Self-adaptive switching method for multi-mode power system of hybrid commercial vehicle

Abstract

The invention discloses a self-adaptive switching method for a multi-mode power system of a hybrid commercial vehicle, and relates to the technical field of automobile power control, and the method comprises the specific steps: collecting multi-dimensional information, obtaining various types of information through a sensor and the like, and processing the information to construct a data set; constructing a five-dimensional model, dividing a five-dimensional parameter quantization interval, and calculating an adaptation degree to generate a candidate set; performing dynamic weight gaming and optimal mode screening, and constructing a three-dimensional objective function to screen an optimal mode; predictive pre-switching adjustment: constructing a model to pre-judge and pre-adjust a power source; dynamically compensating and adjusting the loss of the parts, and monitoring and compensating excessive loss in real time; according to the invention, through a multi-dimensional information acquisition and five-dimensional parameter coupling adaptation algorithm, the traditional limitation is broken through, and the adaptation precision is improved; a prediction model is constructed to pre-judge working condition changes, the response speed is optimized, and a power source is adjusted in advance; partial loss is monitored in real time, a compensation mechanism is triggered, collaborative optimization of performance and reliability is achieved, and guarantee is provided for efficient operation of the hybrid power commercial vehicle.

Description

Technical Field

[0001] This invention relates to the field of automotive power control technology, specifically to an adaptive switching method for a multi-mode power system in a hybrid commercial vehicle. Background Technology

[0002] With the increasing demands for energy consumption control and operational efficiency in the transportation industry, hybrid commercial vehicles, combining the advantages of traditional fuel power and electric drive, have become an important direction for meeting industry needs. These commercial vehicles operate under significantly complex conditions, facing not only frequent load fluctuations and varying road conditions, but also needing to adapt to diverse task modes. Different tasks have significantly different power output requirements. Currently, the industry's performance requirements for hybrid commercial vehicle power systems are no longer limited to basic drive functions, but emphasize the synergy of maximizing energy efficiency, minimizing component losses, and optimizing response speed under different operating conditions. However, in actual commercial vehicle operation, the dynamic changes in operating status and task attributes make it difficult for traditional power mode switching methods to accurately adapt. There is an urgent need for an adaptive switching technology that can integrate multi-dimensional information and achieve dynamic optimization to improve the overall adaptability and operational performance of the power system and meet the operational needs under complex conditions.

[0003] Traditional hybrid commercial vehicle power switching methods have several limitations. First, the information collection dimensions are relatively limited, often focusing only on the power source's own parameters or some operating condition indicators, failing to comprehensively integrate key information such as load, road conditions, component wear, and task modes. This results in an insufficient perception of the system's operating status and task requirements, leading to a lack of sufficient data support for power mode adaptation decisions. Second, the decision-making process often employs fixed-weight or single-objective optimization strategies, making it difficult to dynamically balance the relationship between energy efficiency, wear, and response speed according to actual operating condition changes. This can easily lead to an emphasis on one performance indicator while neglecting others, failing to achieve multi-objective collaborative optimization. Furthermore, the lack of predictive ability for future operating conditions means that the system can only passively respond to changes in current conditions, resulting in power switching lagging behind actual needs and prone to problems with smooth power transitions. At the same time, the compensation mechanism for component wear is not perfect, failing to adjust the operating mode in a timely manner based on task modes and wear status, easily causing excessive component wear and affecting the long-term stability and service life of the system.

[0004] Based on this, in order to solve the above technical problems, this paper proposes an adaptive switching method for a multi-mode power system of hybrid commercial vehicles. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an adaptive switching method for multi-mode power systems in hybrid commercial vehicles. First, multi-dimensional information on load, road surface, power source, component wear, and task mode is collected through sensors and interfaces, and a dataset is constructed after preprocessing. Then, based on the dataset, a five-dimensional parameter quantization interval is divided, and a power mode candidate set is generated using a coupling adaptation algorithm. Subsequently, a three-dimensional objective function is constructed, and the optimal power mode is selected through dynamic weighted game theory. Next, a prediction model is used to predict future operating conditions, and a response correction algorithm is combined to pre-start, pre-shut down, or pre-adjust parameters of the power source. Finally, component wear is monitored in real time, triggering a compensation mechanism to adjust the operating mode.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an adaptive switching method for a multi-mode power system of a hybrid commercial vehicle, the specific steps of which are as follows:

[0007] S100, Multi-dimensional Information Acquisition: Acquires information on load, road surface, power source, component wear and task mode through sensors and interfaces, and performs preprocessing and feature extraction to construct a dataset containing operating status and task attributes;

[0008] S200, Five-dimensional model construction: Based on the dataset, the five-dimensional parameter quantization intervals, including task mode, total load, road surface features, power source efficiency and component loss, are divided. The fit degree between the five-dimensional parameters and the power mode is calculated by the five-dimensional parameter coupling and adaptation algorithm to generate a power mode candidate set.

[0009] The power modes include engine-only drive mode, front motor-only drive mode, rear motor-only drive mode, engine direct drive + motor assist mode, and dual-motor combined drive mode.

[0010] S300, Dynamic Weight Game and Optimal Mode Selection: Construct a three-dimensional objective function that includes energy efficiency, loss, and response speed. Calculate the candidate power mode based on the objective function and select the optimal power mode with the highest objective function value.

[0011] S400, Predictive Pre-Switching Adjustment: Utilizes historical operational data and task attributes to construct a predictive model, predicts future load changes, road surface characteristics, and task demand fluctuations on road sections, calculates response speed correction values ​​using a response correction algorithm, and pre-starts, pre-shuts down, or pre-adjusts parameters of the power source based on the prediction results and the optimal power mode.

[0012] The pre-adjustment specifically includes: if it is necessary to switch to engine direct drive + electric motor auxiliary mode, the engine speed is pre-adjusted from the current speed to 1800±200rpm at an adjustment rate of 500rpm / s, and the motor is pre-excited to 30% of the rated current; if it is necessary to switch to front motor independent drive mode, the engine fuel is cut off 0.3 seconds in advance, and the motor controller is pre-loaded with 20% of the target torque.

[0013] The step of pre-starting, pre-shutting down, or pre-adjusting parameters of the power source based on the prediction result and the optimal power mode specifically includes:

[0014] If it is predicted that future changes in road load, road surface characteristics, or fluctuations in task requirements will necessitate the addition of a power source to drive the road and match the optimal power mode, a pre-start operation will be performed on the new power source to bring it into the ready-to-run state in advance.

[0015] If it is predicted that future changes in road load, road surface characteristics, or fluctuations in task requirements will not require the continuous operation of a specific power source, and the power source is not required for the optimal power mode operation, a pre-shutdown operation will be performed on the specific power source.

[0016] If it is predicted that the load changes, road surface characteristics or task demand fluctuations of the road section in the future need to be adjusted to adapt to the optimal power mode, the operating parameters of the power source are pre-adjusted to ensure that it can meet the operating requirements of the optimal power mode when the working conditions change.

[0017] S500, dynamic compensation and adjustment of component wear: During operation in the optimal power mode, the component wear parameters are monitored in real time and the compensation mechanism is triggered in combination with the current task mode; when the component wear exceeds the standard, the power distribution ratio or operation mode is dynamically adjusted until the wear parameters return to the safe range.

[0018] Furthermore, in S100, during the multi-dimensional information acquisition, four suspension pressure sensors deployed on each of the front and rear axles of the cargo box are used to collect real-time suspension pressure data to calculate the total load value and center of gravity offset coefficient; a 16-line vehicle-mounted LiDAR is fused with a high-precision navigation map to generate real-time road slope, friction coefficient, and pothole level; the engine ECU obtains engine speed and load rate in real-time, and the dual-motor controller obtains torque and speed of the front and rear motors in real-time; the loss monitoring module calculates engine cylinder wear, battery capacity decay rate, and cumulative temperature rise of motor windings in real-time; the user interface of the vehicle-mounted touch screen and mobile APP obtains the manually selected task mode, and the logistics order system API interface automatically identifies task attributes such as delivery time requirements and cargo type in the order.

[0019] Furthermore, in the S200 five-dimensional model construction, the specific content of the five-dimensional parameter quantification range is as follows: the quantification range of the task mode is divided into extreme fuel-saving mode, standard balanced mode, time-limited delivery mode, and component maintenance mode; the quantification range of the total load is divided into no-load, medium-load, and full-load; the quantification range of road surface characteristics is divided into: slope, friction coefficient, and pothole level; the quantification range of power source efficiency is divided into: engine high-efficiency zone and motor high-efficiency zone; and the quantification range of component wear is divided into: normal, critical, and exceeding standard.

[0020] Furthermore, in S200, during the construction of the five-dimensional model, the expression for the five-dimensional parameter coupling and adaptation algorithm is: ,in, Let be the fitness level of the k-th power mode, with a value ranging from 0 to 1. Quantify the factors for the task mode. This is the total load quantization factor. For road surface feature coupling factor, The power source efficiency factor. Quantification factor for component loss. Represents the dynamic weights of the task mode dimension. Dynamic weights for the total load dimension. For the component loss dimension, dynamic weights, and ,filter The dynamic patterns form a candidate set.

[0021] Furthermore, in S300, the expression for the three-dimensional objective function in the dynamic weighted game and optimal mode selection is: ,in, Represents the objective function value. Indicates total energy efficiency. Indicates total loss. Indicates the overall response speed. Indicates energy efficiency weight. Indicates the loss weight. Response speed weight, and .

[0022] Furthermore, in S300, the energy efficiency weight is used in the dynamic weight game and optimal mode selection. The calculation formula is: ,in, This represents the task's energy efficiency sensitivity coefficient. Represents total loss; energy efficiency weight The calculation formula is: ,in, Task loss sensitivity coefficient; response speed weight. The calculation formula is: ,in, This represents the task response coefficient.

[0023] Furthermore, in S400, the predictive pre-switching adjustment uses an LSTM neural network model based on temporal feature fusion, specifically including:

[0024] Input layer parameters: Historical operation data includes load change curves for the same route over the past 3 months, road feature library, and power mode switching history; Current task attributes include current location, mileage traveled, navigation planning data for the remaining route, task mode, and real-time traffic information;

[0025] Feature processing layer: Normalizes the input parameters, assigns weights to different features through an attention mechanism, and extracts temporal features;

[0026] Network structure: It contains 3 layers of LSTM units, each followed by a Dropout layer to prevent overfitting, and the output layer is a fully connected layer;

[0027] Output parameters: load change trend, road surface feature changes, and task requirement fluctuations for the next 500 meters of road section;

[0028] Training process: The sliding window method is adopted, and the parameter adjustment method with adaptive learning rate is used. The initial learning rate is 0.001. The model parameters are continuously adjusted to reduce the mean squared difference between the predicted value and the actual value to a minimum. Training is stopped when the mean squared difference of the validation set is ≤0.02. The model is incrementally updated every 7 days based on the newly added operational data.

[0029] Furthermore, in S400, during the predictive pre-switching adjustment, the training steps of the prediction model are as follows:

[0030] Data preparation: Collect historical and new operational data and integrate them into a training dataset. The historical operational data includes load changes along the same route, road surface feature measurements, and power mode switching records. Clean the data, remove obvious outliers, and retain valid data sets.

[0031] Sliding window partitioning: The training dataset is partitioned using a sliding window method, with a set window size and sliding step; data is selected from each window according to a set ratio as the training set, and the remaining data is used as the validation set;

[0032] Model structure initialization: Set the network to contain multiple layers of time series processing units, with each layer containing a set number of processing nodes; set a random data discard mechanism after each layer and set the discard ratio; set the output nodes for the corresponding prediction parameters in the output layer;

[0033] Iterative training: An adaptive learning rate optimization method is adopted, and an initial learning rate is set. In each iteration, the training set data is input into the model to calculate the prediction results, and the network parameters are adjusted by comparing them with the actual values. The prediction bias is evaluated periodically using validation set data.

[0034] Stop training: When the validation set bias continues to meet the set conditions, stop the current training and save the network parameters at this time as the current version of the model;

[0035] Incremental Updates: Regularly add new operational data to the training dataset, repeat incremental training based on the current model version, and update the network parameters.

[0036] Furthermore, in S400, the expression for the response correction algorithm in the predictive pre-switching adjustment is: ,in, This indicates the corrected total response speed. Indicates the delay time for switching power modes. This represents the prediction deviation coefficient. It is the response correction factor.

[0037] Furthermore, in the S500, the compensation mechanism for dynamic compensation adjustment of component losses is specifically as follows:

[0038] In component maintenance mode: when the wear of the engine cylinder reaches 30%, start alternating load: use the front motor for 3 consecutive starts, and the rear motor for the 4th start, with a switching interval of ≥5 seconds; when the battery capacity decreases by 10%, start equalization charging when parking at night.

[0039] In time-limited delivery mode: when the motor winding temperature rises to 60℃, the torque of the two motors is evenly distributed, and the total torque decrease is ≤10%;

[0040] In standard balance mode: when component wear reaches the critical value, the engine load rate is limited to ≤80% and the motor torque is limited to ≤200N・m.

[0041] Compared with existing technologies, this adaptive switching method for a multi-mode powertrain system in a hybrid commercial vehicle has the following advantages:

[0042] 1. This invention integrates load, road surface, power source, component loss, and task mode information through a multi-dimensional information collection mechanism to construct a dataset containing operating status and task attributes, providing a comprehensive perception foundation for power mode adaptation. It combines a five-dimensional model to construct quantification intervals for key parameters and uses a five-dimensional parameter coupling adaptation algorithm to calculate the adaptation degree and generate a candidate set, overcoming the limitations of single-parameter decision-making. Based on a three-dimensional objective function, it conducts dynamic weighted game theory, comprehensively considering energy efficiency, loss, and response speed to select the optimal mode, ensuring that power output accurately matches current operating conditions while balancing multi-dimensional performance goals. This process effectively solves the problems of poor adaptability and one-sided decision-making in traditional power switching methods, improving the overall coordination and adaptation accuracy of system operation, providing scientific support for efficient operation in different task scenarios, and helping hybrid commercial vehicles achieve performance optimization in complex operating conditions.

[0043] 2. This invention constructs a predictive model based on time-series feature fusion, combining historical operational data with current task attributes to predict future changes in load, road surface, and demand on road segments. It also optimizes response speed using a response correction algorithm, pre-starting, pre-shutting down, or adjusting parameters of the power source in advance. This avoids power transition lag during sudden changes in operating conditions, significantly improving switching smoothness and system response efficiency. Simultaneously, it monitors component wear in real time during optimal operation, triggering targeted compensation mechanisms based on different task modes to dynamically adjust power distribution or operating modes, bringing component wear back to a safe range. This approach ensures long-term system stability while reducing maintenance costs and shortened lifespan due to excessive component wear, achieving synergistic optimization of performance and reliability. This provides strong support for the long-term, economical operation of hybrid commercial vehicles.

[0044] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention.

[0045] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0047] Figure 1 A flowchart illustrating the steps of an adaptive switching method for multi-mode powertrain systems in hybrid commercial vehicles.

[0048] Figure 2 A flowchart showing the input, output, and auxiliary input relationships for each step of the adaptive switching of the multi-mode power system in a hybrid commercial vehicle. Detailed Implementation

[0049] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following description, in conjunction with the accompanying drawings and preferred embodiments, further explains the specific implementation methods, structures, features, and effects of the present invention.

[0050] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0051] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0052] Example 1:

[0053] Urban logistics time-limited delivery scenario.

[0054] S100, Multi-dimensional Information Acquisition:

[0055] By deploying four suspension pressure sensors on each of the front and rear axles of the cargo box, real-time suspension pressure data is collected to calculate the total load value under full load conditions. Simultaneously, the center of gravity offset coefficient is obtained. This data accurately reflects the vehicle's current load status, providing a precise basis for quantifying the total load dimension in the subsequent five-dimensional model. By fusing a 16-line vehicle-mounted LiDAR with a high-precision navigation map, the road surface slope, friction coefficient, and pothole level of the current road segment are generated in real time. This fusion method provides real-time and comprehensive understanding of road conditions, ensuring consistency between subsequent road feature dimension analysis and actual road conditions. The engine ECU obtains the current engine speed and load rate in real time, while the dual-motor controller obtains the torque and speed of the front and rear motors in real time. This power source operating data directly supports the subsequent judgment of the power source efficiency dimension, clarifying the current working condition of the power source. The system monitors the operational status; it calculates engine cylinder wear, battery capacity degradation rate, and cumulative motor winding temperature rise in real time through a loss monitoring module. This module tracks component wear in real time, providing data support for subsequent quantification of component wear dimensions and triggering of compensation mechanisms. The system automatically identifies the delivery time requirements and goods type of the current order through the logistics order system API interface. Since the vehicle touchscreen does not receive manual instructions from the user to adjust the task mode, it defaults to the time-limited delivery mode identified by the order. By combining the API interface with the user interface to obtain the task mode and attributes, it ensures that task requirements are consistent with actual operational goals and avoids deviations in task attribute judgment. Based on all the above operational status data and task attribute data, a dataset containing operational status and task attributes is constructed, laying a complete and reliable data foundation for the subsequent construction of the five-dimensional model. Figure 1 As stated above.

[0056] S200, Five-Dimensional Model Construction:

[0057] Based on the dataset constructed using S100, five-dimensional parameter quantization intervals were defined: the task mode was "time-limited delivery mode," the total load was "full load," the road surface characteristics were "slope ≥ 5°, medium friction coefficient, and slight potholes," the power source efficiency was "inefficient zone" for the engine and "efficient zone" for the motor, and component wear was "normal." This clear quantization interval division allows for more explicit definition of the characteristics of each parameter dimension, providing a unified and controllable analytical standard for subsequent parameter coupling calculations. Subsequently, a five-dimensional parameter coupling adaptation algorithm was used to calculate the compatibility of these five-dimensional parameters with all power modes. The expression for the five-dimensional parameter coupling adaptation algorithm is as follows: ,in, Let be the fitness level of the k-th power mode, with a value ranging from 0 to 1. Quantify the factors for the task mode. This is the total load quantization factor. For road surface feature coupling factor, The power source efficiency factor. Quantification factor for component loss. Represents the dynamic weights of the task mode dimension. Dynamic weights for the total load dimension. For the component loss dimension, dynamic weights, and ,filter The algorithm forms a candidate set of power modes. It can fully consider the impact of various parameters on the power mode and achieve accurate matching between parameters and modes. Power modes with a fit of ≥0.7 are selected, and the final power mode candidate set is generated, including "engine direct drive + motor auxiliary mode" and "dual motor joint drive mode". This screening process can eliminate modes with insufficient fit, narrow down the scope for the next step of optimal mode screening, and ensure that the modes in the candidate set can initially meet the current operation and task requirements.

[0058] S300, Dynamic Weighted Game Theory and Optimal Pattern Selection:

[0059] Construct a three-dimensional objective function that includes energy efficiency, energy loss, and response speed. The expression of the three-dimensional objective function is as follows: ,in, Represents the objective function value. Indicates total energy efficiency. Indicates total loss. Indicates the overall response speed. Indicates energy efficiency weight. Indicates the loss weight. Response speed weight, and This objective function can comprehensively evaluate the performance of the power mode in multiple dimensions, avoiding mode selection bias caused by judging from a single dimension. For the current "time-limited delivery mode," priority must be given to ensuring response speed to meet delivery timeliness. Therefore, the weight proportions of each weight in the three-dimensional objective function are determined through energy efficiency weight calculation formulas, loss weight calculation formulas, and response speed weight calculation formulas. The energy efficiency weight... The calculation formula is: ,in, This represents the task's energy efficiency sensitivity coefficient. Represents total loss; energy efficiency weight The calculation formula is: ,in, Task loss sensitivity coefficient; response speed weight. The calculation formula is: ,in, The task response coefficient is weighted as follows: response speed has the highest weight, followed by energy efficiency, and loss has the lowest weight. This weighting distribution allows the objective function calculation results to better align with the core requirement of time-limited delivery. Based on this three-dimensional objective function, the candidate power mode generated by the S200 is calculated. The objective function values ​​of "engine direct drive + motor auxiliary mode" and "dual motor joint drive mode" are compared. Finally, the "engine direct drive + motor auxiliary mode" with the higher objective function value is selected as the optimal power mode. This selection result ensures that the selected mode meets the requirements of high response speed while taking into account energy efficiency and component loss control, which meets the comprehensive operational needs of time-limited delivery scenarios.

[0060] S400, Predictive Pre-Switch Adjustment:

[0061] A predictive model based on temporal feature fusion is used for forecasting. This model can fully explore the temporal correlation between historical and current data, improving the accuracy of predicting future road conditions. Input parameters include historical operational data and current task attributes, providing comprehensive support for the model's predictions. The model, after calculation through a feature processing layer, three LSTM units, and an output layer, outputs a predictive result for the next 500 meters of road: the load remains at full capacity, the road surface slope remains ≥5°, and the task requirement remains time-limited delivery. This predictive result allows for advance understanding of changes in road conditions and demand, providing direction for pre-adjustment of the power source. Combined with a response correction algorithm, the required delivery time is determined. The power source is pre-adjusted to match the "engine direct drive + motor assist mode". The response correction algorithm can optimize the response speed evaluation results and avoid untimely adjustment caused by prediction deviation. According to the pre-adjustment rules, the engine is pre-adjusted from the current 1200rpm to 1800±200rpm at an adjustment rate of 500rpm / s. At the same time, the motor is pre-excited to 30% of the rated current. This pre-adjustment allows the engine and motor to enter the optimal working state in advance, avoiding power switching delay when entering the climbing section. It ensures that the vehicle can directly operate in the optimal power mode during the climbing process, ensuring that the timeliness of the time-limited delivery is not affected.

[0062] S500, Dynamic Compensation Adjustment for Component Losses:

[0063] During operation in "engine direct drive + motor assist mode", component wear parameters are monitored in real time. Real-time monitoring can promptly detect abnormal component wear and prevent damage caused by excessive wear. When the vehicle reaches the middle of the uphill section, the wear monitoring module shows that the cumulative temperature rise of the motor winding reaches 55℃, and the compensation mechanism is not triggered at this time. Maintaining the current mode ensures stable power output and meets the climbing requirements. After driving for about 300 meters, the temperature rise of the motor winding reaches 60℃, triggering the compensation mechanism. Timely triggering of the compensation mechanism can prevent the motor temperature rise from rising further. According to the requirements of the compensation mechanism, the power distribution ratio is adjusted to implement equal torque distribution between the two motors, and the total torque decrease is ≤10%. This adjustment method can control the motor temperature rise while minimizing power output loss, ensuring that the vehicle can still maintain the power required for climbing. Until the temperature rise of the motor winding falls back to the safe range, the original power distribution ratio is restored. The restoration operation can return to the optimal power output state under the premise of component safety, and continuously meet the operational requirements of time-limited delivery.

[0064] In summary, in urban logistics time-limited delivery scenarios, multi-dimensional information collection and integration of load, road surface, power source, component wear and tear, and task attribute data provide comprehensive support for subsequent analysis. A model is constructed based on a five-dimensional parameter coupling adaptation algorithm, and a candidate set of adaptations is selected to narrow down the range of optimal modes. By utilizing a three-dimensional objective function and dynamic weight allocation, priority is given to ensuring response speed, and the engine direct drive + electric motor auxiliary mode is determined to be optimal. Road conditions are predicted using an LSTM neural network, and the power source is pre-adjusted using a response correction algorithm to avoid switching delays. The expression for the response correction algorithm is: ,in, This indicates the corrected total response speed. Indicates the delay time for switching power modes. This represents the prediction deviation coefficient. It is a response correction coefficient; it monitors component wear in real time and triggers a compensation mechanism to control motor temperature rise. The whole process fits the time-limited delivery requirements, ensuring timeliness while taking into account energy efficiency and component protection.

[0065] Example 2:

[0066] Intercity logistics routine transportation scenarios.

[0067] S100, Multi-dimensional Information Acquisition:

[0068] By deploying four suspension pressure sensors on each of the front and rear axles of the cargo box, real-time suspension pressure data is collected to calculate the total load value under medium load conditions. Simultaneously, the center of gravity offset coefficient is obtained. This data accurately captures the vehicle's current load strength, providing a true and reliable foundation for the quantitative analysis of the total load dimension in the subsequent five-dimensional model, avoiding the impact of load judgment deviations on power mode matching. Utilizing a 16-line vehicle-mounted LiDAR and high-precision navigation map fusion, the road surface slope, friction coefficient, and pothole level of the current intercity road section are generated in real time. This fusion technology can recreate the actual driving conditions on the road surface in real time, ensuring that the subsequent quantitative results of road surface characteristics are highly consistent with actual road conditions, providing accurate road surface basis for power mode adaptation. The engine ECU obtains the current engine speed and load rate in real time, and the dual-motor controller obtains the torque and speed of the front motor and the rear motor in real time. This real-time operating data of the power source directly reflects the working status of the engine and motor, providing core support for the subsequent judgment of the power source efficiency dimension. The system ensures the power source is operating within its high-efficiency range. A loss monitoring module calculates engine cylinder wear, battery capacity degradation rate, and cumulative motor winding temperature rise in real time. This module tracks the wear trends of key components, providing timely and accurate data for quantifying component wear dimensions and triggering compensation mechanisms in the subsequent five-dimensional model. Users manually select "Extreme Fuel Efficiency Mode" as the task mode via the vehicle touchscreen. Simultaneously, the logistics order system API automatically identifies the delivery time requirements and cargo type of the current order. This combination of manual selection and automatic API identification ensures the task mode aligns perfectly with actual operational goals, preventing misjudgments of task attributes from deviating from the core requirement of "extreme fuel efficiency." Based on the aforementioned operational status and task attribute data, a dataset containing operational status and task attributes is constructed. This dataset integrates multi-dimensional information such as vehicle, road surface, and task data, providing complete and systematic data support for the subsequent five-dimensional model construction and ensuring the comprehensiveness of subsequent analysis. Figure 2 As shown.

[0069] S200, Five-Dimensional Model Construction:

[0070] Based on the dataset constructed using S100, five-dimensional parameter quantization intervals were defined: the task mode was "extreme fuel-saving mode," the total load was "medium load," the road surface characteristics were "slope ≤ 1°, high friction coefficient, no potholes," the power source efficiency was "engine in the low-efficiency zone" and the motor in the high-efficiency zone, and component wear was "normal." This clear quantization interval division makes the characteristics of each parameter more concrete, providing a unified and standardized analysis standard for the subsequent calculation of the five-dimensional parameter coupling adaptation algorithm, avoiding the influence of parameter dimension confusion on the adaptation degree calculation results. Subsequently, the five-dimensional parameter coupling adaptation algorithm was used to calculate the adaptation degree of the five-dimensional parameters with all power modes. The expression of the five-dimensional parameter coupling adaptation algorithm is as follows: This algorithm comprehensively considers the impact of five dimensions on the power mode: task, load, road surface, power source efficiency, and component wear. It achieves precise coupling and matching of multiple parameters with the power mode, avoiding adaptation deviations caused by a single dimension. It filters out power modes with an adaptation degree ≥ 0.7 and finally generates a power mode candidate set, including "front motor single drive mode", "rear motor single drive mode", and "engine single drive mode". This filtering process can eliminate power modes with insufficient adaptation, narrowing the scope for the next step of optimal mode selection, and ensuring that each mode in the candidate set can initially meet the operation and task requirements of "extreme fuel efficiency", "medium load", and "flat road section".

[0071] S300, Dynamic Weighted Game Theory and Optimal Pattern Selection:

[0072] Construct a three-dimensional objective function that includes energy efficiency, energy loss, and response speed. The expression of the three-dimensional objective function is as follows: This objective function comprehensively evaluates the overall performance of the power mode from three core dimensions: energy efficiency, component protection, and operational response. This avoids the one-sidedness of mode selection caused by focusing on only a single dimension, ensuring that the selected power mode can balance multiple needs. The current scenario is "extreme fuel-saving mode," requiring priority to ensure energy efficiency to reduce transportation energy consumption. Therefore, the weight proportions of each component in the three-dimensional objective function are determined through energy efficiency weight calculation formulas, loss weight calculation formulas, and response speed weight calculation formulas: Energy Efficiency Weight The calculation formula is: Energy efficiency weighting The calculation formula is: Response speed weight The calculation formula is: The weighting of energy efficiency is the highest, followed by loss, and then response speed. This weighting distribution ensures that the calculation results of the objective function are more aligned with the core requirement of "extreme fuel economy," highlighting the dominant role of energy efficiency. Based on this three-dimensional objective function, the candidate power modes generated by the S200 are calculated. By comparing the objective function values ​​of "front motor-only drive mode," "rear motor-only drive mode," and "engine-only drive mode," the objective function value of "front motor-only drive mode" is the highest, as the motor is currently in the high-efficiency zone and its energy consumption is lower than that of the engine. Therefore, it is selected as the optimal power mode. This result ensures that the selected mode meets the requirement of extreme fuel economy while also taking into account component loss control, which is in line with the operational goals of intercity logistics conventional transportation.

[0073] S400, Predictive Pre-Switch Adjustment:

[0074] A predictive model based on temporal feature fusion is used for forecasting. This model can deeply mine the temporal correlation between historical operating data and current task attributes, improving the accuracy of future road condition predictions through temporal feature analysis and avoiding delays in power adjustments due to forecasting errors. Input parameters include historical operating data and current task attributes. Rich and scenario-appropriate input parameters provide comprehensive and sufficient support for model prediction, ensuring that the prediction results are highly consistent with actual road conditions and task requirements. The model, after calculation through a feature processing layer, three LSTM units, and an output layer, outputs the predicted results for the next 500 meters of road: the load remains medium, the road surface remains flat, and the task requirement remains extreme fuel efficiency. This prediction result allows for advance understanding of changes in operating conditions for subsequent road sections, providing a clear direction for power source pre-adjustment. Combined with a response correction algorithm, it is determined that the power source needs to be pre-adjusted to match the "front motor independent drive mode." The expression for the response correction algorithm is: The response correction algorithm can optimize the evaluation results of response speed and avoid untimely pre-adjustment due to prediction deviation or switching delay. According to the pre-adjustment rules, the engine is shut down by cutting off fuel 0.3 seconds in advance. At the same time, the front motor is preloaded to 20% of the target torque through the motor controller. This pre-adjustment operation allows the engine to stop running in advance and the motor to enter the adaptation state in advance, ensuring that the engine can switch to the optimal power mode smoothly and quickly during subsequent driving, avoiding energy waste during the switching process and meeting the demand for extreme fuel saving.

[0075] S500, Dynamic Compensation Adjustment for Component Losses:

[0076] During operation in "front motor independent drive mode," component wear parameters are monitored in real time. This real-time monitoring can promptly capture changes in component wear, preventing undetected damage due to excessive wear and ensuring vehicle safety. After driving approximately 5 kilometers, the wear monitoring module shows that the battery capacity degradation rate is still 8%, the engine cylinder wear has not changed significantly, and the cumulative temperature rise of the motor windings has reached 32°C. The compensation mechanism has not been triggered. Maintaining the current mode at this point ensures stable power output and low energy consumption, meeting the requirements for extreme fuel economy. Continuing to drive approximately 3 kilometers from the service area, the component wear monitoring shows that the battery capacity degradation rate has risen to 10%, triggering the compensation mechanism. Timely triggering of the compensation mechanism can prevent further battery capacity degradation and extend battery life. Since the vehicle is currently in motion, the equalization charging operation for nighttime parking will not be performed. Instead, the power distribution mode will be adjusted and temporarily switched to "rear motor independent drive mode" to reduce the continuous load on the front motor and battery, thereby reducing battery energy consumption pressure. After the vehicle is parked at a service area, equalization charging will be initiated until the battery capacity degradation rate returns to a safe range. Subsequent driving will then resume "front motor independent drive mode". This adjustment method can protect the battery during driving and complete battery repair when parked, while restoring the optimal power mode to ensure that subsequent driving can still meet the demand for extreme fuel efficiency.

[0077] In summary, in conventional intercity logistics transportation scenarios, multi-dimensional information collection accurately acquires data on vehicles, roads, and tasks, laying the foundation for the construction of a five-dimensional model. A five-dimensional parameter coupling and adaptation algorithm is used to select a candidate set of power modes suitable for the scenario. Through a three-dimensional objective function and weight allocation, energy efficiency is prioritized, determining that the front motor-only drive mode is optimal. LSTM neural networks are used to predict road conditions, combined with a response correction algorithm to pre-adjust the power source, reducing energy waste. Real-time monitoring of component wear triggers a compensation mechanism to switch motors and park the vehicle for charging to protect the battery. Throughout the process, fuel efficiency is the primary goal, achieving highly efficient and energy-saving transportation while controlling component wear.

[0078] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. An adaptive switching method for a multi-mode power system of a hybrid commercial vehicle, characterized in that, The specific steps of this method are as follows: S100, Multi-dimensional Information Acquisition: Acquires information on load, road surface, power source, component wear and task mode through sensors and interfaces, and performs preprocessing and feature extraction to construct a dataset containing operating status and task attributes; S200, Five-dimensional model construction: Based on the dataset, the five-dimensional parameter quantization intervals, including task mode, total load, road surface features, power source efficiency and component loss, are divided. The fit degree between the five-dimensional parameters and the power mode is calculated by the five-dimensional parameter coupling and adaptation algorithm to generate a power mode candidate set. S300, Dynamic Weight Game and Optimal Mode Selection: Construct a three-dimensional objective function that includes energy efficiency, loss, and response speed. Calculate the candidate power mode based on the objective function and select the optimal power mode with the highest objective function value. S400, Predictive Pre-Switching Adjustment: Utilizes historical operational data and task attributes to construct a predictive model, predicts future load changes, road surface characteristics, and task demand fluctuations on road sections, calculates response speed correction values ​​using a response correction algorithm, and pre-starts, pre-shuts down, or pre-adjusts parameters of the power source based on the prediction results and the optimal power mode. S500, dynamic compensation and adjustment of component wear: During operation in the optimal power mode, the component wear parameters are monitored in real time and the compensation mechanism is triggered in combination with the current task mode; when the component wear exceeds the standard, the power distribution ratio or operation mode is dynamically adjusted until the wear parameters return to the safe range.

2. The adaptive switching method for a multi-mode power system of a hybrid commercial vehicle according to claim 1, characterized in that, In the S100 multi-dimensional information acquisition process, four suspension pressure sensors deployed on each of the front and rear axles of the cargo box are used to collect suspension pressure data in real time to calculate the total load value and center of gravity offset coefficient; the road slope, friction coefficient and pothole level are generated in real time by fusing 16-line vehicle-mounted LiDAR with high-precision navigation map; the engine speed and load rate are obtained in real time by the engine ECU; and the torque and speed of the front motor and rear motor are obtained in real time by the dual motor controller. The loss monitoring module calculates engine cylinder wear, battery capacity decay rate, and cumulative motor winding temperature rise in real time; the user interface of the vehicle touch screen and mobile APP obtains the manually selected task mode; and the API interface of the logistics order system automatically identifies the delivery time requirements and goods type in the order.

3. The adaptive switching method for a multi-mode power system of a hybrid commercial vehicle according to claim 1, characterized in that, In the S200 five-dimensional model construction, the specific contents of the five-dimensional parameter quantification range are as follows: the quantification range of the task mode is divided into extreme fuel-saving mode, standard balanced mode, time-limited delivery mode and component maintenance mode; the quantification range of the total load is divided into no-load, medium-load and full-load; the quantification range of road surface characteristics is divided into: slope, friction coefficient and pothole level; the quantification range of power source efficiency is divided into: engine high-efficiency zone and motor high-efficiency zone. The quantitative range of component wear is divided into: normal, critical, and exceeding the standard.

4. The adaptive switching method for a multi-mode power system of a hybrid commercial vehicle according to claim 1, characterized in that, In the S200 five-dimensional model construction, the expression for the five-dimensional parameter coupling adaptation algorithm is: ,in, Let be the fitness level of the k-th power mode, with a value ranging from 0 to 1. Quantify the task mode factors. This is the total load quantization factor. For road surface feature coupling factor, The power source efficiency factor. Quantification factor for component loss. Represents the dynamic weights of the task mode dimension. Dynamic weights for the total load dimension. For the component loss dimension, dynamic weights, and ,filter The dynamic patterns form a candidate set.

5. The adaptive switching method for a multi-mode power system of a hybrid commercial vehicle according to claim 1, characterized in that, In S300, the expression for the three-dimensional objective function in dynamic weighted game and optimal mode selection is: ,in, Represents the objective function value. Indicates total energy efficiency. Indicates total loss. Indicates the total response speed. Indicates energy efficiency weight. Indicates the loss weight. Response speed weight, and .

6. The adaptive switching method for a multi-mode power system of a hybrid commercial vehicle according to claim 5, characterized in that, In the S300 dynamic weight game and optimal mode selection, energy efficiency weight is used. The calculation formula is: ,in, This represents the task's energy efficiency sensitivity coefficient. Represents total loss; energy efficiency weight The calculation formula is: ,in, Task loss sensitivity coefficient; response speed weight. The calculation formula is: ,in, This represents the task response coefficient.

7. The adaptive switching method for a multi-mode power system of a hybrid commercial vehicle according to claim 1, characterized in that, In S400, the predictive pre-switching adjustment uses an LSTM neural network model based on temporal feature fusion, specifically including: Input layer parameters: Historical operation data includes load change curves for the same route over the past 3 months, road feature library, and power mode switching history; Current task attributes include current location, mileage traveled, navigation planning data for the remaining route, task mode, and real-time traffic information; Feature processing layer: Normalizes the input parameters, assigns weights to different features through an attention mechanism, and extracts temporal features; Network structure: It contains 3 layers of LSTM units, each followed by a Dropout layer to prevent overfitting, and the output layer is a fully connected layer; Output parameters: load change trend, road surface feature changes, and task requirement fluctuations for the next 500 meters of road section; Training process: The sliding window method is adopted, and the parameter adjustment method with adaptive learning rate is used. The initial learning rate is 0.

001. The model parameters are continuously adjusted to reduce the mean squared difference between the predicted value and the actual value to a minimum. Training is stopped when the mean squared difference of the validation set is ≤0.

02. The model is incrementally updated every 7 days based on the newly added operational data.

8. The adaptive switching method for a multi-mode power system of a hybrid commercial vehicle according to claim 7, characterized in that, In the S400 predictive pre-switching adjustment, the training steps of the prediction model are as follows: Data preparation: Collect historical and new operational data and integrate them into a training dataset. The historical operational data includes load changes along the same route, road surface feature measurements, and power mode switching records. Clean the data, remove obvious outliers, and retain valid data sets. Sliding window partitioning: The training dataset is partitioned using a sliding window method, with a set window size and sliding step; data is selected from each window according to a set ratio as the training set, and the remaining data is used as the validation set; Model structure initialization: Set the network to contain multiple layers of time series processing units, with each layer containing a set number of processing nodes; set a random data discard mechanism after each layer and set the discard ratio; set the output nodes for the corresponding prediction parameters in the output layer; Iterative training: An adaptive learning rate optimization method is adopted, and an initial learning rate is set. In each iteration, the training set data is input into the model to calculate the prediction results, and the network parameters are adjusted by comparing them with the actual values. The prediction bias is evaluated periodically using validation set data. Stop training: When the validation set bias continues to meet the set conditions, stop the current training and save the network parameters at this time as the current version of the model; Incremental Updates: Regularly add new operational data to the training dataset, repeat incremental training based on the current model version, and update the network parameters.

9. The adaptive switching method for a multi-mode power system of a hybrid commercial vehicle according to claim 1, characterized in that, In S400, during the predictive pre-switching adjustment, the expression for the response correction algorithm is: ,in, This indicates the corrected total response speed. Indicates the delay time for switching power modes. This represents the prediction deviation coefficient. It is the response correction factor.

10. The adaptive switching method for a multi-mode power system of a hybrid commercial vehicle according to claim 1, characterized in that, In the S500, the compensation mechanism for dynamic compensation adjustment of component losses is as follows: In component maintenance mode: when the wear of the engine cylinder reaches 30%, start alternating load: use the front motor for 3 consecutive starts, and the rear motor for the 4th start, with a switching interval of ≥5 seconds; when the battery capacity decreases by 10%, start equalization charging when parking at night. In time-limited delivery mode: when the motor winding temperature rises to 60℃, the torque of the two motors is evenly distributed, and the total torque decrease is ≤10%; In standard balance mode: when component wear reaches the critical value, the engine load rate is limited to ≤80% and the motor torque is limited to ≤200N・m.

Images