Fuel cell heavy truck energy management method and system considering energy-saving driving

Abstract

The invention provides a fuel cell heavy truck energy management method and system considering energy-saving driving. The method comprises the steps that an energy-saving driving index is determined according to driving state parameters obtained in real time; based on the energy-saving driving index, the front road information, the driving motor power and the driving state parameters, utilizing a space time sequence fused graph neural network prediction model to predict an instantaneous vehicle power demand; constructing a multi-objective optimization model and solving the model to obtain the target output power of the fuel cell stack and the target charging and discharging power of the energy storage device; issuing a power instruction to the fuel cell stack and the energy storage device according to the target output power and the target charging and discharging power, and triggering an energy storage compensation mode when the deviation between the actual driving power and the predicted power demand exceeds a preset condition; and the weight of the model is updated based on the deviation and the change of the energy-saving driving index, and energy-saving driving information is fed back to the driver. In this way, the driver is guided to optimize operation so as to further excavate the energy-saving potential, and the energy consumption economy and the driving range of the whole vehicle are improved.

Description

Technical Field

[0001] This invention relates to the field of vehicle power control and energy management technology, and in particular to an energy management method and system for fuel cell heavy trucks that takes into account energy-saving driving. Background Technology

[0002] Existing energy management methods for fuel cell heavy-duty trucks mainly include start-stop control based on empirical rules, dynamic power distribution, fuzzy logic control, and model predictive control. These methods primarily focus on balancing power output between the fuel cell stack and the energy storage device. They optimize energy distribution through statistical analysis of historical driving conditions and prediction using fixed rules or offline models, aiming to achieve a trade-off between vehicle power response and fuel economy under different driving scenarios.

[0003] However, most existing energy management strategies for fuel cell heavy-duty trucks lack effective feedback mechanisms for drivers' actual energy-saving behaviors, failing to further tap the potential energy-saving potential of driving behavior. The existing energy management methods for fuel cell heavy-duty trucks have poor adaptability to operating conditions, relying solely on historical data or fixed rules, making it difficult to respond promptly to changes in road conditions. This results in the fuel cell stack frequently operating in low-efficiency areas, making it difficult to balance range optimization and system durability requirements, leading to a smaller overall vehicle energy economy and driving range. Summary of the Invention

[0004] The purpose of this invention is to provide a fuel cell heavy-duty truck energy management method and system that takes into account energy-saving driving, thereby solving the problems that existing technologies cannot further tap the potential energy-saving energy of driving behavior, as well as the issues of the price difference in energy consumption economy and the short driving range of the whole vehicle.

[0005] To address the aforementioned technical problems, the embodiments of the present invention provide the following technical solutions: The first aspect of this invention provides an energy management method for fuel cell heavy-duty trucks that takes into account energy-saving driving, comprising: The energy-saving driving index is determined based on real-time driving status parameters. The energy-saving driving index is used to indicate and quantify the driver's energy-saving behavior. Based on the energy-saving driving index, road information ahead, drive motor power and driving status parameters, a spatial-temporal fusion graph neural network prediction model is used to predict and output the instantaneous vehicle power demand in the prediction time domain. With fuel cell stack efficiency and energy storage device cycle depth as objectives and instantaneous vehicle power demand as constraints, a multi-objective optimization model is constructed and solved to obtain the target output power of the fuel cell stack and the target charge and discharge power of the energy storage device. Based on the target output power and target charge / discharge power, power commands are issued to the fuel cell stack and energy storage device respectively. When the deviation between the actual driving power and the predicted power demand exceeds the preset conditions, the energy storage device compensation mode is triggered to maintain power output. Based on the changes in deviation and energy-saving driving index, the weight coefficients of the multi-objective optimization model are updated, the energy-saving driving index is converted into corresponding energy-saving driving information, and the information is fed back to the driver through the human-machine interface to achieve real-time adjustment of energy management strategy and closed-loop guidance of driving behavior.

[0006] A second aspect of the present invention provides an energy management system for fuel cell heavy-duty trucks that takes into account energy-saving driving, comprising: The determination module is used to determine the energy-saving driving index based on the real-time driving status parameters. The energy-saving driving index is used to indicate and quantify the driver's energy-saving behavior. The prediction module is used to make predictions based on the energy-saving driving index, road information ahead, drive motor power and driving status parameters, using a spatial-temporal fusion graph neural network prediction model to output the instantaneous vehicle power demand in the prediction time domain. The construction and solution module is used to construct a multi-objective optimization model with the fuel cell stack efficiency and energy storage device cycle depth as objectives and instantaneous vehicle power demand as constraints, and solve the multi-objective optimization model to obtain the target output power of the fuel cell stack and the target charge and discharge power of the energy storage device. The triggering module is used to issue power commands to the fuel cell stack and the energy storage device respectively according to the target output power and the target charge and discharge power, and to trigger the energy storage device compensation mode to maintain power output when the deviation between the actual driving power and the predicted power demand exceeds the preset conditions. The update and feedback module is used to update the weight coefficients of the multi-objective optimization model based on the changes in deviation and energy-saving driving index, convert the energy-saving driving index into corresponding energy-saving driving information, and feed it back to the driver through the human-machine interface to realize real-time adjustment of energy management strategy and closed-loop guidance of driving behavior.

[0007] Compared to existing technologies, the energy management method and system for fuel cell heavy-duty trucks that considers energy-saving driving provided by this invention determines the energy-saving driving index based on real-time acquired driving state parameters; based on the energy-saving driving index, road information ahead, drive motor power, and driving state parameters, a spatial-temporal fusion graph neural network prediction model is used to predict and output the instantaneous vehicle power demand in the predicted time domain; with fuel cell stack efficiency and energy storage device cycle depth as objectives and instantaneous vehicle power demand as constraints, a multi-objective optimization model is constructed and solved to obtain the target output power of the fuel cell stack and the target charge / discharge power of the energy storage device; based on the target output power and target charge / discharge power, power commands are issued to the fuel cell stack and energy storage device respectively, and when the deviation between the real-time monitored actual drive power and the predicted power demand exceeds a preset condition, the energy storage device compensation mode is triggered to maintain power output; based on the deviation and changes in the energy-saving driving index, the weight coefficients of the multi-objective optimization model are updated, the energy-saving driving index is converted into corresponding energy-saving driving information, and feedback is given to the driver through a human-machine interface to achieve real-time adjustment of energy management strategies and closed-loop guidance of driving behavior. In this way, by introducing a graph neural network prediction model based on vehicle-to-everything (V2X) road information and spatial-temporal fusion, the system achieves forward-looking and accurate prediction of instantaneous vehicle power demand. A real-time multi-objective optimization model is constructed with fuel cell stack efficiency and energy storage cycle depth as its core, combined with online adaptive weight updates. This ensures that the fuel cell always operates within its high-efficiency range and the energy storage device maintains shallow cycle operation, significantly reducing hydrogen consumption and slowing down the degradation of the power battery and stack. Simultaneously, in the event of communication failure or sudden changes in operating conditions, the system automatically switches to short-term prediction and triggers the energy storage device's compensation mode, ensuring rapid response and power continuity in complex road conditions. At the driving experience and operational level, the real-time visualization of the energy-saving driving index (EDI) on the human-machine interface forms a "driver-energy management" closed loop, guiding drivers to optimize their operating habits to further tap into energy-saving potential. This improves the overall vehicle's energy economy and driving range, while also reducing system implementation difficulty and maintenance costs, demonstrating good engineering feasibility and promotional value. Attached Figure Description

[0008] The above and other objects, features, and advantages of exemplary embodiments of the present invention will become readily apparent from the following detailed description taken in conjunction with the accompanying drawings. In the drawings, several embodiments of the invention are illustrated by way of example and not limitation, with the same or corresponding reference numerals denoteing the same or corresponding parts, wherein: Figure 1 The flowchart of a fuel cell heavy-duty truck energy management method that takes into account energy-saving driving is illustrated schematically. Figure 1 ; Figure 2 The flowchart of a fuel cell heavy-duty truck energy management method that takes into account energy-saving driving is illustrated schematically. Figure 2 ; Figure 3 A schematic diagram of the data flow of the vehicle system is shown. Figure 4 A schematic diagram illustrating instantaneous power demand forecasting and allocation is provided. Figure 5 A schematic diagram of the energy storage compensation mode is shown. Figure 6 An optimization diagram is shown schematically; Figure 7 A schematic diagram illustrating the calculation of the energy-saving driving index is shown. Figure 8 A schematic diagram of the energy management system for a fuel cell heavy-duty truck that takes into account energy-saving driving is shown. Detailed Implementation

[0009] Exemplary embodiments of the invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the invention and to fully convey the scope of the invention to those skilled in the art.

[0010] It should be noted that, unless otherwise stated, the technical or scientific terms used in this invention should have the ordinary meaning as understood by those skilled in the art.

[0011] The methods in the embodiments of the present invention will be described in detail below.

[0012] Example 1: Figure 1 The flowchart of a fuel cell heavy-duty truck energy management method that takes into account energy-saving driving is illustrated schematically. Figure 1 , Figure 2 The flowchart of a fuel cell heavy-duty truck energy management method that takes into account energy-saving driving is illustrated schematically. Figure 2 , Figure 3 The diagram illustrates the data flow of the in-vehicle system. Figure 4 A schematic diagram illustrating instantaneous power demand forecasting and allocation is shown below. Figures 1 to 4 As shown, the energy management method for fuel cell heavy-duty trucks that takes into account energy-saving driving may include: S101. Determine the energy-saving driving index based on the real-time driving status parameters.

[0013] The energy-saving driving index is used to quantify and indicate a driver's energy-saving behavior. Driving parameters include vehicle speed, longitudinal acceleration, road gradient, and total vehicle mass.

[0014] The expression for the Eco-Driving Index (EDI) is: ; in, For energy-saving driving index, This represents the number of sampling points within the current control cycle. For the first The longitudinal acceleration of each sampling point For the first The vehicle speed at each sampling point Let be the acceleration due to gravity, and take . The lower the EDI value, the more energy-efficient the driving behavior.

[0015] The instantaneous speed of a vehicle (i.e., vehicle speed) affects power demand, and the degree of acceleration or deceleration of the vehicle (i.e., longitudinal acceleration) directly affects the vehicle's power demand. The EDI (Energy-Saving Driving Index) is an indicator calculated based on vehicle speed and longitudinal acceleration, used to quantify the driver's energy-saving driving behavior. A lower EDI value indicates better energy-saving driving behavior, while a higher EDI value indicates poorer energy-saving driving behavior.

[0016] For example, the specific operation is as follows: the vehicle is driving on a smooth suburban road, the control cycle is set to 5 seconds, and the on-board controller synchronously collects a total of 10 sets of data from the Controller Area Network (CAN) bus and the Inertial Measurement Unit (IMU) at a frequency of 10Hz. Table 1 is the data table for calculating the energy-saving driving index. Table 1 gives 10 sets of data, including sampling points, vehicle speed and longitudinal acceleration.

[0017] Table 1. Data Table for Calculating the Energy-Saving Driving Index

[0018] To ensure data quality, all raw signals are first filtered by a fourth-order Butterworth low-pass filter to remove road bumps and interference. Then, the controller calculates the energy-saving driving index before the end of each cycle, with a result of approximately 0.58.

[0019] S102. Based on the energy-saving driving index, road information ahead, drive motor power and driving status parameters, a spatial-temporal fusion graph neural network prediction model is used to predict and output the instantaneous vehicle power demand in the prediction time domain.

[0020] The road information ahead includes the road's longitudinal slope curve, traffic flow density, traffic light phase, and remaining time of the traffic lights.

[0021] Specifically, prior to step S102, the method further includes: Step A1: Obtain the power of the drive motor in real time.

[0022] It is also necessary to obtain the state of charge of the fuel cell in real time, and the real-time obtained state of charge of the fuel cell will be used in the subsequent multi-objective optimization model.

[0023] Step A2: During each control cycle, retrieve road information ahead via the vehicle-to-everything (V2X) communication interface.

[0024] Among them, the road information ahead is the road information within the future preset driving distance for each control cycle.

[0025] Specifically, based on the energy-saving driving index, road information ahead, drive motor power, and driving state parameters, a spatial-temporal fusion graph neural network prediction model is used to predict and output the instantaneous vehicle power demand in the predicted time domain, including: Step B1: Normalize and embed the road longitudinal slope curve, traffic flow density, traffic light phase, vehicle speed, longitudinal acceleration and energy-saving driving index in sequence, and divide the processed data according to the preset time step to construct the corresponding time-series feature vector.

[0026] Step B2: Input the temporal feature vector into the spatial-temporal fusion graph neural network prediction model so that the spatial-temporal fusion graph neural network prediction model can construct the temporal feature vector into a feature vector with dynamic graph nodes and edges, and use graph convolutional layers and temporal coding layers to jointly model the feature vector to complete the correlation prediction of multi-source spatiotemporal features in non-Euclidean road networks, and output the instantaneous vehicle power demand in the prediction time domain.

[0027] Specifically, the temporal feature vectors are input into the spatial-temporal fusion graph neural network prediction model so that the spatial-temporal fusion graph neural network prediction model can construct the temporal feature vectors (i.e., road network topology, traffic flow density, and vehicle driving status) into feature vectors with dynamic graph nodes and edges.

[0028] Following step B2, the following is also included: Step B3: When the interruption time of vehicle-to-everything (V2X) communication exceeds a preset threshold, switch the spatial-temporal fusion graph neural network prediction model to a lightweight graph neural network.

[0029] The lightweight graph neural network is a neural network that uses driving state parameters to predict power. The preset threshold can be 3 seconds, and there can be multiple preset thresholds; no specific limitation is made here.

[0030] Step B4: After the vehicle-to-everything (V2X) communication is restored, the latest road information ahead is reacquired, and the prediction error during the V2X communication interruption is used to update the state of the spatial-temporal fusion graph neural network prediction model to ensure the continuous operation of the prediction process under different communication conditions.

[0031] Specifically, when the data connection of the vehicle network communication interface is lost, the spatial-temporal fusion graph neural network switches to a lightweight graph neural network based only on driving state parameters for short-term prediction, and resynchronizes the cloud information after the communication is restored.

[0032] Specifically, with a control cycle of 5 seconds and a prediction time domain of the next 5 seconds, the vehicle controller retrieves road information ahead every 0.5 seconds via the LTE-V2X communication interface. It then combines local driving status and energy-saving driving index, inputs a spatial-temporal fusion graph neural network prediction model, and outputs instantaneous vehicle power demand with a resolution of 0.5 seconds.

[0033] The specific operation is as follows: at the beginning of the control cycle, the vehicle is located at the kilometer marker of National Highway L202 in the suburbs, and requests road information within a range of 2km ahead from the cloud through the vehicle network interface.

[0034] The road longitudinal slope curve data retrieved in the current control cycle are as follows: +1.2%, +0.8%, -0.5%, 0%, +0.7%, +1.5%, +2.0%, +1.0%, -1.0%, and 0.5%. The corresponding traffic flow densities for the road segments are: 30, 45, 50, 55, 60, 40, 35, 25, 20, and 30. The phases and remaining times of the first two traffic lights are as follows: the green light remains at the first traffic light for 15 seconds, and the red light remains at the first traffic light for 30 seconds; the green light remains at the second traffic light for 10 seconds, and the red light remains at the second traffic light for 25 seconds.

[0035] The road longitudinal slope curve, traffic flow density, traffic light phase, and vehicle speed, longitudinal acceleration, and energy-saving driving index collected at the end of the current control cycle are normalized and embedded locally. The processed data is then divided into 10 0.5s windows according to a preset time step, and these 10 0.5s windows are used as inputs to a spatial-temporal fusion graph neural network prediction model. The spatial-temporal fusion graph neural network prediction model constructs the 10 0.5s windows (i.e., the corresponding road network topology, traffic flow density, and vehicle driving status) into feature vectors of dynamic graph nodes and edges. Through joint modeling of graph convolution and temporal coding, it predicts the correlation of multi-source spatiotemporal features in the non-Euclidean road network to output the instantaneous vehicle power demand in the predicted time domain.

[0036] If V2X communication is lost for more than 3 seconds due to coverage blind spots or signal jitter, it automatically switches to the lightweight graph neural network mode. The lightweight graph neural network uses only vehicle speed, acceleration, and EDI features to predict instantaneous vehicle power demand within 5 seconds. At this time, the prediction resolution is 0.5 seconds, and the running latency is controlled within 10ms.

[0037] Once communication is restored, the controller retrieves the complete 2km road condition data from the cloud and compares the prediction error from the previous 5 seconds in real time. It then updates the state of the spatial-temporal fusion graph neural network prediction model to ensure seamless continuity of predictions.

[0038] In five consecutive tests, the average prediction error using integrated road condition information was ±3kW, while the prediction error using a local lightweight graph neural network was ±8kW. Switching delay control was kept within 50ms, ensuring timely power distribution and reducing hydrogen consumption by approximately 4.2%.

[0039] S103. With fuel cell stack efficiency and energy storage device cycle depth as objectives and instantaneous vehicle power demand as constraints, construct a multi-objective optimization model and solve the multi-objective optimization model to obtain the target output power of fuel cell stack and the target charge and discharge power of energy storage device.

[0040] The expression for the multi-objective optimization model is as follows: ; in, This is the energy consumption weighting coefficient for fuel cells. For the target output power of the fuel cell stack, The unit is kW. This is the weighting coefficient for energy storage cycle depth. The target charge and discharge power of the energy storage device, The unit is kW. For instantaneous vehicle power demand, The unit is kW. This refers to the state of charge of the fuel cell. Dimensionless This is the reference power for the fuel cell stack. The unit is kW. These are the constraints satisfied by the multi-objective optimization model.

[0041] The multi-objective optimization model minimizes the fuel cell output power term and the energy storage discharge depth term, solving for the optimal fuel cell power and energy storage charge / discharge power while satisfying the power balance relationship, thereby achieving efficient energy allocation and energy-saving management of the vehicle. The efficiency of the fuel cell stack is inversely proportional to its target output power, while the cycle depth of the energy storage device is negatively correlated with its target charge / discharge power.

[0042] For example, the specific operation is: instantaneous vehicle power demand. The target output power of the fuel cell stack is 120kW. The target charging and discharging power of the energy storage device is 80kW. The base power of the fuel cell stack is 40kW. The initial weighting coefficient is set to 100kW. , .

[0043] Substituting all the above data into the expression of the multi-objective optimization model, we get: ; The constraints are determined as follows: The target output power of the fuel cell stack in the constraints and the target charge and discharge power of energy storage devices The sum and instantaneous vehicle power demand They are identical and satisfy the constraints.

[0044] S104. Based on the target output power and target charge / discharge power, power commands are issued to the fuel cell stack and energy storage device respectively. When the deviation between the actual driving power and the predicted power demand exceeds the preset conditions, the energy storage device compensation mode is triggered to maintain power output.

[0045] The power command includes a compensation mode for energy storage devices.

[0046] Specifically, when the deviation between the actual driving power and the predicted power demand exceeds a preset condition, the energy storage device compensation mode is triggered to maintain power output, including: Step C1: When the deviation exceeds the preset deviation and the duration of the deviation exceeding the preset deviation reaches the preset duration, the energy storage device compensation mode is triggered.

[0047] Step C2: In the energy storage device compensation mode, control the energy storage device to increase the target charge and discharge power to compensate for the deviation, while keeping the target output power of the fuel cell stack unchanged.

[0048] Step C3: When the deviation is less than the target deviation, exit the energy storage device compensation mode to maintain power output.

[0049] Specifically, when the deviation between the actual driving power obtained from real-time monitoring and the predicted power demand exceeds 5kW for 2 consecutive seconds, the energy storage device compensation mode is triggered. In the energy storage device compensation mode, the energy storage device only bears the transient power gap until the deviation recovers to within 2kW.

[0050] The preset deviation can be 5kW, and there can be multiple preset deviations; no specific limitation is made here. The preset duration can be 2s, and there can be multiple preset durations; no specific limitation is made here. The target deviation can be 2kW, and there can be multiple target deviations; no specific limitation is made here.

[0051] Figure 5 A schematic diagram of the energy storage compensation mode is shown below. Figure 5 As shown, the specific operation is as follows: the energy management controller sends power commands to the fuel cell stack and energy storage device at a high frequency, and then calculates the actual driving power of the vehicle in real time through sensors such as bus current, voltage and wheel torque, and calculates the difference between the actual power and the predicted power demand. The controller continuously tracks this power deviation, and when the deviation exceeds 5kW continuously for 2 seconds, it automatically enters the energy storage device compensation mode.

[0052] In this compensation mode, the energy storage device will increase its target charge and discharge power to make up for the instantaneous power shortfall. During the compensation process, the power output of the fuel cell stack remains constant to prevent the fuel cell from operating in the inefficient region, thereby avoiding efficiency loss due to rapid power adjustment.

[0053] The collaborative working mode between the fuel cell stack and the energy storage device allows both to share the power demand, with the fuel cell stack not participating in additional compensation, maintaining its stable operation to ensure the long-term efficiency of the system.

[0054] For example, if the predicted power demand is 100kW, while the actual drive power calculated in real time is 105kW, the deviation is determined to be 5kW. At this point, the energy storage device can begin to provide an additional 5kW of power to the system to ensure that the vehicle's power output meets the actual demand. When the real-time monitored deviation falls back to within 2kW and remains stable for a short period thereafter, the compensation mode is automatically exited. During the exit process, the discharge power of the energy storage device smoothly transitions back to the normal allocation value, avoiding voltage fluctuations or vibrations perceived by the driver caused by a sudden drop in power.

[0055] Deviation is determined using a combination of a sliding window and a timer. Compensation is only triggered after conditions are met consecutively, preventing malfunctions due to instantaneous fluctuations. Short-time first-order hysteresis filtering is applied during compensation initiation and termination to ensure limited power change rate and prevent system chattering or component overload. In compensation mode, the SOC and temperature of the energy storage device are monitored. If an abnormality is detected, the device is immediately derated and switched to power-limiting protection mode to ensure the safety of the vehicle and its components.

[0056] Through the above steps S104, the instantaneous vehicle power demand can be quickly compensated when the fuel cell stack is not timely, ensuring the continuity of vehicle power. It can also avoid efficiency loss and thermal shock caused by frequent and large-scale fuel cell power adjustment, and achieve robust response to complex road conditions and sudden working conditions.

[0057] S105. Based on the changes in deviation and energy-saving driving index, update the weight coefficients of the multi-objective optimization model, convert the energy-saving driving index into corresponding energy-saving driving information, and feed it back to the driver through the human-machine interface to achieve real-time adjustment of energy management strategy and closed-loop guidance of driving behavior.

[0058] Specifically, the energy-saving driving index is converted into corresponding energy-saving driving information and fed back to the driver through a human-machine interface, including: Step D1: Divide the real-time value of the energy-saving driving index into multiple state intervals.

[0059] Step D2: On the human-machine interface, display the real-time value of the energy-saving driving index and its corresponding status range in the form of a segmented bar chart.

[0060] The display color varies for each status interval, and each status interval is used to provide drivers with energy-saving driving information prompts.

[0061] The human-computer interaction interface displays the energy-saving driving index (EDI) in real time in the form of a segmented bar chart. There are multiple state ranges, which can be three state ranges: 0–0.4, 0.4–0.8, and >0.8. The 0–0.4 state range is green, the 0.4–0.8 state range is yellow, and the >0.8 state range is red.

[0062] Following the example from step S101, the result of the energy-saving driving index is approximately 0.58. At this point, the driver's acceleration and deceleration fluctuations are determined to be at the "moderate energy-saving" level, and the subsequent power prediction and allocation strategy is adjusted accordingly. The EDI of 0.58, along with real-time driving status and road information ahead, is input into a spatial-temporal fusion graph neural network prediction model to output the instantaneous vehicle power demand in the predicted time domain, prioritizing the reduction of the probability of fuel cell inefficient operation.

[0063] In multi-objective optimization, the participation of energy storage devices should be appropriately increased for moderate energy-saving behavior in order to smooth the output power curve of fuel cells.

[0064] The dashboard and voice system prompt the driver with "The current energy efficiency level is moderate. It is recommended to maintain a constant speed or brake gently." The interface also displays the EDI changes in real time with a color bar chart, helping the driver to intuitively understand the effect of the operation.

[0065] The specific classification and feedback of EDI are as follows: The EDI value directly reflects the level of fuel efficiency in driving behavior. Generally, the range of EDI values ​​can be categorized as follows: EDI<0.4: Green indicates good fuel-efficient driving behavior, where the driver maintains smooth driving with minimal acceleration and deceleration.

[0066] 0.4≤EDI<0.8: Yellow indicates moderate fuel-efficient driving, with some fluctuations in the driver's performance during acceleration and braking.

[0067] EDI≥0.8: Red indicates poor fuel economy driving, with the driver frequently accelerating and braking, and driving in a relatively rough manner.

[0068] An EDI value of 0.58 falls within the yellow zone, indicating that the current driving behavior is considered "moderately fuel-efficient." The system can provide fuel-efficient driving suggestions based on the driver's actual behavior. For example, the system might advise the driver to "maintain a constant speed or avoid sudden braking."

[0069] Based on the aforementioned rich data processing and feedback mechanisms, hydrogen consumption of vehicles under the same road conditions can be reduced by 3%–5%, while filtering and calculation delays are controlled within 50 milliseconds, ensuring both calculation timeliness and improving the closed-loop synergistic effect of energy-saving driving of the whole vehicle.

[0070] Figure 6 An optimization diagram is shown schematically; see [link / reference]. Figure 6 As shown, following the example in step S103, the update of the weight coefficients of the multi-objective optimization model depends on the energy-saving driving index. EDI essentially reflects the driver's energy-saving behavior, and the weight coefficients are adjusted according to changes in EDI. The calculation formula for EDI is the same as the expression for the energy-saving driving index in step S101. The adjustment formulas for the fuel cell energy consumption weight coefficient and the energy storage cycle depth weight coefficient are set according to the EDI value as follows: When EDI is low, EDI < 0.4, it indicates that the driver's energy-saving driving behavior is good, and the load on the fuel cell stack should be appropriately reduced, while the participation of the energy storage device should be increased. When EDI is high, EDI ≥ 0.8, it indicates that the driver's energy-saving behavior is poor, and the system should increase the load on the fuel cell stack and reduce the participation of the energy storage device. Fuel Cell Energy Consumption Weight Coefficient And the weighting coefficient of energy storage cycle depth The adjustment formula is: ; ; in, It is an adjustment function based on EDI values, which dynamically calculates the weight ratio of fuel cell stacks and energy storage devices according to the EDI values.

[0071] Specific adjustment function For example: if EDI < 0.4, If 0.4 ≤ EDI < 0.8, If EDI ≥ 0.8, .

[0072] The adjusted weighting coefficients are , .

[0073] Substitute the new weighting coefficients into the expression of the multi-objective optimization model to recalculate the objective function: ; By adjusting the weighting coefficients and The objective function value increased from 50.5 to 57.875. This indicates that by increasing the fuel cell energy consumption weighting coefficient, the system tends to rely more on the fuel cell stack output power, reducing the involvement of energy storage devices and leading to changes in overall efficiency.

[0074] Example 2: This embodiment explains how to quantify driving behavior using the trajectory corresponding to speed and longitudinal acceleration.

[0075] Figure 7 A schematic diagram illustrating the calculation of the energy-saving driving index is shown below. Figure 7 As shown, the vehicle speed during the control cycle is obtained from the onboard system. and longitudinal acceleration data .

[0076] The collected vehicle speed and longitudinal acceleration data are filtered to remove noise and interference, ensuring the smoothness of the vehicle speed and longitudinal acceleration data.

[0077] For each sampling point Calculate the weighted product of longitudinal acceleration and vehicle speed, i.e.: ; Calculate the weighted average of all sampled points, and finally divide by . To standardize: ; The lower the EDI value, the more energy-efficient the driving behavior.

[0078] Classification based on EDI value: EDI < 0.4: Green, energy-saving driving.

[0079] 0.4≤EDI<0.8: Yellow, medium energy saving.

[0080] EDI≥0.8: Red, poor energy saving.

[0081] For example, within one control cycle, data from 5 sampling points are collected. Table 2 shows the collected data, including the sampling points. , No. Vehicle speed at each sampling point and the longitudinal acceleration at each sampling point .

[0082] Table 2 Data collected

[0083] gravitational acceleration The EDI is calculated based on the expression for the energy-saving driving index.

[0084] Calculate the weight of each sampling point: Collection point 1: ; Collection point 2: ; Collection point 3: ; Collection point 4: ; Collection point 5: ; The average value of the EDI data from collection points 1, 2, 3, 4, and 5 is calculated as follows: ; The calculated EDI is 0.724, indicating that the current driving behavior is of moderate energy efficiency.

[0085] Example 3: This embodiment explains how to update the weights of a multi-objective optimization model through real-time feedback to ensure that the mechanism can adapt to changes in driver behavior and road conditions.

[0086] At the end of each control cycle, the current EDI value is calculated and output to assess the driver's fuel efficiency level. EDI < 0.4 indicates good fuel efficiency driving; 0.4 ≤ EDI < 0.8 indicates moderate fuel efficiency driving; and EDI ≥ 0.8 indicates poor fuel efficiency driving behavior.

[0087] Fuel cell stack efficiency ensures that the fuel cell always operates within its high-efficiency range; the energy storage device's cycle depth prevents over-discharge of the energy storage device, extending battery life; and energy consumption ensures that the vehicle's power requirements are met, thereby reducing energy consumption. When a new EDI value is calculated, it is fed back into the multi-objective optimization model, adjusting the weights of different objectives.

[0088] The new EDI value is fed back to the multi-objective optimization model, adjusting its weights. For example, if the EDI value is high, indicating poor energy-saving behavior by drivers, the multi-objective optimization model will focus more on improving fuel cell efficiency and the participation of energy storage devices to achieve greater energy efficiency gains.

[0089] EDI-based weight adjustment: Weights are updated online using a simple linear regression model or a weighted adjustment method. The initial weights are as follows: ; When the EDI value is high, adjust the weights according to empirical rules, for example, change the weight ratio to: ; More resources should be allocated to the efficiency of fuel cell stacks and energy storage devices to address poor fuel economy driving behavior.

[0090] If EDI > 0.8, the weights will be adjusted first to give a larger target weight for fuel cell stack efficiency.

[0091] If the EDI is between 0.4 and 0.8, then the weights of the energy storage device and fuel cell stack efficiency should be adjusted accordingly.

[0092] If EDI < 0.4, the need for energy storage device participation is reduced, and more of the efficiency of the fuel cell stack is utilized.

[0093] At the end of each control cycle, vehicle driving status, energy-saving driving index, and other key factors are fed into a multi-objective optimization model. The actual energy consumption is compared with the target value, and the model's weights are adjusted for optimization.

[0094] Based on the above Figure 1As can be seen from the implementation method, the embodiments of the present invention determine the energy-saving driving index based on the real-time acquired driving state parameters; based on the energy-saving driving index, the road information ahead, the drive motor power, and the driving state parameters, a spatial-temporal fusion graph neural network prediction model is used to predict and output the instantaneous vehicle power demand in the prediction time domain; with the fuel cell stack efficiency and the energy storage device cycle depth as objectives and the instantaneous vehicle power demand as constraints, a multi-objective optimization model is constructed and solved to obtain the target output power of the fuel cell stack and the target charge and discharge power of the energy storage device; according to the target output power and the target charge and discharge power, power commands are issued to the fuel cell stack and the energy storage device respectively, and when the deviation between the real-time monitored actual drive power and the predicted power demand exceeds the preset condition, the energy storage device compensation mode is triggered to maintain power output; based on the deviation and the change of the energy-saving driving index, the weight coefficients of the multi-objective optimization model are updated, the energy-saving driving index is converted into the corresponding energy-saving driving information, and the information is fed back to the driver through the human-machine interface to realize the real-time adjustment of energy management strategy and closed-loop guidance of driving behavior. In this way, by introducing a graph neural network prediction model based on vehicle-to-everything (V2X) road information and spatial-temporal fusion, the system achieves forward-looking and accurate prediction of instantaneous vehicle power demand. A real-time multi-objective optimization model is constructed with fuel cell stack efficiency and energy storage cycle depth as its core, combined with online adaptive weight updates. This ensures that the fuel cell always operates within its high-efficiency range and the energy storage device maintains shallow cycle operation, significantly reducing hydrogen consumption and slowing down the degradation of the power battery and stack. Simultaneously, in the event of communication failure or sudden changes in operating conditions, the system automatically switches to short-term prediction and triggers the energy storage device's compensation mode, ensuring rapid response and power continuity in complex road conditions. In terms of driving experience and operation, the real-time visualization and voice prompts of the Energy-Saving Driving Index (EDI) form a "driver-energy management" closed loop, guiding drivers to optimize their operating habits to further tap into energy-saving potential. Segmented bar charts intuitively display driving behavior status, making it easy for drivers to understand and adjust in real time. Overall, this not only improves the overall vehicle energy economy and driving range but also reduces the system's implementation difficulty and maintenance costs, demonstrating good engineering feasibility and promotional value.

[0095] Based on the same inventive concept, as an implementation of the above-mentioned energy management method for fuel cell heavy-duty trucks that takes into account energy-saving driving, the present invention also provides an energy management system for fuel cell heavy-duty trucks that takes into account energy-saving driving. Figure 8 This is a structural diagram of the fuel cell heavy-duty truck energy management system considering energy-saving driving in an embodiment of the present invention. See also... Figure 8 As shown, the fuel cell heavy-duty truck energy management system that takes into account energy-saving driving may include: The determination module 801 is used to determine the energy-saving driving index based on the real-time acquired driving status parameters. The energy-saving driving index is used to indicate and quantify the driver's energy-saving behavior. The prediction module 802 is used to make predictions based on the energy-saving driving index, the road information ahead, the power of the drive motor and the driving state parameters, using a graph neural network prediction model that integrates spatial and temporal data, so as to output the instantaneous vehicle power demand in the prediction time domain. The construction and solution module 803 is used to construct a multi-objective optimization model with the fuel cell stack efficiency and energy storage device cycle depth as objectives and the instantaneous vehicle power demand as constraints, and solve the multi-objective optimization model to obtain the target output power of the fuel cell stack and the target charge and discharge power of the energy storage device. The trigger module 804 is used to issue power commands to the fuel cell stack and the energy storage device respectively according to the target output power and the target charge and discharge power, and to trigger the energy storage device compensation mode to maintain power output when the deviation between the actual driving power and the predicted power demand exceeds the preset conditions. The update and feedback module 805 is used to update the weight coefficients of the multi-objective optimization model based on the changes in deviation and energy-saving driving index, convert the energy-saving driving index into corresponding energy-saving driving information, and feed it back to the driver through the human-machine interface to realize real-time adjustment of energy management strategy and closed-loop guidance of driving behavior.

[0096] In module 801, the expression for the energy-saving driving index is: ; in, For energy-saving driving index, This represents the number of sampling points within the current control cycle. For the first The longitudinal acceleration of each sampling point For the first The vehicle speed at each sampling point Let be the acceleration due to gravity, and take . .

[0097] The device may further include: a real-time acquisition module, used to acquire the drive motor power in real time before predicting the instantaneous vehicle power demand in the prediction time domain based on the energy-saving driving index, road information ahead, drive motor power, and driving state parameters using a graph neural network prediction model that integrates spatial and temporal data; and a pull module, used to pull the road information ahead through the vehicle network communication interface in each control cycle, wherein the road information ahead is road information within a future preset driving distance for each control cycle.

[0098] The prediction module 802 is specifically used to normalize and embed the road longitudinal slope curve, traffic flow density, traffic light phase, vehicle speed, longitudinal acceleration, and energy-saving driving index sequentially. The processed data is then segmented according to a preset time step to construct corresponding temporal feature vectors. These temporal feature vectors are input into a spatial-temporal fusion graph neural network prediction model. This model constructs the temporal feature vectors as feature vectors with dynamic graph nodes and edges. A graph convolutional layer and a temporal coding layer are used to jointly model the feature vectors, enabling the prediction of the association between multi-source spatiotemporal features in the non-Euclidean road network. The model outputs the instantaneous vehicle power demand in the prediction time domain. The road information ahead includes the road longitudinal slope curve, traffic flow density, traffic light phase, and remaining time of the traffic lights.

[0099] In prediction module 802, the temporal feature vector is input into the spatial-temporal fusion graph neural network prediction model. This model constructs the temporal feature vector into a feature vector with dynamic graph nodes and edges. Graph convolutional layers and temporal coding layers are used to jointly model the feature vector to complete the correlation prediction of multi-source spatiotemporal features in the non-Euclidean road network. After outputting the instantaneous vehicle power demand in the prediction time domain, when the interruption time of vehicle network communication exceeds a preset threshold, the spatial-temporal fusion graph neural network prediction model is switched to a lightweight graph neural network. The lightweight graph neural network is a neural network that uses driving state parameters for power prediction. After vehicle network communication is restored, the latest road information ahead is reacquired, and the prediction error during the vehicle network communication interruption is used to update the state of the spatial-temporal fusion graph neural network prediction model to ensure continuous operation of the prediction process under different communication conditions.

[0100] In the construction and solution module 803, the expression for the multi-objective optimization model is: ; in, This is the energy consumption weighting coefficient for fuel cells. For the target output power of the fuel cell stack, This is the weighting coefficient for energy storage cycle depth. The target charge and discharge power of the energy storage device, For instantaneous vehicle power demand, This refers to the state of charge of the fuel cell. This is the reference power for the fuel cell stack.

[0101] In the trigger module 804, when the deviation between the actual driving power and the predicted power demand monitored in real time exceeds a preset condition, the energy storage device compensation mode is triggered to maintain power output. This includes: triggering the energy storage device compensation mode when the deviation exceeds a preset deviation and the duration of the deviation exceeding the preset deviation reaches a preset duration; in the energy storage device compensation mode, controlling the energy storage device to increase the target charging and discharging power to compensate for the deviation and keeping the target output power of the fuel cell stack unchanged; and exiting the energy storage device compensation mode when the deviation is less than the target deviation to maintain power output.

[0102] In the update and feedback module 805, the energy-saving driving index is converted into corresponding energy-saving driving information and fed back to the driver through the human-machine interface. This includes: dividing the real-time value of the energy-saving driving index into multiple state intervals; displaying the real-time value of the energy-saving driving index and its corresponding state interval in the form of a segmented bar chart on the human-machine interface, with different display colors for each state interval, and each state interval is used to provide energy-saving driving information prompts to the driver.

[0103] It should be noted that the above description of the embodiment of the fuel cell heavy-duty truck energy management system considering energy-saving driving is similar to the description of the above embodiment of the fuel cell heavy-duty truck energy management method considering energy-saving driving, and has similar beneficial effects. For technical details not disclosed in the embodiments of the fuel cell heavy-duty truck energy management system considering energy-saving driving of the present invention, please refer to the description of the embodiment of the fuel cell heavy-duty truck energy management method considering energy-saving driving of the present invention for understanding.

[0104] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A fuel cell heavy truck energy management method considering energy-saving driving, characterized in that, include: The energy-saving driving index is determined based on real-time driving status parameters, and the energy-saving driving index is used to indicate and quantify the driver's energy-saving behavior. Based on the energy-saving driving index, road information ahead, drive motor power and driving status parameters, a spatial-temporal fusion graph neural network prediction model is used to predict and output the instantaneous vehicle power demand in the prediction time domain. With fuel cell stack efficiency and energy storage device cycle depth as objectives and instantaneous vehicle power demand as constraints, a multi-objective optimization model is constructed and solved to obtain the target output power of the fuel cell stack and the target charge and discharge power of the energy storage device. Based on the target output power and the target charge / discharge power, power commands are issued to the fuel cell stack and the energy storage device respectively. When the deviation between the actual driving power and the predicted power demand exceeds a preset condition, the energy storage device compensation mode is triggered to maintain power output. Based on the deviation and the changes in the energy-saving driving index, the weight coefficients of the multi-objective optimization model are updated, the energy-saving driving index is converted into corresponding energy-saving driving information, and the information is fed back to the driver through the human-machine interface to achieve real-time adjustment of energy management strategies and closed-loop guidance of driving behavior.

2. The energy management method for fuel cell heavy duty truck considering energy saving driving according to claim 1, characterized in that, The driving status parameters include vehicle speed, longitudinal acceleration, road gradient, and total vehicle mass.

3. The energy management method for fuel cell heavy duty truck considering energy saving driving according to claim 2, characterized in that, The expression for the energy-saving driving index is: ； in, For energy-saving driving index, This represents the number of sampling points within the current control cycle. For the first The longitudinal acceleration of each sampling point For the first The vehicle speed at each sampling point Let be the acceleration due to gravity, and take . .

4. The energy management method for fuel cell heavy-duty trucks considering energy-saving driving according to claim 1, characterized in that, Before predicting the instantaneous vehicle power demand in the predicted time domain using a spatial-temporal fusion graph neural network prediction model based on the energy-saving driving index, road information ahead, drive motor power, and driving state parameters, the method further includes: The power of the drive motor is obtained in real time; During each control cycle, the road information ahead is retrieved through the vehicle network communication interface. The road information ahead is the road information within the future preset driving distance of each control cycle.

5. The energy management method for fuel cell heavy duty truck considering energy saving driving according to claim 2, characterized in that, The road information ahead includes road longitudinal slope curve, traffic flow density, traffic light phase, and remaining time of the traffic lights; the prediction based on the energy-saving driving index, road information ahead, drive motor power, and driving state parameters, using a spatial-temporal fusion graph neural network prediction model, outputs the instantaneous vehicle power demand in the predicted time domain, including: The road longitudinal slope curve, traffic flow density, traffic light phase, vehicle speed, longitudinal acceleration, and energy-saving driving index are sequentially normalized and embedded in the encoding process. The processed data are then segmented according to a preset time step to construct the corresponding temporal feature vector. The temporal feature vector is input into the spatial-temporal fusion graph neural network prediction model, so that the spatial-temporal fusion graph neural network prediction model constructs the temporal feature vector into a feature vector with dynamic graph nodes and edges, and uses graph convolutional layers and temporal coding layers to jointly model the feature vector to complete the correlation prediction of multi-source spatiotemporal features in non-Euclidean road networks, and outputs the instantaneous vehicle power demand in the prediction time domain.

6. The energy management method for fuel cell heavy duty truck considering energy saving driving according to claim 5, characterized in that, After inputting the temporal feature vector into the spatial-temporal fusion graph neural network prediction model, so that the spatial-temporal fusion graph neural network prediction model constructs the temporal feature vector into a feature vector with dynamic graph nodes and edges, and uses graph convolutional layers and temporal coding layers to jointly model the feature vector to complete the correlation prediction of multi-source spatiotemporal features in non-Euclidean road networks, and outputs the instantaneous vehicle power demand in the prediction time domain, the method further includes: When the interruption time of vehicle network communication is detected to exceed a preset threshold, the spatial-temporal fusion graph neural network prediction model is switched to a lightweight graph neural network, which is a neural network that uses the driving state parameters to predict power. After the vehicle network communication is restored, the latest road information ahead is reacquired, and the state of the spatial-temporal fusion graph neural network prediction model is updated using the prediction error during the vehicle network communication interruption, so as to ensure the continuous operation of the prediction process under different communication conditions.

7. The energy management method for fuel cell heavy duty truck considering energy saving driving according to claim 1, characterized in that, The expression for the multi-objective optimization model is: ； in, This is the energy consumption weighting coefficient for fuel cells. The target output power of the fuel cell stack, This is the weighting coefficient for energy storage cycle depth. The target charge / discharge power of the energy storage device, For the instantaneous vehicle power demand, This refers to the state of charge of the fuel cell. This is the reference power for the fuel cell stack. 8.The energy management method for fuel cell heavy duty truck considering energy saving driving according to claim 1, wherein, When the deviation between the actual driving power and the predicted power demand exceeds a preset condition, the energy storage device compensation mode is triggered to maintain power output, including: When the deviation exceeds a preset deviation and the duration of the deviation exceeding the preset deviation reaches a preset duration, the energy storage device compensation mode is triggered. In the energy storage device compensation mode, the energy storage device is controlled to increase the target charging and discharging power to compensate for the deviation, while keeping the target output power of the fuel cell stack unchanged; When the deviation is less than the target deviation, the energy storage device compensation mode is exited to maintain power output. 9.The fuel cell heavy truck energy management method considering energy-saving driving of claim 1, wherein, The step of converting the energy-saving driving index into corresponding energy-saving driving information and feeding it back to the driver through a human-machine interface includes: The real-time values ​​of the energy-saving driving index are divided into multiple state intervals; The real-time value of the energy-saving driving index and its corresponding state interval are displayed in the form of a segmented bar chart on the human-machine interface. Each state interval is displayed in a different color and is used to provide the driver with energy-saving driving information prompts.

10. A fuel cell heavy duty truck energy management system that accounts for energy efficient driving, characterized in that, include: The determination module is used to determine the energy-saving driving index based on the driving status parameters acquired in real time. The energy-saving driving index is used to indicate and quantify the driver's energy-saving behavior. The prediction module is used to make predictions based on the energy-saving driving index, the road information ahead, the power of the drive motor and the driving state parameters, using a spatial-temporal fusion graph neural network prediction model, so as to output the instantaneous vehicle power demand in the prediction time domain. The construction and solution module is used to construct a multi-objective optimization model with the fuel cell stack efficiency and energy storage device cycle depth as objectives and the instantaneous vehicle power demand as constraints, and solve the multi-objective optimization model to obtain the target output power of the fuel cell stack and the target charge and discharge power of the energy storage device. The triggering module is used to issue power commands to the fuel cell stack and the energy storage device respectively according to the target output power and the target charge and discharge power, and to trigger the energy storage device compensation mode to maintain power output when the deviation between the actual driving power and the predicted power demand exceeds the preset conditions. The update and feedback module is used to update the weight coefficients of the multi-objective optimization model based on the deviation and the changes in the energy-saving driving index, convert the energy-saving driving index into corresponding energy-saving driving information, and feed it back to the driver through the human-machine interface, so as to realize the real-time adjustment of energy management strategy and closed-loop guidance of driving behavior.

Images