Automobile air pump power supply control method, system and automobile air pump

Abstract

The present invention relates to the field of automobile air pump control technology, and discloses a method, system, and automobile air pump power supply control method. The method collects multi-dimensional power supply data of the automobile air pump, extracts dynamic feature vectors, and uses a convolutional neural network to generate a global correlation feature matrix. A power mode knowledge base is constructed, and a genetic algorithm is combined to generate a power control strategy sequence and resource allocation plan. An improved particle swarm optimization algorithm is also used to optimize the strategy, anomalies are detected and the strategy is reconstructed based on a fuzzy logic system, and a knowledge graph is constructed to verify the physical compatibility of the strategy. The present invention can accurately analyze the power supply status, efficiently generate optimization strategies, quickly handle abnormal situations, improve the operating stability and inflation efficiency of the automobile air pump, extend the battery life, and reduce energy consumption.

Description

Technical Field

[0001] The present invention relates to the technical field of automobile accessory automobile air pump control, in particular to an automobile air pump power supply control method and system and an automobile air pump. Background Art

[0002] In daily vehicle use, stable tire pressure is crucial for driving safety and vehicle performance. As a key device for maintaining tire pressure, the performance of automotive air pumps directly impacts the user experience and vehicle safety. However, current automotive air pumps suffer from numerous power supply control issues, severely hindering their further development and application.

[0003] Early automotive air pump power supply control methods were relatively simple, typically relying solely on basic voltage and current sensing. These methods failed to fully account for the complex operating conditions of the air pump. For example, varying ambient temperature and humidity can affect the performance of the pump's internal electronic components, impacting power supply stability and pump efficiency. However, traditional control methods often overlook these factors, leading to air pump failure and shortened service life in high-temperature and high-humidity environments.

[0004] While some air pumps are beginning to focus on more operating parameters with technological advancements, deficiencies remain in data processing and control strategy development. Existing air pumps are unable to effectively extract valuable feature information when collecting multi-dimensional power supply data, resulting in inaccurate judgments about power supply status. For example, simply collecting voltage and current data without in-depth analysis of power supply ripple spectrum characteristics or load fluctuation detection features makes it difficult to quickly and correctly respond to abnormal power supply fluctuations, impacting the normal operation of the air pump.

[0005] Furthermore, current power control strategies lack effective utilization of historical data. When encountering similar power usage or faults, past experience cannot be quickly applied to address them. Instead, different control methods are repeatedly tried, wasting time and potentially leading to more serious problems due to untimely response. For example, when poor battery health affects inflation efficiency, without historical optimization strategies, power output can only be adjusted blindly, failing to achieve optimal matching control between the battery and the inflation pump, further increasing energy consumption.

[0006] Furthermore, existing automotive air pump power control systems suffer from irrational control strategies when dealing with complex hardware topologies. Energy transfer constraints and hardware resource conflicts between various power control operations are not fully considered. This results in component coordination issues during operation, reducing the overall performance of the air pump. For example, motor speed control and battery charging operations may interfere with each other due to resource conflicts, affecting the efficiency and stability of the air pump. Summary of the Invention

[0007] The object of the present invention is to provide a method and system for controlling the power supply of an automobile air pump and an automobile air pump to solve the problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solution: a power supply control method for an automobile air pump, the method comprising:

[0009] Collect multi-dimensional power supply data of automobile air pumps;

[0010] Extracting signal features from the multi-dimensional power supply data to generate dynamic feature vectors of each dimension;

[0011] Mapping the dynamic feature vector to a joint analysis space based on a pre-trained convolutional neural network model to generate a global correlation feature matrix;

[0012] Building a power mode knowledge base that stores historical power usage patterns, failure cases, optimization strategies, and corresponding execution efficiency data;

[0013] According to the global correlation feature matrix and the historical patterns in the knowledge base, a power control strategy sequence and a resource allocation scheme of the current air pump are generated through a genetic algorithm.

[0014] Preferably, the multi-dimensional power supply data includes voltage, current, ambient temperature, humidity, inflation rate, battery health status, user preset pressure value, motor speed and radiator working status;

[0015] The dynamic feature vector includes power ripple spectrum characteristics, load fluctuation detection characteristics, temperature gradient change characteristics, battery internal resistance dynamic distribution characteristics and motor torque response delay characteristics.

[0016] Preferably, the generating of the power supply control strategy sequence and resource allocation scheme of the current air pump by using a genetic algorithm includes:

[0017] generating an initial candidate control strategy set according to the matching degree between the global correlation feature matrix and the fault mode of the knowledge base;

[0018] Building a policy execution dependency graph based on the air pump hardware topology, wherein nodes in the policy execution dependency graph represent power control operations, and edges represent energy transfer constraints or hardware resource conflicts between operations;

[0019] Combined with real-time power supply data, the initial candidate control strategy set is iteratively optimized through an improved particle swarm optimization algorithm to output the final control sequence and resource allocation results.

[0020] Preferably, the convolutional neural network model includes:

[0021] Use multi-scale convolutional layers to extract power signal features in different frequency bands;

[0022] The features of each frequency band are weighted and fused through the channel attention mechanism to generate the global correlation feature matrix.

[0023] Preferably, the improved particle swarm optimization algorithm includes:

[0024] Introducing an adaptive inertia weight adjustment strategy, where the weight is dynamically updated based on the real-time error in the inflation efficiency and the battery energy consumption rate;

[0025] An elite retention mechanism is used to screen the non-inferior solution set in the iterative process, and the global search capability is enhanced through the Cauchy mutation operator.

[0026] Preferably, the signal feature extraction adopts a wavelet transform and time-frequency joint analysis method, specifically including:

[0027] The power signal is decomposed by Morlet wavelet basis function to extract the energy entropy characteristics at each scale;

[0028] The time-frequency matrix is ​​subjected to singular value decomposition to generate dynamic eigenvectors that characterize the signal's sudden changes and steady-state components.

[0029] Preferably, the method further comprises:

[0030] Detect abnormal patterns in real-time power data based on a fuzzy logic system;

[0031] When an anomaly is detected, dynamic reconstruction of the control strategy sequence is triggered, and the resource allocation scheme is updated according to the reconstruction results.

[0032] Preferably, the dynamic reconstruction uses a mixed integer linear programming solver, specifically including:

[0033] Define hard constraints on power supply operation sequence constraints, battery charge and discharge cycle limits, and radiator start and stop time windows;

[0034] A feasible solution set is generated through a column generation algorithm, and the comprehensive optimal solution is screened based on the entropy weight TOPSIS method.

[0035] Preferably, the method further comprises:

[0036] Construct a knowledge graph of power supply component relationships, where nodes represent electrical components of the air pump or fault types, and edges represent energy coupling relationships between components or fault conduction paths;

[0037] In the strategy sequence generation phase, the physical compatibility of the strategy with the knowledge graph is verified through the subgraph isomorphism detection algorithm.

[0038] Preferably, the present invention further includes a power supply control system for an automobile air pump, the system comprising:

[0039] Data acquisition module: used to collect multi-dimensional power supply data of automobile air pumps;

[0040] Feature extraction module: extracts signal features from the multi-dimensional power supply data to generate dynamic feature vectors of each dimension;

[0041] Feature association module: maps the dynamic feature vector to the joint analysis space based on the pre-trained convolutional neural network model to generate a global correlation feature matrix;

[0042] Knowledge base construction module: constructs a power mode knowledge base, which stores historical power usage patterns, failure cases, optimization strategies and corresponding execution efficiency data;

[0043] Strategy generation module: generates a power control strategy sequence and a resource allocation plan for the current air pump through a genetic algorithm according to the global correlation feature matrix and the historical patterns in the knowledge base.

[0044] A car air pump includes an air pump housing body, an air pump assembly and a mobile power supply. The air pump housing body is provided with an assembly slot for the mobile power supply to be movably arranged therein. The assembly slot is provided with an electrical interface terminal electrically connected to the air pump assembly. The mobile power supply is electrically connected to the air pump assembly through the electrical interface terminal. A spring is provided in the assembly slot, one end of the spring is fixed to the bottom of the assembly slot, and the other end is in contact with the mobile power supply. A control button is provided on the surface of the air pump housing body, and the bottom end of the control button is connected to a button bracket in contact with the spring.

[0045] One side of the air pump assembly is connected to a fan assembly, a side wall of the air pump housing body is provided with an air inlet corresponding to the fan assembly, and the other side of the air pump housing body is provided with an air outlet connected to the air pump assembly.

[0046] An illuminating lamp is embedded on the surface of the main body of the air pump housing, and an openable and closable sealing baffle is provided at the opening of the assembly groove.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] At the data acquisition and analysis level, by collecting multi-dimensional power supply data from the car air pump, covering information such as voltage, current, ambient temperature, humidity, inflation rate, battery health status, user-preset pressure value, motor speed, and radiator operating status, a rich data foundation is provided for a comprehensive understanding of the air pump's operating status. Wavelet transform and time-frequency joint analysis methods are used to extract signal features from this data, generating dynamic feature vectors of various dimensions such as power ripple spectrum characteristics and load fluctuation detection characteristics, which can more accurately grasp the real-time status changes of the power supply. This makes the system more sensitive to power supply anomalies. Compared with traditional methods that rely only on basic voltage and current detection, it can detect potential problems in advance, greatly improving the reliability of the air pump's operation and reducing the probability of failures caused by power supply problems.

[0049] A pre-trained convolutional neural network model maps dynamic feature vectors to a joint analysis space, generating a global correlation feature matrix that further explores the underlying connections between the data. This matrix comprehensively reflects the global correlations among features across all dimensions, providing strong support for the subsequent development of precise control strategies. This approach enables the system to more accurately predict power status trends and make proactive adjustments, ensuring the air pump always operates in optimal condition and improving inflation efficiency.

[0050] In terms of control strategy generation, the constructed power mode knowledge base stores historical power usage patterns, fault cases, optimization strategies, and corresponding execution efficiency data. Combining the global correlation feature matrix with the historical patterns in the knowledge base, a genetic algorithm is used to generate the power control strategy sequence and resource allocation plan for the current air pump. This process fully draws on past experience, avoids blind trial and error, and greatly shortens the time required to formulate control strategies. Moreover, the generated strategy can be dynamically adjusted according to the real-time status of the air pump, achieving refined power management. For example, when the battery health is poor, the charging and discharging strategies can be automatically adjusted, ensuring the normal operation of the air pump, extending the battery life, and reducing energy consumption.

[0051] The application of an improved particle swarm optimization algorithm further optimizes the control strategy generation process by introducing an adaptive inertia weight adjustment strategy, an elite retention mechanism, and a Cauchy mutation operator. The adaptive inertia weight is dynamically updated based on the real-time error in inflation efficiency and the battery energy consumption rate, enabling the algorithm to flexibly adjust its search strategy based on actual conditions when searching for the optimal solution, improving the algorithm's convergence speed and solution quality. The elite retention mechanism ensures that excellent solutions are not lost during the iteration process, and the Cauchy mutation operator enhances global search capabilities, preventing the algorithm from being trapped in local optimal solutions. This makes the generated control strategy and resource allocation plan more reasonable, effectively improving the overall performance of the air pump.

[0052] Furthermore, a fuzzy logic system detects abnormal patterns in real-time power data. When an anomaly is detected, a mixed-integer linear programming solver triggers dynamic reconstruction of the control strategy sequence and updates the resource allocation plan. This exception handling mechanism rapidly responds to power anomalies, minimizing their impact on the air pump's operation and ensuring its stable operation. The constructed knowledge graph of power component relationships and the verification of the physical compatibility of the strategy using a subgraph isomorphism detection algorithm during the strategy sequence generation phase ensure that the control strategy conforms to the hardware structure and physical principles of the air pump, avoiding hardware damage or performance degradation caused by unreasonable strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a working principle diagram of the power supply control method of the automobile air pump according to the present invention;

[0054] Figure 2 Schematic diagram of real-time power data anomaly detection and control strategy reconstruction;

[0055] Figure 3 Schematic diagram for dynamic reconstruction solution and optimization;

[0056] Figure 4 Schematic diagram for power component relationship knowledge graph construction and strategy verification;

[0057] Figure 5 This is a schematic diagram of the structure of a car air pump;

[0058] Figure 6 This is a schematic diagram of the internal structure of a car charging pump.

[0059] In the figure: 1. Air pump housing body; 2. Air pump assembly; 3. Mobile power supply; 4. Sealing baffle; 5. Control button; 6. Assembly slot; 7. Shrapnel; 8. Fan assembly; 9. Air inlet; 10. Air outlet; 11. Light; 12. Electrical interface terminal. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0061] See also Figures 1-6 The present invention provides a power supply control method for an automobile air pump, which specifically includes the following steps:

[0062] Using a variety of sensors and detection devices, various power-related data of the automobile air pump during operation are obtained in real time. This data involves multiple dimensions and covers various status information of the power supply when the air pump is working, providing basic data support for subsequent precise analysis and control.

[0063] By using specific signal processing technology, representative features are extracted from the collected multi-dimensional power supply data. These features are integrated to generate dynamic feature vectors of each dimension to more effectively reflect the changing trend of the power supply status.

[0064] Using a convolutional neural network model that has been trained with a large amount of data, the generated dynamic feature vectors are processed and mapped to a joint analysis space, in which a feature matrix is ​​generated that can reflect the global correlation between features of each dimension, thereby mining deep connections in the data.

[0065] Establish a power mode knowledge base to store historical power usage patterns, past failure cases, optimization strategies for different situations, and the execution efficiency data corresponding to these strategies. This knowledge base is an important reference for the subsequent generation of power control strategies.

[0066] The generated global correlation feature matrix is ​​compared and analyzed with the historical patterns in the knowledge base. With the help of genetic algorithm and combined with the actual situation of the current air pump, a set of power control strategy sequences and corresponding resource allocation plans suitable for the current working conditions are generated to achieve optimal control of the power supply of the automobile air pump and improve the working efficiency and performance of the air pump.

[0067] The present invention will be further described below in conjunction with Examples 1 to 6:

[0068] Example 1:

[0069] In actual operation, collecting multi-dimensional power supply data is crucial. Voltage data can be collected using a high-precision voltage sensor, which monitors the input and output voltages of the air pump in real time, providing a basis for determining power supply stability. Current data is acquired using a current transformer, which accurately measures the current in the circuit, allowing for an understanding of the air pump's power consumption during operation.

[0070] Ambient temperature and humidity are collected using temperature and humidity sensors installed in appropriate locations on the air pump. These sensors can sense changes in the surrounding temperature and humidity in real time. This is because ambient temperature and humidity can affect the performance of the air pump and the stability of the power supply. For example, in high temperature and high humidity environments, the performance of electronic components may decline, leading to power supply anomalies.

[0071] The inflation rate is a key indicator of the air pump's efficiency. By installing a flow sensor on the inflation pipe, the volume of gas injected per unit time can be accurately measured, thereby determining the inflation rate. Battery health is monitored by the Battery Management System (BMS). The BMS measures battery voltage, current, temperature, and charge and discharge cycles, then uses a specific algorithm to estimate key information such as remaining capacity and internal resistance, thereby assessing battery health.

[0072] The user-defined pressure value is the inflation target value set by the user based on actual needs. It can be entered through the operation panel on the air pump or a connected mobile device. The motor speed is detected by a speed sensor installed on the motor shaft. Changes in speed reflect the motor's operating status and load. The radiator's operating status is determined by monitoring the radiator's fan speed, temperature, and other information to determine whether the radiator is functioning properly and effectively dissipating heat.

[0073] Signal feature extraction is performed on this collected data. For example, using the power ripple spectrum as an example, a specific signal processing algorithm is used to analyze the voltage signal to obtain the frequency components and amplitude distribution of the power ripple. The load fluctuation detection feature monitors and analyzes changes in the current signal to determine load stability and promptly capture relevant characteristic information when load fluctuations occur.

[0074] The temperature gradient variation feature is extracted by performing time series analysis on the temperature data collected by the temperature and humidity sensors, calculating the rate of change of temperature at adjacent moments, and obtaining the temperature gradient. The dynamic distribution feature of the battery internal resistance utilizes data provided by the battery management system, combined with relevant algorithms, to analyze how the battery internal resistance changes under different charge and discharge conditions. The motor torque response delay feature is generated by comparing the motor control signal with the actual speed change, measuring the delay time of the motor torque response.

[0075] By comprehensively collecting multi-dimensional power supply data and extracting targeted features, we generate dynamic feature vectors for each dimension, providing rich and valuable information for subsequent power supply control strategy development. These feature vectors more accurately reflect the actual operating status of the air pump power supply, helping to achieve precise power supply control and improve the overall performance and reliability of the air pump.

[0076] Example 2:

[0077] When generating the power control strategy sequence and resource allocation plan for the current air pump, the genetic algorithm first generates an initial set of candidate control strategies based on the degree of matching between the global correlation feature matrix and the fault mode in the knowledge base. Each feature in the global correlation feature matrix is ​​then compared with the fault mode features stored in the knowledge base, and the matching degree is calculated. For example, if a feature vector in the global correlation feature matrix is ​​highly similar to a feature vector in a particular fault mode across multiple dimensions, the global correlation feature matrix is ​​considered to have a high degree of matching with that fault mode.

[0078] The fault modes are sorted from high to low according to the matching degree, and the historical control strategies corresponding to the fault modes with higher matching degrees are selected as the initial candidate control strategy set. These initial candidate strategies are based on the strategies adopted in similar situations that have occurred in the past and have certain reference value.

[0079] Next, a policy execution dependency graph was constructed based on the air pump hardware topology. Within the air pump hardware system, different power control operations face energy transfer constraints and hardware resource conflicts. For example, motor speed regulation may affect the battery's discharge current, which in turn affects the battery's energy consumption, reflecting an energy transfer constraint. Furthermore, certain control operations may require the use of the same hardware resource simultaneously. For example, multiple control signals may need to be output through the same microcontroller pin, creating a hardware resource conflict.

[0080] In the policy execution dependency graph, each power control operation is represented as a node, and edges are used to represent energy transfer constraints or hardware resource conflicts between operations. This method clearly demonstrates the interrelationships between different operations and provides an intuitive basis for subsequent optimization strategies.

[0081] Combined with real-time power data, the candidate strategy set is iteratively optimized by the improved particle swarm optimization algorithm. The improved particle swarm optimization algorithm introduces an adaptive inertia weight adjustment strategy, which is dynamically updated according to the real-time error of the charging efficiency and the battery energy consumption rate. Assume that the real-time error of the charging efficiency is , which represents the difference between the current actual inflation efficiency and the target inflation efficiency; the battery energy consumption rate is , which represents the proportion of battery energy consumed per unit time to the total energy. The update formula is ,in and are the maximum and minimum values ​​of the inertia weight, is the maximum number of iterations, is the current iteration number. Thus, when the inflation efficiency error is large and the battery energy consumption rate is high, the inertia weight is adjusted accordingly, making the algorithm more inclined to global search to find a better strategy; conversely, it focuses more on local search to refine the current optimal solution.

[0082] At the same time, an elite retention mechanism is used to screen the set of non-inferior solutions during the iteration process. In each iteration, non-inferior solutions from the current population are saved. These non-inferior solutions represent strategies that performed well in different aspects of the current iteration. The Cauchy mutation operator is then used to enhance global search capabilities. Based on the current solution, the Cauchy mutation operator randomly generates a mutation value according to the Cauchy distribution, perturbing the solution. This increases the probability of the algorithm escaping the local optimal solution, allowing the algorithm to search across a wider range of solution spaces.

[0083] After multiple iterations of optimization, the final output is a control sequence and resource allocation result that conforms to the actual situation of the current air pump, achieving optimized control of the power supply and improving the working efficiency and performance of the air pump.

[0084] Example 3:

[0085] This example focuses on the application of a convolutional neural network model in the power control of an automobile air pump. The convolutional neural network model is designed to more efficiently process dynamic feature vectors and generate a global correlation feature matrix.

[0086] The multi-scale convolution layer of the model is an important part of achieving this goal. The multi-scale convolution layer uses convolution kernels of different sizes to perform convolution operations on the input dynamic feature vector. For example, setting a small size convolution kernel (such as ) is used to extract local detail features and capture subtle changes in the power signal; large-size convolution kernels (such as or ) is used to obtain broader contextual information and grasp the overall trend of the signal. This multi-scale convolution approach can extract power signal features in different frequency bands.

[0087] After extracting features from different frequency bands, these features are weighted and fused using the channel attention mechanism. This mechanism adaptively adjusts the features of different channels by calculating the importance weight of each channel. Specifically, a global average pooling operation is first performed on the feature map in the spatial dimension to compress the features of each channel into a single value, resulting in a channel descriptor. This channel descriptor is then processed through a neural network containing a fully connected layer to generate a weight value for each channel. This weight value reflects the importance of the channel feature within the overall feature set. Finally, the features of each channel are multiplied by the corresponding weight value to achieve weighted fusion of the features across the frequency bands, thereby generating a global correlation feature matrix.

[0088] To illustrate with a simple example, suppose the input dynamic feature vector passes through the multi-scale convolution layer to obtain feature maps of three different frequency bands 、 、 After global average pooling, the channel descriptors are obtained. 、 、 . The corresponding weight value is obtained through neural network processing 、 、 . Then the final fused feature map The calculation method is ,this This is the generated global correlation feature matrix.

[0089] Through the collaborative work of multi-scale convolutional layers and the channel attention mechanism, the convolutional neural network model can fully mine the information in the dynamic feature vector and generate a more representative global correlation feature matrix, providing strong support for the subsequent generation of power control strategies and resource allocation plans based on this matrix, thereby improving the accuracy and effectiveness of the power control of the automotive air pump.

[0090] Embodiment 4:

[0091] In the improved particle swarm optimization algorithm, the adaptive inertia weight adjustment strategy, elite retention mechanism and Cauchy mutation operator work together to effectively improve the algorithm performance.

[0092] The adaptive inertia weight adjustment strategy dynamically updates the inertia weight based on the real-time inflation efficiency error and the battery energy consumption rate. The real-time inflation efficiency error reflects the difference between the actual inflation efficiency of the current pump and the target. When the actual inflation efficiency is lower than the target, the error is positive, and a larger error indicates a greater deviation from the ideal inflation efficiency. The battery energy consumption rate reflects the rate at which battery energy is consumed during the inflation process. Excessive battery energy consumption can affect the pump's endurance and overall performance.

[0093] In actual calculation, it is assumed that the real-time error of the inflation efficiency is By the current actual inflation volume and target inflation volume In unit time The difference within is calculated, that is, Battery energy consumption rate Energy consumed by the current battery and the initial total energy of the battery In unit time The ratio calculation within .

[0094] According to the formula , at the beginning of the algorithm, , inertia weight near , at this time, the algorithm has a strong global search capability, and the particles can explore the solution space in a larger range and find the possible optimal solution area. As the iteration proceeds, if the inflation efficiency error is large and the battery energy consumption rate is high, The larger the value of Will decrease rapidly, making the particles more inclined to global search in order to jump out of the current possible local optimal solution. On the contrary, if the inflation efficiency error is small and the battery energy consumption rate is low, the inertia weight It will be relatively stable at a higher value. The algorithm focuses more on local search and makes fine adjustments to the current optimal solution.

[0095] During the iteration process, the elite retention mechanism screens the population generated at each iteration. Non-inferior solutions—those that are not dominated by other solutions in terms of objective functions (e.g., a combination of multiple objectives such as inflation efficiency and battery energy consumption)—are retained. These non-inferior solutions represent strategies that perform well in different aspects of the current iteration. For example, one solution might excel in inflation efficiency, while another might be better in battery energy consumption; both are retained.

[0096] The Cauchy mutation operator adds the ability to escape from the local optimal solution. In each iteration, a mutation value is generated for some particles according to the Cauchy distribution. The Cauchy distribution has a heavier tail, which can make the particle search far away from the current position compared to the traditional Gaussian mutation. Assume that the current position of the particle is , the variation value is , the position after mutation for Through this mutation operation, the algorithm can explore a wider solution space, avoid falling into the local optimal solution, and increase the probability of finding the global optimal solution.

[0097] By combining these three mechanisms, the improved particle swarm optimization algorithm can more efficiently search the solution space when generating power control strategies and resource allocation plans, improve the algorithm's convergence speed and solution quality, and provide better strategic support for the power control of automotive air pumps.

[0098] Example 5:

[0099] In the process of controlling the power supply of an automobile air pump, signal feature extraction and abnormal pattern detection and processing are important links to ensure the stable operation of the power supply.

[0100] First, the signal feature extraction adopts the wavelet transform and time-frequency joint analysis method. The power signal is decomposed by the Morlet wavelet basis function, and the energy entropy characteristics at each scale are extracted. The Morlet wavelet basis function has good time-frequency localization characteristics and can perform detailed analysis of the power signal at different time and frequency scales. Assume that the power signal is , using Morlet wavelet basis function Perform wavelet transform on it and get the wavelet coefficients ,in is the scale parameter, is the translation parameter. By calculating different scales Energy entropy of wavelet coefficients ,here Represents different translation positions, energy entropy It reflects the complexity and uncertainty of the signal at that scale, thereby extracting the energy entropy characteristics at each scale.

[0101] Next, the time-frequency matrix is ​​subjected to singular value decomposition to generate dynamic eigenvectors that characterize the signal's mutation and steady-state components. The time-frequency matrix is ​​composed of wavelet coefficients obtained by wavelet transform. Let the time-frequency matrix be , perform singular value decomposition on it ,in and is a unitary matrix, It is a diagonal matrix whose diagonal elements are singular values. The size and distribution of the singular values ​​reflect the different characteristics of the signal. Larger singular values ​​correspond to the main components of the signal, while smaller singular values ​​correspond to the secondary components or noise. By selecting appropriate singular values ​​and combining them into dynamic eigenvectors, we can effectively characterize the sudden change and steady-state components of the signal.

[0102] During real-time monitoring, a fuzzy logic system is used to detect abnormal patterns in real-time power supply data. Fuzzy logic systems process uncertain information by defining fuzzy sets and fuzzy rules. For example, for voltage data, fuzzy sets such as "normal voltage," "slightly low voltage," and "excessively low voltage" are defined, along with corresponding membership functions to describe the degree to which the voltage data belongs to each fuzzy set. Based on the fuzzy information of multiple power supply data dimensions (such as voltage, current, and temperature), a series of fuzzy rules are formulated. For example, if the voltage is "excessively low voltage" and the current is "excessively high current," it is determined that an abnormality may exist.

[0103] When an anomaly is detected, dynamic reconfiguration of the control strategy sequence is triggered, and the resource allocation plan is updated based on the reconfiguration results. This dynamic reconfiguration utilizes a mixed-integer linear programming solver. First, power operation sequence constraints are defined to ensure that each power control operation is executed in the correct order, thus avoiding system failures caused by improper operation sequence. For example, before charging the battery, it is necessary to ensure that the charging circuit is properly connected and the relevant protection mechanisms are activated.

[0104] At the same time, consider limiting the number of battery charge and discharge cycles to extend the battery's service life. Batteries have a limited number of charge and discharge cycles, and excessive charge and discharge accelerate battery aging. Therefore, when developing control strategies, avoid frequent and unnecessary charge and discharge operations to keep the number of battery charge and discharge cycles within a reasonable range.

[0105] Furthermore, the rigid constraints of the radiator's on / off time windows must be considered. The radiator's function is to dissipate heat from the air pump, ensuring it operates within a suitable temperature range. In certain situations, such as when the air pump is operating at high load for extended periods, the radiator needs to be activated promptly. However, when temperatures are lower or the air pump is under low load, the radiator can be appropriately shut down to conserve energy.

[0106] A set of feasible solutions is generated using a column generation algorithm. This algorithm gradually expands the solution space by generating new columns (corresponding to different power control strategy options) to find solutions that satisfy various constraints. The optimal overall solution is then selected using the entropy-weighted TOPSIS method. This method weights each metric based on its information entropy. It then ranks feasible solutions by calculating their distance from the ideal and negative ideal solutions, ultimately selecting the optimal overall solution. This allows for rapid adjustment of control strategies and resource allocation plans when power anomalies are detected, ensuring stable operation of the vehicle air pump.

[0107] Example 6:

[0108] In the entire process of automobile air pump power control, building a knowledge graph of power component relationships and verifying the physical compatibility of strategies and knowledge graphs during the strategy sequence generation phase play a key role in ensuring the stable and efficient operation of the air pump power supply.

[0109] When constructing the knowledge graph of power component relationships, the various electrical components in the air pump and the possible fault types are set as nodes. As far as electrical components are concerned, batteries, as key energy supply components of air pumps, store and output electrical energy, and are core nodes in the knowledge graph. They have many properties, such as battery capacity, voltage, internal resistance, etc. These properties directly affect the working time and performance of the air pump. The motor is equally important. It converts electrical energy into mechanical energy to drive the air pump to work. The motor's speed, torque, power and other properties are also key information that the knowledge graph needs to record. In addition, the controller is responsible for regulating the operation of the entire power system, and the radiator ensures that the system works at an appropriate temperature. They all exist in the knowledge graph as independent nodes.

[0110] Fault types are also included in the node category. For example, a battery overcharge fault occurs when the battery charge exceeds the safety limit due to improper charging control; a motor overheating fault is usually caused by prolonged high-load operation or poor heat dissipation. Each fault type node is associated with the relevant electrical component node, and this association reflects the cause and effect relationship between the fault.

[0111] The edges between nodes are used to represent the energy coupling relationship or fault conduction path between components. In terms of energy coupling, there is a close energy transfer connection between the battery and the motor. The electric energy output by the battery is transmitted to the motor through the circuit to drive the motor. In the knowledge graph, a directed edge is drawn from the battery node to the motor node. The attributes of the edge can mark information such as output voltage and current, so as to clearly present the energy coupling between the two. For example, when a motor overheats, the excessive temperature may affect the nearby controller through conduction or heat radiation of metal parts. In the knowledge graph, there will be an edge from the motor overheat fault node to the controller fault node. The attributes of this edge will record the possibility of fault conduction, time delay, and possible controller fault types. In this way, a complete knowledge graph of power component relationships is constructed.

[0112] During the strategy sequence generation phase, a subgraph isomorphism detection algorithm is used to verify the physical compatibility of the strategy with the knowledge graph. When a power control strategy sequence is generated, it is converted into a subgraph. The nodes in the subgraph correspond to the electrical component operations or fault response measures involved in the strategy, and the edges represent the logical relationships between these operations or measures. The subgraph is then compared with a pre-built knowledge graph of power component relationships.

[0113] For example, one control strategy is to start charging when the battery charge is below a certain threshold, while adjusting the motor speed to reduce energy consumption. In the subgraph corresponding to this strategy, the battery node and the motor node will have corresponding operation tags, and there will be edges representing the sequence and mutual influence of the two operations. Through the subgraph isomorphism detection algorithm, check whether this subgraph can find a part isomorphic to it in the knowledge graph. If it can be found, it means that the strategy is reasonable and feasible at the physical level, because it conforms to the actual energy coupling relationship and fault conduction logic between the various components of the air pump; on the contrary, if the isomorphic part cannot be found, it means that the strategy may cause physical conflicts or does not conform to the actual working principle, and the strategy needs to be adjusted or reformulated.

[0114] By constructing a knowledge graph of power component relationships and performing subgraph isomorphism detection, potential problems can be discovered in a timely manner during the strategy formulation stage, ensuring that the generated power control strategy sequence not only meets the control requirements but also conforms to the physical characteristics of the air pump, thereby improving the accuracy and reliability of the power control of the automobile air pump and ensuring the stable operation of the air pump.

[0115] A car air pump includes an air pump housing body 1, an air pump assembly 2 and a mobile power supply 3. The air pump housing body 1 is provided with an assembly groove 6 for the mobile power supply 3 to be movably arranged therein. The assembly groove 6 is provided with an electrical interface terminal 12 electrically connected to the air pump assembly 2. The mobile power supply 3 is electrically connected to the air pump assembly 2 through the electrical interface terminal 12. A spring 7 is provided in the assembly groove 6. One end of the spring 7 is fixed to the bottom of the assembly groove 6, and the other end is in contact with the mobile power supply 3. A control button 5 is provided on the surface of the air pump housing body 1. The bottom end of the control button 5 is connected to a button bracket in contact with the spring 7.

[0116] It should be noted that the button bracket of this embodiment further includes a linkage rod, the two ends of which contact the spring piece 7 and the control button 5 respectively. When in use, the pressing action of the control button 5 is transmitted to the spring piece 7 through the linkage rod. After receiving the thrust of the linkage rod, the free end of the spring piece 7 is pushed upward by its own elastic deformation. When the control button 5 is released, the elastic restoring force of the spring piece 7 causes the free end to return to its initial position, completing a complete linkage operation.

[0117] In addition, the control button 5 and the shrapnel 7 are physically connected through a linkage rod. This connection method is simple and stable, which can ensure the sensitivity and reliability of the button operation. The elastic design of the shrapnel 7 enables it to maintain good performance after multiple operations, reducing the risk of failure due to mechanical fatigue.

[0118] The mobile power supply 3 is electrically connected to the air pump assembly 2 through an electrical interface terminal 12. Specifically, the electrical interface terminal 12 is installed inside the air pump housing body 1 near the installation position of the mobile power supply 3. The electrical interface terminal 12 is composed of multiple metal contact pieces, which correspond to the positive and negative poles of the mobile power supply 3 and the control signal line respectively.

[0119] The contact pieces of the electrical interface terminals 12 have a certain degree of elasticity, providing stable contact pressure when the mobile power supply 3 is inserted, ensuring the reliability of the electrical connection. The side or bottom of the mobile power supply 3 is provided with metal contact points that match the electrical interface terminals 12. The position and shape of these contact points correspond one-to-one with the contact pieces of the electrical interface terminals 12.

[0120] When the mobile power supply 3 is inserted into the interior of the air pump housing body 1, its metal contact points will come into contact with the contact pieces of the electrical interface terminals 12. Due to the elastic design of the electrical interface terminals 12, the contact pieces will automatically clamp the contact points of the mobile power supply 3, ensuring a stable electrical connection between the two.

[0121] Disassembly process: When you need to disassemble the power bank 3, press the control button 5 and use the linkage rod to drive the spring 7 to push the power bank 3 out. Due to the elasticity of the electrical interface terminal 12, the contact piece will automatically release, allowing the power bank 3 to be smoothly removed and the electrical connection to be automatically disconnected.

[0122] Therefore, the detachable design of the power bank 3 allows users to quickly replace a power bank with insufficient power without waiting for charging, thereby improving the efficiency of device use. Users can choose power banks of different capacities according to actual needs to adapt to different usage scenarios.

[0123] The mobile power supply 3 is installed inside the air pump housing body 1, without taking up additional space, making the entire device structure more compact and suitable for use in limited space. The electrical interface terminal 12 integrates the functions of the mobile power supply 3 and the air pump assembly 2, reducing the use of external cables and improving the overall aesthetics and reliability of the device.

[0124] A fan assembly 8 is connected to one side of the air pump assembly 2, and an air inlet 9 corresponding to the fan assembly 8 is provided on the side wall of the air pump housing body 1. An air outlet 10 connected to the air pump assembly 2 is provided on the other side of the air pump housing body 1. When in use, the fan assembly 8 ventilates the interior of the pump body through the air inlet 9 and the air outlet 10 to continuously reduce the temperature.

[0125] In addition, a lighting lamp 11 is embedded on the surface of the air pump housing body 1, which provides temporary lighting to cope with use at night or in locations with poor lighting in underground garages. A closable sealing baffle 4 is provided at the opening of the assembly groove 6 to seal the assembly groove 6 and provide sealing protection.

[0126] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "includes," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0127] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A power supply control method for an automobile air pump, characterized in that: include: Collect multi-dimensional power supply data of automobile air pumps; Extracting signal features from the multi-dimensional power supply data to generate dynamic feature vectors of each dimension; Mapping the dynamic feature vector to a joint analysis space based on a pre-trained convolutional neural network model to generate a global correlation feature matrix; Building a power mode knowledge base that stores historical power usage patterns, failure cases, optimization strategies, and corresponding execution efficiency data; Generate a power control strategy sequence and resource allocation plan for the current air pump through a genetic algorithm based on the global correlation feature matrix and the historical patterns in the knowledge base; The signal feature extraction adopts wavelet transform and time-frequency joint analysis method, which specifically includes: The power signal is decomposed by Morlet wavelet basis function to extract the energy entropy characteristics at each scale; Perform singular value decomposition on the time-frequency matrix to generate dynamic eigenvectors that represent the signal's sudden and steady-state components; The method further comprises: Detect abnormal patterns in real-time power data based on a fuzzy logic system; When an anomaly is detected, the dynamic reconstruction of the control strategy sequence is triggered, and the resource allocation plan is updated according to the reconstruction results; The dynamic reconstruction uses a mixed integer linear programming solver, specifically including: Define hard constraints on power supply operation sequence constraints, battery charge and discharge cycle limits, and radiator start and stop time windows; Generate a feasible solution set through the column generation algorithm, and screen the comprehensive optimal solution based on the entropy weight TOPSIS method; The method further comprises: Construct a knowledge graph of power supply component relationships, where nodes represent electrical components of the air pump or fault types, and edges represent energy coupling relationships between components or fault conduction paths; In the strategy sequence generation phase, the physical compatibility of the strategy with the knowledge graph is verified through the subgraph isomorphism detection algorithm.

2. The power control method according to claim 1, wherein: The multi-dimensional power supply data includes voltage, current, ambient temperature, humidity, inflation rate, battery health status, user-preset pressure value, motor speed and radiator working status; The dynamic feature vector includes power ripple spectrum characteristics, load fluctuation detection characteristics, temperature gradient change characteristics, battery internal resistance dynamic distribution characteristics and motor torque response delay characteristics.

3. The power control method according to claim 1, wherein: The method of generating the power control strategy sequence and resource allocation scheme of the current air pump by using a genetic algorithm includes: generating an initial candidate control strategy set according to the matching degree between the global correlation feature matrix and the fault mode of the knowledge base; Building a policy execution dependency graph based on the air pump hardware topology, wherein nodes in the policy execution dependency graph represent power control operations, and edges represent energy transfer constraints or hardware resource conflicts between operations; Combined with real-time power supply data, the initial candidate control strategy set is iteratively optimized through an improved particle swarm optimization algorithm to output the final control sequence and resource allocation results.

4. The power control method according to claim 3, wherein: The convolutional neural network model includes: Use multi-scale convolutional layers to extract power signal features in different frequency bands; The features of each frequency band are weighted and fused through the channel attention mechanism to generate the global correlation feature matrix.

5. The power control method according to claim 3, wherein: The improved particle swarm optimization algorithm includes: Introducing an adaptive inertia weight adjustment strategy, where the weight is dynamically updated based on the real-time error in the inflation efficiency and the battery energy consumption rate; An elite retention mechanism is used to screen the non-inferior solution set in the iterative process, and the global search capability is enhanced through the Cauchy mutation operator.

6. A power supply control system for an automobile air pump, characterized in that: The power supply control method according to any one of claims 1 to 5 comprises: Data acquisition module: used to collect multi-dimensional power supply data of automobile air pumps; Feature extraction module: extracts signal features from the multi-dimensional power supply data to generate dynamic feature vectors of each dimension; Feature association module: maps the dynamic feature vector to the joint analysis space based on the pre-trained convolutional neural network model to generate a global correlation feature matrix; Knowledge base construction module: constructs a power mode knowledge base, which stores historical power usage patterns, failure cases, optimization strategies and corresponding execution efficiency data; Strategy generation module: generates a power control strategy sequence and a resource allocation plan for the current air pump through a genetic algorithm according to the global correlation feature matrix and the historical patterns in the knowledge base.

7. An automobile air pump, characterized in that: The power supply control system according to claim 6 comprises an air pump housing body (1), an air pump assembly (2) and a mobile power supply (3), characterized in that: an assembly groove (6) is provided on the air pump housing body (1) for the mobile power supply (3) to be movably arranged therein, an electrical interface terminal (12) electrically connected to the air pump assembly (2) is provided in the assembly groove (6), the mobile power supply (3) is electrically connected to the air pump assembly (2) through the electrical interface terminal (12), a spring (7) is provided in the assembly groove (6), one end of the spring (7) is fixed to the bottom of the assembly groove (6), and the other end is in contact with the mobile power supply (3), a control button (5) is provided on the surface of the air pump housing body (1), and the bottom end of the control button (5) is connected to a button bracket in contact with the spring (7).

8. The automobile air pump according to claim 7, characterized in that: One side of the air pump assembly (2) is connected to a fan assembly (8), a side wall of the air pump housing body (1) is provided with an air inlet (9) corresponding to the fan assembly (8), and the other side of the air pump housing body (1) is provided with an air outlet (10) in communication with the air pump assembly (2).

9. The automobile air pump according to claim 7, characterized in that: A lighting lamp (11) is embedded in the surface of the air pump housing body (1), and a sealing baffle (4) that can be opened and closed is provided at the opening of the assembly groove (6).

Images