Search

CN-121980520-A - Driving behavior and working condition coupling durable health management system for intelligent automobile

CN121980520ACN 121980520 ACN121980520 ACN 121980520ACN-121980520-A

Abstract

The invention relates to a driving behavior and working condition coupling durable health management system for an intelligent automobile, and belongs to the technical field of intelligent automobile power system durable management. The system comprises a coupling scene mode dividing unit for collecting driving behavior and working condition characteristics, a multi-source data lightweight fusion unit for outputting a high-quality data set through feature level pre-fusion and decision level lightweight fusion, a coupling factor decoupling and nonlinear prediction unit for stripping independent influence factors and retaining coupling residual factors, a nonlinear model for calculating and dynamically updating a life prediction value, and a prediction control linkage health management unit for dividing health grades, outputting an adaptation control strategy and correcting optimization parameters and strategies through feedback. The driving behavior and working condition coupling scene accurate identification is realized, the multisource data fusion efficiency and the durability prediction precision are improved, and the running reliability and the running safety of the power system are ensured.

Inventors

  • DENG WENBIN

Assignees

  • 慧勒科技(上海)股份有限公司

Dates

Publication Date
20260505
Application Date
20260309

Claims (12)

  1. 1. Driving behavior and working condition coupling durable health management system facing intelligent automobile, which is characterized by comprising: The system comprises a coupling scene mode dividing unit, a clustering analysis unit, a endurance influence rule drawing unit, a driving behavior feature and a working condition feature, wherein the driving behavior feature and the working condition feature are acquired and integrated to form a multidimensional feature vector; The multi-source data lightweight fusion unit is used for carrying out hierarchical fusion processing on the acquired real-time sensor data, driving behavior data and working condition data, carrying out feature level pre-fusion on the sensor data, carrying out redundant information filtering and noise suppression on the sensor data, extracting key features, carrying out decision level lightweight fusion on the sensor data, integrating the preprocessed sensor data with the driving behavior data and the working condition data, and outputting a fusion data set; The coupling factor decoupling and nonlinear prediction unit is used for separating independent influence factors of each factor on the endurance attenuation of the component based on the fusion data set and preserving multi-factor coupling residual factors; The prediction control linkage health management unit divides health grades of the battery pack and the electric drive system based on the life prediction value and a preset health judgment rule, outputs corresponding control strategies aiming at different health grades, acquires part state data after the control strategies are executed, corrects operation related parameters as feedback information, and optimizes subsequent control strategies.
  2. 2. The system of claim 1, wherein the specific process of integrating the two types of features to form the multidimensional feature vector is as follows: firstly, respectively carrying out outlier rejection and standardized pretreatment on the obtained driving behavior characteristics and the working condition characteristics; integrating the numerical range and the time dimension standard of the two types of features, and ordering rules according to preset feature priorities; And sequentially performing dimension splicing on the two types of preprocessed features to form a multidimensional feature vector capable of representing the coupling characteristics of driving behaviors and working conditions.
  3. 3. The system according to claim 1, wherein the specific process of dynamically identifying and dividing the coupled scene modes according to the clustering result is as follows: counting the feature distribution commonality and the discrete degree of each data cluster after clustering, and verifying and eliminating abnormal data clusters through the feature similarity in the clusters; Combining scene characteristics of actual running of the intelligent automobile, and constructing a matching rule of scene types and data cluster characteristics; and identifying the coupling scene types corresponding to the data clusters according to the matching result, and finally dynamically dividing the coupling scene modes according to the differences of the scene types.
  4. 4. The system of claim 1, wherein the specific process of the association corresponding relation between the preset coupling characteristic and the attenuation rate of the battery pack and the electric drive system component is that the internal association rules of the coupling characteristic and the component attenuation rate under different coupling scene modes are mined through a correlation analysis method based on the full life cycle historical operation data and the attenuation record of the key power component, and an association corresponding system of the coupling characteristic and the component attenuation rate is established by setting an association threshold and a corresponding rule based on the association rules.
  5. 5. The system of claim 1, wherein the feature level pre-fusion is performed by performing validity screening on acquired real-time sensor data, removing invalid data beyond a normal value range, suppressing noise interference in the data by adopting an adaptive noise reduction algorithm, and extracting core sensitive features from the processed sensor data based on component durability influence factor analysis to form a feature level pre-fusion result.
  6. 6. The system of claim 1, wherein the decision-level lightweight fusion is characterized by comprising the specific processes of performing time stamp alignment processing on feature level pre-fusion results, driving behavior data and working condition data, eliminating data time sequence deviation, establishing a data importance assessment system, presetting weight proportion of various data to whole vehicle endurance analysis, and sequentially integrating various data by adopting a lightweight fusion algorithm according to the weight proportion to generate a fusion data set capable of reflecting the running state of the whole vehicle.
  7. 7. The system according to claim 1, wherein the specific process of separating the independent influence factors of each factor on the durable attenuation of the component to preserve the coupling residual factors of multiple factors is as follows: processing the fusion data set through a multi-factor decoupling algorithm, and directly stripping independent influence factors of three core factors of driving behavior, working condition and working condition, which act on the durable attenuation of the component respectively; Synchronously extracting superposition influence parts generated by interaction of three types of core factors; and reserving the superposition influence part as a multi-factor coupling residual factor.
  8. 8. The system according to claim 1, wherein the specific process of constructing the nonlinear endurance prediction model is: comprises an input layer, a nonlinear mapping layer and an output layer; the input is a history independent influence factor, a coupling residual factor and corresponding attenuation data of the key power component, and the output is a time sequence prediction result of the attenuation rate of the key power component; The input layer performs dimension alignment and standardization pretreatment on the received characteristic data, the nonlinear mapping layer adopts a multi-layer perceptron framework, embeds nonlinear activation functions to perform iterative conversion layer by layer on the pretreated characteristics, and the output layer performs integration operation on the mapped characteristics to generate a time-sequence prediction result.
  9. 9. The system according to claim 1, wherein the specific process of calculating the life prediction value of the battery pack and the electric drive system is as follows: inputting independent influence factors and coupling residual factors obtained through real-time separation into the nonlinear endurance prediction model, and obtaining attenuation trend data of the key power component through calculation; And calculating the residual life predicted value of the key power component by combining the initial design life parameter and the historical attenuation data of the component through a trend extrapolation method.
  10. 10. The system according to claim 1, wherein the specific process of dividing the health grade of the battery pack and the electric drive system is to set a plurality of gradient health grade intervals and corresponding judging thresholds according to a preset health judging rule and combining the performance parameter thresholds of the key power components, compare the calculated life prediction value with each grade judging threshold one by one, and determine the health grade of the key power components according to the comparison result.
  11. 11. The system of claim 1, wherein the specific process of outputting the corresponding control strategies for different health grades is to construct a matching strategy library of the health grades and the control strategies in advance, wherein the matching strategy library comprises strategy types and basic parameters corresponding to the health grades, and the corresponding matched control strategies are called according to the classified health grades of the key power components, and fine calibration is carried out on the basic parameters of the strategy by combining real-time operation data.
  12. 12. The system of claim 1, wherein the dynamic updating of the input factors according to the real-time fusion data comprises the steps of acquiring a newly generated fusion data set in real time, establishing a factor change monitoring mechanism, tracking the numerical changes of the independent influence factors and the coupling residual factors in real time, replacing old factor data with factor data corresponding to the newly generated fusion data set when the factor numerical change exceeds a preset fluctuation range, and synchronously inputting the updated factors into the nonlinear durable prediction model.

Description

Driving behavior and working condition coupling durable health management system for intelligent automobile Technical Field The invention belongs to the technical field of intelligent automobile power system durability management, and particularly relates to an intelligent automobile-oriented driving behavior and working condition coupling durability health management system. Background Along with the deep transformation of the automobile industry to the electric and intelligent directions, the intelligent electric automobile has become the main development flow in the global traffic field, and the durability of a core power system (mainly comprising a battery pack and an electric drive system) directly determines the running reliability, the safety and the full life cycle use cost of the whole automobile, so that the intelligent electric automobile is one of key indexes for measuring the core competitiveness of the automobile; compared with the traditional fuel oil automobile, the running state of the power component of the intelligent electric automobile is more easily influenced by the coupling of a complex environment and a dynamic working condition, especially the coupling effect of the randomness of driving behaviors (such as rapid acceleration, rapid deceleration, frequent start and stop and the like) and the diversity of driving working conditions (such as urban congestion road conditions, high-speed cruising road conditions, ramp driving road conditions, extreme temperature and humidity environments and the like), the charge and discharge cycle loss of a battery pack, the mechanical abrasion and heat fading of an electric drive system can be remarkably increased, the attenuation rate of the component is caused to present strong nonlinear characteristics, and the serious challenge is brought to the durable management of the power system. At present, the demand of the automobile industry for durability management of a power system is changed from a traditional passive maintenance mode to an active predictive maintenance mode, the purpose of avoiding the failure risk of a part in advance through real-time monitoring and accurate prediction is achieved, therefore, a series of exploration is carried out in the related technical field, a durability management scheme based on sensor monitoring, data analysis and model prediction is provided, and the prior art generally realizes the residual life assessment of the part by collecting sensor data (such as battery voltage, current, temperature, rotation speed, torque, vibration and the like of an electric drive system) in the running process of the vehicle, driving behavior related data and working condition basic data and combining a durability prediction model, and accordingly establishes a maintenance or control strategy. However, in the prior art, when coping with the durability management requirement under the driving behavior and working condition coupling scenario, there are still a plurality of technical defects to be solved urgently, which are specifically expressed in the following aspects: Firstly, the prior proposal singly considers the influence of driving behaviors or working conditions on durability, and fails to fully excavate the characteristic association under the coupling action of the driving behaviors or the working conditions, so that the cognition of a part attenuation mechanism is deviated, and the accuracy of durability prediction is affected; Secondly, the multi-source data fusion link has obvious defects, on one hand, redundant information and noise interference in sensor data cannot be effectively filtered, so that the data quality is uneven, on the other hand, complex algorithms are adopted in the fusion process, the light weight requirement of real-time monitoring of the whole vehicle is difficult to adapt, the coupling characteristics of driving behaviors and working conditions are not fully integrated in the fusion data set, and comprehensive data support cannot be provided for subsequent endurance analysis. In summary, the existing technology for managing the durability of the power system of the intelligent electric automobile is difficult to effectively cope with complex challenges caused by coupling of driving behaviors and working conditions, and has the problems of inaccurate data fusion, insufficient excavation of coupling features, poor suitability of a prediction model, imperfect health management closed loop and the like, so that the durability prediction precision is low, the pertinence of a control strategy is not strong, and the high requirement of the intelligent electric automobile on the durability of the power system cannot be met. Disclosure of Invention In order to solve the problems in the prior art, the invention provides a driving behavior and working condition coupling durable health management system for an intelligent automobile, The aim of the invention can be achieved by the following technical schem