New energy automobile air conditioner multi-sensor cooperative monitoring method and system

Abstract

The invention discloses a multi-sensor cooperative monitoring method and system for an air conditioner of a new energy automobile, and relates to the technical field of cooperative monitoring. An environment situation perception and prediction model is introduced to carry out prospective situation analysis on air quality on a driving path, a dynamic cost function is constructed based on the prospective situation analysis, then the current vehicle state, the predicted environment situation and the dynamic cost function are input into an MPC optimizer, an optimal control sequence is solved in real time through a model prediction control algorithm, and the optimal control sequence is obtained. And finally, the first-step control instruction is extracted and sent to an execution mechanism, and self-adaptive regulation and control of the air conditioning system on the complex driving scene are achieved. In this way, geographic space information and the real-time state of the vehicle are deeply fused, intelligent upgrading of the air conditioning system from passive response to active pre-judgment is achieved, and therefore the intelligence and energy saving performance of the air conditioning system of the new energy vehicle are improved.

Description

Technical Field

[0001] The present application relates to the field of collaborative monitoring technology, and more specifically, to a multi-sensor collaborative monitoring method and system for new energy vehicle air conditioning. Background Art

[0002] With the booming development of the new energy vehicle industry and its continued rise in market penetration, the comfort and health of the driving experience have become core concerns for consumers. As a key component affecting in-vehicle comfort and range, the air conditioning system must precisely regulate in-vehicle temperature, humidity, and air quality in complex and changing driving scenarios (such as urban congestion, highway driving, and varying geographical regions) to balance passenger experience and energy consumption. Traditional air conditioning system control often relies on single sensor data or simple in-vehicle environmental monitoring. These systems lack comprehensive awareness of the external geographical environment (such as road type, regional air quality, altitude, etc.) and dynamic driving path information, resulting in a lack of forward-looking and adaptable control strategies. For example, when entering a road section with poor air quality or a congested area, the inability to predict and adjust the air conditioning filtration and ventilation strategies in advance will not only reduce in-vehicle comfort but also increase energy consumption due to frequent starts and stops or inefficient operation, affecting the range of new energy vehicles.

[0003] In existing technologies, air conditioning system control often only performs simple processing of in-vehicle sensor data (such as temperature, humidity, PM2.5, etc.), lacking deep integration and contextual awareness of geographic data (such as driving routes and regional environmental attributes). For example, when a vehicle travels to different geographical areas (such as industrial areas, residential areas, tunnels, etc.), it is impossible to adjust the air conditioning filter mode or ventilation intensity in advance based on the route attributes, resulting in a delayed response to the in-vehicle air quality. In addition, control strategies are mostly based on static models or simple feedback adjustments, making it difficult to optimize the balance between energy consumption and comfort in real time during dynamic driving. For example, in energy consumption priority mode, excessive restriction of air conditioning power may lead to a decrease in comfort, while in comfort priority mode, energy consumption will increase. This type of air conditioning system control strategy not only makes it difficult to achieve optimal air quality management, but also fails to find the best balance between energy consumption and comfort, making it difficult to meet the high standards of modern new energy vehicles for intelligence, energy saving, and health.

[0004] Therefore, there is an urgent need for an optimized multi-sensor collaborative monitoring method and system for new energy vehicle air conditioning. Summary of the Invention

[0005] In order to solve the above technical problems, this application is proposed.

[0006] According to one aspect of the present application, a multi-sensor collaborative monitoring method for a new energy vehicle air conditioner is provided, comprising:

[0007] Acquire a sequence of standardized status frames, where the standardized status frames include in-vehicle sensor data, out-vehicle sensor data, actuator status, and geographic data;

[0008] Extracting a geographic data sequence from the standardized state frame sequence, and acquiring a path attribute sequence by querying a geographic information database;

[0009] Inputting the geographic data sequence and the path attribute sequence into an environmental context perception and prediction model to obtain an air quality context;

[0010] constructing a dynamic cost function based on the air quality scenario;

[0011] The candidate control sequence, the current normalized state frame and the dynamic cost function are input into the MPC optimizer to obtain the optimal control sequence;

[0012] A control instruction for the first time step is extracted from the optimal control sequence and sent to the fan controller and the damper servo.

[0013] According to another aspect of the present application, a multi-sensor collaborative monitoring system for a new energy vehicle air conditioner is provided, comprising:

[0014] A state frame standardization acquisition module is used to obtain a standardized state frame sequence, which includes in-vehicle sensor data, out-vehicle sensor data, actuator status, and geographic data;

[0015] A geographic data extraction and path query module, configured to extract a geographic data sequence from the standardized state frame sequence and obtain a path attribute sequence by querying a geographic information database;

[0016] an environmental context modeling module, configured to input the geographic data sequence and the path attribute sequence into an environmental context perception and prediction model to obtain an air quality context;

[0017] A dynamic cost function generation module, configured to construct a dynamic cost function based on the air quality scenario;

[0018] An MPC optimization module is used to input the candidate control sequence, the current normalized state frame and the dynamic cost function into the MPC optimizer to obtain the optimal control sequence;

[0019] The control instruction distribution module is used to extract the control instruction of the first time step from the optimal control sequence and send it to the fan controller and the damper servo.

[0020] Compared with existing technologies, the present application provides a multi-sensor collaborative monitoring method and system for new energy vehicle air conditioning. This method acquires path attributes by real-time collection and integration of multi-dimensional sensor data and geographic location information inside and outside the vehicle. It also introduces an environmental context perception and prediction model to conduct a forward-looking contextual analysis of the air quality along the driving path. Based on this, a dynamic cost function is constructed. The current vehicle state, predicted environmental context, and dynamic cost function are then input into an MPC optimizer. The optimal control sequence is solved in real time using a model predictive control algorithm. Finally, the first-step control command is extracted and sent to the actuator, enabling the air conditioning system to adaptively control complex driving scenarios. This deep integration of geographic spatial information and the vehicle's real-time state enables an intelligent upgrade of the air conditioning system from passive response to active prediction, thereby enhancing the intelligence and energy efficiency of the new energy vehicle air conditioning system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0022] Figure 1 This is a flowchart of a multi-sensor collaborative monitoring method for a new energy vehicle air conditioner according to an embodiment of the present application.

[0023] Figure 2 This is a data flow diagram of the multi-sensor collaborative monitoring method for new energy vehicle air conditioning according to an embodiment of the present application.

[0024] Figure 3 This is a flowchart of sub-step S2 of the multi-sensor collaborative monitoring method for new energy vehicle air conditioning according to an embodiment of the present application.

[0025] Figure 4 This is a flowchart of sub-step S4 of the multi-sensor collaborative monitoring method for new energy vehicle air conditioning according to an embodiment of the present application.

[0026] Figure 5 This is a flowchart of sub-step S5 of the multi-sensor collaborative monitoring method for new energy vehicle air conditioning according to an embodiment of the present application.

[0027] Figure 6 This is a block diagram of a multi-sensor collaborative monitoring system for new energy vehicle air conditioning according to an embodiment of the present application. DETAILED DESCRIPTION

[0028] The following describes embodiments of the present disclosure in more detail with reference to the accompanying drawings. While the drawings illustrate certain embodiments of the present disclosure, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0029] In response to the problems in the above-mentioned background technology, this application proposes a multi-sensor collaborative monitoring method for new energy vehicle air conditioners. Figure 1 This is a flowchart of a multi-sensor collaborative monitoring method for a new energy vehicle air conditioner according to an embodiment of the present application. Figure 2 This is a data flow diagram of the multi-sensor collaborative monitoring method for new energy vehicle air conditioners according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the multi-sensor collaborative monitoring method for new energy vehicle air conditioning includes the following steps: S1, obtaining a standardized state frame sequence, the standardized state frame including in-vehicle sensor data, out-of-vehicle sensor data, actuator status and geographic data; S2, extracting a geographic data sequence from the standardized state frame sequence, and obtaining a path attribute sequence by querying a geographic information database; S3, inputting the geographic data sequence and the path attribute sequence into an environmental context perception and prediction model to obtain an air quality context; S4, constructing a dynamic cost function based on the air quality context; S5, inputting an alternative control sequence, a current standardized state frame and a dynamic cost function into an MPC optimizer to obtain an optimal control sequence; S6, extracting a control instruction for the first time step from the optimal control sequence, and sending it to a fan controller and a damper servo.

[0030] In the aforementioned multi-sensor collaborative monitoring method for new energy vehicle air conditioning, step S1 acquires a sequence of standardized status frames, which include in-vehicle sensor data, out-vehicle sensor data, actuator status, and geographic data. Specifically, in-vehicle sensor data includes carbon dioxide concentration, PM2.5 concentration, VOC level, in-vehicle temperature, and in-vehicle humidity; out-vehicle sensor data includes parking space PM2.5 concentration, outside temperature, and outside humidity; actuator status includes fan speed, internal and external circulation damper position, and compressor speed; and geographic data includes GPS coordinates, vehicle speed, and timestamp. It should be understood that vehicle air conditioning system control requires the integration of multi-dimensional information, including in-vehicle and external environments, the vehicle's own operating status, and driving context. Based on this, this application collects and standardizes in-vehicle sensor data, out-of-vehicle sensor data, actuator status and geographic data, realizes the effective integration and standardization of multi-source heterogeneous data, eliminates the interference caused by data noise, format differences and missing values, and forms a standardized state frame sequence containing complete environmental and status information, ensuring that the data is consistent in time sequence, unified in format and reliable in quality, providing a unified input basis for subsequent use, enabling the air-conditioning system to perform forward-looking regulation based on comprehensive and accurate information.

[0031] In particular, in a possible embodiment, the implementation process of step S1 is as follows: first, various sensors in the vehicle are activated to collect vehicle environment data in real time, such as obtaining a current vehicle carbon dioxide concentration of 380ppm through a carbon dioxide sensor and a concentration of 15μg / m 3 The VOC sensor identification level is level 1, the temperature sensor measures 24°C, and the humidity sensor measures 45%. At the same time, the external sensor is activated to collect the PM2.5 concentration outside the vehicle, which is 50μg / m 3, temperature 30°C, and humidity 60%. The current status of the air conditioner actuator is obtained, including fan speed at level 3, recirculation damper position at 40%, and compressor speed at 1800 rpm. The current GPS coordinates obtained from the vehicle's onboard GPS module are (39.9087°, 116.3975°), vehicle speed 30 km / h, and timestamp 14:30:22. The collected raw data is processed. Moving average filtering is used to remove transient fluctuations in the in-vehicle PM2.5 data. The values ​​of five adjacent moments are averaged to obtain the smoothed PM2.5 concentration. Missing values ​​from the external temperature sensor are imputed using linear interpolation based on the temperature values ​​at the preceding and following moments. All data is then calibrated to ensure uniform units of temperature and concentration. Finally, the processed in-vehicle sensor data, external sensor data, actuator status, and geographic data are linked to the same timestamp of 14:30:22 and packaged into a standardized state frame. According to the above process, the acquisition and processing operations are repeated every 10 seconds to continuously obtain standardized state frames at multiple moments to form a standardized state frame sequence.

[0032] In the aforementioned multi-sensor collaborative monitoring method for new energy vehicle air conditioners, step S2 extracts a geographic data sequence from the standardized state frame sequence and obtains a path attribute sequence by querying a geographic information database. It should be understood that a geographic information database stores path attributes associated with geographic location, such as road type, area function, and whether a tunnel is included. By extracting the geographic data sequence and querying the geographic information database, vehicle location information can be converted into environmentally meaningful path features, thus overcoming the limitations of a single sensor's perception of the external environment and providing the necessary path dimension information for subsequent environmental context analysis.

[0033] In particular, in one embodiment, Figure 3 Flowchart of sub-step S2 of the multi-sensor collaborative monitoring method for new energy vehicle air conditioner according to an embodiment of the present application. Figure 3 As shown, the step S2 includes: S21, extracting a GPS coordinate sequence from the standardized state frame sequence; S22, querying path attributes in the geographic information database based on the GPS coordinate sequence to obtain the path attribute sequence.

[0034] Specifically, step S21 extracts a GPS coordinate sequence from the standardized state frame sequence. Specifically, continuous GPS coordinates are separated from the standardized state frame sequence to obtain a time-ordered GPS coordinate sequence, accurately recording the vehicle's geographic location at different times and fully reflecting the vehicle's travel path. This provides clear and continuous location input for subsequent location-based query of path attributes, enabling the extraction of path attributes to closely match the vehicle's dynamic travel trajectory, providing a reliable geospatial foundation for subsequent analysis.

[0035] In particular, in one possible embodiment, step S21 is implemented as follows: Data segments containing GPS coordinates are separated from a continuously generated sequence of standardized state frames in chronological order, with each data segment corresponding to latitude and longitude information at a timestamp. The extracted raw GPS coordinates are preprocessed by first removing obvious outliers using a preset threshold range, such as limiting latitude to -90 to 90 degrees and longitude to -180 to 180 degrees. Data outside this range is marked as invalid and filled in using linear interpolation based on the previous and next valid coordinates. The valid coordinates are then smoothed using a sliding window averaging method, with the window size set to five consecutive state frames. The average of the coordinates within the window is calculated as the GPS coordinate at the center of the window to reduce the impact of instantaneous fluctuations. The processed GPS coordinates are arranged in chronological order by timestamp to form a GPS coordinate sequence. For example, coordinate points such as (39.9087°, 116.3975°, 14:30:22) and (39.9089°, 116.3978°, 14:30:32) are obtained in sequence. Each coordinate point precisely corresponds to the timestamp of the standardized state frame, providing continuous and reliable location information for subsequent path attribute queries.

[0036] Specifically, step S22 queries the geographic information database for path attributes based on the GPS coordinate sequence to obtain the path attribute sequence. Specifically, the present application uses the GPS coordinate sequence to match attributes corresponding to the vehicle's travel path in the geographic information database, such as road type, forward obstacles, and regional functions, to form a path attribute sequence that corresponds to the GPS coordinate sequence in time and space, including information such as road type, regional characteristics, and proximity to tunnels. The geographic location information is converted into specific environmental characteristics, providing input data including path characteristics for the environmental context perception and prediction model to accurately determine the air quality context.

[0037] In particular, in one possible embodiment, step S22 is implemented as follows: First, the GPS coordinates at each moment are extracted from the GPS coordinate sequence in chronological order, and a pre-stored geographic information database is accessed one by one through the vehicle network interface. The database stores mappings between path attributes such as road type, surrounding area function, and tunnel proximity, and GPS coordinates. The system matches the current GPS coordinates with the coordinate range in the database. When the coordinates fall within the coordinate interval of a particular road segment, the path attributes corresponding to that road segment are extracted. For example, when the GPS coordinates (39.9087°, 116.3975°) match the range of a city main road, the path attributes are obtained as an urban road with an industrial area nearby. When the next coordinates (39.9092°, 116.3981°) match near a highway entrance, the path attributes are obtained as a highway entrance and the section about to enter the highway. The path attributes retrieved at each moment are mapped one-to-one with the GPS coordinate sequence according to the timestamp, forming a path attribute sequence. This sequence is updated in real time as the vehicle travels and contains environmental characteristics of the road segment the vehicle is currently and will soon pass through, providing key input for environmental context awareness.

[0038] In the above-mentioned new energy vehicle air conditioning multi-sensor collaborative monitoring method, the step S3 inputs the geographic data sequence and the path attribute sequence into the environmental context perception and prediction model to obtain the air quality context. In a specific example of the present application, the environmental context perception and prediction model is a pre-trained gradient boosting tree, and the air quality context includes clean suburbs, urban congestion, tunnel ahead, highway, and tunnel driving. It should be understood that the environmental context perception and prediction model constructs an integrated learning framework based on multiple decision trees, and forms a strong classifier through iterative training. It includes an input layer, multiple weak classifier (decision tree) layers and an output layer. The input layer receives the feature vectors converted from the geographic data sequence and the path attribute sequence. The weak classifier layer is composed of multiple decision trees. Each decision tree is trained based on different feature subsets and data subsets, and is gradually optimized through a gradient boosting strategy. The output layer integrates the prediction results of all decision trees and outputs the air quality context category. Specifically, this application analyzes geographic data sequences and path attribute sequences through environmental context perception and prediction models to establish a nonlinear spatiotemporal mapping relationship, integrate dynamic driving status and static path characteristics, identify and predict the air quality context that the vehicle is in or will enter in the current and future period, and realize forward-looking judgment of air quality conditions in complex driving scenarios, providing direct and accurate input for subsequent processing, so that the air-conditioning system can adapt to the control needs of different air quality scenarios in advance.

[0039] In particular, in one possible embodiment, the implementation process of step S3 is as follows: first, a large amount of historical driving data is collected, covering geographical data sequences, path attribute sequences and corresponding actual air quality context labels. Then, the collected raw data is preprocessed, the moving average filtering method is used to denoise the instantaneous jump values in the GPS coordinates, the linear interpolation method is used to fill in the occasional missing values in the path attributes, and the classification features such as road types and regional functions are converted into numerical codes to ensure uniform data format and reliable data quality, forming a regular data set. Then, feature engineering is performed to extract the predicted driving trajectory features in the future short time from the geographical data sequences, such as the distance to the next road type transition point calculated based on the GPS coordinate sequence, and to construct derivative features such as the historical PM2.5 fluctuation value in the same period, the correlation between the road congestion probability and the pollution level, etc., to convert the original data into feature vectors containing multi-dimensional features. Then, the data set is divided into a training set (80%) and a test set (20%), the gradient boosting tree model is initialized, the number of trees, the maximum depth, the learning rate and other hyperparameters are set, and the feature vectors and corresponding context labels of the training set are input for training. During training, a simple decision tree is first constructed as the initial model, and then new decision trees are iteratively generated, each new tree aims to minimize the prediction error of the previous model, determines the training direction by calculating the negative gradient of the loss function, focuses on the samples that the previous model predicts inaccurately, and after each new tree is generated, it is integrated into the ensemble model through a weight coefficient, gradually reducing the overall prediction error. During training, 5-fold cross-validation is used to adjust the hyperparameters until the classification accuracy of the air quality context on the test set reaches the preset standard, and finally the trained environmental context perception and prediction model is obtained, which can be used to receive real-time geographical data sequences and path attribute sequences and output the corresponding air quality context.

[0040] In particular, in another possible preferred embodiment, for the feature vectors constructed for the environmental context perception and prediction model, due to their heterogeneous feature characteristics, that is, the features contain various heterogeneous data such as types and numerical values, which makes the model's ability to distinguish when learning the complex patterns of feature combinations decrease relative to single modal data, for example, two numerical values can be easily distinguished, and two categories can also be easily distinguished, but A category + C numerical value has no clear dimension of distinction relative to B category + D numerical value.

[0041] On the other hand, for the air quality context learned by the environmental context perception and prediction model, when determining the PM2.5 weight, the carbon dioxide weight and the energy consumption weight based thereon, it can be seen that the weight values have high distinguishability, therefore, how to optimize the distinguishability enhancement of the feature vectors to adapt the model training to the dynamic cost function construction scenario is an important improvement direction.

[0042] Specifically, the geographic data sequence and path attribute sequence are first converted into feature vectors, and based on the feature vector v i ∈V, construct the first distance matrix and the second distance matrix of the feature vector:

[0043]

[0044] Among them, v i and v j denote the i-th eigenvector and the j-th eigenvector respectively, Represents the eigenvalue of the (i, j) position in a distance matrix, Represents the eigenvalue of the (i, j) position in the two distance matrices.

[0045] Specifically, for the heterogeneous feature vectors input into the environmental context perception and prediction model, the differences between the feature vectors are captured from different dimensions. A distance matrix, for example, reflects differences in underlying scales, such as vehicle speeds, by taking the absolute differences between the elements in the feature vectors. A distance matrix, for example, reflects differences in nonlinear scales, such as the squared difference correlation of PM2.5 concentration fluctuations, by taking the square root of the squared differences of the elements. This compensates for the lack of discriminability of heterogeneous features. This provides multi-scale feature difference information for subsequent processing, enabling a more comprehensive characterization of the spatial distribution of feature vectors and laying the foundation for enhancing the pattern discrimination capabilities of heterogeneous features.

[0046] Then, for the first distance matrix and the second distance matrix, the first gating matrix M1 and the second gating matrix M2 are determined based on the comparison of their matrix values ​​with the corresponding matrices:

[0047]

[0048] Among them, ε1 represents the first distance threshold, ε2 represents the second distance threshold, and m 1(i,j) represents the eigenvalue of the (i, j) position in the first gating matrix, m 2(i,j) Represents the eigenvalue at position (i, j) in the second gating matrix.

[0049] Specifically, the distance threshold here can be taken as the mean of the corresponding one or two distance matrices. The distance threshold is used to filter the elements in the distance matrix, retaining distance values ​​greater than the threshold to highlight the significant differences between the feature vectors, and setting distance values ​​less than or equal to the threshold to 0 to filter out minor differences and noise, thereby enhancing the weight of effective distance information. This strengthens the differences between the feature vectors that have practical distinguishing significance, suppresses irrelevant noise interference, and focuses the distance information more on the parts that play a key role in distinguishing the feature patterns, thereby improving the effective information density of the distance matrix. For example, when a vehicle approaches a tunnel ahead, the path attributes indicate the tunnel ahead, and the GPS coordinates point to the entrance. The threshold is used to filter and retain the significant differences between the "tunnel geographic marker and the PM2.5 warning value", filter out the minor differences in the instantaneous fluctuations of the sensor, strengthen the feature differences between the tunnel ahead and other scenes, and improve the effective information density of the distance matrix.

[0050] Then, the first gating matrix M1 and the second gating matrix M2 are added and mapped to the eigenvector to obtain the enhanced eigenvector V', that is:

[0051]

[0052] in, Indicates adding by position point, represents matrix multiplication, and V' represents the augmented eigenvector.

[0053] Specifically, the effective distance information retained by the two gating matrices is fused, and the resulting distance features are integrated into the feature vector through matrix point addition operations. This improves the discriminability of the feature vector in the spatial distribution dimension and eliminates the lack of discrimination caused by the dependence of heterogeneous feature patterns. The resulting enhanced feature vector significantly enhances spatial distribution differences, improves the ability to distinguish patterns between heterogeneous features, and better reflects the key feature differences in different air quality scenarios, providing higher-quality input for subsequent classification processing.

[0054] Furthermore, for the enhanced feature vector V' that improves the discrimination of the spatial distribution dimension, the class decision discrimination is further improved from the perspective of category determination. That is, the enhanced feature vector is passed through a contrastive Softmax function to obtain a reconstructed feature vector:

[0055]

[0056] Among them, e represents the natural constant, v' i represents the i-th eigenvalue of the enhanced eigenvector, v' i represents the i-th eigenvalue of the reconstructed eigenvector.

[0057] Specifically, through a nonlinear transformation similar to a contrasting Softmax function, the discriminability of feature vectors at the category level is enhanced, the classification decision boundary is optimized, and the category judgment error caused by feature ambiguity is reduced, making the features more suitable for the classification requirements of environmental context perception and prediction models. The resulting reconstructed feature vectors have significantly improved category-level discriminability, and the correlation between features and air quality context categories is strengthened, reducing feature confusion between different categories and improving the effectiveness of features in classification tasks.

[0058] In this way, the spatial distance distribution of heterogeneous features is reconstructed from the eigenvalue level to eliminate the cross-feature pattern dependence of the spatial distance distribution, strengthen the spatial distribution distance of the eigenvalues, and further optimize the classification decision boundary based on the spatial distance reinforcement to improve the overall ambiguity of the eigenvalues ​​based on the classification category (that is, to improve the micro-performance of the features relative to the whole under strong discrimination certainty).

[0059] Finally, the reconstructed feature vector is input into the environmental context perception and prediction model to obtain the air quality context. This optimizes the feature vector for discriminability, allowing the training of the environmental context perception and prediction model to adapt to the dynamic cost function construction scenario. Specifically, the reconstructed feature vector optimized for discriminability improves the environmental context perception and prediction model's accuracy in identifying air quality contexts, enabling the model to more accurately capture air quality characteristics in different driving scenarios. This adapts to the high-precision context judgment requirements of the dynamic cost function construction scenario and improves the accuracy of the context classification based on weight settings.

[0060] In the above-mentioned multi-sensor collaborative monitoring method for air conditioning in new energy vehicles, step S4 constructs a dynamic cost function based on the air quality scenario. It should be understood that due to different air quality scenarios, the priorities of in-vehicle air quality control and energy consumption optimization are significantly different. For example, in urban congestion scenarios, the concentration of pollutants outside the vehicle is high, and it is necessary to prioritize ensuring that the PM2.5 and CO2 concentrations inside the vehicle are within a safe range; in highway scenarios, energy consumption has a significant impact on battery life, and it is necessary to focus on reducing the energy consumption of the air conditioning system; closed scenarios such as tunnels need to take into account both pollutant isolation and in-vehicle air circulation. Therefore, this application constructs a cost function that can dynamically adjust the PM2.5 weight, CO2 weight, and energy consumption weight according to the air quality scenario, so that the total value (J) of the cost function can accurately reflect the priority of air quality control and energy consumption optimization in the current scenario, ensure that the optimization direction is consistent with the core needs of the current scenario, realize adaptive regulation of the air conditioning system in complex driving scenarios, and improve the level of intelligence and energy saving.

[0061] In particular, in one embodiment, Figure 4 Flowchart of sub-step S4 of the multi-sensor collaborative monitoring method for new energy vehicle air conditioner according to an embodiment of the present application. Figure 4 As shown, the step S4 includes: S41, extracting a cost function template; S42, adjusting the PM2.5 weight, carbon dioxide weight and energy consumption weight based on the air quality scenario.

[0062] Specifically, in step S41, a cost function template is extracted, and the cost function template is:

[0063]

[0064] Among them, PM2.5 error PM2.5 error, is the carbon dioxide error, Power is the comprehensive operating power of the air conditioning system, W PM2.5 PM2.5 weight, is the carbon dioxide weight, W e is the energy consumption weight, and J is the total value of the cost function.

[0065] Specifically, step S42 adjusts the PM2.5 weight, carbon dioxide weight, and energy consumption weight based on the air quality scenario. In a specific embodiment, step S42 includes: when the air quality scenario is a highway, the PM2.5 weight, carbon dioxide weight, and energy consumption weight are 1.0, 1.0, and 3.0, respectively; when the air quality scenario is a clean suburban area, the PM2.5 weight, carbon dioxide weight, and energy consumption weight are 1.0, 1.0, and 1.0, respectively; when the air quality scenario is a congested city, the PM2.5 weight, carbon dioxide weight, and energy consumption weight are 50.0, 60.0, and 0.8, respectively; when the air quality scenario is a tunnel ahead, the PM2.5 weight, carbon dioxide weight, and energy consumption weight are 200.0, 5.0, and 0.5, respectively; and when the air quality scenario is driving in a tunnel, the PM2.5 weight, carbon dioxide weight, and energy consumption weight are 150.0, 120.0, and 0.5, respectively.

[0066] In particular, in one possible embodiment, the implementation process of step S4 is as follows: when the environmental situation perception and prediction model outputs the air quality situation as urban congestion, the weights are adjusted according to the settings, the PM2.5 weight is set to 50.0, the carbon dioxide weight is set to 60.0, and the energy consumption weight is set to 0.8. At this time, if the PM2.5 concentration in the car is 75μg / m 3 , target value 35μg / m 3 , the carbon dioxide concentration is 1200ppm, the target value is 800ppm, and the air conditioning power is 2.5kW, then the PM2.5error is 40μg / m 3 , CO 2error is 400ppm, and the template is substituted to calculate J = 50.0×(40) 2+60.0×(400) 2 +0.8×(2.5) 2 =50×1600+60×160000+0.8×6.25=80000+9600000+5=9680005. When the scenario changes to the tunnel ahead, the weights are adjusted to PM2.5 weight 200.0, CO2 weight 5.0, and energy consumption weight 0.5. If the relevant parameters at this time are PM2.5 concentration 40μg / m 3 , target 30μg / m 3 , CO2 concentration 900ppm, target 800ppm, power 2.2kW, then J=200.0×(10) 2 +5.0×(100) 2 +0.5×(2.2) 2 =200×100+5×10000+0.5×4.84=20000+50000+2.42=70002.42. In this way, the cost function can reflect the regulation priority in different situations through weight adjustment.

[0067] In the above-mentioned multi-sensor collaborative monitoring method for air conditioning in new energy vehicles, in step S5, the alternative control sequence, the current standardized state frame and the dynamic cost function are input into the MPC optimizer to obtain the optimal control sequence. It should be understood that the MPC optimizer includes a mathematical model of the in-vehicle environment and a numerical optimization solver, wherein the mathematical model of the in-vehicle environment is used to predict future environmental changes based on the input control sequence and the current state, and the numerical optimization solver is used to solve the optimal control sequence that minimizes the cost function under system constraints. This application uses the MPC optimizer to predict the future impact of the alternative control sequence under the current state, and then combines it with the dynamic cost function for evaluation, and solves the optimal control sequence that can balance air quality control and energy consumption optimization within the system constraints to ensure that the sequence meets the core requirements of the current air quality scenario.

[0068] In particular, in one embodiment, Figure 5 FIG. 5 is a flowchart of sub-step S5 of the multi-sensor collaborative monitoring method for new energy vehicle air conditioners according to an embodiment of the present application. Figure 5As shown, the step S5 comprises: S51, inputting the candidate control sequence and the current normalized state frame into an in-vehicle environment mathematical model of the MPC optimizer to obtain a predicted state trajectory, the predicted state trajectory comprising a predicted carbon dioxide concentration sequence, a PM2.5 concentration sequence and a predicted energy consumption sequence; S52, inputting the predicted state trajectory, system constraints and the dynamic cost function into a numerical optimization solver of the MPC optimizer to obtain the optimal control sequence, the system constraints comprising a fan speed level constraint, an inside-outside circulation damper position constraint, a compressor speed constraint, a mixed damper position constraint, a damper action rate constraint and a wind speed variation smoothness constraint.

[0069] Specifically, the step S51, inputting the candidate control sequence and the current normalized state frame into an in-vehicle environment mathematical model of the MPC optimizer to obtain a predicted state trajectory, the predicted state trajectory comprising a predicted carbon dioxide concentration sequence, a PM2.5 concentration sequence and a predicted energy consumption sequence. It can be understood that the in-vehicle environment mathematical model can quantify the influence of different candidate control sequences on future in-vehicle CO2 concentration, PM2.5 concentration and energy consumption. By the in-vehicle environment mathematical model, the candidate control sequence and the current normalized state frame are converted into a predicted state trajectory for a future period of time, providing a concrete evaluation object for subsequent selection of the optimal control sequence based on the dynamic cost function, ensuring that the optimization process has a basis.

[0070] In particular, in one possible embodiment, the implementation process of the step S51 is as follows: the candidate control sequence comprises fan speed, inside-outside circulation damper position and compressor speed for the next 5 time steps, for example, (3 levels, 30%, 1800 rpm), (4 levels, 20%, 2000 rpm), (4 levels, 10%, 2200 rpm), (5 levels, 0%, 2400 rpm), (5 levels, 0%, 2500 rpm). In the current normalized state frame, the in-vehicle carbon dioxide concentration is 380 ppm, the PM2.5 concentration is 15 μg / m 3 , the in-vehicle temperature is 24℃, the in-vehicle PM2.5 concentration is 15 μg / m 3 , the outside temperature is 30℃, the fan speed is 3 levels, the inside-outside circulation damper position is 40%, the outside circulation proportion is 40%, and the compressor speed is 1800 rpm. After inputting the candidate control sequence and the current normalized state frame into the in-vehicle environment mathematical model, the fan speed is maintained at 3 levels (the ventilation amount is stable) in the first time step, the inside-outside circulation damper position is reduced to 30%, the outside circulation proportion is reduced, the introduction of high PM2.5 outside air is reduced, combined with the outside PM2.5 concentration of 50 μg / m 3 , the predicted in-vehicle PM2.5 concentration is reduced to 12 μg / m 3The carbon dioxide concentration dropped slightly to 370 ppm due to ventilation, and the energy consumption was calculated to be 1.2 kW based on the compressor speed. In the second time step, the fan speed was increased to level 4 (ventilation volume increased), the damper position was 20%, and the external circulation was further reduced. The PM2.5 concentration was predicted to drop to 10 μg / m 3 , the carbon dioxide concentration dropped to 360ppm, and the energy consumption increased to 1.4kW. Subsequent time steps were calculated in sequence according to this logic, and finally the predicted carbon dioxide concentration sequence (370ppm, 360ppm, 350ppm, 340ppm, 330ppm), PM2.5 concentration sequence (12μg / m 3 , 10 μg / m 3 , 8μg / m 3 , 6μg / m 3 , 5μg / m 3 ) and the predicted state trajectory of the predicted energy consumption sequence (1.2kW, 1.4kW, 1.6kW, 1.8kW, 2.0kW).

[0071] Specifically, in step S52, the predicted state trajectory, system constraints and the dynamic cost function are input into the numerical optimization solver of the MPC optimizer to obtain the optimal control sequence, wherein the system constraints include fan speed gear constraints, internal and external circulation damper position constraints, compressor speed constraints, mixing damper position constraints, damper action rate constraints and wind speed change smoothness constraints. It should be understood that the specific mathematical optimization algorithm adopted by the numerical optimization solver is the sequential quadratic programming (SQP) algorithm, which can efficiently solve the optimal control sequence with constraints when dealing with optimization problems containing nonlinear constraints and nonlinear objective functions. Specifically, the present application evaluates the predicted state trajectory through a numerical optimization solver, and calculates the cost function value of each alternative control sequence in combination with the dynamic cost function, such as when the city is congested, the PM2.5 weight is 50.0, the CO2 weight is 60.0, and the energy consumption weight is 0.8, and the control sequence with the minimum total cost is selected as the optimal control sequence to ensure that it meets the hardware operation restrictions and meets the optimization goals of the current situation.

[0072] Specifically, in one possible embodiment, step S52 is implemented as follows: the predicted state trajectory includes CO2 concentration, PM2.5 concentration, and energy consumption data for the next five time steps. System constraints include fan speed limits of 1-5, internal and external circulation damper positions of 0-100%, compressor speed limits of 1000-3000 rpm, damper actuation rates of no more than 20% / s, and smooth wind speed variations requiring speed differences of no more than one level between adjacent time steps. A dynamic cost function is weighted based on the current air quality scenario, i.e., urban congestion, where PM2.5 is weighted 50.0, CO2 is weighted 60.0, and energy consumption is weighted 0.8. After receiving these inputs, the numerical optimization solver evaluates the alternative control sequences and calculates the corresponding cost function value for each sequence. For one alternative sequence, the PM2.5 concentration error in its predicted trajectory is large, resulting in a high cost function value. Another alternative sequence, while having lower energy consumption, exhibits a slow decline in CO2 concentration, resulting in an unsatisfactory cost function value. The solver finds the control sequence that minimizes the cost function value through iterative optimization while satisfying all system constraints: in the first time step, the fan speed is level 3, the internal and external circulation damper position is 35%, and the compressor speed is 1900rpm; in the second time step, the fan speed is level 3, the damper position is 30%, and the compressor speed is 2000rpm; in the third time step, the fan speed is level 4, the damper position is 25%, and the compressor speed is 2100rpm; in the fourth time step, the fan speed is level 4, the damper position is 20%, and the compressor speed is 2200rpm; in the fifth time step, the fan speed is level 5, the damper position is 15%, and the compressor speed is 2300rpm. This sequence is the optimal control sequence.

[0073] In the above-mentioned multi-sensor collaborative monitoring method for air conditioning in new energy vehicles, the step S6 extracts the control instructions for the first time step from the optimal control sequence and sends them to the fan controller and the damper servo. It should be understood that extracting the control instructions for the first time step can ensure the accuracy of the current regulation based on the latest real-time status, while reserving adjustment space for the next round of optimization, avoiding control lags caused by the accumulation of prediction errors, and adapting to the uncertainty of dynamic driving scenarios. Specifically, the present application sends the immediate control instructions that best match the current state in the optimal control sequence to the actuator, so that the air conditioning system immediately responds to the current optimization results and realizes immediate regulation of the vehicle interior environment. At the same time, relying on the rolling optimization characteristics of MPC, it provides a basis for subsequently resolving the optimal control sequence based on the new state, thereby ensuring the continuity and real-time performance of the control strategy.

[0074] In particular, in one possible embodiment, step S6 is implemented as follows: the optimal control sequence is a set of control instructions for the next five time steps, obtained by the MPC optimizer. The system extracts the control instruction for the first time step with the earliest timestamp from this sequence, namely, the instruction for fan speed level 3, internal / external recirculation damper position 35%, and compressor speed 1900 rpm, corresponding to 14:30:32. The system uses the instruction parsing module to separate parameters related to the fan controller, damper servo, and compressor controller. Subsequently, the control signal conversion module converts the extracted parameters into signals recognizable by the actuator: according to a preset mapping relationship, fan speed level 3 corresponds to a 12V DC voltage. After processing by the power module within the fan controller, this voltage can drive the fan to the actuator state within the 3-step speed range; internal / external recirculation damper position 35% corresponds to a PWM signal with a frequency of 50 Hz and a duty cycle of 35%, which meets the control signal requirements of the damper servo; and a compressor speed of 1900 rpm corresponds to a 4-20 mA current signal, which meets the input standards of the compressor controller. The converted electrical signals are transmitted via the vehicle's CAN bus, with the transmission frame conforming to the vehicle communication protocol. The fan control signal frame ID is set to 0x18F00100, and the data segment contains the hexadecimal code 0x0C00, corresponding to 12V. The damper control signal frame ID is set to 0x18F00200, and the data segment contains the hexadecimal code 0x2300, corresponding to a 35% duty cycle. The compressor control signal frame ID is set to 0x18F00300, and the data segment contains the hexadecimal code 0x0764, corresponding to 1900 rpm. After receiving the CAN signal, the fan controller uses its internal voltage regulation circuit to stabilize the input voltage at 12V, driving the fan motor. Simultaneously, a speed sensor monitors the speed in real time. If the speed deviates from the target value corresponding to the three levels, closed-loop control is used to correct the output until the speed stabilizes within the three speed ranges specified by the actuator status. After receiving the PWM signal, the damper servo drives the damper blades through a reduction gear mechanism. An internal position sensor converts the real-time position into a feedback signal, which is compared with the control signal. When the feedback position reaches 35%, the damper stops, ensuring accurate damper position. After receiving the current signal, the compressor controller adjusts the compressor motor speed through the frequency conversion module and, in conjunction with the speed feedback sensor, implements closed-loop control, stabilizing the speed at 1900 rpm. During execution, the actuator status acquisition module simultaneously acquires the current fan speed, the actual damper position, and the real-time compressor speed. This data is encapsulated as a status frame and transmitted back to the main controller via the CAN bus, confirming that the first time-step control command has been executed and that all parameter errors are within the preset range, completing the closed-loop process from command extraction to execution.

[0075] In summary, the multi-sensor collaborative monitoring method for new energy vehicle air conditioning based on the embodiment of the present application is explained. It obtains path attributes by real-time collection and integration of multi-dimensional sensor data and geographic location information inside and outside the vehicle, and introduces an environmental context perception and prediction model to conduct a forward-looking context analysis of the air quality along the driving path. Based on this, a dynamic cost function is constructed. The current vehicle state, the predicted environmental context, and the dynamic cost function are then input into the MPC optimizer. The optimal control sequence is solved in real time through the model predictive control algorithm. Finally, the first-step control instruction is extracted and sent to the actuator to realize the adaptive regulation of the air conditioning system for complex driving scenarios. In this way, the deep integration of geographic spatial information and the real-time status of the vehicle realizes the intelligent upgrade of the air conditioning system from passive response to active prediction, thereby improving the intelligence and energy saving of the air conditioning system of new energy vehicles.

[0076] Figure 6 FIG is a block diagram of a multi-sensor collaborative monitoring system for a new energy vehicle air conditioner according to an embodiment of the present application. Figure 6 As shown, according to an embodiment of the present application, the multi-sensor collaborative monitoring system 100 for the air conditioner of a new energy vehicle includes: a state frame standardization acquisition module 110, which is used to obtain a standardized state frame sequence, and the standardized state frame includes in-vehicle sensor data, out-of-vehicle sensor data, actuator status and geographic data; a geographic data extraction and path query module 120, which is used to extract a geographic data sequence from the standardized state frame sequence, and obtain a path attribute sequence by querying a geographic information database; an environmental scenario modeling module 130, which is used to input the geographic data sequence and the path attribute sequence into an environmental scenario perception and prediction model to obtain an air quality scenario; a dynamic cost function generation module 140, which is used to construct a dynamic cost function based on the air quality scenario; an MPC optimization module 150, which is used to input an alternative control sequence, a current standardized state frame and a dynamic cost function into an MPC optimizer to obtain an optimal control sequence; a control instruction distribution module 160, which is used to extract the control instruction of the first time step from the optimal control sequence and send it to the fan controller and the damper servo.

[0077] As described above, the new energy vehicle air conditioning multi-sensor collaborative monitoring system 100 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a new energy vehicle air conditioning multi-sensor collaborative monitoring algorithm. In one possible implementation, the new energy vehicle air conditioning multi-sensor collaborative monitoring system 100 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the new energy vehicle air conditioning multi-sensor collaborative monitoring system 100 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the new energy vehicle air conditioning multi-sensor collaborative monitoring system 100 can also be one of the many hardware modules of the wireless terminal.

[0078] Alternatively, in another example, the new energy vehicle air conditioning multi-sensor collaborative monitoring system 100 and the wireless terminal may also be separate devices, and the new energy vehicle air conditioning multi-sensor collaborative monitoring system 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0079] Here, those skilled in the art will appreciate that the specific operations of each step in the above-mentioned new energy vehicle air conditioning multi-sensor collaborative monitoring system have been referenced above. Figures 1 to 5 The description of the multi-sensor collaborative monitoring method for new energy vehicle air conditioners has been introduced in detail, and therefore, its repeated description will be omitted.

Claims

1. A multi-sensor collaborative monitoring method for new energy vehicle air conditioners, characterized in that: include: Acquire a sequence of standardized status frames, where the standardized status frames include in-vehicle sensor data, out-vehicle sensor data, actuator status, and geographic data; Extracting a geographic data sequence from the standardized state frame sequence, and acquiring a path attribute sequence by querying a geographic information database; Inputting the geographic data sequence and the path attribute sequence into an environmental situation perception and prediction model to obtain an air quality situation; constructing a dynamic cost function based on the air quality scenario; The candidate control sequence, the current normalized state frame and the dynamic cost function are input into the MPC optimizer to obtain the optimal control sequence; A control instruction for the first time step is extracted from the optimal control sequence and sent to the fan controller and the damper servo.

2. The multi-sensor collaborative monitoring method for new energy vehicle air conditioner according to claim 1 is characterized in that: The in-vehicle sensor data includes carbon dioxide concentration, PM2.5 concentration, VOC level, in-vehicle temperature and in-vehicle humidity; the out-vehicle sensor data includes parking space PM2.5 concentration, outside temperature and outside humidity; the actuator status includes fan speed, internal and external circulation damper position and compressor speed; the geographic data includes GPS coordinates, vehicle speed and timestamp.

3. The multi-sensor collaborative monitoring method for new energy vehicle air conditioner according to claim 2 is characterized in that: Extracting a geographic data sequence from the standardized state frame sequence and acquiring a path attribute sequence by querying a geographic information database includes: extracting a GPS coordinate sequence from the standardized state frame sequence; The path attributes are queried in the geographic information database based on the GPS coordinate sequence to obtain the path attribute sequence.

4. The multi-sensor collaborative monitoring method for new energy vehicle air conditioner according to claim 3 is characterized in that: The environmental context perception and prediction model is a pre-trained gradient boosting tree, and the air quality scenarios include clean suburbs, urban congestion, tunnel ahead, highway, and tunnel driving.

5. The multi-sensor collaborative monitoring method for new energy vehicle air conditioner according to claim 1, characterized in that: Based on the air quality scenario, a dynamic cost function is constructed, including: Extract the cost function template, which is: Among them, PM2.5 error PM2.5 error, is the carbon dioxide error, Power is the comprehensive operating power of the air conditioning system, W PM2.5 PM2.5 weight, is the carbon dioxide weight, W e is the energy consumption weight, J is the total value of the cost function; Based on the air quality scenario, the PM2.5 weight, carbon dioxide weight, and energy consumption weight are adjusted.

6. The multi-sensor collaborative monitoring method for new energy vehicle air conditioner according to claim 5 is characterized in that: Based on the air quality scenario, adjust the PM2.5 weight, CO2 weight, and energy consumption weight, including: When the air quality scenario is a highway, the PM2.5 weight, carbon dioxide weight, and energy consumption weight are 1.0, 1.0, and 3.0, respectively; When the air quality scenario is a clean suburban area, the PM2.5 weight, carbon dioxide weight, and energy consumption weight are 1.0, 1.0, and 1.0, respectively; When the air quality scenario is urban congestion, the PM2.5 weight, carbon dioxide weight, and energy consumption weight are 50.0, 60.0, and 0.8, respectively; When the air quality scenario is the tunnel ahead, the PM2.5 weight, carbon dioxide weight, and energy consumption weight are 200.0, 5.0, and 0.5, respectively; When the air quality scenario is driving in a tunnel, the PM2.5 weight, carbon dioxide weight, and energy consumption weight are 150.0, 120.0, and 0.5, respectively.

7. The multi-sensor collaborative monitoring method for new energy vehicle air conditioner according to claim 1, characterized in that: The candidate control sequence, the current normalized state frame, and the dynamic cost function are input into the MPC optimizer to obtain the optimal control sequence, including: Inputting the alternative control sequence and the current standardized state frame into the in-vehicle environment mathematical model of the MPC optimizer to obtain a predicted state trajectory, the predicted state trajectory including a predicted carbon dioxide concentration sequence, a PM2.5 concentration sequence, and a predicted energy consumption sequence; The predicted state trajectory, system constraints and the dynamic cost function are input into the numerical optimization solver of the MPC optimizer to obtain the optimal control sequence, wherein the system constraints include fan speed gear constraints, internal and external circulation damper position constraints, compressor speed constraints, mixing damper position constraints, damper action rate constraints and wind speed change smoothness constraints.

8. A new energy vehicle air conditioning multi-sensor collaborative monitoring system, characterized in that: include: A state frame standardization acquisition module is used to obtain a standardized state frame sequence, which includes in-vehicle sensor data, out-vehicle sensor data, actuator status, and geographic data; A geographic data extraction and path query module, configured to extract a geographic data sequence from the standardized state frame sequence and obtain a path attribute sequence by querying a geographic information database; an environmental context modeling module, configured to input the geographic data sequence and the path attribute sequence into an environmental context perception and prediction model to obtain an air quality context; A dynamic cost function generation module, configured to construct a dynamic cost function based on the air quality scenario; An MPC optimization module is used to input the candidate control sequence, the current normalized state frame and the dynamic cost function into the MPC optimizer to obtain the optimal control sequence; The control instruction distribution module is used to extract the control instruction of the first time step from the optimal control sequence and send it to the fan controller and the damper servo.

Images