A multi-mode adaptive frequency decoupling power allocation method and system

Abstract

The present application discloses a power allocation method and system with multi-modal adaptive frequency decoupling, which extracts time domain information from a vehicle speed prediction training data set by selecting a eigenvector to characterize the current driving state; sets a first sensitivity factor and a second sensitivity factor according to the current vehicle system state; establishes an evaluation function for the comprehensive operating cost of a non-plug-in hydrogen fuel cell hybrid vehicle, extracts two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under three types of vehicle driving states as reference coefficients, and forms an adjustment coefficient reference matrix; uses "Haar" wavelet to perform three-level wavelet decoupling on the rate of change sequence of power demand to obtain a low-frequency sequence; fuzzifies the adjustment coefficient reference matrix according to the current vehicle driving state to adjust the adjustment coefficient in real time, adaptively adjusts the sensitivity factor, optimizes the low-frequency sequence, and determines the power allocation decision at the next moment. The present application can further reduce the solution complexity of the EMS.

Description

Technical Field

[0001] The present application relates to the technical field of hybrid energy management, and in particular to a power distribution method, system and terminal with multi-mode adaptive frequency decoupling. Background Art

[0002] In order to achieve the "dual carbon" goals and accelerate the decarbonization process of my country's automobile industry, it is necessary to vigorously promote the electrification transformation of heavy-duty truck power systems. Fuel cell heavy-duty trucks have the advantages of zero emissions, no pollution, high energy density, and strong environmental adaptability. They can meet the power needs of long-distance heavy-duty trucks under high-power and long-range conditions, and have the advantages of low carbon, environmental protection, cleanness and high efficiency. They are an important way to achieve my country's carbon neutrality goals in the transportation sector. However, in complex driving scenarios, the imbalance of power distribution control between various energy sources in the hybrid propulsion system will lead to energy waste and shortened vehicle life. Therefore, the study of energy management strategies to solve the power distribution problem between fuel cells and batteries is a key aspect.

[0003] By building an online simulation platform based on an RTU-BOX real-time digital controller and performing processor-in-the-loop (PIL) simulation, the proposed adaptive energy management strategy was verified for functionality and real-time applicability, which will help the strategy more closely match practical application scenarios. However, to successfully embed the strategy into the target hardware, it is necessary to consider the limited computing power of the hardware platform. This makes it difficult to successfully embed the energy management strategy based on quadratic programming (QP) into the target hardware.

[0004] The above analysis reveals the following problems and drawbacks of existing technologies: Currently, the QP-based solution requires multiple matrix transformations and complex calculations, significantly increasing the computational workload and making it difficult to meet the real-time and efficiency requirements of hardware platforms with limited computing power. This makes it impossible to implement the proposed energy management strategy in embedded hardware systems, making it impossible to verify the strategy's functionality and real-time applicability. Summary of the Invention

[0005] The main purpose of this application is to provide a power allocation method, system and terminal for multi-mode adaptive frequency decoupling, aiming to solve the technical problems of the power allocation method for multi-mode adaptive frequency decoupling.

[0006] To achieve the above objectives, the present application provides a multi-modal adaptive frequency decoupling power allocation method, comprising: in an offline phase, selecting a feature vector to extract time domain information from a vehicle speed prediction training dataset to characterize the current driving state, and using a fuzzy C clustering method to classify the driving state into three standard driving states: high-speed cruising, medium-speed driving, and low-speed creeping;

[0007] A first sensitivity factor and a second sensitivity factor are set according to the current vehicle operating state, power demand, fuel cell power, and lithium battery SoC state, wherein the first sensitivity factor is related to the lithium battery SoC state, and the second sensitivity factor is related to the fuel cell output power. An evaluation function for the comprehensive operating cost of non-plug-in hydrogen fuel cell hybrid vehicles is established, and two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under three standard driving conditions are extracted as reference coefficients to form an adjustment coefficient reference matrix. In the online link, the "Haar" wavelet is used to perform a three-level wavelet decoupling on the power demand change rate sequence, extract the third-order approximate coefficients, and reconstruct the third-order approximate coefficients to obtain a low-frequency sequence. The adjustment coefficient reference matrix is ​​fuzzy processed according to the driving state recognition results to obtain the adjustment coefficient corresponding to the current vehicle driving state. The two sensitivity factors are adaptively adjusted to optimize the low-frequency sequence, and the power allocation decision at the next moment is determined based on the optimized low-frequency sequence.

[0008] Optionally, the fuzzy C clustering method is used to divide the driving state into three standard driving states: high-speed cruising, medium-speed driving and low-speed creeping, including: calculating each cluster center in the time domain information and the membership value of each vehicle driving data point relative to all cluster centers by iteratively minimizing the objective function; assigning the current driving state to the corresponding cluster according to the membership value, and each vehicle driving data point is clustered into three standard driving states.

[0009] Optionally, the first sensitivity factor and the second sensitivity factor are set according to the current vehicle operating state, power demand, fuel cell power and lithium battery SoC state, including: using a tangent function in combination with the current vehicle lithium battery SoC state to obtain the first sensitivity factor; using a function related to fuel cell efficiency in combination with the current power point to dynamically generate the second sensitivity factor.

[0010] Optionally, establishing an evaluation function for the comprehensive operating cost of a non-plug-in hydrogen fuel cell hybrid electric vehicle includes establishing an evaluation function for the comprehensive operating cost of a non-plug-in hydrogen fuel cell hybrid electric vehicle based on the sum of hydrogen consumption cost, electricity cost, fuel cell system aging cost, and battery aging cost.

[0011] Optionally, the method extracts two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under the three standard driving conditions as reference coefficients to form an adjustment coefficient reference matrix, including: testing a non-plug-in hydrogen fuel cell hybrid vehicle under the three standard driving conditions to obtain test data, and using the interpolation method to test the data to generate a coefficient distribution diagram of the evaluation function; extracting the sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under the three types of vehicle driving conditions; and constructing an adjustment coefficient reference matrix based on the sensitivity factor adjustment coefficients under the three types of vehicle driving conditions.

[0012] Optionally, the "Haar" wavelet is used to perform a three-level wavelet decoupling on the power demand change rate sequence, extract the third-order approximate coefficients, and reconstruct the third-order approximate coefficients to obtain a low-frequency sequence, including: decomposing the original signal into approximate coefficients and detail coefficients through low-pass and high-pass filters; passing the approximate coefficients and detail coefficients through filters to obtain approximate coefficients of each order, and obtaining a low-frequency sequence based on the third-order approximate coefficients in the approximate coefficients of each order.

[0013] Optionally, the adjustment coefficient of the sensitivity factor includes a first adjustment coefficient and a second adjustment coefficient, and the adjustment coefficient reference matrix is ​​fuzzy processed according to the driving state recognition result to obtain the adjustment coefficient corresponding to the current vehicle driving state, and the two sensitivity factors are adaptively adjusted to optimize the low-frequency sequence, and the power allocation decision at the next moment is determined according to the optimized low-frequency sequence, including: if the value of the low-frequency sequence is greater than or equal to 0, then according to the comparison result of the current vehicle lithium battery SoC state and the SoC reference state and the first adjustment coefficient, the first sensitivity factor value is controlled by a first tangent function related to the vehicle lithium battery SoC state; according to the current vehicle fuel cell power and the second adjustment coefficient, the second sensitivity factor value is controlled by a corresponding second function related to the fuel cell efficiency; if the value of the low-frequency sequence is less than 0, then according to the comparison result of the current vehicle lithium battery SoC state and the SoC reference state and the first adjustment coefficient, the first sensitivity factor value is controlled by a second tangent function related to the vehicle lithium battery SoC state; according to the current vehicle fuel cell power and the second adjustment coefficient, the second sensitivity factor value is controlled by a second function related to the fuel cell efficiency.

[0014] Optionally, the process of establishing the objective function includes: establishing the objective function based on the weighted normalized lithium battery SoC, the normalized fuel cell output power and the normalized fuel cell output power transient; wherein, the first constraint condition of the pre-constructed objective function is used to constrain the lithium battery SoC of the non-plug-in hydrogen fuel cell hybrid vehicle to be near a reference value; the second constraint condition of the pre-constructed objective function is used to constrain the fuel cell output power to be near the reference power; and the third constraint condition of the pre-constructed objective function is used to constrain the fuel cell output power transient to be within a preset range.

[0015] In addition, to achieve the above-mentioned purpose, the present application also provides a multi-modal adaptive frequency decoupling power distribution system, including: a clustering module, which is used to select feature vectors to extract time domain information in the vehicle speed prediction training data set in the offline link to characterize the current driving state, and adopts the fuzzy C clustering method to divide the driving state into three standard driving states: high-speed cruising, medium-speed driving and low-speed creeping; a sensitivity factor calculation module, which is used to set the first sensitivity factor and the second sensitivity factor according to the current vehicle operating state, power demand, fuel cell power and lithium battery SoC state, wherein the first sensitivity factor is related to the lithium battery SoC state, and the second sensitivity factor is related to the fuel cell output power; a coefficient matrix calculation module, which is used to establish a non-plug-in hydrogen fuel An evaluation function for the comprehensive operating cost of battery hybrid vehicles extracts two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under three standard driving conditions as reference coefficients to form an adjustment coefficient reference matrix; a low-frequency sequence reconstruction module is used in the online link to perform three-level wavelet decoupling on the power demand change rate sequence using the "Haar" wavelet, extract the third-order approximate coefficients, and reconstruct the third-order approximate coefficients to obtain a low-frequency sequence; a power allocation module is used to fuzzy the adjustment coefficient reference matrix based on the driving state recognition results to obtain the adjustment coefficient corresponding to the current vehicle driving state, adaptively adjust the two sensitivity factors to optimize the low-frequency sequence, and determine the power allocation decision at the next moment based on the optimized low-frequency sequence.

[0016] To achieve the above objectives, the present application also provides a vehicle power system energy management terminal, which includes a multi-modal adaptive frequency decoupling power distribution system.

[0017] The embodiment of the present application proposes a multi-modal adaptive frequency decoupling power allocation method, system and terminal, which selects a feature vector to extract time domain information in a vehicle speed prediction training data set in an offline link to characterize the current driving state, and adopts a fuzzy C clustering method to divide the driving state into three standard driving states: high-speed cruising, medium-speed driving and low-speed creeping; according to the current vehicle operating state, power demand, fuel cell power and lithium battery SoC state, a first sensitivity factor and a second sensitivity factor are set, wherein the first sensitivity factor is related to the lithium battery SoC state, and the second sensitivity factor is related to the fuel cell output power; an evaluation function for the comprehensive operating cost of a non-plug-in hydrogen fuel cell hybrid vehicle is established, and two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under the three types of vehicle driving states are extracted as reference coefficients , forming a reference matrix of adjustment coefficients; in the online link, the "Haar" wavelet is used to perform three-level wavelet decoupling on the rate of change sequence of power demand, extract the third-order approximate coefficients, and reconstruct the third-order approximate coefficients to obtain a low-frequency sequence. The rate of change of power demand is selected as the decoupling object through the frequency decoupling method, which is conducive to smoothing the power output of the fuel cell and reducing the performance degradation of the fuel cell; according to the vehicle driving state recognition result, the adjustment coefficient reference matrix is ​​fuzzy processed to obtain the adjustment coefficient corresponding to the current vehicle driving state, and the two sensitivity factors are adaptively adjusted to optimize the low-frequency sequence. The power allocation decision at the next moment is determined according to the optimized low-frequency sequence. This application solves the power allocation method of multi-modal adaptive frequency decoupling, simplifies the calculation process of power allocation, reduces the solution complexity of the energy management strategy (Energy Management Strategy, EMS), and maintains a high optimization accuracy. At the same time, compared with other wavelets, the "Haar" wavelet only requires simple addition and subtraction operations in the process of multi-resolution decomposition and reconstruction of data, which simplifies the composition of the program and further improves the execution efficiency of the code. The adaptive energy management strategy proposed in this application is a solution method that can run smoothly in the embedded hardware system. By further reducing the solution complexity of EMS, the control algorithm is downloaded to the constructed processor-in-the-loop simulation PIL platform to verify the functionality and real-time applicability of the proposed energy management strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flow chart of a multi-mode adaptive frequency decoupling method provided by an embodiment of the present invention;

[0019] Figure 2 is a signal processing flow chart of a multi-mode adaptive frequency decoupling method provided by an embodiment of the present invention;

[0020] Figure 3 This is a diagram of the FCM workflow provided by an embodiment of the present invention;

[0021] Figure 4 This is a driving state classification result diagram provided by an embodiment of the present invention;

[0022] Figure 5 This is an example of standard driving state classification provided by an embodiment of the present invention;

[0023] Figure 6 is a distribution diagram of adjustment coefficients a and b under various standard driving conditions provided by an embodiment of the present invention;

[0024] Figure 7 is the relationship between the sensitivity factors K1, K2 and the current power source state provided by the embodiment of the present invention;

[0025] Figure 8 This is a comparison of the real-time power allocation results obtained by QP solution under driving cycle 1 provided by an embodiment of the present invention;

[0026] Figure 9 This is a 1000s power distribution curve for PIL simulation provided by an embodiment of the present invention.

[0027] The purpose, features and advantages of the present invention will be further described with reference to the accompanying drawings and in conjunction with the embodiments. DETAILED DESCRIPTION

[0028] It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0029] One object of the invention is to provide a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs multi-mode adaptive frequency decoupled power allocation.

[0030] Another object of the present application is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute a multi-mode adaptive frequency decoupling power allocation method.

[0031] Another object of the present application is to provide an information data processing terminal, which is used to implement a power allocation method for multi-mode adaptive frequency decoupling.

[0032] Another object of the present application is to provide a power distribution system based on multi-modal adaptive frequency decoupling, the multi-modal adaptive frequency decoupling power distribution system includes: a clustering module for selecting feature vectors to extract time domain information in the vehicle speed prediction training data set in an offline link to characterize the current driving state, and adopting a fuzzy C clustering method to divide the driving state into three standard driving states: high-speed cruising, medium-speed driving and low-speed creeping; a sensitivity factor calculation module for setting a first sensitivity factor and a second sensitivity factor according to the current vehicle operating state, power demand, fuel cell power and lithium battery SoC state, wherein the first sensitivity factor is related to the lithium battery SoC state, and the second sensitivity factor is related to the fuel cell output power; a coefficient matrix calculation module for establishing An evaluation function for the comprehensive operating cost of non-plug-in hydrogen fuel cell hybrid vehicles is developed. Two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under three types of vehicle driving conditions are extracted as reference coefficients to form an adjustment coefficient reference matrix. A low-frequency sequence reconstruction module is used in the online link. The "Haar" wavelet is used to perform three-level wavelet decoupling on the power demand change rate sequence, extract the third-order approximate coefficients, and reconstruct the third-order approximate coefficients to obtain a low-frequency sequence. A power allocation module is used to fuzzy the adjustment coefficient reference matrix based on the vehicle driving state identification results to obtain the adjustment coefficient corresponding to the current vehicle driving state, adaptively adjust the two sensitivity factors to optimize the low-frequency sequence, and determine the power allocation decision at the next moment based on the optimized low-frequency sequence.

[0033] To achieve the above objectives, the present application also provides a vehicle power system energy management terminal, which includes a multi-modal adaptive frequency decoupling power distribution system.

[0034] Another object of the present application is to provide a fuel cell hybrid electric vehicle, which includes a vehicle power system energy management terminal.

[0035] In response to the problems existing in the prior art, the present application provides a multi-mode adaptive frequency decoupling power allocation method, system and terminal.

[0036] This application is implemented as follows: a multi-modal adaptive frequency decoupling power allocation method. The multi-modal adaptive frequency decoupling power allocation method solves the established model predictive control-based energy management optimization problem by introducing the "Haar" wavelet transform and adaptive optimization of sensitivity factors based on driving state identification. First, the "Haar" wavelet is used to perform wavelet decoupling on the rate of change of power demand to extract its low-frequency sequence; then, FCM technology is used to analyze the driving condition data and identify the driving state; finally, the sensitivity factor and corresponding adjustment coefficient are set according to the state of each power source, and the low-frequency sequence is optimized and adjusted to obtain the output power of the fuel cell system.

[0037] Specifically, refer to Figure 1 , Figure 1 This is a schematic diagram of a power allocation method for multi-modal adaptive frequency decoupling provided in the first embodiment of the present application. The method can be executed by a processor of a terminal or a server. The power allocation method for multi-modal adaptive frequency decoupling may include:

[0038] S10. In the offline phase, feature vectors are selected to extract time domain information from the vehicle speed prediction training dataset to characterize the current driving state. The fuzzy C clustering method is used to classify the driving state into three standard driving states: high-speed cruising, medium-speed driving, and low-speed creeping.

[0039] In one embodiment of the present application, using the fuzzy C clustering method to classify the driving state into three standard driving states: high-speed cruising, medium-speed driving, and low-speed creeping may include the following execution process:

[0040] The cluster centers in the time domain information and the membership values ​​of each vehicle driving data point relative to all cluster centers are calculated by iteratively minimizing the objective function.

[0041] The current driving state is assigned to the corresponding cluster according to the membership value, and each vehicle driving data point is clustered into three standard driving states.

[0042] For example, refer to Figure 3 The processor selects appropriate feature vectors to extract the time domain information in the vehicle speed prediction training data set at each moment to characterize the current driving state, and uses the fuzzy C clustering method to divide the driving state into three categories. These three driving states are represented as follows: , They correspond to three standard driving states: high-speed cruising, medium-speed driving, and low-speed creeping. Compared with the single-mode baseline strategy, setting a multi-mode strategy can improve control and prediction accuracy. Therefore, the processor selects the appropriate feature vector to represent the driving state according to the prediction accuracy in the offline stage and classifies the driving cycle through the clustering method. The specific workflow is as follows: Figure 3 In this example, select As the characteristic variable, represents the average speed, represents the standard deviation of acceleration, Indicates the average acceleration, fuzzy C The FCM clustering method divides the driving cycle data into three categories. The specific classification results are as follows: Figure 4 As shown here, the characteristics of the three standard driving conditions are more clearly shown. Figure 5 An example of the classification effect for a 1000-second period of standard driving conditions is shown.

[0043] Specifically, the processor can select the feature vector Among them, it means Normalized average speed, represents the normalized velocity standard deviation, represents the normalized average acceleration, and the fuzzy C clustering method iteratively minimizes the objective function . Calculate the cluster centers And the membership of each vehicle driving data point relative to all cluster centers . It can reflect the fuzzy characteristics of the driving state. The sum of the membership of each data point relative to all cluster centers is 1. Finally, the processor can assign the current vehicle driving state to the appropriate cluster according to the size of the membership value of each data point to obtain the final clustering result. The driving feature vector samples are clustered into three standard driving states according to their maximum membership. Among them, the minimization objective function The calculation formula is as follows:

[0044]

[0045] in, m is the fuzzy index, that is, the number of clusters in the classification process, N is the total number of samples, C is the total number of clusters, Indicates the i data points, For the j The center of the cluster.

[0046] S20. Set a first sensitivity factor and a second sensitivity factor according to the current vehicle operating state, power demand, fuel cell power and lithium battery SoC state, wherein the first sensitivity factor is related to the lithium battery SoC state, and the second sensitivity factor is related to the fuel cell output power.

[0047] In one embodiment of the present application, setting the first sensitivity factor and the second sensitivity factor according to the current vehicle operating state, power demand, fuel cell power, and lithium battery SoC state may include the following execution process:

[0048] The first sensitivity factor is obtained by using the tangent function and combining it with the current vehicle lithium battery SoC state.

[0049] A second sensitivity factor is dynamically generated using a function related to fuel cell efficiency and combined with the current power point.

[0050] For example, according to the current vehicle operating state, power demand , fuel cell power Two sensitivity factors are set for the lithium battery SoC state, and the two sensitivity factors are expressed as: a first sensitivity factor K1 and a second sensitivity factor K2, wherein the first sensitivity factor K1 is related to the lithium battery SoC state, and the second sensitivity factor K2 is related to the fuel cell output power.

[0051] S30. Establish an evaluation function for the comprehensive operating cost of non-plug-in hydrogen fuel cell hybrid vehicles, extract two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under three types of vehicle driving conditions as reference coefficients, and form an adjustment coefficient reference matrix.

[0052] In one embodiment of the present application, the construction of the objective function may include the following execution process:

[0053] An objective function is established based on the weighted normalized lithium battery SoC, the normalized fuel cell output power, and the normalized fuel cell output power transient.

[0054] The first constraint condition of the pre-built objective function is used to constrain the lithium battery SoC of the non-plug-in hydrogen fuel cell hybrid vehicle to be near a reference value.

[0055] The second constraint of the pre-built objective function is used to constrain the fuel cell output power to be near the reference power.

[0056] The third constraint condition of the pre-built objective function is used to constrain the transient output power of the fuel cell to be within a preset range.

[0057] For example, the expressions of the objective function and constraints in the energy management strategy framework applied to non-plug-in hydrogen fuel cell hybrid vehicles are shown in Equations (1) and (2). Among them, Equation (1) is the objective function in the energy management strategy, and items (a), (b), and (c) are the results of normalizing the variables according to Equation (2). Item (a) ensures that the battery state of charge (SoC) of the non-plug-in hydrogen fuel cell hybrid vehicle remains near the reference value. Item (b) controls the fuel cell output power of the non-plug-in hydrogen fuel cell hybrid vehicle to be near the reference power to improve the working efficiency of the fuel cell system and reduce hydrogen consumption. Item (c) suppresses the aging of the fuel cell system by controlling the transient output power of the fuel cell of the non-plug-in hydrogen fuel cell hybrid vehicle.

[0058]

[0059]

[0060]

[0061] Where, k For the current moment, Hp is the future time series number, i The value range is 0 to H p -1. For k+ H p The battery's state of charge at all times, normalized lithium battery SoC . It is the normalized result of formula (a). For k+i Fuel cell output power at the moment, normalized fuel cell output power is the normalized result of formula (b). For k+i The fuel cell output power transient at the moment, the normalized fuel cell output power transient is is the normalized result of formula (c). is the weight factor, are the reference values ​​of fuel cell output power and lithium battery SoC status, They are the upper and lower limits of the lithium battery SoC state, fuel cell output power, and fuel cell output power transient.

[0062] In one embodiment of the present application, the process of establishing the evaluation function may include the following execution process:

[0063] An evaluation function for the comprehensive operating cost of non-plug-in hydrogen fuel cell hybrid vehicles is based on the sum of hydrogen consumption cost, equivalent hydrogen consumption cost, fuel cell system aging cost, and battery aging cost.

[0064] For example, the evaluation function is shown in formula (4):

[0065]

[0066]

[0067] in, The cost of hydrogen consumption, Indicates the conversion of lithium battery energy consumption into equivalent hydrogen consumption, Represents the fuel cell system aging cost, represent k The aging degree of the fuel cell system at all times, Represents the battery aging cost, Represents hydrogen price, Representative fuel cell price, Represents the battery price. For hydrogen low calorific value, for kFuel cell output power at all times, for k Fuel cell system efficiency at all times, Representative k The output power of the fuel cell system at the moment is The efficiency of the fuel cell system. For battery capacity, is the battery voltage. SoH For battery health status, SoH (0), SoH ( N ) are the initial health state of the battery and the final health state of the battery, respectively.

[0068] In one embodiment of the present application, extracting two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under three types of vehicle driving conditions as reference coefficients, and forming an adjustment coefficient reference matrix may include the following execution process:

[0069] A non-plug-in hydrogen fuel cell hybrid vehicle is tested cyclically under three types of vehicle driving conditions, and the test data are interpolated to generate a coefficient distribution diagram of the evaluation function.

[0070] The sensitivity factor adjustment coefficient corresponding to the minimum value of the evaluation function under the three types of vehicle driving conditions is extracted.

[0071] An adjustment coefficient reference matrix is ​​constructed based on the sensitivity factor adjustment coefficients under three types of vehicle driving conditions.

[0072] Specifically, the process of extracting the sensitivity factor adjustment coefficient reference matrix by the processor may include:

[0073] The processor selects a set of discrete candidate values ​​for the weight factor adjustment coefficients , test under three standard driving cycles and record the corresponding evaluation function Finally, the coefficient distribution diagrams under three standard driving cycles are drawn by combining the interpolation method, and the evaluation function is extracted under each standard driving state. Sensitivity factor adjustment coefficient corresponding to the minimum value and , obtain the reference matrix of the adjustment coefficient.

[0074] In S40, in the online link, the “Haar” wavelet is used to perform three-level wavelet decoupling on the power demand change rate sequence, extract the third-order approximate coefficients, and reconstruct the third-order approximate coefficients to obtain a low-frequency sequence.

[0075] The processor decouples the rate of change of power demand through the three-level "Haar" wavelet. The reason for choosing "Haar" is that it has the advantages of simple structure and high time resolution. The decoupling process can be expressed as Equation (8), while the discrete wavelet transform is used to decompose the discrete signal into different resolution levels. N The decomposition and reconstruction of the power change rate requirement are shown in Equations (9) and (10) respectively.

[0076]

[0077]

[0078]

[0079] in, Represents the coefficients of the wavelet basis function at the corresponding position after signal decomposition, Representative k The original sequence signal of seconds, j Indicates the level of decomposition, corresponding to different frequency bandwidths, n Determines the position index of the wavelet basis function, corresponding to the time distribution of the signal at the current level, Decoupling the rate of change of power demand, Representative k The original sequence signal.

[0080] Therefore, at the current moment k The sequence for wavelet transform is: 、 By using the “Haar” wavelet-based n Channel filter bank, original signal x ( t ) through a low-pass filter H0 ( z ) and high-pass filter H1 ( z ) is decomposed into approximate coefficients and detail coefficients. Then it passes through the filter Gk(z), ( k =0, 1, ...) to reconstruct the sequence. Since only the low-frequency part is needed, the detail coefficient [ ], using only the third-order approximation coefficients Reconstruct the low-frequency sequence [ ].

[0081] S50. Fuzzy processing is performed on the adjustment coefficient reference matrix based on the driving state recognition result. The two sensitivity factors are adaptively adjusted through the sensitivity factor adjustment coefficient corresponding to the current vehicle driving state to optimize the low-frequency sequence. The power allocation decision at the next moment is determined based on the optimized low-frequency sequence.

[0082] The power allocation decision is expressed as follows:

[0083]

[0084] Where, y ( k+ 1) is the optimized low-frequency sequence, for k- Fuel cell output power at moment 1.

[0085] Then the processor performs fuzzy processing on the reference matrix of the adjustment coefficient according to the driving state recognition result. The fuzzy processing process of the adjustment coefficient reference matrix by the processor during the online application process is shown in formula (10):

[0086]

[0087] in, Indicates the i The membership relationship of the standard driving state, and Indicates the Nth i Under standard driving conditions, the evaluation function The sensitivity factor adjustment coefficient corresponding to the minimum value.

[0088] In one embodiment of the present application, the sensitivity factor adjustment coefficient includes a first adjustment coefficient and a second adjustment coefficient. The two sensitivity factors are adaptively adjusted according to the sensitivity factor adjustment coefficient corresponding to the current vehicle driving state to optimize the low-frequency sequence. Determining the power allocation decision at the next moment according to the optimized low-frequency sequence may include the following execution process:

[0089] S501: If the value of the low-frequency sequence is greater than or equal to 0, then, based on a comparison result between the current vehicle lithium battery SoC state and the SoC reference state and a first adjustment coefficient, output a value for controlling a first sensitivity factor using a first tangent function related to the vehicle lithium battery SoC state; and based on the current vehicle fuel cell power and a second adjustment coefficient, output a value for controlling a second sensitivity factor using a corresponding second function related to the fuel cell efficiency.

[0090] S502. If the value of the low-frequency sequence is less than 0, then based on the comparison result between the current vehicle lithium battery SoC state and the SoC reference state and the first adjustment coefficient, the output uses a second tangent function related to the vehicle lithium battery SoC state to control the value of the first sensitivity factor; based on the current vehicle fuel cell power and the second adjustment coefficient, the output uses a second function related to the corresponding fuel cell efficiency to control the value of the second sensitivity factor.

[0091] For example, the first sensitivity factor K1 and the second sensitivity factor K2 related to the current state of each power source of the vehicle. Referring to the above, the first sensitivity factor K1 is related to the state of the lithium battery SoC. When , the first tangent function related to the vehicle lithium battery SoC state as shown in formula (12a) is used for control. The second sensitivity factor K2 is related to the fuel cell output power and is controlled by the first function related to the fuel cell efficiency as shown in formula (13a). The interp() interpolation function is used for cubic spline interpolation of dy1, and the current power point is combined with the current power point. , dynamically generate the second sensitivity factor. When , the second tangent function related to the vehicle lithium battery SoC state, as shown in formula (12b), is used for control. The second sensitivity factor K2 is related to the fuel cell output power and is controlled by the second function related to the fuel cell efficiency, as shown in formula (13b).

[0092] when hour:

[0093] (12a)

[0094] (13a)

[0095] when hour ：

[0096] (12b)

[0097] (13b)

[0098] According to the current low-frequency sequence Is it greater than or equal to 0 and the vehicle's lithium battery Status and The solution function of the first sensitivity factor K1 is determined by comparing the reference state results. Whether it is greater than or equal to 0 and the vehicle fuel cell power determines the solution function of the second sensitivity factor K2, Express dy 1Using cubic spline interpolation, in the above formula, a andb Respectively a (k) and b (k).

[0099] Sensitivity factors K1, K2 and current lithium batteries and k -1 moment fuel cell output power, that is When the low frequency sequence of the power demand change rate When it is greater than 0, the power demand will increase at the next moment. The closer , the next moment the fuel cell output power rises The bigger, The rate of change decreases, On the contrary, when it is less than 0, the above adjustments will ultimately ensure that the lithium battery SoC is as stable as possible near the reference value. When greater than 0, The larger the value, the greater the value, the better the result will be. The smaller it is, the closer it is to the reference value of fuel cell output power. ,but The rate of change decreases. When it is less than 0, on the contrary, the above adjustments will ultimately ensure that the fuel cell output is as stable as possible at the highest efficiency point. nearby.

[0100] In another embodiment of the present application, the specific steps of the multi-mode adaptive frequency decoupling power allocation method are as follows:

[0101] Step 1: Compared with the single-mode baseline strategy, setting a multi-mode strategy can improve control and prediction accuracy. Therefore, according to the prediction accuracy, the appropriate feature vector is selected in the offline stage to represent the driving state and the driving cycle is classified by clustering method. The specific workflow is as follows: Figure 3 In this example, select As the characteristic variable, the fuzzy C clustering (FCM) method divides the driving cycle data into three categories. The specific classification results are as follows: Figure 4 As shown here, the characteristics of the three standard driving conditions are more clearly shown. Figure 5 An example of the classification effect for a 1000-second period of standard driving conditions is shown.

[0102] Step 2: Sensitivity factor setting: according to the current vehicle operating status and power demand , fuel cell power Two sensitivity factors are set for the lithium battery SoC state, and the sensitivity factors are expressed as: K1, K2, where K1 is related to the lithium battery SoC state, and K2 is related to the fuel cell output power.

[0103] Step 3: Establish an evaluation function based on the requirements. In this example, in order to improve the economy and durability of the hydrogen fuel cell hybrid system, the hydrogen consumption of the fuel cell, the battery consumption, and the degradation of the fuel cell and battery are considered. The evaluation function in this example is shown in formula (4):

[0104]

[0105] in, The cost of hydrogen consumption, Indicates converting the energy consumption of lithium batteries into equivalent hydrogen consumption, Represents the fuel cell system aging cost, Represents the battery aging cost.

[0106] In this example, we extract the evaluation function under each standard driving state. Adjustment coefficient of the sensitivity factor corresponding to the minimum value and , the specific adjustment coefficient a 、 b The three-dimensional weight diagram of the relationship between the evaluation function is as follows Figure 6 As shown, the adjustment coefficient reference matrix is ​​obtained, as shown in formula (5):

[0107]

[0108] Step 4: Decouple the rate of change of power demand by the three-level “Haar” wavelet. k The sequence of wavelet transform at each moment is: 、 , the prediction time domain of this example is By using an n-channel filter bank based on the "Haar" wavelet, the original signal x ( t ) through a low-pass filter H0 ( z ) and high-pass filter H1 ( z ) is decomposed into approximate coefficients and detail coefficients. Then through the filter Gk ( z ), ( k = 0, 1, ...) to reconstruct the sequence. Since only the low-frequency part is needed, the detail coefficient [ ], using only the third-order approximation coefficients Reconstruct the low-frequency sequence [ ].

[0109] Step 5: Fuzzy process the adjustment coefficient reference matrix according to the driving state recognition result, and adaptively adjust the sensitivity factor through the adjustment coefficient corresponding to the current vehicle driving state. In this example, the relationship between the sensitivity factors K1, K2 and the current power source state is as follows: Figure 7 shown.

[0110] like Figure 8 As shown, under the driving cycle 1 of the non-plug-in hydrogen fuel cell hybrid electric vehicle obtained from ADVISOR (Advanced Vehicle Simulator OR), the comparison of the real-time power distribution results using the multi-mode adaptive frequency decoupling method proposed in this application and the QP solver shows that, although the solution methods are different, the power distribution results using the frequency decoupling method and the distribution results using the QP solver are basically consistent in terms of overall trends and key nodes, especially in terms of the change trends of the fuel cell output power and the lithium battery SoC. This shows that the energy management strategy based on the frequency decoupling method can stably achieve the power distribution optimization goal under different driving conditions and dynamic environments, and the deviation between the results and the QP solver solution is controllable. The multi-mode adaptive frequency solution method proposed in this application can simplify the power distribution calculation process, reduce the solution complexity of the EMS, and ensure that the energy management strategy proposed in this application can run in the embedded hardware system. In driving cycle 1, the adaptive energy management strategy improved by the power allocation method proposed in this application is run in the embedded hardware system using the processor-in-the-loop simulation (PIL) technology. The power allocation results for 1000s recorded by the record module in the RTU-BOX host computer interface Rtunit Studio are as follows: Figure 9 shown.

[0111] Table 1 Comparison of optimization effects of this application example and other EMS under driving cycle 1

[0112]

[0113] In the driving cycle 1 of a non-plug-in hydrogen fuel cell hybrid vehicle obtained from ADVISOR (Advanced Vehicle Simulator OR), the frequency decoupling method proposed in this application is combined with the energy management strategy based on the wavelet neural network solved by the QP solver ( EMS2 Multi-WNNs ), traditional energy management strategy based on model predictive control ( EMS1 BPNN 、 EMS1 LSTM 、 EMS1 FCM_MC ) and fuzzy rule-based energy management strategies ( EMS FLC ). This paper compares the energy efficiency and lifespan optimization results of a fuel cell hybrid heavy-duty truck using a QP solver. While this method performs slightly worse than the adaptive strategy based on a QP solver, it significantly outperforms several other traditional baselines based on model predictive control (MPC) and energy management strategies. This reduces total operation and maintenance costs and significantly improves the economy and durability of hydrogen fuel cell hybrid heavy-duty trucks. A detailed comparison of these components is shown in Table 1.

[0114] in, EMS1 BPNN 、 EMS1 LSTM 、 EMS1 FCM_MC They represent the energy management strategies based on traditional MPC using three traditional speed prediction methods: back propagation neural network (BPNN), long short-term memory network (LSTM), and fuzzy C clustering Markov prediction (FCM-MC).

[0115] On the basis of the above embodiments, the present application also provides a multi-modal adaptive frequency decoupling power distribution system, including: a clustering module for selecting feature vectors to extract time domain information in the vehicle speed prediction training data set in the offline link to characterize the current driving state, and using the fuzzy C clustering method to divide the driving state into three standard driving states: high-speed cruising, medium-speed driving and low-speed creeping; a sensitivity factor calculation module for setting a first sensitivity factor and a second sensitivity factor according to the current vehicle operating state, power demand, fuel cell power and lithium battery SoC state, wherein the first sensitivity factor is related to the lithium battery SoC state, and the second sensitivity factor is related to the fuel cell output power; a coefficient matrix calculation module for establishing a non-plug-in hydrogen fuel cell An evaluation function for the comprehensive operating cost of a hybrid electric vehicle with a power supply is developed. Two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under three standard driving conditions are extracted as reference coefficients to form an adjustment coefficient reference matrix. A low-frequency sequence reconstruction module is used in the online link. The "Haar" wavelet is used to perform three-level wavelet decoupling on the power demand change rate sequence, extract the third-order approximate coefficients, and reconstruct the third-order approximate coefficients to obtain a low-frequency sequence. The power allocation module is used to fuzzy the adjustment coefficient reference matrix based on the driving state recognition results to obtain the adjustment coefficient corresponding to the current vehicle driving state. The two sensitivity factors are adaptively adjusted to optimize the low-frequency sequence, and the power allocation decision at the next moment is determined based on the optimized low-frequency sequence.

[0116] It is not difficult to find that this embodiment is a system embodiment corresponding to the above-mentioned method embodiment, and this embodiment can be implemented in conjunction with the above-mentioned method embodiment. The relevant technical details and technical effects mentioned in the above-mentioned embodiments are still valid in this embodiment, and to reduce repetition, they are not repeated here. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above-mentioned embodiments.

[0117] It is worth mentioning that all modules involved in this embodiment are logical modules. In actual applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, to highlight the innovation of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed by this application. However, this does not mean that other units do not exist in this embodiment.

[0118] Based on the above embodiment, a vehicle power system energy management terminal is provided, which includes the aforementioned multi-modal adaptive frequency decoupled power distribution system.

[0119] The above are only preferred embodiments of the present application and do not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A multi-mode adaptive frequency decoupling power allocation method, characterized in that: include: In the offline phase, feature vectors are selected to extract time domain information from the vehicle speed prediction training dataset to characterize the current driving state. The fuzzy C clustering method is used to classify the driving state into three standard driving states: high-speed cruising, medium-speed driving, and low-speed creeping. Setting a first sensitivity factor and a second sensitivity factor based on the current vehicle operating state, power demand, fuel cell power, and lithium battery SoC state, wherein the first sensitivity factor is related to the lithium battery SoC state, and the second sensitivity factor is related to the fuel cell output power; An evaluation function for the comprehensive operating cost of non-plug-in hydrogen fuel cell hybrid vehicles was established. Two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under three standard driving conditions were extracted as reference coefficients to form an adjustment coefficient reference matrix. In the online phase, the "Haar" wavelet is used to perform three-level wavelet decoupling on the power demand change rate sequence, extract the third-order approximate coefficients, and reconstruct the third-order approximate coefficients to obtain a low-frequency sequence; According to the driving state recognition results, the adjustment coefficient reference matrix is ​​fuzzy processed to obtain the adjustment coefficient corresponding to the current vehicle driving state. The two sensitivity factors are adaptively adjusted to optimize the low-frequency sequence. The power allocation decision at the next moment is determined based on the optimized low-frequency sequence.

2. The multi-mode adaptive frequency decoupling power allocation method according to claim 1, wherein: The fuzzy C clustering method is used to classify the driving state into three standard driving states: high-speed cruising, medium-speed driving and low-speed creeping, including: By iteratively minimizing the objective function, each cluster center in the time domain information and the membership value of each vehicle driving data point relative to all cluster centers are calculated; The current driving state is assigned to the corresponding cluster according to the membership value, and each vehicle driving data point is clustered into three standard driving states.

3. The multi-mode adaptive frequency decoupling power allocation method according to claim 1, wherein: The first sensitivity factor and the second sensitivity factor are set according to the current vehicle operating state, power demand, fuel cell power, and lithium battery SoC state, including: The first sensitivity factor is obtained by using the tangent function and combining it with the current vehicle lithium battery SoC state; A second sensitivity factor is dynamically generated using a function related to fuel cell efficiency and combined with the current power point.

4. The multi-mode adaptive frequency decoupling power allocation method according to claim 1, wherein: Establish an evaluation function for the comprehensive operating cost of non-plug-in hydrogen fuel cell hybrid vehicles, including: An evaluation function for the comprehensive operating cost of non-plug-in hydrogen fuel cell hybrid vehicles is established based on the sum of hydrogen consumption cost, electricity cost, fuel cell system aging cost and battery aging cost.

5. The multi-mode adaptive frequency decoupling power allocation method according to claim 1, wherein: The two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under the three standard driving conditions are extracted as reference coefficients to form an adjustment coefficient reference matrix, including: A non-plug-in hydrogen fuel cell hybrid vehicle is tested under three standard driving conditions to obtain test data, and the test data is interpolated to generate a coefficient distribution diagram of the evaluation function; Extract the sensitivity factor adjustment coefficient corresponding to the minimum value of the evaluation function under the three types of vehicle driving conditions; An adjustment coefficient reference matrix is ​​constructed based on the sensitivity factor adjustment coefficients under three types of vehicle driving conditions.

6. The multi-mode adaptive frequency decoupling power allocation method according to claim 1, wherein: The method uses the "Haar" wavelet to perform three-level wavelet decoupling on the power demand change rate sequence, extracts the third-order approximate coefficients, and reconstructs the third-order approximate coefficients to obtain a low-frequency sequence, including: Decompose the original signal into approximate coefficients and detail coefficients through low-pass and high-pass filters; The approximate coefficients and detail coefficients are passed through filters to obtain approximate coefficients of each order, and a low-frequency sequence is obtained according to the third-order approximate coefficients of each order approximate coefficient.

7. The multi-mode adaptive frequency decoupling power allocation method according to claim 1, wherein: The adjustment coefficient of the sensitivity factor includes a first adjustment coefficient and a second adjustment coefficient. The adjustment coefficient reference matrix is ​​fuzzy processed according to the driving state recognition result to obtain the adjustment coefficient corresponding to the current vehicle driving state. The two sensitivity factors are adaptively adjusted to optimize the low-frequency sequence. The power allocation decision at the next moment is determined according to the optimized low-frequency sequence, including: If the value of the low-frequency sequence is greater than or equal to 0, then based on the comparison result of the current vehicle lithium battery SoC state and the SoC reference state and the first adjustment coefficient, output the value of the first sensitivity factor controlled by the corresponding first tangent function related to the vehicle lithium battery SoC state; Outputting a value of a second sensitivity factor controlled by a corresponding second function related to fuel cell efficiency according to the current vehicle fuel cell power and the second adjustment coefficient; If the value of the low-frequency sequence is less than 0, then based on the comparison result of the current vehicle lithium battery SoC state and the SoC reference state and the first adjustment coefficient, output the value of the first sensitivity factor controlled by the corresponding second tangent function related to the vehicle lithium battery SoC state; According to the current vehicle fuel cell power and the second adjustment coefficient, a second function related to the corresponding fuel cell efficiency is output to control the value of the second sensitivity factor.

8. The multi-mode adaptive frequency decoupling power allocation method according to claim 2, wherein: The process of establishing the objective function includes: Establishing an objective function based on the weighted normalized lithium battery SoC, the normalized fuel cell output power, and the normalized fuel cell output power transient; The first constraint of the pre-built objective function is used to constrain the lithium battery SoC of the non-plug-in hydrogen fuel cell hybrid vehicle to be near a reference value; The second constraint condition of the pre-built objective function is used to constrain the fuel cell output power to be near the reference power; The third constraint condition of the pre-built objective function is used to constrain the transient output power of the fuel cell to be within a preset range.

9. A multi-mode adaptive frequency decoupling power distribution system, characterized in that: include: The clustering module is used to select feature vectors to extract time domain information from the vehicle speed prediction training dataset in the offline process to characterize the current driving state. The fuzzy C clustering method is used to classify the driving state into three standard driving states: high-speed cruising, medium-speed driving, and low-speed creeping. a sensitivity factor calculation module, configured to set a first sensitivity factor and a second sensitivity factor based on the current vehicle operating state, power demand, fuel cell power, and lithium battery SoC state, wherein the first sensitivity factor is related to the lithium battery SoC state, and the second sensitivity factor is related to the fuel cell output power; A coefficient matrix calculation module is used to establish an evaluation function for the comprehensive operating cost of non-plug-in hydrogen fuel cell hybrid vehicles. Two sensitivity factor adjustment coefficients corresponding to the minimum value of the evaluation function under three standard driving conditions are extracted as reference coefficients to form an adjustment coefficient reference matrix. The low-frequency sequence reconstruction module is used in the online link. It uses the "Haar" wavelet to perform three-level wavelet decoupling on the power demand change rate sequence, extracts the third-order approximate coefficients, and reconstructs the third-order approximate coefficients to obtain a low-frequency sequence. The power allocation module is used to fuzzy the adjustment coefficient reference matrix based on the driving state recognition results to obtain the adjustment coefficient corresponding to the current vehicle driving state, adaptively adjust the two sensitivity factors to optimize the low-frequency sequence, and determine the power allocation decision at the next moment based on the optimized low-frequency sequence. 10 . A vehicle power system energy management terminal, comprising the multi-mode adaptive frequency decoupled power distribution system according to claim 9 .

Images