loader is provided, comprising a memory and a processor, wherein the memory is used to store instructions, and the instructions are used to control the processor to operate so as to execute the energy consumption control method for the extended-range electric loader described above.

[0055] In a third aspect, an energy consumption control system for an extended-range electric loader is provided, comprising the energy consumption control device for the extended-range electric loader.

[0056] In a fourth aspect, an extended-range electric loader is provided, including the extended-range electric loader energy consumption control system.

[0057] Beneficial effects: The energy consumption control method and system of an extended-range electric loader provided by the present invention have the following advantages:

[0058] (1) RNN is used to construct a speed prediction model, and the initial parameters of the model are optimized based on the GA / PSOA hybrid optimization algorithm, which can achieve online prediction of vehicle speed with higher accuracy.

[0059] (2) The power-optimal fuel consumption mapping model of the range extender fitted by BPNN can reduce the amount of calculation when the strategy searches for the optimal fuel consumption point online, thus improving the real-time performance of the strategy.

[0060] (3) The MPC-ECMS energy consumption management method designed for extended-range electric loaders uses the rolling optimization idea of ​​MPC and the particle swarm algorithm to solve the optimal equivalent factor of the current prediction window. Based on this, the ECMS algorithm is used in the rolling time domain to calculate the optimal output power combination sequence of the range extender and the battery pack, realizing the online rolling optimization of the equivalent factor. The strategy has good real-time performance and strong adaptability to complex random working conditions.

[0061] The energy management method designed in the present invention realizes real-time adjustment and online rolling optimization of the energy consumption of the whole vehicle under complex working conditions, reduces the instantaneous calculation amount of the strategy and the dependence on working condition information, and expands the application scope of the strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a flow chart of an energy consumption control method for an extended-range electric loader according to an embodiment of the present invention;

[0063] Figure 2A graph for generating a speed prediction model training sample set in an embodiment of the present invention;

[0064] Figure 3 Schematic diagram of the RNN vehicle speed prediction network in an embodiment of the present invention;

[0065] Figure 4 This is a schematic diagram of the BPNN structure in an embodiment of the present invention;

[0066] Figure 5 This is a schematic diagram of an energy consumption control system for an extended-range electric loader according to an embodiment of the present invention;

[0067] Figure 6 It is a schematic diagram of the extended-range electric loader mechanism according to an embodiment of the present invention. DETAILED DESCRIPTION

[0068] The present invention will be further described below in conjunction with specific embodiments.

[0069] The existing MPC-ECMS, on the one hand, adopts a vehicle speed prediction method based on data fitting or data-driven when predicting speed. The speed trajectory predicted by this method has low accuracy and is overly dependent on data completeness, resulting in the calculated power or torque distribution strategy being unable to meet the actual power requirements of the vehicle; on the other hand, when determining the equivalent factor, it mainly relies on the operating condition information to determine the trajectory change curve of the SOC in advance, and on this basis, the equivalent factor is determined with the goal of optimizing energy consumption. In this process, the equivalent factor is highly dependent on the accuracy and integrity of the operating condition information, and it is difficult to meet the requirements of loaders with random operating conditions and unknown future state information. Therefore, it is very necessary to design an MPC-ECMS energy consumption management method that does not rely on the integrity of the operating condition information for extended-range electric loaders with complex and changeable operating conditions.

[0070] In order to overcome the application limitations of the existing ECMS on extended-range electric loaders, the present invention aims at optimizing the energy consumption of the whole vehicle and proposes an MPC-ECMS energy consumption management method and system. First, an RNN speed prediction model is established, and the initial parameters of the model are optimized based on the hybrid optimization algorithm of GA / PSOA to improve the prediction accuracy of the model; then, the power-optimal fuel consumption mapping model of the range extender is fitted by BPNN; on this basis, the rolling optimization idea of ​​MPC and the particle swarm algorithm are used to optimize and solve the equivalent factor of the current prediction window. Based on this, the ECMS algorithm is used in the rolling time domain to calculate the optimal output power combination sequence of the range extender and the battery pack. Under the premise of meeting the power of the whole vehicle, this method adjusts the power output of the range extender and the battery pack in real time through the designed MPC-ECMS, so as to realize the real-time adjustment and online rolling optimization of the energy consumption of the whole vehicle under complex working conditions. The energy management method designed by the present invention reduces the instantaneous calculation amount of the strategy and its dependence on working condition information, and the strategy has strong real-time and working condition adaptability, which is convenient for engineering application.

[0071] Some of the professional terms involved in this application are explained as follows:

[0072] SOC---State of Charge;

[0073] ECMS---equivalent consumption minimum strategy;

[0074] BPNN——Back Propagation Neural Network;

[0075] RNN——Recurrent neural network;

[0076] MPC——Model predictive control.

[0077] Example 1

[0078] A method for controlling energy consumption of an extended-range electric loader, comprising:

[0079] Acquire system state parameters of the vehicle, wherein the system state parameters include vehicle speed v, vehicle acceleration a, and battery pack state of charge SOC;

[0080] According to the current vehicle speed v and vehicle acceleration a, the pre-trained speed prediction model based on recurrent neural network RNN ​​is used to predict the vehicle speed in the future prediction window.

[0081] Vehicle speed in the future prediction window based on the prediction Calculate the required power P of the drive motor in the corresponding prediction window d ;

[0082] Based on the current battery pack state of charge SOC and the calculated drive motor demand power P in the prediction window d and the predetermined range extender power-optimal fuel consumption mapping model, and use the rolling optimization idea of ​​model predictive control MPC and particle swarm algorithm to optimize and solve the pre-built energy consumption optimization objective function to obtain the optimal equivalent factor s in the current prediction window. * ;

[0083] Based on the optimal equivalent factor s obtained in the rolling time domain * , using the equivalent energy consumption minimum strategy ECMS algorithm to calculate the optimal output power combination sequence of the range extender and the battery pack;

[0084] The power output of the range extender and the battery pack in the prediction window is controlled according to the optimal output power combination sequence of the range extender and the battery pack.

[0085] In some specific embodiments, Figure 1 As shown, a method for controlling energy consumption of an extended-range electric loader comprises:

[0086] Step S1: Build a speed prediction model through offline training; specifically including:

[0087] ① Obtain historical information of working condition data offline, including speed and acceleration;

[0088] ② The historical operating condition data is segmented according to the prediction window size τ to form N sets of training samples, such as Figure 2 As shown;

[0089] ③ With the current vehicle speed v t , acceleration a t As input, the speed at the next moment is output, and the speed prediction model is constructed using RNN. Here, the three-layer RNN structure of "1-5-1" is taken as an example. Figure 3 As shown, the RNN vehicle speed prediction network model is constructed:

[0090]

[0091] Where [v t ,a t ] T and Respectively represent the input and output of RNN; s t and Respectively represent the input and output of the hidden layer nodes; ot represents the input of the output layer node; g and σ represent the activation functions of the hidden layer node and the output layer node, respectively, where tanh and purelin functions are selected respectively; U 52 , W 55 、V 15 They represent the weight matrix from the input layer nodes to the hidden layer nodes, the weight matrix between the hidden layer nodes, and the weight matrix from the hidden layer nodes to the output layer nodes respectively; 0 and t f Respectively represent the starting and ending time of the vehicle.

[0092] ④ Replace U in the prediction model formula (1) 52 ,W 55 ,V 15 ,as well as The parameters of the kth individual C are encoded to form the initial parameter set to be optimized, and form individuals and populations. k And the population C consisting of K individuals is marked as:

[0093]

[0094] Where M is U, W, V and The total number of parameters.

[0095] In order to improve the speed prediction accuracy, a hybrid optimization algorithm based on GA / PSOA is designed to optimize the initial parameters of the RNN network, including the following steps:

[0096] 1) Initialize GA and PSOA algorithm parameters, population C and individual C respectively k And determine the fitness function:

[0097]

[0098] Among them, v i and are the true value and predicted value of the speed of the RNN network respectively, and ε is the set training accuracy condition.

[0099] 2) Use formulas (1) and (3) to calculate the fitness values ​​of 2·K individuals in sequence, and select K individuals with small fitness values ​​according to the sequence number to form the initial parent population; and select the individual with the smallest fitness value as the optimal individual of the parent population;

[0100] 3) Based on the same parent population, GA and PSOA update the population according to their respective population update mechanisms to form their own offspring populations;

[0101] 4) The two offspring populations respectively execute equations (1) and (3), compare the fitness values ​​calculated for each individual in the two offspring populations in turn according to the individual sequence number, select the individual with the smallest fitness value in each sequence number, re-form a population consisting of K individuals, and compare the fitness values ​​calculated for each individual with the fitness values ​​calculated for the individuals in the parent population in turn, select the individuals with smaller fitness values ​​to form a new population, and this population is used as the new population for the next iteration of GA and PSOA;

[0102] 5) Select the individual with the smallest fitness value in the new population in step 4, compare it with the fitness value of the best individual in step 2, and select the individual with the smallest fitness value as the new parent of the individual for the next iteration of GA and PSOA;

[0103] 6) Determine whether the number of iterations reaches the upper limit or satisfies the fitness value training accuracy condition ε. If so, the new parent individual output in step 5 is output as the optimal individual; otherwise, GA and PSOA use the population obtained in step 4 as the parent population and return to step 3.

[0104] ⑤ Output the optimized parameter set and assign it to the RNN network as the optimal initial parameters of the RNN network;

[0105] ⑥ Continue to use N groups of samples to train the RNN network loaded with the optimal initial parameters, and U 52 ,W 55 ,V 15 The training is updated according to the gradient descent method with momentum. If the output error Fit of the network satisfies the set value ε or reaches the number of training times, the training is stopped and the RNN speed prediction model is obtained. Otherwise, repeat step ⑥.

[0106] Step S2: According to the working characteristics of the range extender, an off-line determination of a range extender power-optimal fuel consumption mapping model is performed; specifically, the steps include:

[0107] ①According to the engine speed n in the range extender e -Power P e - Fuel consumption rate Corresponding map data, determine the engine power P according to the following expression e -Optimal fuel consumption rate point gather:

[0108]

[0109] ② Use BPNN to build a range extender power-optimal fuel consumption mapping model. Here, take the three-layer BPNN structure of "1-5-1" as an example. Figure 4 As shown, the BPNN network model is constructed:

[0110]

[0111] Among them, H in , H ou Respectively represent the hidden layer input and output; x, y represent the network input and output, respectively, W 1 , W 2 Respectively represent the weight matrices of each node from the input layer to the hidden layer and from the hidden layer to the output layer, θ 1 ,θ 2 Represent the threshold matrices of each node in the hidden layer and the output layer node respectively. The hidden layer node excitation function f(*) and the output layer node excitation function h(*) are selected as tansig and purelin functions respectively. The above variables can be further expressed as:

[0112]

[0113] Using the acquired power P e -Optimal fuel consumption rate point The weights and thresholds in the training formula (6) are combined to fit the range extender power-optimal fuel consumption mapping model.

[0114] Step S3: using the vehicle-mounted data collector to obtain the vehicle's system status parameters in real time, including vehicle speed v, vehicle acceleration a, and battery pack SOC;

[0115] Step S4: Predict the future vehicle speed based on the acceleration and speed prediction model obtained by the vehicle data collector Calculate the required power P of the drive motor d , and set the range extender output power to P e , the battery pack output power is P b The specific required power is calculated as follows:

[0116]

[0117] Where η t is the efficiency of the entire powertrain, the driving force F t It can be expressed as:

[0118]

[0119] Among them, m is the vehicle mass, g is the acceleration of gravity, f is the rolling resistance coefficient, α is the road slope, ρ is the air density, C d is the drag coefficient, A is the frontal area, and δ is the rotation mass conversion factor;

[0120] Step S5: Using the rolling optimization concept of MPC and the particle swarm algorithm, the equivalent factor of the current prediction window is optimized and solved. Based on this, the ECMS algorithm is used in the rolling time domain to calculate the optimal output power combination sequence of the range extender and the battery pack; specifically, it includes:

[0121] ① Determination of energy consumption optimization objective function: The MPC energy consumption optimization problem constructed in the discrete time domain can be expressed as:

[0122]

[0123] The goal of this problem is to find the optimal output power combination sequence u * (t):

[0124]

[0125] in, is the equivalent fuel consumption rate, Q lhv is the fuel calorific value; Δt is the time step, which is taken as 1 here; s is the equivalent factor, and p is the penalty factor. The penalty factor is related to the battery pack SOC. Specifically,

[0126]

[0127] Among them, SOC target is the battery pack SOC target value, SOC max and SOC min They are the maximum and minimum values ​​of the battery pack SOC respectively.

[0128] ②RNN speed prediction model predicts the speed within the future time window τ based on the current speed v(t) and acceleration a(t) collected by the on-board data collector At the same time, the required power sequence P of the drive motor within the time window τ is calculated using step 4. d = {P d (t+1),...,P d (t+τ)};

[0129] ③ Based on the calculated drive motor demand power sequence P of the current prediction window d The range extender power-optimal fuel consumption mapping model established in step 2 is used to solve the optimal equivalent factor s of the current prediction window based on formula (9) by using the rolling optimization idea of ​​MPC and particle swarm algorithm. * ;

[0130] ④ In the rolling time domain, based on the optimal equivalent factor s * , use the ECMS algorithm to calculate the optimal output power combination sequence of the range extender and battery pack

[0131] Step S6: Set u * (t) The actual vehicle is sent to the vehicle, and the vehicle's range extender controller and BMS controller control the power output of the range extender and battery pack respectively. * After (t), time t and vehicle motion state will be updated immediately and enter the next round of MPC adjustment process;

[0132] In particular, during this process, the vehicle will be online to collect operating information. When , the speed prediction model will be updated through online training, where is the training window size, and j is the number of online training times. As the vehicle continues to move, the above control process will continue rolling until the vehicle stops.

[0133] Example 2

[0134] An energy consumption control device for an extended-range electric loader is characterized in that it includes a memory and a processor, the memory is used to store instructions, and the instructions are used to control the processor to operate so as to execute the energy consumption control method for the extended-range electric loader described in Example 1.

[0135] Example 3

[0136] An energy consumption control system for an extended-range electric loader includes the energy consumption control device for an extended-range electric loader described in Example 2.

[0137] In some embodiments, an energy consumption control system for an extended-range electric loader, such as Figure 5As shown, it includes a control layer and an execution layer. The control layer mainly includes an on-board data collector, a speed prediction model, a demand power calculation module, a range extender power-optimal fuel consumption mapping module, an MPC-ECMS energy management controller, a range extender controller, and a BMS controller. Among them, the on-board data collector obtains the system status parameters of the vehicle in real time, including vehicle speed, vehicle acceleration, and battery pack SOC; the speed prediction model predicts the speed of the vehicle at a future time based on the current vehicle speed and acceleration obtained by the on-board data collector; the demand power calculation module can calculate the demand power of the drive motor based on the vehicle acceleration obtained by the on-board data collector and the vehicle speed predicted by the speed prediction model; the range extender power-optimal fuel consumption mapping module can determine the optimal fuel consumption of the range extender at each power point based on the demand power calculated by the demand power calculation module; the MPC-ECMS energy management controller optimizes and solves the equivalent factor of the current prediction window based on the demand power calculated by the demand power calculation module and the battery pack SOC obtained by the on-board data collector in real time, using the rolling optimization idea of ​​MPC and the particle swarm algorithm, and uses the ECMS algorithm in the rolling time domain to calculate the optimal output power combination sequence of the range extender and the battery pack, thereby controlling the output power distribution of the range extender and the battery pack; the range extender controller is used to control the power output of the range extender, and the BMS controller is used to control the power output of the battery pack.

[0138] The execution layer mainly includes the range extender, battery pack, rectifier, inverter, drive assembly, and gearbox. The corresponding loader system configuration is as follows: Figure 6 As shown, the range extender includes an engine and a generator, which are not directly connected to the drive assembly, which greatly reduces the impact and vibration of the range extender caused by the sudden change of load. The range extender drives the assembly after rectification by the rectifier and inversion by the inverter, and can charge the battery pack at the same time; the battery pack plays the role of peak shaving and valley filling. When high power is required, it drives the assembly together with the range extender and stores the recovered energy during braking; the drive assembly includes a motor controller and a drive motor, and the drive motor accurately outputs the drive torque according to the command signal of the motor controller to control the vehicle driving.

[0139] Example 4

[0140] An extended-range electric loader comprises the above-mentioned extended-range electric loader energy consumption control system.

[0141] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0142] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0143] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0144] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for controlling energy consumption of an extended-range electric loader, characterized in that: include: Acquire system state parameters of the vehicle, wherein the system state parameters include vehicle speed v, vehicle acceleration a, and battery pack state of charge SOC; According to the current vehicle speed v and vehicle acceleration a, the pre-trained speed prediction model based on recurrent neural network RNN ​​is used to predict the vehicle speed in the future prediction window. Vehicle speed in the future prediction window based on the prediction Calculate the required power P of the drive motor in the corresponding prediction window d ; Based on the current battery pack state of charge SOC and the calculated drive motor demand power P in the prediction window d and the predetermined range extender power-optimal fuel consumption mapping model, and use the rolling optimization idea of ​​model predictive control MPC and particle swarm algorithm to optimize and solve the pre-built energy consumption optimization objective function to obtain the optimal equivalent factor s in the current prediction window. * ; Based on the optimal equivalent factor s obtained in the rolling time domain * , using the equivalent energy consumption minimum strategy ECMS algorithm to calculate the optimal output power combination sequence of the range extender and the battery pack; Controlling the power output of the range extender and the battery pack in the prediction window according to the optimal output power combination sequence of the range extender and the battery pack; Among them, the method for obtaining the speed prediction model based on the recurrent neural network RNN ​​includes: S21, acquiring historical information of operating condition data, wherein the historical information of operating condition data includes vehicle speed and acceleration; S22, segmenting the historical operating condition data according to the preset prediction window τ to form N sets of training sample sets; S23, at the current vehicle speed v t , acceleration a t is the input, the speed at the next moment As output, a speed prediction model is constructed using a recurrent neural network (RNN); Where [v t ,a t ] T and are the input and output of RNN respectively; s t and Respectively represent the input and output of the hidden layer nodes; o t represents the input of the output layer node; g and σ represent the activation functions of the hidden layer node and the output layer node, respectively, and select tanh and purelin functions; U 52 , W 55 、V 15 They represent the weight matrix from the input layer nodes to the hidden layer nodes, the weight matrix between the hidden layer nodes, and the weight matrix from the hidden layer nodes to the output layer nodes respectively; t0 and t f They represent the starting and ending time of the vehicle respectively, and t is the time; S24, optimizing the initial parameters of the speed prediction model based on a hybrid optimization algorithm of genetic algorithm GA and particle swarm algorithm PSOA to obtain the optimal initial parameters, wherein the initial parameters include: 52 , W 55 、V 15 , S25, using the training sample set to train the speed prediction model loaded with the optimal initial parameters, U 52 , W 55 、V 15 The gradient descent method is used for training and updating until the set conditions are met, and a trained speed prediction model based on the recurrent neural network (RNN) is obtained.

2. The method for controlling energy consumption of an extended-range electric loader according to claim 1, characterized in that: The S24 comprises the following steps: S241, the U in the speed prediction model 52 ,W 55 ,V 15 ,as well as The parameters of the kth individual C are encoded to form the initial parameter set to be optimized, and form individuals and populations; k And the population C consisting of K individuals is marked as: Where M is U, W, V and The total number of parameters; S242, calculating the fitness values ​​Fit of 2·K individuals, and selecting K individuals with small fitness values ​​according to the sequence numbers to form an initial parent population; and selecting the individual with the smallest fitness value as the optimal individual of the parent population; Among them, v i and are the true value and predicted value of the speed of the RNN network, and ε is the set training accuracy condition; S243, based on the same parent population, respectively updating the population according to their respective population update mechanisms to form their respective offspring populations; The two offspring populations S244, GA, and PSOA use the speed prediction model to predict the vehicle speed at the next moment and calculate the fitness value. The fitness values ​​calculated for each individual in the two offspring populations are compared in turn according to the individual sequence number, and the individual with the smallest fitness value in each sequence number is selected to re-form a population consisting of K individuals. The fitness value calculated for each individual in the population is compared with the fitness value calculated for the individual in the parent population in turn, and the individuals with smaller fitness values ​​are selected to form a new population, which is used as the new population for the next iteration of GA and PSOA. S245, select the individual with the smallest fitness value in the new population of step S244, compare it with the fitness value of the best individual in step S242, and select the individual with the smallest fitness value as the new parent of the individual for the next iteration of GA and PSOA; S246. Determine whether the number of iterations reaches the upper limit or satisfies the fitness value training accuracy condition ε. If so, the new parent individual output in step S245 is output as the optimal individual; otherwise, GA and PSOA use the population obtained in step S244 as the parent population and return to execute step S243.

3. The method for controlling energy consumption of an extended-range electric loader according to claim 1, characterized in that: Drive motor power requirement P d The calculation methods include: Among them, η t is the overall powertrain efficiency, F t For driving force; m is the vehicle mass, g is the acceleration of gravity, f is the rolling resistance coefficient, α is the road slope, ρ is the air density, C d is the drag coefficient, A is the frontal area, and δ is the rotational mass conversion coefficient.

4. The method for controlling energy consumption of an extended-range electric loader according to claim 1, characterized in that: The method for determining the range extender power-optimal fuel consumption mapping model includes: According to the engine speed n in the range extender e -Power P e - Fuel consumption rate Corresponding map data, determine the range extender power P e -Optimal fuel consumption rate point gather: The back propagation neural network (BPNN) is used to construct a range extender power-optimal fuel consumption mapping model; Using the range extender power P e -Optimal fuel consumption rate point The collection trains and fits the pre-built range extender power-optimal fuel consumption mapping model to determine the range extender power-optimal fuel consumption mapping model.

5. The energy consumption control method for an extended-range electric loader according to claim 4, characterized in that: The range extender power-optimal fuel consumption mapping model is: Where x and y represent the network input and output, respectively; W1 and W2 represent the weight matrices of the nodes from the input layer to the hidden layer and from the hidden layer to the output layer, respectively; θ1 and θ2 represent the threshold matrices of the nodes in the hidden layer and the output layer, respectively; the hidden layer node activation function f(*) and the output layer node activation function h(*) are selected from tansig and purelin functions, respectively; H in , H ou Respectively represent the input and output of the hidden layer; The variables are represented as:

6. An energy consumption control device for an extended-range electric loader, characterized in that: It comprises a memory and a processor, the memory is used to store instructions, and the instructions are used to control the processor to operate so as to execute the energy consumption control method of the extended-range electric loader according to any one of claims 1-5.

7. An energy consumption control system for an extended-range electric loader, characterized in that: It includes the energy consumption control device for the extended-range electric loader as described in claim 6.

8. An extended-range electric loader, characterized in that: Including the energy consumption control system of the extended-range electric loader as described in claim 7.

Images