Rail vehicle operation processing method and device and rail vehicle

Abstract

The application provides a rail vehicle operation processing method and device and a rail vehicle. The method is applied to a rail vehicle, the rail vehicle uses a hydrogen fuel cell and a lithium battery as a power source, and the method comprises the following steps: calling a pre-trained operation processing network model, wherein the operation processing network model is used to determine driving acceleration of the rail vehicle at a next time and output power of the hydrogen fuel cell according to operation parameters of the rail vehicle at a current time; obtaining the operation parameters of the rail vehicle at the current time, inputting the operation parameters into the operation processing network model, and obtaining the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell output by the operation processing network model, so that, in the case that the rail vehicle is operated according to the driving acceleration and the hydrogen fuel cell outputs power according to the output power, the operation cost of the rail vehicle is minimized. The energy consumption of the hydrogen energy train is effectively reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of rail vehicles, in particular to a rail vehicle operation processing method and device and a rail vehicle. BACKGROUND

[0002] With the enhancement of global environmental protection awareness and the deepening of the concept of sustainable development, reducing the energy consumption and pollution of the railway industry and realizing a green and low-carbon transportation mode have become an important direction for the development of urban transportation. Among many new energy sources, hydrogen is considered an ideal choice to meet the high power demand of the railway industry due to its high energy efficiency, zero carbon emissions, and wide sources. Hydrogen energy trains powered by hydrogen fuel cells have emerged as the times require.

[0003] Therefore, finding a rail vehicle operation processing method that can effectively reduce the energy consumption of hydrogen energy trains has become a current research hotspot. SUMMARY

[0004] The present application provides a rail vehicle operation processing method and device and a rail vehicle, which can effectively reduce the energy consumption of hydrogen energy trains and reduce the operating cost of the rail vehicle.

[0005] The present application provides a rail vehicle operation processing method, which is applied to a rail vehicle powered by a hydrogen fuel cell and a lithium battery. The method comprises the following steps: calling a pre-trained operation processing network model, wherein the operation processing network model is used to determine the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell according to the operation parameters of the rail vehicle at the current time; obtaining the operation parameters of the rail vehicle at the current time and inputting the operation parameters into the operation processing network model to obtain the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell output by the operation processing network model, so that the operating cost of the rail vehicle is minimized when the rail vehicle operates at the driving acceleration and the hydrogen fuel cell outputs power at the output power.

[0006] According to the operation processing method of the rail vehicle provided by the application, the operation processing network model is obtained by the following method: constructing a training data set, wherein the training data set is determined based on an optimal data set, the optimal data set is determined based on an operation curve and a power source curve of a historical operation parameter of the rail vehicle, the operation curve is used to represent a corresponding relationship between a driving acceleration and a time of the rail vehicle under a minimum operation cost of the rail vehicle, and the power source curve is used to represent a corresponding relationship between an output power and a time of the hydrogen fuel cell under the minimum operation cost of the rail vehicle; and training the operation processing network model based on the training data set and according to a proximal policy optimization algorithm to obtain a trained operation processing network model.

[0007] According to the operation processing method of the rail vehicle provided by the application, the operation processing network model includes a critic network model; the operation processing network model is trained based on the training data set and according to the proximal policy optimization algorithm to obtain a trained operation processing network model, and the operation processing network model specifically includes the following steps: determining an operation state of the rail vehicle based on the operation curve and the power source curve in the training data set, wherein the operation state includes an operation speed, a state of charge of the hydrogen fuel cell and an operation distance; training an actor network model corresponding to the critic network model based on the training data set and according to the proximal policy optimization algorithm; inputting the operation state into the actor network model to obtain an action to be performed at a next time point output by the actor network model and corresponding to the operation state, wherein the action to be performed at the next time point includes a predicted driving acceleration of the rail vehicle at the next time point and a predicted output power of the hydrogen fuel cell under the operation state; determining a network loss function based on the action to be performed at the next time point and an operation state at the next time point corresponding to the action to be performed at the next time point, and iteratively training the operation processing network model based on the network loss function until the operation processing network model converges to obtain the trained operation processing network model.

[0008] According to the operation processing method of the rail vehicle provided by the application, the optimal data set is determined by the following method: constructing a target function of the rail vehicle in an operation process, wherein the target function is a function representing an operation cost of the rail vehicle; obtaining historical operation line information of the rail vehicle, wherein the historical operation line information includes speed limit information and station stay duration information of a historical operation line of the rail vehicle; processing the historical operation line information and the target function based on Hermite-Simpson collocation technology to obtain the operation curve and the power source curve; and determining the optimal data set based on the operation curve and the power source curve.

[0009] According to the application, a rail vehicle operation processing method is provided, which processes the historical operation line information and the target function based on the Hermite-Simpson collocation technique to obtain the operation curve and the power source curve, and specifically includes: performing dimension reduction processing on the target function based on the Hermite-Simpson collocation technique to obtain a nonlinear programming corresponding to the target function; and obtaining the operation curve and the power source curve based on the historical operation line information and the nonlinear programming.

[0010] According to the application, a rail vehicle operation processing method is provided, which processes the historical operation line information and the target function based on the Hermite-Simpson collocation technique to obtain the operation curve and the power source curve, and specifically includes: performing dimension reduction processing on the target function based on the Hermite-Simpson collocation technique to obtain a nonlinear programming corresponding to the target function; and obtaining the operation curve and the power source curve based on the historical operation line information and the nonlinear programming.

[0011] According to the application, a rail vehicle operation processing method is provided, which processes the historical operation line information and the target function based on the Hermite-Simpson collocation technique to obtain the operation curve and the power source curve, and specifically includes: performing dimension reduction processing on the target function based on the Hermite-Simpson collocation technique to obtain a nonlinear programming corresponding to the target function; and obtaining the operation curve and the power source curve based on the historical operation line information and the nonlinear programming.

[0012] According to the method, the target function is determined in the following manner: determining a first equivalent hydrogen consumption of the hydrogen fuel cell, a second equivalent hydrogen consumption of the lithium battery, and a first value coefficient of hydrogen; determining a first loss of the hydrogen fuel cell, a second loss of the lithium battery, a second value coefficient of the hydrogen fuel cell, and a third value coefficient of the lithium battery; determining a hydrogen consumption cost of the rail vehicle during operation based on the first equivalent hydrogen consumption, the second equivalent hydrogen consumption, and the first value coefficient; determining a power source life cost of the rail vehicle during operation based on the first loss, the second loss, the second value coefficient, and the third value coefficient; and determining the target function according to the hydrogen consumption cost and the power source life cost.

[0013] The application further provides a rail vehicle operation processing device, which is applied to a rail vehicle, and the rail vehicle uses a hydrogen fuel cell and a lithium battery as power sources. The device comprises a calling module configured to call a pre-trained operation processing network model, wherein the operation processing network model is configured to determine driving acceleration of the rail vehicle at a next time and output power of the hydrogen fuel cell according to operation parameters of the rail vehicle at a current time; and a processing module configured to obtain the operation parameters of the rail vehicle at the current time, input the operation parameters into the operation processing network model, and obtain the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell output by the operation processing network model, so that, when the rail vehicle is operated at the driving acceleration and the hydrogen fuel cell outputs power at the output power, operation cost of the rail vehicle is minimized.

[0014] The application further provides a rail vehicle, which comprises a rail vehicle body and a processor, wherein the processor is configured to execute the rail vehicle operation processing method.

[0015] The application further provides an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the rail vehicle operation processing method according to any one of the above-mentioned rail vehicle operation processing methods when executing the computer program.

[0016] The application further provides a non-transitory computer readable storage medium, which stores a computer program, and the computer program implements the rail vehicle operation processing method according to any one of the above-mentioned rail vehicle operation processing methods when executed by a processor.

[0017] The application further provides a computer program product, which comprises a computer program, and the computer program implements the rail vehicle operation processing method according to any one of the above-mentioned rail vehicle operation processing methods when executed by a processor.

[0018] The application provides a rail vehicle operation processing method and device and a rail vehicle. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0020] Figure 1 Fig. 1 is one of the flowcharts of the rail vehicle operation processing method provided by the application.

[0021] Figure 2 Fig. 3 is the flowchart of training the operation processing network model according to the proximal policy optimization algorithm based on the training data set.

[0022] Figure 3 Fig. 5 is the flowchart of determining the optimal data set.

[0023] Figure 4 Fig. 7 is the flowchart of performing dimensionality reduction processing on the target function based on the Hermite-Simpson collocation technique.

[0024] Figure 5 Fig. 9 is the flowchart of determining the target function.

[0025] Figure 6 Fig. 11 is the structural diagram of the rail vehicle operation processing device provided by the application.

[0026] Figure 7 Fig. 13 is the structural diagram of the electronic device provided by the application. DETAILED DESCRIPTION

[0027] In order to make the objects, technical solutions and advantages of the present application clearer, the following will be combined with the accompanying drawings for the purpose of making the technical solutions in the present application clearer, complete and more comprehensible. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present application.

[0028] The running processing method of the rail vehicle provided by the present application deeply explores the "source-load" coupling characteristics of hydrogen energy driven high-speed trains, further reduces the fuel cell train operation consumption, reasonably allocates the output among the systems of the train, and significantly improves the operation efficiency and durability of the fuel cell power source system, effectively reduces the cost of hydrogen energy trains.

[0029] Figure 1 Figure 1 is one of the flowcharts of the running processing method of the rail vehicle provided by the present application.

[0030] The following will be combined with Figure 1 The process of the running processing method of the rail vehicle provided by the present application will be described.

[0031] In an exemplary embodiment of the present application, the running processing method of the rail vehicle can be applied to the rail vehicle. The rail vehicle can be powered by hydrogen fuel cells and lithium batteries. In combination with Figure 1 It can be known that the running processing method of the rail vehicle can include steps 110 and 120, which will be introduced respectively.

[0032] In step 110, a pre-trained running processing network model is called, wherein the running processing network model is used to determine the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell according to the running parameters of the rail vehicle at the current time.

[0033] In an embodiment, a pre-trained running processing network model can be called. In an example, the running processing network model can be a critic network model. The running processing network model can determine the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell according to the running parameters of the rail vehicle at the current time. It can be understood that if the rail vehicle runs according to the driving acceleration and the hydrogen fuel cell outputs power according to the output power, the running cost of the rail vehicle can be minimized. In another example, the running parameters can be the acceleration of the rail vehicle at the current time and the output power of the hydrogen fuel cell.

[0034] In step 120, the running parameters of the rail vehicle at the current time are obtained, and the running parameters are input into the running processing network model to obtain the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell output by the running processing network model, so that the running cost of the rail vehicle is minimized when the rail vehicle runs at the driving acceleration and the hydrogen fuel cell outputs power at the output power.

[0035] In another embodiment, the running parameters of the rail vehicle at the current time, such as the acceleration of the rail vehicle at the current time and the output power of the hydrogen fuel cell at the current time, can be obtained. Further, the running parameters are input into the running processing network model, and the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell output by the running processing network model can be obtained, so that the running cost of the rail vehicle is minimized when the rail vehicle runs at the driving acceleration and the hydrogen fuel cell outputs power at the output power. The parameters are controlled from the power source end and the load end to effectively reduce the energy consumption of the hydrogen energy train, thereby reducing the running cost of the rail vehicle.

[0036] The running processing method of the rail vehicle provided by the application is applied to the rail vehicle, the rail vehicle uses hydrogen fuel cells and lithium batteries as power sources, and the method comprises the following steps: calling a pre-trained running processing network model; obtaining running parameters of the rail vehicle at the current time, and inputting the running parameters into the running processing network model to obtain driving acceleration of the rail vehicle at the next time and output power of the hydrogen fuel cell output by the running processing network model, so that the running cost of the rail vehicle is minimized when the rail vehicle runs at the driving acceleration and the hydrogen fuel cell outputs power at the output power. The parameters are controlled from the power source end and the load end to effectively reduce the energy consumption of the hydrogen energy train, thereby reducing the running cost of the rail vehicle.

[0037] In another exemplary embodiment of the application, the previously described embodiments are taken as examples, wherein the running processing network model can be trained in the following way:

[0038] A training data set is constructed, wherein the training data set is determined based on an optimal data set, the optimal data set is determined based on a running curve and a power source curve determined based on historical running parameters of the rail vehicle, wherein the running curve is used to represent the corresponding relationship between the driving acceleration of the rail vehicle and the time when the running cost of the rail vehicle is minimized; the power source curve is used to represent the corresponding relationship between the output power of the hydrogen fuel cell and the time when the running cost of the rail vehicle is minimized;

[0039] Based on the training data set, the running processing network model is trained according to the proximal policy optimization algorithm to obtain the trained running processing network model.

[0040] In an embodiment, a training data set can be constructed, wherein the training data set is determined based on an optimal data set. In an example, the optimal data set is determined based on a running curve and a power source curve determined based on historical running parameters of the rail vehicle. The running curve can be used to represent a corresponding relationship between a driving acceleration of the rail vehicle and a time determined under a condition that a running cost of the rail vehicle is minimum; and the power source curve can be used to represent a corresponding relationship between an output power of the hydrogen fuel cell and the time determined under the condition that the running cost of the rail vehicle is minimum.

[0041] Further, the running processing network model is trained according to a proximal policy optimization algorithm based on the training data set, to obtain a trained running processing network model. The running processing network model trained based on the training data set can determine the driving acceleration of the rail vehicle and the output power of the hydrogen fuel cell at a next time according to the running parameters of the rail vehicle at a current time, so that the running cost of the rail vehicle is minimum under the condition that the rail vehicle is operated according to the driving acceleration and the hydrogen fuel cell outputs power according to the output power.

[0042] Figure 2 The process of training the running processing network model according to a proximal policy optimization algorithm based on the training data set to obtain a trained running processing network model is shown in the flowchart provided by the present application.

[0043] The following will be described in combination with Figure 2 The process of training the running processing network model according to a proximal policy optimization algorithm based on the training data set to obtain a trained running processing network model will be described.

[0044] In an exemplary embodiment of the present application, the running processing network model can include a critic network model. The following will be described in combination with Figure 2 It can be seen that the process of training the running processing network model according to a proximal policy optimization algorithm based on the training data set to obtain a trained running processing network model can include steps 210 to 240, which will be introduced respectively.

[0045] In step 210, the running state of the rail vehicle is determined based on the running curve and the power source curve in the training data set, wherein the running state includes the running speed, the state of charge of the hydrogen fuel cell and the running distance.

[0046] In an embodiment, the running state of the rail vehicle can be determined based on the running curve and the power source curve in the training data set. The running state includes the running speed v, the state of charge SoC of the hydrogen fuel cell and the running distance s.

[0047] In step 220, the critic network model is trained based on the training data set according to the proximal policy optimization algorithm.

[0048] In step 230, the running state is input into the actor network model to obtain a next-time execution action corresponding to the running state output by the actor network model, wherein the next-time execution action includes a predicted driving acceleration of the rail vehicle at the next time and a predicted output power of the hydrogen fuel cell under the running state.

[0049] In step 240, a network loss function is determined based on the next-time execution action and a next-time running state corresponding to the next-time execution action, and the running processing network model is iteratively trained based on the network loss function until the running processing network model converges, thereby obtaining a trained running processing network model.

[0050] In yet another embodiment, the critic network model can be trained based on the training data set according to the proximal policy optimization algorithm. Further, the running state is input into the actor network model to obtain a next-time execution action corresponding to the running state output by the actor network model, wherein the next-time execution action includes a predicted driving acceleration of the rail vehicle at the next time and a predicted output power of the hydrogen fuel cell under the running state. Further, a network loss function is determined based on the next-time execution action and a next-time running state corresponding to the next-time execution action, and the running processing network model is iteratively trained based on the network loss function until the running processing network model converges, thereby obtaining a trained running processing network model.

[0051] In yet another embodiment, the critic network model can be trained based on the training data set according to the proximal policy optimization algorithm. Further, the running state is input into the actor network model to obtain a next-time execution action corresponding to the running state output by the actor network model, wherein the next-time execution action includes a predicted driving acceleration of the rail vehicle at the next time and a predicted output power of the hydrogen fuel cell under the running state. Further, a network loss function is determined based on the next-time execution action and a next-time running state corresponding to the next-time execution action, and the running processing network model is iteratively trained based on the network loss function until the running processing network model converges, thereby obtaining a trained running processing network model.

[0052] In another embodiment, the cumulative reward of the current episode can be used to determine whether the current experience is a good experience, and it can be added to the good experience pool. When the good experience pool is full of m episodes, the cumulative reward of the (m+1)th episode is calculated to see if it is greater than the average reward of the previous m episodes. If so, the first episode is discarded and the (m+1)th episode is added to the experience pool. Furthermore, the state set S and action set in the experience pool can be input into the new actor network and the old actor network, respectively, to obtain the corresponding probability distributions πθ and πθ'. old Calculate J PPO (Corresponding to the proximal policy optimization algorithm) is used to update the parameters of the new actor network, thereby obtaining the actor network model.

[0053] Figure 3 This is a flowchart illustrating the process of determining the optimal dataset provided by the present invention.

[0054] The following will combine Figure 3 The process of determining the optimal dataset is explained.

[0055] In an exemplary embodiment of the present invention, combined with Figure 3 As can be seen, determining the optimal dataset may include steps 310 to 340, which will be described in detail below.

[0056] In step 310, an objective function for the operation of the rail vehicle is constructed, wherein the objective function is a function characterizing the operating cost of the rail vehicle.

[0057] In one embodiment, an objective function for the operation of a rail vehicle can be constructed, wherein the objective function can be used to characterize the operating cost of the rail vehicle.

[0058] Figure 5 This is a flowchart illustrating the process of determining the objective function provided by the present invention.

[0059] In yet another exemplary embodiment of the present invention, it can be combined with Figure 5 The process of determining the objective function is explained.

[0060] In an exemplary embodiment of the present invention, combined with Figure 5 As can be seen, determining the objective function may include steps 510 to 550, which will be described in detail below.

[0061] In step 510, the first equivalent hydrogen consumption of the hydrogen fuel cell, the second equivalent hydrogen consumption of the lithium battery, and the first value coefficient of hydrogen are determined.

[0062] In step 520, the first loss of the hydrogen fuel cell, the second loss of the lithium battery, the second value coefficient of the hydrogen fuel cell, and the third value coefficient of the lithium battery are determined.

[0063] In step 530, a hydrogen consumption cost of the rail vehicle during the operation is determined based on the first equivalent hydrogen consumption, the second equivalent hydrogen consumption, and the first value coefficient.

[0064] In step 540, a power source life cost of the rail vehicle during the operation is determined based on the first loss, the second loss, the second value coefficient, and the third value coefficient.

[0065] In step 550, a target function is determined according to the hydrogen consumption cost and the power source life cost.

[0066] In an embodiment, the target function can be expressed by formula (1):

[0067] (1)

[0068] Wherein, t represents the operation time; t f represents the operation time specified by the timetable; J H2 and J life respectively represent the hydrogen consumption cost and the power source life cost during the train running. The hydrogen consumption cost and the power source life cost can be respectively expressed by formula (2):

[0069] (2)

[0070] Wherein, C fcs represents the equivalent hydrogen consumption of the fuel cell (corresponding to the first equivalent hydrogen consumption of the hydrogen fuel cell); C bat represents the equivalent hydrogen consumption of the lithium battery (corresponding to the second equivalent hydrogen consumption of the lithium battery); a H2 represents the value coefficient of hydrogen (corresponding to the first value coefficient of hydrogen); L fcs represents the loss degree of the fuel cell (corresponding to the first loss of the hydrogen fuel cell); L bat represents the loss degree of the lithium battery system (corresponding to the second loss of the lithium battery); a fcs represents the value coefficient of the fuel cell system (corresponding to the second value coefficient of the hydrogen fuel cell); a bat represents the value coefficient of the lithium battery system (corresponding to the third value coefficient of the lithium battery).

[0071] In step 320, historical operation line information of the rail vehicle is obtained, wherein the historical operation line information includes speed limit information and station stay time information of a historical operation line of the rail vehicle.

[0072] In step 330, the historical operation line information and the target function are processed based on the Hermite-Simpson collocation technique to obtain an operation curve and a power source curve.

[0073] In step 340, based on the running curve and the power source curve, the optimal data set is determined.

[0074] In yet another embodiment, historical running line information of the rail vehicle can be acquired, wherein the historical running line information can include speed limit information and station stay duration information of the historical running line of the rail vehicle. Further, based on the Hermite-Simpson collocation technique, the historical running line information and the objective function are processed to obtain the running curve and the power source curve, and based on the running curve and the power source curve, the optimal data set is determined. In the application process, the Hermite-Simpson collocation method with adaptive grid updating can be used to solve under different initial conditions of the train (such as load, train speed limit, etc.), and the train running curve, power source output and other data (corresponding to the power source curve) under different conditions are obtained, thereby forming the optimal data set, thereby laying a foundation for effectively training the running processing network model.

[0075] In yet another exemplary embodiment of the present application, the previous embodiment described above is taken as an example for illustration, wherein based on the Hermite-Simpson collocation technique, the historical running line information and the objective function are processed to obtain the running curve and the power source curve, which can be realized in the following way:

[0076] Based on the Hermite-Simpson collocation technique, the objective function is processed to obtain a nonlinear programming corresponding to the objective function;

[0077] Based on the historical running line information and the nonlinear programming, the running curve and the power source curve are obtained.

[0078] In one embodiment, the objective function obtained above is a complex function with too high dimension of the state space, in order to simplify the operation, based on the Hermite-Simpson collocation technique, the objective function is processed to obtain a nonlinear programming corresponding to the objective function, that is, the original problem is converted into an NLP problem and is solved. Further, based on the historical running line information and the nonlinear programming, the running curve and the power source curve are obtained. Since the original problem (corresponding to the objective function) is converted into an NLP problem and is solved, the running curve and the power source curve can be quickly and efficiently obtained.

[0079] Figure 4 is a flowchart provided by the present application, which shows the process of processing the objective function based on the Hermite-Simpson collocation technique to obtain a nonlinear programming corresponding to the objective function.

[0080] The following will be described in combination with Figure 4 The process of processing the objective function based on the Hermite-Simpson collocation technique to obtain a nonlinear programming corresponding to the objective function will be described.

[0081] In an exemplary embodiment of the present application, in combination Figure 4 It can be known that, based on the Hermite-Simpson collocation technique, the dimension reduction processing is performed on the target function, and a nonlinear programming corresponding to the target function is obtained, which can include steps 410 to 450, which will be introduced respectively.

[0082] In step 410, the historical running line of the rail vehicle is divided into a preset number of running intervals, and the running interval is transformed in the time domain to obtain a transformed running interval, wherein the interval range of the transformed running interval is within a preset range.

[0083] In an embodiment, the train speed v, the position s and the lithium battery SOC can be taken as state variables, and the train acceleration a and the fuel cell system output power P fcs As a control variable, the state space can be represented as formula (3):

[0084] (3)

[0085] Wherein, X can represent a state variable; U can represent a control variable.

[0086] The system transfer equation can be represented as formula (4):

[0087] (4)

[0088] Wherein, I bat is the lithium battery current; Q bat is the lithium battery capacity.

[0089] In addition, the train operation also needs to meet the following condition restrictions, such as formula (5):

[0090] (5)

[0091] Wherein, v max represents the maximum speed; s terminal represents the distance to the end point; SOC max represents the maximum state of charge; SOC min represents the minimum state of charge; a max represents the maximum acceleration; a min represents the minimum acceleration; P max represents the maximum output power; P min represents the minimum output power.

[0092] In yet another embodiment, the train running interval can be divided into N intervals, and the N+1 segment points can be denoted as T0, T1, … T N . And each [T n-1 , T nThe intervals are all transformed to [-1, 1), which means dividing the historical operating lines of the rail vehicles into a preset number of operating intervals and transforming the operating intervals in the time domain to obtain the transformed operating intervals. Among them, the range of the transformed operating intervals can be considered to be in [-1, 1) within the preset range.

[0093] In one example, and each [T] in the time domain n-1 ,T n The interval [-1, 1) can be transformed to the interval [-1, 1) as shown in formula (6):

[0094] (6)

[0095] Where τ represents the operating interval in the time domain.

[0096] In step 420, the Legendre polynomial under the transformed operating interval is obtained based on the transformed operating interval.

[0097] In step 430, the state space of the objective function is discretized based on the Legendre polynomial in the transformed operating interval to obtain the discretized state space.

[0098] In another embodiment, the Legendre polynomial under the transformed operating interval can be obtained based on the transformed operating interval. Furthermore, based on the Legendre polynomial under the transformed operating interval, the state space in the objective function is discretized to obtain a discretized state space.

[0099] In yet another embodiment, a polynomial P can be selected. N (τ)+P N-1 The zeros of (τ) are collocated at τ, where P N (τ) represents an Nth-order Legendre orthogonal polynomial, where P N (τ) can be expressed as formula (7):

[0100] (7)

[0101] Where d represents the differentiation process;

[0102] Furthermore, the state space and action space can be discretized using N+1 and N Hermitian interpolation polynomials respectively, and can be expressed as formulas (8)-(9):

[0103] (8)

[0104] (9)

[0105] Among them, L n,i(τ) can be expressed as formula (10):

[0106] (10)

[0107] wherein P represents the total number of discrete points; i represents the discrete point number; j represents the discrete point number other than i; v(τ n,i ) represents the speed corresponding to the i-th discrete point in the n-th interval; s(τ n,i ) represents the displacement corresponding to the i-th discrete point in the n-th interval; soc(τ n,i ) represents the lithium battery state of charge corresponding to the i-th discrete point in the n-th interval; a(τ n,i ) represents the acceleration corresponding to the i-th discrete point in the n-th interval; P fcs (τ n,i ) represents the fuel cell power corresponding to the i-th discrete point in the n-th interval.

[0108] In step 440, the derivative of the discretized state space is determined based on the discretized state space.

[0109] In step 450, the derivative of the discretized state space is integrated to obtain a nonlinear programming corresponding to the objective function, so as to realize dimension reduction processing of the objective function.

[0110] In another embodiment, the derivative of the discretized state space can be determined based on the discretized state space, and the derivative of the discretized state space is integrated to obtain a nonlinear programming corresponding to the objective function, so as to realize dimension reduction processing of the objective function.

[0111] In an embodiment, after the state variable function is discretized, the derivative function can be expressed as formula (11):

[0112] (11)

[0113] In the formula, L n,i (τ) can be represented by a differential matrix D k,i , and D k,i can be expressed as formula (12):

[0114] (12)

[0115] wherein v(τ n,k ) represents the speed corresponding to the k-th discrete point in the n-th interval; s(τ n,k ) represents the displacement corresponding to the k-th discrete point in the n-th interval; soc(τ n,k ) represents the lithium battery state of charge corresponding to the k-th discrete point in the n-th interval; g(τ n,k) represents the function value corresponding to the kth discrete point in the nth interval, wherein the function expression corresponding to the function value is g(τ);g(τ n,i ) represents the function value corresponding to the ith discrete point in the nth interval;τ n,i represents the ith discrete point in the nth interval;τ n,k represents the kth discrete point in the nth interval.

[0116] Further, the algebraic constraint can be expressed as formula (13):

[0117] (13)

[0118] wherein D k,i represents a differential matrix.

[0119] The original objective function is replaced by Simpson integral to transform the original problem into an NLP problem, and the NLP corresponding to the objective function is obtained by solving, so that the objective function is processed by dimension reduction. The deformation result is formula (14):

[0120] (14)

[0121] In another example embodiment of the application, the above-mentioned embodiments are taken as examples, wherein the preset number of operation intervals can be determined in the following manner: Figure 4 The allowable error of each operation interval is determined;

[0122] In the case that the allowable error of the operation interval is greater than the maximum allowable error, the operation interval is re-divided until the allowable error of the preset number of operation intervals obtained by re-division is less than the maximum allowable error and greater than the minimum allowable error;

[0123] In the case that the allowable error of the operation interval is less than the minimum allowable error, the operation interval is merged until the allowable error of the preset number of operation intervals obtained by merging is less than the maximum allowable error and greater than the minimum allowable error.

[0124] In the case that the allowable error of the operation interval is less than the minimum allowable error, the operation interval is merged until the allowable error of the preset number of operation intervals obtained by merging is less than the maximum allowable error and greater than the minimum allowable error.

[0125] In an embodiment, the maximum allowable error εmax can be set in each independent interval, and when the relative allowable error (corresponding to the allowable error of the operation interval) ε of the interval is greater than εmax, the interval needs to be re-refined or the order of the interpolation polynomial is increased, otherwise the adjacent intervals can be merged or the order of the interpolation polynomial is reduced. The relative allowable error ε (which can correspond to ε(τ) in the formula) of the interval can be expressed as formula (15):

[0126] (15)

[0127] wherein LN(ζ) is the N-th derivative of L n,i (τ) of the N-th order, ζ is a number in [t0, t f ] and X(τ) represents the function value corresponding to the point τ, wherein X(.) represents the aforementioned speed, acceleration, SOC, displacement, fuel cell power, etc.

[0128] wherein the maximum allowable error and the minimum allowable error can be adjusted according to the actual situation, and the maximum allowable error and the minimum allowable error are not specifically limited in the embodiment.

[0129] According to the foregoing description, the operation processing method of the rail vehicle provided by the application is applied to the rail vehicle, the rail vehicle uses hydrogen fuel cells and lithium batteries as power sources, and the method comprises the following steps: calling a pre-trained operation processing network model; acquiring operation parameters of the rail vehicle at a current time, inputting the operation parameters into the operation processing network model, and obtaining driving acceleration of the rail vehicle at a next time and output power of the hydrogen fuel cells output by the operation processing network model, so that the operation cost of the rail vehicle is minimized under the condition that the rail vehicle is operated according to the driving acceleration and the hydrogen fuel cells output power according to the output power. The parameters are controlled from the power source end and the load end to effectively reduce the energy consumption of the hydrogen energy train, thereby reducing the operation cost of the rail vehicle.

[0130] The operation processing device of the rail vehicle provided by the application will be described below. The operation processing device of the rail vehicle described below can be correspondingly referred to the operation processing method of the rail vehicle described above.

[0131] Figure 6 is a structural schematic diagram of the operation processing device of the rail vehicle provided by the application.

[0132] The operation processing device of the rail vehicle provided by the application will be described below. Figure 6 The operation processing device of the rail vehicle provided by the application will be described below.

[0133] In an exemplary embodiment of the application, the device can be applied to a rail vehicle, and the rail vehicle uses hydrogen fuel cells and lithium batteries as power sources. The operation processing device of the rail vehicle provided by the application will be described below. Figure 6 It can be known that the operation processing device of the rail vehicle can comprise a calling module 610 and a processing module 620, which will be introduced respectively.

[0134] The calling module 610 can be configured to call a pre-trained operation processing network model, wherein the operation processing network model is used to determine driving acceleration of the rail vehicle at a next time and output power of the hydrogen fuel cells according to operation parameters of the rail vehicle at a current time;

[0135] The processing module 620 can be configured to acquire a running parameter of the rail vehicle at a current time, input the running parameter into the running processing network model, and obtain a driving acceleration of the rail vehicle at a next time and an output power of the hydrogen fuel cell output by the running processing network model, so that the running cost of the rail vehicle is minimized when the rail vehicle runs at the driving acceleration and the hydrogen fuel cell outputs power at the output power.

[0136] In an example embodiment of the present application, the calling module 610 can implement the training of the running processing network model in the following manner:

[0137] A training data set is constructed, wherein the training data set is determined based on an optimal data set, and the optimal data set is determined based on a running curve and a power source curve determined based on historical running parameters of the rail vehicle, wherein the running curve is used to represent a corresponding relationship between a driving acceleration of the rail vehicle and time determined when the running cost of the rail vehicle is minimized; and the power source curve is used to represent a corresponding relationship between an output power of the hydrogen fuel cell and time determined when the running cost of the rail vehicle is minimized.

[0138] The running processing network model is trained based on the training data set and according to a proximal policy optimization algorithm to obtain a trained running processing network model.

[0139] In an example embodiment of the present application, the running processing network model includes a critic network model; the calling module 610 can implement the training of the running processing network model based on the training data set and according to the proximal policy optimization algorithm to obtain a trained running processing network model in the following manner:

[0140] The running state of the rail vehicle is determined based on the running curve and the power source curve in the training data set, wherein the running state includes a running speed, a state of charge of the hydrogen fuel cell, and a running distance.

[0141] An actor network model corresponding to the critic network model is trained based on the training data set and according to the proximal policy optimization algorithm.

[0142] The running state is input into the actor network model to obtain a next-time execution action corresponding to the running state output by the actor network model, wherein the next-time execution action includes a predicted driving acceleration of the rail vehicle at a next time and a predicted output power of the hydrogen fuel cell under the running state.

[0143] Based on the next time point action and the next time point running state corresponding to the next time point action, a network loss function is determined, and the running processing network model is iteratively trained based on the network loss function until the running processing network model converges, to obtain a trained running processing network model.

[0144] In an example embodiment of the present application, the calling module 610 can determine the optimal data set in the following manner:

[0145] A target function of the rail vehicle in the running process is constructed, wherein the target function is a function representing the running cost of the rail vehicle;

[0146] Historical running line information of the rail vehicle is obtained, wherein the historical running line information includes speed limit information and station stay duration information of the historical running line of the rail vehicle;

[0147] Based on the Hermite-Simpson collocation point technology, the historical running line information and the target function are processed to obtain the running curve and the power source curve;

[0148] Based on the running curve and the power source curve, the optimal data set is determined.

[0149] In an example embodiment of the present application, the calling module 610 can implement processing of the historical running line information and the target function based on the Hermite-Simpson collocation point technology to obtain the running curve and the power source curve in the following manner:

[0150] Based on the Hermite-Simpson collocation point technology, the target function is processed in a dimension reduction manner to obtain a nonlinear programming corresponding to the target function;

[0151] Based on the historical running line information and the nonlinear programming, the running curve and the power source curve are obtained.

[0152] In an example embodiment of the present application, the calling module 610 can implement processing of the target function based on the Hermite-Simpson collocation point technology to obtain a nonlinear programming corresponding to the target function in the following manner:

[0153] The historical running line of the rail vehicle is divided into a preset number of running intervals, and the running intervals are processed in a time domain to obtain transformed running intervals, wherein the interval range of the transformed running intervals is within a preset range;

[0154] Based on the transformed running intervals, a Legendre polynomial under the transformed running intervals is obtained;

[0155] Discretize the state space in the target function based on the Legendre polynomials in the transformed operation intervals, to obtain a discretized state space;

[0156] Determine the derivative of the discretized state space based on the discretized state space;

[0157] Integrate the derivative of the discretized state space to obtain a nonlinear programming corresponding to the target function, to realize dimension reduction processing of the target function.

[0158] In an exemplary embodiment of the present application, the calling module 610 can determine the preset number of operation intervals in the following manner:

[0159] Determine the allowable error of each operation interval;

[0160] In the case where the allowable error of the operation interval is greater than the maximum allowable error, the operation interval is re-divided until the allowable error of the preset number of operation intervals obtained by re-division is less than the maximum allowable error and greater than the minimum allowable error;

[0161] In the case where the allowable error of the operation interval is less than the minimum allowable error, the operation interval is merged until the allowable error of the preset number of operation intervals obtained by merging is less than the maximum allowable error and greater than the minimum allowable error.

[0162] In an exemplary embodiment of the present application, the calling module 610 can determine the target function in the following manner:

[0163] Determine the first equivalent hydrogen consumption of the hydrogen fuel cell, the second equivalent hydrogen consumption of the lithium battery, and the first value coefficient of hydrogen;

[0164] Determine the first loss of the hydrogen fuel cell, the second loss of the lithium battery, the second value coefficient of the hydrogen fuel cell, and the third value coefficient of the lithium battery;

[0165] Determine the hydrogen consumption cost of the rail vehicle in the running process based on the first equivalent hydrogen consumption, the second equivalent hydrogen consumption, and the first value coefficient;

[0166] Determine the power source life cost of the rail vehicle in the running process based on the first loss, the second loss, the second value coefficient, and the third value coefficient;

[0167] Determine the target function according to the hydrogen consumption cost and the power source life cost.

[0168] Based on the same inventive concept, the present application also provides a rail vehicle, which will be described below in connection with the following embodiments.

[0169] In an exemplary embodiment of the present application, the rail vehicle can comprise a rail vehicle body and a processor, wherein the processor is configured to execute the operation processing method of the rail vehicle as described in any one of the preceding embodiments. Through this embodiment, parameter control is performed from the power source end and the load end to effectively reduce the energy consumption of the hydrogen energy train, thereby reducing the operation cost of the rail vehicle.

[0170] Figure 7 An example of an electronic device is shown in the physical structure diagram of the electronic device as shown in Figure 7 The electronic device can include a processor 710, a communications interface 720, a memory 730, and a communications bus 740, wherein the processor 710, the communications interface 720, and the memory 730 communicate with each other through the communications bus 740. The processor 710 can invoke the logic instructions in the memory 730 to execute the operation processing method of the rail vehicle, which is applied to a rail vehicle powered by a hydrogen fuel cell and a lithium battery, and the method comprises: invoking a pre-trained operation processing network model, wherein the operation processing network model is configured to determine the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell according to the operation parameters of the rail vehicle at the current time; obtaining the operation parameters of the rail vehicle at the current time, and inputting the operation parameters into the operation processing network model to obtain the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell output by the operation processing network model, so that the rail vehicle has the minimum operation cost when it operates according to the driving acceleration and the hydrogen fuel cell outputs power according to the output power.

[0171] In addition, the logic instructions in the memory 730 described above can be implemented in the form of a software function unit and sold or used as an independent product, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in various embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0172] In another aspect, the present application also provides a computer program product, the computer program product comprising a computer program, the computer program being stored on a non-transitory computer readable storage medium, and the computer program being executable by a processor to cause a computer to perform the operation processing method of the rail vehicle provided by the above method, the method being applied to a rail vehicle powered by a hydrogen fuel cell and a lithium battery, and the method comprising: calling a pre-trained operation processing network model, wherein the operation processing network model is used to determine a driving acceleration of the rail vehicle at a next time and an output power of the hydrogen fuel cell according to an operation parameter of the rail vehicle at a current time; obtaining the operation parameter of the rail vehicle at the current time, and inputting the operation parameter into the operation processing network model to obtain the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell output by the operation processing network model, so that the rail vehicle has a minimum operation cost when the rail vehicle is operated according to the driving acceleration and the hydrogen fuel cell outputs power according to the output power.

[0173] In yet another aspect, the present application also provides a non-transitory computer-readable storage medium having stored thereon a computer program, which, when executed by a processor, implements the operation processing method of the rail vehicle provided by the above method, the method being applied to a rail vehicle powered by a hydrogen fuel cell and a lithium battery, and the method comprising: calling a pre-trained operation processing network model, wherein the operation processing network model is configured to determine a driving acceleration of the rail vehicle at a next time and an output power of the hydrogen fuel cell according to operation parameters of the rail vehicle at a current time; obtaining the operation parameters of the rail vehicle at the current time, and inputting the operation parameters into the operation processing network model to obtain the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell output by the operation processing network model, so that the operation cost of the rail vehicle is minimized when the rail vehicle is operated at the driving acceleration and the hydrogen fuel cell outputs power at the output power.

[0174] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0175] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and necessary universal hardware platforms, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions, essentially or in terms of contribution to the prior art, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as a ROM / RAM, a magnetic disk, or an optical disk, and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0176] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method of operation processing of a railway vehicle, characterized by, The method is applied to a rail vehicle powered by a hydrogen fuel cell and a lithium battery, and the method comprises: calling a pre-trained operation processing network model, wherein the operation processing network model is used to determine a driving acceleration of the rail vehicle at a next time and an output power of the hydrogen fuel cell according to operation parameters of the rail vehicle at a current time; obtaining the operation parameters of the rail vehicle at the current time, and inputting the operation parameters into the operation processing network model to obtain the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell output by the operation processing network model, so that the operation cost of the rail vehicle is minimized when the rail vehicle operates at the driving acceleration and the hydrogen fuel cell outputs power at the output power, wherein the operation processing network model is obtained by training a training data set, the training data set is determined based on an optimal data set, and the optimal data set is determined based on an operation curve and a power source curve of the rail vehicle, wherein the operation curve is used to represent a corresponding relationship between the driving acceleration of the rail vehicle and time when the operation cost of the rail vehicle is minimized, and the power source curve is used to represent a corresponding relationship between the output power of the hydrogen fuel cell and time when the operation cost of the rail vehicle is minimized.

2. The operation processing method of a rail vehicle according to claim 1, characterized by, The operation processing network model is trained in the following manner: based on the training data set, the operation processing network model is trained according to a proximal policy optimization algorithm to obtain a trained operation processing network model.

3. The operation processing method of a rail vehicle according to claim 2, characterized by, The operation processing network model comprises a critic network model; and the operation processing network model is trained based on the training data set according to a proximal policy optimization algorithm to obtain a trained operation processing network model, specifically comprising: determining an operation state of the rail vehicle based on the operation curve and the power source curve in the training data set, wherein the operation state comprises an operation speed, a state of charge of the hydrogen fuel cell and an operation distance; training an actor network model corresponding to the critic network model according to a proximal policy optimization algorithm based on the training data set; inputting the operation state into the actor network model to obtain a next time execution action corresponding to the operation state output by the actor network model, wherein the next time execution action comprises a predicted driving acceleration of the rail vehicle at the next time and a predicted output power of the hydrogen fuel cell under the operation state; determining a network loss function based on the next time execution action and a next time operation state corresponding to the next time execution action, and iteratively training the operation processing network model based on the network loss function until the operation processing network model converges to obtain a trained operation processing network model.

4. The operation processing method of a rail vehicle according to claim 2 or 3, characterized by, The optimal data set is determined in the following manner: constructing a target function of the rail vehicle in a running process, wherein the target function is a function representing a running cost of the rail vehicle; obtaining historical running line information of the rail vehicle, wherein the historical running line information comprises speed limit information and station stay duration information of a historical running line of the rail vehicle; processing the historical running line information and the target function based on a Hermite-Simpson collocation technique to obtain the running curve and the power source curve; determining the optimal data set based on the running curve and the power source curve.

5. The operation processing method of a rail vehicle according to claim 4, characterized by, The processing of the historical running line information and the target function based on the Hermite-Simpson collocation technique to obtain the running curve and the power source curve specifically comprises: dimension reduction processing of the target function based on the Hermite-Simpson collocation technique to obtain a nonlinear programming corresponding to the target function; obtaining the running curve and the power source curve based on the historical running line information and the nonlinear programming.

6. The operation processing method of a rail vehicle according to claim 5, characterized by The dimension reduction processing of the target function based on the Hermite-Simpson collocation technique to obtain a nonlinear programming corresponding to the target function specifically comprises: dividing the historical running line of the rail vehicle into a preset number of running intervals, and performing transformation processing on the running intervals in the time domain to obtain transformed running intervals, wherein the interval range of the transformed running intervals is within a preset range; obtaining Legendre polynomials under the transformed running intervals based on the transformed running intervals; discretization processing of a state space in the target function based on the Legendre polynomials under the transformed running intervals to obtain a discretized state space; determining the derivative of the discretized state space based on the discretized state space; integral processing of the derivative of the discretized state space to obtain a nonlinear programming corresponding to the target function, so as to realize the dimension reduction processing of the target function.

7. The operation processing method of a rail vehicle according to claim 6, characterized by, The preset number of running intervals is determined in the following manner: determining the allowable error of each running interval; in the case that the allowable error of a running interval is greater than the maximum allowable error, re-dividing the running interval until the allowable error of the preset number of running intervals obtained by re-division is less than the maximum allowable error and greater than the minimum allowable error; in the case that the allowable error of a running interval is less than the minimum allowable error, performing merging processing on the running interval until the allowable error of the preset number of running intervals obtained after merging processing is less than the maximum allowable error and greater than the minimum allowable error.

8. The operation processing method of a rail vehicle according to claim 5, characterized by, The target function is determined in the following manner: determining the first equivalent hydrogen consumption of the hydrogen fuel cell, the second equivalent hydrogen consumption of the lithium battery, and the first value coefficient of hydrogen; determining the first loss of the hydrogen fuel cell, the second loss of the lithium battery, the second value coefficient of the hydrogen fuel cell, and the third value coefficient of the lithium battery; determining the hydrogen consumption cost of the rail vehicle in the running process based on the first equivalent hydrogen consumption, the second equivalent hydrogen consumption, and the first value coefficient. determine a power source life cost of the rail vehicle during operation based on the first loss, the second loss, the second value coefficient, and the third value coefficient; determine the target function based on the hydrogen consumption cost and the power source life cost.

9. An operation processing device of a railway vehicle, characterized by, The device is applied to a rail vehicle, the rail vehicle taking a hydrogen fuel cell and a lithium battery as power sources, and the device is used to implement the operation processing method of the rail vehicle according to any one of claims 1 to 8, and the device comprises: a calling module configured to call a pre-trained operation processing network model, wherein the operation processing network model is configured to determine driving acceleration of the rail vehicle at a next time and output power of the hydrogen fuel cell according to operation parameters of the rail vehicle at a current time; a processing module configured to obtain the operation parameters of the rail vehicle at the current time, input the operation parameters into the operation processing network model, and obtain the driving acceleration of the rail vehicle at the next time and the output power of the hydrogen fuel cell output by the operation processing network model, so that, in a case where the rail vehicle operates at the driving acceleration and the hydrogen fuel cell outputs power at the output power, an operation cost of the rail vehicle is minimized.

10. A rail vehicle, characterized by The rail vehicle comprises: a rail vehicle body, and a processor configured to execute the operation processing method of the rail vehicle according to any one of claims 1 to 8.

Images