Multi-objective optimization hydrogen fuel cell hybrid vehicle energy management method

Abstract

The present application relates to a kind of multi-objective optimization hydrogen fuel cell hybrid vehicle energy management method, belong to hybrid electric vehicle energy management technical field, by driving behavior is integrated into hydrogen fuel cell hybrid vehicle energy management method, online hybrid adaptive anti-noise clustering algorithm and a heuristic self-learning marking algorithm HSL-SVM / NN obtain driving behavior identification model, and with driving behavior identification model as foundation, combined with PMP of pontiac minimum principle, obtain the multi-objective optimization A-ECMS energy management strategy of adaptation driving behavior, simultaneously, combined with electric motor load demand power and its change rate and / or energy storage system weighted SOC, comprehensive consideration is included in hydrogen fuel cell hybrid electric vehicle energy management, obtain the optimal vehicle energy management system corresponding, solve the technical problem that present energy management strategy mostly only considers external driving condition, leading to unreasonable distribution to demand power, cause energy waste and power supply service life short.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of hybrid electric vehicle energy management, and particularly relates to a hydrogen fuel cell hybrid electric vehicle energy management method based on multi-objective optimization. BACKGROUND

[0002] With the increasing demand for urban living environment, hybrid electric vehicles have received extensive attention in the field of automobile research. Since oil-electric hybrid vehicles still cannot get rid of the problem of environmental pollution caused by tail gas emission, and pure electric vehicles have the problem of insufficient endurance, hydrogen fuel cell hybrid electric vehicles stand out. Hydrogen fuel cell hybrid electric vehicles include fuel cells, lithium batteries and super capacitors as three energy sources, and are welcomed by the society as a new energy hybrid electric vehicle with energy saving, no pollution and elimination of range anxiety.

[0003] The energy management research of such hybrid electric vehicles mainly focuses on how to coordinate the output power of each energy source, so as to achieve the best economy and practical performance of the whole vehicle. However, most of the existing energy management strategies only consider external driving conditions, which easily leads to unreasonable allocation of demand power, resulting in energy waste and short service life of power supply. SUMMARY

[0004] In order to solve the technical problem that most of the existing energy management strategies only consider external driving conditions, which leads to unreasonable allocation of demand power, resulting in energy waste and short service life of power supply, the present application provides an A-ECMS energy management strategy which integrates driving behavior into the energy management strategy and forms a multi-objective optimization A-ECMS energy management strategy adapting to driving behavior. At the same time, the size of the motor load demand power and its change rate are collected, or the SOC of the energy storage system is adjusted within a reasonable range, so as to maximize the service life of the energy storage system and the fuel cell system, and to real-time allocate the load demand power between the fuel cell and the energy storage system, so as to obtain the optimal vehicle energy management strategy corresponding to the optimized demand.

[0005] In order to achieve the above object, the specific scheme adopted by the present application is: a multi-objective optimization hydrogen fuel cell hybrid vehicle energy management method, a hydrogen fuel cell hybrid vehicle energy management system model is established, a driving behavior identifier for online identification of driving behavior is obtained; a multi-objective optimization A-ECMS energy management strategy suitable for driving behavior is established; based on the multi-objective optimization A-ECMS energy management strategy suitable for driving behavior, equivalent hydrogen consumption and energy storage system SOC are taken as learning targets to obtain the corresponding optimal vehicle energy management strategy; based on the multi-objective optimization A-ECMS energy management strategy suitable for driving behavior, the optimal power distribution vehicle energy management strategy is obtained according to the change rate of demand power, lithium battery and super capacitor SOC data; the energy of the hydrogen fuel cell hybrid vehicle is managed according to the obtained different vehicle energy management strategies.

[0006] As an optimization scheme of the above-mentioned multi-objective optimization hybrid vehicle energy management method, the multi-objective optimization hydrogen fuel cell hybrid vehicle energy management method comprises the following steps:

[0007] S1, a hydrogen fuel cell hybrid vehicle energy management system model is established;

[0008] S2, pedal parameters of the hydrogen fuel cell hybrid vehicle are collected, data under different driving conditions are obtained, driving behavior is classified by using a hybrid clustering algorithm, and clustering centers and all historical sample data are labeled offline, a real-time self-labeling method HSL-SVM / NN based on a support vector machine is used to obtain a driving behavior identifier for online identification of driving behavior;

[0009] S3, based on the optimal equivalent factor EF table of the Pontryagin minimum principle PMP, driving behavior identification is performed on the simulation data of different driving conditions and the driving behavior identifier, and the optimal equivalent factor under each specific condition is combined for weighted average, so as to obtain the equivalent factor of each driving behavior, a multi-objective optimization A-ECMS strategy suitable for driving behavior is established, and the optimal vehicle energy management strategy under the driving behavior is obtained;

[0010] S4, the data under different driving conditions are processed by using a nearest neighbor algorithm, a demand power transfer probability matrix with universality is obtained based on a Markov decision process, and the energy storage system SOC and the total hydrogen consumption A-ECMS are taken as learning targets to obtain the corresponding optimal vehicle energy management strategy;

[0011] S5, based on the hydrogen fuel cell vehicle energy management system model, the optimal vehicle energy management strategy suitable for power distribution of the hydrogen fuel cell hybrid vehicle is obtained according to the change rate of demand power, lithium battery and super capacitor SOC data by combining steps S3 and S4;

[0012] S6, the optimal vehicle energy management strategy of any hydrogen fuel cell hybrid vehicle obtained from S3 to S5 manages the energy of the hydrogen fuel cell hybrid vehicle.

[0013] As another optimization scheme of the above-mentioned one based on multi-objective optimization, in the step S1, the hydrogen fuel cell hybrid vehicle energy management system model comprises

[0014] S101, a fuel cell voltage model is established:

[0015] V FC = n cell × (E cell -V act.loss -V ohm.loss ) … … … … … (1)

[0016] In the formula, n cell is the number of fuel cell sheets in the fuel cell stack, E cell is the voltage of a single fuel cell, V act.loss is the activation loss voltage of a single fuel cell, V ohm.loss is the internal resistance loss voltage of a single fuel cell;

[0017] S102, a lithium battery voltage model is established:

[0018]

[0019]

[0020] In the formula, SOC BAT.ini is the initial SOC of the lithium battery, i BAT is the lithium battery current, β = ±1 is the lithium battery charging and discharging state selection coefficient, and β is negative when charging and positive when discharging, C nom is the rated capacity of the lithium battery, V(SOC BAT ) BAT.oc and r(SOC BAT ) are the open circuit voltage and internal resistance of the lithium battery when the lithium battery SOC is SOC BAT , P BAT is the electric power of the lithium battery;

[0021] S103, a super capacitor voltage model is established:

[0022] V UC.oc = SOC UC · (V UC.max -V UC.min ) + V UC.min … … … (4)

[0023]

[0024] where V UC.max and V UC.min are the maximum and minimum output voltage of the supercapacitor respectively, R UC is the equivalent internal resistance of the supercapacitor, P UC is the electric power of the supercapacitor.

[0025] S104, establishing a three-energy-source energy management system model:

[0026] P demand = P FC + P BAT + P UC (6)

[0027] where P demand is the load demand power, P FC is the fuel cell power, P BAT is the lithium battery power, P UC is the supercapacitor power, P BAT and P UC are positive when the lithium battery and supercapacitor are discharging, and P BAT and P UC are negative when the lithium battery and supercapacitor are charging.

[0028] As another optimization scheme of the above-mentioned hybrid vehicle energy management method based on multi-objective optimization, in step S2, data under different driving conditions is collected, pedal parameters of the hydrogen fuel cell hybrid vehicle are collected, including brake pedal voltage and accelerator pedal voltage, an offline training database is obtained, and driving behaviors are divided into rapid acceleration, normal acceleration, cruising, normal braking and emergency braking through a hybrid clustering algorithm.

[0029] In step S2, the process of obtaining an online driving behavior recognizer includes a driving behavior recognition clustering process and establishing a recognizer for online recognition, and the driving behavior recognition clustering process is as follows:

[0030] S211, dividing all collected data sample spaces into n*n blocks;

[0031] S212, traversing all blocks to find all non-empty blocks containing the least sample data;

[0032] S213, merging N adjacent blocks to establish a new block and calculating the cost function by formula

[0033]

[0034] S214, repeat step S213 until the cost is less than a given threshold and the number of data points in the current block is not greater than the number of data points in the initial block, otherwise delete the current block and its data points;

[0035] S215, place a sample data in the initial cluster center set C;

[0036] S216, calculate the sample x i with the minimum distance minD(x i ) to the samples in the current cluster center set C and calculate the sum sum(minD(x i ));

[0037] S217, select a random number R from 0 to sum(minD(x i )), calculate R = R-minD(x i );

[0038] S218, repeat steps S216 and S217 until R≤0, place the sample x i in C, repeat steps S215 to S218 until the number of cluster centers is greater than K, and execute step S219;

[0039] S219, find the sample x belonging to the center domain C k by p ;

[0040] S220, repeat step S219 until all data samples have their own clusters;

[0041] An identifier for online recognition is established, and a real-time self-labeling method HSL-SVM / NN based on a support vector machine (SVM) is obtained, which includes the following steps:

[0042] S221, divide the input sample data into three categories: test set, training set, and cross-validation set;

[0043] S222, regard the multi-class SVM model as a series of quadratic programming problems:

[0044]

[0045]

[0046] S223, solve the solution vector under the constraint condition in the calculation formula, obtain a decision function for classification with a radial basis function (RBF) kernel, and obtain the label of each data;

[0047] S224, train a neural network model using an SGD method based on the training set, obtain the label of each data in the test set and the validation set, and repeat until the recognition accuracy meets the requirements.

[0048] S225, obtaining a trained driving behavior recognition model.

[0049] As another optimization scheme of the above-mentioned hybrid vehicle energy management method based on multi-objective optimization, in step S3, the multi-objective optimization A-ECMS strategy adapted to driving behavior includes the following steps:

[0050] S301, collecting a large amount of different working condition data, through offline training and simulation of driving working condition data, using the Pontryagin Minimum Principle PMP to obtain the optimal equivalent factor EF under each specific driving working condition;

[0051] S302, through offline simulation of driving working condition data and driving behavior recognition of the driving behavior recognizer, combining the optimal equivalent factor under each specific working condition to obtain the equivalent factor of each driving behavior through weighted average;

[0052] S303, establishing an equivalent factor database based on driving behavior, and matching the appropriate equivalent factor through real-time recognition of driving behavior;

[0053] S304, combining the above-mentioned driving behavior recognizer with the multi-objective optimization A-ECMS strategy to design an energy management strategy, so as to obtain the optimal power distribution under the driving behavior, and achieve the purpose of minimizing the vehicle energy consumption and prolonging the service life of the energy source.

[0054] As another optimization scheme of the above-mentioned hybrid vehicle energy management method based on multi-objective optimization, in step S4, taking the energy storage system SOC and the total hydrogen consumption A-ECMS as the learning target, the optimal vehicle energy management strategy of the hydrogen fuel cell hybrid vehicle is obtained, including the following steps:

[0055] S401, obtaining a large amount of data of different working conditions to obtain a more universal hydrogen fuel cell hybrid vehicle energy management strategy;

[0056] S402, processing the data obtained in S401 under different driving conditions through the nearest neighbor algorithm, based on Markov decision process, and setting the five driving behaviors obtained in step S2 as one of the application driving condition set, to obtain a demand power transfer probability matrix with universality;

[0057] S403, in order to obtain the optimal energy management strategy of the hydrogen fuel cell hybrid vehicle, taking the energy storage system SOC and the total hydrogen consumption A-ECMS as the learning target;

[0058]

[0059]

[0060] where P BAT·max , P BAT·min are the maximum and minimum power of the lithium battery, respectively, FC·max , P FC·min are the maximum and minimum power of the fuel cell, respectively, is the maximum rate of change of the lithium battery SOC, is the maximum rate of change of the super capacitor SOC, BAT·max , SOC BAT·min are the maximum and minimum SOC of the lithium battery, respectively, UC·max , SOC UC·min are the maximum and minimum SOC of the super capacitor, respectively.

[0061] S404, by writing a corresponding adaptive fast deep learning program with lower computational complexity, the optimal vehicle energy management strategy of the hydrogen fuel cell hybrid vehicle is obtained.

[0062] As another optimization scheme of the above-mentioned multi-objective optimization-based hybrid vehicle energy management method, in step S5, the vehicle energy management of the fuel cell hybrid vehicle is carried out in combination with the rate of change of the demand power, the lithium battery and super capacitor SOC and other factors, including the following steps:

[0063] S501, according to the change amplitude of the vehicle accelerator pedal, the driving motor load demand power P demand is obtained;

[0064] S502, according to the obtained driving motor load demand power P demand , an adaptive low-pass filter is used, the high-frequency power is borne by the super capacitor, and the low-frequency power is borne by the fuel cell and the lithium battery together;

[0065] S503, according to the low-frequency power obtained in the second step, in combination with the multi-objective optimization-based hybrid vehicle A-ECMS strategy and the optimal hybrid vehicle energy management strategy obtained in step S4, the power distribution of the fuel cell power P FC , the lithium battery power P BAT and the super capacitor power P UC is obtained.

[0066] Advantages:

[0067] 1. The multi-objective optimization hydrogen fuel cell hybrid vehicle energy management method of the present application, by integrating driving behavior into the hydrogen fuel cell hybrid vehicle energy management method, obtaining a driving behavior recognition model through an online hybrid adaptive anti-noise clustering algorithm and a heuristic self-learning labeling algorithm HSL-SVM / NN, and obtaining a multi-objective optimization A-ECMS energy management strategy that adapts to driving behavior based on the driving behavior recognition model combined with the Pontryagin Minimum Principle PMP, while considering the load demand power of the electric motor and its rate of change and / or the weighted SOC of the energy storage system in the hydrogen fuel cell hybrid vehicle energy management, obtaining the optimal vehicle energy management system, and improving the operating efficiency and power source life of the hydrogen fuel cell hybrid vehicle.

[0068] 2. The multi-objective optimization hydrogen fuel cell hybrid vehicle energy management method of the present application, based on the online hybrid adaptive anti-noise clustering algorithm and the heuristic self-learning labeling algorithm HSL-SVM / NN, obtains an accurate driving behavior recognition model and can recognize driving behavior in real time online, while obtaining the optimal power distribution of the hybrid vehicle through the multi-objective optimization A-ECMS energy management strategy, and improving energy efficiency.

[0069] 3. The multi-objective optimization hydrogen fuel cell hybrid vehicle energy management method of the present application, in the strategy optimization process, the maximum charging and discharging current is constrained to prevent overcharging and discharging of the lithium battery, so that the SOC of the energy storage system fluctuates smoothly in a reasonable range, prolonging the service life. BRIEF DESCRIPTION OF DRAWINGS

[0070] Figure 1 is the structure diagram of the hybrid power system of the present application;

[0071] Figure 2 is the driving behavior recognition flowchart of the driving behavior recognizer of the present application;

[0072] Figure 3 is the A-ECMS flowchart based on driving behavior of the present application;

[0073] Figure 4 is the structure diagram of the hybrid vehicle energy management system of the present application. DETAILED DESCRIPTION

[0074] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0075] In the description of this invention, the terms A-ECMS refer to Adaptive Equivalent Fuel Consumption Minimization Strategy, and EMS refer to Vehicle Energy Management Strategy.

[0076] like Figure 1 , 2 As shown in Figures 3 and 4, a multi-objective optimized energy management method for hydrogen fuel cell hybrid vehicles includes a hybrid power system structure of a hydrogen fuel cell, a lithium battery, and a supercapacitor. The fuel cell is connected in parallel to the system bus via a unidirectional DC / DC converter, while the lithium battery and supercapacitor are connected via a bidirectional DC / DC converter to provide energy to the drive motor load. During operation, the hydrogen fuel cell system is the main power source, the supercapacitor provides or absorbs the instantaneous peak power that the fuel cell and lithium battery cannot provide or absorb, and the lithium battery provides or absorbs the remaining power.

[0077] A hydrogen fuel cell hybrid vehicle energy management system model was established. Pedal parameters were collected to obtain sample data under different driving conditions. The sample data was clustered and labeled to obtain an online driving behavior recognizer. Based on the optimal equivalent factor (EF) table of the Pontryagin Minimum Principle (PMP), the equivalent factor for each driving behavior was obtained, and a multi-objective optimization A-ECMS energy management strategy adapted to driving behavior was established. Based on the multi-objective optimization A-ECMS energy management strategy adapted to driving behavior, the optimal vehicle energy management strategy was obtained by using equivalent hydrogen consumption and the storage system's State of Charge (SOC) as learning objectives. Based on the multi-objective optimization A-ECMS energy management strategy adapted to driving behavior, the optimal power allocation vehicle energy management strategy was obtained based on the rate of change of demand power, lithium battery, and supercapacitor SOC data. The energy of the hydrogen fuel cell hybrid vehicle was managed according to the obtained different vehicle energy management strategies.

[0078] A multi-objective optimization method for energy management of hydrogen fuel cell hybrid vehicles, the specific implementation steps of which are as follows:

[0079] S1. Establish a model for the energy management system of hydrogen fuel cell hybrid vehicles;

[0080] S101. Establish the fuel cell voltage model:

[0081] V FC =n cell ×(E cell -V act.loss -V ohm.loss )……………………………………(12)

[0082]

[0083]

[0084]

[0085]

[0086] where n cell is the number of fuel cell pieces in the fuel cell stack, E cell is the single piece fuel cell voltage, V act.loss is the single piece fuel cell activation loss voltage, V ohm.loss is the single piece fuel cell internal resistance loss voltage, T is the catalyst layer temperature, T c = 298.15 K is the temperature compensation, R = 8.314 J·(mol·K) -1 is the gas constant, F = 96485 C·mol -1 is the Faraday constant, and are the surface pressures of the fuel cell cathode and anode, respectively, a is the charge transfer coefficient, I FC is the fuel cell current, S cata is the catalyst layer cross-sectional area, I0 is the exchange current density, l is the exchange membrane thickness, S mem is the exchange membrane surface area, Γ(T mem , λ(z)) is the exchange membrane local resistivity, T mem is the exchange membrane temperature, λ(z), z ∈ [0, l] is the water content in the exchange membrane.

[0087] S102, establish a lithium battery voltage model:

[0088]

[0089]

[0090] where SOC BAT.ini is the initial SOC of the lithium battery, i BAT is the lithium battery current, β = ±1 is the lithium battery charging and discharging state selection coefficient, negative for charging and positive for discharging, C nom is the rated capacity of the lithium battery, V(SOC BAT ) BAT.oc and r(SOC BAT ) are the open circuit voltage and internal resistance of the lithium battery when the SOC of the lithium battery is SOC BAT , P BAT is the electric power of the lithium battery.

[0091] S103, establish a supercapacitor voltage model:

[0092] V UC.oc = SOC UC ·(V UC.max -V UC.min )+VUC.min ………………………(19)

[0093]

[0094] where V UC.max and V UC.min are the maximum and minimum output voltage of the supercapacitor, respectively, R UC is the equivalent internal resistance of the supercapacitor, and P UC is the electric power of the supercapacitor.

[0095] S104, a hydrogen fuel cell hybrid vehicle energy management system model of three energy sources is established;

[0096] P demand = P FC + P BAT + P UC …………………………………………(21)

[0097] where P demand is the load demand power, P FC is the fuel cell power, P BAT is the lithium battery power, and P UC is the supercapacitor power. When the lithium battery and the supercapacitor are discharging, P BAT and P UC are positive; when the lithium battery and the supercapacitor are charging, P BAT and P UC are negative.

[0098] S2, pedal parameters of a hydrogen fuel cell hybrid vehicle are collected, data under different driving conditions are obtained, driving behaviors are classified by using a hybrid clustering algorithm, and clustering centers and all historical sample data are labeled offline, a real-time self-labeling method HSL-SVM / NN based on a support vector machine is used to obtain a driving behavior recognizer for online recognition of driving behaviors;

[0099] comprising the following steps:

[0100] S201, offline training data collection, pedal parameters of a fuel cell hybrid vehicle are collected, brake pedal voltage and accelerator pedal voltage are collected; a large amount of data under different conditions is collected to support the acquisition of a more universal hybrid vehicle energy management strategy;

[0101] S202, driving behaviors are divided into rapid acceleration, normal acceleration, cruising, normal braking, and emergency braking by using a hybrid clustering algorithm HADANOC;

[0102] S202, offline labeling is performed on the clustering center and all historical sample data, and the labeled data is well trained and fully utilized, and then a real-time self-labeling method HSL-SVM / NN based on a support vector machine (SVM) is established, and then an online recognition driving behavior recognizer is obtained;

[0103] The process of obtaining the online recognition driving behavior recognizer includes a driving behavior recognition clustering process and an establishment of an recognizer facing online recognition;

[0104] The driving behavior recognition clustering process includes the following steps:

[0105] S211, all data sample spaces collected are divided into n*n blocks;

[0106] S212, all blocks are traversed to find all non-empty blocks containing the least sample data;

[0107] S213, N adjacent blocks are swallowed to establish a new block, and a cost function is calculated by formula

[0108]

[0109] S214, step S213 is repeated until the cost is less than a given threshold and the number of data points in the current block is not greater than that in the initial block, otherwise the current block and its data points are deleted;

[0110] S215, a sample data is placed in an initial clustering center set C;

[0111] S216, a sample x i is calculated to find the minimum distance minD(x i ) with the sample in the current clustering center set C, and a sum sum(minD(x i )) is calculated;

[0112] S217, a random number R is selected from 0 to sum(minD(x i )), and R=R-minD(x i ) is calculated;

[0113] S218, steps S216 and S217 are repeated until R≤0, the sample x i is placed in C, and steps S215 to S218 are repeated until the number of clustering centers is greater than K, and step S219 is executed;

[0114] S219, by a sample x k belonging to the center definition domain C p is found;

[0115] S220. Repeat step S219 until all data samples have their own clusters.

[0116] The process of establishing a recognizer for online recognition and obtaining a real-time self-labeling method HSL-SVM / NN based on Support Vector Machine (SVM) includes the following steps:

[0117] S221. Divide the input sample data into three categories: test set, training set, and cross-validation set;

[0118] S222. Treat the multi-class SVM model as a series of quadratic programming problems:

[0119]

[0120]

[0121] S223. Solve for the solution vector in the calculation formula under the constraints to obtain the decision function with the radial basis function (RBF) kernel for classification, and obtain the label of each data point;

[0122] S224. Train the neural network model using the SGD method based on the training set to obtain the label of each data in the test set and validation set until the recognition accuracy meets the requirements.

[0123] S225. Obtain a well-trained driving behavior recognition model.

[0124] S204. Based on the optimal equivalent factor EF table based on the Pontryagin minimum principle (PMP), in order to extend the service life of the power source of hybrid electric vehicles and improve energy efficiency, an A-ECMS framework strategy based on multi-objective optimization that can adapt to driving behavior is designed to obtain the optimal power allocation of hybrid electric vehicles.

[0125] S3. Based on the Pontryagin Minimum Principle (PMP), the optimal equivalent factor (EF) table is obtained by using simulation data of different driving conditions and driving behavior recognition to identify driving behavior. The optimal equivalent factor under each specific condition is weighted and averaged to obtain the equivalent factor for each driving behavior. A multi-objective optimization A-ECMS strategy adapted to driving behavior is established to obtain the optimal vehicle energy management strategy under the driving behavior.

[0126] The A-ECMS strategy for multi-objective optimization of driving behavior includes the following steps:

[0127] S301. Collect a large amount of data under different working conditions. Through offline training and simulation of driving working condition data, use the Pontryagin Minimum Principle (PMP) to obtain the optimal equivalent factor (EF) for each specific driving working condition.

[0128] S302, obtaining the equivalent factor of each driving behavior by driving behavior recognition through offline simulation of driving condition data and driving behavior recognizer, and combining the optimal equivalent factor under each specific condition for weighted average;

[0129] S303, establishing a database of equivalent factors based on driving behavior, and matching the appropriate equivalent factor by real-time recognition of driving behavior;

[0130] S304, combining the driving behavior recognizer described above with the A-ECMS strategy of multi-objective optimization to design an energy management strategy, so as to obtain the optimal power distribution under the driving behavior, and achieve the purpose of minimizing the energy consumption of the whole vehicle and prolonging the service life of the energy source.

[0131] S4, processing data under different driving conditions by nearest neighbor algorithm, obtaining a demand power transfer probability matrix with universality based on Markov decision process, taking the energy storage system SOC and total hydrogen consumption A-ECMS as learning goals, and obtaining the corresponding optimal vehicle energy management strategy;

[0132] Taking the energy storage system SOC and total hydrogen consumption A-ECMS as learning goals, the optimal vehicle energy management strategy of hydrogen fuel cell hybrid vehicle is obtained, including the following steps:

[0133] S401, obtaining a large amount of data under different conditions to obtain a more universal energy management strategy of hydrogen fuel cell hybrid vehicle;

[0134] S402, processing data obtained in S401 by nearest neighbor algorithm to process data under different driving conditions, and obtaining a demand power transfer probability matrix with universality based on Markov decision process, while setting the five driving behaviors obtained in step S2 as one of the application driving condition sets;

[0135] S403, in order to obtain the optimal energy management strategy of hydrogen fuel cell hybrid vehicle, taking the energy storage system SOC and total hydrogen consumption A-ECMS as learning goals;

[0136]

[0137]

[0138] In the formula, P BAT·max , P BAT·min are the maximum and minimum power of lithium battery, P FC·max , P FC·min are the maximum and minimum power of fuel cell, is the maximum change rate of lithium battery SOC, is the maximum change rate of super capacitor SOC, BAT·max , SOCBAT·min are the maximum and minimum SOC of lithium battery, SOC UC·max , SOC UC·min are the maximum and minimum SOC of super capacitor, respectively.

[0139] S404, by writing a corresponding adaptive fast deep learning program with lower computational complexity, the optimal vehicle energy management strategy of hydrogen fuel cell hybrid vehicle is obtained.

[0140] S5, based on the hydrogen fuel cell vehicle energy management system model, combined with steps S3 and S4, according to the change rate of demand power, lithium battery and super capacitor SOC data, the optimal vehicle energy management strategy suitable for hydrogen fuel cell hybrid vehicle power distribution is obtained;

[0141] Combined with the change rate of demand power, lithium battery and super capacitor SOC and other factors, the energy management of fuel cell hybrid vehicle is carried out, including the following steps:

[0142] S501, according to the change amplitude of vehicle accelerator pedal, the driving motor load demand power P demand is obtained;

[0143] S502, according to the obtained driving motor load demand power P demand , an adaptive low-pass filter is adopted, the high-frequency power is borne by super capacitor, and the low-frequency power is borne by fuel cell and lithium battery together;

[0144] S503, according to the low-frequency power obtained in the second step, combined with the multi-objective optimization of hybrid vehicle A-ECMS strategy and step S4, the optimal hybrid vehicle energy management strategy is obtained, and the power distribution of fuel cell power P FC , lithium battery power P BAT and super capacitor power P UC is obtained.

[0145] S6, any vehicle energy management strategy of hydrogen fuel cell hybrid vehicle obtained from S3 to S5 manages the energy of hydrogen fuel cell hybrid vehicle.

Claims

1. A multi-objective optimized energy management method for hydrogen fuel cell hybrid vehicles, characterized in that: Establish a hydrogen fuel cell hybrid vehicle energy management system model and obtain a driving behavior recognizer for online identification of driving behavior; establish an A-ECMS energy management strategy that adapts to driving behavior and optimizes multiple objectives. The A-ECMS energy management strategy based on multi-objective optimization of adaptive driving behavior uses equivalent hydrogen consumption and energy storage system SOC as learning objectives to obtain the corresponding optimal vehicle energy management strategy. The A-ECMS energy management strategy, which is based on multi-objective optimization to adapt to driving behavior, obtains the corresponding optimal power allocation vehicle energy management strategy according to the rate of change of demand power, lithium battery and supercapacitor SOC data; and manages the energy of hydrogen fuel cell hybrid vehicles according to the different vehicle energy management strategies obtained. Includes the following steps: S1. Establish a model for the energy management system of hydrogen fuel cell hybrid vehicles; S2. Collect pedal parameters of the hydrogen fuel cell hybrid vehicle to obtain data under different driving conditions. Use a hybrid clustering algorithm to classify driving behavior and label the cluster centers and all historical sample data offline. Based on the real-time self-labeling method HSL-SVM / NN of support vector machine, obtain a driving behavior recognizer for online recognition of driving behavior. In step S2, collect data under different driving conditions. By collecting pedal parameters of the hydrogen fuel cell hybrid vehicle, including brake pedal voltage and accelerator pedal voltage, obtain an offline training database. Use a hybrid clustering algorithm to classify driving behavior into rapid acceleration, normal acceleration, cruising, normal braking and emergency braking. S3. Based on the Pontryagin Minimum Principle (PMP), the optimal equivalent factor (EF) table is obtained by using simulation data of different driving conditions and driving behavior recognition to identify driving behavior. The optimal equivalent factor under each specific condition is weighted and averaged to obtain the equivalent factor for each driving behavior. A multi-objective optimization A-ECMS strategy adapted to driving behavior is established to obtain the optimal vehicle energy management strategy under the driving behavior. In step S3, the A-ECMS strategy for multi-objective optimization adapted to driving behavior is established, including the following steps: S301. Collect a large amount of data under different working conditions. Through offline training and simulation of driving working condition data, use the Pontryagin Minimum Principle (PMP) to obtain the optimal equivalent factor (EF) for each specific driving working condition. S302. By offline simulation of driving condition data and driving behavior recognition, driving behavior is identified, and the optimal equivalent factor under each specific condition is weighted and averaged to obtain the equivalent factor for each driving behavior. S303. Establish an equivalent factor database based on driving behavior, and match appropriate equivalent factors by identifying driving behavior in real time; S304. Combine the above-mentioned driving behavior recognizer with the multi-objective optimization A-ECMS strategy to design an energy management strategy, thereby obtaining the optimal power allocation under the driving behavior, so as to minimize the energy consumption of the whole vehicle and extend the life of the energy source. S4. The nearest neighbor algorithm is used to process data under different driving conditions. Based on the Markov decision process, a universal demand power transition probability matrix is ​​obtained. The energy storage system SOC and total hydrogen consumption A-ECMS are used as learning targets to obtain the corresponding optimal vehicle energy management strategy. In step S4, the optimal vehicle energy management strategy for hydrogen fuel cell hybrid vehicles is obtained by using the energy storage system's SOC and total hydrogen consumption A-ECMS as learning targets, including the following steps: S401. Acquire a large amount of data under different operating conditions to obtain a more universal energy management strategy for hydrogen fuel cell hybrid vehicles; S402. The data obtained in S401 is processed by the nearest neighbor algorithm under different driving conditions. Based on the Markov decision process, the five driving behaviors obtained in step S2 are set as one of the application driving condition sets to obtain a universal demand power transition probability matrix. S403. In order to obtain the optimal energy management strategy for hydrogen fuel cell hybrid vehicles, the SOC of the energy storage system and the total hydrogen consumption A-ECMS are used as learning targets. min C total (t)=k FC (η FC )C FC (t)+s(j)·k BAT (SOC BAT )C BST (t) In the formula, P BAT·max P BAT·min These are the maximum and minimum power of the lithium battery, P. FC·max P FC·min These are the maximum and minimum power of the fuel cell, respectively. This is the maximum rate of change of the SOC of a lithium battery. This is the maximum rate of change of the State of Charge (SOC) of a supercapacitor. BAT·max SOC BAT·min It is the maximum and minimum SOC of a lithium battery. UC·max SOC UC·min These are the maximum and minimum SOC of the supercapacitor; S404. By writing a corresponding adaptive fast deep learning program with low computational complexity, the optimal vehicle energy management strategy for hydrogen fuel cell hybrid vehicles is obtained. S5. Based on the hydrogen fuel cell vehicle energy management system model, and in conjunction with steps S3 and S4, obtain the optimal vehicle energy management strategy suitable for power allocation of hydrogen fuel cell hybrid vehicles according to the rate of change of demand power, lithium battery and supercapacitor SOC data. In step S5, the energy management of the fuel cell hybrid vehicle is carried out by combining the rate of change of demand power, the SOC factors of the lithium battery and the supercapacitor, including the following steps: S501. Based on the change amplitude of the vehicle's accelerator pedal, obtain the drive motor load demand power P. demand ; S502, Based on the obtained drive motor load demand power P demand An adaptive low-pass filter is used, with high-frequency power handled by a supercapacitor and low-frequency power handled by a fuel cell and a lithium battery. S503. Based on the low-frequency power obtained in S502, combined with the multi-objective optimization hybrid vehicle A-ECMS strategy and step S4, the optimal hybrid vehicle energy management strategy is obtained, and the fuel cell power P is obtained. FC Lithium battery power P BAT and supercapacitor power P UC Power allocation; S6. The energy management strategy of any hydrogen fuel cell hybrid vehicle obtained from S3 to S5 is used to manage the energy of the hydrogen fuel cell hybrid vehicle.

2. The multi-objective optimized energy management method for hydrogen fuel cell hybrid vehicles according to claim 1, characterized in that: In step S1, the hydrogen fuel cell hybrid vehicle energy management system model includes... S101. Establish the fuel cell voltage model: V FC =n cell ×(E cell -V act.loss -V ohm.loss )............(1) In the formula, n cell E is the number of fuel cell plates in a fuel cell stack. cell It is the voltage of a single fuel cell, V. act.loss It is the activation loss voltage of a single fuel cell, V. ohm.loss It is the internal resistance loss voltage of a single fuel cell; S102. Establish a lithium battery voltage model: In the formula, SOC BAT.ini It is the initial SOC of the lithium battery, i BAT This refers to the lithium battery current, where β = ±1 is the lithium battery charge / discharge state selection coefficient, with β being negative during charging and positive during discharging. C nom This refers to the rated capacity of a lithium battery, V(SOC). BAT ) BAT.oc and r(SOC) BAT These are the values ​​when the SOC of the lithium battery is SOC. BAT The open-circuit voltage and internal resistance of a lithium battery, P BAT It refers to the electrical power of the lithium battery; S103. Establish the supercapacitor voltage model: V UC.oc =SOC UC ·(V UC.max -V UC.min )+V UC.min ………………………………(4) In the formula, V UC.max and V UC.min These are the maximum and minimum output voltages of the supercapacitor, R. UC It is the equivalent internal resistance of the supercapacitor, P UC It is the electrical power of the supercapacitor; S104. Establish a three-energy-source energy management system model: P demand =P FC +P BAT +P UC (6) In the formula, P demand It is the load power requirement, P FC It is the fuel cell power, P BAT It refers to the power of the lithium battery, P. UC It refers to the power of the supercapacitor. When the lithium battery and the supercapacitor discharge, P... BAT and P UC The value is positive when the lithium battery and supercapacitor are charged, P... BAT and P UC It is negative.

3. The multi-objective optimized energy management method for hydrogen fuel cell hybrid vehicles according to claim 1, characterized in that: In step S2, the process of obtaining an online driving behavior recognizer includes a driving behavior recognition and clustering process and the establishment of a recognizer for online recognition. The driving behavior recognition and clustering process steps are as follows: S211. Divide the space of all collected data samples into n×n blocks; S212. Traverse all blocks and find all non-empty blocks that contain the fewest sample data. S213. Merge N adjacent blocks to create a new block and calculate the cost function using the formula. S214. Repeat step S213 until the cost is less than the given threshold and the number of data points in the current block is not greater than the number of data points in the initial block; otherwise, delete the current block and its data points. S215. Place a sample data in the initial cluster center set C; S216, Calculate sample x i The minimum distance between the sample and the current cluster center set C is min D(x). i And calculate the summation sum(minD(x)). i )); S217, From 0 to sum(min D(x) i Choose a random number R from the given information, and calculate R = R - min D(x). i S218. Repeat steps S216 and S217 until R ≤ 0, and then process the sample x. i Place it into C, repeat steps S215 to S218 until the number of cluster centers is greater than K, then execute step S219; S219, Through Find the domain C. k Example x p ; S220. Repeat step S219 until all data samples have their own clusters; To establish a recognizer for online recognition, a real-time self-labeling method based on Support Vector Machine (SVM) HSL-SVM / NN is obtained, which includes the following steps: S221. Divide the input sample data into three categories: test set, training set, and cross-validation set; S222. Treat the multi-class SVM model as a series of quadratic programming problems: S223. Solve for the solution vector in the calculation formula under the constraints to obtain the decision function with the radial basis function (RBF) kernel for classification, and obtain the label of each data point; S224. Train the neural network model using the SGD method based on the training set to obtain the label of each data in the test set and validation set until the recognition accuracy meets the requirements. S225. Obtain a well-trained driving behavior recognition model.

Images