Multi-target energy management strategy weight factor dynamic matching method, system and application

Abstract

This invention belongs to the field of fuel cell energy management technology, and discloses a method, system, and application for dynamic matching of weight factors in a multi-objective energy management strategy. In the offline stage, fuzzy C-clustering is used to classify driving cycle states. The objective function terms are processed according to the battery and fuel cell specifications and value ranges to ensure they are on the same order of magnitude before weight allocation. A weight distribution map for each driving state is built offline using dynamic programming, and the optimal weight allocation for each driving cycle state is selected. In the online stage, real-time pattern recognition results are used to perform fuzzy optimization based on membership degrees to obtain the real-time optimal weight allocation, which is then input into a quadratic programming solver. This invention is advantageous for adapting to different driving conditions; the real-time fuzzy optimization of weight coefficients followed by dynamic weight allocation adjustment effectively improves control performance.

Description

Technical Field

[0001] This invention belongs to the field of fuel cell energy management technology, and particularly relates to a method, system and application for dynamic matching of weight factors in multi-objective energy management strategies. Background Technology

[0002] Currently, hydrogen fuel cells possess advantages such as zero emissions, no pollution, high energy density, and strong environmental adaptability, thus holding immense potential in the transportation sector. Fuel cell hybrid electric vehicles (FCHEVs) for long-haul heavy-duty trucks represent a highly valuable research direction. However, single fuel cells cannot effectively handle transient power demand and energy recovery. Lithium-ion batteries, with their high energy density and rapid charge / discharge capabilities, are typically chosen as auxiliary energy sources in fuel cell hybrid systems. Therefore, research into energy management strategies that address power distribution between fuel cells and batteries is a crucial aspect.

[0003] In the research process of solving the power allocation decision of a hydrogen fuel cell hybrid power system in real time based on speed prediction results, setting a multi-objective cost function that considers the economy and durability of the fuel cell hybrid power system according to demand is crucial. However, a review of existing literature reveals that the selection of appropriate weighting coefficients for each term of the objective function is often overlooked in this process.

[0004] Based on the above analysis, the existing technology has the following problems and shortcomings: the weighting of each term in the multi-objective cost function of current hybrid electric vehicle energy management strategies. The weight coefficient chosen for each term of the objective function will greatly affect the control effect. The traditional weight matching process is usually set based on experience and then adjusted multiple times based on test results. This process has the following drawbacks: First, this parameter tuning method not only requires a lot of experience but also greatly prolongs the development cycle; second, the control effect obtained by the parameters selected through multiple parameter tunings is not necessarily optimal; finally, the selected parameters may have limitations and lack robustness. In practical applications, when facing complex driving conditions and the driving cycle data changes, the set weight parameters will greatly reduce the control effect, leading to increased hydrogen consumption and reduced durability of fuel cell hybrid electric vehicles.

[0005] The paper "Multi-mode predictive energy management for fuel cell hybrid electric vehicles using Markov driving pattern recognizer" describes a Markov pattern recognizer that classifies driving states into one of three predefined modes to improve prediction accuracy. Similarly, dynamically matching appropriate weight factors in real time under different driving states can effectively improve the system's optimization performance and robustness. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides a method, system, and application for dynamic matching of weight factors in multi-objective energy management strategies.

[0007] This invention is implemented as follows: a dynamic matching method for weight factors of a multi-objective energy management strategy. In the offline stage, the method uses fuzzy C-clustering to classify driving cycle states; it processes the terms of the objective function according to the specifications and value ranges of the battery and fuel cell, ensuring they are on the same order of magnitude, and then assigns weights; it uses dynamic programming to offline establish a weight distribution map for each driving state and selects the optimal weight allocation result for each driving cycle state; in the online stage, it uses real-time pattern recognition results to perform fuzzy optimization based on membership degrees to obtain the real-time optimal weight allocation, which is then input into a quadratic programming solver.

[0008] Furthermore, the dynamic matching method for weight factors of the multi-objective energy management strategy specifically includes the following steps:

[0009] Step 1: In the offline phase, based on the prediction accuracy, appropriate feature vectors are selected from the existing driving cycle database to represent the driving state, and the data is divided into 3-5 sub-driving cycle data using the fuzzy C-clustering method. The first to nth sub-driving cycle data are represented as follows: Each type of sub-driving cycle data represents a specific driving state;

[0010] Step 2: Normalize and downgrade each variable in the objective function according to the fuel cell and battery specifications and design requirements, so that they are all on the same order of magnitude, thereby reducing the workload of weight selection and testing.

[0011] Step 3: Based on the requirements of economic efficiency and durability of hydrogen fuel cell hybrid electric vehicles, an evaluation function is established. The established evaluation function is used as the dependent variable, and different weight distributions are used as independent variables. The dynamic programming (DP) method is used to train the function under the different driving state standards obtained in Step 1. Finally, the interpolation method is used to fit the measured data and draw the weight distribution map of each driving state.

[0012] Step 4: Based on the weight map, select the weight allocation corresponding to the minimum value of the evaluation function under each driving state standard as the optimal weight allocation decision, and construct the optimal weight allocation matrix accordingly. ,in These refer to the weight values ​​of a single weight factor in the objective function for the first, second, ..., nth class of standard driving conditions. These refer to the weight values ​​of the two weighting factors in the objective function for the first, second, ..., nth class of standard driving conditions. These refer to the weight values ​​of a single weight factor in the objective function for the first, second, ..., nth standard sub-driving states, respectively.

[0013] Step 5: In the online phase, the speed information collected by the vehicle speed sensor is transmitted to the controller. The controller receives the current vehicle speed information and combines it with the historical vehicle speed information within the previous Hm seconds to determine the driving state, where Hm is the number of historical time series. A membership matrix is ​​obtained by comparing this data with several specific driving cycle state standards obtained in Step 1. ,in These refer to the membership degree of the current vehicle state relative to the first, second, ..., nth standard sub-driving states, respectively. The optimal weight allocation matrix is ​​then subjected to fuzzy optimization using the membership matrix to obtain the real-time optimal weight matching result.

[0014] Step Six: The optimal weight allocation result obtained in Step Five is dynamically input into the Quadratic Programming (QP) solver in real time to obtain the optimal power allocation decision within the next Hp seconds, where Hp is the number of future time series. A rolling optimization method is adopted, and only the allocation result of the first second is used each time to control the output power of the fuel cell by adjusting the duty cycle.

[0015] Furthermore, the overall energy management strategy framework for hydrogen fuel cell hybrid electric vehicles based on the dynamic matching method of weight factors for the multi-objective energy management strategy adopts the Markov prediction method, and equation (1) is the objective function in the energy management strategy.

[0016]

[0017]

[0018] in, , , This is the result of normalizing the variables according to equation (2); This ensures that the battery SoC of non-plug-in hydrogen fuel cell hybrid vehicles remains near the reference value; The goal is to control the fuel cell output power near the reference power to improve the efficiency of the fuel cell system and reduce hydrogen consumption; It suppresses fuel cell system aging by controlling the transient output power of the fuel cell system; k is the current time, Hp is the future time series number, and i takes the value of a positive integer from 1 to Hp. Let k be the state of charge of the battery at time k+Hp. for The result of normalization using equation (a); To determine the fuel cell output power at time k+i, for The result after normalization using equation (b); For the fuel cell output power fluctuation at time k+i-1, for The result after normalization using equation (c); As a weighting factor, For adjustment coefficients, These are reference values ​​for fuel cell output power and battery state of charge, respectively. These are the upper and lower limits of the power transient of the fuel cell, respectively. These are the upper and lower limits of the battery's state of charge. The upper limit of fuel cell output power.

[0019] Furthermore, in step one, the average velocity, the standard deviation of acceleration, and the average acceleration are selected as follows: As a feature variable, the fuzzy C-clustering method divides driving cycle data into 3 to 5 categories.

[0020] Furthermore, in step two, an adjustment coefficient is set based on the selection of the fuel cell and battery to perform degradation processing, wherein the adjustment coefficient... The method for determining the value is shown in equation (3):

[0021]

[0022] in It refers to the range of battery state of charge that is desired to be maintained during a driving cycle; This refers to the capacity of the selected battery, expressed in Ah. This is the bus voltage.

[0023] Furthermore, the evaluation function for step three is shown in equation (4):

[0024]

[0025] in, For hydrogen consumption costs, For fuel cell system degradation, Due to battery degradation, This is the final value of the battery SoC at the end of the driving cycle. To evaluate the adjustment coefficients of the function, each design objective item can be adjusted according to the design requirements.

[0026] Step four involves establishing a weight distribution map for each driving state obtained through training using the dynamic programming method, and defining the weight distribution factors. As the independent variable, it satisfies equation (5); the grid is set to 0.05 for training, and the evaluation function E is used. _total As the dependent variable, it can be calculated using equation (4). Finally, the interpolation method is used to fit the measured data and draw the weight distribution map of each driving state. The weight distribution map of each driving cycle state is obtained as follows:

[0027]

[0028] Furthermore, in step five: based on the weight distribution Map of each state obtained in step four, select the evaluation function E. _total The W value corresponding to the minimum point is used as the most suitable weight for each driving cycle state; the driving cycle is divided into three types, and the optimal weight allocation matrix is: .

[0029] In step six, during the online phase, the driving status is assessed in real time, and the weight coefficients are fuzzy optimized using equation (6) based on the membership relationship between the current driving status and the set driving cycle status to obtain the optimal weight allocation:

[0030] (6)

[0031] in This is the membership matrix between the current driving state and the three standard driving cycles obtained in step one in this example.

[0032] One objective of this invention is to provide a method for selecting weighting factors in a multi-objective energy management strategy for fuel cell hybrid electric vehicles. This method does not rely on experience, does not require repeated parameter tuning based on test results, can shorten the development cycle, and improve control performance.

[0033] Another object of the present invention is to provide a controller device for a fuel cell hybrid electric vehicle, the controller device including a memory and a processor, the memory storing historical driving cycle information, standard driving cycle state information, an optimal weight allocation matrix and an offline computer program, the computer program being executed by the processor causing the processor to execute the multi-objective energy management strategy weight factor dynamic matching method.

[0034] The controller device for the fuel cell hybrid electric vehicle includes the following modules:

[0035] The data update module periodically updates the standard driving cycle status information based on recent driver historical driving cycle information, using fuzzy C-clustering to classify driving cycle states. The updated weight distribution map, generated using dynamic programming, further updates the optimal weight allocation matrix. The standard driving cycle status information and the optimal weight allocation matrix are stored in the controller device's memory for real-time application in online processes.

[0036] The speed prediction module uses Markov prediction to predict the speed within the next Hp seconds based on the speed information obtained from the vehicle's speed detection sensor over the past Hm seconds.

[0037] The pattern recognition module is used in the online process to compare the speed prediction results within Hp seconds with the standard driving cycle states stored in the optimal weight allocation matrix update module to perform real-time pattern recognition, obtain the membership matrix of the current driving state relative to each standard driving cycle state, and transmit it to the weight matching module.

[0038] The weight matching module performs fuzzy optimization on the optimal weight allocation matrix based on the membership matrix to obtain the final result, and then inputs it into the power allocation module.

[0039] The power allocation module uses a quadratic programming solver to solve for the optimal power allocation decision within the Hp range based on the speed prediction results and the real-time optimal weight allocation results. It performs rolling optimization by only using the allocation result of the first second each time.

[0040] Based on the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solution to be protected by this invention are as follows:

[0041] First, this invention employs fuzzy C-clustering to classify driving cycle states offline. Based on the battery and fuel cell specifications and value ranges, the objective function terms are processed to ensure they are on the same order of magnitude, facilitating weight allocation. A weight distribution map for each driving state is established offline using dynamic programming, and the optimal weight allocation for each driving cycle state is selected from this map. In the online phase, real-time pattern recognition results are used to perform fuzzy optimization based on the membership matrix to obtain the real-time optimal weight allocation, which is then input into the quadratic programming solver. Verification shows that this method improves weight matching efficiency and effectively enhances control performance.

[0042] Secondly, this invention first implements a method for selecting weighting factors, providing a methodological reference for the selection of weighting coefficients in energy management strategies based on model predictive control. This helps to select the optimal parameters to achieve the desired control effect, improving the control effect while greatly reducing the difficulty of parameter tuning. The invention also classifies driving cycle states and selects the most suitable weight allocation for each driving cycle state, which helps to adapt to different driving conditions. Furthermore, the invention dynamically adjusts the weight allocation after real-time fuzzy optimization of the weighting coefficients, which can effectively improve the control effect.

[0043] Third, as supplementary evidence of the inventive step of the claims of this invention, it is also reflected in the following important aspects:

[0044] In the selection of weighting factors in the multi-objective energy management strategy of fuel cell hybrid electric vehicles, traditional weight matching usually requires setting based on experience and repeated parameter tuning based on test results. The method provided in this invention overcomes the shortcomings of traditional methods, eliminating the need for reliance on experience and repeated parameter tuning based on test results, thus shortening the development cycle.

[0045] In traditional control strategies, the weighting factor is usually a fixed value. Unlike traditional prediction-based energy management strategies, this method adjusts the weighting factor in real time, which can effectively improve the sensitivity of the control system and further enhance the control effect.

[0046] The method employed in this invention is highly robust. In practical applications, when faced with complex driving conditions and changes in driving cycle data, unreasonable weight parameter settings can significantly reduce control effectiveness, leading to increased hydrogen consumption and reduced durability in fuel cell hybrid electric vehicles. This method, by periodically updating the standard driving cycle state and the optimal weight allocation matrix, can better adapt to various driving cycle states; it dynamically matches the optimal weight allocation in real time, improving the dynamics of control, better achieving the objective, and reducing hydrogen consumption and system losses.

[0047] Fourth, the specific significant technological advancements achieved in each step are as follows:

[0048] Step 1: In the offline phase, select appropriate feature vectors to represent the driving state and classify the driving cycle using the fuzzy C-clustering method.

[0049] Technological advancements: This step employs feature vectors to characterize driving states, which may involve selecting and processing multiple key parameters of the driving cycle to more accurately describe the driving state. Furthermore, fuzzy C-clustering is used to classify driving cycles. Fuzzy C-clustering has advantages in handling fuzzy data and uncertainties, and can better adapt to the diversity of actual driving conditions.

[0050] Step 2: Process the variables in the objective function according to the fuel cell and battery specifications to make them all of the same order of magnitude, thereby reducing the workload in the selection of weighting factors.

[0051] Technological advancement: This step processes the terms in the objective function to ensure they are on the same order of magnitude, tailored to different energy management goals and specifications. This simplifies the selection of weighting factors, prevents some terms from being too large or too small, and makes the allocation of weighting factors more reasonable and effective.

[0052] Step 3: Based on the requirements of economic efficiency and durability of hydrogen fuel cell hybrid electric vehicles, establish an evaluation function, and use dynamic programming to establish a weight distribution map under the several driving state standards obtained in Step 1.

[0053] Technological advancements: This step introduces requirements for the economy and durability of hydrogen fuel cell hybrid electric vehicles, which may involve multiple indicators such as vehicle performance, fuel consumption, and battery life. By employing dynamic programming methods, a weighted distribution map can be established based on multiple driving state standards, thereby better reflecting the weight combinations under various driving states and achieving multi-objective energy management optimization.

[0054] Step 4: Select the optimal weight allocation matrix for each driving state standard based on the weight map.

[0055] Technological advancement: This step selects the optimal weight allocation matrix through a weight map, which may involve comparing and evaluating multiple weight distribution results. In this way, the optimal weight combination can be found for each driving condition standard, providing an accurate reference for real-time energy management decisions.

[0056] Step 5: During the online phase, the driving status is judged in real time, and the weight coefficients are fuzzy optimized based on the membership relationship between the current driving status and the several driving cycle status standards obtained in Step 1.

[0057] Technological advancement: This step utilizes real-time pattern recognition results to determine the current driving state and performs fuzzy optimization based on the membership relationship between the driving state and the driving cycle state standard obtained in the offline stage to obtain the real-time optimal weight allocation. This real-time optimization method can better adapt to the energy management needs under different driving states, improving the flexibility and real-time performance of energy management.

[0058] Step 6: The optimal weight matching result obtained in Step 5 is dynamically input into the quadratic programming solver QP in real time to obtain the optimal power allocation result.

[0059] Technological advancement: This step dynamically inputs the real-time optimal weight matching results into a quadratic programming solver to obtain the optimal power allocation result. This dynamic, real-time solution method ensures the timeliness and accuracy of energy management decisions, meeting the energy optimization needs of vehicles under different conditions.

[0060] In summary, the dynamic matching method for weight factors in this multi-objective energy management strategy has achieved significant technological advancements at each step. From feature extraction, fuzzy C-clustering, dynamic programming, fuzzy optimization to online real-time decision solving, it forms a relatively complete and efficient energy management scheme as a whole. Attached Figure Description

[0061] Figure 1 This is a flowchart of the dynamic matching method for weight factors of multi-objective energy management strategy provided in this embodiment of the invention;

[0062] Figure 2 This is an overall framework diagram of the energy management strategy provided in the embodiments of the present invention;

[0063] Figure 3 This is a flowchart of the dynamic matching weight factor method provided in the embodiments of the present invention;

[0064] Figure 4 This is an example diagram illustrating the driving cycle classification effect provided in an embodiment of the present invention;

[0065] Figure 5 This is an example diagram of establishing a weight distribution Map under various driving states provided in the embodiments of the present invention;

[0066] Figure 6 This is the real-time dynamic matching result of the weight factors in the example provided by the present invention;

[0067] Figure 7 This is a comparison of the SoC and total power consumption function of the present invention with those of the conventional method under driving cycle 1;

[0068] Figure 8 This is a comparison of the SoC and total power consumption function of the present invention with those of the traditional method under the CWTVC driving cycle. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0070] like Figure 1 As shown, the multi-objective energy management strategy weight factor dynamic matching method provided in this embodiment of the invention includes the following steps:

[0071] S101: In the offline stage, fuzzy C-clustering method is used to classify the driving cycle state; the terms of the objective function are processed according to the battery and fuel cell specifications and value ranges to make them all of the same order of magnitude and weights are assigned.

[0072] S102: Use dynamic programming to build a weight distribution map of each driving state offline and select the best weight allocation result for each driving cycle state from it;

[0073] S103: In the online stage, the real-time pattern recognition results are used to perform fuzzy optimization based on the membership degree to obtain the real-time optimal weight allocation, which is then input into the quadratic programming solver.

[0074] The multi-objective energy management strategy weight factor dynamic matching method provided in this embodiment of the invention specifically includes the following steps:

[0075] Step 1: In the offline phase, based on the prediction accuracy, appropriate feature vectors are selected from the existing driving cycle database to represent the driving state, and the data is divided into 3-5 sub-driving cycle data using the fuzzy C-clustering method. The first to nth sub-driving cycle data are represented as follows: Each type of sub-driving cycle data represents a specific driving state;

[0076] Step 2: Normalize and downgrade each variable in the objective function according to the fuel cell and battery specifications and design requirements, so that they are all on the same order of magnitude, thereby reducing the workload of weight selection and testing.

[0077] Step 3: Based on the requirements of economic efficiency and durability of hydrogen fuel cell hybrid electric vehicles, an evaluation function is established. The established evaluation function is used as the dependent variable, and different weight distributions are used as independent variables. The dynamic programming (DP) method is used to train the function under the different driving state standards obtained in Step 1. Finally, the interpolation method is used to fit the measured data and draw the weight distribution map of each driving state.

[0078] Step 4: Based on the weight map, select the weight allocation corresponding to the minimum value of the evaluation function under each driving state standard as the optimal weight allocation decision, and construct the optimal weight allocation matrix accordingly. ,in These refer to the weight values ​​of a single weight factor in the objective function for the first, second, ..., nth class of standard driving conditions. These refer to the weight values ​​of the two weighting factors in the objective function for the first, second, ..., nth class of standard driving conditions. These refer to the weight values ​​of a single weight factor in the objective function for the first, second, ..., nth standard sub-driving states, respectively.

[0079] Step 5: In the online phase, the speed information collected by the vehicle speed sensor is transmitted to the controller. The controller receives the current vehicle speed information and combines it with the historical vehicle speed information within the previous Hm seconds to determine the driving state, where Hm is the number of historical time series. A membership matrix is ​​obtained by comparing this data with several specific driving cycle state standards obtained in Step 1. ,in These refer to the membership degree of the current vehicle state relative to the first, second, ..., nth standard sub-driving states, respectively. The optimal weight allocation matrix is ​​then subjected to fuzzy optimization using the membership matrix to obtain the real-time optimal weight matching result.

[0080] Step Six: The optimal weight allocation result obtained in Step Five is dynamically input into the Quadratic Programming (QP) solver in real time to obtain the optimal power allocation decision within the next Hp seconds, where Hp is the number of future time series. A rolling optimization method is adopted, and only the allocation result of the first second is used each time to control the output power of the fuel cell by adjusting the duty cycle.

[0081] This invention, in a model predictive control-based energy management strategy for hybrid electric vehicles, obtains prediction results in real time within HP seconds using a selected prediction method. A quadratic programming (QP) solver is then used to obtain the optimal power allocation decision at each backtracking level. A rolling optimization approach is employed, and the decision results are input into the hydrogen fuel cell hybrid system for control, thereby achieving the desired control effect. The proposed method addresses the weight setting problem of the objective function in the energy management strategy. The overall framework of the hydrogen fuel cell hybrid electric vehicle energy management strategy is as follows: Figure 2 As shown, the Markov prediction method is adopted, and equation (1) is the objective function in the energy management strategy.

[0082]

[0083]

[0084] in, , ,

[0085] This is the result of normalizing the variables according to equation (2).

[0086] This ensures that the battery SoC of non-plug-in hydrogen fuel cell hybrid vehicles remains near the reference value; Controlling the fuel cell output power to be near the reference power can improve the efficiency of the fuel cell system and reduce hydrogen consumption;

[0087] It suppresses fuel cell system aging by controlling the transient changes in the output power of the fuel cell system.

[0088] The specific steps of the weight factor dynamic matching method for the multi-objective energy management strategy of hybrid electric vehicles according to this invention are as follows:

[0089] Step 1: Compared to a single-mode baseline strategy, setting a multi-mode strategy can improve control and prediction accuracy. Therefore, based on prediction accuracy, appropriate feature vectors are selected offline to represent the driving state, and driving cycles are classified using clustering methods. In this example, we select... As a feature variable, the Fuzzy C-Clustering (FCM) method divides the driving cycle data into three categories. For ease of viewing, examples of the classification results from 600s to 2000s are shown here. Figure 4 As shown;

[0090] Step Two: Adjust the variables in the multi-objective cost function according to the fuel cell and battery specifications to ensure they are on the same order of magnitude, reducing the workload in selecting weighting factors. Use adjustment coefficients based on the selected fuel cell and battery type. Make adjustments. Value retrieval method:

[0091]

[0092] in It refers to the range of battery state of charge that is desired to be maintained during a driving cycle; This refers to the capacity of the selected battery, expressed in Ah. This refers to the bus voltage, in this example. Take values ​​of 1, 1 / 1000, and 1 / 1000 respectively.

[0093] Step 3: Establish the evaluation function based on the requirements. In this example, to improve the economy and durability of the hydrogen fuel cell hybrid power system, the hydrogen consumption of the fuel cell, battery consumption, and degradation of the fuel cell and battery are considered. The evaluation function in this example is as follows:

[0094]

[0095] in, For hydrogen consumption costs, For fuel cell system degradation, For battery degradation, the adjustment coefficient of the evaluation function is used in this example. Take values ​​of 10, 1, and 2 respectively.

[0096] Step 4: Establish a weight distribution map for each driving state obtained through training using the dynamic programming method, with weight allocation factors. As the independent variable, it satisfies equation (5). The grid size is set to 0.05 for training, and the evaluation function E... _total The dependent variable can be calculated using equation (4) above. Finally, the interpolation method is used to fit the measured data and draw the weight distribution map for each driving state. The weight distribution map for each driving cycle state obtained in this example is as follows: Figure 5 As shown:

[0097]

[0098] Step 5: Based on the weight distribution map of each state obtained in Step 4, select the evaluation function E. _total The W value corresponding to the minimum value is used as the most suitable weight for each driving cycle state. In this example, the driving cycle is divided into three types, and the optimal weight allocation matrix is: The specific value is .

[0099] Step Six: During the online phase, the driving status is assessed in real time, and the weight coefficients are fuzzy optimized using equation (6) based on the membership relationship between the current driving status and the set driving cycle status to obtain the optimal weight allocation:

[0100]

[0101] in This is the membership matrix between the current driving state and the three standard driving cycles obtained in step one in this example.

[0102] The multi-objective energy management strategy weight factor dynamic matching system provided in this embodiment of the invention includes:

[0103] The data update module is used to classify driving cycle states offline using fuzzy C-clustering based on recent driver historical driving cycle information, and periodically update the standard driving cycle state information. It uses dynamic programming to offline build a weight distribution map for each standard driving state and selects the optimal weight allocation result for each driving cycle state to update the optimal weight allocation matrix. The standard driving cycle state information and the optimal weight allocation matrix are stored in the controller device's storage for real-time application in the online phase.

[0104] The speed prediction module uses Markov prediction to predict the speed within the next Hp seconds based on the speed information obtained from the vehicle's speed detection sensor over the past Hm seconds.

[0105] The pattern recognition module is used in the online process to compare the speed prediction results within Hp seconds with the standard driving cycle states stored in the optimal weight allocation matrix update module to perform real-time pattern recognition, obtain the membership matrix of the current driving state relative to each standard driving cycle state, and transmit it to the weight matching module.

[0106] The weight matching module performs fuzzy optimization on the optimal weight allocation matrix based on the membership matrix to obtain the final result, and then inputs it into the power allocation module.

[0107] The power allocation module uses a quadratic programming solver to solve for the optimal power allocation decision within the Hp range based on the speed prediction results and the real-time optimal weight allocation results. It performs rolling optimization by only using the allocation result of the first second each time.

[0108] As shown in the training evaluation function (7), the desired control effect in this example is to reduce hydrogen consumption and losses in the hydrogen fuel cell hybrid electric vehicle, namely fuel cell system losses and battery losses, while simultaneously controlling the stability of the battery SoC. To verify the control effect obtained in the online application stage, this example sets a total consumption function C_total based on the economy and durability of the fuel cell hybrid electric vehicle, as shown in Equation (7).

[0109]

[0110] in It's the price of hydrogen fuel. It's hydrogen consumption; This refers to the unit price of a hydrogen fuel cell system. It is the degradation of the hydrogen fuel cell system; That's the unit price of the battery pack. This is battery degradation. Since the degradation items in fuel cell systems are generally small, this example magnifies them by 10 times for better observation.

[0111] Figure 7 The proposed method was compared with traditional prediction-based SoC and total power consumption function methods under driving cycle 1 of a heavy-duty truck obtained from ADVISOR (Advanced Vehicle Simulator). The results show that the proposed method can better control the battery SoC to stay near the reference value and has better control performance. To verify the robustness of the proposed method, the driving cycle was modified and tested under the CWTVC (World Transient Vehicle Cycle) driving cycle for comparison. Figure 8This comparison examines the proposed method versus the traditional prediction-based SoC (System-on-Chips) and its evaluation function after altering the driving cycle. It is evident that the battery SoC obtained using this method exhibits better stability, significantly improving the economy and durability of the hydrogen fuel cell hybrid vehicle. Detailed comparisons of each component are shown in Table 1.

[0112] Table 1 provides a detailed comparison between the examples of this invention and traditional methods.

[0113] Driving Cycle Total consumption Hydrogen consumption (kg) fuel cell degradation Battery degradation Driving Cycle 1 Method 17.72 4.398 0.0922 3.605 Driving Cycle 1 Traditional MPC 19.42 5.393 0.1747 3.858 CWTVC This Method 11.39 2.663 0.06575 1.773 CWTVC Traditional MPC 12.29 3.211 0.07783 1.827

[0114] Example 1: Energy Management Strategy for Hydrogen Fuel Cell Hybrid Bus

[0115] 1. Feature Selection and Driving Cycle Classification: In the offline phase, appropriate feature vectors are selected to describe the bus's driving state, such as vehicle speed, acceleration, and battery SOC. Then, fuzzy C-clustering is used to classify the bus's driving cycles, dividing different types of driving cycles into several categories.

[0116] 2. Objective Function Processing and Weight Allocation: Based on the bus's fuel cell and battery specifications, the objective function for energy management is processed to ensure all objectives are on the same order of magnitude. Then, a dynamic programming method is used to establish a weight distribution map under different driving conditions, from which the optimal weight allocation matrix for each driving condition is selected to meet the requirements of bus economy and durability.

[0117] 3. Online Fuzzy Optimization and Quadratic Programming: In the online phase, the current driving state of the bus is determined in real time, and the weight coefficients are fuzzy optimized based on the membership relationship between the real-time driving state and the driving cycle state standard obtained offline. Finally, the obtained optimal weights are dynamically input into the quadratic programming solver to obtain the real-time optimal power allocation result, which is used to control the bus's power system.

[0118] Example 2: Energy Management Strategy for Hydrogen Fuel Cell Hybrid Electric Vehicles

[0119] 1. Feature Selection and Driving Cycle Classification: In the offline phase, appropriate feature vectors are selected to describe the driving state of the car, such as parameters like vehicle speed, acceleration, and throttle opening. Then, fuzzy C-clustering is used to classify the car's driving cycles, dividing different types of driving cycles into several categories.

[0120] 2. Objective Function Processing and Weight Allocation: Based on the fuel cell and battery specifications of the car, the objective function for energy management is processed to ensure that all objectives are on the same order of magnitude. Then, a dynamic programming method is used to establish a weight distribution map under different driving conditions, from which the optimal weight allocation matrix for each driving condition is selected to meet the requirements of the car's economy and durability.

[0121] 3. Online Fuzzy Optimization and Quadratic Programming: In the online phase, the current driving state of the car is determined in real time, and the weight coefficients are fuzzy optimized based on the membership relationship between the real-time driving state and the driving cycle state standard obtained offline. Finally, the obtained optimal weights are dynamically input into the quadratic programming solver to obtain the real-time optimal power allocation result, which is used to control the car's power system.

[0122] In these embodiments, specific methods such as feature selection, fuzzy C-clustering, dynamic programming, fuzzy optimization, and quadratic programming can be implemented according to specific circumstances, such as using appropriate algorithms and programming languages ​​to implement related functions. Furthermore, the dynamic matching method for weight factors involved in the embodiments shows promise for application in different types of hydrogen fuel cell hybrid vehicles, and can be adjusted and optimized accordingly based on the characteristics and needs of the vehicle.

[0123] Embodiments of the present invention can be implemented in hardware, software, or a combination of both. The hardware portion can be implemented using dedicated logic; the software portion can be stored in memory and executed by a suitable instruction execution system, such as a microprocessor or dedicated-design hardware. Those skilled in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and / or included in processor control code, for example, such code provided on a carrier medium such as a disk, CD, or DVD-ROM, a programmable memory such as read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuitry such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field-programmable gate arrays, programmable logic devices, etc., or by software executed by various types of processors, or by a combination of the above-described hardware circuitry and software, such as firmware.

[0124] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions, and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention, and within the spirit and principles of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A dynamic matching method for weight factors in a multi-objective energy management strategy, characterized in that, In the offline phase, fuzzy C-clustering is used to classify driving cycle states; the objective function terms are processed according to the battery and fuel cell specifications and value ranges to ensure they are on the same order of magnitude, and weights are assigned; a weight distribution map of each driving state is built offline using dynamic programming, and the optimal weight assignment result for each driving cycle state is selected from it. In the online process, the real-time pattern recognition results are used to perform fuzzy optimization based on membership degree to obtain the real-time optimal weight allocation, which is then input into the quadratic programming solver. The dynamic matching method for weight factors of the multi-objective energy management strategy specifically includes the following steps: Step 1: In the offline phase, based on the prediction accuracy, appropriate feature vectors are selected from the existing driving cycle database to represent the driving state, and the data is divided into 3-5 sub-driving cycle data using the fuzzy C-clustering method. The first to nth sub-driving cycle data are represented as follows: Each type of sub-driving cycle data represents a specific driving state; Step 2: Normalize and downgrade each variable in the objective function according to the fuel cell and battery specifications and design requirements, so that they are all on the same order of magnitude, thereby reducing the workload of weight selection and testing. Step 3: Establish an evaluation function based on the economic and durability requirements of fuel cell hybrid electric vehicles. Use the established evaluation function as the dependent variable and different weight distributions as independent variables. Train the function under the driving conditions obtained in Step 1 using dynamic programming (DP) method. Finally, use interpolation to fit the measured data and draw the weight distribution map for each driving condition. Step 4: Based on the weight map, select the weight allocation corresponding to the minimum value of the evaluation function under each driving state standard as the optimal weight allocation decision, and construct the optimal weight allocation matrix accordingly. ,in These refer to the weight values ​​of a single weight factor in the objective function for the first, second, ..., nth class of standard driving conditions. These refer to the weight values ​​of the two weighting factors in the objective function for the first, second, ..., nth class of standard driving conditions. These refer to the weight values ​​of a single weight factor in the objective function for the first, second, ..., nth standard sub-driving states, respectively. Step 5: In the online phase, the speed information collected by the vehicle speed sensor is transmitted to the controller. The controller receives the current vehicle speed information and combines it with the historical vehicle speed information within the previous Hm seconds to determine the driving state, where Hm is the number of historical time series. A membership matrix is ​​obtained by comparing this data with several specific driving cycle state standards obtained in Step 1. ,in These refer to the membership degree of the current vehicle state relative to the first, second, ..., nth standard sub-driving states, respectively; the membership degree matrix is ​​used to perform fuzzy optimization on the optimal weight allocation matrix as the real-time optimal weight matching result; Step 6: The optimal weight allocation result obtained in Step 5 is dynamically input into the quadratic programming solver QP in real time to obtain the optimal power allocation decision within the next Hp seconds, where Hp is the number of future time series. The rolling optimization method is adopted, and only the allocation result of the first second is used each time to control the output power of the fuel cell by adjusting the duty cycle.

2. The dynamic matching method for weight factors of multi-objective energy management strategies as described in claim 1, characterized in that, The overall energy management strategy framework for hydrogen fuel cell hybrid electric vehicles, based on the dynamic matching method of weight factors for the multi-objective energy management strategy, adopts the Markov prediction method. Equation (1) is the objective function in the energy management strategy. (1); ; in, , , This is the result of normalizing the variables according to equation (2); This ensures that the battery SoC of non-plug-in hydrogen fuel cell hybrid vehicles remains near the reference value; The goal is to control the fuel cell output power near the reference power to improve the efficiency of the fuel cell system and reduce hydrogen consumption; This is achieved by controlling the transient output power of the fuel cell system to suppress the aging of the fuel cell system; k is the current time, Hp is the future time series number, and i takes the value of a positive integer from 1 to Hp; Let k be the state of charge of the battery at time k+Hp. for The result of normalization using equation (a); To determine the fuel cell output power at time k+i, for The result after normalization using equation (b); For the fuel cell output power fluctuation at time k+i-1, for The result after normalization using equation (c); As a weighting factor, For adjustment coefficients, These are reference values ​​for fuel cell output power and battery state of charge, respectively. These are the upper and lower limits of the battery's state of charge. Upper and lower limits of fuel cell power transients Maximum output power of fuel cells.

3. The dynamic matching method for weight factors of multi-objective energy management strategies as described in claim 1, characterized in that, The first step is to select As a feature variable, the fuzzy C-clustering method divides driving cycle data into 3 to 5 categories.

4. The dynamic matching method for weight factors of multi-objective energy management strategies as described in claim 1, characterized in that, Step two involves setting adjustment coefficients for degradation processing based on the selection of fuel cells and batteries, wherein the adjustment coefficients... Value retrieval method: (3), among which, It is something we hope to maintain. scope; This refers to the capacity of the selected battery, expressed in Ah. This is the bus voltage.

5. The dynamic matching method for weight factors of multi-objective energy management strategies as described in claim 1, characterized in that, The evaluation function in step three: ;in, For hydrogen consumption costs, For fuel cell system degradation, Due to battery degradation, This is the final value of the battery SoC at the end of the driving cycle. The adjustment coefficients for the evaluation function can be adjusted for each design objective item according to the design requirements; Step four involves establishing a weight distribution map for each driving state obtained through training using the dynamic programming method, and defining the weight distribution factors. As the independent variable, it satisfies equation (5); the grid is set to 0.05 for training, E _total As the dependent variable, it can be calculated using equation (4). Finally, the interpolation method is used to fit the measured data and draw the weight distribution map of each driving state. The weight distribution map of each driving cycle state is obtained as follows: 。 6. The method for dynamic matching of weight factors in a multi-objective energy management strategy as described in claim 1, characterized in that, Step 5: Based on the weight distribution Map obtained in Step 4, select the W value corresponding to the minimum of the evaluation function E_total as the most suitable weight for each driving cycle state. The driving cycle is divided into three types, and the optimal weight allocation matrix is ​​as follows: ; In step six, during the online phase, the driving status is assessed in real time, and the weight coefficients are fuzzy optimized using equation (6) based on the membership relationship between the current driving status and the set driving cycle status to obtain the optimal weight allocation: （6）； in, This is the membership matrix between the current driving state and the three standard driving cycles obtained in step one; Step six takes the real-time dynamics processed by fuzzy optimization in equation (6) as a variable and inputs it into the QP solver to solve the optimal power allocation decision within the Hp range. Only the allocation result of the first second is used each time for rolling optimization.

7. A computer device, characterized in that, The computer device includes a memory and a processor. The memory stores a computer program. When the computer program is executed by the processor, the processor performs the dynamic matching method for weight factors of the multi-objective energy management strategy as described in claims 1 to 6.

8. A system for dynamic matching of weight factors for multi-objective energy management strategies based on the dynamic matching method for weight factors of multi-objective energy management strategies according to any one of claims 1 to 6, characterized in that, include: The data update module is used to classify driving cycle states offline using fuzzy C-clustering based on recent driver historical driving cycle information, and periodically update standard driving cycle state information; it also uses dynamic programming to offline build a weight distribution map of each standard driving state and selects the best weight allocation result for each driving cycle state to update the optimal weight allocation matrix. The standard driving cycle state information and the optimal weight allocation matrix are stored in the controller device's storage for real-time application in online processes; The speed prediction module uses a multi-step Markov prediction method to predict the speed within the next Hp seconds based on the speed information obtained from the vehicle's speed detection sensor within the past Hm seconds. The pattern recognition module is used in the online process to compare the speed prediction results within Hp seconds with the standard driving cycle states stored in the optimal weight allocation matrix update module to perform real-time pattern recognition, obtain the membership matrix of the current driving state relative to each standard driving cycle state, and transmit it to the weight matching module. The weight matching module performs fuzzy optimization on the optimal weight allocation matrix based on the membership matrix to obtain the final result, and then inputs it into the power allocation module. The power allocation module uses a quadratic programming solver to solve for the optimal power allocation decision within the Hp range based on the speed prediction results and the real-time optimal weight allocation results. It performs rolling optimization by only using the allocation result of the first second each time.

9. A hydrogen fuel cell hybrid electric vehicle, characterized in that, The hydrogen fuel cell hybrid electric vehicle includes the multi-objective energy management strategy weight factor dynamic matching system as described in claim 8.

Images