Air purification method and vehicle

Abstract

The invention provides an air purification method and a vehicle, the method is applied to the technical field of vehicle environment control, and the method comprises the steps that environment air state data of the vehicle, filter element state data of an air purification system in the vehicle and operation energy consumption data of the air purification system are obtained; determining a comfort score value based on the ambient air state data; determining filter element performance parameters based on the filter element state data; determining a target purification strategy based on the comfort level score value, the filter element performance parameters and the operation energy consumption data; and controlling the air purification system to work by adopting the target purification strategy. According to the method, personalized intelligent balance control among the air comfort perceived by a user, the energy consumption of the air purification system and the performance of the filter element can be achieved, the air quality in the vehicle can be effectively controlled under the different requirements that the air comfort is preferential, the energy consumption is preferential or the service life loss of the filter element is preferential, and therefore the driving experience of the user is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of vehicle environment control, and more particularly, to an air purification method and a vehicle in the technical field of vehicle environment control. BACKGROUND

[0002] With the popularity of vehicles, cars have become one of the main means of transportation for people's daily travel. In the use process of the vehicle, the air quality in the vehicle will affect the driving experience of the driver and the passenger.

[0003] At present, the air quality in the vehicle is mainly monitored in real time by installing an air quality sensor in the vehicle, such as monitoring particulate matter 2.5 (PM2.5), and comparing the monitored air quality index in the vehicle with a preset concentration threshold to switch the purification strategy of the air purification system in the vehicle.

[0004] However, the above-mentioned air purification strategy based on the concentration threshold is difficult to effectively control the air quality in the vehicle under different needs, thereby affecting the driving experience of the user. SUMMARY

[0005] The present application provides an air purification method and a vehicle, which can effectively control the air quality in the vehicle under different needs such as air comfort priority, energy consumption priority or filter core life loss priority, thereby improving the driving experience of the user.

[0006] In a first aspect, an air purification method is provided, applied to a vehicle, the method comprising: obtaining environmental air state data of the vehicle, filter core state data of an air purification system in the vehicle, and running energy consumption data of the air purification system; determining a comfort score value based on the environmental air state data, the comfort score value being used to reflect the satisfaction degree of the user to the air quality in the vehicle; determining a filter core performance parameter based on the filter core state data, the filter core performance parameter being used to reflect the purification ability, pressure drop characteristic and service life of the purification filter core in the air purification system; determining a target purification strategy based on the comfort score value, the filter core performance parameter and the running energy consumption data; and controlling the air purification system to work by using the target purification strategy.

[0007] Through the above technical solution, the target purification strategy can be determined based on the comfort score value, the filter core performance parameter and the running energy consumption data, realizing the individualized intelligent balance control between the air comfort perceived by the user, the energy consumption of the air purification system and the filter core performance. It can effectively control the air quality in the vehicle under different needs such as air comfort priority, energy consumption priority or filter core life loss priority, so as to realize the individualized, dynamic and efficient control of the air quality in the vehicle, thereby improving the driving experience of the user.

[0008] In a possible implementation manner of the first aspect, the determining the comfort score value based on the environment air state data comprises: processing the environment air state data by using a comfort representation model to obtain a feature latent variable vector corresponding to the environment air state data; and processing the feature latent variable vector by using a score prediction model to obtain the comfort score value.

[0009] By the above technical solution, the comfort score value can be accurately measured based on the comfort representation model and the score prediction model.

[0010] In a possible implementation manner of the first aspect and the above implementation manner, the processing the environment air state data by using the comfort representation model to obtain the feature latent variable vector corresponding to the environment air state data comprises: processing the environment air state data by using an encoder in the comfort representation model to obtain a mean vector corresponding to the environment air state data and a standard deviation vector corresponding to the environment air state data; and performing reparameterization sampling on the mean vector and the standard deviation vector by using a reparameterization layer in the comfort representation model to obtain the feature latent variable vector.

[0011] By the above technical solution, the feature latent variable vector can be determined based on the encoder and the reparameterization layer in the comfort representation model, so as to implicitly capture the change mode in the environment air state data that has a decisive effect on the subjective perception of the user.

[0012] In a possible implementation manner of the first aspect and the above implementation manner, the filter state data comprises first state data, second state data and third state data, the first state data is related to a purification capability of the purification filter, the second state data is related to a pressure drop characteristic of the purification filter, and the third state data is related to a service life of the purification filter; the determining the filter performance parameter based on the filter state data comprises: processing the first state data by using a purification efficiency model to obtain a purification efficiency percentage, the purification efficiency percentage being used to reflect the purification capability of the purification filter; processing the second state data by using a pressure drop characteristic model to obtain a pressure drop level, the pressure drop level being used to reflect the pressure drop characteristic of the purification filter; and processing the third state data by using a life consumption model to obtain a filter remaining life estimate, the filter remaining life estimate being used to reflect the service life of the purification filter.

[0013] By the above technical solution, the filter performance parameter can be accurately determined based on the purification efficiency model, the pressure drop characteristic model and the life consumption model.

[0014] In a possible implementation of the first aspect and the above implementation, based on the comfort score value, the filter performance parameter, and the operation energy consumption data, the target purification strategy is determined, including: obtaining a plurality of purification strategy combinations, each of which includes a plurality of purification strategy individuals; determining a non-dominated solution corresponding to each of the purification strategy combinations based on each of the purification strategy combinations, the comfort score value, the filter performance parameter, and the operation energy consumption data; determining a non-dominated solution set based on the non-dominated solutions corresponding to each of the plurality of purification strategy combinations; and determining the target purification strategy from the non-dominated solution set, the target purification strategy being one of the non-dominated solutions in the non-dominated solution set.

[0015] Through the above technical solution, the non-dominated solution set can be determined based on the multi-objective optimization model, so that a most suitable non-dominated solution can be selected from the non-dominated solution set as the target purification strategy according to the user preference configuration or the current working mode of the air purification system.

[0016] In a possible implementation of the first aspect and the above implementation, the plurality of purification strategy combinations include an initial purification strategy combination and at least one target purification strategy combination; each of the target purification strategy combinations is obtained by performing a crossover operation and / or a mutation operation on the initial purification strategy combination, the crossover operation being used to exchange part of the control parameters in any two purification strategy individuals in the initial purification strategy combination, and the mutation operation being used to change part of the control parameters in at least one purification strategy individual in the initial purification strategy combination.

[0017] Through the above technical solution, the crossover operation and / or the mutation operation can be used to ensure the population diversity and expand the search space during the search process.

[0018] In a possible implementation of the first aspect and the above implementation, based on each of the purification strategy combinations, the comfort score value, the filter performance parameter, and the operation energy consumption data, the non-dominated solution corresponding to each of the purification strategy combinations is determined, including: determining a first feedback value corresponding to each of the purification strategy individuals based on each of the purification strategy individuals in the purification strategy combination and the comfort score value; determining a second feedback value corresponding to each of the purification strategy individuals based on each of the purification strategy individuals in the purification strategy combination and the filter performance parameter; determining a third feedback value corresponding to each of the purification strategy individuals based on each of the purification strategy individuals in the purification strategy combination and the operation energy consumption data; and performing non-dominated sorting on each of the purification strategy individuals in the purification strategy combination based on the first feedback value, the second feedback value, and the third feedback value to obtain the non-dominated solution corresponding to the purification strategy combination.

[0019] By the technical solution, based on the first feedback value, the second feedback value and the third feedback value, each purification strategy individual in the purification strategy combination is non-dominantly sorted, so as to filter out an optimal solution in which one purification strategy individual is not suppressed by other purification strategy individuals in all target dimensions, and a non-dominant solution corresponding to the purification strategy combination is obtained.

[0020] In combination with the first aspect and the above implementation manners, in some possible implementation manners, the target purification strategy is determined from the non-dominant solution set, including: based on a user preference configuration or a current working mode of the air purification system, the target purification strategy is determined from the non-dominant solution set.

[0021] By the technical solution, the target purification strategy can be determined according to different requirements.

[0022] In combination with the first aspect and the above implementation manners, in some possible implementation manners, the target purification strategy includes at least one of the following: an air circulation mode, a fan air volume level, a switching time interval of the air purification strategy, and a duration of the purification mode.

[0023] In a second aspect, an air purification device is provided, which is applied to a vehicle. The device includes: a data acquisition module, configured to acquire environmental air state data of the vehicle, filter element state data of an air purification system in the vehicle, and operation energy consumption data of the air purification system; a first determination module, configured to determine a comfort score value based on the environmental air state data, the comfort score value being used to reflect a satisfaction degree of a user on air quality in the vehicle; a second determination module, configured to determine a filter element performance parameter based on the filter element state data, the filter element performance parameter being used to reflect a purification ability, a pressure drop characteristic and a service life of a purification filter element in the air purification system; a third determination module, configured to determine a target purification strategy based on the comfort score value, the filter element performance parameter and the operation energy consumption data; and a control module, configured to control the air purification system to work by using the target purification strategy.

[0024] In combination with the second aspect, in some possible implementation manners, the first determination module is specifically configured to process the environmental air state data by using a comfort representation model to obtain a feature latent variable vector corresponding to the environmental air state data; and process the feature latent variable vector by using a score prediction model to obtain the comfort score value.

[0025] In combination with the second aspect and the above implementation manners, in some possible implementation manners, the first determination module is specifically configured to process the environmental air state data by using an encoder in the comfort representation model to obtain a mean vector corresponding to the environmental air state data and a standard deviation vector corresponding to the environmental air state data; and perform reparameterization sampling on the mean vector and the standard deviation vector by using a reparameterization layer in the comfort representation model to obtain the feature latent variable vector.

[0026] With reference to the second aspect and the foregoing implementation manners, in some possible implementation manners, the filter state data includes first state data, second state data, and third state data, the first state data is related to a purification capability of the purification filter, the second state data is related to a pressure drop characteristic of the purification filter, and the third state data is related to a service life of the purification filter; the second determining module is specifically configured to process the first state data by using a purification efficiency model to obtain a purification efficiency percentage, the purification efficiency percentage being used to reflect the purification capability of the purification filter; process the second state data by using a pressure drop characteristic model to obtain a pressure drop level, the pressure drop level being used to reflect the pressure drop characteristic of the purification filter; and process the third state data by using a service life consumption model to obtain a filter residual life estimate, the filter residual life estimate being used to reflect the service life of the purification filter.

[0027] With reference to the second aspect and the foregoing implementation manners, in some possible implementation manners, the third determining module is specifically configured to obtain a plurality of purification strategy combinations, each of which includes a plurality of purification strategy individuals; determine, based on each of the purification strategy combinations, the comfort score value, the filter performance parameter, and the operation energy consumption data, a non-dominated solution corresponding to each of the purification strategy combinations; determine, based on the non-dominated solution corresponding to each of the plurality of purification strategy combinations, a non-dominated solution set; and determine, from the non-dominated solution set, a target purification strategy, the target purification strategy being one of the non-dominated solutions in the non-dominated solution set.

[0028] With reference to the second aspect and the foregoing implementation manners, in some possible implementation manners, the plurality of purification strategy combinations include an initial purification strategy combination and at least one target purification strategy combination; wherein each of the target purification strategy combinations is obtained by performing a crossover operation and / or a mutation operation on the initial purification strategy combination, the crossover operation being used to exchange part of the control parameters in any two of the purification strategy individuals in the initial purification strategy combination, and the mutation operation being used to change part of the control parameters in at least one of the purification strategy individuals in the initial purification strategy combination.

[0029] With reference to the second aspect and the foregoing implementation manners, in some possible implementation manners, the third determining module is specifically configured to determine, based on each of the purification strategy individuals in the purification strategy combination and the comfort score value, a first feedback value corresponding to each of the purification strategy individuals; determine, based on each of the purification strategy individuals in the purification strategy combination and the filter performance parameter, a second feedback value corresponding to each of the purification strategy individuals; determine, based on each of the purification strategy individuals in the purification strategy combination and the operation energy consumption data, a third feedback value corresponding to each of the purification strategy individuals; and perform non-dominated sorting on each of the purification strategy individuals in the purification strategy combination based on the first feedback value, the second feedback value, and the third feedback value, to obtain the non-dominated solution corresponding to the purification strategy combination.

[0030] In a possible implementation of the second aspect and the foregoing implementation, the third determining module is specifically configured to determine the target purification strategy from the non-dominated solution set based on a user preference configuration or a current working mode of the air purification system.

[0031] In a possible implementation of the second aspect and the foregoing implementation, the target purification strategy includes at least one of an air circulation mode, a fan air volume level, a switching time interval of the air purification strategy, and a duration of the purification mode.

[0032] In a third aspect, a vehicle is provided, including: a memory configured to store executable program code; and a processor configured to invoke and run the executable program code from the memory, so that the vehicle executes the method in the first aspect or any possible implementation of the first aspect.

[0033] In a fourth aspect, a computer program product is provided, which includes computer program code configured to cause a computer to execute the method in the first aspect or any possible implementation of the first aspect when the computer program code is run on the computer.

[0034] In a fifth aspect, a computer-readable storage medium is provided, which stores computer program code configured to cause a computer to execute the method in the first aspect or any possible implementation of the first aspect when the computer program code is run on the computer.

[0035] The possible implementations of the second aspect to the fifth aspect have similar effects to those of the first aspect and the possible implementations of the first aspect, and will not be described here. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 is an application scenario diagram of an air purification method provided by an embodiment of the present application; Figure 2 is a flowchart of an air purification method provided by an embodiment of the present application; Figure 3 is a structural diagram of an air purification device provided by an embodiment of the present application; Figure 4 is a structural diagram of a vehicle provided by an embodiment of the present application. DETAILED DESCRIPTION

[0037] The technical solutions in the present application will be described clearly and exhaustively below with reference to the drawings. In the description of the embodiments of the present application, unless otherwise specified, " / " represents the meaning of or, for example, A / B can represent A or B: "and / or" in the text is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which can represent: A exists alone, A and B exist together, and B exists alone, in addition, in the description of the embodiments of the present application, "multiple" means two or more than two.

[0038] Hereinafter, the terms "first" and "second" are used only for descriptive purposes and cannot be understood as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first" and "second" can explicitly or implicitly include one or more of the features.

[0039] In order to facilitate the understanding of the technical solutions of the embodiments of the present application, some terms involved in the embodiments of the present application are simply explained below.

[0040] Variational auto-encoder (VAE) model: a deep learning-based generative model that models the latent representation of data by introducing a probabilistic framework, which not only realizes data compression and feature extraction, but also generates new samples similar to the training data.

[0041] Among them, the variational auto-encoder model can include an encoder, a reparameterization layer and a decoder. Specifically, the input data is converted into the distribution of the latent space by the encoder. The encoder is usually composed of several layers of neural networks, and outputs the mean and variance of the latent variable. The reparameterization layer performs reparameterization sampling on the mean and variance output by the encoder to generate the latent variable, which enables the variational auto-encoder model to perform back propagation during training. The decoder receives the latent variable and converts it back to the distribution of the original data. The decoder is also composed of a neural network, and the purpose is to reconstruct the input data.

[0042] Support vector regression (SVR) model: a regression method based on support vector machine (SVM), the core idea of which is to find an optimal hyperplane that makes the distance (i.e. error) of all data points to the hyperplane as small as possible, while allowing a certain error range (i.e. ε-insensitive band).

[0043] Non-dominated Sorting Genetic Algorithm II (NSGA-II): A highly efficient multi-objective optimization algorithm, mainly solving the optimization problem of mutual conflict between multiple objectives.

[0044] With the popularity of vehicles, cars have become one of the main means of transportation for people's daily travel. During the use of the vehicle, the air quality in the vehicle will affect the driving experience of the driver and passenger.

[0045] At present, the air quality in the vehicle is mainly monitored in real time by installing an air quality sensor in the vehicle, such as monitoring fine particulate matter PM2.5, and comparing the monitored air quality index in the vehicle with the preset concentration threshold to switch the purification strategy of the air purification system in the vehicle.

[0046] However, the above-mentioned air purification strategy based on the concentration threshold lacks comprehensive perception and optimization capability for user subjective experience, filter state and energy consumption of the air purification system, which leads to the inability to dynamically adjust the strategy according to the perception difference of different users on air freshness, and often results in insufficient comfort or resource waste. The root cause of this problem is that the current scheme does not introduce a modeling mechanism between the environmental air state and the user perception, and does not model and control the nonlinear relationship between the filter life and the purification efficiency. When adjusting the purification strategy of the air purification system, it lacks intelligent trade-off and multi-objective optimization mechanism, and it is difficult to effectively control the air quality in the vehicle under different demands, thereby affecting the driving experience of the user.

[0047] Based on the above problems, the embodiment of the present application provides an air purification method, which comprises the following steps: obtaining environmental air state data of a vehicle, filter state data of an air purification system in the vehicle, and running energy consumption data of the air purification system; determining a comfort score value based on the environmental air state data, the comfort score value being used to reflect the satisfaction degree of a user on the air quality in the vehicle; determining a filter performance parameter based on the filter state data, the filter performance parameter being used to reflect the purification ability, pressure drop characteristic and service life of the purification filter; determining a target purification strategy based on the comfort score value, the filter performance parameter and the running energy consumption data; and controlling the air purification system to work by using the target purification strategy. Therefore, the embodiment of the present application can determine the target purification strategy based on the comfort score value, the filter performance parameter and the running energy consumption data, and realize individualized intelligent balance control between the air comfort perceived by the user, the energy consumption of the air purification system and the filter performance, which can effectively control the air quality in the vehicle under different demands such as air comfort priority, energy consumption priority or filter life loss priority, so as to realize individualized, dynamic and efficient control of the air quality in the vehicle, thereby improving the driving experience of the user.

[0048] In order to better understand the embodiments of the present application, the application scenarios of the air purification method in the embodiments of the present application are schematically described below.

[0049] Schematically, Figure 1 is a schematic diagram of an application scenario of the air purification method provided by the embodiments of the present application. As shown in Figure 1 , the application scenario includes a vehicle 100 and a cloud 200.

[0050] Among them, the vehicle 100 and the cloud 200 are connected through a network, and the vehicle 100 can interact with the cloud 200 through the network to receive or send messages, etc. The network provides a communication link medium between the vehicle 100 and the cloud 200. The network can include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0051] In some embodiments, the models involved in the air purification method provided by the embodiments of the present application, such as the comfort representation model, the score prediction model and the multi-objective optimization model, etc. can be trained and updated in the cloud 200.

[0052] After the training or updating of the models involved in the air purification method provided by the embodiments of the present application is completed in the cloud 200, these models can be deployed on the vehicle 100, so that the vehicle 100 can execute the air purification method provided by the embodiments of the present application based on these models.

[0053] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems are described in detail below with specific embodiments.

[0054] Figure 2 is a flowchart of an air purification method provided by the embodiments of the present application. The air purification method can be applied to a vehicle, such as Figure 2 , the air purification method can specifically include the following steps: S201, obtaining environmental air state data of the vehicle, filter element state data of the air purification system in the vehicle and running energy consumption data of the air purification system.

[0055] Among them, the environmental air state data is used to reflect the environmental air condition of the inside and outside of the vehicle, the filter element state data is used to reflect the performance condition of the purification filter element of the air purification system in the vehicle, and the running energy consumption data is used to reflect the power consumption level of the air purification system in the vehicle.

[0056] Specifically, a plurality of groups of air quality sensors are deployed in the vehicle, each group of air quality sensors collecting environmental air conditions inside and outside the vehicle in real time. The types of air quality sensors include, but are not limited to, PM2.5 sensors, PM10 sensors, total volatile organic compound (TVOC) sensors, CO2 sensors, temperature sensors, humidity sensors, wind speed sensors, odor sensors, filter pressure difference sensors, and air duct state sensors. Therefore, the environmental air conditions inside and outside the vehicle can include PM2.5 concentration, PM10 concentration, TVOC concentration, carbon dioxide concentration, temperature, humidity, wind speed, gas concentration of odor, pressure drop value of the purification filter in the air purification system, and air duct state in the air purification system, etc.

[0057] The collection frequency of each group of air quality sensors can be consistent, such as the collection frequency of each group of air quality sensors can be uniformly set to 1 Hz (Hertz), to ensure that rapidly changing environmental air condition data can be captured, facilitating subsequent modeling analysis of micro-changed environmental air condition data.

[0058] In some embodiments, the air purification system (also referred to as an air filtration system) refers to a system for adjusting the air conditions inside the vehicle, and the air purification system includes air conditioners, air purifiers, and other mechanisms that can change the air conditions inside the vehicle. In the case where the air purification system is in an activated state, the air purification system can purify the air inside the vehicle and / or provide fresh air to the space inside the vehicle.

[0059] The air purification system can include a purification filter, and the filter state data of the purification filter can include the difference in particulate matter concentration (such as the difference in PM2.5 concentration, the difference in PM10 concentration, etc.) between inside and outside the vehicle, the air volume of the air purification system, the air speed of the air purification system, the frequency and intensity corresponding to the purification mode of the air purification system, the pressure drop value of the purification filter, the fan speed of the air purification system, the air duct structure in the air purification system, the cumulative use time of the purification filter, the amount of particulate matter filtered by the purification filter in the past, and the historical cleaning record or historical replacement record of the purification filter. The pressure drop value of the purification filter represents the pressure difference between the two sides of the purification filter.

[0060] The operation energy consumption data of the air purification system includes the energy consumption corresponding to the fan gear of the air purification system, the motor operating state, and the battery load, etc.

[0061] S202, determining a comfort score value based on the environmental air condition data, the comfort score value reflecting the user's satisfaction with the air quality inside the vehicle.

[0062] After obtaining the ambient air state data, the ambient air state data can be preprocessed. The preprocessing mainly includes elimination of abnormal data in the ambient air state data, filling of missing data, time window segmentation, normalization processing, etc.

[0063] For example, the data in the ambient air state data exceeds the upper limit of the data collected by the sensor, or the data in the ambient air state data is a fixed value for a long time, so that the data is identified as abnormal data, and the abnormal data in the ambient air state data is eliminated. When some data in the ambient air state data is missing, the missing data can be completed by using a forward filling or linear interpolation method. The normalization processing is based on the vehicle enterprise standard environment reference interval, to ensure that the numerical ranges of each channel dimension input into the comfort representation model are consistent.

[0064] Therefore, the preprocessed ambient air state data can be composed of multiple channels, including PM2.5 concentration, PM10 concentration, TVOC concentration, carbon dioxide concentration, temperature, humidity, wind speed, gas concentration of existing odor, pressure drop value of the purification filter element, and air duct state, etc. Ten-dimensional data. Each dimension includes sequence data of 300 consecutive time points, which represents the evolution trajectory of the ambient air state data sampled every second within the previous 5 minutes before determining the comfort score value.

[0065] Generally, the user (i.e. the driver or passenger) feels the air quality in the vehicle not only affected by objective parameters such as particulate matter, odor, temperature, humidity, etc., but also interfered by subjective factors such as individual sensitivity, driving habits, and tolerance to odor. The preprocessed ambient air state data can be used as input features of the comfort representation model to describe the influence process of air changes on user's feelings during the trip.

[0066] Therefore, in one possible implementation, the above S202 "determining a comfort score value based on the ambient air state data" can specifically include the following steps: processing the ambient air state data using a comfort representation model to obtain a feature latent variable vector corresponding to the ambient air state data; processing the feature latent variable vector using a score prediction model to obtain the comfort score value.

[0067] Further, the above "processing the ambient air state data using a comfort representation model to obtain a feature latent variable vector corresponding to the ambient air state data" can specifically include the following steps: processing the ambient air state data using an encoder in the comfort representation model to obtain a mean vector corresponding to the ambient air state data and a standard deviation vector corresponding to the ambient air state data; reparameterizing the mean vector and the standard deviation vector using a reparameterization layer in the comfort representation model to obtain the feature latent variable vector.

[0068] In one example, the comfort representation model can be a variational autoencoder model. In the variational autoencoder model, the encoder adopts a stacked convolutional structure combined with a fully connected layer to gradually compress the time series information. In the first stage, a one-dimensional convolutional layer is used to extract the local time-dependent features of the preprocessed environmental air state data, and the convolution kernel size and step length are designed as learnable parameters to adapt to different signal frequency characteristics. Then, a pooling layer is introduced for time dimension downsampling to retain key trends. The second stage is a fully connected neural network layer, which takes the downsampled feature matrix as input and outputs two independent vectors representing the mean vector and standard deviation vector of the latent variable, respectively, to construct a multi-dimensional Gaussian distribution. Then, the reparameterization layer in the variational autoencoder model reparameterizes the mean vector and standard deviation vector to generate a feature latent variable vector. The feature latent variable vector is a low-dimensional vector, and its dimension is usually set between 32 and 64, such as 32 or 64. The dimension of the feature latent variable vector can be flexibly adjusted according to the tuning process of the comfort representation model.

[0069] It should be noted that after the trained comfort representation model is used to process the environmental air state data, a set of stably distributed feature latent variable vectors can be obtained, which can be used as the encoded representation of the original environmental air state data, and implicitly capture the change patterns in the environmental air state data that affect the user's subjective perception. These feature latent variable vectors are referred to as comfort latent factors, which will be used as the main input of the score prediction model to establish the nonlinear relationship between the air state and the user's score. Moreover, the feature latent variable vector can accommodate the sensitivity differences of different users and does not directly depend on a single sensor value, so that the model can still maintain its discriminant ability in cases where the air state is similar but the user's feedback is significantly different.

[0070] In some implementations, each user can have their own corresponding encoder weight parameters, i.e., the weight parameters included in the encoder of the comfort representation model corresponding to different users can be different. In the initial stage, the comfort representation model is uniformly pre-trained by relevant personnel of the vehicle manufacturer and loaded with general parameters. In the subsequent use process, environmental air state data and actual user score data can be continuously collected to fine-tune the encoder weight parameters in the comfort representation model.

[0071] Generally, the comfort representation model can be deployed in the vehicle's on-board system chip, with an inference delay of less than 300 milliseconds, meeting the real-time requirements of vehicle-mounted systems. The feature latent variable vector obtained by processing the environmental air state data using the comfort representation model can be saved in the local cache of the vehicle for a certain period of time (such as 24 hours), avoiding frequent repeated calculation of the feature latent variable vector, thereby reducing the computational power consumption.

[0072] Therefore, the comfort representation model provides a low-dimensional compressed expression of perception data for the entire intelligent adjustment system, laying a foundation for implementing personalized air strategy control.

[0073] Due to the nonlinear mapping relationship between the objective parameters of air quality and the user subjective score, limited by individual differences of users and fluctuations of perception thresholds, a traditional linear model cannot accurately fit the variation trend of such subjective experience data.

[0074] Therefore, in the embodiments of the present application, a support vector regression model can be used to construct a score prediction model, so that the score prediction model can learn the nonlinear mapping relationship between the feature hidden variable vector output by the comfort representation model and the user subjective score, and realize the approximate fitting of the user score. The score prediction model can be used as a score predictor of the subsequent air purification strategy, and the comfort score value output by the score prediction model will participate in the multi-objective optimization control.

[0075] It should be noted that, since the feature hidden variable vector already contains the compressed air parameter variation features in the vehicle and the signal expression associated with the user comfort, since the feature hidden variable vector input to the score prediction model is a stable structured numerical vector, it does not need to be normalized or feature engineered again, and can be directly input to the score prediction model.

[0076] In one example, the score prediction model can be a support vector regression model, which can select a radial basis kernel function as a nonlinear mapping kernel function, suitable for a feature space with moderate dimensions but complex boundary distribution.

[0077] In addition, in order to improve the stability and generalization ability of the score prediction model in actual application, a minimum score sample number threshold can be set, and only when the cumulative number of valid score samples of the user exceeds the specified threshold, the personalized score prediction model is enabled. For users with insufficient samples, a unified pre-trained model or a regional average model is used to provide score prediction, ensuring that there is a basic prediction ability in the early stage. The training process is regularly performed in the private cloud of the vehicle manufacturer, and the model is updated after a certain batch of user scores are uploaded each time. The model weight version is managed in the background and supports over-the-air (OTA) distribution to vehicles.

[0078] The output of the score prediction model is a comfort score value, which can reach one decimal place in accuracy, and is used to represent the user's perceived comfort of the vehicle under the current air state. The comfort score value is not directly displayed to the user, but is used as one of the main targets of the multi-objective optimization model to participate in the dynamic control selection of the air purification strategy.

[0079] When the comfort score value output by the score prediction model is close to 5, it represents that the user is satisfied with the current air state, and the vehicle can maintain the current air purification strategy in the energy-saving mode. When the comfort score value output by the score prediction model is close to 1, it represents that the user is not satisfied with the current air state, and the vehicle can trigger a higher intensity air purification strategy.

[0080] It should be noted that the score prediction model deployed in the vehicle is a lightweight model, and the inference process takes no more than 50 milliseconds, which can quickly respond to changes in the air state in the vehicle. The input of the score prediction model comes from the cached latest feature hidden variable vector, without the need for repeated encoding, further reducing the system load. In each evaluation period of the air purification strategy, the score prediction model reevaluates the comfort score value in the current state and outputs the comfort score value to the multi-objective optimization model as one of the optimization directions to guide the generation of the target purification strategy.

[0081] The continuous operation of the score prediction model also supports a dynamic correction mechanism for user profiling. The vehicle will periodically analyze the difference between the predicted comfort score value and the actual user feedback score. If there is a persistent deviation, the personalized retraining task of the score prediction model is automatically triggered to improve the adaptability of the score prediction model to user preferences, making the entire score prediction system flexible, real-time and scalable, providing a solid data-driven foundation for personalized air purification strategies.

[0082] S203, determine a filter performance parameter based on the filter state data, the filter performance parameter being used to reflect a purification ability, a pressure drop characteristic and a service life of the purification filter in the air purification system.

[0083] In the embodiments of the present application, after obtaining the filter state data of the air purification system in the vehicle, the filter performance parameter can be determined based on the filter state data.

[0084] The filter state data includes first state data, second state data and third state data, the first state data being related to the purification ability of the purification filter, the second state data being related to the pressure drop characteristic of the purification filter, and the third state data being related to the service life of the purification filter. Therefore, in a possible implementation manner, the above S203 of determining the filter performance parameter based on the filter state data can specifically include the following steps: processing the first state data by using a purification efficiency model to obtain a purification efficiency percentage, the purification efficiency percentage being used to reflect the purification ability of the purification filter; processing the second state data by using a pressure drop characteristic model to obtain a pressure drop level, the pressure drop level being used to reflect the pressure drop characteristic of the purification filter; and processing the third state data by using a life consumption model to obtain a filter remaining life estimate, the filter remaining life estimate being used to reflect the service life of the purification filter.

[0085] Specifically, the first state data can include the concentration difference of particulate matter inside and outside the vehicle, the air volume of the air purification system, the air speed of the air purification system, and the frequency and intensity corresponding to the purification mode of the air purification system, etc. The second state data can include the pressure drop value of the purification filter element, the fan speed of the air purification system, and the air duct structure in the air purification system, etc. The third state data can include the cumulative use time of the purification filter element, the historical filtered particulate matter amount of the purification filter element, and the historical cleaning record or historical replacement record of the purification filter element, etc.

[0086] As the core component in the air purification system, the filter element state data of the purification filter element directly affects the purification efficiency, air volume performance, and vehicle energy consumption load, etc. The purification filter element shows performance degradation characteristics in different use stages, and its filtering capacity, pressure drop coefficient, and particulate matter capture ability all dynamically change with factors such as cumulative use time, through particulate matter mass, and external environmental pollution level. In order to realize intelligent adjustment of the air purification strategy, the filter element state data needs to be accurately modeled, a quantifiable performance evaluation index (i.e. filter element performance parameter) is constructed, and used in the subsequent multi-objective optimization model as a constraint variable and performance feedback.

[0087] Among them, the model obtained after modeling the filter element state data mainly includes three dimensions, which are the purification efficiency model, the pressure drop characteristic model, and the life consumption model.

[0088] The first case, the purification efficiency model processes the first state data to obtain the purification efficiency percentage.

[0089] Specifically, the purification efficiency model compares the concentration difference of particulate matter inside and outside the vehicle, and combines data such as the air volume of the air purification system, the air speed of the air purification system, and the frequency and intensity corresponding to the purification mode of the air purification system, etc. to learn how many particulate matters (i.e. particulate matter removal efficiency) the purification filter element can filter out under different working conditions, thereby outputting the current purification efficiency percentage.

[0090] The second case, the pressure drop characteristic model processes the second state data to obtain the pressure drop level.

[0091] Specifically, the pressure drop characteristic model is based on the pressure drop value of the purification filter element, and combines the fan speed of the air purification system and the air duct structure in the air purification system, etc. to fit the relationship between the blockage degree and the airflow resistance, and then divide the resistance into several levels to generate the current pressure drop level.

[0092] The third case, the life consumption model processes the third state data to obtain the filter element remaining life estimate.

[0093] Specifically, the life consumption model is based on the cumulative use time of the purification filter element, the historical filtered particulate matter amount of the purification filter element, and the historical cleaning record or historical replacement record of the purification filter element, etc., to estimate the remaining life estimate of the filter element.

[0094] Therefore, the vehicle can integrate the real-time outputs of the purification efficiency model, the pressure drop characteristic model, and the life consumption model, i.e., the filter element performance parameters, i.e., the filter element performance parameters can include the purification efficiency percentage, the pressure drop level, and the filter element remaining life estimate.

[0095] These filter element performance parameters can be used as a decision basis and participate in the generation process of the target purification strategy. When the purification efficiency percentage starts to fluctuate or continuously decreases, the vehicle determines whether the purification filter element enters the performance recession period. When the pressure drop level exceeds the set threshold, it indicates that the air duct may be blocked and affect air circulation. The filter element remaining life estimate can also be used to determine whether the purification filter element is close to the replacement threshold. When the filter element remaining life estimate is lower than the set replacement threshold, the vehicle can issue a replacement reminder or adjust the target purification strategy to delay the use of the load.

[0096] It should be noted that, in order to improve the real-time performance and stability of the purification efficiency model, the pressure drop characteristic model, and the life consumption model, some key performance parameters can be predicted online through interpolation regression to avoid relying on offline model updates.

[0097] In addition, the purification efficiency model, the pressure drop characteristic model, and the life consumption model can use an embedded model engine to reside in the vehicle main control chip, and use a hybrid method of rules combined with machine learning for state monitoring. The model calculation complexity is controlled within the range of real-time response, ensuring continuous updating of filter element performance parameters in a dynamic driving environment, providing reliable boundary conditions and cost evaluation basis for optimization algorithms.

[0098] In addition, the purification efficiency model, the pressure drop characteristic model, and the life consumption model also support linkage with the vehicle manufacturer's background database, automatically match the optimal parameter curve template according to different vehicle models, filter element types, and use environments, and realize generalization adaptation across vehicle models and climate environments. At the same time, it supports OTA online updating of filter element performance parameter table and model version, ensuring seamless support of the system when new filter element models go online. The complete filter element state and performance modeling capability provides bottom-layer performance support for the entire air purification strategy, ensuring dynamic balance between the effect, energy consumption, and life of the target purification strategy.

[0099] S204, determining the target purification strategy based on the comfort score value, the filter element performance parameters, and the running energy consumption data.

[0100] In the embodiments of the present application, after the comfort score value, the filter performance parameter and the operation energy consumption data are obtained, the comfort score value, the filter performance parameter and the operation energy consumption data can be input into the multi-objective optimization model, so that the multi-objective optimization model outputs the target purification strategy. In one example, the multi-objective optimization model can be an NSGA-II model.

[0101] In one possible implementation, the above S204 "determining a target purification strategy based on the comfort score value, the filter performance parameter and the operation energy consumption data" can specifically include the following steps: obtaining a plurality of purification strategy combinations, each of which includes a plurality of purification strategy individuals; determining a non-dominated solution corresponding to each purification strategy combination based on each purification strategy combination, the comfort score value, the filter performance parameter and the operation energy consumption data; determining a non-dominated solution set based on the non-dominated solution corresponding to each purification strategy combination in the plurality of purification strategy combinations; and determining a target purification strategy from the non-dominated solution set, the target purification strategy being one of the non-dominated solutions in the non-dominated solution set.

[0102] Intelligent adjustment of the air purification strategy of the air purification system not only needs to consider the current air purification effect, but also needs to comprehensively evaluate the mutual relationship between the filter life loss, the electric energy consumption and the user subjective comfort. The traditional switching control strategy based on a single target or a fixed concentration threshold cannot simultaneously consider the user experience and the system cost. Therefore, a multi-objective optimization framework is introduced, a non-dominated sorting genetic algorithm is used as a core scheduler to perform global search and dynamic adjustment on the air purification strategy, and an optimal control scheme is output to ensure that the overall performance is optimal under multi-dimensional performance indicators.

[0103] The input of the multi-objective optimization model is composed of three parts: first, the comfort score value output from the score prediction model, which is used to reflect the user's satisfaction with the current air quality in the vehicle; second, the purification efficiency percentage, the pressure drop level and the filter remaining life estimate output from the purification efficiency model, the pressure drop characteristic model and the life consumption model, which are used to reflect the use cost and physical state of the air purification system; and third, the operation energy consumption data, including the energy consumption corresponding to the fan gear in the air purification system, the motor working state and the battery load, which are used to estimate the electric energy consumption level under different air purification strategies.

[0104] The optimization variable is designed as a combination of adjustable parameters in the control system, including air circulation modes (external circulation, internal circulation, high-efficiency filtration, etc.), fan air volume levels (including low, medium, and high levels), switching time intervals of air purification strategies, and duration of purification modes, etc. The combination of these adjustable parameters constitutes a purification strategy individual, which is a chromosome in the NSGA-II population and participates in the global search process. Each purification strategy individual represents a set of air purification strategies, and the execution effect of each purification strategy individual under the current air state is simulated through a multi-objective optimization model to evaluate its performance in the objective function space.

[0105] In some embodiments, the plurality of purification strategy combinations includes an initial purification strategy combination and at least one target purification strategy combination; wherein each target purification strategy combination is obtained by performing a crossover operation and / or a mutation operation on the initial purification strategy combination, the crossover operation is used to exchange part of the control parameters in any two purification strategy individuals in the initial purification strategy combination, and the mutation operation is used to change part of the control parameters in at least one purification strategy individual in the initial purification strategy combination.

[0106] Moreover, the "determining a non-dominated solution corresponding to each purification strategy combination based on each purification strategy combination, the comfort score value, the filter performance parameter, and the operation energy consumption data" can specifically include the following steps: determining a first feedback value corresponding to each purification strategy individual based on each purification strategy individual in the purification strategy combination and the comfort score value; determining a second feedback value corresponding to each purification strategy individual based on each purification strategy individual in the purification strategy combination and the filter performance parameter; determining a third feedback value corresponding to each purification strategy individual based on each purification strategy individual in the purification strategy combination and the operation energy consumption data; and performing non-dominated sorting on each purification strategy individual in the purification strategy combination based on the first feedback value, the second feedback value, and the third feedback value to obtain the non-dominated solution corresponding to the purification strategy combination.

[0107] In addition, the "determining a target purification strategy from the non-dominated solution set" can specifically include the following steps: determining a target purification strategy from the non-dominated solution set based on a user preference configuration or a current working mode of the air purification system.

[0108] The optimization process of the multi-objective optimization model can be divided into six stages: initialization, evaluation, selection, crossover, mutation, and elite reservation.

[0109] In the initialization stage, a set of purification strategy individuals are randomly generated to form a population (i.e., an initial purification strategy combination). Then, each purification strategy individual in the initial purification strategy combination is evaluated by three objective functions, which are maximizing user comfort score, minimizing filter life consumption rate, and minimizing electrical energy consumption level, respectively. The vehicle simulates the purification effect, life consumption, and energy consumption data of the purification strategy individual after the execution of the purification strategy individual according to the current air state, to obtain numerical feedback of the three objectives. That is, when each purification strategy individual is evaluated, the comfort score value, filter performance parameters, and running energy consumption data are brought in, respectively, so as to calculate the first feedback value corresponding to each purification strategy individual, the second feedback value corresponding to each purification strategy individual, and the third feedback value corresponding to each purification strategy individual.

[0110] Among them, the comfort score value is used to calculate the "comfort target", and the numerical feedback corresponding to the "comfort target" is the first feedback value. The higher the comfort score value, the better the comfort. The purification efficiency percentage, pressure drop level, and filter remaining life estimate are collectively used to calculate the "filter life consumption target", and the numerical feedback corresponding to the "filter life consumption target" is the second feedback value. The higher the pressure drop level, the more clogged the purification filter, and this purification strategy individual will make the filter more tired and the life drop faster, so the pressure drop level is an important input of the filter life consumption target. The running energy consumption data is used to calculate the "energy consumption target", and the numerical feedback corresponding to the "energy consumption target" is the third feedback value. The larger the fan air volume, the higher the energy consumption.

[0111] After that, the selection stage and the elite reservation stage are entered in turn. Specifically, based on the first feedback value corresponding to each purification strategy individual, the second feedback value corresponding to each purification strategy individual, and the third feedback value corresponding to each purification strategy individual, each purification strategy individual in the initial purification strategy combination is non-dominantly sorted to filter out the optimal solution that is not suppressed by other purification strategy individuals in all target dimensions, to obtain the non-dominant solution (i.e., the pareto frontier) corresponding to the initial purification strategy combination, and to ensure that the optimal solution will not be discarded.

[0112] Then, the crossover and mutation stage is entered, and the crossover operation and the mutation operation are performed on the dimensions of the purification strategy individual, to ensure the diversity of the population and expand the search space during the search process. Among them, the crossover operation is used to randomly select two purification strategy individuals in the initial purification strategy combination, exchange part of the control parameters of the two purification strategy individuals, to generate a new purification strategy individual; the mutation operation is used to disturb (i.e., change) part of the control parameters in at least one purification strategy individual in the initial purification strategy combination, such as changing the fan air volume level in a certain purification strategy individual from medium to high. In this way, after each round of crossover operation and / or mutation operation on the initial purification strategy combination, a target purification strategy combination can be obtained.

[0113] After each iteration (i.e. each time the crossover operation and / or mutation operation is performed on the initial purification strategy combination), a non-dominated solution corresponding to each target purification strategy combination is determined based on the comfort score value, the filter performance parameter and the operating energy consumption data of each target purification strategy combination. That is, the objective function is evaluated again for each purification strategy individual in the target purification strategy combination to obtain the first feedback value, the second feedback value and the third feedback value corresponding to each purification strategy individual in the target purification strategy combination, and the non-dominated sorting is performed on each purification strategy individual in the target purification strategy combination to screen the non-dominated solution corresponding to the target purification strategy combination.

[0114] The above iteration process can continue for multiple iteration cycles until convergence or reaching the computing resource limit of the vehicle, so as to finally output a non-dominated solution set, i.e. the non-dominated solution set is a set of non-dominated solutions corresponding to each purification strategy combination in the multiple purification strategy combinations. Each non-dominated solution in the non-dominated solution set forms a different trade-off between comfort, energy consumption and service life.

[0115] Finally, a most suitable non-dominated solution in the non-dominated solution set can be selected as the target purification strategy according to the user preference configuration or the current working mode of the air purification system, and the target purification strategy is the execution scheme of the current air purification strategy.

[0116] It should be noted that the user preference configuration refers to the air purification strategy preference selected by the user in the vehicle machine, which can include "comfort first", "energy saving first" or "filter protection", etc. The current working mode of the air purification system can be automatically identified by the vehicle, which can include "normal mode", "high pollution emergency mode", "low power energy saving mode" and "filter degradation protection mode", etc., for dynamically changing the optimization constraints according to the power of the battery in the vehicle, the pollution degree of the external environment where the vehicle is located, the filter state of the purification filter, etc., to ensure safety and stability.

[0117] It should be understood that the selection mechanism of the target purification strategy supports different priority settings, for example, prioritizing comfort when the user is allergic or children are riding, and prioritizing energy consumption control when the power of the vehicle is limited.

[0118] It should also be understood that the high pollution emergency mode can be determined according to the pollution degree of the external environment of the location where the vehicle is located. For example, when the pollution degree of the external environment of the location where the vehicle is located is high pollution, the high pollution emergency mode is triggered. Specifically, the location where the vehicle is located can be determined based on a global positioning system (GPS) module and a map application programming interface (API) interface in the vehicle, and the pollution degree of the external environment of the location where the vehicle is located can be determined based on the location where the vehicle is located and a local city pollution heat map information source. In addition, whether to trigger the low battery energy saving mode can also be determined according to the power of the battery in the vehicle. For example, when the power of the battery in the vehicle is lower than a preset power, the low battery energy saving mode is triggered.

[0119] Since the target purification strategy is actually a non-dominated solution, and the non-dominated solution is a purification strategy individual that is not suppressed in all target dimensions by other purification strategy individuals, the target purification strategy includes at least one of the following: air circulation mode, fan air volume level, switching time interval of air purification strategy, and duration of purification mode.

[0120] It should be noted that the switching time interval refers to how long the current target purification strategy is executed before allowing reevaluation and switching of the target purification strategy, thereby avoiding frequent switching of the target purification strategy causing air volume mutation and discomfort. The duration of the purification mode refers to the execution duration that a specific purification mode (such as the high-efficiency filtration mode) plans to maintain, which is an internal parameter in the target purification strategy and determines the duration of the purification operation itself.

[0121] In one example, the multi-objective optimization model can be deployed in a vehicle-mounted edge computing platform, and the combination of offline training and online inference is used to improve real-time performance. The standard operation cycle is to evaluate the target purification strategy once every five minutes, and the inference time does not exceed 500 milliseconds, ensuring timely control response. In order to reduce the demand for computing power at the vehicle end, the population size and iteration rounds in the multi-objective optimization model are adjusted to remain within a controllable range of resources. In addition, strategy space optimization and parameter updating of the multi-objective optimization model through the cloud are also supported, ensuring the evolvability of the algorithm during long-term operation.

[0122] The multi-objective optimization model establishes a complete air purification strategy evaluation and selection mechanism by fusing comfort score values, filter performance parameters, and operating energy consumption data. As the core algorithm engine, the multi-objective optimization model not only improves the decision-making intelligence of the air purification strategy, but also brings personalized, high-efficiency, and sustainable operation optimization capabilities to the vehicle-mounted system, becoming a key control link in the air purification system.

[0123] S205, control the air purification system to work according to the target purification strategy.

[0124] After determining the target purification strategy corresponding to the air purification system, the air purification system can be controlled to work according to the target purification strategy to purify the air in the vehicle.

[0125] Therefore, the embodiments of the present application can determine the target purification strategy based on the comfort score value, the filter performance parameter and the operation energy consumption data, and realize the personalized intelligent balance control between the air comfort perceived by the user, the energy consumption of the air purification system and the filter performance, which can effectively control the air quality in the vehicle under different requirements such as air comfort priority, energy consumption priority or filter life loss priority, to realize personalized, dynamic and efficient control of the air quality in the vehicle, thereby improving the driving experience of the user.

[0126] It should be noted that the intelligent adjustment function of the air purification system depends on the cooperative operation of multiple modules, including the comfort representation model, the score prediction model and the multi-objective optimization model. In order to realize end-to-end strategy closed-loop control, an efficient and stable fusion module needs to be constructed to effectively interface the outputs of the above models in the system architecture, and to complete the purification mode switching and the issuance and execution of the air volume control command in the vehicle control logic in the lowest delay manner. The fusion module as the strategy logic center is responsible for completing five core functions of data reception, model reasoning, result fusion, strategy analysis and execution scheduling.

[0127] The input of the fusion module includes the output results of the comfort representation model, the score prediction model and the multi-objective optimization model. The first part is the feature hidden variable vector generated by the comfort representation model, which represents the mapping expression of the current air state in the user's subjective space, and is used to update the system perception layer state. The second part is the comfort score value output by the score prediction model, which is used as the quantitative evaluation of the user's current air experience. The third part is the purification strategy combination in the processing process of the multi-objective optimization model. Each purification strategy individual in the purification strategy combination corresponds to a set of air purification control parameter combination, including air circulation mode, fan air volume level, switching time interval of air purification strategy, duration of purification mode and other key control variables.

[0128] The fusion module can maintain the environmental air state data and model inference results of the last cycle in the internal buffer area, ensuring that the current target purification strategy is generated based on the latest perception. Whenever the environmental air state data changes beyond a threshold, or every time the scheduling period of the target purification strategy is reached, the system immediately starts a complete target purification strategy update process. After the process starts, the comfort representation model reads the latest environmental air state data to update the feature hidden variable vector, and the updated feature hidden variable vector is passed into the score prediction model to calculate the comfort score value. Subsequently, the multi-objective optimization model is called to search for a new optimal non-dominated solution set based on the current comfort score value as the first target and the filter life and energy consumption state as the remaining targets, to obtain the target purification strategy.

[0129] The strategy selection mechanism is controlled by the weight allocation logic built into the fusion module. By default, the system sorts the non-dominated solutions according to the fixed priorities of comfort, life, and energy consumption, and selects the purification strategy individual with the optimal overall score as the current target purification strategy. Users can select preference modes such as "comfort first," "energy saving first," or "filter protection" through the vehicle machine setting interface to affect the weight configuration of each objective function in the multi-objective optimization model, thereby reflecting individual differences in strategy selection.

[0130] The selected target purification strategy will be parsed into executable vehicle control instructions, including switching the air purification system (such as the air conditioning system) to internal or external circulation, whether the high-efficiency filtration module is turned on, the fan air volume level, the switching time interval of the air purification strategy, and the duration of the purification mode, etc. The fusion module is responsible for converting these control variables into a data frame format recognizable by the vehicle control bus, and sending them to the controller in the air purification system through the controller area network (CAN) or local interconnect network (LIN) communication interface for execution. In addition, the strategy command carrying the data frame format can be accompanied by a unique identity document (ID) and a verification code to ensure that the controller in the air purification system can correctly identify and feedback the execution status, preventing instruction loss or repeated issuance.

[0131] The fusion module also supports strategy transition buffer design to prevent wind volume sudden changes or driving experience interference caused by frequent target purification strategy changes. For example, before switching to the high-efficiency filtration mode, the system automatically reserves a wind speed buffer interval, and adjusts the air volume output in a gradient manner within the control period, improving user comfort and reducing fan noise perception. At the same time, the system sets a cooling time window for the target purification strategy to avoid frequent switching of the target purification strategy due to short-term pollution level fluctuations, improving control stability.

[0132] The fusion module is built-in with abnormal detection and safety fallback mechanism. If a model occurs calculation abnormality or output exceeds the reasonable interval during inference, the system automatically calls the default rule control logic, based on the preset PM concentration threshold, etc. to perform simplified control, ensuring that the air purification ability is not affected by the model state. In addition, the system records the target purification strategy execution results and user feedback comfort scores of each round, which are used for subsequent model fine-tuning and strategy evolution, forming a complete data closed loop.

[0133] The whole fusion module is deployed in the vehicle-mounted main control unit, supporting linkage with the vehicle air conditioning system, central gateway, data recording system, etc. All control records, execution results, feedback states can be uploaded to the vehicle data platform in real time for remote monitoring, strategy upgrade and model optimization. The overall design of the module meets the real-time and reliability requirements of vehicle regulations, with redundancy mechanism, running state monitoring and hot switching capability, ensuring long-term stable operation under multiple working conditions.

[0134] In addition, different users have significant differences in the perception of indoor air quality. Some users are extremely sensitive to odors, while some users are slow to react to changes in particulate matter concentration. In order to achieve long-term adaptability and individualized strategy matching of the air purification system in a multi-user environment, a complete personalized adaptive adjustment mechanism needs to be built, so that the system can dynamically optimize the model structure, parameter configuration and control strategy according to the actual feedback of each user, and continuously ensure the balance between subjective satisfaction and system performance of air purification under complex conditions such as multi-user shared vehicles, different regional climates, different filter usage states, etc.

[0135] The adaptive adjustment mechanism is based on user identity recognition as the entrance. The system realizes automatic user switching through in-vehicle face recognition, Bluetooth key ID or vehicle App binding user account. After recognition, the system loads the personalized model parameter file corresponding to the user, including the encoder weight parameters in the exclusive comfort representation model, the exclusive score prediction model, the exclusive multi-objective optimization model preference weight configuration and control parameter cache. These parameters are kept locally. If the user is using the system for the first time, the vehicle pre-trained default model is loaded as the initialization version, and customized updates are gradually made during use.

[0136] In use, the comfort score data fed back by each user will be collected separately and bound with the air quality in the vehicle to build a personal data set. The system sets a scoring trigger mechanism, which will mark the current environmental air state data as a sample and store it when the user actively scores or the system detects behavioral deviation (such as an increase in the number of manual mode switching). As the scoring samples accumulate, the system triggers a local fine-tuning mechanism to perform lightweight incremental training on the encoder in the comfort representation model and the scoring prediction model, making them more in line with the user's sensitivity characteristics and scoring habits for air changes.

[0137] To prevent model overfitting or long-term deviation from system stability, the system sets a model fine-tuning constraint mechanism, including a minimum sample size threshold, error change monitoring, and a training frequency upper limit. After each fine-tuning, the system evaluates the prediction error of the new model on historical samples. If the error is reduced and the generalization ability is improved, the original model is replaced and the version number is recorded. If the model performs worse than the previous version, it is automatically rolled back to ensure that the model remains in the optimal state during system operation.

[0138] The individualized performance of the optimization strategy is mainly reflected in the dynamic adjustment of the target function weights of the multi-objective optimization model. The system records the historical trajectory of the scoring performance of each user after the execution of different target purification strategies, analyzes the correlation between the comfort level change trend and the target purification strategy configuration, and automatically adjusts the weight proportion of the target function. For example, when the user's score is always high under the high purification efficiency strategy, the system prioritizes increasing the weight of the comfort level corresponding target function, and is more inclined to select the non-dominated solution that maximizes the user's comfort level score. Or, when the user is more sensitive to energy saving, the priority of comfort level is gradually reduced, and energy consumption and life control are emphasized.

[0139] The system also supports an environment adaptive mechanism, which automatically switches to the model of the corresponding region when there are significant differences in climate, pollution type and frequency in the region where the vehicle is located. For example, the TVOC sensitive factors in high-humidity coastal cities and dry northern regions may be different, and the system automatically matches the most suitable feature extraction template to improve the fitting degree of the score and the response accuracy of the strategy. The environment adaptation strategy can also run in parallel with the user adaptive model, forming a double adaptation at the control layer.

[0140] The life cycle management of the individualization mechanism is maintained by the local model management module, and each user has an independent model configuration file, supporting OTA remote update and backup. Model updates can be performed centrally by the cloud, generating new version models by aggregating similar user data and distributing them to vehicles. The system provides a version backtracking mechanism, and users can restore the historical model state. When a vehicle is replaced or a user transfers an account, the model can be migrated at the account level, ensuring that the long-term individualized experience remains available.

[0141] The whole adaptive mechanism does not require active intervention of the user, and all personalized modeling and strategy adjustment processes are automatically executed in the background, which maximally reduces the operation burden, and realizes fine experience regulation and system resource optimization. This mechanism enables the air purification system to have long-term evolution ability, which can not only cope with complex environmental changes, but also accurately respond to the air quality demand of each user, and build a people-centered active intelligent control closed loop.

[0142] It should be noted that the comfort degree representation model, the score prediction model, the purification efficiency model, the pressure drop characteristic model, the life consumption model and the multi-objective optimization model introduced in the embodiments of the present application are all pre-trained models, which need to be trained before actual application.

[0143] The intelligent adjustment strategy of the air purification system depends on the accurate matching of high-quality environmental perception data and user subjective feedback comfort score data. The comfort score data comes from the user active scoring mechanism. At the end of each vehicle use, the air quality experience feedback interface will pop up on the vehicle-mounted central control screen or the vehicle enterprise mobile terminal App interface, and the user can score the air freshness of this trip, with a score range of 1 to 5. To ensure the subjective consistency of the comfort score data, the comfort score data is suggested to be prompted to combine the multi-dimensional subjective perception of odor perception, stuffy feeling, irritability, etc. for comprehensive scoring. The system automatically marks the environmental air state data within the last 5 minutes before the scoring time point as the environmental state interval of the current scoring sample, and establishes the label relationship between the comfort score and the objective environmental air state data. The comfort score data is uploaded to the vehicle enterprise private cloud through the vehicle-mounted telematics box (T-Box), and is used to gather the labeled data set. If the user does not actively score, the system will default that the trip has no label, and only participates in the unsupervised pre-training stage, and does not participate in the supervised training of the regression model.

[0144] To expand the scale of available data, the system sets a scoring incentive strategy to improve the user's willingness to participate. At the same time, in a multi-user driven vehicle, the system distinguishes different user scores through face recognition or App binding user ID mechanism, to ensure the label accuracy of personalized model construction. Each data record content includes timestamp, environmental air state data, current vehicle working condition, filter element state data, comfort score data, user ID, etc. The record structure is uniformly stored in the edge computing node, and is regularly uploaded to the main data platform in the cloud for unified preprocessing, model training and personalized model version management.

[0145] After data labeling is completed, a pretreatment process is entered, mainly including outlier elimination, missing value filling, time window segmentation, normalization processing, etc., and finally a sample set is formed with the user's comfort score data as the label and the vehicle's ambient air state data as the feature, for subsequent model feature extraction and regression modeling. Among them, the comfort score data is the subjective score of the user on the vehicle's ambient air state data in the corresponding time period, the score value is a real number, the value range is between 1 and 5, and a decimal is allowed, which reflects the user's fine feedback in different environments.

[0146] Therefore, the comfort representation model can be trained based on the ambient air state data of each training sample in the sample set. Among them, the comfort representation model to be trained can include an encoder, a reparameterization layer and a decoder, and part of the structure of the decoder is symmetrical to the encoder. First, the encoder in the comfort representation model to be trained can process the ambient air state data of the training sample to obtain the mean vector of the training sample and the standard deviation vector of the training sample; then, the reparameterization layer in the comfort representation model to be trained is used to reparameterize the mean vector and the standard deviation vector to obtain the feature latent variable vector of the training sample; then, the decoder in the comfort representation model to be trained is used to input the feature latent variable vector into a fully connected layer first, so as to restore the feature latent variable vector to an intermediate feature matrix, and then the intermediate feature matrix is gradually reconstructed into the original multi-channel time series data through the deconvolution layer in the decoder.

[0147] In the training process of the comfort representation model, the reconstruction error between the ambient air state data of the training sample and the reconstructed multi-channel time series data is minimized, and the relative entropy (kullback leibler, KL) divergence constraint is introduced to keep the stability of the feature latent variable vector to prevent overfitting. The loss function is composed of a reconstruction error term and a regularization term, and the system sets the training number of rounds, the learning rate, the batch size and other hyperparameters, adjusts according to the stability of the sample set, until the model converges, thereby training the comfort representation model.

[0148] Further, the score prediction model can be trained based on the feature latent variable vector output by the comfort representation model and the comfort score data of each training sample in the sample set. In the training phase of the score prediction model, the regression error between the comfort score value output by the score prediction model to be trained and the comfort score data actually evaluated by the user is minimized, and the maximum interval boundary of the score prediction model to be trained in the feature space is kept, so as to realize good fitting ability for the nonlinear score distribution. In the training process, the cross-validation method is introduced to optimize the hyperparameters of the score prediction model to be trained, including the penalty coefficient, the kernel function width and the tolerance error range, so as to prevent underfitting or overfitting problems.

[0149] For the modeling of the purification efficiency model, the pressure drop characteristic model, and the life consumption model, the required data comes from the multi-source sensors in the air purification system in the vehicle and the historical operation log. The core input parameters include the cumulative use time of the purification filter element, the historical filtered particulate matter amount of the purification filter element, the pressure drop value of the purification filter element, the fan speed of the air purification system, the particulate matter concentration difference between the inside and outside of the vehicle, the temperature and humidity environment of the purification filter element, the air volume of the air purification system, the air speed of the air purification system, the frequency and intensity corresponding to the purification mode of the air purification system, the air duct structure in the air purification system, and the historical cleaning record or historical replacement record of the purification filter element, etc. The vehicle continuously collects and records the above parameters to form a sample set, and the operation trajectory of the filter element in the life cycle is formed based on each training sample in the sample set, which is used to construct the filter element performance change curve, including the purification efficiency model, the pressure drop characteristic model, and the life consumption model.

[0150] The modeling method of the purification efficiency model, the pressure drop characteristic model, and the life consumption model adopts a multi-channel data fusion strategy, and uses multi-dimensional input to construct a regression prediction structure for each sub-target. The particulate matter concentration difference between the inside and outside of the vehicle, the air volume of the air purification system, the air speed of the air purification system, and the frequency and intensity corresponding to the purification mode of the air purification system, etc. corresponding to each training sample in the sample set are fitted by a regression neural network or a gradient boosting tree to obtain the purification efficiency model. The air volume and the pressure drop value of the purification filter element corresponding to each training sample in the sample set are used to capture the nonlinear response under different air volumes and blockage degrees by a fitting function, and the resistance change trend is judged based on the pressure drop value of the purification filter element at each sampling point to obtain the pressure drop characteristic model. The complete life cycle data of the filter element (including the cumulative use time of the purification filter element, the historical filtered particulate matter amount, and the historical cleaning record or historical replacement record, etc.) is processed in a time sequence manner to construct a dynamic decreasing function to obtain the life consumption model for real-time calculation of the remaining life estimation value of the filter element.

[0151] The purification efficiency model, the pressure drop characteristic model, and the life consumption model are all trained and calibrated by the data of the vehicle test vehicle of the vehicle manufacturer to ensure deployability and high adaptability.

[0152] In addition, based on the comfort score value output by the score prediction model, the filter element performance parameters output by the purification efficiency model, the pressure drop characteristic model, and the life consumption model, and in combination with the operation energy consumption data corresponding to each training sample in the sample set and the calibrated target purification strategy, a multi-objective optimization model can be trained.

[0153] It is worth noting that after each model is deployed to the vehicle end and the whole vehicle is put into operation, the collection and uploading of real vehicle data can be realized based on OTA remote communication, and the cloud end can also fine-tune each model based on OTA real vehicle data.

[0154] The above is described in combination with Figure 2 The air purification method provided by the embodiments of the present application is described, and the device for executing the method is described below.

[0155] Figure 3 is a structural schematic diagram of an air purification device provided by the embodiments of the present application, which can be applied to a vehicle. As Figure 3 indicated, the air purification device 300 can include a data acquisition module 301, a first determination module 302, a second determination module 303, a third determination module 304, and a control module 305.

[0156] The data acquisition module 301 is configured to acquire environmental air state data of the vehicle, filter core state data of an air purification system in the vehicle, and operation energy consumption data of the air purification system. The first determination module 302 is configured to determine a comfort score value based on the environmental air state data, the comfort score value being used to reflect a satisfaction degree of a user to an in-vehicle air quality of the vehicle. The second determination module 303 is configured to determine a filter core performance parameter based on the filter core state data, the filter core performance parameter being used to reflect a purification ability, a pressure drop characteristic, and a service life of a purification filter core in the air purification system. The third determination module 304 is configured to determine a target purification strategy based on the comfort score value, the filter core performance parameter, and the operation energy consumption data. The control module 305 is configured to control the air purification system to work by using the target purification strategy.

[0157] In a possible implementation, the first determination module 302 is specifically configured to process the environmental air state data by using a comfort representation model to obtain a feature latent variable vector corresponding to the environmental air state data, and process the feature latent variable vector by using a score prediction model to obtain the comfort score value.

[0158] In a possible implementation, the first determination module 302 is specifically configured to process the environmental air state data by using an encoder in the comfort representation model to obtain a mean vector corresponding to the environmental air state data and a standard deviation vector corresponding to the environmental air state data, and perform reparameterization sampling on the mean vector and the standard deviation vector by using a reparameterization layer in the comfort representation model to obtain the feature latent variable vector.

[0159] In a possible implementation, the filter state data includes first state data, second state data and third state data, the first state data is related to the purification capability of the purification filter, the second state data is related to the pressure drop characteristic of the purification filter, and the third state data is related to the service life of the purification filter; the second determining module 303 is specifically configured to process the first state data by using a purification efficiency model to obtain a purification efficiency percentage, the purification efficiency percentage being used to reflect the purification capability of the purification filter; process the second state data by using a pressure drop characteristic model to obtain a pressure drop level, the pressure drop level being used to reflect the pressure drop characteristic of the purification filter; and process the third state data by using a life consumption model to obtain a filter residual life estimate, the filter residual life estimate being used to reflect the service life of the purification filter.

[0160] In a possible implementation, the third determining module 304 is specifically configured to obtain a plurality of purification strategy combinations, each of which includes a plurality of purification strategy individuals; determine a non-dominated solution corresponding to each of the purification strategy combinations based on each of the purification strategy combinations, the comfort score value, the filter performance parameter and the operation energy consumption data; determine a non-dominated solution set based on the non-dominated solution corresponding to each of the plurality of purification strategy combinations; and determine the target purification strategy from the non-dominated solution set, the target purification strategy being one of the non-dominated solutions in the non-dominated solution set.

[0161] In a possible implementation, the plurality of purification strategy combinations includes an initial purification strategy combination and at least one target purification strategy combination; wherein each of the target purification strategy combinations is obtained by performing a crossover operation and / or a mutation operation on the initial purification strategy combination, the crossover operation being used to exchange part of the control parameters in any two of the purification strategy individuals in the initial purification strategy combination, and the mutation operation being used to change part of the control parameters in at least one of the purification strategy individuals in the initial purification strategy combination.

[0162] In a possible implementation, the third determining module 304 is specifically configured to determine a first feedback value corresponding to each of the purification strategy individuals based on each of the purification strategy individuals in the purification strategy combination and the comfort score value; determine a second feedback value corresponding to each of the purification strategy individuals based on each of the purification strategy individuals in the purification strategy combination and the filter performance parameter; determine a third feedback value corresponding to each of the purification strategy individuals based on each of the purification strategy individuals in the purification strategy combination and the operation energy consumption data; and perform non-dominated sorting on each of the purification strategy individuals in the purification strategy combination based on the first feedback value, the second feedback value and the third feedback value to obtain the non-dominated solution corresponding to the purification strategy combination.

[0163] In a possible implementation, the third determining module 304 is specifically configured to determine the target purification strategy from the non-dominated solution set based on the user preference configuration or the current working mode of the air purification system.

[0164] In a possible implementation, the target purification strategy includes at least one of the following: an air circulation mode, a fan air volume level, a switching time interval of the air purification strategy, and a duration of the purification mode.

[0165] Figure 4 is a structural schematic diagram of a vehicle provided by an embodiment of the present application. As shown in the example, Figure 4 The vehicle 400 includes a memory 401 and a processor 402, where the memory 401 stores executable program code 4011, and the processor 402 is configured to invoke and execute the executable program code 4011 to perform an air purification method.

[0166] In addition, an apparatus provided by an embodiment of the present application can include a memory and a processor, where the memory stores executable program code, and the processor is configured to invoke and execute the executable program code to perform an air purification method provided by an embodiment of the present application.

[0167] The embodiment can divide the apparatus into functional modules according to the above method examples, for example, each functional module can be provided, or two or more functions can be integrated into one processing module. The integrated module can be implemented in the form of hardware. It should be noted that the division of modules in the embodiment is illustrative, and is only a logical functional division. In actual implementation, another division manner can be used.

[0168] In the case of dividing each functional module according to each function, the apparatus can further include a data acquisition module, a first determination module, a second determination module, a third determination module, and a control module, etc. It should be noted that all related contents of each step involved in the above method embodiments can be referred to the function description of the corresponding functional module, which will not be repeated here.

[0169] It should be understood that the apparatus provided by the embodiment is used to perform the above air purification method, and thus can achieve the same effect as the above implementation method.

[0170] In the case of using an integrated unit, the apparatus can include a processing module and a storage module. When the apparatus is applied to a vehicle, the processing module can be used to control and manage the actions of the vehicle. The storage module can be used to support the vehicle to execute related program codes, etc.

[0171] The processing module can be a processor or a controller, which can implement or execute various exemplary logical blocks, modules, and circuits shown in combination with the disclosure. The processor can also be a combination of computing functions, such as one or more microprocessor combinations, digital signal processing (DSP) and microprocessor combinations, etc. The storage module can be a memory.

[0172] In addition, the device provided by the embodiments of the present application can be a chip, a component or a module, and the chip can include a processor and a memory connected thereto. The memory is used to store instructions, and when the processor calls and executes the instructions, the chip can perform the air purification method provided by the above embodiments.

[0173] The embodiments also provide a computer-readable storage medium having computer program codes stored therein, which, when executed on a computer, causes the computer to perform the above-mentioned related method steps to implement the air purification method provided by the above embodiments.

[0174] The embodiments also provide a computer program product, which, when executed on a computer, causes the computer to perform the above-mentioned related steps to implement the air purification method provided by the above embodiments.

[0175] The device, computer-readable storage medium, computer program product or chip provided by the embodiments can be used to execute the corresponding method provided above, and thus the beneficial effects achieved thereby can refer to the beneficial effects of the corresponding method provided above, which will not be described here.

[0176] Through the above description of the embodiments, those skilled in the art can understand that, for the convenience and brevity of description, only the above-mentioned division of functional modules is taken as an example for illustration, and in actual application, the above-mentioned functions can be completed by different functional modules according to needs, i.e. the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.

[0177] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, and the division of modules or units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another device, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the shown or discussed mutual elements can be indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0178] The above merely provides the specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. An air purification method, characterized in that, When applied to vehicles, the method includes: Acquire ambient air condition data of the vehicle, filter status data of the air purification system in the vehicle, and operating energy consumption data of the air purification system; A comfort score is determined based on the ambient air quality data. The comfort score reflects the user's satisfaction with the air quality inside the vehicle. The filter performance parameters are determined based on the filter status data. These parameters reflect the purification capacity, pressure drop characteristics, and service life of the filter in the air purification system. Based on the comfort score, the filter performance parameters, and the operating energy consumption data, a target purification strategy is determined. The air purification system is controlled to operate using the target purification strategy.

2. The method according to claim 1, characterized in that, The process of determining the comfort score based on the ambient air condition data includes: The ambient air state data is processed using a comfort characterization model to obtain the feature latent variable vector corresponding to the ambient air state data; The comfort score is obtained by processing the feature latent variable vector using a rating prediction model.

3. The method according to claim 2, characterized in that, The process of using a comfort characterization model to process the ambient air state data yields a feature latent variable vector corresponding to the ambient air state data, including: The encoder in the comfort characterization model is used to process the ambient air state data to obtain the mean vector and the standard deviation vector corresponding to the ambient air state data. The mean vector and the standard deviation vector are reparameterized using the reparameterization layer in the comfort representation model to obtain the feature latent variable vector.

4. The method according to claim 1, characterized in that, The filter element status data includes first status data, second status data, and third status data. The first status data is related to the purification capacity of the purification filter element, the second status data is related to the pressure drop characteristics of the purification filter element, and the third status data is related to the service life of the purification filter element. Determining filter element performance parameters based on the filter element status data includes: The first state data is processed using a purification efficiency model to obtain a purification efficiency percentage, which is used to reflect the purification capacity of the purification filter element. The second state data is processed using a pressure drop characteristic model to obtain a pressure drop level, which is used to reflect the pressure drop characteristics of the purification filter element. The third-state data is processed using a lifespan consumption model to obtain an estimated remaining lifespan of the filter element, which is used to reflect the service life of the purification filter element.

5. The method according to claim 1, characterized in that, The process of determining the target purification strategy based on the comfort score, the filter performance parameters, and the operating energy consumption data includes: Obtain multiple purification strategy combinations, each of which includes multiple individual purification strategies; Based on each of the purification strategy combinations, the comfort score, the filter performance parameters, and the operating energy consumption data, determine the non-dominated solution corresponding to each of the purification strategy combinations; Based on the non-dominated solution corresponding to each of the plurality of purification strategy combinations, determine the non-dominated solution set; A target purification strategy is determined from the set of non-dominated solutions, wherein the target purification strategy is a non-dominated solution in the set of non-dominated solutions.

6. The method according to claim 5, characterized in that, The multiple purification strategy combinations include an initial purification strategy combination and at least one target purification strategy combination; Each of the target purification strategy combinations is obtained by performing crossover and / or mutation operations on the initial purification strategy combination. The crossover operation is used to exchange some control parameters in any two purification strategy individuals in the initial purification strategy combination, and the mutation operation is used to change some control parameters in at least one purification strategy individual in the initial purification strategy combination.

7. The method according to claim 5, characterized in that, The process of determining the non-dominated solution corresponding to each purification strategy combination based on each purification strategy combination, the comfort score, the filter performance parameters, and the operating energy consumption data includes: Based on each individual purification strategy in the purification strategy combination and the comfort score, a first feedback value corresponding to each individual purification strategy is determined. Based on each individual purification strategy in the purification strategy combination and the filter performance parameters, a second feedback value corresponding to each individual purification strategy is determined. Based on each individual purification strategy in the purification strategy combination and the operating energy consumption data, a third feedback value corresponding to each individual purification strategy is determined. Based on the first feedback value, the second feedback value, and the third feedback value, the individual purification strategies in the purification strategy combination are non-dominated and sorted to obtain the non-dominated solution corresponding to the purification strategy combination.

8. The method according to claim 5, characterized in that, Determining the target purification strategy from the non-dominated solution set includes: The target purification strategy is determined from the non-dominated solution set based on user preference configuration or the current operating mode of the air purification system.

9. The method according to any one of claims 1 to 8, characterized in that, The target purification strategy includes at least one of the following: air circulation mode, fan airflow level, air purification strategy switching time interval, and purification mode duration.

10. A vehicle, characterized in that, The vehicles include: Memory, used to store executable program code; A processor for calling and running the executable program code from the memory, causing the vehicle to perform the method as described in any one of claims 1 to 9.

Images