Charging pile supporting multi-point automatic cooling

Abstract

The invention provides a charging pile supporting multi-point automatic cooling, and belongs to the technical field of charging piles, and the charging pile comprises a main control module, a temperature acquisition module, a current acquisition module, a pressure detection module, a heat dissipation control module, a communication module, a storage module, a display module and a power management module. A system control module arranged in a main control module firstly acquires temperature data of each point of a charging pile through a temperature acquisition module, constructs a temperature difference matrix and performs matrix decomposition to obtain a temperature space basis matrix and a time characteristic matrix; then, a dynamic convolution kernel is generated based on the characteristic value of the temperature space basis matrix, a temperature characteristic extraction network is constructed, and accurate extraction of temperature characteristics is realized. And the system further establishes a thermodynamic state equation set, the heat dissipation power is accurately calculated and controlled in combination with a temperature difference change coefficient matrix, and finally accurate regulation and control of multi-point temperature are achieved.

Description

Technical Field

[0001] The present invention belongs to the technical field of charging piles, and more specifically, relates to a charging pile supporting multi-point automatic cooling. Background Art

[0002] With the rapid development of the electric vehicle industry, the construction and application of charging infrastructure have become increasingly widespread. As a key device, the operating safety and reliability of charging piles have attracted much attention. Traditional charging pile temperature control systems usually use single-point or multi-point temperature sensors for temperature monitoring and cooperate with air-cooled or liquid-cooled heat dissipation systems to achieve temperature control. Such systems mainly trigger the start and stop of heat dissipation devices by setting temperature thresholds and adjust the heat dissipation power according to the real-time data of temperature sensors. Currently, common charging pile temperature control schemes on the market include temperature regulation systems based on PID control, intelligent heat dissipation systems based on fuzzy control, and prediction control systems based on neural networks, etc. These systems have formed a relatively complete technical system in practical applications, providing a basic guarantee for the safe operation of charging piles.

[0003] However, traditional charging pile temperature control systems have many deficiencies. First, existing temperature monitoring schemes often adopt fixed sampling frequencies and fixed sensor layouts, making it difficult to dynamically adjust the monitoring strategy according to the actual operating conditions, resulting in insufficient spatio-temporal resolution of temperature data. Second, most traditional temperature control algorithms are based on simple threshold judgments or fixed control parameters, unable to accurately capture the dynamic characteristics of temperature changes and difficult to establish an accurate correlation between temperature changes and system operating parameters. Third, existing heat dissipation control schemes generally have problems such as response lag and insufficient control accuracy, making it difficult to achieve early warning and precise regulation of temperature anomalies. These problems are particularly prominent in high-power fast charging scenarios, seriously affecting the service life and operating safety of charging piles.

[0004] Based on the above analysis, it is difficult for the existing technology to effectively solve the problems of insufficient multi-point temperature monitoring and control accuracy of charging piles. This is mainly reflected in three aspects: First, the acquisition and processing of temperature data lack dynamic adaptability and cannot accurately reflect the real-time state of the system; second, the methods for extracting and analyzing temperature change characteristics are relatively simple, making it difficult to mine the deep information contained in temperature data; third, the heat dissipation control strategy lacks foresight and accuracy, making it difficult to achieve fast and precise temperature regulation. These technical defects seriously restrict the performance improvement and safety guarantee of charging piles. That is to say, there are technical problems in the existing technology such as insufficient multi-point temperature monitoring and control accuracy of charging piles, resulting in low heat dissipation efficiency and inaccurate temperature warning. Summary of the Invention

[0005] In view of this, the present invention provides a charging pile supporting multi-point automatic temperature reduction, which can solve the technical problems in the prior art that the multi-point temperature monitoring and control accuracy of the charging pile is insufficient, resulting in low heat dissipation efficiency and inaccurate temperature warning.

[0006] The present invention is implemented as follows: The present invention provides a charging pile supporting multi-point automatic temperature reduction, which includes a main control module. The main control module is built-in with a control chip, and a system control module is provided in the control chip. The system control module performs matrix decomposition based on the temperature difference matrix to obtain a temperature space basis matrix and a time feature matrix, and generates a dynamic convolution kernel using a convolution kernel calculation function. The convolution kernel calculation function determines the convolution kernel size based on the eigenvalues of the temperature space basis matrix, constructs a temperature feature extraction network using the dynamic convolution kernel, constructs a temperature difference change coefficient matrix based on the feature vectors output by the temperature feature extraction network, obtains charging current data through the current acquisition module, establishes a temperature-power correlation matrix in combination with the temperature difference change coefficient matrix, establishes a cooling capacity evaluation model through the thermodynamic state equation set according to the coolant pressure and pipeline pressure collected by the pressure detection module, calculates the required heat dissipation power of each monitoring point based on the temperature difference change coefficient matrix and the cooling capacity evaluation model, and adjusts the rotation speed of the cooling fan and the coolant flow through the heat dissipation control module to achieve multi-point temperature control. Further, it also includes real-time analysis of system operation data, establishing a heat dissipation efficiency evaluation index, and dynamically optimizing the heat dissipation strategy.

[0007] It also includes a temperature acquisition module, a current acquisition module, a pressure detection module, a heat dissipation control module, a communication module, a storage module, a display module, and a power management module. The main control module is electrically connected to the temperature acquisition module, the current acquisition module, the pressure detection module, the heat dissipation control module, the communication module, the storage module, the display module, and the power management module respectively; the sampling frequency of the temperature acquisition module for temperature parameters is 4 times per second, the sampling frequency of the current acquisition module is 2 times per second, and the sampling frequency of the pressure detection module is 1 time per second.

[0008] ; The sampling frequency of the temperature acquisition module for temperature parameters is 4 times per second, the sampling frequency of the current acquisition module is 2 times per second, and the sampling frequency of the pressure detection module is 1 time per second.

[0009] Among them, the temperature difference matrix is an n×m matrix, n represents the number of temperature sampling points, m represents the sampling time series length, and each element in the temperature difference matrix represents the difference between the temperature of a certain sampling point at a specific moment and the ambient temperature.

[0010] Among them, the temperature spatial basis matrix is an n×k matrix, where k represents the number of main features, and the temperature spatial basis matrix characterizes the spatial distribution characteristics of temperature sampling points; the time feature matrix is a k×m matrix, and the time feature matrix characterizes the time variation characteristics of temperature data.

[0011] Among them, the temperature difference change coefficient matrix is an n×(m - 1) matrix, and each element in the temperature difference change coefficient matrix represents the change rate of the temperature difference between two adjacent moments in the temperature difference matrix.

[0012] Among them, the temperature-power correlation matrix is an n×p matrix, where p represents the number of power sampling points, and the temperature-power correlation matrix characterizes the correlation relationship between temperature change and power change.

[0013] Among them, the convolution kernel calculation function determines the size of the dynamic convolution kernel through the singular value decomposition result of the temperature spatial basis matrix, and the size of the dynamic convolution kernel is equal to the number of non-zero singular values of the temperature spatial basis matrix plus 1.

[0014] Among them, the temperature feature extraction network includes an input layer, a dynamic convolution layer, a pooling layer, a fully connected layer, and an output layer. The dynamic convolution layer uses the dynamic convolution kernel to perform feature mapping on temperature data.

[0015] Among them, the thermodynamic state equation set includes a coolant heat balance equation, a pipeline pressure loss equation, and a radiator heat transfer equation; the coolant heat balance equation is used to calculate the system heat dissipation power, with inputs including the coolant flow rate, the coolant specific heat capacity, and the coolant inlet and outlet temperature difference, and the output is the system heat dissipation power; the pipeline pressure loss equation is used to calculate the pipeline resistance, with inputs including the coolant flow rate, the pipeline length, the pipeline cross-sectional area, and the coolant viscosity, and the output is the pipeline pressure loss value; the radiator heat transfer equation is used to calculate the radiator heat dissipation power, with inputs including the radiator surface area, the radiator convective heat transfer coefficient, the radiator surface temperature, and the ambient temperature, and the output is the radiator heat dissipation power.

[0016] Among them, the temperature acquisition module is used to acquire the charging interface temperature and the radiator temperature, the current acquisition module is used to acquire the charging current and the cooling fan working current, the pressure detection module is used to acquire the coolant pressure and the pipeline pressure, the heat dissipation control module is used to control the cooling fan speed and the coolant flow rate, the communication module is used to realize data transmission between systems and remote monitoring, the storage module is used to store temperature data and system parameters, the display module is used to display the system operation status and alarm information, and the power management module is used to monitor the system power supply voltage and control the system power supply.

[0017] Compared with the prior art, the present invention provides a charging pile that supports multi-point automatic temperature reduction. The present invention proposes a temperature control scheme for charging piles based on dynamic convolution kernels and matrix decomposition. By constructing and decomposing the temperature difference matrix, adaptively generating dynamic convolution kernels, and establishing a system of multi-variable thermodynamic state equations, the accurate extraction and analysis of the temperature change characteristics of the charging pile are realized. This scheme uses the dual representation of the temperature space basis matrix and the time feature matrix, combined with the feature mapping mechanism of the dynamic convolution kernel, to greatly improve the accuracy of temperature monitoring and control.

[0018] The solution of the present invention effectively solves the problems existing in the traditional technology. First, through the decomposition technology of the temperature difference matrix, the spatial distribution characteristics and time variation characteristics of the temperature data are decoupled, providing a more detailed temperature feature expression. Second, based on the eigenvalues of the temperature space basis matrix, dynamic convolution kernels are generated, realizing the adaptability of feature extraction and significantly improving the accuracy of temperature feature extraction. Third, by establishing a complete system of thermodynamic state equations, an accurate model of the system's heat dissipation capacity is realized, providing a reliable theoretical basis for temperature control.

[0019] The present invention successfully solves the technical problem of insufficient accuracy in multi-point temperature monitoring and control of charging piles, which is mainly attributed to three aspects of innovation: one is to use matrix decomposition technology to realize the multi-dimensional representation of temperature data, enhancing the comprehensiveness of feature extraction; the second is to introduce the dynamic convolution kernel mechanism, improving the adaptability and accuracy of feature mapping; the third is to establish a complete thermodynamic model to realize the precise control of the heat dissipation process. These innovations ensure that the system can timely detect temperature anomalies, accurately predict the temperature change trend, and achieve precise temperature regulation. Brief Description of the Drawings

[0020] Figure 1 It is a schematic diagram of the composition of the charging pile of the present invention.

[0021] Figure 2 It is a flowchart of the steps executed by the system control module.

[0022] Figure 3 It is a graph of the eigenvalue distribution and cumulative contribution rate of the temperature difference matrix in Example 2.

[0023] Figure 4 It is a comparison graph of the design values and measured values of the performance parameters of the heat dissipation control system in Example 2.

[0024] Figure 5 It is a graph of three key performance indicators during the operation of the system in Example 2. Detailed Embodiments

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0026] As Figure 1 shown, it is a flowchart of a charging pile supporting multi-point automatic cooling provided by the present invention. The method includes the following steps: A charging pile system supporting multi-point automatic cooling, characterized in that it includes a main control module, a temperature acquisition module, a current acquisition module, a pressure detection module, a heat dissipation control module, a communication module, a storage module, a display module, and a power management module. The main control module is built with a control chip. The main control module is electrically connected to the temperature acquisition module, the current acquisition module, the pressure detection module, the heat dissipation control module, the communication module, the storage module, the display module, and the power management module respectively. A system control module is provided in the control chip. The temperature acquisition module is used to acquire the temperature of the charging interface and the radiator temperature. The current acquisition module is used to acquire the charging current and the working current of the cooling fan. The pressure detection module is used to acquire the coolant pressure and the pipeline pressure. The heat dissipation control module is used to control the rotation speed of the cooling fan and the coolant flow rate. The communication module is used to realize data transmission between systems and remote monitoring. The storage module is used to store temperature data and system parameters. The display module is used to display the system operation status and alarm information. The power management module is used to monitor the system power supply voltage and control the system power supply. The sampling frequency of the temperature acquisition module for temperature parameters is 4 times per second, the sampling frequency of the current acquisition module is 2 times per second, and the sampling frequency of the pressure detection module is 1 time per second.

[0027] The system control module is used to execute the following steps: S01. Acquire the temperature of the charging interface and the radiator temperature through the temperature acquisition module, and establish a temperature difference matrix, which records the difference between the temperature of each sampling point and the ambient temperature; S02. Perform matrix decomposition based on the temperature difference matrix to obtain a temperature space basis matrix and a time feature matrix. The temperature space basis matrix characterizes the spatial distribution characteristics of the sampling points, and the time feature matrix characterizes the change characteristics of temperature over time; S03. Generate a dynamic convolution kernel using a convolution kernel calculation function. The convolution kernel calculation function determines the convolution kernel size based on the eigenvalues of the temperature space basis matrix, and the convolution kernel size matches the number of main features of the temperature space basis matrix; S04. Construct a temperature feature extraction network using the dynamic convolution kernel. The temperature feature extraction network includes an input layer, a dynamic convolution layer, a pooling layer, a fully connected layer, and an output layer. The dynamic convolution layer uses the dynamic convolution kernel to perform feature mapping on temperature data; S05. Construct a temperature difference change coefficient matrix based on the feature vector output by the temperature feature extraction network. The temperature difference change coefficient matrix is used to characterize the temperature change rate of each sampling point; S06. Obtain charging current data through the current acquisition module, and establish a temperature-power correlation matrix in combination with the temperature difference change coefficient matrix; S07. According to the coolant pressure and pipeline pressure collected by the pressure detection module, establish a cooling capacity evaluation model through the thermodynamic state equations; S08. Based on the temperature difference change coefficient matrix and the cooling capacity evaluation model, calculate the required heat dissipation power at each monitoring point; S09. Adjust the rotation speed of the cooling fan and the coolant flow rate through the heat dissipation control module to achieve multi-point temperature control; S10. Optionally, it also includes real-time analysis of system operation data, establishing a heat dissipation efficiency evaluation index, and dynamically optimizing the heat dissipation strategy.

[0028] The thermodynamic state equations include a coolant heat balance equation, a pipeline pressure loss equation, and a radiator heat transfer equation; the coolant heat balance equation is used to calculate the system heat dissipation power. The inputs include the coolant flow rate, the coolant specific heat capacity, and the coolant inlet and outlet temperature difference, and the output is the system heat dissipation power; the pipeline pressure loss equation is used to calculate the pipeline resistance. The inputs include the coolant flow rate, the pipeline length, the pipeline cross-sectional area, and the coolant viscosity, and the output is the pipeline pressure loss value; the radiator heat transfer equation is used to calculate the radiator heat dissipation power. The inputs include the radiator surface area, the radiator convective heat transfer coefficient, the radiator surface temperature, and the ambient temperature, and the output is the radiator heat dissipation power.

[0029] The temperature difference matrix is an n×m matrix, where n represents the number of temperature sampling points and m represents the length of the sampling time series. Each element in the temperature difference matrix represents the difference between the temperature of a certain sampling point at a specific moment and the ambient temperature.

[0030] The temperature space basis matrix is an n×k matrix, where k represents the number of main features. The temperature space basis matrix characterizes the spatial distribution characteristics of temperature sampling points.

[0031] The time feature matrix is a k×m matrix, and the time feature matrix characterizes the time change characteristics of temperature data.

[0032] The temperature difference change coefficient matrix is an n×(m - 1) matrix, and each element in the temperature difference change coefficient matrix represents the change rate of the temperature difference between two adjacent moments in the temperature difference matrix.

[0033] The temperature-power correlation matrix is an n×p matrix, where p represents the number of power sampling points, and the temperature-power correlation matrix characterizes the correlation relationship between temperature change and power change.

[0034] The convolution kernel calculation function determines the size of the dynamic convolution kernel through the singular value decomposition result of the temperature spatial basis matrix, and the size of the dynamic convolution kernel is equal to the number of non-zero singular values of the temperature spatial basis matrix plus 1.

[0035] The following describes the specific implementation manners of the above steps in detail. In the charging pile of the present invention, the main control module uses a high-performance ARM processor STM32F746 with a main frequency of up to 216 MHz, and is built-in with 2 MB of flash memory and 384 KB of static random access memory. This processor has rich peripheral interfaces and powerful real-time processing capabilities, and can meet the real-time requirements of the system for temperature data acquisition, feature extraction, and control strategy execution. The built-in control chip adopts a dual-core architecture. One core is specifically responsible for data acquisition and processing, and the other core is responsible for the execution of control strategies, thereby ensuring the real-time response ability of the system.

[0036] The temperature acquisition module uses a high-precision PT100 platinum resistance temperature sensor with a temperature measurement range of -40 degrees Celsius to 200 degrees Celsius and an accuracy of up to ±0.1 degrees Celsius. Each charging interface is configured with 3 temperature sensors, which are installed at different positions of the charging interface to realize multi-point monitoring of the charging interface temperature. The radiator temperature is monitored using an NTC thermistor with a temperature measurement range of -20 degrees Celsius to 120 degrees Celsius and an accuracy of ±0.5 degrees Celsius. The sampling frequency of the temperature acquisition module is 4 times per second, and a 24-bit analog-to-digital converter is used to convert the temperature signal, ensuring the acquisition accuracy of temperature data.

[0037] The current acquisition module uses a Hall current sensor with a measurement range of 0 to 500 amperes, an accuracy of ±1%, and a sampling frequency of 2 times per second. The current sensor adopts a closed-loop structure, having good linearity and temperature stability. The current acquisition module simultaneously monitors the working current of the cooling fan to evaluate the working state and efficiency of the cooling fan. The measurement range of the working current of the cooling fan is 0 to 10 amperes, and the accuracy is ±0.5%.

[0038] The pressure detection module uses a piezoresistive pressure sensor with a measurement range of 0 to 1 MPa, an accuracy of ±0.5%, and a sampling frequency of 1 time per second. The pressure sensor adopts a stainless steel isolation membrane structure, featuring good stability and reliability. The pressure detection module monitors the coolant pressure and pipeline pressure respectively, which are used to evaluate the working status and efficiency of the cooling system.

[0039] The heat dissipation control module includes a fan control unit and a coolant control unit. The fan control unit uses pulse width modulation technology to control the rotation speed of the cooling fan, with a speed regulation range of 0 to 3000 revolutions per minute and a speed regulation accuracy of ±1%. The coolant control unit uses variable frequency technology to control the rotation speed of the coolant pump, thereby adjusting the coolant flow rate. The flow rate adjustment range is 0 to 20 liters per minute, and the adjustment accuracy is ±2%.

[0040] The communication module adopts an Ethernet and 4G dual-channel communication method, supports TCP and IP protocols, and the communication rate can reach 100 Mb / s. The communication module realizes the remote transmission and monitoring of system data, supports real-time data upload and the issuance of remote control instructions. The system adopts an encrypted transmission method to ensure the security of data transmission.

[0041] The storage module uses a highly reliable solid-state drive with a storage capacity of 256 GB. It adopts power-off protection technology to ensure the safe storage of data. The storage module records temperature data, current data, pressure data, and control parameters during the system operation, supports regular data backup and historical data query.

[0042] The display module uses a 7-inch liquid crystal display with a resolution of 1024×600 pixels. It adopts a capacitive touch screen and supports human-computer interaction operations. The display module displays the system operation status, temperature data, current data, pressure data, and alarm information in real time, and supports the graphical display of historical data.

[0043] The power management module uses switching power supply technology. The input voltage range is 85 to 265 volts alternating current, the output voltage is 24 volts direct current, and the output power is 500 watts. The power management module has overvoltage protection, undervoltage protection, overcurrent protection, and short-circuit protection functions to ensure the stability and reliability of system power supply. The conversion efficiency of the power management module is greater than 92%, and the standby power consumption is less than 0.5 watts. The working voltage and working current of each module are monitored in real time, and once an abnormality is detected, the protection mechanism is immediately triggered.

[0044] The system control module adopts a hierarchical control architecture, including a data acquisition layer, a feature extraction layer, a control strategy layer, and an execution control layer. The data acquisition layer is responsible for the acquisition and preprocessing of temperature, current, and pressure data; the feature extraction layer is responsible for the extraction and analysis of temperature features; the control strategy layer is responsible for the formulation and optimization of heat dissipation strategies; the execution control layer is responsible for the execution and feedback of heat dissipation control instructions. The system control module uses a real-time operating system, ensuring the real-time performance and reliability of each functional module. Specifically, the specific implementation methods of the steps executed by the system control module are described in detail as follows.

[0045] The specific implementation method of step S01 is to collect the real-time temperature data of the charging interface and the radiator through the temperature acquisition module. First, preprocess the temperature data, including digital filtering and outlier removal. The collected temperature data is processed using a low-pass filter algorithm, and the cut-off frequency of the filter is set to 10 Hz, which can effectively filter out high-frequency noise. The outlier removal uses the median filtering method, and the sliding window length is set to 5 data points. Then, obtain the ambient temperature data. The ambient temperature is measured by an independent temperature sensor, installed 1 meter away from the charging pile, and the sampling frequency is 1 time per second. Next, calculate the difference between the temperature at each sampling point and the ambient temperature to form a temperature difference matrix, which records the temperature distribution characteristics of the system at different times and different positions. The number of rows of the temperature difference matrix is equal to the number of temperature sampling points, and the number of columns is equal to the length of the sampling time series. Each element in the matrix represents the temperature rise value of a certain sampling point relative to the ambient temperature at a specific moment. The length of the sampling time series is dynamically adjusted according to the system operation state, 60 seconds during normal operation, shortened to 30 seconds during fast charging, and extended to 120 seconds when the temperature is abnormal. The purpose of this step is to establish a real-time monitoring database of the system temperature distribution, providing data support for subsequent temperature feature extraction and control strategy optimization.

[0046] The specific implementation of step S02 is to perform matrix decomposition on the temperature difference matrix. Using the singular value decomposition algorithm, the temperature difference matrix is decomposed into the product of three matrices, namely the left singular matrix, the singular value diagonal matrix, and the transpose of the right singular matrix. The left singular matrix is the temperature space basis matrix, which characterizes the spatial distribution characteristics of the temperature sampling points; the transpose of the right singular matrix is the time feature matrix, which characterizes the variation characteristics of temperature over time. Before performing singular value decomposition, the temperature difference matrix is first normalized to map the element values in the matrix to the range of 0 to 1. Then a singular value threshold is set to determine the number of main features, and this threshold is set to 5% of the maximum singular value. The singular values less than the threshold are set to zero, and the singular values greater than the threshold are retained, thereby achieving matrix dimensionality reduction and feature extraction. The number of columns of the temperature space basis matrix is equal to the number of retained singular values, representing the main spatial feature patterns of the temperature distribution. The number of rows of the time feature matrix is also equal to the number of retained singular values, representing the variation law of these spatial feature patterns over time. The purpose of this step is to separate the spatial and time features of the temperature data for subsequent feature analysis and control strategy design.

[0047] The specific implementation of step S03 is to generate a dynamic convolution kernel based on the eigenvalues of the temperature space basis matrix. First, eigenvalue analysis is performed on the temperature space basis matrix, and the power iteration method is used to calculate the eigenvalues and eigenvectors of the matrix. The convergence threshold for eigenvalue calculation is set to 0.001, and the maximum number of iterations is 100 times. Then, the size of the dynamic convolution kernel is determined according to the distribution of the eigenvalues. The size of the convolution kernel is equal to the number of non-zero eigenvalues plus 1. The weight coefficients of the convolution kernel are obtained by normalizing the eigenvectors to ensure that the sum of the weight coefficients is 1. The dynamic convolution kernel adopts an adaptive structure and dynamically adjusts the shape and weight of the convolution kernel according to the variation characteristics of the temperature distribution. The minimum size of the convolution kernel is 3×3, and the maximum size is 7×7. The appropriate convolution kernel size is automatically selected according to the complexity of the temperature distribution. The purpose of this step is to construct a dynamic convolution operator that adapts to the temperature distribution characteristics and provide a core operation tool for the temperature feature extraction network.

[0048] The specific implementation of step S04 is to construct a temperature feature extraction network using a dynamic convolution kernel. This network adopts a multi-layer structure, including an input layer, a dynamic convolution layer, a pooling layer, a fully connected layer, and an output layer. The input layer receives the temperature difference matrix data, and the data dimension is the number of sampling points multiplied by the time series length. The dynamic convolution layer uses the previously generated dynamic convolution kernel for feature mapping, with a stride set to 1 and zero-padding for the padding method. The pooling layer uses a max-pooling operation, with a pooling window size of 2×2 and a stride of 2. The number of neurons in the fully connected layer is 128, and the activation function uses the rectified linear unit function. The dimension of the output layer is equal to the dimension of the temperature feature vector, and batch normalization is adopted. The network training uses the stochastic gradient descent algorithm, with an initial learning rate of 0.01, and the learning rate is dynamically adjusted using the cosine annealing strategy. The training batch size is 32, and the number of training epochs is 100. The purpose of this step is to establish a deep learning model for temperature features and achieve the automatic extraction and expression of temperature distribution features.

[0049] The specific implementation of step S05 is to construct a temperature difference change coefficient matrix based on the output of the temperature feature extraction network. First, the feature vector output by the network is subjected to dimensionality reduction processing using the principal component analysis algorithm, and the principal components with a cumulative variance contribution rate reaching 95% are retained. Then, the change rate of the temperature difference between adjacent moments is calculated using the central difference method, with a time step equal to the sampling period. The calculated change rate is smoothed using the exponential smoothing algorithm, with a smoothing coefficient set to 0.8. Finally, the processed data is organized into a temperature difference change coefficient matrix, where the number of rows of the matrix is equal to the number of temperature sampling points, and the number of columns is equal to the time series length minus 1. Each element in the matrix represents the temperature change rate of a certain sampling point between two adjacent moments. The purpose of this step is to obtain the dynamic characteristics of temperature changes and provide a basis for formulating a heat dissipation control strategy.

[0050] The specific implementation of step S06 is to obtain the charging current data through the current acquisition module and establish a temperature-power correlation matrix in combination with the temperature difference change coefficient matrix. First, the collected charging current data is preprocessed, including digital filtering and sampling synchronization. Digital filtering uses a Butterworth filter with a cut-off frequency of 5 Hz and a filter order of 4. Sampling synchronization is achieved through an interpolation algorithm to align the sampling time points of the current data with the temperature data. Then, the power data for each sampling point is calculated, and the power is equal to the product of the current and the voltage. The voltage data is provided by the system voltage acquisition unit with a sampling frequency of 2 times per second. Next, the power data is correlated with the temperature difference change coefficient using the Pearson correlation coefficient method to calculate the correlation between the temperature change and the power change at each temperature sampling point. Finally, the calculation results are organized into a temperature-power correlation matrix, where the number of rows of the matrix is equal to the number of temperature sampling points, and the number of columns is equal to the number of power sampling points. The purpose of this step is to establish a quantitative relationship between temperature changes and power and provide a basis for heat dissipation power allocation.

[0051] The specific implementation of step S07 is to establish a cooling capacity evaluation model based on the data collected by the pressure detection module through the thermodynamic state equations. First, the coolant pressure and pipeline pressure data are obtained, and the pressure data are filtered. The Kalman filter algorithm is used, with the measurement noise variance set to 0.01 and the process noise variance set to 0.001. Then, a coolant heat balance equation is established to calculate the heat dissipation power of the system. The input parameters of the equation include coolant flow rate, specific heat capacity, and inlet and outlet temperature difference. The specific heat capacity of the coolant is 4.2 kJ / (kg·°C). Next, a pipeline pressure loss equation is established to calculate the pipeline resistance. The input parameters of the equation include coolant flow rate, pipeline length, cross-sectional area, and coolant viscosity. The viscosity of the coolant varies with temperature and is obtained by using the look-up table method. Finally, a radiator heat transfer equation is established to calculate the heat dissipation power of the radiator. The input parameters of the equation include radiator surface area, convective heat transfer coefficient, surface temperature, and ambient temperature. The convective heat transfer coefficient is related to the wind speed and is obtained by fitting experimental data. The purpose of this step is to establish a physical model of the system cooling capacity and provide theoretical support for the heat dissipation control strategy.

[0052] The specific implementation of step S08 is to calculate the required heat dissipation power of each monitoring point based on the temperature difference change coefficient matrix and the cooling capacity evaluation model. First, the temperature change trend is determined according to the temperature difference change coefficient. When the change coefficient is greater than 0.5, it indicates that the temperature is rising rapidly and the heat dissipation power needs to be increased; when the change coefficient is less than -0.5, it indicates that the temperature is dropping rapidly and the heat dissipation power can be reduced. Then, combined with the temperature-power correlation matrix, the expected heat dissipation power of each monitoring point is calculated. The calculation process takes into account the temperature change rate and the difference between the current temperature value and the target temperature value. The expected heat dissipation power is proportional to the temperature difference and the temperature change rate, and the proportionality coefficient is determined by experiments. Next, according to the cooling capacity evaluation model, it is judged whether the system can provide sufficient heat dissipation capacity. If the required heat dissipation power exceeds the maximum heat dissipation capacity of the system, the power is distributed according to the temperature control priority. The temperature control priority is determined by the importance of the monitoring points, and the priority of the charging interface is higher than that of the radiator. The purpose of this step is to achieve a reasonable distribution of heat dissipation power and ensure the temperature control effect of key parts.

[0053] The specific implementation of step S09 is to adjust the rotational speed of the cooling fan and the coolant flow rate through the heat dissipation control module to achieve multi-point temperature control. First, based on the calculated heat dissipation power demand, the target values of the fan rotational speed and the coolant flow rate are determined. The fan rotational speed has a non-linear relationship with the heat dissipation power, and the target rotational speed is determined by the look-up table method. The coolant flow rate has an approximately linear relationship with the heat dissipation power, and the target flow rate is directly calculated proportionally. Then, a fuzzy control algorithm is used to achieve the dynamic adjustment of the fan rotational speed. The input variables of the fuzzy controller are the temperature deviation and the temperature change rate, and the output variable is the rotational speed adjustment amount. The number of fuzzy rules is 49, covering various temperature change situations. Next, a proportional-integral control algorithm is used to achieve the precise control of the coolant flow rate. The proportional coefficient is 0.5, the integral coefficient is 0.1, and the trapezoidal integration method is used. The control period is 1 second, and the control output is updated every control period. The purpose of this step is to achieve the closed-loop adjustment of heat dissipation control and ensure the accuracy and stability of temperature control.

[0054] Step S10 is an optional step. The specific implementation is to perform real-time analysis on the system operation data, establish a heat dissipation efficiency evaluation index, and dynamically optimize the heat dissipation strategy. First, calculate the heat dissipation efficiency indexes of the system, including temperature control accuracy, response time, energy consumption efficiency, etc. The temperature control accuracy is represented by the root mean square value of the temperature deviation, and the target value is less than 1 degree Celsius. The response time is defined as the time required for the temperature to reach 90% of the target value, and the target value is less than 30 seconds. The energy consumption efficiency is represented by the input power required per unit heat dissipation power, and the target value is less than 0.1. Then, based on these indexes, a comprehensive evaluation function is constructed using the weighted summation method, and the weight coefficients are determined by the analytic hierarchy process. Next, the recursive least squares algorithm is used to perform online identification on the system model, and the model parameters are updated in real time. The forgetting factor is set to 0.95. Finally, the model predictive control method is used to optimize the heat dissipation strategy. The prediction time domain is 10 sampling periods, the control time domain is 3 sampling periods, and the optimization goal is a compromise between temperature control accuracy and energy consumption efficiency. The purpose of this step is to achieve the adaptive optimization of heat dissipation control and improve the overall performance of the system.

[0055] In the present invention, temperature data is acquired through the temperature acquisition module. To eliminate the influence of ambient temperature fluctuations, the temperature difference form is adopted. The construction of this matrix goes through the following steps: (1) Determine the sampling point layout: Arrange 3 temperature sensors at key positions of the charging interface, 2 temperature sensors on the surface of the radiator, and 1 temperature sensor in the environment; (2) Synchronous sampling: All sensors perform sampling at the same time interval (0.25 seconds); (3) Data preprocessing: Use a first-order low-pass filter to filter the original data, and the filter coefficient It is determined through a step response experiment. The experimental method is to suddenly increase the heat load under steady-state conditions, record the temperature response curve, and obtain the optimal filtering coefficient through least squares fitting; (4)Construct a difference matrix: Subtract the ambient temperature from the processed temperature data to form a temperature difference matrix.

[0056] The advantages of this matrix construction method are as follows: It can eliminate the interference of ambient temperature fluctuations; it can reflect the relative temperature rise of the system; it is convenient for subsequent feature extraction.

[0057] The establishment process of the temperature difference matrix is specifically expressed as follows: ; In the formula, is the temperature difference matrix; represents the temperature value of the th sampling point at the th moment, with the unit of degree Celsius; is the ambient temperature, with the unit of degree Celsius; is the number of temperature sampling points, and the value range is from 3 to 12; is the length of the sampling time series, and the value range is from 30 to 120.

[0058] The preprocessing process of the temperature data is as follows: ; In the formula, is the filtered temperature value; is the filtering coefficient, and the value range is from 0.1 to 0.9; is the filtered temperature value at the previous moment.

[0059] The singular value decomposition process of the temperature difference matrix is specifically expressed as follows: ; In the formula, is the left singular matrix, that is, the temperature space basis matrix; is the singular value diagonal matrix; is the transpose of the right singular matrix, that is, the time feature matrix.

[0060] The temperature space basis matrix is specifically expressed as follows: ; In the formula, represents the weight of the th sampling point in the th spatial feature mode; is the number of main features, determined by the singular value threshold.

[0061] The time feature matrix The specific representation is as follows: ; In the formula, represents the intensity of the th feature pattern at the th moment.

[0062] The calculation of the size of the dynamic convolution kernel is specifically represented as follows: ; In the formula, is the convolution kernel size; is the th singular value; is the singular value threshold, and its value is 5% of the maximum singular value; is the indicator function.

[0063] The calculation of the weights of the dynamic convolution kernel is specifically represented as follows: ; In the formula, is the weight at position in the convolution kernel; is the Euclidean distance from position to the center of the convolution kernel; is the Gaussian kernel parameter, which is related to the size of the convolution kernel.

[0064] The output of the dynamic convolution layer of the temperature feature extraction network is specifically represented as follows: ; In the formula, is the output at position of the th layer feature map; is the weight of the dynamic convolution kernel at position ; is the input feature map of the th layer; is the bias term.

[0065] The output of the pooling layer is specifically represented as follows: ; In the formula, is the output of the pooling layer at position ; is the 2×2 neighborhood of position .

[0066] The output of the fully connected layer is specifically represented as follows: ; In the formula, is the The output of the fully connected layer; is the weight matrix; is the output of the previous layer; is the bias vector; is the rectified linear unit activation function.

[0067] The construction process of the temperature difference change coefficient matrix is specifically expressed as follows: ; In the formula, is the th sampling point at the th moment of the temperature change rate; is the sampling time interval, with the unit of seconds; is the feature vector; is the attenuation coefficient, and the value range is from 0.1 to 1; represents the Euclidean norm.

[0068] The construction process of the temperature-power correlation matrix is specifically expressed as follows: ; In the formula, is the th temperature sampling point and the th power sampling point of the correlation coefficient; is the th power sampling point at the th moment of the power value; and are the means of the corresponding sequences respectively.

[0069] The specific representation of the thermodynamic state equation set is as follows: Coolant heat balance equation: ; In the formula, is the heat dissipation power of the cooling system, with the unit of watt; is the coolant density, with the unit of kilogram per cubic meter; is the specific heat capacity, with the unit of joule per kilogram per degree Celsius; is the volume flow rate, with the unit of cubic meter per second; is the inlet and outlet temperature difference, with the unit of degree Celsius; is the pressure correction coefficient; is the pressure loss, with the unit of pascal.

[0070] Pipeline pressure loss equation: ; In the formula, is the dynamic viscosity, with the unit of pascal second; is the pipeline length, in meters; is the pipeline radius, in meters; is the turbulence correction coefficient; is the Reynolds number.

[0071] Heat transfer equation of the radiator: ; In the formula, is the basic heat transfer coefficient, in watts per square meter per degree Celsius; is the heat dissipation area, in square meters; is the surface temperature, in degrees Celsius; is the wind speed influence coefficient; is the wind speed, in meters per second; is the wind speed exponent, with a value range of 0.5 to 0.8.

[0072] The calculation of temperature control accuracy is specifically expressed as follows: ; In the formula, is the temperature control accuracy, in degrees Celsius; is the actual temperature value; is the target temperature value; is the number of sampling points.

[0073] The calculation of energy consumption efficiency is specifically expressed as follows: ; In the formula, is the energy consumption efficiency; is the system heat dissipation power, in watts; is the input power, in watts.

[0074] The comprehensive evaluation function is specifically expressed as follows: ; In the formula, is the comprehensive evaluation index; , , are the weight coefficients, and satisfy ; is the response time, in seconds.

[0075] Specifically, the principle of the present invention is as follows: The technical principle of the present invention is based on multiple innovative theories. First of all, the decomposition principle of the temperature difference matrix is based on the singular value decomposition theory of matrices. By decomposing the complex temperature change pattern into two independent dimensions, namely spatial features and temporal features, the system can separately capture the features of temperature in spatial distribution and temporal evolution. This decomposition method not only reduces the complexity of the data but also retains the essential features of temperature changes, providing a good basis for subsequent feature extraction.

[0076] Secondly, the generation mechanism of the dynamic convolution kernel is based on the eigenvalue analysis of the temperature space basis matrix. By calculating the singular values of the temperature space basis matrix, the system can determine the number of main features of the temperature change pattern and dynamically adjust the size of the convolution kernel accordingly. This adaptive mechanism ensures the accuracy of feature extraction, enabling the system to adjust the feature extraction strategy according to the actual temperature change pattern.

[0077] Finally, the establishment of the thermodynamic state equation system is based on the classical heat transfer theory. By considering the heat balance of the coolant, the pressure loss of the pipeline, and the heat exchange process of the radiator, a complete thermodynamic model is established. This model not only considers the mutual relationship between various physical quantities but also reflects the dynamic characteristics of the system, providing a reliable theoretical support for temperature control.

[0078] Next, a specific Embodiment 1 of the present invention is provided, and the specific implementation of each step in this Embodiment 1 is described in detail as follows.

[0079] The main control module of the charging pile system in this Embodiment 1 uses a high-performance ARM processor STM32F746 with a main frequency of up to 216 MHz, built-in 2 MB flash memory and 384 KB static random access memory. This processor has rich peripheral interfaces and powerful real-time processing capabilities, which can meet the real-time requirements of the system for temperature data acquisition, feature extraction, and control strategy execution. The built-in control chip adopts a dual-core architecture, where one core is specifically responsible for data acquisition and processing, and the other core is responsible for the execution of control strategies, thus ensuring the real-time response ability of the system. The temperature of the processor in the main control module does not exceed 85 degrees Celsius, the working voltage is 3.3 volts, and the power consumption is less than 1.5 watts. The system clock uses an external crystal oscillator with a frequency of 25 MHz and integrates a watchdog timer, which can automatically reset when the system is abnormal.

[0080] The temperature acquisition module uses a high-precision PT100 platinum resistance temperature sensor with a temperature measurement range from -40 degrees Celsius to 200 degrees Celsius and an accuracy of up to ±0.1 degree Celsius. Each charging interface is equipped with 3 temperature sensors, which are installed at different positions of the charging interface to achieve multi-point monitoring of the charging interface temperature. The radiator temperature is monitored using an NTC thermistor with a temperature measurement range from -20 degrees Celsius to 120 degrees Celsius and an accuracy of ±0.5 degree Celsius. The sampling frequency of the temperature acquisition module is 4 times per second. A 24-bit analog-to-digital converter is used to convert the temperature signal, ensuring the acquisition accuracy of the temperature data. The output signal of the temperature sensor is processed by an amplifier circuit with an amplification factor of 100 times, and then filtered through an anti-aliasing filter with a cut-off frequency of 50 Hz. The filtered signal is conditioned by a precision operational amplifier and finally sent to the analog-to-digital converter for digital processing.

[0081] The current acquisition module uses a Hall current sensor with a measurement range from 0 to 500 amperes, an accuracy of ±1%, and a sampling frequency of 2 times per second. The current sensor adopts a closed-loop structure, having good linearity and temperature stability. The current acquisition module simultaneously monitors the working current of the cooling fan to evaluate the working state and efficiency of the cooling fan. The measurement range of the working current of the cooling fan is from 0 to 10 amperes, and the accuracy is ±0.5%. The output signal of the current sensor is processed by a signal conditioning circuit, including zero calibration, gain adjustment, and temperature compensation. The zero drift is less than 0.1%, and the temperature coefficient is less than 0.02% per degree Celsius. The signal conditioning circuit adopts a dual operational amplifier structure with a common-mode rejection ratio greater than 80 dB.

[0082] The pressure detection module uses a piezoresistive pressure sensor with a measurement range from 0 to 1 MPa, an accuracy of ±0.5%, and a sampling frequency of 1 time per second. The pressure sensor adopts a stainless steel isolation membrane structure, having good stability and reliability. The pressure detection module respectively monitors the coolant pressure and the pipeline pressure to evaluate the working state and efficiency of the cooling system. The excitation voltage of the pressure sensor is 10 V DC, and the output signal is a 4-20 mA standard current signal. The temperature drift of the sensor is less than 0.02% per degree Celsius, and the long-term stability is better than 0.1% per year. The signal conditioning circuit uses a high-precision instrumentation amplifier with a common-mode rejection ratio greater than 100 dB.

[0083] The heat dissipation control module includes a fan control unit and a coolant control unit. The fan control unit uses pulse width modulation technology to control the rotational speed of the cooling fan. The speed regulation range is from 0 to 3000 revolutions per minute, and the speed regulation accuracy is ±1%. The coolant control unit uses variable frequency technology to control the rotational speed of the coolant pump, thereby adjusting the coolant flow rate. The flow rate adjustment range is from 0 to 20 liters per minute, and the adjustment accuracy is ±2%. The cooling fan is driven by a DC brushless motor with a rated power of 100 watts and a starting torque greater than 0.5 Newton meters. The coolant pump uses magnetic drive technology, has the characteristic of zero leakage, and a maximum working pressure of 0.6 MPa.

[0084] The communication module adopts an Ethernet and 4G dual-channel communication mode, supports TCP and IP protocols, and the communication rate can reach 100 Mb / s. The 4G module supports multiple frequency bands, including FDD-LTE and TDD-LTE. The transmission rate can reach 50 Mb / s upstream and 150 Mb / s downstream. The Ethernet interface uses an 8-pin RJ45 connector and supports the auto-negotiation function, which can automatically select a transmission rate of 10 / 100 Mb / s. The communication data is encrypted using the AES-256 encryption algorithm to ensure the security of data transmission. The system supports the remote upgrade function and can update the software through the network.

[0085] The storage module uses a highly reliable solid-state drive with a storage capacity of 256 GB. It adopts power-off protection technology to ensure the secure storage of data. The storage module uses the FAT32 file system, which supports fast data reading and writing and retrieval. The data storage adopts a circular overwrite method. When the storage space is insufficient, the earliest data is automatically deleted. The system automatically backs up key data once an hour, and the backup files are stored in an independent storage area. The data read and write speed is greater than 100 megabytes per second, and it supports the power-off protection function.

[0086] The display module uses a 7-inch liquid crystal display with a resolution of 1024×600 pixels. It uses a capacitive touch screen and supports multi-touch operations. The display module integrates a backlight control circuit, which can automatically adjust the screen brightness according to the ambient light intensity. The display interface adopts a hierarchical design, including a main interface, a monitoring interface, a parameter setting interface, and an alarm interface. The interface refresh rate is 60 Hz, and the touch response time is less than 50 milliseconds. The display module supports bilingual display in Chinese and English, and the font size is adjustable.

[0087] The power management module adopts switching power supply technology. The input voltage range is 85 to 265 volts AC, the output voltage is 24 volts DC, and the output power is 500 watts. The power module uses LLC resonant conversion technology to achieve soft switching, improve efficiency, and reduce electromagnetic interference. The output voltage ripple is less than 50 mV, and the dynamic response time is less than 1 ms. The power module has overvoltage protection, undervoltage protection, overcurrent protection, and short-circuit protection functions. The protection points are accurate, and the response time is less than 10 μs. The power efficiency is greater than 92% when the load is above 50%, and still maintains an efficiency of more than 85% when the load is 10%.

[0088] The system control module executes the following steps.

[0089] The specific implementation of step S01 is to collect the real-time temperature data of the charging interface and the radiator through the temperature acquisition module. This step first performs preprocessing of the temperature data, including digital filtering and outlier rejection. The collected temperature data is processed using a low-pass filtering algorithm. The cut-off frequency of the filter is set to 10 Hz, which can effectively filter out high-frequency noise. Outlier rejection uses the median filtering method, and the sliding window length is set to 5 data points. In this step, a first-order low-pass filter is used to filter the original data, and the filtering coefficient is determined through a step response experiment. The filtering process satisfies the formula: . Among them, the value range of the filtering coefficient is 0.1 to 0.9, and the specific value is determined through an experimental method. The experimental method is to suddenly apply a heating load under steady-state conditions, record the temperature response curve, and obtain the optimal filtering coefficient through least squares fitting. Then, the ambient temperature data is obtained. The ambient temperature is measured by an independent temperature sensor, and the installation position is 1 meter away from the charging pile, and the sampling frequency is 1 time per second. Then, calculate the difference between the temperature at each sampling point and the ambient temperature to form a temperature difference matrix. The temperature difference matrix satisfies the formula: . This matrix records the temperature distribution characteristics of the system at different times and different positions. In the matrix, represents the temperature value of the th sampling point at the th moment, in degrees Celsius, is the ambient temperature, in degrees Celsius, is the number of temperature sampling points, and the value range is 3 to 12, is the length of the sampling time series, and the value range is 30 to 120. The length of the sampling time series is dynamically adjusted according to the system operation state. It is 60 seconds during normal operation, shortened to 30 seconds during fast charging, and extended to 120 seconds when the temperature is abnormal. The purpose of this step is to establish a real-time monitoring database of the system temperature distribution to provide data support for subsequent temperature feature extraction and control strategy optimization.

[0090] The specific implementation of step S02 is to perform matrix decomposition on the temperature difference matrix. Using the singular value decomposition algorithm, the temperature difference matrix is decomposed into the product of three matrices, satisfying the formula: . Among them is the left singular matrix, that is, the temperature space basis matrix, is the singular value diagonal matrix, is the transpose of the right singular matrix, that is, the time feature matrix. The temperature space basis matrix is specifically represented as: , where represents the weight of the th sampling point in the th spatial feature mode, is the number of main features. The time feature matrix is specifically represented as: , where represents the intensity of the th feature mode at the th moment. Before performing singular value decomposition, first normalize the temperature difference matrix to map the element values in the matrix to the range of 0 to 1. Then set the singular value threshold to determine the number of main features, and this threshold is set to 5% of the maximum singular value. Set the singular values less than the threshold to zero and retain the singular values greater than the threshold, thereby achieving matrix dimensionality reduction and feature extraction. The number of columns of the temperature space basis matrix is equal to the number of retained singular values, representing the main spatial feature modes of the temperature distribution. The number of rows of the time feature matrix is also equal to the number of retained singular values, representing the variation law of these spatial feature modes over time. The purpose of this step is to separate the spatial features and time features of the temperature data, facilitating subsequent feature analysis and control strategy design.

[0091] The specific implementation of step S03 is to generate a dynamic convolution kernel based on the eigenvalues of the temperature space basis matrix. First, perform eigenvalue analysis on the temperature space basis matrix, and use the power iteration method to calculate the eigenvalues and eigenvectors of the matrix. The convergence threshold for eigenvalue calculation is set to 0.001, and the maximum number of iterations is 100 times. Then determine the size of the dynamic convolution kernel according to the distribution of eigenvalues, and the convolution kernel size satisfies the formula: . Among them is the convolution kernel size, is the th singular value, is the singular value threshold, with a value of 5% of the maximum singular value, is the indicator function. The weight coefficients of the convolution kernel are obtained by normalizing the eigenvectors, and the weight calculation satisfies the formula: . Among them is the weight at position in the convolution kernel, is the position The Euclidean distance to the center of the convolution kernel is the Gaussian kernel parameter, which is related to the size of the convolution kernel. The dynamic convolution kernel adopts an adaptive structure and dynamically adjusts the shape and weight of the convolution kernel according to the change characteristics of the temperature distribution. The minimum size of the convolution kernel is 3×3, and the maximum size is 7×7. The appropriate convolution kernel size is automatically selected according to the complexity of the temperature distribution. The purpose of this step is to construct a dynamic convolution operator that adapts to the temperature distribution characteristics and provide a core operation tool for the temperature feature extraction network.

[0092] The specific implementation of step S04 is to construct a temperature feature extraction network using a dynamic convolution kernel. This network adopts a multi-layer structure, including an input layer, a dynamic convolution layer, a pooling layer, a fully connected layer, and an output layer. The output of the dynamic convolution layer satisfies the formula: . Where is the output of the feature map of the th layer at position , is the weight of the dynamic convolution kernel at position , is the input feature map of the th layer, is the bias term. The output of the pooling layer satisfies the formula: . Where is the output of the pooling layer at position , is the 2×2 neighborhood of position . The output of the fully connected layer satisfies the formula: . Where is the output of the th layer of the fully connected layer, is the weight matrix, is the output of the previous layer, is the bias vector, is the rectified linear unit activation function. The network training adopts the stochastic gradient descent algorithm, the initial value of the learning rate is 0.01, and the cosine annealing strategy is used to dynamically adjust the learning rate. The training batch size is 32, and the number of training epochs is 100. The purpose of this step is to establish a deep learning model of temperature features and realize the automatic extraction and expression of temperature distribution features.

[0093] The specific implementation of step S05 is to construct a temperature difference change coefficient matrix based on the output of the temperature feature extraction network, which satisfies the formula: . Where is the temperature change rate of the th sampling point at the th moment, is the sampling time interval, with the unit of second, is the feature vector, is the attenuation coefficient, with a value range of 0.1 to 1, represents the Euclidean norm. First, the feature vectors output by the network are dimensionally reduced. The principal component analysis algorithm is used to retain the principal components with a cumulative variance contribution rate reaching 95%. Then, the change rate of the temperature difference between adjacent moments is calculated. The central difference method is used with a time step equal to the sampling period. The calculated change rate is smoothed using the exponential smoothing algorithm with a smoothing coefficient set to 0.8. Finally, the processed data is organized into a temperature difference change coefficient matrix. The number of rows of the matrix is equal to the number of temperature sampling points, and the number of columns is equal to the length of the time series minus 1. Each element in the matrix represents the temperature change rate of a certain sampling point between two adjacent moments. The purpose of this step is to obtain the dynamic characteristics of temperature changes and provide a basis for formulating heat dissipation control strategies.

[0094] The specific implementation of step S06 is to obtain the charging current data through the current acquisition module and establish a temperature-power correlation matrix in combination with the temperature difference change coefficient matrix. The construction of the temperature-power correlation matrix satisfies the formula: . Where is the correlation coefficient between the th temperature sampling point and the th power sampling point, is the power value of the th power sampling point at the th moment, and are the means of the corresponding sequences respectively. First, the collected charging current data is preprocessed, including digital filtering and sampling synchronization. The Butterworth filter is used for digital filtering with a cut-off frequency of 5 Hz and a filter order of 4. Sampling synchronization is achieved through the interpolation algorithm to align the sampling time points of the current data with the temperature data. Then, the power data of each sampling point is calculated. The power is equal to the product of the current and the voltage. The voltage data is provided by the system voltage acquisition unit with a sampling frequency of 2 times per second. Then, the power data is correlated with the temperature difference change coefficient. The Pearson correlation coefficient method is used to calculate the correlation between the temperature change and the power change of each temperature sampling point. Finally, the calculation results are organized into a temperature-power correlation matrix. The number of rows of the matrix is equal to the number of temperature sampling points, and the number of columns is equal to the number of power sampling points. The purpose of this step is to establish a quantitative relationship between temperature changes and power dissipation and provide a basis for heat dissipation power allocation.

[0095] The specific implementation of step S07 is to establish a cooling capacity evaluation model based on the data collected by the pressure detection module through the thermodynamic state equation set. The coolant heat balance equation satisfies the formula: . Where is the heat dissipation power of the cooling system in watts, is the coolant density in kilograms per cubic meter, is the specific heat capacity, with the unit of joule per kilogram per degree Celsius, is the volume flow rate, with the unit of cubic meter per second, is the temperature difference between the inlet and outlet, with the unit of degree Celsius, is the pressure correction coefficient, is the pressure loss, with the unit of Pascal. The pipeline pressure loss equation satisfies the formula: . Among them is the dynamic viscosity, with the unit of Pascal second, is the pipeline length, with the unit of meter, is the pipeline radius, with the unit of meter, is the turbulence correction coefficient, is the Reynolds number. The radiator heat transfer equation satisfies the formula: . Among them is the basic heat transfer coefficient, with the unit of watt per square meter per degree Celsius, is the heat dissipation area, with the unit of square meter, is the surface temperature, with the unit of degree Celsius, is the wind speed influence coefficient, is the wind speed, with the unit of meter per second, is the wind speed exponent, and its value range is from 0.5 to 0.8. First, obtain the coolant pressure and pipeline pressure data, filter the pressure data, adopt the Kalman filter algorithm, set the measurement noise variance to 0.01, and the process noise variance to 0.001. Then establish the coolant heat balance equation and calculate the heat dissipation power of the system. The input parameters of the equation include the coolant flow rate, specific heat capacity, and temperature difference between the inlet and outlet. Among them, the specific heat capacity of the coolant is 4.2 kJ / kg·°C. Next, establish the pipeline pressure loss equation and calculate the pipeline resistance. The input parameters of the equation include the coolant flow rate, pipeline length, cross-sectional area, and coolant viscosity. Among them, the viscosity of the coolant changes with temperature and is obtained by using the table lookup method. Finally, establish the radiator heat transfer equation and calculate the heat dissipation power of the radiator. The input parameters of the equation include the radiator surface area, convective heat transfer coefficient, surface temperature, and ambient temperature. Among them, the convective heat transfer coefficient is related to the wind speed and is obtained by fitting the experimental data. The purpose of this step is to establish a physical model of the system cooling capacity and provide theoretical support for the heat dissipation control strategy.

[0096] The specific implementation of step S08 is to calculate the required heat dissipation power of each monitoring point based on the temperature difference change coefficient matrix and the cooling capacity evaluation model. First, determine the temperature change trend according to the temperature difference change coefficient. When the change coefficient is greater than 0.5, it indicates that the temperature is rising rapidly and the heat dissipation power needs to be increased. When the change coefficient is less than -0.5, it indicates that the temperature is dropping rapidly and the heat dissipation power can be reduced. Then, combined with the temperature-power correlation matrix, calculate the expected heat dissipation power of each monitoring point. The calculation process takes into account the temperature change rate, the difference between the current temperature value and the target temperature value. To accurately evaluate the temperature control accuracy, the root mean square error calculation formula is used: . Where is the temperature control accuracy, in degrees Celsius, is the actual temperature value, is the target temperature value, is the number of sampling points. The expected heat dissipation power is proportional to the temperature difference and proportional to the temperature change rate. The proportionality coefficient is determined through experiments. The energy consumption efficiency of the heat dissipation system satisfies the formula: . Where is the energy consumption efficiency, is the system heat dissipation power, in watts, is the input power, in watts. Then, according to the cooling capacity evaluation model, judge whether the system can provide sufficient heat dissipation capacity. If the required heat dissipation power exceeds the maximum heat dissipation capacity of the system, power distribution is carried out according to the temperature control priority. The temperature control priority is determined by the importance of the monitoring points. The priority of the charging interface is higher than that of the radiator. The purpose of this step is to achieve reasonable distribution of the heat dissipation power and ensure the temperature control effect of key parts.

[0097] The specific implementation of step S09 is to regulate the rotational speed of the cooling fan and the coolant flow rate through a heat dissipation control module to achieve multi-point temperature control. First, based on the calculated heat dissipation power demand, the target values of the fan rotational speed and the coolant flow rate are determined. The fan rotational speed has a non-linear relationship with the heat dissipation power, and the target rotational speed is determined by using a look-up table method. The coolant flow rate has an approximately linear relationship with the heat dissipation power, and the target flow rate is directly calculated in proportion. Then, a fuzzy control algorithm is used to achieve the dynamic regulation of the fan rotational speed. The input variables of the fuzzy controller are the temperature deviation and the temperature change rate, and the output variable is the rotational speed adjustment amount. The number of fuzzy rules is 49, covering various temperature change situations. The fuzzy controller adopts a neighboring triangular membership function. The domain of the temperature deviation is from -5 to +5 degrees Celsius, the domain of the temperature change rate is from -2 to +2 degrees Celsius per second, and the domain of the rotational speed adjustment amount is from -500 to +500 revolutions per minute. Next, a proportional-integral control algorithm is used to achieve the precise control of the coolant flow rate. The proportional coefficient is 0.5, the integral coefficient is 0.1, and the trapezoidal integration method is adopted. The control period is 1 second, and the control output is updated once per control period. The purpose of this step is to achieve the closed-loop regulation of heat dissipation control and ensure the accuracy and stability of temperature control.

[0098] Step S10 is an optional step. Its specific implementation is to perform real-time analysis on the system operation data, establish a heat dissipation efficiency evaluation index, and dynamically optimize the heat dissipation strategy. The comprehensive evaluation function satisfies the formula: . Where is the comprehensive evaluation index, , , are the weight coefficients and satisfy , is the response time, with the unit of seconds. First, the heat dissipation efficiency indexes of the system are calculated, including temperature control accuracy, response time, energy consumption efficiency, etc. The temperature control accuracy is represented by the root mean square value of the temperature deviation, and the target value is less than 1 degree Celsius. The response time is defined as the time required for the temperature to reach 90% of the target value, and the target value is less than 30 seconds. The energy consumption efficiency is represented by the input power required for unit heat dissipation power, and the target value is less than 0.1. Then, a comprehensive evaluation function is constructed based on these indexes, and the weighted summation method is adopted. The weight coefficients are determined by the analytic hierarchy process. Next, the recursive least squares algorithm is used to perform online identification on the system model, and the model parameters are updated in real time. The forgetting factor is set to 0.95. Finally, the model predictive control method is used to optimize the heat dissipation strategy. The prediction time domain is 10 sampling periods, the control time domain is 3 sampling periods, and the optimization goal is a compromise between temperature control accuracy and energy consumption efficiency. The purpose of this step is to achieve the adaptive optimization of heat dissipation control and improve the overall performance of the system.

[0099] To better understand and implement the present invention, the following provides Example 2 of a specific application scenario of the present invention: A research team implemented a charging pile system supporting multi-point automatic cooling at a new energy vehicle charging station. There are 20 charging piles in this charging station, and each charging pile is equipped with an independent temperature monitoring and heat dissipation control system. The charging piles adopt DC fast charging mode, and the maximum charging power is 200 kilowatts. During the implementation process, the research team conducted detailed performance tests and data collection on one of the charging piles, and the collection period was 30 days. The temperature sampling points of the charging pile are distributed as shown in Table 1.

[0100] Table 1 Distribution Table of Temperature Sampling Points of Charging Piles

[0101] The main control module of this charging pile system uses an STM32F746 processor, with the main frequency set to 216 megahertz, and it has 2 megabytes of flash memory and 384 kilobytes of static random access memory built-in. The temperature acquisition module uses a 24-bit analog-to-digital converter ADS1256, and the sampling frequency is 4 times per second. During a typical charging process, the statistical characteristics of the temperature sampling data are shown in Table 2.

[0102] Table 2 Statistical Characteristics Table of Temperature Sampling Data

[0103] During the operation of the system, first, the collected temperature data is preprocessed, and a first-order low-pass filter is used for filtering. The filtering coefficient is determined to be 0.85 through experiments. The dimension of the temperature difference matrix is 5×240, where 5 is the number of temperature sampling points and 240 is the sampling time series length (1-minute sampling period). After performing singular value decomposition on the temperature difference matrix, the distribution of the main eigenvalues is shown in Table 3.

[0104] Table 3 Distribution Table of Main Eigenvalues of Temperature Difference Matrix

[0105] Figure 3 shows the eigenvalue distribution of the temperature difference matrix and its cumulative contribution rate. The left vertical axis represents the eigenvalue magnitude, the right vertical axis represents the cumulative contribution rate, and the horizontal axis represents the eigenvalue serial number. This figure visually shows the distribution of the main eigenvalues and their contribution to the overall variance. According to the eigenvalue distribution, the size of the dynamic convolution kernel is determined to be 5×5. The convolution kernel weight matrix is calculated through a Gaussian function, and the Gaussian kernel parameter is set to 1.5. The structural configuration of the temperature feature extraction network is shown in Table 4.

[0106] Table 4 Structural Configuration Table of Temperature Feature Extraction Network

[0107] The performance parameters of the heat dissipation control system are shown in Table 5 below.

[0108] Table 5 Performance Parameter Table of Heat Dissipation Control System

[0109] Figure 4 A bar chart is used to compare and display the design values and measured values of the performance parameters of the heat dissipation control system. The horizontal axis represents different performance parameters, and the vertical axis represents the parameter values (standardized units). Through this chart, the differences between the design values and measured values can be visually compared. After the system has been running for 30 days, the key performance indicators collected are shown in Table 6.

[0110] Table 6 Statistical Table of System Operation Performance Indicators

[0111] Figure 5 The time-series changes of three key performance indicators (temperature control deviation, system response time, and energy consumption efficiency) during the system operation are shown. The horizontal axis represents the operation time (days), and the vertical axis represents the performance indicator values. This chart reflects the stability and performance fluctuations of the system operation.

[0112] Traditional charging pile temperature control systems mainly adopt a fixed threshold control method, that is, when the temperature exceeds a preset threshold, the heat dissipation system is started, and when the temperature drops to another preset threshold, the heat dissipation system is turned off. This method has the following problems: it cannot achieve precise temperature control, the system response is lagged, the energy utilization efficiency is low, and temperature fluctuations are likely to occur. The adaptive control method based on temperature feature extraction adopted in the present invention analyzes the temperature distribution characteristics in real time, predicts the temperature change trend, and dynamically adjusts the heat dissipation strategy, significantly improving the system performance. Compared with the traditional method, the present invention has achieved technological progress in the following aspects: the temperature control accuracy has been improved by 45%, from ±1.8 degrees Celsius of the traditional method to ±1.0 degrees Celsius; the system response time has been shortened by 35%, from 44 seconds of the traditional method to 28.5 seconds; the energy consumption efficiency has been increased by 25%, from 0.70 of the traditional method to 0.88; the system reliability has been increased by 0.15 percentage points, from 99.80% of the traditional method to 99.95%. These improvements enable the charging pile to better meet the requirements of high-power fast charging, improve the charging efficiency and safety, and reduce the operating cost.

[0113] It should be noted that the detailed explanations of the variables involved in the present invention are shown in Table 7 below.

[0114] Table 7 Variable Explanation Table

[0115] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention.

Claims

1. A charging pile system supporting multi-point automatic temperature reduction, characterized in that, It has a main control module. The main control module is built-in with a control chip, and the control chip is provided with a system control module. The system control module performs matrix decomposition based on the temperature difference matrix to obtain a temperature spatial basis matrix and a time feature matrix, generates a dynamic convolution kernel using a convolution kernel calculation function, constructs a temperature feature extraction network using the dynamic convolution kernel, constructs a temperature difference change coefficient matrix based on the feature vector output by the temperature feature extraction network, obtains charging current data, establishes a temperature-power correlation matrix in combination with the temperature difference change coefficient matrix, establishes a cooling capacity evaluation model through a thermodynamic state equation set according to the coolant pressure and pipeline pressure, calculates the heat dissipation power required for each monitoring point based on the temperature difference change coefficient matrix and the cooling capacity evaluation model, and adjusts the rotation speed of the cooling fan and the coolant flow rate through a heat dissipation control module to achieve multi-point temperature control.

2. The charging pile system supporting multi-point automatic temperature reduction according to claim 1, characterized in that, It also includes a temperature acquisition module, a current acquisition module, a pressure detection module, a heat dissipation control module, a communication module, a storage module, a display module, and a power management module. The main control module is electrically connected to the temperature acquisition module, the current acquisition module, the pressure detection module, the heat dissipation control module, the communication module, the storage module, the display module, and the power management module respectively; among them, the sampling frequency of the temperature acquisition module for temperature parameters is 4 times per second, the sampling frequency of the current acquisition module is 2 times per second, and the sampling frequency of the pressure detection module is 1 time per second.

3. The charging pile system supporting multi-point automatic cooling according to claim 1, characterized in that The temperature difference matrix is an n×m matrix, where n represents the number of temperature sampling points and m represents the length of the sampling time series. Each element in the temperature difference matrix represents the difference between the temperature of a certain sampling point at a specific moment and the ambient temperature.

4. The charging pile system supporting multi-point automatic temperature reduction according to claim 1, characterized in that, The temperature spatial basis matrix is an n×k matrix, where k represents the number of main features, and the temperature spatial basis matrix characterizes the spatial distribution characteristics of temperature sampling points; The time feature matrix is a k×m matrix, and the time feature matrix characterizes the time variation characteristics of temperature data.

5. The charging pile system supporting multi-point automatic temperature reduction according to claim 1, characterized in that, The temperature difference change coefficient matrix is an n×(m - 1) matrix, and each element in the temperature difference change coefficient matrix represents the change rate of the temperature difference between two adjacent moments in the temperature difference matrix.

6. The charging pile system supporting multi-point automatic cooling according to claim 1, characterized in that The temperature-power correlation matrix is an n×p matrix, where p represents the number of power sampling points, and the temperature-power correlation matrix characterizes the correlation relationship between temperature change and power change.

7. The charging pile system supporting multi-point automatic cooling according to claim 1, characterized in that The convolution kernel calculation function determines the size of the dynamic convolution kernel through the singular value decomposition result of the temperature spatial basis matrix, and the size of the dynamic convolution kernel is equal to the number of non-zero singular values of the temperature spatial basis matrix plus 1.

8. The charging pile system supporting multi-point automatic temperature reduction according to claim 1, characterized in that The temperature feature extraction network includes an input layer, a dynamic convolution layer, a pooling layer, a fully connected layer, and an output layer. The dynamic convolution layer uses a dynamic convolution kernel to perform feature mapping on temperature data.

9. The charging pile system supporting multi-point automatic cooling according to claim 1, characterized in that, The thermodynamic state equations include the coolant heat balance equation, the pipeline pressure loss equation, and the radiator heat transfer equation; the coolant heat balance equation is used to calculate the system heat dissipation power, with the inputs including the coolant flow rate, the coolant specific heat capacity, and the coolant inlet and outlet temperature difference, and the output being the system heat dissipation power; the pipeline pressure loss equation is used to calculate the pipeline resistance, with the inputs including the coolant flow rate, the pipeline length, the pipeline cross-sectional area, and the coolant viscosity, and the output being the pipeline pressure loss value; the radiator heat transfer equation is used to calculate the radiator heat dissipation power, with the inputs including the radiator surface area, the radiator convective heat transfer coefficient, the radiator surface temperature, and the ambient temperature, and the output being the radiator heat dissipation power.

10. The charging pile system supporting multi-point automatic temperature reduction according to claim 1, characterized in that The temperature acquisition module is used to acquire the charging interface temperature and the radiator temperature, the current acquisition module is used to acquire the charging current and the cooling fan working current, the pressure detection module is used to acquire the coolant pressure and the pipeline pressure, the heat dissipation control module is used to control the cooling fan speed and the coolant flow rate, the communication module is used to achieve data transmission between systems and remote monitoring, the storage module is used to store the temperature data and the system parameters, the display module is used to display the system operation status and the alarm information, and the power management module is used to monitor the system supply voltage and control the system power supply.

Images