New energy power system modeling and hydrogen consumption prediction analysis method

Abstract

The invention relates to a new energy power system modeling and hydrogen consumption prediction analysis method, and belongs to the technical field of new energy automobile power systems. Aiming at the problems of low model precision, large energy consumption prediction deviation and high experiment cost caused by nonlinear coupling characteristics of an existing fuel cell hybrid power system, the invention provides a nonlinear regression model construction method based on a least square support vector machine (LSSVM). A model is trained by collecting differential energy consumption delta E, transient motor speed n, torque T and accumulated energy consumption E (t) data of a vehicle CAN bus, the current and voltage of a direct-current bus are predicted, and accurate energy consumption prediction is achieved in combination with a hydrogen consumption formula and an SOC calculation model. According to the method, the dependence on a mechanism model is reduced through black box modeling, the nonlinear characteristics are processed by adopting the RBF kernel function, the prediction precision is remarkably improved, the experiment cost is reduced, and the method is suitable for energy management optimization of the extended-range fuel cell vehicle.

Description

Technical Field

[0001] The present invention belongs to the technical field of new energy vehicle power systems, and relates to a new energy power system modeling and hydrogen consumption prediction analysis method. Background Art

[0002] Currently, due to the characteristics of slow dynamic response and energy recovery during faults of fuel cell vehicles (FCHVs), an energy storage system (battery and / or supercapacitor bank) is usually integrated with a fuel cell system (FCS) to form a hybrid power system.

[0003] The range-extended fuel cell hybrid vehicle has become a competitive solution at present due to its advantages such as convenient refueling, extended driving range, and reduced cost of the fuel cell stack.

[0004] However, the power system of the range-extended FCHV has multiple non-linear coupling characteristics, which are difficult to accurately describe by traditional mathematical models, resulting in deviations in the design of energy management strategies and energy consumption prediction, and unable to accurately evaluate the energy consumption of FCHVs.

[0005] Existing technologies mostly use simplified empirical models or single control variables (such as battery SOC) for power distribution, without fully considering the non-linear efficiency characteristics of the powertrain, resulting in a large difference between simulation and measured energy consumption.

[0006] Existing energy consumption prediction methods rely on a large amount of experimental data, with high costs and long cycles.

[0007] The methods based on simplified empirical models or single variables cannot accurately reflect the actual operation of the vehicle, and the prediction accuracy is insufficient.

[0008] The present invention aims to solve the above problems, and provides a new energy power system modeling and hydrogen consumption prediction analysis method to improve the energy consumption prediction accuracy and make the energy consumption prediction results of the range-extender FCHV as consistent with the actual energy consumption as possible.

[0009] By establishing a non-linear regression model based on the LSSVM intelligent algorithm, integrating the non-linear characteristics of the powertrain, reducing the dependence on the mechanism models of complex components, and reducing the experimental costs and improving the prediction efficiency. Summary of the Invention

[0010] In view of this, the purpose of the present invention is to provide a new energy power system modeling and hydrogen consumption prediction analysis method. This model regards the powertrain as an overall "black box", establishes a non-linear regression model based on an intelligent algorithm, and then calculates the energy consumption of the fuel cell, improves the energy consumption prediction accuracy, and makes the energy consumption prediction results of the range-extender FCHV as consistent with the actual energy consumption as possible.

[0011] To achieve the above purpose, the present invention provides the following technical solutions:

[0012] A new energy power system modeling and hydrogen consumption prediction analysis method, comprising the following steps:

[0013] S1: Establish a non-linear regression model of the new energy vehicle power system based on the least squares support vector machine (LSSVM) algorithm model, where the model includes a first sub-model for predicting the DC bus current and a second sub-model for predicting the DC bus voltage;

[0014] S2: Collect time series data from the vehicle CAN bus, including differential energy consumption ΔE, transient motor speed n, transient motor torque T, and cumulative energy consumption E(t), construct a training data set, and train the non-linear regression model;

[0015] S3: Input the real-time data of the power system to be measured into the trained model to obtain the predicted DC bus current I(t) and DC bus voltage U(t);

[0016] S4: According to the predicted DC bus current I(t) and voltage U(t), combined with the hydrogen consumption calculation formula, calculate the hydrogen consumption m of the fuel cell H2 and the state of charge SOC(t) of the battery.

[0017] Further, in S1, the input parameters of the first sub-model include differential energy consumption ΔE, transient motor speed n, and transient motor torque T; the input parameters of the second sub-model include differential energy consumption ΔE, transient motor speed n, transient motor torque T, and cumulative energy consumption E(t).

[0018] Further, the LSSVM algorithm model uses the radial basis function (RBF) as the kernel function to map the input data to a high-dimensional feature space to achieve non-linear regression.

[0019] Further, in S4, the hydrogen consumption is calculated by the formula:

[0020]

[0021] where is the molar mass of hydrogen, n cell is the number of cells in the fuel cell stack, I FC is the output current of the fuel cell, and F is the Faraday constant.

[0022] Further, in S4, the formula for calculating the state of charge SOC(t) of the battery is:

[0023]

[0024] where SOC initialis the initial state of charge, Q is the battery capacity, and I(t) is the transient DC bus current.

[0025] Furthermore, the training data of the non-linear regression model further includes the internal resistance R of the lithium battery BAT , the mapping relationship between the battery capacity Q, the open-circuit voltage OCV(t) and SOC(t).

[0026] Furthermore, in S3, the prediction of the DC bus current I(t) is further based on the following relational expression:

[0027] I(t) = f(T(t), n(t), ΔE)

[0028] where, T(t) is the transient motor torque, n(t) is the transient motor speed, and ΔE is the differential energy consumption.

[0029] Furthermore, in S3, the prediction of the DC bus voltage U(t) is further based on the following relational expression:

[0030] U(t) = h(E(t), I(t))

[0031] where, E(t) is the cumulative energy consumption, and I(t) is the transient DC bus current.

[0032] Furthermore, in S2, the time series data is obtained through a discretized lithium battery model, and the discretized model includes:

[0033] U(t) = OCV(t) - R BAT ·I(t)

[0034]

[0035] where, OCV(t) is the open-circuit voltage, R BAT is the internal resistance of the lithium battery, ΔSOC is the differential state of charge, and Δt is the time interval.

[0036] Furthermore, the hidden layer of the LSSVM algorithm model includes at least one first hidden layer containing 50 neurons and a second hidden layer containing 80 neurons, which is used to improve the prediction accuracy of non-linear regression.

[0037] The beneficial effects of the present invention are as follows:

[0038] (1) The LSSVM model has better performance in predicting the DC bus current and voltage than traditional BP networks, GABP, etc., and significantly improves the accuracy of fuel cell energy consumption calculation.

[0039] (2) The present invention reduces the dependence on the mechanism models of complex components of fuel cell vehicles through "black box" modeling. Only data such as differential energy consumption, transient motor speed n, transient motor torque T, and cumulative energy consumption E(t) are required to complete model training, reducing experimental costs and improving prediction efficiency.

[0040] (3) Based on the principle of structural risk minimization, LSSVM can not only fit the training data well but also maintain good prediction performance on new data, showing strong applicability.

[0041] (4) The present invention constructs a model by regarding the nonlinear subsystem of the powertrain as an overall black box, fully considering the nonlinear characteristics of the powertrain. Compared with traditional empirical models, it can more accurately reflect the actual operation of the vehicle and make the energy consumption prediction results closer to the actual values.

[0042] Other advantages, objectives, and features of the present invention will be described to some extent in the subsequent specification, and to some extent, will be obvious to those skilled in the art based on the study of the following text, or can be taught from the practice of the present invention. The objectives and other advantages of the present invention can be achieved and obtained through the following specification. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in detail with reference to the accompanying drawings, where:

[0044] Figure 1 is a schematic diagram of the system flow;

[0045] Figure 2 is a diagram of the model network architecture; (a) is the least squares method model for current prediction; (b) is the least squares method model for voltage prediction; (c) is the support vector machine model for current prediction; (d) is the support vector machine model for voltage prediction;

[0046] Figure 3 is a schematic diagram of the layout of the hybrid powertrain. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0047] The following specific examples illustrate the embodiments of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the diagrams provided in the following embodiments only illustrate the basic concept of the present invention schematically, and the following embodiments and the features in the embodiments can be combined with each other without conflict.

[0048] Among them, the attached drawings are only for illustrative purposes, showing only schematic diagrams rather than physical diagrams, and should not be construed as a limitation on the present invention; for better illustration of the embodiments of the present invention, some components in the attached drawings will be omitted, enlarged or reduced, which do not represent the dimensions of the actual product; for those skilled in the art, it is understandable that some well-known structures and their descriptions in the attached drawings may be omitted.

[0049] In the attached drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "rear", etc. indicating the orientation or positional relationship, they are based on the orientation or positional relationship shown in the attached drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the attached drawings are only for illustrative purposes and should not be construed as a limitation on the present invention. For those of ordinary skill in the art, the specific meanings of the above terms can be understood according to specific circumstances.

[0050] As Figure 1 shown, the present invention provides a method for establishing a new energy vehicle power system model based on the LSSVM intelligent algorithm and carrying out an extended-range fuel cell hybrid power and energy consumption prediction method in combination with the constructed model, including:

[0051] 1. Establish a new energy vehicle power system model based on the LSSVM intelligent algorithm

[0052] 2. Collect time series data from the vehicle's CAN bus and establish a model training database to train the model;

[0053] 3. Import the data of the battery to be measured into the established model to obtain the predicted DC bus current and voltage;

[0054] 4. Calculate the energy consumption through the predicted DC bus current and voltage.

[0055] Least Squares Support Vector Machine (LSSVM) is an improved support vector machine based on statistical theory. It simplifies the problem-solving process by transforming the quadratic optimization problem into the solution of a linear equation system. The goal of the LSSVM model is to find a function that maps the input space to the feature space for linear regression in the feature space. Assuming the input sample is x ∈ Rn and the output is y ∈ R, the functional form of the LSSVM model is:

[0056]

[0057] Among them: K(x, x i) is the kernel function, which realizes the function of mapping the data in the input space to a high-dimensional feature space, making the data that is non-linearly separable in the original input space become linearly separable in the feature space. Common kernel functions include the radial basis kernel function (RBF). a i are the Lagrange multipliers, where i = 1, 2, …, N and N is the number of training samples, and b is the bias term.

[0058] Determine a i and b by solving an optimization problem, so that the sum of squared errors of the model on the training data is minimized while satisfying certain constraint conditions, thereby obtaining the optimal function f(x) for predicting new input data.

[0059] Establish a non-linear regression model according to the LSSVM algorithm, as Figure 2 shown, Figure 2 (a) and Figure 2 (b) show the neural network architecture of the non-linear regression model. Figure 2 (a) is used to predict the DC bus current. Figure 2 (b) is used to predict the DC bus voltage. In Figure 2 (a), there are three inputs and one output in the network. The first hidden layer has 50 neurons, and the second hidden layer has 80 neurons. In Figure 2 (b), there are four inputs and one output in the network, and the architecture of the hidden layer is the same as Figure 2 (a). Figure 2 (c) and Figure 2 (d) show the LSSVM architecture of the non-linear regression model, where Figure 2 (c) is used to predict the DC bus current. Figure 2 (d) is used to predict the DC bus voltage. The kernel function of LSSVM is the radial basis function (RBF) kernel. The RBF kernel can map the data in the low-dimensional input space to a high-dimensional feature space, making it possible for the data to become linearly separable in the high-dimensional space. It transforms the non-linear relationship that is difficult to handle in the low-dimensional space into a linear relationship in the high-dimensional space by calculating the similarity between sample points, so that LSSVM can effectively process this data. The centers of the hidden nodes are the support vectors and the Lagrange multipliers, which determine the relative significance of the input data to the output data. LSSVM replaces the inequality constraints of the SVM problem with a set of linear equality constraints, which leads to the simplification of the solution of the Lagrange multipliers.

[0060] The energy consumption of the hydrogen fuel cell can be expressed using the hydrogen consumption and the state of charge (SOC) of the fuel cell. From Equation (2) and Equation (3) according to Figure 3For power balance and current balance in a hybrid system, it is necessary to know the current and voltage on the DC bus. The energy consumption of an extended-range fuel cell can be calculated according to the following formula.

[0061]

[0062] In the formula, is the hydrogen consumption, is the molar mass of hydrogen, n cell is the number of batteries, I FC is the output current of the fuel cell, F is the Faraday constant, SOC(t) is the transient charge state, SOC initial is the initial state of charge, Q is the battery capacity, and I(t) is the transient DC bus current.

[0063] To predict the DC bus current and voltage, appropriate process variables need to be selected as the inputs of the model. The differential energy consumption ΔE, the transient motor speed n, and the transient motor torque T can be used as the inputs of an intelligent algorithm to predict the transient DC bus current; the differential energy consumption ΔE, the transient motor speed n, the transient motor torque T, and the cumulative energy consumption E(t) are used to predict the transient DC bus voltage. The following is the specific derivation process:

[0064] First, discretize the lithium battery model as follows:

[0065]

[0066] R BAT represents the internal resistance. The differential state of discharge during Δt is ΔSOD, and the decrease in the difference of SOC can be written as ΔSOC.

[0067] From Equation (4-b) and Equation (4-c), the transient DC bus current I(t) is related to the battery capacity Q, the differential time Δt, and the differential charge state ΔSOC (or differential discharge state ΔSOD), and can be given by the following abstract expression:

[0068] I(t) = f(Q, Δt, ΔSOC) (5)

[0069] The decrease in SOC means the energy consumption of the vehicle. The power consumed by the vehicle can be calculated as follows:

[0070]

[0071] where T(t) is the transient torque of the motor and n(t) is the transient speed of the motor.

[0072] Then the differential energy consumption of the vehicle during Δt can be given by the following formula:

[0073] ΔE = P require (t)·Δt (7)

[0074] Therefore, the relationship between the differential charge state ΔSOC and the differential energy consumption ΔE can be constructed as follows:

[0075]

[0076] Substituting Equation (8) into Equation (5), the transient DC bus current can be rewritten as:

[0077] I(t) = f(T(t), n(t), ΔE) (9)

[0078] According to Equation (6) and Equation (7), the differential energy consumption ΔE is determined by the transient motor torque T(t), and the transient motor speed n(t) and the differential time Δt can be given by the following abstract expression:

[0079] ΔE = g(T(t), n(t), Δt) (10)

[0080] Substituting Equation (10) into Equation (9), the transient DC bus current can be rewritten as:

[0081] I(t) = f(Q, Δt, g(T(t), n(t), Δt) (11)

[0082] Then Equation (10) can be derived into the following form:

[0083] I(t) = f(Q, T(t), n(t), Δt) (12)

[0084] Therefore, the transient DC bus current is related to the motor torque and the motor speed. The independent variable Δt can be explained according to Equation (7). The independent variable Q is a constant and can be ignored in Equation (12) because it does not affect the dynamic properties. Finally, the transient DC bus current can be given by the following formula:

[0085] I(t) = f(T(t), n(t), ΔE) (13)

[0086] From Equation (13), the differential energy consumption, the transient motor speed, and the transient motor torque can be used as inputs of the intelligent algorithm to predict the transient DC bus current. The desired function of the transient DC bus current curve can be expressed in the standard form as:

[0087] I(t) = f(T(t), n(t), ΔE; β) + ε (14)

[0088] where β = (β1, β2, …, βm) is the regression coefficient and ε is the random error.

[0089] To predict the transient DC bus voltage U(t), the model requires more input variables, and the derivation of these variables is as follows.

[0090] There is a relationship between the transient OCV(t) and the transient SOC(t), as described in Equation (4-d). Substituting Equation (4-d) into Equation (4-a), the abstract relationship between the transient DC bus voltage U(t), the transient SOC(t), and the transient DC bus current I(t) can be described as:

[0091] U(t) = h(SOC(t), I(t)) (15)

[0092] The transient SOC(t) can be expressed as the cumulative reduction of SOC from the initial moment at each moment, SOCinitial. Then, the cumulative reduction of SOC can be replaced by the cumulative energy consumption E(t) at each moment relative to the initial moment. Therefore, the relationship between the transient SOC(t) and the cumulative energy consumption E(t) can be expressed as:

[0093]

[0094] The cumulative energy consumption can be calculated by the following formula:

[0095]

[0096] Substituting Equation (13) and Equation (16) into Equation (15), the transient DC bus voltage can be rewritten as:

[0097] U(t) = h(E(t), f(T(t), n(t), ΔE)) (18)

[0098] From Equation (18), T(t), n(t), ΔE, and the cumulative energy consumption E(t) are required as inputs for predicting the transient DC bus voltage. The desired function of the transient DC bus voltage curve can be expressed in standard form as:

[0099] U(t) = H(T(t), n(t), ΔE, E(t); β) + ε (19)

[0100] where β = (β1, β2, …, βm)' is the regression coefficient, and e is the random error.

[0101] Based on the DC bus current and voltage predicted by the model, the energy consumption prediction data of the fuel cell are obtained according to Equations (2) and (3).

[0102] 1. Establish a new energy vehicle power system model based on LSSVM and train the model using real fuel cell data; 2. Test to obtain the differential energy consumption ΔE, the transient motor speed n, the transient motor torque T, and the cumulative energy consumption E(t) of the current fuel cell;

[0103] 3. Taking the differential energy consumption, transient motor speed n, and transient motor torque T as model inputs to predict the transient DC bus current; taking the differential energy consumption ΔE, transient motor speed n, transient motor torque T, and cumulative energy consumption E(t) as model inputs to predict the transient DC bus voltage, and obtaining the predicted DC bus current and voltage values;

[0104] 4. Calculating the predicted fuel cell energy consumption using the predicted DC bus current and voltage values.

[0105] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the purpose and scope of the present technical solution, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A new energy power system modeling and hydrogen consumption prediction and analysis method, characterized in that: Including the following steps: S1: Establish a non-linear regression model for the power system of new energy vehicles based on the least squares support vector machine (LSSVM) algorithm model. The model includes a first sub-model for predicting the DC bus current and a second sub-model for predicting the DC bus voltage; S2: Collect time series data from the vehicle CAN bus, including differential energy consumption ΔE, transient motor speed n, transient motor torque T, and cumulative energy consumption E(t), construct a training data set and train the non-linear regression model; S3: Input the real-time data of the power system to be measured into the trained model to obtain the predicted DC bus current I(t) and DC bus voltage U(t); S4: Calculate the hydrogen consumption m of the fuel cell and the state of charge SOC(t) of the battery according to the predicted DC bus current I(t) and voltage U(t), in combination with the hydrogen consumption calculation formula. H2 ​ 2. The new energy power system modeling and hydrogen consumption prediction and analysis method according to claim 1, characterized in that: In S1, the input parameters of the first sub-model include differential energy consumption ΔE, transient motor speed n, and transient motor torque T; the input parameters of the second sub-model include differential energy consumption ΔE, transient motor speed n, transient motor torque T, and cumulative energy consumption E(t).

3. The new energy power system modeling and hydrogen consumption prediction and analysis method according to claim 1, characterized in that: The LSSVM algorithm model uses the radial basis function (RBF) as the kernel function to map the input data to a high-dimensional feature space to achieve non-linear regression.

4. The new energy power system modeling and hydrogen consumption prediction analysis method according to claim 1, characterized in that: In S4, the hydrogen consumption is calculated by the formula: Wherein, is the molar mass of hydrogen, and n cell is the number of cells in the fuel cell stack, I FC is the output current of the fuel cell, and F is the Faraday constant.

5. The new energy power system modeling and hydrogen consumption prediction analysis method according to claim 1, characterized in that: In S4, the calculation formula for the state of charge (SOC) of the battery is: where SOC initial is the initial state of charge, Q is the battery capacity, and I(t) is the transient DC bus current.

6. The new energy power system modeling and hydrogen consumption prediction analysis method according to claim 1, wherein: The training data of the non-linear regression model further includes the internal resistance R of the lithium battery BAT , the mapping relationship between the battery capacity Q, the open circuit voltage OCV(t) and the SOC(t).

7. The new energy power system modeling and hydrogen consumption prediction analysis method according to claim 1, characterized in that: In S3, the prediction of the DC bus current I(t) is further based on the following relationship: I(t) = f(T(t), n(t), ΔE) where T(t) is the transient motor torque, n(t) is the transient motor speed, and ΔE is the differential energy consumption.

8. The new energy power system modeling and hydrogen consumption prediction analysis method according to claim 1, characterized in that: In S3, the prediction of the DC bus voltage U(t) is further based on the following relationship: U(t) = h(E(t), I(t)) where E(t) is the cumulative energy consumption and I(t) is the transient DC bus current.

9. The new energy power system modeling and hydrogen consumption prediction analysis method according to claim 1, wherein: In S2, the time series data is obtained through a discretized lithium battery model. The discretized model includes: U(t) = OCV(t) - R BAT ·I(t) Among them, OCV(t) is the open-circuit voltage, R BAT is the internal resistance of the lithium battery, ΔSOC is the differential state of charge, and Δt is the time interval.

10. The new energy power system modeling and hydrogen consumption prediction and analysis method according to any one of claims 1 to 9, characterized in that: The hidden layer of the LSSVM algorithm model includes at least one first hidden layer containing 50 neurons and a second hidden layer containing 80 neurons to improve the prediction accuracy of non-linear regression.

Images