Cloud-end interaction electric vehicle transmission system dynamic characteristic positioning and analysis method

Abstract

This invention discloses a cloud-edge interactive method for locating and analyzing dynamic features of electric vehicle transmission systems. The method includes: collecting multi-source data; constructing fused data; uploading the fused data to a cloud platform based on a triggering mechanism; on the cloud platform, constructing behavior vectors based on the fused data, constructing dynamic feature indicators based on the behavior vectors, and constructing dynamic feature vectors from the dynamic feature indicators; constructing a feature expectation base value and calculating the deviation of the dynamic feature indicators from the feature expectation base value; calculating the individual contribution value of each component in the behavior vector to the deviation, and extracting the data component with the largest individual contribution value; constructing a comprehensive driving cost function; minimizing the comprehensive driving cost function value based on the extracted data component with the largest individual contribution value; and generating a strategy parameter set. This invention enables the localization of dynamic features of the transmission system, improves the driving efficiency and power tracking accuracy of the electric vehicle transmission system, and enhances the smoothness of gear shifting.

Description

Technical Field

[0001] This invention belongs to the field of intelligent connected vehicle technology, specifically relating to a method for dynamic feature localization and analysis of electric vehicle transmission systems with cloud-edge interaction. Background Technology

[0002] As the core assembly for power output and energy conversion in electric vehicles, the transmission system's operating characteristics directly impact the vehicle's energy consumption, power response, and driving comfort. With the continuous development of vehicle technology, accurately locating and analyzing the dynamic characteristics of the transmission system using digital means has become crucial for improving the level of vehicle intelligence.

[0003] Accurate positioning and analysis of the dynamic characteristics of the transmission system are crucial for electric vehicles. They can reveal the operating rules of the transmission system in actual driving and complex road conditions, and are also the core prerequisite for improving vehicle energy utilization efficiency, improving power response, and enhancing driving comfort. For example, the invention patent CN202310133136.9, "A method for analyzing the information features of a hybrid transmission system based on vehicle-cloud collaboration," fails to deeply link the features of "human-vehicle-road" to achieve dynamic feature localization of the transmission system. Similarly, the invention patent CN202411245770.2, "Digital twin system, method, and device for traction transmission system of rail transit vehicles," focuses on digital twin modeling and health status early warning of key electrical components inside the rail transit traction system, without adopting a trigger-based mechanism to upload data to optimize communication efficiency. Furthermore, the invention patent CN201410212562.2, "Method and device for softening the power transmission system using vehicle-to-cloud-to-vehicle system," mainly relies on vehicle position and historical road condition data to adjust components such as torque converter bypass clutches in advance to suppress transmission system disturbances. At the same time, it focuses on improving smoothness in advance for specific road conditions, without constructing a comprehensive driving cost function that integrates economy, power, and comfort, and cannot achieve multi-objective dynamic weighted optimization and closed-loop control for the operating status of the transmission system.

[0004] In summary, existing methods have limitations in dealing with the complex dynamics of electric vehicle drive systems, including insufficient dynamic feature analysis and a lack of multi-objective optimization mechanisms. Summary of the Invention

[0005] The purpose of this invention is to provide a method for locating and analyzing the dynamic features of an electric vehicle transmission system using cloud-edge interaction.

[0006] The technical solution for achieving the objective of this invention is: a cloud-edge interactive method for locating and analyzing the dynamic features of an electric vehicle transmission system, characterized by the following steps:

[0007] Step (1): Collect data from multiple sources;

[0008] Step (2): Spatiotemporal registration and fusion of the multi-source data collected in step (1) are performed to construct fused data;

[0009] Step (3): Upload the fused data to the cloud platform based on the trigger condition mechanism;

[0010] Step (4): On the cloud platform, based on the uploaded fusion data, construct micro-behavior vectors, intentional behavior vectors, and environmental behavior vectors. Based on the behavior vectors, construct dynamic feature indicators representing the economy, comfort, and power of the transmission system respectively. Construct dynamic feature vectors from the dynamic feature indicators. Construct the feature expectation base value to represent the operating state of the transmission system and calculate the deviation of the dynamic feature indicators from the feature expectation base value. Calculate the individual contribution value of each component in the behavior vector to the deviation value and extract the data component with the largest individual contribution value.

[0011] Step (5): Construct sub-cost functions for economy, comfort and power; merge the sub-cost functions to construct a comprehensive driving cost function; minimize the comprehensive driving cost function value based on the data component with the largest single contribution value extracted in step (4); generate a set of strategy parameters including energy management weight, shift rule correction factor and torque compensation coefficient based on the optimization results;

[0012] Step (6): The vehicle terminal receives the strategy parameter set generated in step (5) and updates the vehicle control parameters.

[0013] Furthermore, the multi-source data in step (1) includes: transmission system data as the main data, and on-board data including drive system data and vehicle status data as auxiliary data, as well as roadside data composed of roadside driving environment data.

[0014] Furthermore, step (2) specifically involves:

[0015] Step (21): Time registration: Based on the Precision Time Protocol (PTP), the clock reference of the unified transmission system and each acquisition device is established;

[0016] Step (22): Spatial registration: Establish a unified vehicle coordinate system, map the data collected by the sensors to the vehicle coordinate system, and use map matching technology to associate the real-time position of the vehicle with the map data;

[0017] Step (23): Obtain bus direct acquisition data from the spatiotemporally registered data, including gear position signal, motor angular velocity, motor output torque, wheel end angular velocity, vehicle speed, longitudinal acceleration, steering wheel angle, pedal position, distance to the stop line of the traffic light ahead, and the remaining time of the green light of the traffic light ahead. Estimate state information including wheel end torque, road slope, curvature, and adhesion coefficient based on the direct acquisition data.

[0018] Step (24): Standardize and encapsulate the directly collected data and estimated status information to construct fused data for dynamic feature analysis of the transmission system.

[0019] Furthermore, step (21) specifically involves:

[0020] Step (211): Propagation delay estimation: The average propagation delay is calculated using a weighted statistical estimation method within a sliding window.

[0021] ;

[0022] In the formula, T delay The average propagation delay is denoted by N; the number of samples within the preset observation window is denoted by α; the asymmetry factor of the communication link path is denoted by α, which is used to compensate for the delay offset caused by the difference in uplink and downlink bandwidth; k is the sampling sequence of the synchronization period; T is the average propagation delay. 1,k T 2,k These represent the times when the master device sends the k-th synchronization message and when the slave device receives the k-th synchronization message, respectively; T 3,k T 4,k These are the times when the slave device sends the k-th delay request and when the master device receives the k-th delay request, respectively.

[0023] Step (212): Clock offset calculation: The clock offset is calculated using an error feedback correction model.

[0024] ;

[0025] In the formula, offset k λ is the clock offset; λ∈(0,1] is the filter gain coefficient, used to balance the weights of historical synchronization trajectories and current sampled values ​​to filter out transient network noise; offset k-1 This is the clock offset from the previous synchronization cycle;

[0026] Step (213): Time correction and frequency drift compensation: A frequency drift function is introduced to eliminate the cumulative timing error caused by the difference in hardware crystal oscillator frequency between synchronization cycles. The final correction formula is defined as follows:

[0027] ;

[0028] In the formula, T corrected The precise time synchronized with the master clock after device calibration; T local t is the device's current local time; t is the system's current time; t sync ξ is the reference time of the last synchronization operation; ξ is the integral variable, representing the time elapsed from the last synchronization to the current time; ρ(ξ) is the frequency drift function characterizing the difference in hardware crystal oscillators.

[0029] Furthermore, step (4) specifically involves:

[0030] Step (41): Extract gear position signal, wheel end torque, wheel end angular velocity, motor output torque, and motor angular velocity information to construct a micro behavior vector; extract longitudinal acceleration information to construct an intentional behavior vector; extract pedal position, distance to the stop line of the traffic light ahead, remaining green light time of the traffic light ahead, road slope, curvature, and adhesion coefficient information to construct an environmental behavior vector;

[0031] Step (42): Construct dynamic characteristic indicators representing the economy of the transmission system based on the micro-behavioral vector, including the driving efficiency of the transmission system and the gear holding trend score; construct dynamic characteristic indicators representing the comfort of the transmission system based on the intentional behavior vector, including longitudinal impact, which is determined based on longitudinal acceleration; construct dynamic characteristic indicators representing the power performance of the transmission system based on the environmental behavior vector, including driving force deviation, which is determined based on the difference between the expected driving force and the real-time output driving force of the vehicle.

[0032] Step (43): Construct a dynamic feature vector composed of dynamic feature indicators of economy, comfort and power; retrieve and match historical samples that match the current operating conditions in the historical database of the cloud platform; calculate the expected base value of the feature based on the actual recorded values ​​of the dynamic feature indicators in the dynamic feature vector of the matched historical samples and the similarity weight with the current operating conditions; calculate the deviation of the dynamic feature indicators from the expected base value of the feature; calculate the individual contribution value of each component in the behavior vector to the deviation; and take the component with the largest individual contribution value as the dominant factor.

[0033] Furthermore, the drive efficiency of the transmission system in step (42) is calculated as follows:

[0034] ;

[0035] In the formula, η t For the drive efficiency of the transmission system; T wheel ω is the wheel end torque; wheel T is the wheel end angular velocity; motor ω is the output torque of the motor. motor This refers to the angular velocity of the motor.

[0036] The method for extracting the gear position trend score is as follows:

[0037] ;

[0038] In the formula, S gear The gear position maintains a trend score; f(·) is a normalized membership function that maps the input to the [0,1] interval; ε1 and ε2 are weighting coefficients; Δω motor ω represents the motor angular velocity deviation.opt (T motor (t) represents the current output torque T. motor Theoretically optimal economic angular velocity for the lower motor; i g (t) represents the actual gear ratio in the current gear signal of the system; i opt (t) represents the theoretically optimal transmission ratio;

[0039] The longitudinal impact intensity in step (42) is calculated as follows:

[0040] ;

[0041] In the formula, j represents the longitudinal impact intensity; Δa represents the change in longitudinal acceleration; and Δt represents the time interval.

[0042] The calculation process for "driving force deviation" in step (42) is as follows:

[0043] The comprehensive road resistance potential is constructed based on the environmental behavior vector and calculated as follows:

[0044] ;

[0045] In the formula, denoted as , where is the comprehensive road resistance potential; s0 is the current position; L is the aiming distance; ω(s) is the spatial weighting function varying with s; G(s) is the road slope at s; C(s) is the road curvature at s; Ф μ (s) represents the adhesion coefficient related term; μ ref The reference adhesion coefficient is μ(s); the adhesion coefficient at position s is ε. g ε c and ε μ These are the weighting coefficients for road surface slope, curvature, and adhesion coefficient, respectively.

[0046] The driving intention factor is derived from the standard deviation of the rate of change of pedal position over a short period of time, as follows:

[0047] ;

[0048] In the formula, The rate of change of pedal position; Pedal i Let be the position of the pedal sampled at time t; σ represents the mean rate of change of pedal position; σ is the driving intention factor.

[0049] The traffic light traffic status is calculated as a vector with two elements, as follows:

[0050] ;

[0051] In the formula, VTL Traffic light status; χ TL For valid traffic light indication; L stopline T is the distance to the stop line of the traffic light ahead. green_rem C represents the remaining green time of the traffic light ahead; max To presuppose a sufficiently large constant;

[0052] Define the valid indicator for traffic lights:

[0053] ;

[0054] In the formula, χ TL The value is 1 when traffic light information is available and can be obtained; the value is 0 when there is no traffic light or it is not available.

[0055] The cloud platform derives the desired acceleration based on traffic light status, and the calculation is as follows:

[0056] ;

[0057] In the formula, a des V is the desired acceleration. current Current vehicle speed;

[0058] The desired driving force is derived by combining the driving intention factor and calculated as follows:

[0059] ;

[0060] In the formula, F des Driven by expectations; is the intent sensitivity constant; m is the vehicle mass; F resist The driving resistance is calculated using the comprehensive road resistance potential.

[0061] The driving force deviation is calculated as follows:

[0062] ;

[0063] In the formula, ΔF is the driving force deviation; i g i is the gear ratio; i0 is the main reduction ratio; r is the effective radius of the wheel.

[0064] Furthermore, step (43) is as follows:

[0065] Step (431): Under synchronous time, the matrix form of the dynamic feature vector is expressed as follows:

[0066] ;

[0067] In the formula, Y(t) is the dynamic characteristic vector; η(t) is the drive efficiency of the transmission system at the current moment; S gear(t) represents the gear holding trend score at the current moment; j(t) represents the longitudinal impact at the current moment; ΔF(t) represents the driving force deviation at the current moment;

[0068] The matrix form of the behavior vector is expressed as follows:

[0069] ;

[0070] In the formula, X(t) is the action vector; X micro (t) is the microscopic behavior vector; X intent (t) is the intentional behavior vector; X env (t) represents the environmental behavior vector;

[0071] The components of the behavior vector are expanded as follows:

[0072] ;

[0073] In the formula, i g (t) represents the actual gear ratio in the current gear signal of the system; T wheel (t) represents the current wheel-end torque; ω wheel (t) represents the current wheel end angular velocity; T motor (t) represents the motor output torque at the current moment; ω motor (t) represents the current angular velocity of the motor; a(t) represents the current longitudinal acceleration; Pedal(t) represents the current pedal position; G(s) represents the road gradient at the current vehicle position s; C(s) represents the road curvature at the current vehicle position s; μ(s) represents the road adhesion coefficient at the current vehicle position s, s = s(t); L stopline (t) represents the distance between the current vehicle and the stop line at the traffic light ahead; T green_rem (t) represents the remaining green time of the traffic light ahead;

[0074] Step (432): The expected base value of the dynamic feature index is calculated as follows:

[0075] ;

[0076] In the formula, y n,exp (t) represents the expected baseline value of the feature; H represents the total number of samples retrieved and matched by the cloud platform from the historical database that are in a similar working condition to the current time t; h represents the index of the matched historical sample number; ρ h y represents the high-dimensional similarity weight between the h-th historical matching sample and the current real-time operating condition; n (h) This represents the actual recorded value of the nth dynamic feature indicator in the hth historical matching sample.

[0077] Step (433): The deviation of the dynamic feature index from the expected base value is calculated as follows:

[0078] ;

[0079] In the formula, Δy n (t) represents the deviation; y n (t) represents the actual calculated value of the nth feature index component in the dynamic feature vector Y(t) of the system at the current moment; y n,exp (t) represents the expected base value of the nth feature index component. Similarly, the expected base value x of the ith data component in the behavior vector is obtained. i,exp (t);

[0080] Step (434): The individual contribution value of the i-th data component in the behavior vector X(t) to the deviation of the corresponding feature index is calculated as follows:

[0081] ;

[0082] In the formula, S n,i (t) represents the local sensitivity coefficient of the i-th data component of the behavior vector X(t) with respect to the n-th feature index component of the dynamic feature vector Y(t), and the right-hand side represents the partial derivative of the n-th feature index function with respect to the i-th behavior component; C n,i (t) represents the individual contribution value; x i (t) represents the i-th data component in the behavior vector X(t); x i,exp (t) represents the expected base value of the i-th data component obtained by weighting based on historical matching samples; N x S represents the total dimension of the behavior vector X(t); n,q (t) is the local sensitivity coefficient of the q-th action vector component; x q (t) represents the actual collected value of the q-th data component; x q,exp (t) represents the characteristic expectation base value of the q-th data component.

[0083] Furthermore, step (5) is as follows:

[0084] Step (51): Construct sub-cost functions for economy, comfort, and dynamics:

[0085] The economic cost function is based on the drive efficiency η of the transmission system. t Maintaining trend with gear position rating S gear Perform nonlinear modeling:

[0086] ;

[0087] In the formula, J econε3 is the cost function of economic efficiency; ε4 are the distribution coefficients of economic efficiency. ε3 is used to adjust the penalty of the system when the transmission efficiency deviates from the ideal range, and ε4 is used to adjust the degree of suppression of frequent gear shifting.

[0088] The comfort cost function is constructed based on the longitudinal impact j:

[0089] ;

[0090] In the formula, J comf ε is the comfort cost function; ε5 and ε6 are comfort smoothness operators, where ε5 is the penalty coefficient used to adjust the absolute amplitude of a single longitudinal impact, and ε6 is the integral operator used to penalize the rate of change of impact; j(t) is the longitudinal impact at the current moment; j(τ) is the predicted value of the longitudinal impact.

[0091] The dynamic cost function takes the driving force deviation ΔF as its core input:

[0092] ;

[0093] In the formula, J perf ε7 and ε8 are dynamic performance cost functions; ε7 is the basic penalty weight used to adjust the driving force deviation ΔF under normal cruise conditions, and ε8 is the nonlinear gain term dynamically adjusted based on the current vehicle speed and the desired acceleration.

[0094] Step (52): The cloud platform integrates the sub-cost functions of economy, comfort, and power performance into a comprehensive driving cost function through a dynamic weighting module:

[0095] ;

[0096] In the formula, J total For the comprehensive driving cost function; ε econ ε comf ε perf These are the weighting coefficients for economy, comfort, and power output by the dynamic weighting module, and the sum of these weighting coefficients is 1.

[0097] Step (53): The cloud platform minimizes the comprehensive driving cost function based on the data with the largest single contribution value extracted in step (4); the strategy parameter set is obtained by the cloud platform by minimizing the comprehensive driving cost function value and is mapped to the dynamic characteristic index in step (4), including energy management weight, shifting rule correction factor and torque compensation coefficient.

[0098] Compared with the prior art, the significant advantages of this invention are:

[0099] This invention employs cloud-edge interactive association of "human-vehicle-road" features to locate the dynamic features of the transmission system. Based on the optimization of the comprehensive driving cost function value, a set of strategy parameters is generated and distributed, which improves the driving efficiency and power tracking accuracy of the electric vehicle transmission system and enhances the smoothness of gear shifting. Attached Figure Description

[0100] Figure 1 This is a flowchart of the cloud-edge interactive method for locating and analyzing dynamic features of an electric vehicle transmission system according to the present invention.

[0101] Figure 2 This is a schematic diagram illustrating the implementation principle of the cloud-edge interactive method for dynamic feature localization and analysis of electric vehicle transmission systems according to the present invention.

[0102] Figure 3 This is a schematic diagram comparing the drive efficiency of the transmission system in an embodiment of the present invention and a comparative example;

[0103] Figure 4 This is a schematic diagram comparing the longitudinal impact of embodiments and comparative examples of the present invention;

[0104] Figure 5 This is a schematic diagram comparing the driving force deviations of embodiments and comparative examples of the present invention. Detailed Implementation

[0105] The present invention will now be described in further detail with reference to the accompanying drawings.

[0106] In this invention, "dynamic characteristics" refers to the change pattern of the state of the electric vehicle transmission system over time under the coupling effect of "human-vehicle-road". This characteristic is represented by a behavior vector composed of on-board and roadside data at the bottom layer, and is constructed by indicators that characterize the economy, power and comfort of the transmission system on the cloud platform.

[0107] like Figures 1-2 As shown, the method for dynamic feature localization and analysis of electric vehicle transmission systems using cloud-edge interaction includes:

[0108] Step (1): Collect multi-source data with transmission system data as the core. The data includes: transmission system data as the main data, and on-board data including drive system data and vehicle status data as auxiliary data, as well as roadside data composed of roadside driving environment data.

[0109] The multi-source data in step (1) specifically includes: transmission system data as the main data, including gear position signal and output shaft speed. The gear position signal is used to verify the transmission ratio, and the output shaft speed is used to calculate the wheel end angular velocity; drive system data as auxiliary data, including motor speed and motor output torque. The motor speed is used to extract the motor angular velocity; and vehicle state data used to characterize the vehicle state, including vehicle speed, longitudinal acceleration, steering wheel angle, and pedal position; and roadside driving environment data, including the distance to the stop line of the traffic light ahead and the remaining green light time of the traffic light ahead, which together construct the traffic light passage status.

[0110] Step (2): Spatiotemporal registration and fusion of the multi-source data collected in step (1) are performed to construct fused data;

[0111] Step (21): Time registration is based on the PTP precise time protocol, with the vehicle controller or GNSS module equipped with high-precision time synchronization as the global master clock and each distributed acquisition device as a slave clock; the time synchronization process is completed based on a four-step handshake mechanism, and the four moments of information exchange between the master and slave devices are T. 1,k T 2,k T 3,k T 4,k This leads to the construction of a dynamic synchronization model based on multi-sampling statistical filtering and frequency drift compensation. The slave clock device calculates its time offset relative to the master clock based on this dynamic synchronization model and dynamically corrects its local time, achieving high-precision data synchronization under a unified time base. Specifically, the process of calculating the propagation delay and clock offset of the communication link for time correction based on four moments from both the master and slave devices is as follows:

[0112] Step (211): Propagation delay estimation. To suppress the interference of vehicular network communication link jitter on synchronization accuracy, a weighted statistical estimation method within a sliding window is used to calculate the average propagation delay:

[0113] ;

[0114] In the formula, T delay The average propagation delay is denoted by N; the number of samples within the preset observation window is denoted by α; the asymmetry factor of the communication link path is denoted by α, which is used to compensate for the delay offset caused by the difference in uplink and downlink bandwidth; k is the sampling sequence of the synchronization period; T is the average propagation delay. 1,k T 2,k These represent the times when the master device sends the k-th synchronization message and when the slave device receives the k-th synchronization message, respectively; T 3,k T 4,k These represent the times when the slave device sends the k-th delay request and when the master device receives it, respectively.

[0115] Step (212): Clock skew calculation. To achieve smooth tracking of the clock skew between the master and slave devices, a clock skew based on an error feedback correction model is used:

[0116] ;

[0117] In the formula, offset k λ is the clock offset; λ∈(0,1] is the filter gain coefficient, used to balance the weights of historical synchronization trajectories and current sampled values ​​to filter out transient network noise; offset k-1 This is the clock offset of the previous synchronization cycle.

[0118] Step (213): Time correction and frequency drift compensation. A frequency drift function is introduced to eliminate the cumulative timing error caused by the difference in hardware crystal oscillator frequency between synchronization cycles. The final correction formula is defined as follows:

[0119] ;

[0120] In the formula, T corrected The precise time synchronized with the master clock after device calibration; T local t is the device's current local time; t is the system's current time; t sync ξ is the reference time of the last synchronization operation; ξ is the integral variable, representing the time elapsed from the last synchronization to the current time; ρ(ξ) is the frequency drift function characterizing the difference in hardware crystal oscillators.

[0121] Step (22): Spatial registration aims to build a unified data benchmark. Specifically, a vehicle coordinate system is established as the target reference system, and map matching technology is used to associate the real-time vehicle location information with map data including road feature information; at the same time, the acquired traffic light information is mapped to specific spatial nodes under the vehicle coordinate system to achieve spatial alignment of multi-source data under a unified coordinate system.

[0122] Step (23): Obtain bus direct acquisition data from the spatiotemporally registered data, including gear position signal, motor angular velocity, motor output torque, wheel end angular velocity, longitudinal acceleration, steering wheel angle, pedal position, distance to the stop line of the traffic light ahead, and the remaining green light time of the traffic light ahead. Estimate state information including wheel end torque, road slope, curvature, and adhesion coefficient based on the direct acquisition data. Standardize and encapsulate the direct acquisition data and estimated information to construct fusion data for dynamic feature analysis of the transmission system.

[0123] Step (3): Upload the fused data to the cloud platform based on the trigger condition mechanism;

[0124] Step (4): On the cloud platform, based on the uploaded fusion data, construct micro-behavior vectors, intentional behavior vectors, and environmental behavior vectors. Based on the behavior vectors, construct dynamic feature indicators that characterize the economy, comfort, and power of the transmission system. Construct dynamic feature vectors from the dynamic feature indicators. Construct the feature expectation base value to characterize the operating state of the transmission system and calculate the deviation of the dynamic feature indicators from the feature expectation base value. Calculate the individual contribution value of each component in the behavior vector to the deviation value. Extract the data component with the largest individual contribution value as the dominant factor to indicate the optimization of the comprehensive driving cost function in step (5).

[0125] Step (41): Extract gear position signal, wheel end torque, wheel end angular velocity, motor output torque, and motor angular velocity information to construct a micro behavior vector; extract longitudinal acceleration information to construct an intentional behavior vector; extract pedal position, distance to the stop line of the traffic light ahead, remaining green light time of the traffic light ahead, road slope, curvature, and adhesion coefficient information to construct an environmental behavior vector;

[0126] Step (42): Construct dynamic characteristic indicators representing the economy of the transmission system based on the micro-behavioral vectors, including transmission system driving efficiency and gear holding trend score. The transmission system driving efficiency is calculated as follows:

[0127] ;

[0128] In the formula, η t For the drive efficiency of the transmission system; T wheel ω is the wheel end torque; wheel T is the wheel end angular velocity; motor ω is the output torque of the motor. motor This represents the angular velocity of the motor.

[0129] The gear holding trend score is used to quantitatively evaluate the rationality of the transmission system in maintaining the current gear. It is a dynamic index characterizing the system's shifting stability, and its extraction method is as follows:

[0130] ;

[0131] In the formula, S gear The gear position maintains a trend score; f(·) is a normalized membership function that maps the input to the [0,1] interval; ε1 and ε2 are weighting coefficients; Δω motor ω represents the motor angular velocity deviation. opt (T motor (t) represents the current output torque T. motor Theoretically optimal economic angular velocity for the lower motor; i g (t) represents the actual gear ratio in the current gear signal of the system; i opt (t) represents the theoretically optimal transmission ratio.

[0132] Dynamic characteristic indicators representing the comfort of the transmission system are constructed based on the intentional behavior vector, including longitudinal impact, and are calculated as follows:

[0133] ;

[0134] In the formula, j represents the longitudinal impact intensity; Δa represents the change in longitudinal acceleration; and Δt represents the time interval.

[0135] The comprehensive road resistance potential is used to quantify the induced effect of the environment on the load. It is constructed based on the environmental behavior vector and calculated as follows:

[0136] ;

[0137] In the formula, denoted as , where is the comprehensive road resistance potential; s0 is the current position; L is the aiming distance; ω(s) is the spatial weighting function varying with s; G(s) is the road slope at s; C(s) is the road curvature at s; Ф μ (s) represents the adhesion coefficient related term; μ ref The reference adhesion coefficient is μ(s); the adhesion coefficient at position s is ε. g ε c and ε μ These are the weighting coefficients for road surface slope, curvature, and adhesion coefficient, respectively.

[0138] The driving intention factor, as a quantitative indicator of current driving style, is derived from the standard deviation of the rate of change of pedal position over a short period of time, and is calculated as follows:

[0139] ;

[0140] In the formula, The rate of change of pedal position; Pedal i Let be the position of the pedal sampled at time t; σ represents the mean rate of change of pedal position; σ is the driving intention factor.

[0141] The traffic light permitting status is a feature vector used to quantify the traffic constraints at the intersection ahead. It is a two-element vector describing the feasibility of the traffic light ahead, calculated as follows:

[0142] ;

[0143] In the formula, V TL Traffic light status; χ TL For valid traffic light indication; L stopline T is the distance to the stop line of the traffic light ahead. green_rem C represents the remaining green time of the traffic light ahead; maxThis is a pre-set, sufficiently large constant.

[0144] Define the valid indicator for traffic lights:

[0145] ;

[0146] In the formula, χ TL The value is 1 when there is and the traffic light information is available; it is 0 when there is no traffic light or the information is unavailable.

[0147] Based on environmental baselines, the cloud platform derives the desired acceleration by combining traffic light conditions, as calculated below:

[0148] ;

[0149] In the formula, a des V is the desired acceleration. current This represents the current vehicle speed.

[0150] The desired driving force is derived by combining the driving intention factor and calculated as follows:

[0151] ;

[0152] In the formula, F des Driven by expectations; is the intent sensitivity constant; m is the vehicle mass; F resist The driving resistance is calculated using the comprehensive road resistance potential.

[0153] The driving force deviation is calculated as follows:

[0154] ;

[0155] In the formula, ΔF is the driving force deviation; i g i is the gear ratio; i0 is the main reduction ratio; r is the effective radius of the wheel.

[0156] Step (43): Under synchronous time, the matrix form of the dynamic feature vector is expressed as follows:

[0157] ;

[0158] In the formula, Y(t) is the dynamic characteristic vector; η(t) is the drive efficiency of the transmission system at the current moment; S gear (t) represents the gear holding trend score at the current moment; j(t) represents the longitudinal impact at the current moment; ΔF(t) represents the driving force deviation at the current moment.

[0159] The matrix form of the behavior vector is expressed as follows:

[0160] ;

[0161] In the formula, X(t) is the action vector; X micro (t) is the microscopic behavior vector; X intent (t) is the intentional behavior vector; X env (t) is the environmental behavior vector.

[0162] The components of the behavior vector are expanded as follows:

[0163] ;

[0164] In the formula, i g (t) represents the actual gear ratio in the current gear signal of the system; T wheel (t) represents the current wheel-end torque; ω wheel (t) represents the current wheel end angular velocity; T motor (t) represents the motor output torque at the current moment; ω motor (t) represents the current angular velocity of the motor; a(t) represents the current longitudinal acceleration; Pedal(t) represents the current pedal position; G(s) represents the road gradient at the current vehicle position s; C(s) represents the road curvature at the current vehicle position s; μ(s) represents the road adhesion coefficient at the current vehicle position s, s = s(t); L stopline (t) represents the distance between the current vehicle and the stop line at the traffic light ahead; T green_rem (t) represents the remaining green time of the traffic light ahead.

[0165] The expected base value of the dynamic feature index is calculated as follows:

[0166] ;

[0167] In the formula, y n,exp (t) represents the expected baseline value of the feature; H represents the total number of samples retrieved and matched by the cloud platform from the historical database that are in a similar working condition to the current time t; h represents the index of the matched historical sample number; ρ h y represents the high-dimensional similarity weight between the h-th historical matching sample and the current real-time operating condition; n (h) This represents the actual recorded value of the nth dynamic feature index in the hth historical matching sample. Similarly, the expected base value of the behavior vector can be obtained.

[0168] The deviation of the dynamic characteristic index from the expected baseline value is calculated as follows:

[0169] ;

[0170] In the formula, Δy n (t) represents the deviation; y n(t) represents the actual calculated value of the nth feature index component in the dynamic feature vector Y(t) of the system at the current moment; y n,exp (t) represents the expected base value of the nth feature index component. Similarly, the expected base value x of the ith data component in the behavior vector is obtained. i,exp (t).

[0171] The individual contribution value of the i-th data component in the behavior vector X(t) to the deviation of the corresponding feature index is calculated as follows:

[0172] ;

[0173] In the formula, S n,i (t) represents the local sensitivity coefficient of the i-th data component of the behavior vector X(t) with respect to the n-th feature index component of the dynamic feature vector Y(t), and the right-hand side represents the partial derivative of the n-th feature index function with respect to the i-th behavior component; C n,i (t) represents the individual contribution value; x i (t) represents the i-th data component in the behavior vector X(t); x i,exp (t) represents the expected base value of the i-th data component obtained by weighting based on historical matching samples; N x S represents the total dimension of the behavior vector X(t); n,q (t) is the local sensitivity coefficient of the q-th action vector component; x q (t) represents the actual collected value of the q-th data component; x q,exp (t) represents the characteristic expectation base value of the q-th data component.

[0174] Step (5): Construct sub-cost functions for economy, comfort and power; merge the sub-cost functions to construct a comprehensive driving cost function; optimize the comprehensive driving cost function value by minimizing the dominant factors in step (4); generate a set of strategy parameters including energy management weights, shift rule correction factors and torque compensation coefficients based on the optimization results, and send them to the vehicle terminal.

[0175] Step (51): Construct the sub-cost functions for economy, comfort, and dynamics. The specific steps are as follows:

[0176] The economic cost function aims to evaluate energy transfer efficiency and the rationality of gear selection decisions, based on the drive efficiency η of the transmission system. t Maintaining trend with gear position rating S gear Perform nonlinear modeling:

[0177] ;

[0178] In the formula, J econε3 is the cost function of economic efficiency; ε4 are the distribution coefficients of economic efficiency. ε3 is used to adjust the penalty of the system for transmission efficiency deviating from the ideal range, and ε4 is used to adjust the degree of suppression of frequent gear shifting.

[0179] The comfort sub-cost function is used to evaluate and suppress longitudinal shocks induced by transmission system regulation, and is constructed based on the longitudinal shock degree j in step (42):

[0180] ;

[0181] In the formula, J comf ε is the comfort cost function; ε5 and ε6 are comfort smoothness operators, ε5 is the penalty coefficient used to adjust the absolute amplitude of a single longitudinal impact, and ε6 is the integral operator used to penalize the rate of change of impact; j(t) is the longitudinal impact at the current moment; j(τ) is the predicted value of the longitudinal impact.

[0182] The dynamic performance cost function is used to quantify the real-time response accuracy of the vehicle's power request, with the driving force deviation ΔF derived in step (416) as the core input:

[0183] ;

[0184] In the formula, J perf ε7 and ε8 are dynamic performance cost functions; ε7 is the basic penalty weight used to adjust the driving force deviation ΔF under normal cruise conditions, and ε8 is a nonlinear gain term that is dynamically adjusted based on the current vehicle speed and the desired acceleration.

[0185] Step (52): The cloud platform integrates the sub-cost functions of economy, comfort, and power performance into a comprehensive driving cost function through a dynamic weighting module:

[0186] ;

[0187] In the formula, J total For the comprehensive driving cost function; ε econ ε comf ε perf These are the weighting coefficients for economy, comfort, and power output by the dynamic weighting module, and the sum of these weighting coefficients is 1.

[0188] Step (53): Based on the dominant factors in step (4), the cloud platform adjusts the weights of the comprehensive driving cost function and searches for the strategy parameter set with the minimum value of the comprehensive driving cost function in the prediction time domain. The strategy parameter set establishes a mapping relationship with the dynamic characteristic indicators. When the dynamic characteristic indicator located in step (4) is longitudinal impact, the cloud platform adjusts the comfort cost weights in the comprehensive driving cost function to obtain the shifting rule correction factor; when the dynamic characteristic indicator located is driving force deviation, the cloud platform adjusts the power cost weights in the comprehensive driving cost function to obtain the torque compensation coefficient; when the dynamic characteristic indicator located is transmission system driving efficiency, the cloud platform adjusts the economy cost weights in the comprehensive driving cost function to obtain the energy management weight.

[0189] Step (6): The vehicle terminal receives the energy management weights, shift rule correction factors, and torque compensation coefficients generated in step (5) and performs a safety check to update the vehicle control parameters. The energy management weights are used to optimize the drive efficiency of the transmission system to improve its economy; the shift rule correction factors are used to reduce longitudinal shock to improve comfort; and the torque compensation coefficients are used to compensate for drive force deviations to improve the power performance of the transmission system.

[0190] Example

[0191] Under continuous acceleration conditions from 0 km / h to 100 km / h, including but not limited to this condition, a rule-based shift control strategy is set proportionally. The embodiment utilizes cloud-edge interaction and multi-objective optimization to perform real-time optimization of the transmission system performance.

[0192] In economic indicators (J econ )aspect, Figure 3 In the comparative example, due to the lack of predictive adjustment at shift points of 2.5 seconds and 7.0 seconds, the drive efficiency of the transmission system experienced a sharp drop of 15%. The embodiment utilizes a cloud platform to predict that the longitudinal impact is relatively small under the current operating conditions. To maximize system energy utilization, it triggers the weight adjustment of the dynamic weighting module, adjusting the economic weight coefficient ε in the comprehensive driving cost function. econ Adjusted to 0.6, under this weight, the cloud platform combines gear hold trend scores to generate strategy parameters, increasing the peak drive efficiency of the transmission system to 91%; in terms of comfort index (J... comf )aspect, Figure 4 The peak longitudinal impact force at the moment of gear shift is nearly 13 m / s. 3 Far exceeding 10 m / s 3 The comfort threshold; the example demonstrates better active damping control performance. When the cloud platform predicts that the longitudinal impact exceeds the threshold, it triggers dynamic weighted adjustment, adjusting the comfort weight coefficient ε. comfThe amplitude was adjusted to 0.8 for intervention. Around 2.9 seconds, the amplitude abrupt change in the curve of the example was due to the algorithm predicting that the transmission system was about to enter the resonance zone. The system adjusted the control variable to damping gain to rapidly dissipate the mechanical resonance, thereby suppressing the peak longitudinal impact force throughout the entire process to 10 m / s². 3 Within; in dynamic indicators (J perf )aspect, Figure 5 The example demonstrates the ability to accurately track the desired driving force. The example is adjusted due to the optimization process, but the peak value of the driving force deviation is significantly reduced compared to the comparative example.

Claims

1. A method for dynamic feature localization and analysis of electric vehicle transmission systems using cloud-edge interaction, characterized in that, Includes the following steps: Step (1): Collect data from multiple sources; Step (2): Spatiotemporal registration and fusion of the multi-source data collected in step (1) are performed to construct fused data; Step (3): Upload the fused data to the cloud platform based on the trigger condition mechanism; Step (4): On the cloud platform, based on the uploaded fusion data, construct micro-behavior vectors, intentional behavior vectors, and environmental behavior vectors. Based on the behavior vectors, construct dynamic feature indicators representing the economy, comfort, and power of the transmission system respectively. Construct dynamic feature vectors from the dynamic feature indicators. Construct the feature expectation base value to represent the operating state of the transmission system and calculate the deviation of the dynamic feature indicators from the feature expectation base value. Calculate the individual contribution value of each component in the behavior vector to the deviation value and extract the data component with the largest individual contribution value. Step (5): Construct sub-cost functions for economy, comfort and power; merge the sub-cost functions to construct a comprehensive driving cost function; minimize the comprehensive driving cost function value based on the data component with the largest single contribution value extracted in step (4); generate a set of strategy parameters including energy management weight, shift rule correction factor and torque compensation coefficient based on the optimization results; Step (6): The vehicle terminal receives the strategy parameter set generated in step (5) and updates the vehicle control parameters.

2. The method for dynamic feature localization and analysis of electric vehicle transmission systems with cloud-edge interaction according to claim 1, characterized in that, The multi-source data in step (1) includes: transmission system data as the main data, and on-board data including drive system data and vehicle status data as auxiliary data, as well as roadside data composed of roadside driving environment data.

3. The method for dynamic feature localization and analysis of electric vehicle transmission systems with cloud-edge interaction according to claim 2, characterized in that, Step (2) specifically involves: Step (21): Time registration: Based on the Precision Time Protocol (PTP), the clock reference of the unified transmission system and each acquisition device is established; Step (22): Spatial registration: Establish a unified vehicle coordinate system, map the data collected by the sensors to the vehicle coordinate system, and use map matching technology to associate the real-time position of the vehicle with the map data; Step (23): Obtain bus direct acquisition data from the spatiotemporally registered data, including gear position signal, motor angular velocity, motor output torque, wheel end angular velocity, vehicle speed, longitudinal acceleration, steering wheel angle, pedal position, distance to the stop line of the traffic light ahead, and the remaining time of the green light of the traffic light ahead. Estimate state information including wheel end torque, road slope, curvature, and adhesion coefficient based on the direct acquisition data. Step (24): Standardize and encapsulate the directly collected data and estimated status information to construct fused data for dynamic feature analysis of the transmission system.

4. The method for dynamic feature localization and analysis of cloud-edge interactive electric vehicle transmission system according to claim 3, characterized in that, Step (21) is as follows: Step (211): Propagation delay estimation: The average propagation delay is calculated using a weighted statistical estimation method within a sliding window. ； In the formula, T delay The average propagation delay is denoted by N; the number of samples within the preset observation window is denoted by α; the asymmetry factor of the communication link path is denoted by α, which is used to compensate for the delay offset caused by the difference in uplink and downlink bandwidth; k is the sampling sequence of the synchronization period; T is the average propagation delay. 1,k T 2,k These represent the times when the master device sends the k-th synchronization message and when the slave device receives the k-th synchronization message, respectively; T 3,k T 4,k These are the times when the slave device sends the k-th delay request and when the master device receives the k-th delay request, respectively. Step (212): Clock offset calculation: The clock offset is calculated using an error feedback correction model. ； In the formula, offset k λ is the clock offset; λ∈(0,1] is the filter gain coefficient, used to balance the weights of historical synchronization trajectories and current sampled values ​​to filter out transient network noise; offset k-1 This is the clock offset from the previous synchronization cycle; Step (213): Time correction and frequency drift compensation: A frequency drift function is introduced to eliminate the cumulative timing error caused by the difference in hardware crystal oscillator frequency between synchronization cycles. The final correction formula is defined as follows: ； In the formula, T corrected The precise time synchronized with the master clock after device calibration; T local t is the device's current local time; t is the system's current time; t sync ξ is the reference time of the last synchronization operation; ξ is the integral variable, representing the time elapsed from the last synchronization to the current time; ρ(ξ) is the frequency drift function characterizing the difference in hardware crystal oscillators.

5. The cloud-edge interactive method for dynamic feature localization and analysis of electric vehicle transmission systems according to claim 4, characterized in that, Step (4) is as follows: Step (41): Extract gear position signal, wheel end torque, wheel end angular velocity, motor output torque, and motor angular velocity information to construct a micro behavior vector; extract longitudinal acceleration information to construct an intentional behavior vector; extract pedal position, distance to the stop line of the traffic light ahead, remaining green light time of the traffic light ahead, road slope, curvature, and adhesion coefficient information to construct an environmental behavior vector; Step (42): Construct dynamic characteristic indicators representing the economy of the transmission system based on the micro-behavioral vector, including the driving efficiency of the transmission system and the gear holding trend score; construct dynamic characteristic indicators representing the comfort of the transmission system based on the intentional behavior vector, including longitudinal impact, which is determined based on longitudinal acceleration; construct dynamic characteristic indicators representing the power performance of the transmission system based on the environmental behavior vector, including driving force deviation, which is determined based on the difference between the expected driving force and the real-time output driving force of the vehicle. Step (43): Construct a dynamic feature vector composed of dynamic feature indicators of economy, comfort and power; retrieve and match historical samples that match the current operating conditions in the historical database of the cloud platform; calculate the expected base value of the feature based on the actual recorded values ​​of the dynamic feature indicators in the dynamic feature vector of the matched historical samples and the similarity weight with the current operating conditions; calculate the deviation of the dynamic feature indicators from the expected base value of the feature; calculate the individual contribution value of each component in the behavior vector to the deviation; and take the component with the largest individual contribution value as the dominant factor.

6. The method for dynamic feature localization and analysis of cloud-edge interactive electric vehicle transmission system according to claim 5, characterized in that, The drive efficiency of the transmission system in step (42) is calculated as follows: ； In the formula, η t For the drive efficiency of the transmission system; T wheel ω is the wheel end torque; wheel T is the wheel end angular velocity; motor ω is the output torque of the motor. motor This refers to the angular velocity of the motor. The method for extracting the gear position trend score is as follows: ； In the formula, S gear The gear position maintains a trend score; f(·) is a normalized membership function that maps the input to the [0,1] interval; ε1 and ε2 are weighting coefficients; Δω motor ω represents the motor angular velocity deviation. opt (T motor (t) represents the current output torque T. motor Theoretically optimal economic angular velocity for the lower motor; i g (t) represents the actual gear ratio in the current gear signal of the system; i opt (t) represents the theoretically optimal transmission ratio; The longitudinal impact intensity in step (42) is calculated as follows: ； In the formula, j represents the longitudinal impact intensity; Δa represents the change in longitudinal acceleration; and Δt represents the time interval. The calculation process for "driving force deviation" in step (42) is as follows: The comprehensive road resistance potential is constructed based on the environmental behavior vector and calculated as follows: ； In the formula, denoted as , where is the comprehensive road resistance potential; s0 is the current position; L is the aiming distance; ω(s) is the spatial weighting function varying with s; G(s) is the road slope at s; C(s) is the road curvature at s; Ф μ (s) represents the adhesion coefficient related term; μ ref The reference adhesion coefficient is μ(s); the adhesion coefficient at position s is ε. g ε c and ε μ These are the weighting coefficients for road surface slope, curvature, and adhesion coefficient, respectively. The driving intention factor is derived from the standard deviation of the rate of change of pedal position over a short period of time, as follows: ； In the formula, The rate of change of pedal position; Pedal i Let be the position of the pedal sampled at time t; σ represents the mean rate of change of pedal position; σ is the driving intention factor. The traffic light traffic status is calculated as a vector with two elements, as follows: ； In the formula, V TL Traffic light status; χ TL For valid traffic light indication; L stopline T is the distance to the stop line of the traffic light ahead. green_rem C represents the remaining green time of the traffic light ahead; max To presuppose a sufficiently large constant; Define the valid indicator for traffic lights: ； In the formula, χ TL The value is 1 when traffic light information is available and can be obtained; the value is 0 when there is no traffic light or it is not available. The cloud platform derives the desired acceleration based on traffic light status, and the calculation is as follows: ； In the formula, a des V is the desired acceleration. current Current vehicle speed; The desired driving force is derived by combining the driving intention factor and calculated as follows: ； In the formula, F des Driven by expectations; is the intent sensitivity constant; m is the vehicle mass; F resist The driving resistance is calculated using the comprehensive road resistance potential. The driving force deviation is calculated as follows: ； In the formula, ΔF is the driving force deviation; i g i is the gear ratio; i0 is the main reduction ratio; r is the effective radius of the wheel.

7. The cloud-edge interactive method for dynamic feature localization and analysis of electric vehicle transmission systems according to claim 6, characterized in that, Step (43) is as follows: Step (431): Under synchronous time, the matrix form of the dynamic feature vector is expressed as follows: ； In the formula, Y(t) is the dynamic characteristic vector; η(t) is the drive efficiency of the transmission system at the current moment; S gear (t) represents the gear holding trend score at the current moment; j(t) represents the longitudinal impact at the current moment; ΔF(t) represents the driving force deviation at the current moment; The matrix form of the behavior vector is expressed as follows: ； In the formula, X(t) is the action vector; X micro (t) is the microscopic behavior vector; X intent (t) is the intentional behavior vector; X env (t) represents the environmental behavior vector; The components of the behavior vector are expanded as follows: ； In the formula, i g (t) represents the actual gear ratio in the current gear signal of the system; T wheel (t) represents the current wheel-end torque; ω wheel (t) represents the current wheel end angular velocity; T motor (t) represents the motor output torque at the current moment; ω motor (t) represents the current angular velocity of the motor; a(t) represents the current longitudinal acceleration; Pedal(t) is the current pedal position; G(s) is the road gradient at the vehicle's current position s; C(s) is the road curvature at the vehicle's current position s; μ(s) is the road adhesion coefficient at the vehicle's current position s, s = s(t); L stopline (t) represents the distance between the current vehicle and the stop line at the traffic light ahead; T green_rem (t) represents the remaining green time of the traffic light ahead; Step (432): The expected base value of the dynamic feature index is calculated as follows: ； In the formula, y n,exp (t) represents the expected baseline value of the feature; H represents the total number of samples retrieved and matched by the cloud platform from the historical database that are in a similar working condition to the current time t; h represents the index of the matched historical sample number; ρ h y represents the high-dimensional similarity weight between the h-th historical matching sample and the current real-time operating condition; n (h) This represents the actual recorded value of the nth dynamic feature indicator in the hth historical matching sample. Step (433): The deviation of the dynamic feature index from the expected base value is calculated as follows: ； In the formula, Δy n (t) represents the deviation; y n (t) represents the actual calculated value of the nth feature index component in the dynamic feature vector Y(t) of the system at the current moment; y n,exp (t) represents the expected base value of the nth feature index component, from which the expected base value x of the i-th data component in the behavior vector is derived. i,exp (t); Step (434): The individual contribution value of the i-th data component in the behavior vector X(t) to the deviation of the corresponding feature index is calculated as follows: ； In the formula, S n,i (t) represents the local sensitivity coefficient of the i-th data component of the behavior vector X(t) with respect to the n-th feature index component of the dynamic feature vector Y(t), and the right-hand side represents the partial derivative of the n-th feature index function with respect to the i-th behavior component; C n,i (t) represents the individual contribution value; x i (t) represents the i-th data component in the behavior vector X(t); x i,exp (t) represents the feature expectation base value of the i-th data component obtained by weighting based on historical matching samples; N x S represents the total dimension of the behavior vector X(t); n,q (t) is the local sensitivity coefficient of the q-th action vector component; x q (t) represents the actual collected value of the q-th data component; x q,exp (t) represents the characteristic expectation base value of the q-th data component.

8. The method for dynamic feature localization and analysis of electric vehicle transmission systems with cloud-edge interaction according to claim 7, characterized in that, Step (5) is as follows: Step (51): Construct sub-cost functions for economy, comfort, and dynamics: The economic cost function is based on the drive efficiency η of the transmission system. t Maintaining trend with gear position rating S gear Perform nonlinear modeling: ； In the formula, J econ ε3 is the cost function of economic efficiency; ε4 are the distribution coefficients of economic efficiency. ε3 is used to adjust the penalty of the system when the transmission efficiency deviates from the ideal range, and ε4 is used to adjust the degree of suppression of frequent gear shifting. The comfort cost function is constructed based on the longitudinal impact j: ； In the formula, J comf ε is the comfort cost function; ε5 and ε6 are comfort smoothness operators, where ε5 is the penalty coefficient used to adjust the absolute amplitude of a single longitudinal impact, and ε6 is the integral operator used to penalize the rate of change of impact; j(t) is the longitudinal impact at the current moment; j(τ) is the predicted value of the longitudinal impact. The dynamic cost function takes the driving force deviation ΔF as its core input: ； In the formula, J perf ε7 and ε8 are dynamic performance cost functions; ε7 is the basic penalty weight used to adjust the driving force deviation ΔF under normal cruise conditions, and ε8 is the nonlinear gain term dynamically adjusted based on the current vehicle speed and the desired acceleration. Step (52): The cloud platform integrates the sub-cost functions of economy, comfort, and power performance into a comprehensive driving cost function through a dynamic weighting module: ； In the formula, J total For the comprehensive driving cost function; ε econ ε comf ε perf These are the weighting coefficients for economy, comfort, and power output by the dynamic weighting module, and the sum of these weighting coefficients is 1. Step (53): The cloud platform minimizes the comprehensive driving cost function based on the data with the largest single contribution value extracted in step (4); the strategy parameter set is obtained by the cloud platform by minimizing the comprehensive driving cost function value and is mapped to the dynamic characteristic index in step (4), including energy management weight, shifting rule correction factor and torque compensation coefficient.

Images