Vehicle-mounted refrigerator temperature control method and system

Abstract

The invention provides a vehicle-mounted refrigerator temperature control method and system, and relates to the technical field of intelligent networked automobiles, and the method comprises the steps: firstly collecting the historical travel basic data of an automobile and the state data of the automobile, and enabling a starting point to be a reference starting area and an end point to be a reference destination area according to the coordinates of a starting point and a destination; and mapping the end point as a reference destination area and taking the nearest charging pile as a destination area. Generating a travel route based on the reference departure area and the destination area, and counting driving time; estimating predicted driving time when navigation is started, predicting a travel route by adopting the trained classification model, extracting parameters strongly related to travel charging through PCA, inputting the parameters into a logistic regression model to complete charging behavior prediction, and finally, according to a navigation state and a charging prediction result, calculating the charging behavior. And calculating the residual electric quantity after reaching the target area by combining the predicted driving time, the refrigerator power consumption parameter and the driving energy consumption per unit mileage of the vehicle, and dividing the travel into multiple types of typical operation scenes so as to control the operation mode of a refrigerator compressor.

Description

Technical Field

[0001] This invention relates to the field of intelligent connected vehicle technology, specifically to a method and system for controlling the temperature of an in-vehicle refrigerator. Background Technology

[0002] With the popularization of new energy vehicles and the increasing demand from consumers for travel quality, the requirements for the rationality and stability of power supply for in-vehicle refrigerators are constantly increasing.

[0003] In the prior art, such as the patent document with publication number CN119436724A, a power supply control method for a vehicle refrigerator is disclosed. This method obtains the trip plan and the remaining battery capacity, determines whether the battery supports the planned trip, and dynamically adjusts the on / off state and power supply of the refrigerator according to the power threshold after the trip ends and the working time of the refrigerator under different power levels. This realizes basic power supply planning based on power and trip, which optimizes the energy consumption management of the refrigerator to a certain extent while ensuring the power supply of the vehicle.

[0004] However, existing power supply control logic primarily revolves around the battery's current remaining capacity and the trip's inherent power requirements, failing to adequately incorporate user charging behavior planning or prediction. Furthermore, it doesn't fully connect the destination area definition to the core element of the charging scenario: the location of nearby charging stations. In real-world travel scenarios, users may have a clear plan to charge mid-journey or after reaching their destination. In such cases, even with low battery capacity, the anticipated power replenishment based on charging behavior theoretically eliminates the need to sacrifice user experience by reducing refrigerator power or even shutting it off. Existing methods, by not incorporating charging behavior into power supply control decision-making, may lead to excessive energy consumption restrictions on the refrigerator even in scenarios with planned charging, impacting the user experience. Therefore, a technical solution is needed that can dynamically adapt the refrigerator's power supply strategy based on user charging behavior prediction, better balancing refrigerator user experience and energy management while ensuring vehicle range.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a method and system for controlling the temperature of a vehicle-mounted refrigerator, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for controlling the temperature of a vehicle-mounted refrigerator, comprising the following steps: Step 1: Collect basic historical trip data and vehicle status data of the vehicle and perform anomaly cleaning. The trip data includes the coordinates of the departure point, the coordinates of the destination, the departure time and the arrival time. The vehicle status data includes the remaining battery power, charging records and refrigerator power consumption parameters. Step 2: Based on the origin and destination coordinates in the historical data, use a density clustering algorithm to classify nearby and dense origins into reference origin areas and destinations into reference destination areas. When navigation is enabled, the destination is directly mapped to the reference destination area, and the location of the nearest charging station in the reference destination area is defined as the destination area. Step 3: Generate travel routes based on reference departure and destination areas in historical data, calculate the travel time for each route, and when navigation is enabled, estimate the actual speed and calculate the estimated travel time based on the legal speed limit information of the road. Step 4: When navigation is not enabled, the trained classification model is used. The current reference departure area, departure time and remaining battery power are used as input features, and the route is used as the classification label to output the prediction result. The average driving time of the route is associated as the estimated driving time for this trip. Based on the reference departure area, destination area, departure time and remaining battery power in historical data, PCA is used to extract parameters that are strongly correlated with the trip charging, and the parameters are input into the logistic regression model to complete the charging behavior prediction. Step 5: Based on the navigation status and charging prediction results, combined with the estimated travel time, refrigerator power consumption parameters and vehicle energy consumption per unit mile, calculate the remaining power after the vehicle reaches the destination area, and compare it with the charging tolerance threshold and non-charging energy saving threshold respectively. Divide the current travel route into multiple typical operating scenarios to control the refrigerator compressor's operating mode.

[0008] Furthermore, the specific steps for collecting historical vehicle mileage data and vehicle status data, and performing anomaly cleanup, are as follows: When the vehicle remains stationary and off for a duration exceeding At the specified time, the destination is reached. ; The specific definitions and collection specifications for trip basic data and vehicle status data are as follows: Departure and destination coordinates: Departure and destination coordinates are obtained in real time through vehicle-mounted maps, i.e., the geographical coordinates of the start and end points of the trip; Departure and arrival times: Record the departure and arrival times, i.e., the start and end points of the trip with corresponding timestamps, and mark them in conjunction with the holiday database; Remaining battery power: Collected in real time by the vehicle battery management system, in percentage form, recorded every 5 minutes with a corresponding timestamp; Charging history: Records whether charging was performed each time the device reached its destination; Refrigerator power consumption parameters: collected in real time by the vehicle refrigerator control system, recording the energy consumption per unit time, recorded every 5 minutes and accompanied by a corresponding timestamp; Data with remaining power values ​​less than 0% or greater than 100%, refrigerator power consumption negative, or exceeding the rated power consumption range of the equipment will be excluded.

[0009] Furthermore, based on the latitude and longitude of the origin and destination, a density clustering algorithm is used to group nearby and densely packed origins into a reference origin region and destinations into a reference destination region. The normalized x-axis coordinate value of a point is obtained by subtracting the minimum x-axis coordinate value from the original x-axis coordinate value of a single point. This difference is then divided by the difference between the maximum and minimum x-axis coordinate values ​​of all points. A departure coordinate set is constructed based on the departure coordinates, and a destination coordinate set is constructed based on the destination coordinates. The density clustering algorithm DBSCAN is used to perform spatial clustering analysis on the departure and destination coordinates respectively, where a cluster radius parameter is set. Set a minimum number of points threshold In the clustering process, for each point in the origin coordinate set and the destination coordinate set, a neighborhood with a radius of 0.01 is constructed with that point as the center. The number of coordinate points contained in the neighborhood is counted. If a point's neighborhood contains no less than 5 coordinate points, then the point is determined to be a core point, and all points in its neighborhood constitute an initial clustering unit. If a point is not a core point but falls within the neighborhood of a core point, then the point is assigned to the cluster of the corresponding core point. If a point is neither a core point nor falls within the neighborhood of any core point, then it is determined to be an outlier and assigned to the nearest reference destination region. After clustering is completed, the clustering results of the departure coordinate set are integrated: all coordinate points contained in each cluster constitute a reference departure region. The geometric mean of the normalized coordinates of all coordinate points in this region is calculated and used as the centroid coordinate of the reference departure region to represent the reference departure region. The same operation is performed on the clustering results of the destination coordinate set to form multiple reference destination regions and their corresponding centroid coordinates to represent each reference destination region. The spatial location of the region is represented by numbering the departure region and the destination region.

[0010] Furthermore, when navigation is activated, the method for estimating the actual speed and calculating the estimated travel time based on legal speed limit information is as follows: After the navigation system generates the route, it first breaks down the route into road segments, extracts the road type and corresponding legal speed limit for each segment, and estimates the actual driving speed of each segment by multiplying the legal speed limit by a speed correction factor for different road types. Simultaneously, it obtains the mileage distance of each segment, calculates the driving time of a single segment, and sums up the driving times of all segments to finally obtain the estimated total driving time for this trip.

[0011] Furthermore, the method of using a trained classification model, with the departure area, departure time, and remaining battery power as input features, and the route as the classification label, to output the prediction result is as follows: The XGBoost classification model was trained, and samples were generated by extracting reference departure areas, departure timestamps, remaining battery power at departure, actual driving routes, and associated holiday information from historical data. Features were then extracted. Reference departure region characteristics: The region number of the reference departure region is used as a discrete feature; Departure time features: Decompose the departure timestamp into an integer feature of the hour, an integer feature of the day of the week, and a binary feature of whether it is a weekday; Remaining battery charge at departure: The percentage of the battery's total capacity at departure is retained, with a value ranging from 0 to 1. Holiday characteristics: Based on the associated holiday database, generate three binary characteristics: whether it is a holiday, whether it is 1 day before the holiday, and whether it is 1 day after the holiday; The actual routes of historical trips are classified using unique identifiers as labels. An XGBoost classification model is used. For core functional parameters, the objective function is set to directly output the class, the number of classes corresponds to the total number of unique identifiers of the trip routes, and the evaluation metric is set to multi-class log loss. For tree structure parameters, the maximum tree depth is set to 5, the minimum loss reduction required for splitting is set to 0.5, and the feature sampling ratio is set to 0.8. For training control parameters, the learning rate is set to 0.1, and the number of trees is set to 300. For regularization parameters, the L1 regularization coefficient is set to 0.1, the L2 regularization coefficient is set to 1, and the number of early stop rounds is set to 50.

[0012] Furthermore, based on the reference departure area, destination area, departure time, and remaining battery power, the method for extracting parameters related to trip charging intensity using PCA is as follows: Use unique hot coding for the area codes of the reference departure area and destination area; Periodically convert the hour and day of the week in the timestamp of the departure time: In the formula, This indicates the periodic conversion characteristic of hours in the timestamp. This indicates the periodic switching characteristic of the days of the week in the timestamp. The hour represents the timestamp. Indicates the day of the week in the timestamp. ; Then, the unique hot coding characteristics of the departure area, the unique hot coding characteristics of the destination area, the characteristics of the periodic conversion of the departure time by hour, the characteristics of the periodic conversion of the departure time by day of the week, whether it is a weekday, whether it is a holiday, whether it is one day before a holiday, whether it is one day after a holiday, and the characteristics of the remaining battery power at the time of departure will be integrated into a feature matrix: In the formula, Represents the characteristic matrix, Indicates the first The first sample Each reference starting region has a one-hot encoded feature, taking values ​​of 0 or 1. Indicates the sample index. , Indicates the number of samples. This indicates the reference starting area one-hot encoded index. , This indicates the total number of categories in the reference departure area; Indicates the first The first sample Each target region has a one-hot encoded feature, taking values ​​of 0 or 1. This represents the one-hot encoded index of the destination region. , This indicates the total number of categories in the reference departure area. Indicates the first The hourly periodicity characteristics of each sample. Indicates the first The periodic switching characteristics of the days of the week for each sample. Indicates the first Whether a sample is a weekday or not Indicates the first Whether a sample is a holiday or not Indicates the first Whether a sample is from the day before the holiday. Indicates the first Whether a sample is one day after the holiday is a characteristic. Indicates the first The remaining battery power characteristics of each sample at the start time, with the binary features all taking values ​​of 0 or 1; For matrix Centralize by column: In the formula, This represents the mean of each feature column. This represents the eigenvalue matrix after centering. Calculate the covariance matrix and perform eigenvalue decomposition to obtain eigenvalues ​​and eigenvectors. The eigenvectors are arranged in descending order according to their corresponding eigenvalues. Then, based on a preset variance explanation threshold... Determine the number of principal components, i.e., the cumulative number of components. The ratio of the sum of individual eigenvalues ​​to the sum of all eigenvalues ​​must be no less than the variance explained rate threshold. Select the smallest one that meets the cumulative contribution ratio requirement. The value is used to determine the number of principal components to be retained. These principal components are the extracted parameters related to the stroke charge intensity.

[0013] Furthermore, the method for using a logistic regression model to predict charging behavior is as follows: Extract the principal component parameter matrix that is correlated with the charge intensity during the stroke as the input feature matrix. The charging records corresponding to the trip route are used as label vectors. , Indicates the first Each sample is charged after arriving at its destination. Indicates the first If a sample arrives at its destination but is not charged, a binary classification logistic regression model is used, with the objective function set as the cross-entropy loss function: In the formula, Indicates the first Predicted charging probability value for each sample. , This represents the Sigmoid activation function. Represents the feature weight parameters. The bias term is introduced, and L2 regularization is used to suppress overfitting. The regularization coefficient is set to 0.01, and the number of iterations is set to 500. Early stopping is triggered when the validation set loss does not decrease for 10 consecutive iterations, and the optimal parameters are saved. and , For the predicted travel route, the principal component feature vector is constructed using the charging-related parameters of the travel route. Input the principal component feature vectors into the trained logistic regression model to calculate the charging probability: In the formula, This indicates the charging probability, with a pre-set charging probability threshold. ,when It is predicted that the vehicle will recharge upon reaching its destination.

[0014] Furthermore, based on the navigation status and charging prediction results, combined with the estimated driving time, refrigerator power consumption parameters, and vehicle energy consumption per unit mile, the method for calculating the remaining battery power after the vehicle reaches the destination area is as follows: The total distance traveled on the route is determined according to the following rules: When navigation is enabled, the actual road distance from the current departure location to the nearest charging station in the destination area is obtained and extracted and output by the navigation system in real time; when navigation is disabled, the road distance from the reference departure area to the nearest charging station corresponding to the predicted destination area is calculated, i.e., the distance from the reference departure area's centroid coordinates to the nearest charging station corresponding to the destination area's centroid coordinates. The power consumption of a refrigerator is determined according to the following rules: From the historical refrigerator power consumption database, the energy consumption per unit time that is consistent with the refrigerator's set temperature range and the current travel setting is filtered out. The arithmetic mean of the extracted unit time energy consumption data is taken to obtain the refrigerator's average power consumption. The standard operating condition energy consumption specified by the manufacturer is used to obtain the vehicle's driving energy consumption per unit mileage. Based on the estimated travel time and the refrigerator's average power consumption, calculate the total energy consumption of the refrigerator for this trip. Based on the total travel distance and the driving energy consumption per unit mile, calculate the total driving energy consumption for this trip. Add the total refrigerator energy consumption and the total driving energy consumption for this trip to obtain the total energy consumption. Then, the remaining power after reaching the destination area is: In the formula, This indicates the remaining battery power after reaching the destination area. This indicates the remaining battery power at the start of the journey. Indicates total energy consumption. This indicates the total capacity of the vehicle's battery.

[0015] Furthermore, by comparing the current travel route with the charging tolerance threshold and the non-charging energy-saving threshold respectively, the method for controlling the refrigerator compressor's operating mode by classifying the current travel route into multiple typical operating scenarios is as follows: Preset power threshold Non-charging energy saving threshold ,and Define the high power range as The medium power range is Low battery range ; When navigation is enabled: if the battery is high when arriving at the destination, the compressor operates in full-power cooling mode; if the battery is medium when arriving at the destination, the compressor operates at a reduced frequency. When the compressor reaches the destination area with low battery, its operating frequency decreases. ; When navigation is not enabled: If the battery is high upon arrival at the destination and the route indicates predicted charging, the compressor operates in full-power cooling mode; if the battery is medium upon arrival at the destination and the route indicates predicted charging, the compressor operates in full-power cooling mode; if the battery is low upon arrival at the destination and the route indicates predicted charging, the compressor operates at a reduced frequency. When the battery level is high upon arrival at the destination area, and the route is predicted not to require charging, the compressor operates in full-power cooling mode. When the battery level is medium upon arrival at the destination area, and the route is predicted not to require charging, the compressor operates at a reduced frequency. When the battery is low upon arrival at the destination area, and the route is predicted to not require charging, the compressor's operating frequency decreases. ,in, .

[0016] Additionally, a vehicle-mounted refrigerator temperature control system is provided, characterized in that: the system is used to execute the aforementioned vehicle-mounted refrigerator temperature control method, including: The data acquisition module is used to collect basic historical travel data and vehicle status data of the vehicle and perform anomaly cleaning. The basic travel data includes the coordinates of the departure point, the coordinates of the destination, the departure time and the arrival time. The vehicle status data includes the remaining battery power, charging records and refrigerator power consumption parameters. The destination area definition module is used to classify nearby and densely packed starting points as reference starting areas and ending points as reference destination areas based on the origin and destination coordinates in historical data and through density clustering algorithms. When navigation is enabled, the destination is directly mapped to the reference destination area, and the location of the nearest charging station in the reference destination area is defined as the destination area. The route time estimation module is used to generate travel routes based on reference departure and destination areas in historical data, calculate the travel time of each route, and estimate the actual speed and calculate the estimated travel time based on the legal speed limit information of the road when navigation is enabled. The charging behavior prediction module is used when navigation is not enabled. It uses a trained classification model with the current reference departure area, departure time and remaining battery power as input features and the route as the classification label to output the prediction result. It also associates the average travel time of the route as the estimated travel time for this trip. Based on the reference departure area, destination area, departure time and remaining battery power in historical data, it uses PCA to extract parameters that are strongly correlated with the trip charging and inputs them into the logistic regression model to complete the charging behavior prediction. The operation mode control module is used to calculate the remaining power of the vehicle after it reaches the destination area based on the navigation status and charging prediction results, combined with the estimated driving time, refrigerator power consumption parameters and vehicle driving energy consumption per unit mile. It then compares the remaining power with the charging tolerance threshold and the non-charging energy saving threshold, and divides the current travel route into multiple typical operating scenarios to control the refrigerator compressor's operating mode.

[0017] Compared with the prior art, the beneficial effects of the present invention are: This invention uses a density clustering algorithm to divide the origin and destination into regions, defining the nearest charging station location at the destination as the destination area. Combined with navigation status, it accurately generates the route and estimates the travel time, ensuring that route generation, time estimation, and subsequent energy consumption assessment are all closely related to the charging scenario. This ensures that the calculation of various parameters based on this region is more in line with the user's actual charging needs. This invention also uses PCA to extract strongly correlated charging parameters and a logistic regression model to predict charging behavior. Finally, based on the navigation status, charging prediction results, and the comparison of remaining power with a threshold, it divides the operating scenario to control the refrigerator compressor mode, avoiding sacrificing the user experience by reducing power or turning off the refrigerator when the user has a charging plan. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a graph showing the remaining battery power at the start of this invention. Figure 3 This is the charging probability threshold evaluation diagram for the present invention; Figure 4 This is a diagram showing the selection of charging probability thresholds for this invention. Figure 5 This is a power consumption statistics chart for the refrigerator of this invention; Figure 6 This is a schematic diagram of the overall system structure of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0020] It should be noted that, unless otherwise defined, the technical or scientific terms used in this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0021] Example: Please see Figures 1 to 5 The present invention provides a technical solution: A method for controlling the temperature of a vehicle-mounted refrigerator, comprising the following steps: Step 1: Collect basic historical trip data and vehicle status data of the vehicle and perform anomaly cleaning. The trip data includes the coordinates of the departure point, the coordinates of the destination, the departure time and the arrival time. The vehicle status data includes the remaining battery power, charging records and refrigerator power consumption parameters. When the vehicle switches from an off state to an on state, the trip is considered to have started. The two-dimensional coordinates at this moment are obtained in real time through the onboard map as the starting coordinates of the trip. If the vehicle remains stationary and in an off state for a duration longer than [a certain period], the trip will proceed. At the specified time, the destination is reached. ; Trip baseline data and vehicle status data are extracted from historical vehicle data. The specific definitions and collection specifications for trip baseline data and vehicle status data are as follows: Departure and destination coordinates: Departure and destination coordinates are obtained in real time through vehicle-mounted maps, i.e., the geographical coordinates of the start and end points of the trip; Departure and arrival times: Record the departure and arrival times, i.e., the start and end points of the trip with corresponding timestamps, and mark them in conjunction with the holiday database; Remaining battery power: Collected in real time by the vehicle battery management system, in percentage form, recorded every 5 minutes with a corresponding timestamp; Charging history: Records whether charging was performed each time the device reached its destination; Refrigerator power consumption parameters: collected in real time by the vehicle refrigerator control system, recording the energy consumption per unit time, recorded every 5 minutes and accompanied by a corresponding timestamp; Finally, the trip data and vehicle status data are cleaned for anomalies. For example, data with remaining battery power less than 0% or greater than 100%, refrigerator power consumption negative, or exceeding the device's rated power consumption range are removed. If a single trip lasts more than 24 hours, the records are split by calendar day. If the vehicle is started but not actually driven (e.g., only the air conditioner is turned on or the vehicle status is checked), the trip data and vehicle status data for that trip are removed if the speed is consistently 0 km / h or the time interval from start to stop is less than 30 minutes. If the vehicle travels a very short distance (e.g., moving the car within a residential area or adjusting a parking space), the trip data and vehicle status data for that trip are removed if the distance is less than 1 kilometer and the travel time is less than 10 minutes. The removal conditions can be set according to the actual situation to ensure data purity. Step 2: Based on the origin and destination coordinates in the historical data, use a density clustering algorithm to classify nearby and dense origins into reference origin areas and destinations into reference destination areas. When navigation is enabled, the destination is directly mapped to the reference destination area, and the location of the nearest charging station in the reference destination area is defined as the destination area. Firstly, in constructing the coordinate system, the coordinate systems that can be used include the National 2000 Coordinate System and local plane coordinate systems. The latitude and longitude coordinates within the region are converted into plane rectangular coordinates through Gauss-Kruger projection, so that the x-axis is in the east-west direction and the y-axis is in the north-south direction, with the unit being meters. This avoids the deviation caused by the spherical surface in calculating Euclidean distance and is more in line with the physical meaning of proximity in density clustering. For the obtained origin and destination coordinates, the x and y axis values ​​of the planar coordinates may vary significantly. Direct clustering would result in a greater weighting of the x and y axes, deviating from the actual spatial relationship. Therefore, it is necessary to perform normalization processing on the x and y axes of the planar coordinates separately: Subtract the minimum value of the x-axis coordinates of all coordinate points from the original x-axis coordinate value of a single coordinate point to obtain the difference between the x-axis coordinate of that point and the minimum value; then divide this difference by the difference between the maximum and minimum values ​​of the x-axis coordinates of all coordinate points, and the result is the normalized coordinate value of the x-axis of that point; Subtract the minimum value of the y-axis coordinates of all coordinate points from the original y-axis coordinate value of a single coordinate point, and then divide the resulting difference by the difference between the maximum and minimum values ​​of the y-axis coordinates to obtain the normalized coordinate value of the y-axis. This completes the normalization processing for each coordinate point, and the normalized coordinates are then directly used to replace the original coordinates for subsequent density clustering and region analysis. A set of origin coordinates is constructed based on the origin coordinates, and a set of destination coordinates is constructed based on the destination coordinates. The density clustering algorithm DBSCAN is used to perform spatial clustering analysis on the origin coordinates and destination coordinates respectively. DBSCAN does not require pre-specifying the number of clusters, can automatically identify the number of areas that users actually visit frequently, is not sensitive to noise such as remote locations visited by chance, and is suitable for the irregular distribution characteristics of geographic coordinates. Among them, the cluster radius parameter is set. A value of 0.01 corresponds to an actual distance of approximately 500 meters from the original coordinates. In practice, this is generally sufficient to cover nearby locations, such as different entrances to the same residential complex or office building. This prevents the same geographical target, which the user actually starts from or arrives at, from being incorrectly divided into multiple independent areas. A minimum point threshold is set accordingly. To ensure the representativeness of the region, the reference destination area is a frequently visited location, such as home or office, avoiding accidental visits like temporary shopping stops that would form independent clusters. Specifically, during clustering, for each point in both the origin and destination coordinate sets, a neighborhood with a radius of 0.01 is constructed centered on that point. The number of coordinate points within this neighborhood is counted. If a point's neighborhood contains at least 5 coordinate points, it is identified as a core point, and all points within its neighborhood form an initial cluster unit. If a point is not a core point but falls within the neighborhood of a core point, it is assigned to the corresponding core point's cluster. If a point is neither a core point nor falls within the neighborhood of any core point, it is identified as an outlier and assigned to the nearest reference destination area, preventing isolated points from affecting subsequent analysis. After clustering is completed, the clustering results of the departure coordinate set are integrated: all coordinate points contained in each cluster constitute a reference departure region. The geometric mean of the normalized coordinates of all coordinate points in this region is calculated and used as the centroid coordinate of the reference departure region to represent the reference departure region. The same operation is performed on the clustering results of the destination coordinate set to form multiple reference destination regions and their corresponding centroid coordinates to represent each reference destination region. Using the centroid coordinates, multidimensional discrete data can be compressed into a single coordinate point, realizing the dimensionality reduction representation of regional features and reducing the computational load of subsequent algorithms. As an average value, it can also offset the influence of local extreme values. Then, the spatial location of the region is represented by numbering the departure region and the destination region. In subsequent steps such as route prediction and charging behavior analysis, there is no need to repeatedly call the lengthy coordinate data. Table 1 shows 40 sets of basic data on the vehicle's historical trips and vehicle status data, covering the trip route number, reference departure area coordinates, destination area coordinates, departure timestamp, remaining battery power at departure, whether charging, refrigerator set temperature, and refrigerator power consumption. The data is then compiled into a historical data table. In the "Whether charging" field, 1 indicates yes and 0 indicates no.

[0022] Table 1 Historical Data Table like Figure 2 As shown, red dots correspond to the behavior of not charging, and blue dots correspond to the behavior of charging. This shows that when the remaining battery level is high at the start, that is, when the remaining battery level is greater than 50%, no charging behavior occurs. Conversely, when the remaining battery level is low at the start, that is, when the remaining battery level is less than 50%, charging behavior occurs. This intuitively reflects the behavior logic of charging when the battery level is low and not charging when the battery level is high.

[0023] Step 3: Generate travel routes based on reference departure and destination areas in historical data, calculate the travel time for each route, and when navigation is enabled, estimate the actual speed and calculate the estimated travel time based on the legal speed limit information of the road. The route generated based on the reference departure area and destination area outputs the centroid coordinates of the reference departure area, the centroid coordinates of the destination area, and the actual driving distance from the start point to the end point, which can be obtained directly through the navigation system. In scenarios where vehicle navigation is activated, the core premise is that the destination of the trip is uncertain or infrequent, requiring the navigation system to obtain accurate route information. When the vehicle navigation is activated, after the navigation system generates the trip route, it first breaks down the route into road segments, extracts the road type and corresponding legal speed limit for each segment, and estimates the actual driving speed of each segment by multiplying the legal speed limit by a speed correction factor for different road types. For example, 0.8 is used for highways, 0.6 for main roads, and 0.5 for side roads. Alternatively, the correction factor can be determined by the ratio of the historical average speed of the same road segment over the past 3 months to the legal speed limit. Simultaneously, the mileage distance of each segment is obtained, the driving time of a single segment is calculated, and the driving times of all segments are summed to obtain the estimated total driving time for this trip. This provides a reliable time benchmark for subsequent time-based battery assessment and refrigerator operation mode control.

[0024] Step 4: When navigation is not enabled, the trained classification model is used. The current reference departure area, departure time and remaining battery power are used as input features, and the route is used as the classification label to output the prediction result. The average driving time of the route is associated as the estimated driving time for this trip. Based on the reference departure area, destination area, departure time and remaining battery power in historical data, PCA is used to extract parameters that are strongly correlated with the trip charging, and the parameters are input into the logistic regression model to complete the charging behavior prediction. The XGBoost classification model was trained, and samples were generated by extracting reference departure areas, departure timestamps, remaining battery power at departure, actual driving routes, and associated holiday information from historical data. Features were then extracted. Reference starting region features: The region number of the reference starting region is used as a discrete feature, which can be converted into a 0 / 1 vector through one-hot encoding. Departure time features: Decompose the departure timestamp into an integer feature of the hour, an integer feature of the day of the week, and a binary feature of whether it is a weekday, where 1 indicates yes and 0 indicates no; Remaining battery charge at start time: The percentage of the battery charge at start time relative to the total battery capacity is retained, with a value range of 0-1. For example, if the battery charge percentage is 30%, it is normalized to a floating-point number of 0.3 (0-1). Holiday characteristics: Based on the associated holiday database, generate three binary features: whether it is a holiday, whether it is 1 day before the holiday, and whether it is 1 day after the holiday. 1 indicates yes and 0 indicates no. In the reference departure area feature, different reference departure areas, such as home and company, correspond to different travel purposes and route preferences. Discrete features can capture this spatial correlation. In the time feature, travel time may directly affect route selection and charging needs. For example, evening travel may require charging. Breaking it down into hourly, weekly, and weekday features can accurately capture periodic patterns. For example, there are different patterns during Monday morning rush hour and weekend afternoon. The remaining battery power at departure determines whether users need to charge at their destination and is the core feature for predicting charging behavior. In the holiday feature, for some people, travel during holidays usually involves longer distances and more complex routes, and the charging needs are significantly different from weekdays. Adding this feature can improve the model's adaptability to special scenarios.

[0025] The actual routes of historical trips are classified using unique identifiers as classification labels, and the XGBoost classification model is adopted. This is because the selection of trip routes is affected by the interaction of multiple factors, such as a reference departure area and the morning rush hour. The tree structure of XGBoost can capture this non-linear relationship, which is better than the linear model. By integrating multiple trees, it can reduce the noise impact of a single tree and effectively address the problem that there may be a small number of anomalies in the vehicle data. For core functional parameters, the objective function is set to directly output the category, the number of categories corresponds to the total number of unique identifiers for the route, and the evaluation metric is set to multi-class log loss. For tree structure parameters, the maximum tree depth is set to 5 to limit the tree depth and avoid overfitting, the minimum loss reduction required for splitting is set to 0.5 to avoid ineffective splits that only slightly improve the loss, and the feature sampling ratio is set to 0.8, randomly sampling 80% of the features for each split to increase the diversity between trees and improve the ensemble effect. For training control parameters, the learning rate is set to 0.1, and the number of trees is set to 300. For regularization parameters, the L1 regularization coefficient is set to 0.1, and the L2 regularization coefficient is set to 1, which can effectively suppress overfitting together. The number of early stop rounds is set to 50 to avoid ineffective iterations and overfitting. The features of the current reference departure area, departure timestamp, remaining battery power at departure, actual driving route, and associated holiday information are extracted using the same method and input into the trained XGBoost classification model to obtain the predicted route.

[0026] Next, based on the reference departure area, destination area, departure time, and remaining battery power, PCA is used to extract parameters strongly correlated with the trip's charging strength. One-hot encoding is then applied to the region numbers of the reference departure and destination areas. Directly inputting incrementally increasing numerical numbers into the model would misinterpret it as a linear relationship between regions. One-hot encoding assigns an independent feature dimension to each region, ensuring that the uniqueness of each region is not compressed. The hour and day of the week in the departure time timestamp are periodically converted. In the formula, This indicates the periodic conversion characteristic of hours in the timestamp. This indicates the periodic switching characteristic of the days of the week in the timestamp. The hour represents the timestamp. Indicates the day of the week in the timestamp. To avoid issues such as 11 PM and 1 PM being close but having large numerical differences, the same applies to days of the week; Then, the unique hot coding characteristics of the departure area, the unique hot coding characteristics of the destination area, the characteristics of the periodic conversion of the departure time by hour, the characteristics of the periodic conversion of the departure time by day of the week, whether it is a weekday, whether it is a holiday, whether it is one day before a holiday, whether it is one day after a holiday, and the characteristics of the remaining battery power at the time of departure will be integrated into a feature matrix: In the formula, Represents the characteristic matrix, Indicates the first The first sample Each reference starting region has a one-hot encoded feature, taking values ​​of 0 or 1. Indicates the sample index. , Indicates the number of samples. This indicates the reference starting area one-hot encoded index. , This indicates the total number of categories in the reference departure area; Indicates the first The first sample Each target region has a one-hot encoded feature, taking values ​​of 0 or 1. This represents the one-hot encoded index of the destination region. , This indicates the total number of categories in the reference departure area. Indicates the first The hourly periodicity characteristics of each sample. Indicates the first The periodic switching characteristics of the days of the week for each sample. Indicates the first Whether a sample is a weekday or not Indicates the first Whether a sample is a holiday or not Indicates the first Whether a sample is from the day before the holiday. Indicates the first Whether a sample is one day after the holiday is a characteristic. Indicates the first The remaining battery power characteristics of each sample at the start time, with the binary features all taking values ​​of 0 or 1; Table 2 shows the PCA input feature table, which is a feature matrix sample formed by extracting features from the original travel information. Here, 1 indicates yes and 0 indicates no. The table is generated after summarizing 40 sets of data. These features provide data for subsequent PCA dimensionality reduction and model training by quantifying information such as region, time, and scene.

[0027] Table 2 PCA Input Feature Table PCA dimensionality reduction is performed on the original feature matrix, which contains correlations, such as a high correlation between holidays and non-working days. PCA transforms these correlations into uncorrelated principal components through orthogonal transformation. The original feature matrix has a high dimensionality, but PCA compression allows it to centrally reflect the comprehensive information most relevant to charging behavior. Centralize by column: In the formula, This represents the mean of each feature column. This represents the centered feature matrix, eliminating dimensional differences. Calculate the covariance matrix and perform eigenvalue decomposition to obtain eigenvalues ​​and eigenvectors. The eigenvectors are arranged in descending order according to their corresponding eigenvalues. Then, based on a preset variance explanation threshold... Determine the number of principal components, i.e., the cumulative number of components. The ratio of the sum of individual eigenvalues ​​to the sum of all eigenvalues ​​must be no less than the variance explained rate threshold. Select the smallest one that meets the cumulative contribution ratio requirement. The value is used to determine the number of principal components to be retained. These principal components are the extracted parameters related to the stroke charge intensity.

[0028] Charging behavior is a typical binary classification problem, and logistic regression is a classic model designed for binary classification tasks. Its core function is to map the input features to probability values ​​between 0 and 1 using the sigmoid function, directly matching the target of "predicting whether to charge". Predicting charging behavior can determine in advance whether there are charging conditions after the vehicle arrives at its destination. If charging is predicted, the refrigerator can maintain a high power consumption while driving without deliberately saving power; if charging is predicted, the power consumption mode can be adjusted as needed to prioritize the vehicle's range.

[0029] The samples were divided into training and validation sets in a 7:3 ratio, ensuring a consistent ratio of charging / non-charging labels during the partitioning. Principal component parameter matrices strongly correlated with trip charging were extracted as the input feature matrix. The charging records corresponding to the trip route are used as label vectors. , Indicates the first Each sample is charged after arriving at its destination. Indicates the first If a sample arrives at its destination but is not charged, a binary classification logistic regression model is used, with the objective function set as the cross-entropy loss function: In the formula, Indicates the first Predicted charging probability value for each sample. , This represents the Sigmoid activation function. This represents the feature weight parameters, which are initialized to the dimension. The zero vector is used to avoid unstable training due to excessively large initial values. This represents the bias term, initialized to 0. The Adam optimizer is used, which converges faster and is more robust to the learning rate. The learning rate is set to 0.01 by default, and L2 regularization is introduced to suppress overfitting. In the formula, This represents the regularization coefficient, with a value of 0.01. The number of iterations is set to 500. In each iteration, a batch of samples is randomly drawn from the training set. For example, if the batch size is chosen to be 32 or 64, the calculation is performed. Predicted charging probability value for each sample ,loss and gradient (for) and The partial derivatives of the parameter are used to update the parameters through the optimizer. Early stopping is triggered when the validation set loss does not decrease for 10 consecutive iterations, and the optimal parameters are saved. and For the predicted route, the principal component feature vector is constructed using the charging-related parameters of the route. Input the principal component feature vectors into the trained logistic regression model to calculate the charging probability: In the formula, This indicates the charging probability, with a pre-set charging probability threshold. ,when The system predicts that the vehicle will charge upon reaching the destination area; otherwise, it predicts that it will not charge upon reaching the destination area. The specific value can be determined by calculating the F1 score under different values, and selecting the highest F1 value as the charging probability threshold. For example, in the validation set... When F1 is optimal, this value of 0.6 is fixed as the charging probability threshold. Table 3 shows the analysis of whether the prediction of trip charging behavior is correct when the charging probability threshold is 0.6. Among the 40 samples, 38 groups were predicted correctly, and the model prediction results are highly consistent with the actual scenario.

[0030] Table 3 Logistic Regression Validation Table like Figures 3-4 As shown, as the charging probability threshold increases from 0.4 to 0.8, the precision gradually increases, but the recall continues to decrease. A high threshold will miss more samples that actually need charging. The F1 score is the harmonic mean of precision and recall, which comprehensively measures the classification performance of the model. The F1 score is the highest when the charging probability threshold is 0.6, which achieves the optimal balance between avoiding misjudgment and covering real-world scenarios.

[0031] Step 5: Based on the navigation status and charging prediction results, combined with the estimated travel time, refrigerator power consumption parameters and vehicle unit mileage driving energy consumption, calculate the remaining power after the vehicle reaches the destination area, and compare it with the charging tolerance threshold and non-charging energy saving threshold respectively. Divide the current travel route into multiple typical operating scenarios to control the refrigerator compressor operation mode. The total distance traveled on the route is determined according to the following rules: When navigation is enabled, the actual road distance from the current departure location to the nearest charging station in the destination area is obtained. This is the actual road distance that generates the recommended driving route that conforms to actual traffic rules. The navigation system extracts and outputs this distance in real time. This distance includes real-time traffic condition corrections, such as detours to avoid congestion, thus outputting the total distance of the actual road path. When navigation is disabled, the road distance from the reference departure area to the nearest charging station corresponding to the predicted destination area is used. This is the road distance from the reference centroid coordinates of the departure area to the nearest charging station corresponding to the centroid coordinates of the destination area. This is the optimal actual road distance from historical data, ensuring the accuracy of the total energy consumption calculation. The power consumption of a refrigerator is determined according to the following rules: From historical refrigerator power consumption data, energy consumption per unit time that matches the refrigerator's set temperature range and the current travel setting is selected to ensure consistency between the extracted historical power consumption and the current usage scenario. The arithmetic mean of the extracted unit time energy consumption data is then calculated to obtain the refrigerator's average power consumption. Within the selected unit time energy consumption data that matches the refrigerator's set temperature range and the current travel setting, the set temperature can be divided into several intervals based on the refrigerator's adjustment settings. For example, a regular refrigeration mode above 3℃, a rapid cooling mode from 0℃ to 3℃, a low-temperature refrigeration mode from -5℃ to 0℃, a light freezing mode from -18℃ to -5℃, and a deep freezing mode from -18℃ to -22℃. This allows for precise matching of power consumption characteristics under different cooling intensities. Figure 5 As shown, the power consumption of a refrigerator is affected by a variety of factors in actual use. The power consumption bars in each mode are not completely consistent and fluctuate to some extent. Different cooling modes correspond to different cooling intensities and power consumption logic. By dividing the mode into temperature ranges and filtering the data, the average value is calculated, which is more in line with the actual power consumption of the device under specific cooling needs. The standard operating condition energy consumption specified by the manufacturer is used to obtain the vehicle's driving energy consumption per unit mileage. For example, the WLTP condition in the vehicle's factory parameters is the energy consumption per 100 kilometers under the globally unified light vehicle test procedure. It is the mainstream test standard in the world and is closer to the actual road driving conditions. After conversion, the driving energy consumption per unit mileage of the vehicle is obtained. Based on the estimated travel time and the refrigerator's average power consumption, calculate the total energy consumption of the refrigerator for this trip: In the formula, The total energy consumption of the refrigerator represents the travel route. This indicates the average power consumption of the refrigerator. Indicates the estimated travel time for the route; Calculate the total driving energy consumption for this trip based on the total driving distance and the driving energy consumption per unit mileage: In the formula, This represents the total driving energy consumption for the entire journey. This indicates the total distance traveled. This represents the vehicle's driving energy consumption per unit distance. The total energy consumption is obtained by adding the total energy consumption of the refrigerator and the total driving energy consumption for this trip. In the formula, Indicates total energy consumption; The remaining battery power after reaching the destination area is: In the formula, This indicates the remaining battery power after reaching the destination area. This indicates the remaining battery power at the start of the journey. This indicates the total capacity of the vehicle's battery.

[0032] Preset power threshold Non-charging energy saving threshold ,and Define the high power range as The medium power range is Low battery range The remaining battery power is divided into high, medium, and low ranges, providing a clear quantitative basis for the refrigerator's operating mode. For example, it can be set... A value of 0.2 and 20% represents the "charging anxiety threshold" for most users. Below this value, the risk of range anxiety increases, necessitating the prioritization of limiting unnecessary energy consumption to ensure access to charging stations. With a power rating of 0.5, more than 50% of the battery is sufficient, enough to meet the needs of short-distance travel even without charging. There is no need to limit the power consumption of the refrigerator, and user experience can be prioritized. When navigation is enabled: When the battery is high upon arrival at the destination, the compressor operates in full-power cooling mode. At this time, the battery is fully charged, so energy saving is unnecessary, and cooling performance is prioritized. When the battery is medium upon arrival at the destination, the compressor operates at a reduced frequency. When the battery level is moderate, moderate energy saving is implemented to balance user experience and battery life; when the battery level is low upon arrival at the destination, the compressor operates at a reduced frequency. With power supplies tight, it is necessary to significantly reduce power consumption and prioritize ensuring that vehicles reach charging stations. When navigation is not enabled: If the battery is high upon arrival at the destination and the route indicates a predicted charge, the compressor operates in full-power cooling mode; if the battery is medium upon arrival and the route indicates a predicted charge, the compressor operates in full-power cooling mode. Since there will be subsequent charging, energy saving is not necessary, prioritizing user experience; if the battery is low upon arrival and the route indicates a predicted charge, the compressor operates at a reduced frequency. Since charging can replenish the battery, only moderate energy saving is needed; when the battery is high upon arrival at the destination and the route is predicted not to require charging, the compressor operates in full-power cooling mode, and no energy saving is needed as the battery is sufficient; when the battery is medium upon arrival at the destination and the route is predicted not to require charging, the compressor operating frequency decreases. At this point, the battery level is moderate, so moderate energy saving is recommended to balance user experience and battery life. When the battery level is low upon arrival at the destination, and the route is predicted to not require charging, the compressor's operating frequency will decrease. With power supplies tight, a significant reduction in power consumption is needed, including... Adjust the values ​​according to the actual situation, for example This ensures basic cooling and prioritizes energy conservation, ensuring that the refrigerator operates in accordance with actual battery levels and charging needs, ultimately achieving the optimal balance between battery life and user experience.

[0033] Please see Figure 6 The present invention also provides a vehicle refrigerator temperature control system for executing the above-described vehicle refrigerator temperature control method, comprising: The data acquisition module is used to collect basic historical travel data and vehicle status data of the vehicle and perform anomaly cleaning. The basic travel data includes the coordinates of the departure point, the coordinates of the destination, the departure time and the arrival time. The vehicle status data includes the remaining battery power, charging records and refrigerator power consumption parameters. The destination area definition module is used to classify nearby and densely packed starting points as reference starting areas and ending points as reference destination areas based on the origin and destination coordinates in historical data and through density clustering algorithms. When navigation is enabled, the destination is directly mapped to the reference destination area, and the location of the nearest charging station in the reference destination area is defined as the destination area. The route time estimation module is used to generate travel routes based on reference departure and destination areas in historical data, calculate the travel time of each route, and estimate the actual speed and calculate the estimated travel time based on the legal speed limit information of the road when navigation is enabled. The charging behavior prediction module is used when navigation is not enabled. It uses a trained classification model with the current reference departure area, departure time and remaining battery power as input features and the route as the classification label to output the prediction result. It also associates the average travel time of the route as the estimated travel time for this trip. Based on the reference departure area, destination area, departure time and remaining battery power in historical data, it uses PCA to extract parameters that are strongly correlated with the trip charging and inputs them into the logistic regression model to complete the charging behavior prediction. The operation mode control module is used to calculate the remaining power of the vehicle after it reaches the destination area based on the navigation status and charging prediction results, combined with the estimated driving time, refrigerator power consumption parameters and vehicle driving energy consumption per unit mile. It then compares the remaining power with the charging tolerance threshold and the non-charging energy saving threshold, and divides the current travel route into multiple typical operating scenarios to control the refrigerator compressor's operating mode.

[0034] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0035] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution.

[0036] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0037] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for controlling the temperature of a vehicle refrigerator, characterized by, The specific steps include: Step 1: Collecting automobile historical travel basic data and vehicle state data and performing abnormal cleaning, the travel basic data including departure coordinate, destination coordinate, departure time and arrival time, and the vehicle state data including remaining power, charging record and refrigerator power consumption parameter; Step 2: According to the departure coordinate and destination coordinate in the historical data, adjacent and dense starting points are classified as reference departure area and ending points are classified as reference destination area by density clustering algorithm, when navigation is started, the destination is directly mapped to the reference destination area, and the charging pile position closest to the reference destination area is defined as the destination area; Step 3: Generating travel route based on the reference departure area and destination area in the historical data, and counting the travel time of each travel route, when navigation is started, the actual speed is estimated according to the legal speed limit information of the road and the estimated travel time is calculated; Step 4: When navigation is not started, the trained classification model is used, the current reference departure area, departure time and remaining power are taken as input features, the travel route is taken as classification label, the prediction result is output, and the average travel time of the travel route is taken as the estimated travel time of this time, based on the reference departure area, destination area, departure time and remaining power in the historical data, the parameters strongly related to travel charging are extracted by PCA, and the charging behavior prediction is completed by inputting the logistic regression model; Step 5: According to the navigation state and charging prediction result, combining the estimated travel time, refrigerator power consumption parameter and vehicle unit mileage driving energy consumption, the remaining power of the vehicle after reaching the destination area is calculated, and compared with the charging tolerance threshold and non-charging energy saving threshold respectively, the current travel route is divided into multiple typical operation scenes to control the operation mode of the refrigerator compressor.

2. The temperature control method of a vehicle-mounted refrigerator according to claim 1, characterized by: The specific steps of collecting automobile historical travel basic data and vehicle state data and performing abnormal cleaning are as follows: When the vehicle is in a stationary and turned off state for a duration higher than one minute, it is noted as reaching the end point, wherein, ; The specific definition and collection specification of travel basic data and vehicle state data are as follows: Departure coordinate and destination coordinate: obtain the departure coordinate data and destination coordinate data through real-time positioning of vehicle-mounted map, that is, the geographic coordinates of the starting point and ending point of the trip; Departure time and arrival time: record the time of departure and arrival at the destination, that is, the starting point and ending point of the trip are accompanied by corresponding time stamp, and are marked by associating with holiday database; Remaining power: collected by vehicle battery management system in real time, in percentage unit, recorded every 5 minutes with corresponding time stamp; Charging record: record whether charging is performed every time the destination is reached; Refrigerator power consumption parameter: collected by vehicle-mounted refrigerator control system in real time, record the energy consumption per unit time, recorded every 5 minutes with corresponding time stamp; Eliminate data with remaining power less than 0% or more than 100%, negative refrigerator power consumption or exceeding the rated power consumption range of the device.

3. The temperature control method of a vehicle-mounted refrigerator according to claim 2, characterized by: The method of classifying adjacent and dense starting points as reference departure area and ending points as reference destination area according to the departure coordinate and destination coordinate in the historical data by density clustering algorithm is as follows: The x-axis original coordinate value of a single coordinate point is subtracted by the minimum value in all coordinate points x-axis coordinates, to obtain the difference value of the x-axis coordinate of the point and the minimum value; then the difference value is divided by the difference value between the maximum value and the minimum value in all coordinate points x-axis coordinates, and the calculation result is the normalized coordinate value of the x-axis of the point; the y-axis original coordinate value of a single coordinate point is subtracted by the minimum value in all coordinate points y-axis coordinates, and the obtained difference value is divided by the difference value between the maximum value and the minimum value of the y-axis coordinates, to obtain the normalized coordinate value of the y-axis, and the normalization processing of each coordinate point is completed; A departure location coordinate set is constructed based on the departure location coordinates, a destination location coordinate set is constructed based on the destination location coordinates, and a density clustering algorithm DBSCAN is used to perform spatial clustering analysis on the departure location coordinates and the destination location coordinates respectively , a minimum point threshold is set, that is, in the clustering process, for each point in the departure location coordinate set and the destination location coordinate set, a neighborhood with a radius of 0.01 is constructed with the point as the center, the number of coordinate points contained in the neighborhood is counted, if the neighborhood of a point contains not less than 5 coordinate points, the point is determined as a core point, and all points in the neighborhood of the core point constitute an initial clustering unit; if a point is not a core point but falls within the neighborhood of a core point, the point is attributed to the clustering of the corresponding core point; if a point is neither a core point nor falls within the neighborhood of any core point, the point is determined as an outlier point and is attributed to the nearest reference destination region; After clustering is completed, the clustering results of the departure coordinate set are integrated: all coordinate points contained in each cluster form a reference departure area, the geometric mean of the normalized coordinates of all coordinate points in the area is calculated as the centroid coordinate of the reference departure area, representing the reference departure area, and the same operation is performed on the clustering results of the destination coordinate set to form multiple reference destination areas and corresponding centroid coordinates, representing each reference destination area, and the spatial positions of the areas are represented by numbering the departure areas and the destination areas.

4. The temperature control method of a vehicle-mounted refrigerator according to claim 1, characterized by: When the navigation is started, the method for estimating the actual speed and calculating the predicted travel time according to the legal speed limit information is: After the navigation system generates a travel route, the route is first split by road section granularity, the road type and corresponding legal speed limit value of each road section are extracted, and for different road types, the actual travel speed of each road section is estimated by multiplying the legal speed limit value by the speed correction coefficient, the mileage distance of each road section is simultaneously obtained, the travel time of a single road section is calculated, the travel times of all road sections are accumulated and summed, and finally the predicted total travel time of this trip is obtained.

5. The temperature control method of a vehicle-mounted refrigerator according to claim 2, characterized by: The method for using the trained classification model to input the current reference departure area, departure time and remaining power as input features, and output the travel route as a classification label to predict the result is: The XGBoost classification model is trained, the reference departure area, departure timestamp, remaining power at departure, actual travel route and associated holiday information in the historical data are extracted to generate samples, and the features are extracted: Reference departure area feature: the area number of the reference departure area as a discrete feature; Departure time feature: the departure timestamp is decomposed into an integer feature of hours, an integer feature of the day of the week and a binary feature of whether it is a weekday; Remaining power at departure feature: the proportion of the actual power at departure to the total capacity of the battery is retained, with a value range of 0-1; Holiday feature: based on the associated holiday database, three binary features of whether it is a holiday, whether it is one day before the holiday and whether it is one day after the holiday are generated; The actual travel route of the historical trip is classified by a unique identifier as a classification label, for the core function parameters, the objective function is set to directly output the category, the number of categories is set to correspond to the total number of unique identifiers of the travel route, and the evaluation index is set to the multi-class logarithmic loss; for the tree structure parameters, the maximum depth of the tree is set to 5, the minimum loss reduction required for splitting is set to 0.5, and the feature sampling ratio is set to 0.8; for the training control parameters, the learning rate is set to 0.1, and the number of trees is set to 300; for the regularization parameters, the L1 regularization coefficient is set to 0.1, and the L2 regularization coefficient is set to 1; the early stopping round is set to 50; The current reference departure area, departure timestamp, remaining power at departure, actual driving route, and associated holiday information are extracted using the same method, and input into the trained XGBoost classification model to obtain the prediction result of the travel route.

6. The temperature control method of a vehicle-mounted refrigerator according to claim 5, characterized by: Based on the reference departure area, destination area, departure time, and remaining power in the historical data, the method for extracting parameters strongly related to travel charging using PCA is as follows: The area numbers of the reference departure area and the destination area are one-hot encoded; The hour and day of the week in the departure time timestamp are periodically converted: In the formula, represents a periodic conversion feature in the time stamp for hours, represents a periodic conversion feature in the time stamp for the day of the week, represents the hours of the time stamp, represents the day of the week of the time stamp, ; Then, the reference departure area one-hot encoded feature, the destination area one-hot encoded feature, the hour periodicity converted feature in the departure time, the day of the week periodicity converted feature in the departure time, whether it is a weekday, whether it is a holiday, whether it is a day before the holiday, whether it is a day after the holiday, and the remaining power at departure feature are integrated into a feature matrix: In the formula, represents a characteristic matrix, represents the first reference departure area one-hot encoding characteristic of the first sample, taking values 0 or 1, represents a sample index, , represents a sample quantity, represents a reference departure area one-hot encoding index, , represents a total number of reference departure area categories; represents the first destination area one-hot encoding characteristic of the first sample, taking values 0 or 1, represents a destination area one-hot encoding index, , represents a total number of reference departure area categories, represents an hourly periodicity conversion characteristic of the first sample, represents a day-of-week periodicity conversion characteristic of the first sample, represents a weekday feature of the first sample, represents a holiday feature of the first sample, represents a pre-holiday 1 day feature of the first sample, represents a post-holiday 1 day feature of the first sample, represents a departure-time remaining power characteristic of the first sample, taking values 0 or 1 for binary characteristics. On the matrix Centering by column: wherein, denotes the mean of each column of features, denotes the centered feature matrix; The covariance matrix is calculated and eigen decomposition is performed to obtain eigenvalues and eigenvectors, the eigenvectors are arranged in descending order according to the corresponding eigenvalues, and a preset variance explanation rate threshold is obtained The number of principal components is determined, that is, the ratio of the sum of the first The sum of eigenvalues to the sum of all eigenvalues needs to be not less than the variance explanation rate threshold The minimum value meeting the cumulative contribution ratio requirement is screened out, so as to determine the number of principal components to be retained These principal components are the extracted parameters strongly related to the travel charging.

7. The temperature control method of a vehicle-mounted refrigerator according to claim 6, characterized by: The method for inputting a logistic regression model to complete charging behavior prediction is as follows: Extract the principal component parameter matrix strongly related to the trip charging as the input feature matrix , and the charging record corresponding to the trip route as the label vector , indicates that the first sample charges after arriving at the destination, indicates that the first sample does not charge after arriving at the destination, a binary classification logistic regression model is used, and the objective function is set as the cross-entropy loss function: In the formula, The charging probability prediction value of the first sample is represented by , The Sigmoid activation function is represented by The feature weight parameter is represented by The bias term is represented by L2 regularization is introduced to suppress overfitting, the regularization coefficient is set to 0.01, the number of iterations is set to 500, early stopping is triggered when the validation set loss does not decrease for 10 consecutive iterations, and the optimal parameters are saved and , For the predicted travel route, a principal component feature vector is constructed using parameters strongly related to charging of the travel route The principal component feature vector is input into the trained logistic regression model to calculate the charging probability: In the formula, represents the charging probability, and a charging probability threshold is preset , when , it is predicted that charging will be performed after reaching the destination area in the travel route.

8. The temperature control method of a vehicle-mounted refrigerator according to claim 7, characterized by: According to the navigation state and the charging prediction result, combined with the estimated driving time, the refrigerator power consumption parameter, and the vehicle unit mileage driving energy consumption, the method for calculating the remaining power of the vehicle after reaching the destination area is as follows: The total driving distance of the travel route is determined according to the following rules: When the navigation is turned on, the actual road distance from the current departure location to the nearest charging pile location of the destination area is obtained, which is extracted and output in real time by the navigation system; when the navigation is not turned on, the road distance from the reference departure area to the nearest charging pile location of the destination area, i.e. the centroid coordinates of the reference departure area, to the nearest charging pile location of the centroid coordinates of the destination area, is obtained; The refrigerator power consumption is determined according to the following rules: From the historical refrigerator power consumption database, the energy consumption per unit time is selected when the refrigerator set temperature range is consistent with the current trip setting, and the arithmetic mean of the extracted unit time energy consumption data is taken to obtain the average power consumption of the refrigerator; The standard working condition energy consumption marked by the manufacturer is used to obtain the vehicle unit mileage driving energy consumption; Based on the estimated driving time and the average power consumption of the refrigerator, the total energy consumption of the refrigerator on the current travel route is calculated, based on the total driving distance and the unit mileage driving energy consumption, the total driving energy consumption of the current travel route is calculated, and the total energy consumption of the current travel route is obtained by adding the total energy consumption of the refrigerator and the total driving energy consumption, then the remaining power after reaching the destination area is: In the formula, represents the remaining power after reaching the destination area, represents the remaining power at the time of departure, represents the total energy consumption, represents the total capacity of the vehicle battery.

9. The temperature control method of a vehicle-mounted refrigerator according to claim 8, characterized by: The method for comparing the current travel route with the charging tolerance threshold and the non-charging energy saving threshold to divide the current travel route into multiple typical operating scenarios to control the operation mode of the refrigerator compressor is as follows: Pre-set power threshold and non-charging energy saving threshold , and , define the high power interval as , the medium power interval as , and the low power interval as ; In the scenario of navigation starting: when reaching the destination area at high power, the compressor adopts full-power refrigeration mode; when reaching the destination area at medium power, the compressor runs at a reduced frequency ; Compressor operating frequency is reduced when arriving at the destination area with low battery ; When the destination area is reached at high power and the travel route is predicted to be charged, the compressor adopts full-power refrigeration mode; When the destination area is reached at medium power and the travel route is predicted to be charged, the compressor adopts full-power refrigeration mode; Arriving at the destination area with low battery and the trip route is predicted to be charging, the compressor operating frequency is reduced ; When the destination area is reached at high power and the travel route is predicted not to be charged, the compressor adopts full-power refrigeration mode; Arriving at the destination area with medium battery level and the trip route is predicted to not be charging, the compressor run frequency is reduced ; Compressor run frequency reduced when arriving at destination area with low charge and trip route predicted to not be charging wherein, .

10. A vehicle-mounted refrigerator temperature control system, characterized by: The system is used to perform the temperature control method of the vehicle refrigerator according to any one of claims 1-9: The data acquisition module is used to collect historical travel basic data and vehicle state data of the automobile and perform abnormal cleaning, the travel basic data includes departure coordinates, destination coordinates, departure time and arrival time, and the vehicle state data includes residual power, charging record and refrigerator power consumption parameters; The destination area definition module is used to define the reference departure area and the reference destination area according to the departure coordinates and the destination coordinates in the historical data through the density clustering algorithm, when the navigation is turned on, the end point is directly mapped to the reference destination area, and the charging pile position closest to the reference destination area is defined as the destination area; The route time estimation module is used to generate a travel route based on the reference departure area and the destination area in the historical data, to count the driving time of each travel route, and to estimate the actual speed and calculate the expected driving time according to the legal speed limit information of the road when the navigation is turned on; The charging behavior prediction module is used to use the trained classification model to output the prediction result when the navigation is not turned on, taking the current reference departure area, departure time and residual power as input features, and taking the travel route as a classification label, and to associate the average driving time of the travel route as the expected driving time, to extract the parameters strongly related to the travel charging based on the reference departure area, the destination area, the departure time and the residual power in the historical data, and to input the logistic regression model to complete the charging behavior prediction; The running mode control module is used to calculate the residual power of the vehicle after reaching the destination area according to the navigation state and the charging prediction result, the expected driving time, the refrigerator power consumption parameters and the vehicle unit mileage driving energy consumption, and to compare the residual power with the charging tolerance threshold and the non-charging energy saving threshold respectively, to divide the current travel route into multiple typical running scenarios to control the refrigerator compressor running mode.

Images