Automobile residual value management and prediction system

Abstract

The application discloses a car residual value management and prediction system and belongs to the technical field of artificial intelligence. The appearance image of a car is collected, YOLO algorithm is used for car appearance damage detection, and a car appearance damage set is obtained. An association between a car appearance damage standard library and the car appearance damage set is established, the damage degree of different positions of the car is evaluated, and the car damage depreciation rate is calculated. A deep neural network model is constructed and trained to predict the battery predicted residual value of the car. The value loss after depreciation is calculated based on the car damage depreciation rate, and the value of the car after depreciation is obtained. The proportion of the battery in the value of the car is determined, and the predicted residual value of the car is calculated. Based on the initial value, service life and predicted residual value of the car, the residual value after different service lives is calculated using the annual depreciation amount, and the residual value of the car is obtained. The transaction value of the same type of car is compared with the residual value of the car, and a comparison conclusion is obtained. The effect of market supply and demand on the residual value of the car is evaluated, and the residual value of the car is adjusted.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an automobile residual value management and prediction system. Background Art

[0002] With the rapid development of the automotive industry, particularly the explosive growth of the new energy vehicle market, auto transactions are becoming increasingly frequent. This has led to a sharp increase in demand for vehicle residual value assessments in scenarios such as used car transactions, vehicle mortgages, and insurance claims. Accurately assessing vehicle residual value is not only of vital importance to consumers but also crucial for relevant industry players such as automakers, dealers, and financial institutions.

[0003] Existing technologies for assessing exterior damage in vehicles rely heavily on manual judgment, which is highly subjective and inefficient. When evaluating the residual value of vehicles involving batteries, particularly new energy vehicles, existing methods suffer from numerous shortcomings. For one thing, there are limited methods for detecting battery location, damage level, and health. Furthermore, a lack of effective data processing and analysis methods makes it difficult to accurately predict residual values. Traditional methods for calculating residual values ​​often employ simple depreciation models, such as straight-line depreciation, which fail to comprehensively consider the impact of complex factors on residual value, such as exterior damage, battery condition, and market supply and demand. Summary of the Invention

[0004] The purpose of the present invention is to provide a vehicle residual value management and prediction system to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] In a first aspect, the present invention provides a vehicle residual value management and prediction system, comprising:

[0007] Damage depreciation rate calculation module: The damage detection unit collects vehicle exterior images and uses the YOLO algorithm to detect exterior damage, generating an exterior damage set. The depreciation rate calculation unit associates the exterior damage standard library with the exterior damage set, assesses the extent of damage at different locations on the vehicle, and calculates the damage depreciation rate.

[0008] Value depreciation calculation module: The model building and training unit obtains the damage level, battery health, and basic battery information of the vehicle battery location, performs feature engineering, builds and trains a deep neural network model, and predicts the estimated residual value of the vehicle battery. The value depreciation calculation unit collects the initial value of the vehicle, calculates the value depreciation based on the vehicle damage depreciation rate, and obtains the vehicle's value after depreciation.

[0009] Estimated residual value calculation module: The estimated residual value calculation unit determines the proportion of batteries in the vehicle value and calculates the estimated residual value of the vehicle based on the estimated residual value of the battery and the depreciated vehicle value;

[0010] Vehicle residual value calculation module: The depreciation calculation unit collects the vehicle's service life and calculates the annual depreciation based on the vehicle's initial value, service life, and estimated residual value. The vehicle residual value calculation unit uses the annual depreciation to calculate the residual value after different service years to obtain the vehicle's residual value;

[0011] Vehicle residual value adjustment module: The market data comparison and analysis unit collects market data and compares the transaction value of similar models with the residual value of vehicles through statistical analysis to obtain comparative conclusions. Based on the comparative conclusions, the vehicle residual value adjustment unit evaluates the impact of market supply and demand on the residual value of vehicles and adjusts the residual value of vehicles.

[0012] In conjunction with the first aspect, in a first implementation of the first aspect of the present application, a YOLO algorithm is used to detect vehicle exterior damage, and a vehicle exterior damage set is obtained, including:

[0013] Guide the car into the inspection area and park it in a pre-marked position, ensuring that the car is level and fixed in position; start the camera acquisition program, and when the car is stationary, take multi-angle photos of the car's exterior, including the four sides of the car, the roof, the bottom of the car, and various details of the car, to collect the car's appearance images; organize an image annotation team to annotate and classify the collected appearance images; preprocess the annotated appearance images, and use the YOLO algorithm to train the YOLO model based on the preprocessed appearance images; during the training process, adjust the hyperparameters of the YOLO model; apply the trained YOLO model to the appearance image of the car to be inspected; the YOLO model detects the damage location, type, and size of the car's exterior, outputs the detection results in the form of bounding boxes and category labels, and integrates them to obtain the car appearance damage set.

[0014] In conjunction with the first aspect, in a second implementation of the first aspect of the present application, establishing an association between the automobile appearance damage standard library and the automobile appearance damage set, evaluating the degree of damage at different locations of the automobile, and calculating the automobile damage depreciation rate includes:

[0015] For each damage record in the vehicle exterior damage set, the corresponding category is found in the vehicle exterior damage standard library based on the damage type; the damage location in the vehicle exterior damage set is matched with the descriptions and examples of different locations in each damage type in the vehicle exterior damage standard library; for single damage, the damage size is compared between the vehicle exterior damage set and the vehicle exterior damage standard library. When the damage size in the vehicle exterior damage set is between two levels in the vehicle exterior damage standard library, the damage degree is determined based on the principle of approaching the more serious level; for composite damage, several factors involved in the damage situation in the vehicle exterior damage set are obtained. According to the assessment process for composite damage in the vehicle exterior damage standard library, the damage degree of each damage is first determined separately, and then the overall damage degree is determined;

[0016] According to the damage type, damage degree and damage location, the corresponding depreciation rate coefficient is searched in the standard library of automobile appearance damage; the depreciation rate coefficients of all damages in the automobile appearance damage set are accumulated to obtain the automobile damage depreciation rate.

[0017] In conjunction with the first aspect, in a third implementation of the first aspect of the present application, obtaining the damage degree, battery health, and basic battery information of the vehicle battery location, performing feature engineering, constructing and training a deep neural network model, and predicting the expected residual value of the vehicle battery include:

[0018] The damage degree of the vehicle battery location is obtained based on the association between the appearance damage standard library and the vehicle appearance damage set. The battery health and basic battery information are obtained from the vehicle's battery management system. Data preprocessing is performed to extract original features related to the battery residual value. Derivative features are constructed by performing mathematical operations and logical combinations on the original features. Correlation analysis is used to screen out key features that have an impact on the battery residual value.

[0019] Design the network structure of the deep neural network model, including the input layer, hidden layer and output layer, and initialize the weights and biases of the deep neural network model; divide the key features into a training set, a validation set and a test set, select the mean square error as the loss function, use stochastic gradient descent to optimize the model, and use the training set and validation set to train the model; use the test set data to evaluate the trained model; input the battery-related data of the vehicle to be tested into the trained model, and output the estimated residual value of the vehicle's battery.

[0020] In combination with the first aspect, in a fourth implementation of the first aspect of the present application, calculating the value loss based on the automobile damage depreciation rate to obtain the automobile value after loss includes:

[0021] For new cars, the initial value of the car is determined based on the invoice value at the time of purchase; for used cars, the initial value of the car is determined through a used car evaluation platform; the value loss is calculated based on the car's damage depreciation rate, and the calculation formula is: D = V0 × r, where D is the value loss, V0 is the initial value, and r is the car's damage depreciation rate; subtract the value loss from the initial value to obtain the car's value after depreciation.

[0022] In combination with the first aspect, in a fifth implementation of the first aspect of the present application, determining the proportion of the battery in the vehicle value and calculating the estimated vehicle residual value based on the estimated battery residual value and the vehicle value after depreciation include:

[0023] Collect vehicle transaction values ​​and battery-related data from used car appraisal platforms, car dealer databases, and relevant industry statistics websites for vehicles of the same make, model, age, and mileage as the target vehicle, within a deviation threshold. Classify the collected battery-related data into different battery status categories based on the actual battery status of the vehicle. Calculate the average vehicle transaction value for each battery status data subset, and compare the average transaction value of vehicles with the best battery health status with the average transaction value of vehicles in other categories to determine a value difference. Analyze the relationship between the value difference and the battery-related data to establish a value impact coefficient model.

[0024] Taking the vehicle with the best battery health as a benchmark, the proportion of its battery in the total vehicle value is set as an initial value. Based on the value impact coefficient model, the proportion of the battery in the vehicle value is adjusted for different battery states;

[0025] According to the proportion of the battery in the value of the car, the theoretical value of the battery value in the value of the car after depreciation is calculated. The calculation formula is: V b =V c ×p, where V b is the theoretical value of the battery, V c is the value of the car after depreciation, and p is the proportion of the battery in the value of the car;

[0026] When the estimated residual value of the battery is greater than or equal to the theoretical value of the battery, the calculation formula for the estimated residual value of the car is as follows: c =V c -V b +R b , where R c is the estimated residual value of the car, R b is the estimated residual value of the battery; when the estimated residual value of the battery is less than the theoretical value of the battery, the calculation formula for the estimated residual value of the car is as follows: R c =V c -(V b -R b ).

[0027] In conjunction with the first aspect, in a sixth implementation of the first aspect of the present application, the depreciation amount calculation unit collects the service life of the automobile and calculates the annual depreciation amount based on the initial value, service life, and estimated residual value of the automobile. The automobile residual value calculation unit uses the annual depreciation amount to calculate the residual value after different service lives to obtain the automobile residual value, including:

[0028] Subtract the estimated residual value from the car's initial value to get the total depreciation over the car's entire useful life. Divide the total depreciation by the useful life to get the annual depreciation. Starting from the car's initial value, subtract the annual depreciation each year until you reach the given useful life to get the car's residual value.

[0029] In conjunction with the first aspect, in a seventh implementation of the first aspect of the present application, collecting market data and comparing the transaction value and residual value of similar vehicle models through statistical analysis to obtain a comparative conclusion includes:

[0030] Identify the target vehicle model, collect the current actual transaction value of similar vehicles, record the time of each transaction and the number of transactions of similar vehicles within the set time range; calculate the average current actual transaction value of transactions of similar vehicles as a reference benchmark for the transaction value of similar vehicles in the market; draw a value histogram to analyze the distribution of transaction value, and analyze the change trend of transaction value over time based on the transaction time sequence; integrate the statistical analysis data to obtain the transaction value of similar vehicles;

[0031] The residual value of the car is matched and compared with the transaction value of similar models according to the model configuration and age. The difference between the residual value of the car and the transaction value of similar models is calculated, and the size of the difference is determined. By calculating the percentage difference, the degree of deviation between the two is evaluated to draw a comparative conclusion between the residual value of the car and the transaction value of similar models.

[0032] In conjunction with the first aspect, in an eighth implementation of the first aspect of the present application, drawing a value histogram to analyze the distribution of transaction values, and analyzing the trend of changes in transaction values ​​over time based on the transaction time sequence, includes:

[0033] Traverse each transaction record, analyze the minimum and maximum transaction values, determine the value range based on the transaction value, and count the number of transactions in each range; use data analysis software to create a histogram, labeling the value range on the horizontal axis and the vertical axis representing the number of transactions in each range. Each value range corresponds to a rectangular bar, and the height of the rectangular bar is proportional to the number of transactions in the range; the histogram is used to obtain the distribution of transaction values ​​in different ranges;

[0034] Arrange transaction data in ascending order by transaction time, determine the time interval unit, and group the transaction data to obtain several time interval groups; for each time interval group, calculate the average transaction value; use data analysis software to draw a time series graph, with the horizontal axis representing time and the vertical axis representing the average transaction value; plot the average transaction value points corresponding to each time interval in sequence, and use line segments to connect the average transaction value points to form a trend line that shows the change in value over time; based on the direction of the trend line, determine whether the transaction value is increasing, decreasing, or remaining relatively stable.

[0035] In conjunction with the first aspect, in a ninth implementation of the first aspect of the present application, the step of evaluating the effect of market supply and demand on the residual value of the vehicle based on the comparison conclusion and adjusting the residual value of the vehicle includes:

[0036] Obtain the market supply and demand situation; when the market is in short supply for this model and the residual value of the car is lower than the market transaction value, it means that the residual value of the car has not fully considered the positive impact of market supply and demand, and the residual value of the car will be adjusted upward by adjusting the amount of increase based on the ratio of the market transaction value exceeding the residual value of the car; when the market is in oversupply for this model and the residual value of the car is higher than the market transaction value, the residual value of the car will be adjusted downward by adjusting the amount of decrease based on the ratio of the market transaction value below the residual value of the car.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] 1. This invention uses the YOLO algorithm to detect vehicle exterior images, which can quickly, comprehensively, and accurately identify various types of exterior damage. By establishing an association with a standard library of vehicle exterior damage, it can accurately assess the degree of damage at different locations and calculate the depreciation rate, thereby improving the accuracy of exterior damage assessment.

[0039] 2. The present invention comprehensively considers the damage degree of the vehicle battery location, battery health and basic battery information, and uses a deep neural network model for training to more accurately predict the expected residual value of the vehicle battery.

[0040] 3. The present invention calculates the value loss based on the vehicle's damage depreciation rate and initial value, and comprehensively considers the proportion of the battery in the vehicle's value, the estimated residual value of the battery, and the value of the vehicle after depreciation to calculate the estimated residual value of the vehicle. This residual value calculation model that comprehensively considers multiple factors can more realistically reflect the value changes of the vehicle in actual use, making the evaluation results closer to the actual market value. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Fig. 1 This is a system structure diagram of the automobile residual value management and prediction system of the present invention;

[0042] Fig. 2 It is a schematic diagram of the steps for calculating the residual value of an automobile in the automobile residual value management and prediction system of the present invention. DETAILED DESCRIPTION

[0043] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0044] Example: Figs. 1-2 As shown, the present invention provides a technical solution.

[0045] like Fig. 1 As shown in the system structure diagram of the automobile residual value management and prediction system of the present invention, the present invention provides an automobile residual value management and prediction system, including:

[0046] Damage depreciation rate calculation module: The damage detection unit collects vehicle exterior images and uses the YOLO algorithm to detect exterior damage, generating an exterior damage set. The depreciation rate calculation unit associates the exterior damage standard library with the exterior damage set, assesses the extent of damage at different locations on the vehicle, and calculates the damage depreciation rate.

[0047] Specifically, guide the car into the inspection area and park it in a pre-marked position to ensure that the car is level and fixed; start the camera acquisition program, and when the car is stationary, take multi-angle photos of the car's appearance, including the four sides of the car, the roof, the bottom of the car, and various details of the car, to collect the car's appearance images; organize an image annotation team to annotate and classify the collected appearance images; preprocess the annotated appearance images, and use the YOLO algorithm to train the YOLO model based on the preprocessed appearance images; during the training process, adjust the hyperparameters of the YOLO model; apply the trained YOLO model to the appearance image of the car to be inspected; the YOLO model detects the damage location, type, and size of the car's appearance, outputs the detection results in the form of bounding boxes and category labels, and integrates them to obtain a car appearance damage set.

[0048] Furthermore, for each damage record in the automobile exterior damage set, a corresponding category is found in the automobile exterior damage standard library based on the damage type; the damage location in the automobile exterior damage set is matched with the descriptions and examples of different locations in each damage type in the automobile exterior damage standard library; for single damage, the damage size is compared between the automobile exterior damage set and the automobile exterior damage standard library; when the damage size in the automobile exterior damage set is between two levels in the automobile exterior damage standard library, the damage degree is determined based on the principle of being closer to the more serious level; for composite damage, several factors involved in the damage situation in the automobile exterior damage set are obtained, and according to the composite damage assessment process in the automobile exterior damage standard library, the damage degree of each damage is first determined separately, and then the overall damage degree is determined;

[0049] According to the damage type, damage degree and damage location, the corresponding depreciation rate coefficient is searched in the standard library of automobile appearance damage; the depreciation rate coefficients of all damages in the automobile appearance damage set are accumulated to obtain the automobile damage depreciation rate.

[0050] In a specific embodiment, a professional image annotation team is organized to use the LabelImg tool to annotate and classify the collected images. During the annotation process, a total of 5 exterior damages were identified, including: a scratch with a length of 12 cm on the left front door; a dent with a diameter of 6 cm on the right rear door; a paint chip with an area of ​​15 square centimeters on the front bumper; a slight collision deformation on the roof; and a scratch with a length of 8 cm on the right rear fender. These damages are classified and stored according to the damage type (scratches, dents, paint chipping, collision deformation) and location. The annotated images are preprocessed, the images are uniformly adjusted to a size of 640×640 pixels, and normalized to convert the pixel values ​​from the range of 0-255 to the range of 0-1. At the same time, the annotated data is converted into the format required by the YOLO algorithm, including bounding box coordinates and category labels.

[0051] Based on the preprocessed image data, the YOLO model was trained using the YOLOv5 algorithm. During training, the learning rate was set to 0.001, the batch size was set to 16, and the number of iterations was set to 500. By continuously adjusting the model's hyperparameters and observing the model's performance indicators on the validation set (such as mean average precision (mAP), accuracy, and recall), the model ultimately achieved good performance. The trained YOLO model was applied to the vehicle exterior image to be inspected. The model detected the location, type, and size of damage on the vehicle exterior and output the detection results in the form of bounding boxes and category labels. After testing, the model accurately identified the five damages mentioned above, which was consistent with the manual annotation results, and successfully integrated into the vehicle exterior damage set.

[0052] For each damage record in the vehicle exterior damage collection, we match the damage type to a corresponding category in the standard vehicle exterior damage database. For example, a scratch on the left front door corresponds to the "scratches" category, and a dent on the right rear door corresponds to the "dents" category. We also match the damage location with the descriptions and examples of different locations under each damage type in the standard vehicle exterior damage database.

[0053] The 12cm scratch on the left front door is classified as moderate according to the standard library for vehicle exterior damage, which classifies scratches between 8 and 15cm in length. The 6cm dent on the right rear door is classified as moderate, as the standard library defines dents between 5 and 8cm in diameter as moderate. The 15cm squared paint chip on the front bumper is classified as moderate, as the standard library defines paint chipping between 10 and 20cm squared as moderate. The 8cm scratch on the right rear fender is classified as light (3-8cm is considered light in the standard library). The slight deformation of the roof is classified as light within the single damage category (according to the standard library's description of deformation severity).

[0054] Based on the damage type, severity, and location, the corresponding depreciation coefficient is searched in the standard database for vehicle exterior damage. For example, the depreciation coefficient for a moderate scratch on the left front door is 0.4%, a moderate dent on the right rear door is 0.5%, a moderate paint chipping on the front bumper is 0.4%, a minor impact deformation on the roof is 0.3%, and a minor scratch on the right rear fender is 0.2%. Adding up the depreciation coefficients for all damage in the vehicle exterior damage set yields the following depreciation rate: 0.4% + 0.5% + 0.4% + 0.3% + 0.2% = 1.8%.

[0055] Value depreciation calculation module: The model building and training unit obtains the damage level, battery health, and basic battery information of the vehicle battery location, performs feature engineering, builds and trains a deep neural network model, and predicts the estimated residual value of the vehicle battery. The value depreciation calculation unit collects the initial value of the vehicle, calculates the value depreciation based on the vehicle damage depreciation rate, and obtains the vehicle's value after depreciation.

[0056] Specifically, the damage degree of the vehicle battery location is obtained based on the association between the appearance damage standard library and the vehicle appearance damage set. The battery health and basic battery information are obtained from the vehicle's battery management system. Data preprocessing is performed to extract original features related to the battery residual value. Derived features are constructed by performing mathematical operations and logical combinations on the original features. Correlation analysis is used to screen out key features that have an impact on the battery residual value.

[0057] Design the network structure of the deep neural network model, including the input layer, hidden layer and output layer, and initialize the weights and biases of the deep neural network model; divide the key features into a training set, a validation set and a test set, select the mean square error as the loss function, use stochastic gradient descent to optimize the model, and use the training set and validation set to train the model; use the test set data to evaluate the trained model; input the battery-related data of the vehicle to be tested into the trained model, and output the estimated residual value of the vehicle's battery.

[0058] Furthermore, for new cars, the initial value of the car is determined based on the invoice value at the time of purchase; for used cars, the initial value of the car is determined through a used car evaluation platform; the value depreciation is calculated based on the car damage depreciation rate, and the calculation formula is: D = V0 × r, where D is the value depreciation, V0 is the initial value, and r is the car damage depreciation rate; the value depreciation is subtracted from the initial value to obtain the car value after depreciation.

[0059] In a specific embodiment, based on the association between the previously established standard library of automobile appearance damage and the automobile appearance damage set, it was found upon inspection that the battery casing of the vehicle had a slight dent with a diameter of 3 cm. In the standard library of appearance damage, this type of battery casing dent is judged to be mild damage. The following data was obtained from the battery management system (BMS) of the vehicle: the battery health state (SOH) is 80%. The current capacity of the battery is 40Ah (the initial capacity is 50Ah). The battery type is a ternary lithium battery. The battery manufacturer is Company X. The battery has been in use for 3 years and the vehicle has a cumulative mileage of 60,000 kilometers.

[0060] The acquired battery data is cleaned to remove outliers and noise. For example, transient abnormal voltage values ​​occasionally transmitted by the BMS are corrected. All numerical data is standardized to a mean of 0 and a standard deviation of 1. For example, data such as battery capacity and SOH value are standardized. Raw features related to battery residual value are extracted, including battery capacity, SOH value, age, mileage, battery type (converted to a binary vector using one-hot encoding), battery manufacturer (which can also be encoded), and battery location damage level (mild is set to 1, moderate is set to 2, and severe is set to 3). Derived features are constructed by performing mathematical operations and logical combinations on the raw features. For example, the battery capacity decay rate is calculated as (initial capacity - current capacity) / initial capacity, which is (50-40) / 50 = 0.2; the average annual mileage is calculated as mileage / age, which is 60,000 / 3 = 20,000 kilometers / year. Using correlation analysis, it was found that the battery capacity decay rate, SOH value, service life and degree of damage to the battery location were highly correlated with the battery residual value. These key features were screened out for subsequent model training.

[0061] Based on the number of key features identified, the number of neurons in the input layer was set to 4. Two hidden layers were set, with the first hidden layer having 8 neurons and the second hidden layer having 4 neurons. Fully connected layers were used, and the activation function was ReLU. One neuron was set in the output layer to output the predicted residual battery value. A linear function was used as the activation function. The weights of the deep neural network model were initialized using a Gaussian distribution with a mean of 0 and a standard deviation of 0.01, and the bias was initialized to 0.01. The key features were divided into a training set (70%), a validation set (15%), and a test set (15%). For example, given 100 data sets, 70 were used for training, 15 for validation, and 15 for testing. Mean squared error (MSE) was selected as the loss function, and the stochastic gradient descent (SGD) algorithm was used for model optimization, with a learning rate of 0.001 and 300 iterations. During training, forward and backward propagation were performed using the training set data, continuously adjusting the weights and biases. Model performance was also evaluated on the validation set to prevent overfitting. The trained model was evaluated using the test set data, and the root mean square error (RMSE) and mean absolute error (MAE) were calculated. The RMSE was 0.2 and the MAE was 0.15, indicating that the model has good predictive performance.

[0062] The trained model inputs key battery feature data from the vehicle under test, outputting an estimated residual value of 40,000 yuan. Since the vehicle is used, its initial value was determined to be 100,000 yuan through a professional used car appraisal platform, taking into account factors such as the vehicle's brand, model, configuration, condition, and market conditions. Given the vehicle's damage depreciation rate r = 1.8% (derived from a previous exterior damage assessment), the depreciation calculation formula D = V0 × r yields a depreciation of D = 100,000 × 1.8% = 1,800 yuan. Subtracting the depreciation from the initial value yields a post-depreciation value of 100,000 yuan - 1,800 yuan = 98,200 yuan.

[0063] Estimated residual value calculation module: The estimated residual value calculation unit determines the proportion of batteries in the vehicle value and calculates the estimated residual value of the vehicle based on the estimated residual value of the battery and the depreciated vehicle value;

[0064] Specifically, the vehicle transaction values ​​of vehicles of the same brand, model, age, and mileage as the target vehicle, within a deviation threshold, and battery-related data are collected from used car appraisal platforms, car dealer databases, and relevant industry statistics websites. The collected battery-related data is classified into different battery status categories based on the actual status of the vehicle batteries. For each data subset of vehicle transaction values ​​for each battery status category, the average vehicle transaction value is calculated, and the average transaction value of vehicles with the best battery health status is compared with the average transaction value of vehicles in other categories to obtain a value difference value. By analyzing the relationship between the value difference value and the battery-related data, a value impact coefficient model is established.

[0065] Taking the vehicle with the best battery health as a benchmark, the proportion of its battery in the total vehicle value is set as an initial value. Based on the value impact coefficient model, the proportion of the battery in the vehicle value is adjusted for different battery states;

[0066] According to the proportion of the battery in the value of the car, the theoretical value of the battery value in the value of the car after depreciation is calculated. The calculation formula is: V b =V c ×p, where V b is the theoretical value of the battery, V c is the value of the car after depreciation, and p is the proportion of the battery in the value of the car;

[0067] When the estimated residual value of the battery is greater than or equal to the theoretical value of the battery, the calculation formula for the estimated residual value of the car is as follows: c =V c -V b +R b , where R c is the estimated residual value of the car, R b is the estimated residual value of the battery; when the estimated residual value of the battery is less than the theoretical value of the battery, the calculation formula for the estimated residual value of the car is as follows: R c =V c -(V b -R b ).

[0068] In one specific embodiment, the collected battery-related data is divided into three categories based on the actual state of the vehicle battery: Good battery health (SOH ≥ 85%): 20 vehicles were classified as such. Medium battery health (70% ≤ SOH < 85%): 40 vehicles were classified as such. Poor battery health (SOH < 70%): 20 vehicles were classified as such. Average transaction value calculation: For vehicles with good battery health, the average transaction value was 110,000 yuan. For vehicles with medium battery health, the average transaction value was 90,000 yuan. For vehicles with poor battery health, the average transaction value was 70,000 yuan. Value difference calculation: The value difference from medium to good battery health = 110,000 yuan - 90,000 yuan = 20,000 yuan. The value difference from poor to good battery health = 110,000 yuan - 70,000 yuan = 40,000 yuan. By analyzing the relationship between the value difference and battery-related data (such as SOH and battery capacity decay rate), a value impact coefficient model is established. Analysis found that for every 10% decrease in SOH, the average transaction value of the car decreased by 15,000 yuan; for every 0.1 increase in battery capacity attenuation rate, the average transaction value of the car decreased by 10,000 yuan.

[0069] Taking the vehicle with the best battery health as the benchmark, assume that its battery accounts for 40% of the vehicle's value. For vehicles with medium battery health, according to the value impact coefficient model, the SOH decreases by 5% (from 85% to 80%) and the battery capacity decay rate increases by 0.05 (from 0.1 to 0.15). The battery proportion is adjusted to 40% - (1.5 ÷ 11) × 40% - (1 ÷ 11) × 40% × 0.05 ÷ 0.1 = 33.64%. For vehicles with poor battery health, the SOH decreases by 15% (from 85% to 70%) and the battery capacity decay rate increases by 0.15 (from 0.1 to 0.25). The battery proportion is adjusted to 40% - (1.5 ÷ 11) × 40% × 1.5 - (1 ÷ 11) × 40% × 0.15 ÷ 0.1 = 25.45%.

[0070] The previously calculated car value after depreciation is 98,200 yuan. For the vehicle we tested, the battery health status is medium, and the battery accounts for p = 33.64% of the car value. According to the battery theoretical value calculation formula, the battery theoretical value V battery理论 =98200×33.64%≈32934 yuan.

[0071] The estimated residual value of the battery is 40,000 yuan, which is greater than the theoretical value of the battery, 32,934 yuan. According to the formula for calculating the estimated residual value of the car, the estimated residual value of the car is V. 汽车预计残值 =98200-32934+40000=105266 yuan.

[0072] like Fig. 2The automobile residual value calculation steps of the automobile residual value management and prediction system of the present invention are shown in the automobile residual value calculation module: the depreciation amount calculation unit collects the service life of the automobile, calculates the annual depreciation amount based on the initial value, service life and expected residual value of the automobile, and the automobile residual value calculation unit uses the annual depreciation amount to calculate the residual value after different service years to obtain the automobile residual value;

[0073] Specifically, subtract the estimated residual value from the car's initial value to get the total depreciation over the car's entire useful life. Divide the total depreciation by the useful life to get the annual depreciation. Starting from the car's initial value, subtract the annual depreciation year by year until the given useful life is calculated to get the car's residual value.

[0074] Vehicle residual value adjustment module: The market data comparison and analysis unit collects market data and compares the transaction value of similar models with the residual value of vehicles through statistical analysis to obtain comparative conclusions. Based on the comparative conclusions, the vehicle residual value adjustment unit evaluates the impact of market supply and demand on the residual value of vehicles and adjusts the residual value of vehicles.

[0075] Specifically, the target vehicle model is determined, and the current actual transaction value of similar vehicles is collected. The time of each transaction and the number of transactions of similar vehicles within the set time range are recorded. The average current actual transaction value of transactions of similar vehicles is calculated as a reference benchmark for the transaction value of similar vehicles in the market. A value histogram is drawn to analyze the distribution of transaction value. Based on the transaction time sequence, the trend of transaction value over time is analyzed. The statistical analysis data is integrated to obtain the transaction value of similar vehicles.

[0076] The residual value of the car is matched and compared with the transaction value of similar models according to the model configuration and age. The difference between the residual value of the car and the transaction value of similar models is calculated, and the size of the difference is determined. By calculating the percentage difference, the degree of deviation between the two is evaluated to draw a comparative conclusion between the residual value of the car and the transaction value of similar models.

[0077] Furthermore, each transaction record is traversed, the minimum and maximum transaction values ​​are analyzed, the value range to which the transaction value belongs is determined based on the transaction value, and the number of transactions in each range is counted; a histogram is created using data analysis software, with the value range marked on the horizontal axis and the number of transactions in each range on the vertical axis. Each value range corresponds to a rectangular bar, and the height of the rectangular bar is proportional to the number of transactions in the range; the distribution of transaction values ​​in different ranges is obtained through the histogram;

[0078] Arrange transaction data in ascending order by transaction time, determine the time interval unit, and group the transaction data to obtain several time interval groups; for each time interval group, calculate the average transaction value; use data analysis software to draw a time series graph, with the horizontal axis representing time and the vertical axis representing the average transaction value; plot the average transaction value points corresponding to each time interval in sequence, and use line segments to connect the average transaction value points to form a trend line that shows the change in value over time; based on the direction of the trend line, determine whether the transaction value is increasing, decreasing, or remaining relatively stable.

[0079] Furthermore, the market supply and demand situation is obtained; when the market is in short supply for this model and the residual value of the car is lower than the market transaction value, it means that the residual value of the car has not fully considered the positive impact of market supply and demand, and the residual value of the car is adjusted upward, and the increase is adjusted based on the ratio of the market transaction value exceeding the residual value of the car; when the market is in oversupply for this model and the residual value of the car is higher than the market transaction value, the residual value of the car is adjusted downward, and the decrease is adjusted based on the ratio of the market transaction value below the residual value of the car.

[0080] In a specific embodiment, the transaction value data of 100 collected similar models are calculated, and the total is 9.5 million yuan. Therefore, the average value is 950 ÷ 100 = 95,000 yuan, which is used as a reference benchmark for the transaction value of similar models in the market.

[0081] Go through each transaction record and analyze the transaction value. The minimum is 80,000 yuan and the maximum is 120,000 yuan. Divide the value range into several intervals, such as 80,000-85,000 yuan, 85,000-90,000 yuan, 90,000-95,000 yuan, 95,000-100,000 yuan, 100,000-105,000 yuan, 105,000-110,000 yuan, 110,000-115,000 yuan, and 115,000-120,000 yuan. Count the number of transactions in each interval. For example, there are 10 transactions in the 80,000-85,000 yuan range, 15 transactions in the 85,000-90,000 yuan range, 20 transactions in the 90,000-95,000 yuan range, 25 transactions in the 95,000-100,000 yuan range, 15 transactions in the 100,000-105,000 yuan range, 10 transactions in the 105,000-110,000 yuan range, 3 transactions in the 110,000-115,000 yuan range, and 2 transactions in the 115,000-120,000 yuan range. A histogram was created using data analysis software, with the aforementioned value ranges plotted on the horizontal axis and the number of transactions within each range on the vertical axis. Each value range is represented by a rectangular bar, with the height of the bar proportional to the number of transactions within that range. The histogram clearly shows that the transaction value is most concentrated in the 90,000-100,000 yuan range.

[0082] Arrange the transaction data in ascending order by transaction time, averaging one week as the time interval. Group the transaction data to obtain 12 time interval groups (since data was collected for three months, approximately 12 weeks). Calculate the average transaction value for each time interval group. For example, the total value of the five transactions in week 1 was 450,000 yuan, with an average transaction value of 45 ÷ 5 = 90,000 yuan; the total value of the eight transactions in week 2 was 760,000 yuan, with an average transaction value of 76 ÷ 8 = 95,000 yuan, and so on. Use data analysis software to plot a time series graph, with the horizontal axis representing time (in weeks) and the vertical axis representing average transaction value. Plot the average transaction value points corresponding to each time interval in turn, and connect these average transaction value points with line segments to form a trend line showing the change in value over time. Observation shows that the trend line shows a slight upward trend, indicating that the overall market transaction value of this vehicle model has been on an upward trend over the past three months.

[0083] Looking back at the previous calculations, the residual value of the target vehicle model is 105,266 yuan. The residual value is then compared with the transaction value of similar vehicles based on the vehicle configuration (this model has a single configuration) and age (3 years). The difference between the residual value and the average transaction value of similar vehicles (95,000 yuan) is calculated as 105,266 yuan - 95,000 yuan = 10,266 yuan. The percentage difference is calculated as (10,266 ÷ 95,000 yuan) × 100% ≈ 10.81%. The positive difference and significant percentage difference indicate that the residual value of the vehicle is higher than the average transaction value of similar vehicles, indicating a significant deviation. This leads to the conclusion that the residual value of the vehicle is relatively high relative to the market transaction value.

[0084] Based on industry reports, feedback from auto dealers, and data from market research institutions, we know that this model is currently in short supply in the local market, with strong demand and relatively insufficient supply. Due to this shortfall and the vehicle's residual value exceeding the average market transaction value, a downward adjustment is necessary. The reduction is calculated based on the ratio of the market transaction value to the vehicle's residual value. Given a residual value of 105,266 yuan and an average market transaction value of 95,000 yuan, the difference is 10,266 yuan. The downward adjustment is calculated as (10,266 ÷ 105,266) × 100% ≈ 9.75%. Therefore, the adjusted residual value is 105,266 × (1 - 9.75%) ≈ 95,935 yuan. This adjustment brings the vehicle's residual value more in line with market conditions and takes into account the impact of market supply and demand on vehicle value.

[0085] It will be apparent to those skilled in the art that the application is not limited to the details of the above-exemplified embodiments and that the present application can be implemented in other particular forms without departing from the spirit or essential characteristics of the present application. The embodiments should therefore be considered in all respects as illustrative and not restrictive, the scope of the application being indicated by the appended claims rather than by the above description, and all changes which come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein. No reference signs in the claims should be considered as limiting the scope of the claims with respect to the figures of the patent document.

Claims

1. Automobile residual value management and prediction system, characterized by: include: Damage depreciation rate calculation module: The damage detection unit collects vehicle exterior images and uses the YOLO algorithm to detect exterior damage, generating an exterior damage set. The depreciation rate calculation unit associates the exterior damage standard library with the exterior damage set, assesses the extent of damage at different locations on the vehicle, and calculates the damage depreciation rate. Value depreciation calculation module: The model building and training unit obtains the damage level, battery health, and basic battery information of the vehicle battery location, performs feature engineering, builds and trains a deep neural network model, and predicts the estimated residual value of the vehicle battery. The value depreciation calculation unit collects the initial value of the vehicle, calculates the value depreciation based on the vehicle damage depreciation rate, and obtains the vehicle's value after depreciation. Estimated residual value calculation module: The estimated residual value calculation unit determines the proportion of batteries in the vehicle value and calculates the estimated residual value of the vehicle based on the estimated residual value of the battery and the depreciated vehicle value; Collect vehicle transaction values ​​and battery-related data from used car appraisal platforms, car dealer databases, and relevant industry statistics websites for vehicles of the same make, model, age, and mileage as the target vehicle, within a deviation threshold. Classify the collected battery-related data into different battery status categories based on the actual battery status of the vehicle. Calculate the average vehicle transaction value for each battery status data subset, and compare the average transaction value of vehicles with the best battery health status with the average transaction value of vehicles in other categories to determine a value difference. Analyze the relationship between the value difference and the battery-related data to establish a value impact coefficient model. Taking the vehicle with the best battery health as a benchmark, the proportion of its battery in the total vehicle value is set as an initial value. Based on the value impact coefficient model, the proportion of the battery in the vehicle value is adjusted for different battery states; Based on the proportion of the battery in the car's value, the theoretical value of the battery in the car's value after depreciation is calculated using the following formula: ,in, is the theoretical value of the battery, is the value of the car after depreciation, and p is the proportion of the battery in the value of the car; When the estimated residual value of the battery is greater than or equal to the theoretical value of the battery, the calculation formula for the estimated residual value of the vehicle is as follows: ,in, is the estimated residual value of the car, is the estimated residual value of the battery; when the estimated residual value of the battery is less than the theoretical value of the battery, the calculation formula for the estimated residual value of the car is as follows: ; Vehicle residual value calculation module: The depreciation calculation unit collects the vehicle's service life and calculates the annual depreciation based on the vehicle's initial value, service life, and estimated residual value. The vehicle residual value calculation unit uses the annual depreciation to calculate the residual value after different service years to obtain the vehicle's residual value; Subtract the estimated residual value from the car's initial value to get the total depreciation over its entire useful life. Divide the total depreciation by the useful life to get the annual depreciation. Starting from the car's initial value, subtract the annual depreciation each year until you reach the given useful life to get the residual value. Vehicle residual value adjustment module: The market data comparison and analysis unit collects market data and, through statistical analysis, compares the transaction value of similar models with the residual value of the vehicle to obtain a comparative conclusion. Based on the comparative conclusion, the vehicle residual value adjustment unit assesses the impact of market supply and demand on the residual value of the vehicle and adjusts the residual value accordingly. Identify the target vehicle model, collect the current actual transaction value of similar vehicles, record the time of each transaction and the number of transactions of similar vehicles within the set time range; calculate the average current actual transaction value of transactions of similar vehicles as a reference benchmark for the transaction value of similar vehicles in the market; draw a value histogram to analyze the distribution of transaction value, and analyze the change trend of transaction value over time based on the transaction time sequence; integrate the statistical analysis data to obtain the transaction value of similar vehicles; The residual value of the car is matched and compared with the transaction value of similar models according to the model configuration and age. The difference between the residual value of the car and the transaction value of similar models is calculated, and the size of the difference is determined. By calculating the percentage difference, the degree of deviation between the two is evaluated to draw a comparative conclusion between the residual value of the car and the transaction value of similar models.

2. The automobile residual value management and prediction system according to claim 1, characterized in that: The vehicle appearance images are collected, and the vehicle appearance damage detection is performed using the YOLO algorithm to obtain a vehicle appearance damage set, including: Guide the car into the inspection area and park it in a pre-marked position, ensuring that the car is level and fixed in position; start the camera acquisition program, and when the car is stationary, take multi-angle photos of the car's exterior, including the four sides of the car, the roof, the bottom of the car, and various details of the car, to collect the car's appearance images; organize an image annotation team to annotate and classify the collected appearance images; preprocess the annotated appearance images, and use the YOLO algorithm to train the YOLO model based on the preprocessed appearance images; during the training process, adjust the hyperparameters of the YOLO model; apply the trained YOLO model to the appearance image of the car to be inspected; the YOLO model detects the damage location, type, and size of the car's exterior, outputs the detection results in the form of bounding boxes and category labels, and integrates them to obtain the car appearance damage set.

3. The automobile residual value management and prediction system according to claim 1, characterized in that: The establishment of the association between the automobile appearance damage standard library and the automobile appearance damage set, the assessment of the damage degree of different positions of the automobile, and the calculation of the automobile damage depreciation rate include: For each damage record in the vehicle exterior damage set, the corresponding category is found in the vehicle exterior damage standard library based on the damage type; the damage location in the vehicle exterior damage set is matched with the descriptions and examples of different locations in each damage type in the vehicle exterior damage standard library; for single damage, the damage size is compared between the vehicle exterior damage set and the vehicle exterior damage standard library. When the damage size in the vehicle exterior damage set is between two levels in the vehicle exterior damage standard library, the damage degree is determined based on the principle of approaching the more serious level; for composite damage, several factors involved in the damage situation in the vehicle exterior damage set are obtained. According to the assessment process for composite damage in the vehicle exterior damage standard library, the damage degree of each damage is first determined separately, and then the overall damage degree is determined; According to the damage type, damage degree and damage location, the corresponding depreciation rate coefficient is searched in the standard library of automobile appearance damage; the depreciation rate coefficients of all damages in the automobile appearance damage set are accumulated to obtain the automobile damage depreciation rate.

4. The automobile residual value management and prediction system according to claim 1, characterized in that: The method of obtaining the damage degree, battery health, and basic battery information of the vehicle battery location, performing feature engineering, building and training a deep neural network model, and predicting the expected residual value of the vehicle battery includes: The damage degree of the vehicle battery location is obtained based on the association between the appearance damage standard library and the vehicle appearance damage set. The battery health and basic battery information are obtained from the vehicle's battery management system. Data preprocessing is performed to extract original features related to the battery residual value. Derivative features are constructed by performing mathematical operations and logical combinations on the original features. Correlation analysis is used to screen out key features that have an impact on the battery residual value. Design the network structure of the deep neural network model, including the input layer, hidden layer and output layer, and initialize the weights and biases of the deep neural network model; divide the key features into a training set, a validation set and a test set, select the mean square error as the loss function, use stochastic gradient descent to optimize the model, and use the training set and validation set to train the model; use the test set data to evaluate the trained model; input the battery-related data of the vehicle to be tested into the trained model, and output the estimated residual value of the vehicle's battery.

5. The automobile residual value management and prediction system according to claim 1, characterized in that: The initial value of the collected car is calculated based on the car damage depreciation rate to obtain the car value after depreciation, including: For new cars, the initial value of the car is determined based on the invoice value at the time of purchase; for used cars, the initial value of the car is determined through a used car evaluation platform; the value loss is calculated based on the car's damage depreciation rate, and the calculation formula is: , where D is the value loss, is the initial value, r is the car damage depreciation rate; subtract the value loss from the initial value to obtain the car value after depreciation.

6. The automobile residual value management and prediction system according to claim 1, characterized in that: The above-mentioned process of drawing a value histogram to analyze the distribution of transaction value and analyzing the trend of transaction value over time according to the transaction time sequence includes: Traverse each transaction record, analyze the minimum and maximum transaction values, determine the value range based on the transaction value, and count the number of transactions in each range; use data analysis software to create a histogram, labeling the value range on the horizontal axis and the vertical axis representing the number of transactions in each range. Each value range corresponds to a rectangular bar, and the height of the rectangular bar is proportional to the number of transactions in the range; the histogram is used to obtain the distribution of transaction values ​​in different ranges; Arrange transaction data in ascending order by transaction time, determine the time interval unit, and group the transaction data to obtain several time interval groups; for each time interval group, calculate the average transaction value; use data analysis software to draw a time series graph, with the horizontal axis representing time and the vertical axis representing the average transaction value; plot the average transaction value points corresponding to each time interval in sequence, and use line segments to connect the average transaction value points to form a trend line that shows the change in value over time; based on the direction of the trend line, determine whether the transaction value is increasing, decreasing, or remaining relatively stable.

7. The automobile residual value management and prediction system according to claim 1, characterized in that: Based on the comparative conclusions, the impact of market supply and demand on the residual value of the vehicle is evaluated and the residual value of the vehicle is adjusted, including: Obtain the market supply and demand situation; when the market is in short supply for this model and the residual value of the car is lower than the market transaction value, it means that the residual value of the car has not fully considered the positive impact of market supply and demand, and the residual value of the car will be adjusted upward by adjusting the amount of increase based on the ratio of the market transaction value exceeding the residual value of the car; when the market is in oversupply for this model and the residual value of the car is higher than the market transaction value, the residual value of the car will be adjusted downward by adjusting the amount of decrease based on the ratio of the market transaction value below the residual value of the car.

Images