An outdoor intelligent charging pile automatic heat dissipation adjusting method based on internet of things technology

Abstract

The application discloses an outdoor intelligent charging pile automatic heat dissipation adjusting method based on Internet of Things technology, which comprises the following steps: S1, collecting external environment parameters and charging pile operation parameters of the outdoor intelligent charging pile; S2, uploading the parameters to a cloud platform, performing data cleaning and normalization processing, and unifying the input format; S3, using a FEDformer temperature prediction model to predict the temperature value inside the equipment; S4, calling a Mamdani fuzzy control algorithm to perform inference calculation to obtain a control output value; S5, controlling the heat dissipation subsystem according to the control output value; and S6, when the environmental temperature and the load rate are lower than a threshold value, automatically switching to a low-energy-consumption operation mode. The application combines Internet of Things and AI temperature control prediction, realizes intelligent adjustment of the heat dissipation of the charging pile, and has the advantages of fast response, low energy consumption and fine control.

Description

Technical Field

[0001] This invention relates to the field of thermal management system control, and more particularly to an automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things (IoT) technology. Background Technology

[0002] With the rapid growth of new energy vehicle ownership, outdoor smart charging piles, as an important part of electric vehicle infrastructure, directly affect user experience and energy efficiency through their reliability and operational stability. During the charging process, due to high power output and continuous operation, key components such as the power module, charging module, and control module inside the charging pile will generate a lot of heat. If heat dissipation is not timely and effective, it will lead to equipment overheating, system performance degradation, and even safety hazards. Therefore, designing an efficient and intelligent heat dissipation and regulation system has become a key technological direction for the development of smart charging piles.

[0003] Existing outdoor charging stations typically employ a passive cooling strategy driven by temperature thresholds. This means that when the internal temperature of the equipment exceeds a set threshold, the air-cooling or liquid-cooling system is activated to lower the temperature. This strategy relies on data from a single temperature sensor and cannot comprehensively analyze multiple factors such as equipment load status and changes in the external environment. This control mechanism has a lag response and is prone to insufficient or redundant heat dissipation, affecting system stability and causing energy waste. While some advanced systems have introduced fuzzy control or multi-parameter judgment mechanisms, they generally lack temperature trend prediction capabilities and cannot model and predict temperature rises in advance, resulting in a control strategy primarily based on "lagging response."

[0004] Furthermore, existing cooling systems mostly employ a unified control strategy, meaning the entire system uses a single cooling solution. This lacks the ability to optimize and schedule the cooling of different modules based on their thermal characteristics, making it difficult to achieve fine-grained control under varying operating conditions. Especially under low load and low temperature environments, traditional systems cannot automatically enter energy-saving mode and maintain full cooling power, further reducing energy efficiency.

[0005] Therefore, there is an urgent need to propose a novel heat dissipation regulation method that combines IoT sensing, AI temperature trend prediction, and multi-module fuzzy adaptive control. This method can perform fine-grained prediction and response to the thermal load of multiple modules inside outdoor smart charging piles, realize independent control and low-energy operation of the modular heat dissipation system, and thus improve the stability, response speed and energy efficiency of the charging pile system in complex environments. Summary of the Invention

[0006] One objective of this invention is to propose an automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things (IoT) technology. This invention integrates IoT data acquisition technology with AI temperature trend prediction algorithms to construct a heat dissipation intelligent decision-making system based on the FEDformer model and Mamdani fuzzy control. It achieves differentiated and independent control based on the heat load characteristics of multiple modules inside the charging pile, and has the advantages of timely response, low energy consumption, strong adaptability and high heat dissipation efficiency, thereby improving the operational stability and energy utilization efficiency of outdoor smart charging piles in complex environments.

[0007] An automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things (IoT) technology, according to an embodiment of the present invention, includes the following steps: S1. Collect external environmental parameters and charging pile operating parameters of outdoor smart charging piles; S2. Using IoT technology, external environmental parameters and charging pile operating parameters are uploaded to the cloud platform for data cleaning and normalization, and the input format is standardized to form an input dataset. S3. Input the input dataset into the FEDformer temperature prediction model, perform multivariate time series modeling and frequency domain trend decomposition, and output the predicted temperature value inside the device. S4. Based on the predicted temperature value and the current charging pile operating parameters, call the Mamdani fuzzy control algorithm to perform inference calculations on the heat dissipation control output and obtain the control output value. S5. Control multiple heat dissipation subsystems installed in the outdoor smart charging pile according to the control output value; S6. When the ambient temperature is lower than the first temperature threshold and the load rate is lower than the first load threshold, automatically switch to low-energy operation mode, shut down some independent heat dissipation units, and only retain the passive heat dissipation path to maintain normal heat dissipation needs.

[0008] Optionally, the external environmental parameters include ambient temperature, ambient humidity, and wind speed, and the charging pile operating parameters include internal temperature, charging current, charging voltage, and load rate.

[0009] Optionally, the FEDformer temperature prediction model includes an embedding mapping layer, a time series decomposition layer, a frequency domain modeling layer, a general attention layer, a frequency domain thermogenic attention layer, and a sequence decoding layer.

[0010] Optionally, S3 specifically includes: S31. Generate a unified time series vector based on the input dataset; S32. Input the time series vector into the embedding mapping layer, and use a multilayer perceptron to perform a unified dimension mapping to obtain the embedding matrix; S33. Perform trend and period decomposition on the time dimension for each variable sequence in the time series vector, and use the moving average method to extract the trend component and disturbance component: ; in, Indicates the first Each variable at time step The trend component, Indicates the first Each variable at time step The perturbation component, Indicates the first Each variable at time step The variable sequence, Indicates the first Each variable at time step The variable sequence, The window width representing the moving average. Indicates the index of the sliding steps; S34. Input the trend component and the perturbation component into the embedding mapping layer respectively, and concatenate the results with the embedding matrix to form the enhanced feature matrix; S35. Input the time series vector into the frequency domain modeling layer, perform a fast Fourier transform to capture the periodicity and non-stationarity in the temperature rise change, obtain the spectrum vector of each variable, and concatenate them into a frequency domain matrix along the variable dimension. S36. Introduce a thermal factor-guided attention mechanism in the frequency domain thermal factor attention layer, calculate the historical linear contribution of each variable to the internal temperature, calculate the covariance correlation and normalize it to obtain the thermal factor-guided weight: ; in, Indicates the first Each variable has a heat-dependent weighting effect on the internal temperature. This represents the covariance between the variable series and the target temperature series. This represents the absolute value operation. Represents the internal temperature sequence. Indicates the total number of variable categories. Indicates the first A sequence of variables, Indicates the first A sequence of variables; S37. In the frequency domain heat source attention layer, the frequency domain matrix is ​​mapped to a query, key, and value matrix; in the ordinary attention layer, the enhanced feature matrix is ​​mapped to a query, key, and value matrix. ; ; in, These represent the query, key, and value matrices of the frequency domain matrix and the enhanced feature matrix, respectively. These represent the weight matrices of the query, key, and value matrices of the frequency domain matrix and the enhanced feature matrix, respectively. Represents the frequency domain matrix. Represents the enhanced feature matrix; The heat source guiding weights are incorporated into the attention function to obtain the heat source attention weights, and the heat source context matrix is ​​output: ; in, Indicates the first Attention weights of each variable Indicates the embedding dimension. Indicates the scaling factor. Represents the normalization function. Indicates the transpose operation; ; in, Represents the thermal context matrix; Based on the query, key, and value matrices of the enhanced feature matrix, ordinary attention weights are obtained, and a feature representation matrix is ​​output: ; in, Indicates the first Ordinary attention weights for each variable; ; in, The characteristic representation matrix; The thermal context matrix and the feature representation matrix are concatenated to generate a comprehensive feature matrix; S38. Input the comprehensive feature matrix into the sequence decoding layer, and calculate and generate the temperature prediction value for the future time step through the linear prediction head, wherein the linear prediction head uses a multilayer perceptron. S39. During model training, the internal true temperature value sequence is used as supervision to construct a mean squared error loss function, and the temperature prediction value is optimized by minimizing the mean squared error loss function. ; in, This represents the mean squared error loss function. Indicates time step Temperature forecast, Indicates time step The actual internal temperature value, It represents the number of steps in the future time frame.

[0011] Optionally, S4 specifically includes: S41. Three parallel fuzzy controllers are constructed for the charging module, power supply module and main control module of the outdoor smart charging pile respectively. Each fuzzy controller corresponds to an independent heat dissipation subsystem to realize heat source zoning control. Each fuzzy controller receives the predicted temperature value inside the device, the current charging current, the charging voltage and the load rate as input variables to form an input vector. The fuzzy controller includes a fuzzification module, a fuzzy rule base, an inference engine and a defuzzification module. The fuzzification module performs fuzzification processing on each dimension of the input vector, converting it into the membership degree of the corresponding fuzzy set; The fuzzy rule base is independently configured in each fuzzy controller to define control logic strategies, such as "if high temperature, high current, high voltage, and high load → strong cooling"; The inference engine selects matching fuzzy rules based on the current fuzzified input variables, calculates the rule activation degree of each fuzzy rule, and merges its output results into an output fuzzy set. The defuzzification module converts the output fuzzy set into a specific control output value using the centroid method; S42. Each fuzzy controller performs fuzzification processing on each dimension of the input vector, and uses triangular membership functions to model the fuzzy set of the input: ; in, Indicates input value Membership degree in a fuzzy set Represents the initial value of the fuzzy set. Represents the center value of a fuzzy set. This represents the termination value of the fuzzy set. Represents a fuzzy set; Each input variable is divided into three fuzzy states: low, medium, and high, corresponding to three functions respectively; S43. Each fuzzy controller is configured with an independent fuzzy rule base, which contains several fuzzy rules. Each fuzzy rule is as follows: ; in, Indicates belonging to the first The fuzzy controller's first A fuzzy rule, Indicates the first The fuzzy set to which each input value belongs. Indicates the first The fuzzy set to which each input value belongs. Indicates the first The fuzzy set to which each input value belongs. Indicates the first The fuzzy set to which each input value belongs. Indicates the first The output fuzzy set of the fuzzy rule Indicates the first The heat dissipation adjustment output value corresponding to each fuzzy controller. This represents the predicted temperature value. Indicates the charging current. Indicates the charging voltage. Indicates load rate; S44. Introduce weighting coefficients to perform weighted inference on each fuzzy rule in conjunction with its membership degree, and calculate the rule activation degree: ; in, This represents the activation level of the fuzzy rule. This indicates taking the minimum value. This indicates that the predicted temperature value is in the fuzzy set. Membership degree below Indicates charging current in fuzzy set Membership degree below Indicates the charging voltage in the fuzzy set Membership degree below Indicates load rate in fuzzy set Membership degree below Indicates the first In the module controller, the first The weight coefficients of the fuzzy rules are initially set to 1. When the activation degree is greater than 0, it means that the fuzzy rule is activated. If it is equal to 0, it means that at least one condition does not match at all, and this fuzzy rule does not participate in the reasoning. S45. Merge the fuzzy sets of all activated fuzzy rules to obtain the control output fuzzy set corresponding to each module. The output fuzzy set is the membership function distribution of the output control quantity over its value range, representing the current fuzzy decision preference. ; in, This indicates the output fuzzy set, describing the first... The membership function distribution of the output fuzzy set of each module controller at the output value of the heat dissipation regulation. This indicates that the output value of the heat dissipation regulation is in the fuzzy set. Membership degree below This represents the output variable of the module controller, describing the current cooling intensity to be adjusted. This indicates taking the maximum value; S46. Perform a demodulation operation on the output fuzzy set and use the centroid method to obtain the control output value: ; in, Indicates the first The control output values ​​of each module controller These represent the lower and upper limits of the output range, respectively. S47. After executing the control command, the system obtains the actual response temperature within the current cycle and calculates the prediction error: ; in, Indicates the first The prediction error of the individual module controller Indicates the actual response temperature; Based on this, the performance index values ​​are calculated as follows: ; in, Indicates the first Performance metrics of each module controller; By updating the weight coefficients of corresponding fuzzy rules in the fuzzy rule base using performance index values, a dynamic learning mechanism for rule credibility is formed, enabling the fuzzy controller to have adaptive optimization capabilities. ; in, Indicates at time step Time In the module controller, the first The weight coefficients of the fuzzy rules, Indicates at time step Time In the module controller, the first The weight coefficients of the fuzzy rules, This represents the conservatism coefficient, which describes the trade-off between historical and current errors.

[0012] Optionally, S5 specifically includes: S51. The control output value obtained based on the fuzzy controller is applied to the independent heat dissipation units of the charging module, power supply module and main control module to independently control multiple heat dissipation subsystems of the outdoor smart charging pile. The heat dissipation subsystem includes independent heat dissipation units for the charging module, power supply module and main control module respectively. S52. Based on the thermal load status of each module, obtain the actual execution control output of each heat dissipation unit: ; in, Indicates the first The actual execution control output of each module's heat dissipation unit. Indicates the first The control output values ​​of each module controller This indicates the current normalized heat load value of the module. This represents the response coefficient of the control output. This represents the heat load correction factor; S53. If the current heat load value of a certain module exceeds the set upper limit, the actual execution control output will be temporarily increased to the maximum value to achieve overheat protection. S54. The main control module sends the actual execution control output to the corresponding air-cooled unit or liquid-cooled pump control terminal of each module to adjust the fan PWM duty cycle, liquid-cooled pump frequency or solenoid valve opening to ensure that each heat dissipation unit responds independently according to the real-time status of its own module. S55 continuously records the actual control outputs and module temperature response data during the heat dissipation control process, and uploads them to the cloud via the Internet of Things communication protocol.

[0013] Optionally, S6 specifically includes: S61. Real-time collection of the current ambient temperature and load rate of the outdoor smart charging pile. The load rate is calculated based on the ratio of the current charging power to the rated maximum power to assess the current operating intensity of the equipment. S62. The first temperature threshold and the first load threshold are used as the start conditions for the low energy consumption operation mode. When the external ambient temperature is lower than the set first temperature threshold and the load rate is lower than the preset first load threshold, it is determined that the current state is in a low heat risk state and the conditions for entering the low energy consumption operation mode are met. S63. After entering the low-power operation mode, the power of the currently running independent heat dissipation unit will be adjusted or turned off, leaving only a small number of low-power fans or natural heat sinks to maintain basic heat dissipation capacity. S64. During the continuous low-energy operation mode, continue to periodically collect ambient temperature and load rate, and compare them with the preset first temperature threshold and first load threshold in real time. If any monitored parameter exceeds the threshold, the low-energy operation mode will be immediately interrupted and the system will switch back to the conventional fuzzy control adjustment strategy.

[0014] The beneficial effects of this invention are: First, this invention provides an automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things (IoT) technology. Addressing issues such as lag response, high energy consumption, and insufficient control granularity in existing charging pile heat dissipation systems, it proposes a system optimization scheme centered on "prediction-based adjustment + fuzzy regulation + independent module control." This improves the system's intelligence level and operational efficiency. By using multi-dimensional parameters such as ambient temperature, humidity, wind speed, and equipment operating status as inputs, and introducing the FEDformer temperature prediction model to model and predict internal temperature trends, the system can prepare for adjustments before temperature rise occurs, avoiding the overheating risk and energy waste caused by traditional passive response mechanisms.

[0015] Secondly, the Mamdani fuzzy control algorithm is used to perform multi-condition fusion reasoning on temperature prediction values ​​and operating parameters, and differentiated control commands are output according to the thermal load status of different modules. This enables zoned independent heat dissipation control for heat sources such as charging modules, power supply modules, and control modules. Compared with the traditional "one-size-fits-all" heat dissipation method, this invention achieves a more granular and adaptive control strategy, which not only improves heat dissipation efficiency but also extends the service life of key components of the equipment. In addition, by introducing a dynamic fuzzy rule weight adjustment mechanism and combining actual temperature feedback data to continuously optimize the control strategy, this invention enables the fuzzy controller to have self-learning and self-adaptive capabilities, further enhancing the system's robustness and self-optimization capabilities to different operating environments.

[0016] Finally, under low load or low temperature conditions, the present invention can automatically switch to a low energy consumption mode according to the environment and load status, shut down some active heat dissipation units, and retain only the basic heat dissipation path to maintain stable system operation, reduce the overall energy consumption, and conform to the green and energy-saving technology development trend. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of an automatic heat dissipation adjustment method for an outdoor smart charging pile based on Internet of Things technology proposed in this invention; Figure 2 This is a schematic diagram of the FEDformer temperature prediction model for an automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things technology proposed in this invention. Figure 3 This is a flowchart illustrating the calculation process for fuzzy control rule activation and control output generation in an automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things technology proposed in this invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figure 1-3 An automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things (IoT) technology includes the following steps: S1. Collect external environmental parameters and charging pile operating parameters of outdoor smart charging piles; S2. Using IoT technology, external environmental parameters and charging pile operating parameters are uploaded to the cloud platform for data cleaning and normalization, and the input format is standardized to form an input dataset. S3. Input the input dataset into the FEDformer temperature prediction model, perform multivariate time series modeling and frequency domain trend decomposition, and output the predicted temperature value inside the device. S4. Based on the predicted temperature value and the current charging pile operating parameters, call the Mamdani fuzzy control algorithm to perform inference calculations on the heat dissipation control output and obtain the control output value. S5. Control multiple heat dissipation subsystems installed in the outdoor smart charging pile according to the control output value; S6. When the ambient temperature is lower than the first temperature threshold and the load rate is lower than the first load threshold, automatically switch to low-energy operation mode, shut down some independent heat dissipation units, and only retain the passive heat dissipation path to maintain normal heat dissipation needs.

[0020] This invention constructs an automatic heat dissipation regulation method for outdoor smart charging piles that integrates IoT data acquisition, cloud data processing, temperature trend prediction, and fuzzy control inference. This method transforms the control mode from passive response to proactive prediction. During the operation of the charging pile, the method can collect environmental and equipment parameters in real time, predict the temperature rise trend through AI models, and activate reasonable heat dissipation strategies in advance. This avoids the overheating or energy waste problems caused by response lag in traditional solutions, thereby improving the timeliness and accuracy of the system's heat dissipation response and enhancing the stability, lifespan, and environmental adaptability of the equipment.

[0021] In this embodiment, the external environmental parameters include ambient temperature, ambient humidity, and wind speed, and the charging pile operating parameters include internal temperature, charging current, charging voltage, and load rate.

[0022] This invention provides comprehensive and multi-dimensional raw data support for subsequent AI modeling and heat dissipation decisions by clearly collecting external parameters such as ambient temperature, humidity, and wind speed, as well as internal operating parameters such as temperature, current, voltage, and load rate. These parameter settings cover key physical variables that determine thermal behavior, which is conducive to establishing a more accurate dynamic temperature rise model, improving prediction accuracy and control precision. Compared with traditional methods that rely solely on internal temperature values, this invention achieves three-dimensional perception of the thermal state, providing a stronger data foundation for heat dissipation control strategies, and effectively improving the system's predictive capabilities and the level of intelligent heat dissipation regulation.

[0023] In this embodiment, the FEDformer temperature prediction model includes an embedding mapping layer, a time series decomposition layer, a frequency domain modeling layer, a general attention layer, a frequency domain thermal attention layer, and a sequence decoding layer.

[0024] The FEDformer temperature prediction model structure proposed in this invention integrates multiple modules such as embedding mapping, time series decomposition, frequency domain modeling, general and thermal attention mechanisms, and sequence decoding, forming an efficient prediction architecture suitable for non-stationary temperature data. This model can accurately capture the periodicity, trend, and key variable influencing factors in temperature rise data, thereby providing reliable prediction outputs before equipment temperature changes, providing a basis for the early initiation of intelligent heat dissipation strategies. Compared with traditional ARIMA or LSTM models, the FEDformer structure is more suitable for sequence modeling in complex multidimensional thermal environments, and has stronger generalization ability and inference performance.

[0025] In this embodiment, S3 specifically includes: S31. Generate a unified time series vector based on the input dataset; S32. Input the time series vector into the embedding mapping layer, and use a multilayer perceptron to perform a unified dimension mapping to obtain the embedding matrix; S33. Perform trend and period decomposition on the time dimension for each variable sequence in the time series vector, and use the moving average method to extract the trend component and disturbance component: ; in, Indicates the first Each variable at time step The trend component, Indicates the first Each variable at time step The perturbation component, Indicates the first Each variable at time step The variable sequence, Indicates the first Each variable at time step The variable sequence, The window width representing the moving average. Indicates the index of the sliding steps; S34. Input the trend component and the perturbation component into the embedding mapping layer respectively, and concatenate the results with the embedding matrix to form the enhanced feature matrix; S35. Input the time series vector into the frequency domain modeling layer, perform a fast Fourier transform to capture the periodicity and non-stationarity in the temperature rise change, obtain the spectrum vector of each variable, and concatenate them into a frequency domain matrix along the variable dimension. S36. Introduce a thermal factor-guided attention mechanism in the frequency domain thermal factor attention layer, calculate the historical linear contribution of each variable to the internal temperature, calculate the covariance correlation and normalize it to obtain the thermal factor-guided weight: ; in, Indicates the first Each variable has a heat-dependent weighting effect on the internal temperature. This represents the covariance between the variable series and the target temperature series. This represents the absolute value operation. Represents the internal temperature sequence. Indicates the total number of variable categories. Indicates the first A sequence of variables, Indicates the first A sequence of variables; S37. In the frequency domain heat source attention layer, the frequency domain matrix is ​​mapped to a query, key, and value matrix; in the ordinary attention layer, the enhanced feature matrix is ​​mapped to a query, key, and value matrix. ; ; in, These represent the query, key, and value matrices of the frequency domain matrix and the enhanced feature matrix, respectively. These represent the weight matrices of the query, key, and value matrices of the frequency domain matrix and the enhanced feature matrix, respectively. Represents the frequency domain matrix. Represents the enhanced feature matrix; The heat source guiding weights are incorporated into the attention function to obtain the heat source attention weights, and the heat source context matrix is ​​output: ; in, Indicates the first Attention weights of each variable Indicates the embedding dimension. Indicates the scaling factor. Represents the normalization function. Indicates the transpose operation; ; in, Represents the thermal context matrix; Based on the query, key, and value matrices of the enhanced feature matrix, ordinary attention weights are obtained, and a feature representation matrix is ​​output: ; in, Indicates the first Ordinary attention weights for each variable; ; in, The characteristic representation matrix; The thermal context matrix and the feature representation matrix are concatenated to generate a comprehensive feature matrix; S38. Input the comprehensive feature matrix into the sequence decoding layer, and calculate and generate the temperature prediction value for the future time step through the linear prediction head, wherein the linear prediction head uses a multilayer perceptron. S39. During model training, the internal true temperature value sequence is used as supervision to construct a mean squared error loss function, and the temperature prediction value is optimized by minimizing the mean squared error loss function. ; in, This represents the mean squared error loss function. Indicates time step Temperature forecast, Indicates time step The actual internal temperature value, It represents the number of steps in the future time frame.

[0026] This invention discloses in detail the specific implementation process of the FEDformer temperature prediction model, covering steps such as the construction of multivariate time series, embedding mapping, trend period decomposition, frequency domain feature extraction, heat factor attention weight guidance, and attention fusion mechanism. It effectively improves the modeling depth and physical relevance of temperature rise prediction. The heat factor-guided attention mechanism enhances the model's sensitivity to key variables by introducing the covariance weight between variables and target temperature, making the model pay more attention to input factors with actual heat factor connections. It realizes the transformation from "all-variable attention" to "heat factor-weighted attention", which greatly improves the accuracy and efficiency of prediction.

[0027] In this embodiment, S4 specifically includes: S41. Three parallel fuzzy controllers are constructed for the charging module, power supply module and main control module of the outdoor smart charging pile respectively. Each fuzzy controller corresponds to an independent heat dissipation subsystem to realize heat source zoning control. Each fuzzy controller receives the predicted temperature value inside the device, the current charging current, the charging voltage and the load rate as input variables to form an input vector. The fuzzy controller includes a fuzzification module, a fuzzy rule base, an inference engine and a defuzzification module. The fuzzification module performs fuzzification processing on each dimension of the input vector, converting it into the membership degree of the corresponding fuzzy set; The fuzzy rule base is independently configured in each fuzzy controller to define control logic strategies, such as "if high temperature, high current, high voltage, and high load → strong cooling"; The inference engine selects matching fuzzy rules based on the current fuzzified input variables, calculates the rule activation degree of each fuzzy rule, and merges its output results into an output fuzzy set. The defuzzification module converts the output fuzzy set into a specific control output value using the centroid method; S42. Each fuzzy controller performs fuzzification processing on each dimension of the input vector, and uses triangular membership functions to model the fuzzy set of the input: ; in, Indicates input value Membership degree in a fuzzy set Represents the initial value of the fuzzy set. Represents the center value of a fuzzy set. This represents the termination value of the fuzzy set. Represents a fuzzy set; Each input variable is divided into three fuzzy states: low, medium, and high, corresponding to three functions respectively; S43. Each fuzzy controller is configured with an independent fuzzy rule base, which contains several fuzzy rules. Each fuzzy rule is as follows: ; in, Indicates belonging to the first The fuzzy controller's first A fuzzy rule, Indicates the first The fuzzy set to which each input value belongs. Indicates the first The fuzzy set to which each input value belongs. Indicates the first The fuzzy set to which each input value belongs. Indicates the first The fuzzy set to which each input value belongs. Indicates the first The output fuzzy set of the fuzzy rule Indicates the first The heat dissipation adjustment output value corresponding to each fuzzy controller. This represents the predicted temperature value. Indicates the charging current. Indicates the charging voltage. Indicates load rate; S44. Introduce weighting coefficients to perform weighted inference on each fuzzy rule in conjunction with its membership degree, and calculate the rule activation degree: ; in, This represents the activation level of the fuzzy rule. This indicates taking the minimum value. This indicates that the predicted temperature value is in the fuzzy set. Membership degree below Indicates charging current in fuzzy set Membership degree below Indicates the charging voltage in the fuzzy set Membership degree below Indicates load rate in fuzzy set Membership degree below Indicates the first In the module controller, the first The weight coefficients of the fuzzy rules are initially set to 1. When the activation degree is greater than 0, it means that the fuzzy rule is activated. If it is equal to 0, it means that at least one condition does not match at all, and this fuzzy rule does not participate in the reasoning. S45. Merge the fuzzy sets of all activated fuzzy rules to obtain the control output fuzzy set corresponding to each module. The output fuzzy set is the membership function distribution of the output control quantity over its value range, representing the current fuzzy decision preference. ; in, This indicates the output fuzzy set, describing the first... The membership function distribution of the output fuzzy set of each module controller at the output value of the heat dissipation regulation. This indicates that the output value of the heat dissipation regulation is in the fuzzy set. Membership degree below This represents the output variable of the module controller, describing the current cooling intensity to be adjusted. This indicates taking the maximum value; S46. Perform a demodulation operation on the output fuzzy set and use the centroid method to obtain the control output value: ; in, Indicates the first The control output values ​​of each module controller These represent the lower and upper limits of the output range, respectively. S47. After executing the control command, the system obtains the actual response temperature within the current cycle and calculates the prediction error: ; in, Indicates the first The prediction error of the individual module controller Indicates the actual response temperature; Based on this, the performance index values ​​are calculated as follows: ; in, Indicates the first Performance metrics of each module controller; By updating the weight coefficients of corresponding fuzzy rules in the fuzzy rule base using performance index values, a dynamic learning mechanism for rule credibility is formed, enabling the fuzzy controller to have adaptive optimization capabilities. ; in, Indicates at time step Time In the module controller, the first The weight coefficients of the fuzzy rules, Indicates at time step Time In the module controller, the first The weight coefficients of the fuzzy rules, This represents the conservatism coefficient, which describes the trade-off between historical and current errors.

[0028] This invention constructs a modular and parallel Mamdani fuzzy controller system, which operates separately on the charging, power supply, and main control modules of the charging pile. This achieves independent heat dissipation regulation and control for multiple heat source zones, avoiding the energy redundancy problem of traditional unified air-cooling strategies. The fuzzy controller is input with key parameters such as predicted temperature, current, voltage, and load rate, and performs inference based on a rule base and weight mechanism, possessing adaptive adjustment capabilities for different operating conditions. The introduced fuzzy rule weight update mechanism can also dynamically optimize the control logic based on control errors, enabling the system to learn and self-evolve, improving its long-term intelligence and energy efficiency.

[0029] In this embodiment, S5 specifically includes: S51. The control output value obtained based on the fuzzy controller is applied to the independent heat dissipation units of the charging module, power supply module and main control module to independently control multiple heat dissipation subsystems of the outdoor smart charging pile. The heat dissipation subsystem includes independent heat dissipation units for the charging module, power supply module and main control module respectively. S52. Based on the thermal load status of each module, obtain the actual execution control output of each heat dissipation unit: ; in, Indicates the first The actual execution control output of each module's heat dissipation unit. Indicates the first The control output values ​​of each module controller This indicates the current normalized heat load value of the module. This represents the response coefficient of the control output. This represents the heat load correction factor; S53. If the current heat load value of a certain module exceeds the set upper limit, the actual execution control output will be temporarily increased to the maximum value to achieve overheat protection. S54. The main control module sends the actual execution control output to the corresponding air-cooled unit or liquid-cooled pump control terminal of each module to adjust the fan PWM duty cycle, liquid-cooled pump frequency or solenoid valve opening to ensure that each heat dissipation unit responds independently according to the real-time status of its own module. S55 continuously records the actual control outputs and module temperature response data during the heat dissipation control process, and uploads them to the cloud via the Internet of Things communication protocol.

[0030] This invention introduces a module-level thermal load correction mechanism into the heat dissipation execution strategy, enabling the control output to not only consider the fuzzy inference results but also integrate the current thermal load status of each module. This achieves more precise and dynamic heat dissipation power allocation. Through fine adjustment of execution components such as air-cooled units, liquid-cooled pumps, and solenoid valves, it ensures that the heat dissipation system of each module responds independently according to its thermal intensity, effectively improving the overall heat dissipation efficiency. This method supports real-time control command reporting and feedback recording, facilitating cloud monitoring and historical behavior learning, further supporting fuzzy controller optimization, and effectively avoiding over- or under-heat dissipation.

[0031] In this embodiment, S6 specifically includes: S61. Real-time collection of the current ambient temperature and load rate of the outdoor smart charging pile. The load rate is calculated based on the ratio of the current charging power to the rated maximum power to assess the current operating intensity of the equipment. S62. The first temperature threshold and the first load threshold are used as the start conditions for the low energy consumption operation mode. When the external ambient temperature is lower than the set first temperature threshold and the load rate is lower than the preset first load threshold, it is determined that the current state is in a low heat risk state and the conditions for entering the low energy consumption operation mode are met. S63. After entering the low-power operation mode, the power of the currently running independent heat dissipation unit will be adjusted or turned off, leaving only a small number of low-power fans or natural heat sinks to maintain basic heat dissipation capacity. S64. During the continuous low-energy operation mode, continue to periodically collect ambient temperature and load rate, and compare them with the preset first temperature threshold and first load threshold in real time. If any monitored parameter exceeds the threshold, the low-energy operation mode will be immediately interrupted and the system will switch back to the conventional fuzzy control adjustment strategy.

[0032] This invention introduces a low-energy-consumption operation mode control mechanism. When the external ambient temperature and load rate are both at low levels, the system can intelligently determine that the current thermal risk is low and automatically shut down some heat dissipation subsystems, keeping only the basic heat dissipation channels running. This strategy not only avoids energy waste in non-high-heat scenarios, but also extends the service life of the air-cooled device. The system has continuous monitoring and rapid switching capabilities, ensuring that normal heat dissipation control is immediately restored when the heat load rises again, effectively balancing heat dissipation efficiency and energy use, and achieving a dynamic operation strategy that emphasizes both high performance and energy saving.

[0033] Example 1: To verify the applicability and feasibility of the present invention in a real-world environment, the intelligent heat dissipation regulation method of the present invention was deployed in an outdoor intelligent charging pile system in a large parking lot. The average temperature of this site is high in summer, and the charging equipment operates around the clock. Especially during the high-temperature period in the afternoon, the heat load of the equipment fluctuates drastically, facing serious heat dissipation pressure and overheating risks. During the implementation process, IoT data acquisition terminals integrating temperature sensors, anemometers, humidity sensors, etc., were deployed to continuously monitor key operating parameters such as ambient temperature, humidity, wind speed around the charging pile, as well as current, voltage, internal temperature, and load rate of the equipment itself, and upload them to the cloud platform in real time.

[0034] The intelligent heat dissipation regulation method of the present invention first uses the FEDformer model to perform multivariate modeling on the above-mentioned multi-source heterogeneous data, and combines frequency domain trends and disturbance characteristics to predict future internal temperature trends. Based on the prediction results, the Mamdani fuzzy controller performs decision reasoning. Unlike the traditional empirical rule-based fan start logic, this controller performs targeted regulation for each independent heat dissipation module (charging module, power supply module, main control module). The controller uses temperature, current, voltage, and load rate as input variables, and combines self-learning fuzzy rule weights to adjust the output intensity. The generated control output realizes heat dissipation control through liquid cooling pump frequency adjustment, solenoid valve opening adjustment, and fan PWM control.

[0035] Table 1 is a comparative data table of heat dissipation performance of smart charging piles, showing the comparison data of average internal temperature, energy consumption and overheating of traditional heat dissipation solutions and the solution of the present invention under different environmental conditions.

[0036] Table 1 Comparison of Heat Dissipation Performance of Smart Charging Piles

[0037] Table 1 reflects the key performance differences between the traditional heat dissipation system and the intelligent heat dissipation regulation scheme based on IoT and AI proposed in this invention during 10 days of typical high-temperature operation. In terms of ambient temperature, the average daily external temperature fluctuated between 33.9℃ and 39.1℃ during the test period, which presented a significant challenge in terms of heat load.

[0038] Regarding the control of average internal temperature, the internal temperature of traditional heat dissipation methods is generally maintained above 69°C, reaching a maximum of 72.1°C. However, the solution of this invention can control the average internal temperature below 67°C under the same environmental conditions, and in some cases as low as 62.3°C. This indicates that the present invention is superior to traditional methods in terms of temperature control response and thermal regulation accuracy, effectively delaying the thermal aging process of the equipment.

[0039] Based on the energy consumption data of the heat dissipation system, the daily energy consumption of the traditional solution is between 3.9 and 4.7 kWh, while the intelligent adjustment strategy of the present invention reduces the daily energy consumption to 2.8 to 3.6 kWh, with an energy saving effect of 20% to 30%, which is more obvious on days with lower external temperatures or lighter loads, demonstrating the dynamic energy efficiency optimization capability under fuzzy control and prediction mechanism.

[0040] In the comparison of the number of module overheating events, traditional heat dissipation frequently resulted in more than 3 module overheating events on high-temperature days, even reaching 6 times. In contrast, the solution of this invention controls the overheating risk to less than 2 times per day, and in most cases, it remains at 1 time or 0 times. This fully demonstrates the effectiveness of the adaptive heat dissipation strategy and module partition control, which greatly improves the stability and safety of system operation.

[0041] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. An automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things (IoT) technology, characterized in that, Includes the following steps: S1. Collect external environmental parameters and charging pile operating parameters of outdoor smart charging piles; S2. Using IoT technology, external environmental parameters and charging pile operating parameters are uploaded to the cloud platform for data cleaning and normalization, and the input format is standardized to form an input dataset. S3. Input the input dataset into the FEDformer temperature prediction model, perform multivariate time series modeling and frequency domain trend decomposition, and output the predicted temperature value inside the device. S4. Based on the predicted temperature value and the current charging pile operating parameters, call the Mamdani fuzzy control algorithm to perform inference calculations on the heat dissipation control output and obtain the control output value. S5. Control multiple heat dissipation subsystems installed in the outdoor smart charging pile according to the control output value; S6. When the ambient temperature is lower than the first temperature threshold and the load rate is lower than the first load threshold, automatically switch to low-energy operation mode, shut down some independent heat dissipation units, and only retain the passive heat dissipation path to maintain normal heat dissipation needs.

2. The automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things technology according to claim 1, characterized in that, The external environmental parameters include ambient temperature, ambient humidity, and wind speed, while the charging pile operating parameters include internal temperature, charging current, charging voltage, and load rate.

3. The automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things technology according to claim 1, characterized in that, The FEDformer temperature prediction model includes an embedding mapping layer, a time series decomposition layer, a frequency domain modeling layer, a general attention layer, a frequency domain thermogenic attention layer, and a sequence decoding layer.

4. The automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things technology according to claim 1, characterized in that, S3 specifically includes: S31. Generate a unified time series vector based on the input dataset; S32. Input the time series vector into the embedding mapping layer, and use a multilayer perceptron to perform a unified dimension mapping to obtain the embedding matrix; S33. Perform trend and period decomposition on the time dimension for each variable sequence in the time series vector, and use the moving average method to extract the trend component and disturbance component. S34. Input the trend component and the perturbation component into the embedding mapping layer respectively, and concatenate the results with the embedding matrix to form the enhanced feature matrix; S35. Input the time series vector into the frequency domain modeling layer, perform a fast Fourier transform to capture the periodicity and non-stationarity in the temperature rise change, obtain the spectrum vector of each variable, and concatenate them into a frequency domain matrix along the variable dimension. S36. Introduce a thermal factor-guided attention mechanism in the frequency domain thermal factor attention layer, calculate the historical linear contribution of each variable to the internal temperature, calculate the covariance correlation and normalize it to obtain the thermal factor-guided weight. S37. In the frequency domain heat source attention layer, the frequency domain matrix is ​​mapped to a query, key, and value matrix, and in the ordinary attention layer, the enhanced feature matrix is ​​mapped to a query, key, and value matrix. The heat factor guiding weights are incorporated into the attention function to obtain the heat factor attention weights, and the heat factor context matrix is ​​output. The query, key, and value matrices based on the enhanced feature matrix are used to obtain ordinary attention weights, and the feature representation matrix is ​​output. The thermal context matrix and the feature representation matrix are concatenated to generate a comprehensive feature matrix; S38. Input the comprehensive feature matrix into the sequence decoding layer, and calculate and generate the temperature prediction value for the future time step through the linear prediction head, wherein the linear prediction head uses a multilayer perceptron. S39. During model training, the internal temperature true value sequence is used as supervision to construct a mean squared error loss function, and the temperature prediction value is optimized by minimizing the mean squared error loss function.

5. The automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things technology according to claim 1, characterized in that, S4 specifically includes: S41. Three parallel fuzzy controllers are constructed for the charging module, power module and main control module of the outdoor smart charging pile respectively. Each fuzzy controller corresponds to an independent heat dissipation subsystem. Each fuzzy controller receives the predicted temperature value inside the device, the current charging current, the charging voltage and the load rate as input variables to form an input vector. The fuzzy controller includes a fuzzification module, a fuzzy rule base, an inference engine and a defuzzification module. S42. Each fuzzy controller performs fuzzification processing on each dimension of the input vector and uses triangular membership functions to model the fuzzy set of the input. Each input variable is divided into three fuzzy states: low, medium, and high, corresponding to three functions respectively; S43. Each fuzzy controller is configured with an independent fuzzy rule base, which contains several fuzzy rules; S44. Introduce weighting coefficients to perform weighted inference on each fuzzy rule in conjunction with its membership degree, and calculate the rule activation degree: ； in, This represents the activation level of the fuzzy rule. This indicates taking the minimum value. This indicates that the predicted temperature value is in the fuzzy set. Membership degree below Indicates charging current in fuzzy set Membership degree below Indicates the charging voltage in the fuzzy set Membership degree below Indicates load rate in fuzzy set Membership degree below Indicates the first In the module controller, the first The weight coefficients of the fuzzy rules are initially set to 1. When the activation degree is greater than 0, it means that the fuzzy rule is activated. If it is equal to 0, it means that at least one condition does not match at all, and this fuzzy rule does not participate in the reasoning. S45. Merge the fuzzy sets of all activated fuzzy rules to obtain the control output fuzzy set corresponding to each module. S46. Perform a demodeling operation on the output fuzzy set and use the centroid method to obtain the control output value; S47. After executing the control command, the system obtains the actual response temperature in the current cycle and calculates the prediction error. Based on this, the performance index values ​​are calculated; The weight coefficients of corresponding fuzzy rules in the fuzzy rule base are updated using performance index values, forming a dynamic learning mechanism for rule credibility: ； in, Indicates at time step Time In the module controller, the first The weight coefficients of the fuzzy rules, Indicates at time step Time In the module controller, the first The weight coefficients of the fuzzy rules, This represents the conservatism coefficient, describing the trade-off between historical and current errors. Indicates the first Performance metrics for each module controller.

6. The automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things technology according to claim 1, characterized in that, S5 specifically includes: S51. The control output value obtained based on the fuzzy controller is applied to the independent heat dissipation units of the charging module, power supply module and main control module to independently control multiple heat dissipation subsystems of the outdoor smart charging pile. S52. Based on the thermal load status of each module, obtain the actual execution control output of each heat dissipation unit; S53. If the current heat load value of a certain module exceeds the set upper limit, the actual execution control output will be temporarily increased to the maximum value to achieve overheat protection. S54. The main control module sends the actual execution control output to the corresponding air-cooled unit or liquid-cooled pump control terminal of each module to adjust the fan PWM duty cycle, liquid-cooled pump frequency or solenoid valve opening. S55 continuously records the actual control outputs and module temperature response data during the heat dissipation control process, and uploads them to the cloud via the Internet of Things communication protocol.

7. The automatic heat dissipation adjustment method for outdoor smart charging piles based on Internet of Things technology according to claim 1, characterized in that, S6 specifically includes: S61. Real-time collection of the current ambient temperature and load rate of outdoor smart charging piles; S62. The first temperature threshold and the first load threshold are used as the start conditions for the low energy consumption operation mode. When the external ambient temperature is lower than the set first temperature threshold and the load rate is lower than the preset first load threshold, it is determined that the current state is in a low heat risk state and the conditions for entering the low energy consumption operation mode are met. S63. After entering the low-power operation mode, the power of the currently running independent heat dissipation unit will be adjusted or turned off, leaving only a small number of low-power fans or natural heat sinks to maintain basic heat dissipation capacity. S64. During the continuous low-energy operation mode, continue to periodically collect ambient temperature and load rate, and compare them with the preset first temperature threshold and first load threshold in real time. If any monitored parameter exceeds the threshold, the low-energy operation mode will be immediately interrupted and the system will switch back to the conventional fuzzy control adjustment strategy.

Images