Method and system for predicting quantity of urban decommissioned power batteries based on multi-dimensional feature ML

Abstract

The invention discloses a city-level decommissioned power battery quantity prediction method and system based on a multi-dimensional feature ML. The method comprises the following steps: obtaining a new energy automobile risk quantity and N city feature values in a preset time period to construct a data set; analyzing the N kinds of city characteristic values, and determining city key characteristic values and weight values of the city key characteristic values which influence the risk amount of the new energy automobile; screening a data set according to the urban key characteristic values and the weight values thereof, and training a plurality of optimized machine learning models by using the screened data set; evaluating the prediction performance of each machine learning model, and selecting an optimal model for predicting the insurance amount of the passenger vehicle and the commercial vehicle; inputting the future city key characteristic value into the optimal model to predict the risk amount of the new energy automobile; and based on the predicted risk amount of the new energy vehicle, predicting the urban-level decommissioned power battery amount by utilizing the battery type proportion and the battery characteristics of the new energy vehicle and combining Weibull life distribution. According to the method, the prediction accuracy and practicability of the urban-level retired power battery quantity can be realized.

Description

Technical Field

[0001] The present invention relates to the field of new energy vehicle market and retired power battery prediction technology, and in particular to a city-level retired power battery quantity prediction method and system based on multi-dimensional feature ML. Background Art

[0002] With the dual pressures of global climate change and environmental protection, the development of new energy vehicles, as a key tool to reduce greenhouse gas emissions and promote energy transformation, has received global attention. Power batteries are the core of new energy vehicles. Although when the energy of power batteries decays to below 70%-80%, they can basically not meet travel needs, they still have a high recycling value. Since recycling key metals in power batteries can not only supplement material supply, but also effectively reduce the environmental impact of resource mining, the recycling of power batteries has received widespread attention. Due to the significant differences between cities in industrial structure, resource endowment, economic development level, transportation demand and policy effects, regional forecasts of retired power batteries can provide more refined data support.

[0003] The inventor of this patent discovered at least the following technical problems in the prior art during the process of implementing the method for predicting the insurance volume of new energy vehicles in the present invention: (1) Single data dimension: Most existing technologies predict the amount of retired power batteries based on historical single-dimensional data. For example, existing technologies do not make detailed distinctions between the types of new energy vehicles, especially between passenger cars and commercial vehicles, which are very different in terms of usage purposes, user groups, and market demand patterns; existing technologies do not distinguish between battery types, and the installed capacity of each type of battery will continue to update as the market, technology, climate, and consumer demand change. The single-dimensional data in existing technologies cannot fully reflect the actual retired weight and capacity of retired power batteries, resulting in inaccurate prediction results.

[0004] (2) Failure to consider differences in urban characteristics: Existing technologies make predictions based on provincial or national levels, and do not fully consider the significant differences between cities in terms of economic development level, population structure, policy environment, etc. Cities are the main consumption areas for new energy vehicles. The unique geographical location of each city determines the differences in its resource endowment and industrial layout, which in turn affects the source and type of retired power batteries.

[0005] (3) The model is single and has poor applicability: The core assumption of the linear regression model used in existing technologies is based on the stationarity and linear relationship of the data, and relies on historical laws to build a prediction framework. In the scale transition from macro to micro, the model exposes significant adaptability issues.

[0006] (4) The model lacks dynamic adaptability: When faced with rapid changes in the market environment, the existing forecasting model is unable to adjust the model parameters and structure in a timely manner, making it difficult to accurately predict the impact of these changes on the insurance volume of new energy vehicles, and lacks dynamic adaptability.

[0007] In summary, the existing retired power battery quantity prediction methods cannot meet actual needs. Summary of the invention

[0008] The embodiments of the present invention provide a city-level retired power battery quantity prediction method and system based on multi-dimensional feature ML, which solves the problem that the existing retired power battery quantity prediction method cannot meet actual needs.

[0009] An embodiment of the present invention provides a method for predicting the amount of retired power batteries at the city level based on multi-dimensional feature ML, including: Obtain the number of new energy vehicles insured within a preset time period and N types of city characteristic values ​​to construct a data set; Analyze the N types of city characteristic values ​​to determine the key city characteristic values ​​and their weight values ​​that affect the insurance volume of new energy vehicles; Filtering data sets according to the key characteristic values ​​of the cities and their weight values, and using the filtered data sets to train multiple tuned machine learning models; Evaluate the prediction performance of each machine learning model and select the best model for predicting the insurance volume of passenger cars and commercial vehicles respectively; Input the key characteristic values ​​of future cities into the optimal model to predict the insurance volume of new energy vehicles; Based on the predicted number of new energy vehicles insured, the proportion of battery types and battery characteristics of new energy vehicles, combined with the Weibull life distribution, the number of retired power batteries at the city level is predicted.

[0010] Optionally, the N types of city characteristic values ​​include city population, city GDP, urbanization rate, Baidu Index, and city new energy vehicle promotion subsidy amount.

[0011] Optionally, the method for analyzing the N types of city characteristic values ​​to determine the city's key characteristic values ​​that affect the insurance volume of new energy vehicles is specifically a cluster analysis method, and the method for determining the weight values ​​of the city's key characteristics is specifically a principal component analysis method.

[0012] Optionally, the machine learning model includes an LSTM model, a random forest model, an XGBoost model, an ANN model and a stacked regression model, and the tuning method is specifically a Bayesian optimization method or a hyperparameter tuning method.

[0013] Optionally, the evaluation indicators used to evaluate the predictive performance of each machine learning model include mean square error and determination coefficient.

[0014] Optionally, the key characteristic values ​​of the future city include the future city GDP, the future city population, the future urbanization rate, the future Baidu Index, and the future city new energy vehicle promotion subsidy amount.

[0015] Optionally, the battery types specifically include 24 types consisting of LFP batteries, NCM111 batteries, NCM523 batteries, NCM622 batteries, NCM811 batteries, and NCA batteries for pure electric passenger cars, hybrid electric passenger cars, pure electric commercial vehicles, and hybrid electric commercial vehicles; the battery characteristics specifically include battery life, battery weight, and battery capacity.

[0016] On the other hand, an embodiment of the present invention further provides a city-level retired power battery quantity prediction system based on multi-dimensional feature ML, comprising: The acquisition module is used to obtain the number of new energy vehicles insured within a preset time period and N types of city characteristic values ​​to construct a data set; A determination module, used to analyze the N types of city characteristic values, and determine the key city characteristic values ​​and their weight values ​​that affect the insurance volume of new energy vehicles; A training module, used for filtering data sets according to the key characteristic values ​​of the cities and their weight values, and using the filtered data sets to train a variety of tuned machine learning models; The selection module is used to evaluate the prediction performance of each machine learning model and select the best model for predicting the insurance volume of passenger cars and commercial vehicles respectively; The processing module is used to input the key characteristic values ​​of the future city into the optimal model to predict the insurance volume of new energy vehicles. Based on the predicted insurance volume of new energy vehicles, the proportion of battery types and battery characteristics of new energy vehicles are used, combined with the Weibull life distribution, to predict the amount of retired power batteries in the city.

[0017] In a third aspect, an embodiment of the present invention further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a method for predicting the amount of retired power batteries at the city level based on multi-dimensional features ML in the aforementioned embodiment are implemented.

[0018] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for predicting the amount of retired power batteries at the city level based on multi-dimensional features ML in the aforementioned embodiment.

[0019] One or more technical solutions provided in the embodiments of the present invention have at least the following technical effects or advantages: First, in terms of data processing and analysis, through comprehensive consideration of N types of city characteristic values, we have broken through the limitation of traditional prediction of new energy vehicle insurance volume relying only on a single or a small number of data dimensions. By determining the key characteristic values ​​of cities, we can accurately identify the factors that play a key role in the insurance volume of new energy vehicles in different cities, effectively reducing data redundancy and irrelevant information interference. At the same time, determining the weight value of each key feature further highlights the influence of important factors in the prediction model, making the model input data more targeted and representative, and laying a solid foundation for subsequent accurate predictions.

[0020] Secondly, in the model construction and training phase, a variety of machine learning models were used and optimized. After optimization, the determination coefficient of the best model increased by more than 0.85, and the mean square error decreased to nearly 0.

[0021] Furthermore, in terms of the accuracy and practicality of the prediction, model evaluation and optimal model selection are carried out for passenger cars and commercial vehicles respectively, taking into full account the differences in urban characteristics between these two types of vehicles. The present invention has undergone multiple sets of rigorous cross-validation and time series analysis experiments. Stable and reliable prediction results have been obtained in different future city key characteristic value data. This makes the prediction results more in line with the insurance situation of different models in the actual market, and thus obtains accurate annual consumer demand for new energy vehicles.

[0022] Finally, based on the precise demand for new energy passenger vehicles and commercial vehicles, it is possible to accurately predict the amount of retired power batteries at the city level, effectively improve the accuracy and practicality of the "city-provincial-regional" retired power battery forecast, and promote the rational allocation of resources and healthy development of the power battery recycling industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a flow chart of a method for predicting the amount of retired power batteries at the city level based on multi-dimensional feature ML in one embodiment of the present invention; Figure 2 A schematic diagram of cluster analysis results in one embodiment of the present invention; Figure 3 This is a schematic diagram of the principal component analysis result in one embodiment of the present invention; Figure 4 A performance comparison diagram of a machine learning model in one embodiment of the present invention; Figure 5 A diagram showing the installed capacity ratio of 24 battery types in a certain period of time in an embodiment of the present invention; Figure 6 1 is a distribution diagram of Weibull life model parameters for 24 battery types in one embodiment of the present invention. DETAILED DESCRIPTION

[0024] The embodiments of the present invention provide a city-level retired power battery quantity prediction method and system based on multi-dimensional feature ML, which solves the problem that the existing retired power battery quantity prediction method cannot meet actual needs.

[0025] The technical solution of the embodiment of the present invention is to solve the above problems, and the overall idea is as follows: First, in terms of data collection, the present invention is committed to obtaining comprehensive and multi-dimensional data. Data on the number of new energy vehicles insured within a preset time period are collected from multiple reliable data sources. These data can directly reflect the use of new energy vehicles in the actual market. At the same time, the present invention will widely collect N types of city characteristic values, which cover the key factors affecting the number of new energy vehicles insured.

[0026] Secondly, conduct in-depth analysis on the collected data. Use data mining and analysis techniques to process the city characteristic values. For example, use cluster analysis and other methods to reasonably cluster cities, dig out the differences in various characteristics of different city clusters, and then determine the key characteristic values ​​of cities that have a significant impact on the insurance volume of new energy vehicles. This step can filter out the data most relevant to the prediction target from a large number of complex data, reduce data noise and interference from irrelevant information, and improve data quality.

[0027] Then, the data set is further screened based on the key characteristic values ​​of the screened cities, retaining those data that are valuable for predicting the insurance volume of new energy vehicles and removing redundant data to obtain a refined and highly relevant data set.

[0028] Next, the machine learning model is trained using the filtered data set. The present invention is not limited to a single machine learning model, but comprehensively considers a variety of different types of machine learning models, such as long short-term memory network model, random forest model, XGboost model, ANN neural network model and stacked regression model. By inputting the filtered data sets into these different machine learning models for training, the machine learning model learns the intrinsic relationship between the key characteristic values ​​of the city and the insurance volume of new energy vehicles.

[0029] After the model training is completed, each machine learning model is comprehensively and carefully evaluated. Using scientific and reasonable evaluation indicators such as mean square error and determination coefficient, the prediction performance of each machine learning model is quantitatively analyzed, so as to accurately compare the advantages and disadvantages of each machine learning model.

[0030] Thirdly, the best performing model is selected from the evaluated machine learning models, and the key characteristic values ​​of future cities are set based on reasonable assumptions about possible future scenarios. These set key characteristic values ​​of future cities are input into the best model to accurately predict the future insurance volume of new energy vehicles.

[0031] Finally, after obtaining the forecast results of the insurance volume of new energy passenger cars and commercial vehicles in each city, based on the predicted insurance volume of new energy vehicles, the proportion of battery types and battery characteristics of new energy vehicles, combined with the Weibull life distribution, the forecast of the number of retired power batteries in each city is realized.

[0032] In summary, through this systematic and comprehensive technical idea, the present invention can effectively overcome the problems existing in the prior art in predicting the amount of retired power batteries, such as single data dimension, single model with poor applicability, and failure to consider differences in urban characteristics, and provide reliable demand forecasting data support for the development of the power battery recycling industry.

[0033] like Figure 1 As shown, the present invention provides a city-level retired power battery quantity prediction method based on multi-dimensional feature ML, including steps S1 to S6.

[0034] Step S1: Obtain the number of new energy vehicles insured within a preset time period and N types of city characteristic values ​​to construct a data set. The N types of city characteristic values ​​include city population, city GDP, urbanization rate, Baidu index, and city new energy vehicle promotion subsidy amount.

[0035] In the specific implementation process, for example: obtain the insurance volume data of new energy vehicles within a preset time period from the database of relevant insurance institutions, and collect various city characteristic values, such as the city population, urbanization rate, and city GDP from the city statistics department, obtain the attention of urban netizens to new energy vehicles from Baidu Index, obtain environmental protection policy documents from relevant government departments and quantify them into the amount of subsidies for the promotion of new energy vehicles in the city, etc. Finally, the above data constitute a data set.

[0036] Step S2: Analyze the N types of city characteristic values ​​to determine the city key characteristic values ​​and their weight values ​​that affect the insurance volume of new energy vehicles. The method for analyzing the N types of city characteristic values ​​to determine the city key characteristic values ​​that affect the insurance volume of new energy vehicles is specifically a cluster analysis method, and the method for determining the weight values ​​of the city key characteristics is specifically a principal component analysis method.

[0037] In the specific implementation process, for example, cluster analysis is performed on the collected city feature values, and the K-means algorithm is used to divide the cities into several groups according to feature similarity. The objective function of the K-means algorithm is expressed as: , In the formula, Represents the objective function of clustering. The smaller its value is, the better the clustering effect is, that is, the smaller the sum of the distances from the sample points in the cluster to the center of the cluster to which it belongs is; : represents the number of clusters, that is, the number of groups into which the data is to be divided; : is an index variable used to traverse clusters, from 1 to ; :Indicates the The set of sample points contained in a cluster; : represents a sample point in the data set; :Indicates the The cluster center of a cluster is a vector with the same dimension as the sample point; Represents sample points To Cluster Centers The Euclidean distance of , the distance is squared in the objective function.

[0038] like Figure 2 As shown, the economic GDP and urban population in urban characteristics are divided into 7 groups, including cluster 1 to cluster 7.

[0039] The clustering effect is evaluated by the Calinski-Harabasz index indicator, and the optimal number of clusters is selected. Principal component analysis is performed on each city group after clustering to extract the main components and reduce the feature dimension. The mathematical principle of principal component analysis: Let the original data matrix be X, X is A data matrix with each column is a sample, each row is a feature. Compute the covariance matrix: , T represents the transpose of the matrix. Covariance matrix is a symmetric matrix that reflects the linear relationship and variance between the variables in the original data. Calculate the eigenvalues ​​and eigenvectors of the covariance matrix: , yes The characteristic value of is the eigenvector of the matrix corresponding to the eigenvalue. ,in is the eigenvector The matrix is ​​arranged in columns, that is, , is a The matrix, each column Indicates The score of the principal component, that is, the original data is The projection values ​​on the principal components.

[0040] like Figure 3 As shown in the figure, under different clusters, the correlation between the insurance volume of new energy passenger vehicles and commercial vehicles and various urban characteristics varies significantly. For example, for the urban agglomeration in cluster 6, the weight of urbanization rate on the insurance volume of commercial vehicles is higher than that on the insurance volume of passenger vehicles.

[0041] Step S3: Filter the data set according to the key characteristic values ​​of the city and their weight values, and use the filtered data set to train multiple optimized machine learning models. The machine learning models include LSTM model, random forest model, XGBoost model, ANN model and stacked regression model, and the tuning method is specifically Bayesian optimization method or hyperparameter tuning method.

[0042] In the specific implementation process, for example, in order to improve the generalization ability of the model, the Bayesian optimization method was used to fine-tune the hyperparameters during the training process, and a variety of advanced ML (Machine Learning) algorithms were introduced, including: LSTM (Long Short-Term Memory), random forest, XGBoost (eXtreme Gradient Boosting), ANN (Artificial Neural Network) and stacked regression, in order to better capture the changing trend of the insurance volume of new energy vehicles. In addition, due to the significant differences between different cities, it is necessary to analyze the special conditions of the city before applying it to a specific city, and adjust the structure or parameters of the machine learning model accordingly to ensure that the prediction results are both widely applicable and in line with local reality. In order to achieve the final training effect, the following steps S31 to S33 are also included.

[0043] Step S31: construct an LSTM model, a random forest model, an XGBoost model, an ANN model and a stacked regression model.

[0044] LSTM model: Use the Keras library to build an LSTM network. The output calculation formula of the LSTM unit is: , , , , , In the formula, is the input gate, which is a value vector between 0 and 1; It is the sigmoid activation function, which maps the input between 0 and 1; is the weight matrix input to the input gate; is the input data at time step; is the weight matrix from the hidden state to the input gate at the previous time step; is the hidden state at the previous time step; is the bias vector of the input gate, which is used to adjust the output of the input gate; It is the Gate of Oblivion; Represents the weight matrix input to the forget gate; is the hidden state at the previous time step To the weight matrix of the forget gate; is the bias vector of the forget gate; is the time step The updated cell state is the core of LSTM long-term memory; Represents element-wise multiplication; is the cell state at the previous time step; represents the hyperbolic tangent activation function, which maps the input between -1 and 1; Is input to The weight matrix of the cell state update part; is the weight matrix of the updated part from the hidden state to the cell state in the previous time step; is the bias vector of the cell state update part; is at the time step The output gate of Yes Input The weight matrix to the output gate; is the hidden state at the previous time step The weight matrix to the output gate; is the bias vector of the output gate; At time step The hidden state of is one of the outputs of LSTM and will be passed to the next time step or used for subsequent tasks.

[0045] Random forest model: Use the Scikit-learn library to configure a random forest classifier. A random forest consists of multiple decision trees, and the construction process of each tree is as follows: randomly select a subset of samples; randomly select a subset of features at each node; perform the best split based on the selected feature subset; repeat the above process until the stopping condition is met.

[0046] XGBoost model: A distributed gradient boosting library based on the optimization implementation of the gradient boosting framework. It builds the final model by gradually optimizing the objective function through multiple rounds of iterations. Its objective function includes the loss function and regularization term : , In the formula, Indicated in The objective function value at the round iteration; represents the number of samples; is the sample index; It is The true label value of each sample; is the loss function, It is The predicted value of the round, is the predicted value of the newly added tree.

[0047] ANN model: Define a simple fully connected neural network structure through TensorFlow or PyTorch. Assume that the network has two hidden layers, and the output of each layer can be expressed as: , , In the formula, It is The weighted inputs of the layer; and They are the weight matrix and bias vector, which are continuously updated according to the optimization algorithm used to minimize the loss function; is the activation function; No. The activation output of the layer.

[0048] Stacked regression model: Use the stacked regression model as the base estimator and integrate their outputs through another meta-model (such as linear regression). The prediction process of stacked regression can be formalized as: , In the formula, represents the predicted values ​​of the stacked regression model; middle is the base estimator (it can be multiple base models, such as multiple different regression models), is the input data (a data set containing multiple features). The base estimator Process and predict to get a set of intermediate results. These intermediate results can be regarded as a feature transformation or preliminary prediction of the original data; is a meta-model, usually a simple model (such as linear regression) that combines the underlying estimators The output of is taken as input, and this information is further integrated to finally output the predicted value. .

[0049] Step S32: Divide the data set into a training set and a test set, and use the training set to train the machine learning model.

[0050] Step S33: Optimize the machine learning model, including Bayesian optimization and hyperparameter tuning.

[0051] Bayesian optimization: Use Bayesian optimization methods to tune the model's hyperparameters, such as learning rate, hidden layer size, batch size, etc. The core idea of ​​Bayesian optimization is to approximate the objective function by building a probability model, and use the probability model for sampling to gradually approach the optimal solution.

[0052] Hyperparameter tuning: Further optimize the model's hyperparameters through methods such as grid search or random search.

[0053] Step S4: Evaluate the prediction performance of each machine learning model and select the best model for predicting the insurance volume of passenger cars and commercial vehicles. The evaluation indicators used to evaluate the prediction performance of each machine learning model include mean square error and determination coefficient.

[0054] In the specific implementation process, for example: through the determination coefficient The prediction performance of each machine learning model is evaluated by the mean square error (MSE) indicator, and the best models suitable for the insurance volume of new energy passenger vehicles and new energy commercial vehicles are comprehensively selected.

[0055] Coefficient of determination The expression is as follows: , The expression of mean square error MSE is as follows: , In the formula, represents the number of samples, No. The actual observed value of samples, The model predicts the The predicted value of samples, is the average of all sample actual observations, is the residual sum of squares, is the total sum of squares.

[0056] like Figure 4 As shown, through the coefficient of determination From the MSE and mean square error results, it can be concluded that the best model for predicting the insurance volume of new energy passenger vehicles is the LSTM model, and the best model for predicting the insurance volume of new energy commercial vehicles is the stacked regression model.

[0057] Step S5: Input the key characteristic values ​​of the future city into the optimal model to predict the insurance volume of new energy vehicles. The key characteristic values ​​of the future city include the future city GDP, the future city population, the future urbanization rate, the future Baidu index, and the future city new energy vehicle promotion subsidy amount.

[0058] In the specific implementation process, for example, it is necessary to accurately obtain and organize the key characteristic values ​​of various future cities. The GDP of future cities is determined by comprehensive economic experts' forecasts, regional economic development plans, and macroeconomic trend analysis. The future urban population is determined by referring to census data, population migration trend research, and population growth targets in urban planning. The future urbanization rate is estimated based on information such as the city's urbanization promotion plan and land use planning. The Baidu Index reflects the changing trend of public attention to new energy vehicles, and predicts its future direction by continuously monitoring search engine data and social media heat analysis. The amount of subsidies for the promotion of new energy vehicles in future cities is estimated based on the direction of urban environmental protection subsidy policies.

[0059] After accurately inputting these future key characteristic parameters covering urban economy, population, policy and attention into the best model that has been carefully trained and screened, the best model will perform efficient calculations and in-depth reasoning based on its internal complex algorithm architecture and the inherent logical relationship between the characteristic values ​​of each city and the number of new energy vehicles insured during the previous training process. The best model will comprehensively consider the interaction and synergistic influence between various parameters, accurately predict and calculate the number of new energy vehicles insured in different cities under specific scenarios in the future, and output the predicted value of the number of new energy vehicles insured.

[0060] Step S6: Based on the predicted number of new energy vehicles insured, using the battery type proportion and battery characteristics of new energy vehicles, combined with Weibull life distribution, predict the number of retired power batteries at the city level. Among them, the battery types specifically include 24 types consisting of LFP batteries, NCM111 batteries, NCM523 batteries, NCM622 batteries, NCM811 batteries, and NCA batteries for pure electric passenger cars, hybrid passenger cars, pure electric commercial vehicles, and hybrid commercial vehicles; battery characteristics specifically include battery life, battery weight, and battery capacity.

[0061] In the specific implementation process, for example: 24 types of new energy vehicle batteries include: passenger car pure electric LFP battery, passenger car pure electric NCM111 battery, passenger car pure electric NCM523 battery, passenger car pure electric NCM622 battery, passenger car pure electric NCM811 battery, passenger car pure electric NCA battery, passenger car hybrid LFP battery, passenger car hybrid NCM111 battery, passenger car hybrid NCM523 battery, passenger car hybrid NCM622 battery, passenger car hybrid NCM811 battery Batteries, passenger car hybrid NCA batteries, commercial vehicle pure electric LFP batteries, commercial vehicle pure electric NCM111 batteries, commercial vehicle pure electric NCM523 batteries, commercial vehicle pure electric NCM622 batteries, commercial vehicle pure electric NCM811 batteries, commercial vehicle pure electric NCA batteries, commercial vehicle hybrid LFP batteries, commercial vehicle hybrid NCM111 batteries, commercial vehicle hybrid NCM523 batteries, commercial vehicle hybrid NCM622 batteries, commercial vehicle hybrid NCM811 batteries, commercial vehicle hybrid NCA batteries.

[0062] The installed capacity of 24 new energy vehicle battery types in a certain period of time Figure 5 As shown. The two-parameter Weibull distribution model that is closest to the actual use of power batteries is used for estimation. , the Weibull distribution function can be expressed as ,in Represents time, the probability density function formula of Weibull distribution is: , in: Is the model The shape parameter determines the shape of the distribution, affecting the curvature and retirement mode of the function. The value can indicate different failure characteristics. For example, when When , it usually indicates an early retirement mode, that is, the vehicle battery has a higher probability of retirement in the early stage of use; when When , the Weibull distribution degenerates into an exponential distribution, which can be used to describe random retirement; when When , it usually indicates a wear-out retirement mode, that is, the power battery retirement increases with time. Is the model The scale parameter is a proportional parameter related to the scale of the distribution and determines the range of the distribution, such as the peak position and width of the distribution. Figure 6 As shown, for each type of battery, different and values ​​to reflect the differences in life span of various types of batteries.

[0063] The cumulative distribution function of the Weibull distribution is expressed in terms of the calculation time The probability of power battery retirement within 1 year, cumulative distribution function The formula is: .

[0064] Based on the insurance data of new energy vehicles, The year when new energy vehicles are insured. To calculate the year, For the city In the insurance year of new energy vehicles insured, Year of insurance Medium Model The proportion of For car models On the insurance The probability of Weibull retirement after 2017 is: For car models The weight of the power battery, For car models Power battery capacity. City In calculation year Medium Model Weight of retired batteries The calculation formula is: , City In calculation year Medium Model Retired battery capacity The calculation formula is: , Weibull retirement probability The calculation formula is: , After calculating the weight and capacity of each type of retired power batteries in each city each year, the retired power batteries from passenger cars or commercial vehicles, the retired power batteries of LFP or NMC types, the total retired power batteries in the city, etc. can be calculated based on demand. The refined results can provide data science support for the company's layout strategy at the "city-province-region" level.

[0065] It should be noted that the present invention can predict the amount of retired power batteries in 364 inland cities, including 293 prefecture-level cities, 4 municipalities, 7 regions, 30 autonomous prefectures, 3 leagues, and 27 counties.

[0066] In a second aspect, an embodiment of the present invention further provides a city-level retired power battery quantity prediction system based on multi-dimensional feature ML, comprising: The acquisition module is used to obtain the number of new energy vehicles insured within a preset time period and N types of city characteristic values ​​to construct a data set; A determination module is used to analyze N types of city characteristic values ​​and determine the key characteristic values ​​of cities and their weight values ​​that affect the insurance volume of new energy vehicles; The training module is used to filter the data set according to the key characteristic values ​​of the city and their weight values, and use the filtered data set to train multiple fine-tuned machine learning models; The selection module is used to evaluate the prediction performance of each machine learning model and select the best model for predicting the insurance volume of passenger cars and commercial vehicles respectively; The processing module is used to input the key characteristic values ​​of the future city into the optimal model to predict the insurance volume of new energy vehicles. Based on the predicted insurance volume of new energy vehicles, the proportion of battery types and battery characteristics of new energy vehicles are used, combined with the Weibull life distribution, to predict the amount of retired power batteries in the city.

[0067] From the perspective of system and application, the city-level retired power battery quantity prediction system based on multi-dimensional feature ML provided by the present invention realizes efficient integration and convenient application of technical solutions. Each module in the system has a clear division of labor and works together, forming a complete closed-loop process from data acquisition, feature analysis, model training to final prediction processing.

[0068] In a third aspect, an embodiment of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a method for predicting the amount of retired power batteries at the city level based on multi-dimensional features ML in the aforementioned embodiment are implemented.

[0069] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for predicting the amount of retired power batteries at the city level based on multi-dimensional features ML in the aforementioned embodiment.

[0070] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for predicting the amount of retired power batteries at the city level based on multi-dimensional feature ML, characterized in that: include: Obtain the number of new energy vehicles insured within a preset time period and N types of city characteristic values ​​to construct a data set; Analyze the N types of city characteristic values ​​to determine the key city characteristic values ​​and their weight values ​​that affect the insurance volume of new energy vehicles; Filtering data sets according to the key characteristic values ​​of the cities and their weight values, and using the filtered data sets to train multiple tuned machine learning models; Evaluate the prediction performance of each machine learning model and select the best model for predicting the insurance volume of passenger cars and commercial vehicles respectively; Input the key characteristic values ​​of future cities into the optimal model to predict the insurance volume of new energy vehicles; Based on the predicted number of new energy vehicles insured, the proportion of battery types and battery characteristics of new energy vehicles, combined with the Weibull life distribution, the number of retired power batteries at the city level is predicted.

2. The method according to claim 1, characterized in that The N types of city characteristic values ​​include city population, city GDP, urbanization rate, Baidu index, and city new energy vehicle promotion subsidy amount.

3. The method according to claim 1, characterized in that The method for analyzing the N types of city characteristic values ​​to determine the city key characteristic values ​​that affect the insurance volume of new energy vehicles is specifically a cluster analysis method, and the method for determining the weight values ​​of the city key characteristics is specifically a principal component analysis method.

4. The method according to claim 1, characterized in that The machine learning models include LSTM models, random forest models, XGBoost models, ANN models and stacked regression models, and the tuning method is specifically a Bayesian optimization method or a hyperparameter tuning method.

5. The method according to claim 1, characterized in that The evaluation indicators used to evaluate the prediction performance of each machine learning model include mean square error and determination coefficient.

6. The method according to claim 1, characterized in that The key characteristic values ​​of the future city include the future city GDP, the future city population, the future urbanization rate, the future Baidu index, and the future city new energy vehicle promotion subsidy amount.

7. The method according to claim 1, characterized in that The battery types specifically include 24 types consisting of LFP batteries, NCM111 batteries, NCM523 batteries, NCM622 batteries, NCM811 batteries, and NCA batteries for pure electric passenger cars, hybrid electric passenger cars, pure electric commercial vehicles, and hybrid electric commercial vehicles; the battery characteristics specifically include battery life, battery weight, and battery capacity.

8. A city-level retired power battery quantity prediction system based on multi-dimensional feature ML, characterized in that: include: The acquisition module is used to obtain the number of new energy vehicles insured within a preset time period and N types of city characteristic values ​​to construct a data set; A determination module, used to analyze the N types of city characteristic values, and determine the key city characteristic values ​​and their weight values ​​that affect the insurance volume of new energy vehicles; A training module, used for filtering data sets according to the key characteristic values ​​of the cities and their weight values, and using the filtered data sets to train a variety of tuned machine learning models; The selection module is used to evaluate the prediction performance of each machine learning model and select the best model for predicting the insurance volume of passenger cars and commercial vehicles respectively; The processing module is used to input the key characteristic values ​​of the future city into the optimal model to predict the insurance volume of new energy vehicles. Based on the predicted insurance volume of new energy vehicles, the proportion of battery types and battery characteristics of new energy vehicles are used, combined with the Weibull life distribution, to predict the amount of retired power batteries in the city.

9. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images