A method and system for optimizing energy efficiency of an electric vehicle based on failure prediction

Abstract

The application provides an energy efficiency optimization method and system for an electric vehicle based on fault prediction, the method comprising: collecting running data of the electric vehicle in real time, predicting a fault probability of the electric vehicle according to the running data, constructing an optimization objective function comprising minimization of energy consumption cost, minimization of battery life attenuation and minimization of travel time, and setting constraint conditions of the optimization objective function according to the fault probability; solving the optimization objective function by using a local model of the electric vehicle to obtain a charging strategy and a driving strategy of the electric vehicle, and pushing the charging strategy and the driving strategy to a user end; recording a user's adoption rate and adjustment of the pushed strategy as a log file, removing location information and user information in the log file to obtain desensitized data, uploading the desensitized data to a cloud end, calculating a difference weight of each user, aggregating local models of all users according to the difference weight, and updating the local models of all electric vehicles according to knowledge of the aggregated models.

Description

Technical Field

[0001] This invention belongs to the field of new energy vehicle technology, specifically relating to an energy efficiency optimization method and system for electric vehicles based on fault prediction. Background Technology

[0002] The global new energy vehicle industry is expanding rapidly, leading to a surge in demand for intelligent operation and maintenance for electric vehicles. As the most critical component of a vehicle, the accuracy of fault warnings and the level of energy efficiency management in the power battery system directly determine the vehicle's safety and economy. However, achieving a balance between efficient and accurate fault warnings and comprehensive intelligent energy efficiency optimization, and ensuring real-time collaboration between the two, remains a pressing issue for the industry.

[0003] Traditional battery management systems (BMS) have long relied on threshold alarm mechanisms based on single parameters such as voltage and temperature. This fails to capture abnormal signals in the early stages of major faults like thermal runaway, often resulting in irreversible battery damage by the time an alarm is triggered. Industry statistics show that false alarm rates consistently exceed 15% due to sensor noise and environmental interference, eroding user trust. These isolated warning logics lack effective integration with the vehicle's overall energy efficiency strategy. When the battery is faulty, the system cannot adjust charging power or driving recommendations in a timely manner, potentially accelerating battery degradation or exacerbating faults through extreme energy-saving strategies. Energy efficiency management also suffers from a single-dimensional approach: charging plans and driving prompts only consider static state of charge (SOC), ignoring dynamic electricity price fluctuations and lacking quantitative assessments of driving behaviors such as rapid acceleration and frequent braking. Single-objective optimization algorithms struggle to balance energy consumption costs, travel time, and battery life, leading to strategies that deviate from real-world driving scenarios. Furthermore, traditional solutions generally fail to incorporate real-time health status and potential fault risks into energy efficiency decisions, highlighting the conflict between energy saving and safety. To overcome the aforementioned limitations, in recent years the industry has begun to adopt cloud-based AI models on a large scale as a new technological approach. By aggregating massive amounts of vehicle data, cloud platforms can train fault diagnosis and energy efficiency optimization models in parallel, and then use OTA to feed the models or strategies back to the vehicle to achieve higher-dimensional joint analysis.

[0004] While cloud-based intelligent solutions can handle massive amounts of data, they still have some problems: network latency caused by full data transmission increases fault diagnosis response time to over 200ms, failing to meet automotive-grade real-time requirements; deploying high-performance computing units supporting the original model locally would drastically increase BMS hardware costs, hindering solution implementation. Furthermore, the heavy-load cloud architecture disrupts the high-frequency collaboration between fault warning and energy efficiency optimization—the vehicle can only periodically retrieve strategies, unable to dynamically adjust energy-saving targets based on instantaneous health status. Summary of the Invention

[0005] This invention proposes a method and system for optimizing the energy efficiency of electric vehicles based on fault prediction, which solves the problem of poor real-time performance of existing cloud-based intelligent solutions.

[0006] To address the aforementioned technical problems, this invention provides an energy efficiency optimization method for electric vehicles based on fault prediction, comprising the following steps:

[0007] Step S1: Collect real-time operating data of electric vehicles, predict the failure probability of electric vehicles based on the operating data, construct an optimization objective function that includes minimizing energy consumption cost, minimizing battery life degradation and minimizing travel time, and set constraints on the optimization objective function based on the failure probability;

[0008] Step S2: Solve the optimization objective function using the local model of the electric vehicle to obtain the charging strategy and driving strategy of the electric vehicle, and push the charging strategy and driving strategy to the user terminal.

[0009] Step S3: Collect the user's adoption rate and adjustment records of the push strategy as a log file, remove the location information and user information from the log file to obtain de-identified data, upload the de-identified data to the cloud, calculate the difference weight of each user based on the de-identified data, aggregate the local models of all users based on the difference weight, and update the local models of all electric vehicles based on the cross-vehicle knowledge in the aggregated model.

[0010] Preferably, the real-time acquisition of electric vehicle operating data in step S1 includes the following steps: acquiring raw data of electric vehicle battery voltage, temperature distribution and acceleration, normalizing the raw data using a sliding window, and constructing a time-series data matrix.

[0011] Preferably, in step S1, the electrochemical and thermodynamic features of the operating data are extracted using a local model of the electric vehicle, and the failure probability of the electric vehicle is predicted based on the electrochemical and thermodynamic features. The expression for predicting the failure probability is as follows:

[0012] ;

[0013] In the formula, This represents the probability of failure. For activation functions; This is the weight matrix of the fully connected layers in the local model; It is characterized by electrochemical properties; These are thermodynamic eigenvectors; This indicates the concatenation of electrochemical and thermodynamic eigenvectors; This is the bias term for the fully connected layer.

[0014] Preferably, the expression for the optimization objective function in step S2 is:

[0015] ;

[0016] ;

[0017] In the formula, To optimize the objective function; , , They are respectively , , The weights; This represents minimizing energy consumption costs; This indicates minimizing battery life degradation; This indicates minimizing the travel time; This represents the total discrete time steps within the planning time period; For time steps The charging power; For time step Real-time electricity price; This represents the average growth rate of the battery's internal resistance. The average state of charge during a charging or discharging cycle; For the driving time of electric vehicles; These are the weighting coefficients; The total charging time within the planned time period.

[0018] Preferably, in step S2, a fuzzy controller is used to dynamically adjust the weights in the objective function based on the battery health status and the user-set driving mode. The fuzzy rules used by the fuzzy controller are as follows:

[0019] (1) If the battery health status is healthy and the user's driving mode is economy priority, then increase The weight is reduced. and The weights;

[0020] (2) If the battery health status is healthy and the user's driving mode is balanced, then maintain , and The weights;

[0021] (3) If the battery health status is healthy and the user's driving mode is performance priority, then improve The weight is reduced. and The weights;

[0022] (4) If the battery health status is slightly degraded and the user's driving mode is set to economy priority, then increase The weight, maintain The weight is reduced. The weights;

[0023] (5) If the battery health status is slightly degraded and the user's driving mode is balanced, then increase The weight, maintain and The weights;

[0024] (6) If the battery health status is slightly degraded and the user's driving mode is performance mode, then increase and The weight is reduced. The weights;

[0025] (7) If the battery health status is severely degraded and the user's driving mode is set to economy priority, then reduce the power consumption. Weighting, increasing The weight, maintain The weights;

[0026] (8) If the battery health status is severely degraded and the user's driving mode is balanced, then increase The weight is reduced. and The weights;

[0027] (9) If the battery health status is severely degraded and the user's driving mode is performance priority, then increase The weight is reduced. and The weight.

[0028] Preferably, in step S3, after uploading the user's adoption rate and adjustment records of the push strategy as a log file to the cloud, reinforcement learning is used to periodically update the fuzzy rules based on the log file. The expression for updating the fuzzy rules is:

[0029] ;

[0030] In the formula, To enhance learning Value table; For state, , The state space includes battery health status, user driving mode preferences, current vehicle operating conditions, and the adoption rate of the last optimization suggestion. For action, , The action space includes the weights of energy consumption optimization targets, battery life targets, and range efficiency targets. The learning rate; As a reward; Discount factor; Indicates that the current action is being performed. The state at the next moment obtained after the environmental transition; Indicates the state The action chosen for the next moment.

[0031] Preferably, the constraints in step S1 include:

[0032] (1) When the failure probability When the faulty battery cell is located, the upper limit of the discharge current of the branch is lowered.

[0033] (2) When the failure probability At this time, the energy recovery strategy is paused, limiting the accelerator pedal response limit;

[0034] (3) When the failure probability At that time, the high-voltage relay is disconnected, controlling the liquid cooling system to operate at full power.

[0035] Preferably, the expression for calculating the difference weight of each user based on the de-identified data in step S3 is as follows:

[0036] ;

[0037] In the formula, For users i Difference weights; , , They are respectively , , Weighting coefficients; For users i The relative data volume factor of the vehicles in the current training period; For users i The quality reliability factor of vehicle-uploaded data; For users i The validation loss improvement factor after updating the local model of the vehicle; For users i The number of valid samples uploaded; These are weighting coefficients; For users i The proportion of outliers in the uploaded data; For users i Signal noise fluctuations in uploaded data; For users i The sampling interval missing rate; The validation loss of the aggregation model before downloading; For users i The validation loss after updating the local model of the vehicle.

[0038] Preferably, in step S3, knowledge across vehicles is extracted from the aggregated model, and an update package for the local model of all electric vehicles is generated using a differential compression algorithm. This package is then downloaded to the user's vehicle via over-the-air (OTA) technology, thereby updating the local model of the electric vehicle.

[0039] The present invention also provides an energy efficiency optimization system for electric vehicles based on fault prediction, which is based on the above-mentioned energy efficiency optimization method for electric vehicles based on fault prediction, and includes: a data acquisition module, a fault prediction module, an optimization strategy generation module, a cloud collaborative update module, and a human-computer interaction module.

[0040] The data acquisition module acquires real-time data on the vehicle's battery pack voltage, temperature distribution, and acceleration.

[0041] The fault prediction module extracts the vehicle's electrochemical and thermodynamic characteristics from the data collected by the data acquisition module and calculates the vehicle's fault probability.

[0042] The optimization strategy generation module: constructs a multi-objective optimization function, realizes multi-objective dynamic optimization under fault constraints, and outputs vehicle charging period suggestions, vehicle speed control range, power response parameters, etc.

[0043] The cloud-based collaborative update module calculates the user's differential weights based on the user's strategy preferences, aggregates the local models of all vehicles, updates the global model in the cloud, generates an incremental update package, and sends it to the vehicle to update the vehicle's local model.

[0044] The human-computer interaction module provides users with a visual representation of the vehicle's fault level heatmap, energy consumption optimization schemes, and battery health trends. It also allows users to adjust vehicle speed range offsets, manually modify charging periods, record user strategy adoption rates and behavioral deviations, and update the user's driver profile.

[0045] The beneficial effects of the present invention include at least the following:

[0046] 1. Optimized calculations are performed locally on the vehicle, avoiding network latency caused by uploading all data and meeting automotive-grade real-time requirements; at the same time, there is no need to stack high-performance computing units on the vehicle, significantly reducing hardware costs;

[0047] 2. By collecting adoption rates and adjustment records through a push-feedback closed loop, real user preferences are quantified into differential weights, thereby allowing the local model to continuously adapt to individual driving habits and reduce the gap in strategy implementation;

[0048] 3. The cloud only processes the desensitized weight and gradient information, which protects privacy and can use federated aggregation to generalize the driving habits of each user into cross-vehicle knowledge, feed it back into the local model of each vehicle, and realize the continuous upgrading of the local vehicle model without having to send back the original operating condition big data. Attached Figure Description

[0049] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram illustrating the decision logic of the multi-objective optimization algorithm according to an embodiment of the present invention;

[0051] Figure 3 This is a system architecture diagram of an embodiment of the present invention;

[0052] Figure 4 This is a schematic diagram illustrating the collaborative interaction between the fault prediction module and the optimization strategy generation module in an embodiment of the present invention. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0054] like Figure 1 As shown in the figure, this embodiment of the invention provides an energy efficiency optimization method for electric vehicles based on fault prediction, including the following steps:

[0055] Step S1: Collect real-time operating data of electric vehicles, predict the failure probability of electric vehicles based on the operating data, construct an optimization objective function that includes minimizing energy consumption cost, minimizing battery life degradation and minimizing travel time, and set constraints on the optimization objective function based on the failure probability.

[0056] Specifically, the system connects to the electric vehicle's powertrain bus via the OBD-II interface of the on-board diagnostic system (OBD-II) to collect real-time data on battery pack voltage, temperature distribution including individual cell temperature, and vehicle acceleration. The raw data is then processed using sliding window normalization to eliminate sensor dimensional differences and construct a time-series feature matrix. The system continuously monitors and records these parameters, providing data support for subsequent average temperature calculations and historical baseline construction. The data acquisition process follows the SAE J1939 protocol standard, and acceleration data from an inertial measurement unit (IMU) is used for auxiliary compensation of non-standard parameters.

[0057] Electrochemical and thermodynamic features are extracted from the operational data using a local model of electric vehicles. Preprocessed voltage / current time-series data is input into a one-dimensional convolutional layer (Conv1D, kernel_size=5), and local fluctuation features are captured using a sliding window.

[0058] ;

[0059] In the formula, These are the trainable convolutional kernel weights; for t Voltage sample value at time t.

[0060] After performing max pooling (Pool_size=3) to reduce the dimensionality of the output feature map, a trend map of internal resistance variation is generated.

[0061] Rearrange the two-dimensional spatial data of the temperature sensing array into a spatiotemporal sequence:

[0062] ;

[0063] In the formula, This represents a complete temperature time-series dataset consisting of observations from all temperature detection units at continuous time steps. Indicates the first Each temperature detection unit at time step The temperature measurement value reflects the dynamic change of the unit over time; This represents the total number of time steps obtained from sampling, which is the length of the time series data.

[0064] The aforementioned spatiotemporal sequence is input into a bidirectional LSTM network for spatiotemporal feature fusion. Through the outputs of the forward and backward networks of the bidirectional LSTM, a prediction matrix for the heat propagation rate is finally obtained. This matrix reflects the spatiotemporal evolution of the thermal runaway process, from... Key thermodynamic features such as maximum heat propagation rate, average heat propagation rate, location of heat diffusion center, and thermal runaway trend are extracted and integrated into a low-dimensional vector through dimensionality reduction processing such as principal component analysis or autoencoder. This vector comprehensively characterizes the thermal diffusion state and trend of the battery. Feature extraction is performed to obtain thermodynamic feature vectors. .

[0065] Meanwhile, based on the historical stable temperature data of the battery pack under different operating conditions continuously collected by the system, a dynamically changing temperature baseline is constructed. This baseline represents the expected average temperature under non-fault conditions based on historical operating data. It is determined by the battery's current state of charge (SOC) and... Dynamically adjust temperature warning threshold :

[0066] ;

[0067] In the formula, , As experimental calibration coefficients, in this embodiment of the invention, they are set , ; This represents the deviation of the current average temperature of the battery pack from its historical normal operating temperature baseline.

[0068] - ;

[0069] In the formula, The average temperature of the current battery pack is obtained in real time.

[0070] By integrating electrochemical and thermodynamic eigenvectors, the failure probability of electric vehicles is predicted through a fully connected layer of a local model.

[0071] ;

[0072] In the formula, This represents the probability of failure. The activation function is Sigmoid, which maps the output value to the range (0,1). This is the weight matrix of the fully connected layer; It is characterized by electrochemical properties; These are thermodynamic eigenvectors; This is the bias term for the fully connected layer.

[0073] Depending on the probability of a fault, the system triggers different response actions to ensure timely response to potential fault threats. Specifically, when the probability of a fault is between 0.4 and 0.6, the system sends a Diagnostic Trouble Code (DTC) via the CAN bus, requiring an execution time of less than 100 milliseconds; when the probability of a fault is between 0.6 and 0.8, the system triggers a voice warning and flashes a red indicator light on the instrument panel, requiring a response time of less than 50 milliseconds; when the probability of a fault is greater than or equal to 0.8, the system immediately disconnects the high-voltage relay and starts the liquid cooling system, requiring a response time of less than 10 milliseconds.

[0074] After completing the multimodal feature fusion and risk probability assessment through the above process, the diagnostic results are not only used to trigger the graded response mechanism, but also directly serve as one of the constraints for energy efficiency optimization. Specifically, when the fault probability exceeds 0.4, the system will temporarily lower the upper limit of the discharge current of the branch where the faulty cell is located, and simultaneously adjust the vehicle power output curve to limit the occurrence of high-load conditions; when the battery thermal diffusion trend is detected to exceed the preset threshold, the system will suspend the energy recovery strategy and adjust the cooling cycle interval to reduce system thermal stress. The above mechanisms ensure that the energy efficiency optimization strategy always operates within a controllable risk range, avoiding optimization behavior from exacerbating fault risks, thereby constructing a linkage control path of diagnosis-constraint-optimization to achieve the coupled synergy of vehicle operation safety and economy.

[0075] The optimization module is deployed on the vehicle-mounted embedded AI processor. Under safety constraints, it executes a multi-objective collaborative decision-making process with inputs such as battery health, electricity price fluctuations, driving behavior characteristics, and navigation road condition data to achieve comprehensive optimization of energy consumption costs, battery life, and trip efficiency.

[0076] like Figure 2 The diagram shown illustrates the decision logic of a multi-objective optimization algorithm according to an embodiment of the present invention. The multi-objective optimization function incorporates four types of dynamic parameters to construct an input vector space, including:

[0077] a) Battery health parameters: including battery capacity decay rate, internal resistance growth coefficient and polarization voltage offset;

[0078] b) Economic parameters: Time-of-use electricity price data are obtained from the power grid API, and an electricity price-location correlation matrix is ​​constructed by combining the location distribution of charging piles;

[0079] c) Behavioral characteristic parameters: Based on the user's historical driving data, the frequency of rapid acceleration, average braking intensity, and cruise speed preference are statistically analyzed.

[0080] d) Traffic parameters: Integrate real-time traffic flow, gradient changes, and path length information provided by the navigation system.

[0081] The battery health parameters mentioned above are directly quantified as the degree of battery degradation and used as safety constraints; the economic parameters are directly involved in the calculation and optimization of energy consumption costs; and the behavioral characteristic parameters and road condition parameters provide key basis for the dynamic weight adjustment of the optimization model, the generation of driving strategies, and the planning of charging routes.

[0082] The constructed input vector space is the underlying data foundation for the entire dynamic energy efficiency optimization strategy. It works together in the collaborative optimization process by directly using it as a variable in the objective function, influencing the weight of the objective function, applying optimization constraints, or serving as the basis for strategy generation, thereby achieving comprehensive and dynamic optimization of energy consumption, battery life, and range efficiency.

[0083] The expression for the constructed optimization objective function is:

[0084] ;

[0085] ;

[0086] In the formula, To optimize the objective function; , , They are respectively , , The weights; This represents minimizing energy consumption costs; This indicates minimizing battery life degradation; This indicates minimizing the travel time; This represents the total discrete time steps within the planning time period; For time step The charging power; For time step Real-time electricity price; The average growth rate of battery internal resistance is a key indicator of battery aging. This parameter is directly provided by the internal resistance growth coefficient in the battery health parameters. The average state of charge during a charging or discharging cycle; The driving time of electric vehicles is affected by road condition parameters such as real-time traffic flow and path length. This is a weighting coefficient used to balance the relative importance of charging time and driving time in the total trip time. Its value can be dynamically adjusted based on the user's driving mode preferences. The total charging time within the planned time period.

[0087] Step S2: Solve the objective function using the local model of the electric vehicle to obtain the charging strategy and driving strategy of the electric vehicle, and push the charging strategy and driving strategy to the user terminal.

[0088] Specifically, to further improve the accuracy of optimization decisions, a fuzzy logic controller dynamically adjusts the weights of the three optimization objectives. The fuzzy inputs for the objective weights include battery health status and user driving mode preferences. The fuzzy logic controller outputs real-time weight coefficients for the three objective functions based on a rule base, and adjusts the relative importance of each objective according to user mode and battery health status. Battery health status includes: healthy, slightly degraded, and severely degraded; user driving mode preferences include: economy priority, balanced mode, and performance priority. These two combinations form nine fuzzy rule terms as shown in Table 1.

[0089] Table 1 Fuzzy Rule Base

[0090]

[0091] In actual operation, the system continuously collects vehicle operation data, user behavior feedback, and data on the effectiveness of optimization strategies. Machine learning methods are used to iteratively optimize and adaptively adjust the initial rule base. By analyzing the impact of weight allocation under different battery states and driving behaviors on overall benefits or user satisfaction in historical data, the system automatically learns and refines the rules. Specifically, by analyzing a large amount of real-world driving data, the system identifies significant impacts on battery life under specific temperature ranges or load modes, and then dynamically adjusts relevant rules to achieve refined weight allocation.

[0092] For fuzzy inputs and fuzzy outputs, corresponding fuzzy sets are defined, and membership functions are set for each fuzzy set. Specifically, health status can be divided into good, moderate, and poor; weight adjustment can be divided into significantly increased, increased, maintained, decreased, and significantly decreased. The membership function maps precise input values ​​to the membership degrees of the fuzzy sets, enabling the rule base to handle continuously changing inputs.

[0093] The rule base, constructed and continuously optimized using the aforementioned methods, enables the fuzzy logic controller to dynamically and intelligently adjust the weights of each objective function in multi-objective optimization based on the vehicle's real-time operating conditions, thereby generating energy efficiency optimization strategies more suited to the current scenario. Based on this, the weights of the aforementioned optimization objectives are integrated, and a multi-objective Pareto solution set is constructed using the improved fast elite multi-objective genetic algorithm NSGA-II. Through a human-machine interface terminal, customized driving and charging strategies, including recommended speed ranges, charging schedules, and power response adjustment parameters, are recommended to the driver.

[0094] Simultaneously, based on real-time data, optimization calculations are performed every 5 minutes. The optimization process generates driving speed range and charging period suggestions based on the latest battery status, environmental conditions, and user preferences. These suggestions are displayed through the in-vehicle human-machine interface (HMI), allowing users to manually adjust parameters as needed, such as the offset of the driving speed range or charging period. The user's adjustment deviation (δ) is recorded and used for policy updates via reinforcement learning. Reinforcement learning algorithms are used to adjust the weight parameters or membership function shape of the fuzzy rules, thereby making the weight allocation of the fuzzy controller more aligned with user preferences and overall performance goals in the next run. The expression for updating the rule base through reinforcement learning is:

[0095] ;

[0096] In the formula, To enhance learning Value table; For state, , The state space includes battery health status, user driving mode preferences, current vehicle operating conditions, and the adoption rate of the last optimization suggestion. For action, , The action space includes the weights of energy consumption optimization targets, battery life targets, and range efficiency targets. The learning rate; As a reward; Discount factor; Indicates that the current action is being performed. The state at the next moment obtained after the environmental transition; Indicates the agent's state The next action is selected based on the strategy.

[0097] To ensure that optimization actions are performed within a controlled range of failure risks, risk factors are loaded simultaneously during the optimization process. Specifically, risk factors It is the probability of failure. The semantic mapping is used as a risk constraint indicator in the energy efficiency optimization module. The system calculates the failure probability value, which is then combined with the dynamic early warning threshold. Risk factors are calculated using the following functional relationship. :

[0098] ;

[0099] Among them, the function This represents the risk response strategy set in the actual system, mainly manifested as segmented control or conditional judgment, including:

[0100] (1) When the failure probability When the faulty battery cell is located, the upper limit of the discharge current of the branch is lowered. Specifically, the upper limit of the branch discharge current is reduced by 20% to 35% linearly, the regenerative braking power is limited to less than 20% of the original value, the upper limit torque of the drive pedal is reduced by 10% to 15%, the cooling pump duty cycle is increased to 70% to 80%, the fan duty cycle is increased to 60% to 70%, and the charging rate is limited to below 0.5C.

[0101] (2) When the failure probability At this time, the energy recovery strategy is suspended, the accelerator pedal response limit is restricted, and the battery SOC is maintained between 20% and 90%, with the cell temperature controlled between 5°C and 45°C. Specifically, this is manifested as follows: regenerative braking power is directly shut off; the upper limit of drive torque is limited to 60% to 75% of the rated value, and the upper limit of effective pedal travel is limited to 60% to 70%; the branch discharge current is further reduced to 40% to 60%; the cooling pump speed is increased to 90% duty cycle, the fan to 85% duty cycle, and a uniform temperature cycle is implemented; the charging rate is limited to below 0.3C, and the DC fast charging power is limited to below 30kW.

[0102] (3) When the failure probability When the high-voltage relay is activated, the liquid cooling system will operate at full power. Immediately disconnect the main positive / main negative relays and keep the pre-charge relay disconnected; the liquid cooling pump and fan will operate at full power, and if compressor cooling is available, the frequency will be increased to the maximum until the cell temperature drops below 40°C; all drive and charging requests will be turned off, leaving only necessary low-load functions such as defrosting / dehumidification in the cabin; at the same time, a voice alarm will be triggered and the red light on the instrument panel will flash continuously, and at least 30 seconds of voltage, current, temperature, and DTC diagnostic data will be saved as a fault record.

[0103] Step S3: Collect the user's adoption rate and adjustment records of the push strategy as a log file, remove the location information and user information from the log file to obtain de-identified data, upload the de-identified data to the cloud, calculate the difference weight of each user based on the de-identified data, aggregate the local models of all users based on the difference weight, and update the local models of all electric vehicles based on the cross-vehicle knowledge in the aggregated model.

[0104] Existing battery life prediction methods are limited to static analysis based on a single data source and have not yet established a cross-vehicle data collaboration and continuous model iteration mechanism. The prediction results deviate significantly from the actual degradation patterns, which not only weakens the forward-looking judgment of energy efficiency optimization on battery health but also restricts the continuous evolution of fault diagnosis models.

[0105] To enhance the generalization ability and dynamic adaptability of fault diagnosis and optimization models, this invention proposes a cloud-based model training and distribution mechanism based on a fusion of federated learning and transfer learning. This mechanism aggregates parameters and transfers knowledge from multiple vehicle models without uploading sensitive original user data, and synchronizes model updates to the vehicle-mounted embedded AI processor via over-the-air (OTA) updates, enabling the self-evolution of the edge intelligence module.

[0106] The cloud-based model update mechanism includes the following steps:

[0107] Step S31: Upload data summaries daily, including abnormal segments, energy efficiency strategy execution records, and user preference adjustment logs. The data structure does not contain location information or user identification to ensure privacy compliance.

[0108] Step S32: The cloud platform, based on the FedAvg algorithm, fuses local model gradients from multiple vehicles to train a unified fault diagnosis model and energy efficiency optimization model. A difference weighting coefficient is introduced during the model aggregation process. This reflects the differences in training data volume, data quality, and model effectiveness for each vehicle. Specifically, it involves calculating the difference weighting coefficient. The table is as follows:

[0109] ;

[0110] In the formula, For users i Difference weights; , , These are weighting coefficients, used to balance the contributions of different indicators in the calculation of differential weights, and to balance the impact of data volume, data quality, and model effectiveness. The sum of these three factors is 1, and they remain stable throughout the training period, only being adjusted when significant data fluctuations occur or business objectives change. Specifically, This represents the weight assigned to the relative data volume factor, reflecting the user's... The importance of the scale of uploaded data in aggregation; This represents the weight assigned to the data quality confidence factor, used to highlight the contribution of high-quality, low-noise data to the model aggregation results; This represents the weights assigned to the validation loss improvement factor, used to emphasize the user. The effectiveness of the local model in improving overall validation performance after the update.

[0111] For users i The relative data volume factor of the vehicles in the current training period; For users i The quality reliability factor of vehicle-uploaded data; For users i The validation loss improvement factor after updating the local model of the vehicle; For users i The number of valid samples uploaded; These are weighting coefficients; For users i The proportion of outliers in the uploaded data; For users i Signal noise fluctuations in uploaded data; For users i The sampling interval missing rate; The validation loss of the aggregation model before downloading; For usersi The validation loss after updating the local model of the vehicle.

[0112] Step S33: Based on the global model, extract the battery aging pattern, typical user driving behavior model and charging mode preference to achieve knowledge transfer across vehicle platforms.

[0113] Step S34: The updated results of the trained model are compressed into an update package using a differential compression algorithm. This update package contains only the changes in model weights and is pushed to the corresponding vehicle terminal via the vehicle-to-everything (V2X) communication link, where it is automatically loaded by the OTA upgrade subsystem.

[0114] Step S35: After the vehicle receives and loads the update, a quick test needs to be performed to ensure model compatibility and stability. Once the test is passed, the original model version will be automatically replaced. The update process is transparent to the user.

[0115] Through the aforementioned cloud-based collaboration mechanism, the system can continuously iterate and optimize the model while maintaining efficient edge response, ensuring that the fault identification logic and energy efficiency strategy always align with the latest vehicle operating conditions and user behavior characteristics, forming a continuous learning closed loop that integrates cloud and edge.

[0116] To enhance the system's interpretability and user acceptability, a human-computer interface (HMI) with feedback capabilities is incorporated into the system. This HMI visualizes diagnostic results and energy efficiency optimization suggestions for the user and records user responses, allowing user feedback to participate in the system model's adaptive learning process. The HMI module supports hierarchical information display technology and primarily includes the following functional components:

[0117] Risk Status Display Interface: Displays warning information such as the current fault level, abnormal cell temperature location, and battery health trend graph in a graphical manner.

[0118] Energy efficiency recommendation interface: dynamically presents the current recommended vehicle speed range, feasible charging window, SOC prediction curve and target optimization weight.

[0119] User preference adjustment module: Supports users to manually adjust suggested parameters, such as vehicle speed upper and lower limits, charging time period offset, etc. All user modification operations are recorded in real time and sent back to the optimization module as strategy offset.

[0120] Behavior adoption record module: The system builds driver behavior profiles based on user adoption rates and links them with the system's policy adoption reward mechanism to dynamically update and optimize the state-action mapping relationship of the model, thereby improving personalized adaptability.

[0121] When the system detects that user behavior deviates continuously from the policy recommendations, such as ignoring energy-saving prompts three times in a row or frequently increasing the acceleration limit, it will trigger an adaptive learning mechanism to reconstruct the behavior profile and adjust the fuzzy controller rule base, thereby realizing the evolution of policy recommendations from global general to personalized customization.

[0122] like Figure 3 As shown, this embodiment of the invention also provides an energy efficiency optimization system for electric vehicles based on fault prediction, which is implemented based on the above-mentioned energy efficiency optimization method for electric vehicles based on fault prediction, and includes: a data acquisition module, a fault prediction module, an optimization strategy generation module, a cloud collaborative update module, and a human-computer interaction module.

[0123] Data acquisition module: Real-time acquisition of vehicle battery pack voltage, temperature distribution, and acceleration data.

[0124] Fault prediction module: Extracts the vehicle's electrochemical and thermodynamic characteristics from the data collected by the data acquisition module, and calculates the vehicle's fault probability.

[0125] Optimization strategy generation module: Constructs a multi-objective optimization function to achieve multi-objective dynamic optimization under fault constraints, and outputs vehicle charging time period suggestions, vehicle speed control range, power response parameters, etc.

[0126] like Figure 4 As shown, this embodiment of the invention establishes an optimization strategy generation module that integrates battery health (SOH), real-time electricity price, driving behavior characteristics, and navigation road conditions under the safety state constraints provided by the fault prediction module. An improved NSGA-II algorithm is used to generate a Pareto optimal solution set, and a fuzzy logic controller dynamically adjusts the optimization weights of energy consumption cost, battery life, and travel time. During vehicle operation, the edge computing unit analyzes driving behavior data in real time, combining cloud-based electricity price prediction information and road condition updates, iteratively generating recommended driving speed ranges and charging time suggestions every 5 minutes. This optimization strategy fully integrates battery health information from the fault prediction module, ensuring that battery safety and lifespan are prioritized during the optimization process. It effectively balances the combined needs of economy, safety, and user experience, achieving a technological leap from static planning to dynamic adaptation and fault risk avoidance in global energy efficiency management.

[0127] Cloud-based collaborative update module: Calculates the user's difference weight based on the user's strategy preferences, aggregates the local models of all vehicles, updates the global model in the cloud, generates an incremental update package and sends it to the vehicle, thereby updating the vehicle's local model.

[0128] Human-computer interaction module: Visualizes vehicle fault level heatmaps, energy consumption optimization schemes, and battery health trends for users; allows users to adjust vehicle speed range offsets and manually modify charging periods; records user strategy adoption rates and behavioral deviations; and updates user driver profiles.

[0129] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described; only preferred embodiments of the present invention are illustrated. The descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the present invention. As long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0130] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this invention should be determined by the appended claims.

Claims

1. A method for optimizing the energy efficiency of electric vehicles based on fault prediction, characterized in that, Includes the following steps: Step S1: Collect real-time operating data of electric vehicles, predict the failure probability of electric vehicles based on the operating data, construct an optimization objective function that includes minimizing energy consumption cost, minimizing battery life degradation and minimizing travel time, and set constraints on the optimization objective function based on the failure probability; Step S2: Solve the optimization objective function using the local model of the electric vehicle to obtain the charging strategy and driving strategy of the electric vehicle, and push the charging strategy and driving strategy to the user terminal; Step S3: Collect the user's adoption rate and adjustment records of the push strategy as a log file, remove the location information and user information from the log file to obtain de-identified data, upload the de-identified data to the cloud, calculate the difference weight of each user based on the de-identified data, aggregate the local models of all users based on the difference weight, and update the local models of all electric vehicles based on the cross-vehicle knowledge in the aggregated model.

2. The energy efficiency optimization method for electric vehicles based on fault prediction according to claim 1, characterized in that: The real-time acquisition of electric vehicle operating data in step S1 includes the following steps: acquiring raw data of electric vehicle battery voltage, temperature distribution and acceleration, normalizing the raw data using a sliding window, and constructing a time-series data matrix.

3. The energy efficiency optimization method for electric vehicles based on fault prediction according to claim 1, characterized in that: In step S1, the electrochemical and thermodynamic features of the operating data are extracted using a local model of the electric vehicle. Based on these features, the failure probability of the electric vehicle is predicted. The expression for predicting the failure probability is as follows: ； In the formula, This represents the probability of failure. For activation functions; This is the weight matrix of the fully connected layers in the local model; It is characterized by electrochemical properties; These are thermodynamic eigenvectors; This indicates the concatenation of electrochemical and thermodynamic eigenvectors; This is the bias term for the fully connected layer.

4. The energy efficiency optimization method for electric vehicles based on fault prediction according to claim 1, characterized in that: The expression for the optimization objective function mentioned in step S2 is: ； ； In the formula, To optimize the objective function; , , They are respectively , , The weights; This represents minimizing energy consumption costs; This indicates minimizing battery life degradation; This indicates minimizing the travel time; This represents the total discrete time steps within the planning time period; For time step The charging power; For time step Real-time electricity price; This represents the average growth rate of the battery's internal resistance. The average state of charge during a charging or discharging cycle; The driving time of the electric vehicle; These are the weighting coefficients; The total charging time within the planned time period.

5. The energy efficiency optimization method for electric vehicles based on fault prediction according to claim 4, characterized in that: In step S2, a fuzzy controller is used to dynamically adjust and optimize the weights in the objective function based on the battery health status and the user-set driving mode. The fuzzy rules used by the fuzzy controller are as follows: (1) If the battery health status is healthy and the user's driving mode is economy priority, then increase The weight is reduced. and The weights; (2) If the battery health status is healthy and the user's driving mode is balanced, then maintain , and The weights; (3) If the battery health status is healthy and the user's driving mode is performance priority, then improve The weight is reduced. and The weights; (4) If the battery health status is slightly degraded and the user's driving mode is set to economy priority, then increase The weight, maintain The weight is reduced. The weights; (5) If the battery health status is slightly degraded and the user's driving mode is balanced, then increase The weight, maintain and The weights; (6) If the battery health status is slightly degraded and the user's driving mode is performance mode, then increase and The weight is reduced. The weights; (7) If the battery health status is severely degraded and the user's driving mode is set to economy priority, then reduce the power consumption. Weighting, increasing The weight, maintain The weights; (8) If the battery health status is severely degraded and the user's driving mode is balanced, then increase The weight is reduced. and The weights; (9) If the battery health status is severely degraded and the user's driving mode is performance priority, then increase The weight is reduced. and The weight.

6. The energy efficiency optimization method for electric vehicles based on fault prediction according to claim 5, characterized in that: In step S3, after uploading the user's adoption rate and adjustment records of the push strategy as a log file to the cloud, reinforcement learning is used to periodically update the fuzzy rules based on the log file. The expression for updating the fuzzy rules is: ； In the formula, To enhance learning Value table; For state, , The state space includes battery health status, user driving mode preferences, current vehicle operating conditions, and the adoption rate of the last optimization suggestion. For action, , The action space includes the weights of energy consumption optimization targets, battery life targets, and range efficiency targets. The learning rate; As a reward; Discount factor; Indicates that the current action is being performed. The state at the next moment obtained after the environmental transition; Indicates the state The action chosen for the next moment.

7. The energy efficiency optimization method for electric vehicles based on fault prediction according to claim 1, characterized in that: The constraints mentioned in step S1 include: (1) When the failure probability When the faulty battery cell is located, the upper limit of the discharge current of the branch is lowered. (2) When the failure probability At this time, the energy recovery strategy is paused, limiting the accelerator pedal response limit; (3) When the failure probability At that time, the high-voltage relay is disconnected, controlling the liquid cooling system to operate at full power.

8. The energy efficiency optimization method for electric vehicles based on fault prediction according to claim 1, characterized in that: The expression for calculating the difference weight of each user based on the anonymized data in step S3 is as follows: ； In the formula, For users i Difference weights; , , They are respectively , , Weighting coefficients; For users i The relative data volume factor of the vehicles in the current training period; For users i The quality reliability factor of vehicle-uploaded data; For users i The validation loss improvement factor after updating the local model of the vehicle; For users i The number of valid samples uploaded; These are weighting coefficients; For users i The proportion of outliers in the uploaded data; For users i Signal noise fluctuations in uploaded data; For users i The sampling interval missing rate; The validation loss of the aggregation model before downloading; For users i The validation loss after updating the local model of the vehicle.

9. The energy efficiency optimization method for electric vehicles based on fault prediction according to claim 1, characterized in that: In step S3, knowledge across vehicles is extracted from the aggregated model, and an update package for the local model of all electric vehicles is generated using a differential compression algorithm. This package is then downloaded to the user's vehicle via over-the-air (OTA) technology, thus updating the local model of the electric vehicle.

10. An energy efficiency optimization system for electric vehicles based on fault prediction, implemented based on the energy efficiency optimization method for electric vehicles based on fault prediction as described in any one of claims 1 to 9, characterized in that, include: The module includes a data acquisition module, a fault prediction module, an optimization strategy generation module, a cloud-based collaborative update module, and a human-computer interaction module. The data acquisition module acquires real-time data on the vehicle's battery pack voltage, temperature distribution, and acceleration. The fault prediction module extracts the vehicle's electrochemical and thermodynamic characteristics from the data collected by the data acquisition module and calculates the vehicle's fault probability. The optimization strategy generation module: constructs a multi-objective optimization function, realizes multi-objective dynamic optimization under fault constraints, and outputs vehicle charging period suggestions, vehicle speed control range, power response parameters, etc. The cloud-based collaborative update module calculates the user's differential weights based on the user's strategy preferences, aggregates the local models of all vehicles, updates the global model in the cloud, generates an incremental update package, and sends it to the vehicle to update the vehicle's local model. The human-computer interaction module provides users with a visual representation of the vehicle's fault level heatmap, energy consumption optimization schemes, and battery health trends. It also allows users to adjust vehicle speed range offsets, manually modify charging periods, record user strategy adoption rates and behavioral deviations, and update the user's driver profile.

Images