Heat dissipation heat energy recovery method for power electronic converter

Abstract

The invention discloses a power electronic converter radiating heat energy recovery method, which relates to the technical field of converter heat energy recovery, and comprises the following steps of: firstly, regularly collecting working state data of a converter by adopting a time driving strategy to ensure the continuity and the monitoring performance of the data; according to the scheme, the data acquisition period is dynamically adjusted through the deep learning model, and the response efficiency and economical efficiency of the heat energy recovery system are improved. When the load is increased suddenly, the system detects load fluctuation in real time, automatically increases the acquisition frequency, quickly responds to temperature change and recovers waste heat, and heat waste and overheating risks are avoided; and when the load is stable, the existing monitoring frequency is maintained, and resource waste is avoided. The mechanism optimizes resource allocation and reduces unnecessary energy consumption while improving the energy utilization efficiency, thereby remarkably enhancing the overall energy efficiency, economical efficiency and return on investment of the system.

Description

Technical Field

[0001] The invention relates to the technical field of converter heat energy recovery, and in particular to a method for recovering heat energy from a power electronic converter. Background Art

[0002] Heat recovery of power electronic converter heat dissipation refers to the use of waste heat (heat energy of the heat dissipation part) generated by the converter during operation through a heat recovery system, converting the heat originally dissipated to the environment into usable energy, thereby improving the energy efficiency of the overall system. Power electronic converters inevitably generate losses when converting electrical energy, and this part of the loss is dissipated in the form of heat energy. If it is not recovered, the heat dissipation system will consume additional electrical energy to cool the equipment. The heat recovery technology converts this waste heat back into electrical energy or heat energy storage through thermoelectric conversion modules, heat exchangers or waste heat utilization devices for auxiliary power supply, heating or other industrial purposes. This technology can not only reduce the energy consumption of the heat dissipation system, but also reduce environmental thermal pollution, improve the energy utilization rate of the converter system and the economy of overall operation.

[0003] The prior art has the following deficiencies:

[0004] In order to maximize the efficiency of heat recovery, modern heat recovery systems are usually equipped with intelligent control systems. In the prior art, intelligent control systems often adopt a periodic scheduling monitoring strategy, which regularly collects the working status data of the converter, such as temperature changes, environmental conditions, load fluctuations and heat dissipation requirements, through time-driven, and analyzes and judges based on these data, dynamically adjusts the operation mode of the heat recovery equipment to improve the heat recovery efficiency, stability and economy of the system. However, in the case of a sudden increase in load, the temperature of the converter may rise rapidly, and periodic monitoring cannot capture this change in time, resulting in the failure of the heat recovery system to respond in time. As a result, the waste heat that should have been recovered cannot be effectively converted into usable energy because the thermoelectric conversion module and the heat exchange system cannot be adjusted quickly to cope with the excessive temperature difference. This delayed response not only leads to energy waste and reduces the energy utilization efficiency of the system, but also reduces economic benefits, because the waste heat that cannot be used in time affects the overall return on investment of the converter system.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a method for recovering heat energy from heat dissipation of a power electronic converter. By dynamically adjusting the data acquisition cycle through a deep learning model, the scheme improves the response efficiency and economy of the heat recovery system. When the load surges, the system detects load fluctuations in real time, automatically increases the acquisition frequency, quickly responds to temperature changes and recovers waste heat to avoid heat waste and overheating risks; when the load is stable, the existing monitoring frequency is maintained to avoid resource waste. This mechanism improves energy utilization efficiency while optimizing resource allocation and reducing unnecessary energy consumption, thereby significantly enhancing the overall energy efficiency, economy and return on investment of the system to solve the problems in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a method for recovering heat energy from heat dissipation of a power electronic converter, comprising the following steps:

[0008] Firstly, a time-driven strategy is used to regularly collect the working status data of the converter to ensure the continuity and controllability of the data;

[0009] Preprocessing the acquired raw data, and after the data preprocessing is completed, extracting key features reflecting the surge in converter load;

[0010] The extracted key features are analyzed in detail under the detection window, and the load changes of the converter are quantified by analyzing the key features;

[0011] The quantified feature data is input into a pre-trained deep learning model, and the load change of the converter is intelligently evaluated through the deep learning model;

[0012] Based on the evaluation results of the deep learning model, the load change of the converter is divided into two states: load surge and load stability.

[0013] For stable load, continue to adopt the time-driven periodic data collection strategy and maintain the existing monitoring frequency to ensure that the heat recovery system maintains efficient operation under stable load conditions;

[0014] In the case of load surges, the data collection cycle is dynamically adjusted based on the evaluation results of the deep learning model, the collection frequency is increased, and load fluctuations and heat changes are responded to in a timely manner, thereby ensuring that waste heat can be recovered in a timely manner.

[0015] Preferably, key features reflecting the surge in converter load are extracted therefrom, the extracted features including the converter internal impedance change and the on-state maintenance time of the converter internal switching elements, the extracted converter internal impedance change and the on-state maintenance time of the converter internal switching elements are analyzed under the detection window, and an internal impedance change rate reference value and a converter current on-time reference value are generated respectively, and the converter load change is quantified by the internal impedance change rate reference value and the converter current on-time reference value.

[0016] Preferably, the specific steps of analyzing the internal impedance change of the converter in the detection window to generate the reference value of the internal impedance change rate are as follows:

[0017] First, the original working status data is collected from the sensors of the converter. In order to ensure the accuracy of the data, the original data is preprocessed and the instantaneous impedance of the converter is calculated by the acquired data. The calculation expression is as follows:

[0018]

[0019] Where V(t) is the instantaneous voltage, which represents the value of the voltage at the output of the converter at time t, I(t) is the instantaneous current, which represents the value of the current at the output of the converter at time t, and Z(t) is the instantaneous impedance;

[0020] Calculate the change rate of the internal impedance of the converter. The change rate of the impedance is an important indicator to measure the converter's response to load fluctuations. The calculation expression is as follows:

[0021]

[0022] In the formula, is the impedance change rate, Z(t-1) is the instantaneous impedance at the moment before time t, α is the weighting coefficient, and n is the total number of time points;

[0023] The impedance change rate is converted into a characteristic value, and a change rate characteristic function is constructed to obtain a characteristic value reflecting load fluctuations, thereby enhancing the sensitivity to the change rate when the sudden load surges. The expression is as follows:

[0024]

[0025] Where C(t) is the characteristic value of the rate of change, and β is the expansion coefficient;

[0026] The internal impedance change rate reference value is generated by taking a time-weighted average of the change rate characteristic value C(t) and combining it with the current load state. The calculation expression is as follows:

[0027]

[0028] In the formula, R Z is the reference value of the internal impedance change rate, and γ is the weight coefficient, which is used to adjust the contribution of the change rate characteristic value C(t) to the total change.

[0029] Preferably, the specific steps of analyzing the on-state maintenance time of the switch element inside the converter to generate the converter current on-time reference value are as follows:

[0030] First, relevant switch element status data is collected in real time from the converter control system. The acquired data provides a basis for subsequent analysis. By monitoring the precise moment of each switch operation, the on-time of the switch element is extracted, that is, the duration from on to off of each switch. The formula is as follows:

[0031] T conduct (i) = ∫ i i+1 State(t)dt

[0032] Where, T conduct (i) is the conduction time of the i-th switch operation, State(t) is a time function, which represents the conduction state of the switch element at any time t, ∫ i i+1 State(t)dt is the calculation of the on-time of the switch element between the i-th to the i+1-th switching operations;

[0033] After obtaining the on-time, the on-time fluctuation analysis is performed to calculate the cumulative change of the on-time. The calculation expression is as follows:

[0034]

[0035] Where, T conduct (i-1) is the on-time of the i-1th switch operation, i.e., the on-time of the previous switch operation, m is the total number of switch operations within the detection window, ω is an exponential coefficient, ΔT conduct is the on-time variation;

[0036] According to the on-time change ΔT conduct Identify the load surge, introduce a dynamic weight coefficient to adjust the importance of the on-time change, and automatically optimize the weight coefficient based on the historical load pattern. The expression is as follows:

[0037]

[0038] Where W conduct is the on-time weight coefficient, ΔT conduct (i) is the change in on-time of the i-th switching operation, ΔT conduct (j) is the change in on-time of the jth switching operation, is the exponential magnification factor;

[0039] Based on the on-time T conduct (i) and the on-time weight coefficient W conduct Generate the converter current conduction time reference value, the generation formula is as follows:

[0040]

[0041] Where γ is the exponential weighting factor, T ref It is the reference value of the converter current conduction time.

[0042] Preferably, the internal impedance change rate reference value and the converter current conduction time reference value generated after analysis are input into a pre-learned deep learning model, a load fluctuation index is generated by the deep learning model, and the converter load change is intelligently evaluated by the load fluctuation index.

[0043] Preferably, the load fluctuation index generated when the load change of the converter is intelligently evaluated by the pre-learned deep learning model is compared and analyzed with a pre-set load fluctuation index reference threshold, and the load change of the converter is divided, and the division steps are as follows:

[0044] If the load fluctuation index is greater than the load fluctuation index reference threshold, the load change of the converter is classified as a load surge;

[0045] If the load fluctuation index is less than or equal to the load fluctuation index reference threshold, the load change of the converter is classified as load stability.

[0046] Preferably, for a load surge, based on the deep learning model evaluation results, the data collection cycle is dynamically adjusted, the collection frequency is increased, and the load fluctuation and heat change are responded to in a timely manner, thereby ensuring that the waste heat can be recovered in a timely manner. The specific steps are as follows:

[0047] When the load fluctuation index Load flu When it is greater than the reference threshold of the load fluctuation index, the data collection cycle is dynamically adjusted according to the evaluation results to ensure that the real-time monitoring system can respond quickly to load fluctuations and heat changes. The formula is as follows:

[0048]

[0049] In the formula, Load flu is the load fluctuation index, Load ref is the reference threshold of load fluctuation index, e is the weaving and dyeing base, θ is the adjustment coefficient, and f base is the basic sampling frequency, f sampling is the frequency of new data collection;

[0050] With the dynamic adjustment of data acquisition frequency, the heat recovery system also dynamically responds to load fluctuations and heat changes. The heat recovery efficiency adjustment formula is as follows:

[0051]

[0052] Where η recovery is the adjusted heat recovery efficiency, ηnominal is the heat recovery efficiency under normal conditions, and μ is the heat recovery response coefficient.

[0053] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0054] The present invention dynamically adjusts the data collection cycle based on a deep learning model. This solution significantly improves the response speed and efficiency of the heat recovery system, while optimizing the economy of the system. In the case of a load surge, the system can detect load fluctuations in time and automatically increase the frequency of data collection, thereby quickly responding to temperature changes and power fluctuations, ensuring that waste heat can be recovered in time, and avoiding heat waste and overheating risks. When the load is stable, the system maintains the existing regular monitoring strategy to avoid resource waste and unnecessary energy consumption. Through this dynamic adjustment mechanism, the heat recovery system not only improves energy utilization efficiency when the load surges and reduces the potential risks caused by overheating, but also optimizes the use of resources and computing resources when the load is stable, thereby significantly improving the overall energy efficiency, economy and return on investment of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0056] Figure 1 The present invention is a flow chart of the method for recovering heat energy from heat dissipation of a power electronic converter. DETAILED DESCRIPTION

[0057] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0058] The present invention provides Figure 1 The heat recovery method for heat dissipation of a power electronic converter shown comprises the following steps:

[0059] Firstly, a time-driven strategy is used to regularly collect the working status data of the converter to ensure the continuity and controllability of the data;

[0060] The time-driven strategy of periodically collecting the working status data of the converter refers to continuously obtaining various types of monitoring data of the converter during operation at predetermined time intervals (such as every second, every minute, etc.). Specifically, the system automatically collects the working parameters of the converter according to the set time period, including temperature, current, voltage, load, power and other indicators. This periodic collection method ensures the consistency and traceability of the data, allowing the monitoring system to continuously track the operating status of the converter, thereby reflecting its working conditions in real time. This strategy is often used in systems that require continuous observation, especially when the load fluctuations are not large or the environmental changes are relatively stable. The time-driven method simplifies the data collection process and reduces the computational burden.

[0061] This strategy ensures data continuity and system monitorability, and maintains a comprehensive understanding of the converter's operating status even when the load is stable. In this way, the system can establish detailed operating history data, providing a solid foundation for subsequent data analysis and feature extraction.

[0062] Preprocessing the acquired raw data, and after the data preprocessing is completed, extracting key features reflecting the surge in converter load;

[0063] After collecting the raw data, data preprocessing is a crucial step. Preprocessing includes data cleaning, denoising, missing value filling, and standardization to ensure the accuracy and consistency of the data. Specifically, the cleaning step removes obvious outliers and erroneous data, the denoising step uses filtering technology to reduce the interference of sensor noise, the missing value filling fills the data gaps through interpolation or other methods, and the standardization process converts data of different dimensions into a unified scale. The purpose of this stage is to improve data quality so that subsequent feature extraction and analysis are more reliable and effective.

[0064] The extracted key features are analyzed in detail under the detection window, and the load changes of the converter are quantified by analyzing the key features;

[0065] Key features reflecting the surge in converter load are extracted therefrom, the extracted features including the converter internal impedance change and the on-state maintenance time of the converter internal switching elements. The extracted converter internal impedance change and the on-state maintenance time of the converter internal switching elements are analyzed under the detection window, and an internal impedance change rate reference value and a converter current on-time reference value are generated respectively. The converter load change is quantified by the internal impedance change rate reference value and the converter current on-time reference value.

[0066] The internal impedance of the converter shows an exponential growth, which usually indicates that the converter is in a state of load surge. When the load surges, the power conversion devices (such as IGBT, MOSFET) inside the converter will be driven by the sudden high load, causing their working conditions to change dramatically. In this process, the internal thermal effects and current fluctuations of the power converter will change the material properties of the electrical components, especially its internal impedance. With the sharp increase in load, the current demand inside the converter rises sharply, resulting in a significant increase in the current density flowing in the converter's switching elements and conductors, which increases the internal resistance. The increase in current density is directly related to the rate of current change, and the increase in impedance often shows nonlinear, exponential growth characteristics, indicating that when the converter responds to the load surge, the internal electrical components' impedance resistance to the current increases significantly. The exponential growth of impedance means that the converter's internal response capability encounters a bottleneck, the current flow becomes more difficult, the efficiency of power conversion decreases, and may cause additional heat generation and current fluctuations. Therefore, the exponential growth of the converter's internal impedance directly reflects the surge in load, and suggests that the converter may face problems such as reduced efficiency and increased thermal management pressure in this state.

[0067] The specific steps for analyzing the internal impedance change of the converter in the detection window to generate the reference value of the internal impedance change rate are as follows:

[0068] First, the original working state data is collected from the sensors of the converter, especially the current and voltage parameters. In order to ensure the accuracy of the data, the original data is preprocessed. In this way, accurate instantaneous current and voltage waveforms are obtained. The instantaneous impedance of the converter is calculated by the acquired data. The calculation expression is as follows:

[0069]

[0070] Where V(t) is the instantaneous voltage, which represents the value of the voltage at the output of the converter at time t; I(t) is the instantaneous current, which represents the value of the current at the output of the converter at time t; Z(t) is the instantaneous impedance, which represents the ratio between the voltage and current of the converter at time point t;

[0071] This step can directly reflect the impedance of the converter under different loads and is the basis for subsequent analysis.

[0072] Calculate the change rate of the internal impedance of the converter. The change rate of the impedance is an important indicator to measure the response degree of the converter to load fluctuations. In order to accurately capture the dynamic changes of the internal impedance of the converter, the weighted difference method is used to calculate the impedance. This method not only considers the instantaneous changes, but also enhances the correlation between the changes between the previous and next time points, thereby reducing the error. The calculation expression is as follows:

[0073]

[0074] In the formula, is the impedance change rate, which means the change rate of the internal impedance of the converter at time t, Z(t-1) is the instantaneous impedance at the moment before time t, α is the weighting coefficient, and n is the total number of time points;

[0075] This step can capture the instantaneous dynamics of impedance changes and enhance the response sensitivity to sudden or surging loads.

[0076] In the formula, α is the weighting coefficient, which represents the weight of the differential term between different time points when calculating the impedance change rate of the converter. Specifically, α adjusts the degree of influence of impedance changes on the overall rate calculation in different time intervals. Its function is to assign different weights according to the importance of changes in different time periods, ensuring that in the impedance change rate calculation, more important time periods (for example, drastic changes occurring in a short period of time) can obtain higher weights and thus be considered more. For parts with longer time intervals and slower changes, smaller weights are assigned. By reasonably setting and adjusting α, the impact between instantaneous changes and long-term changes can be effectively balanced, making the calculation of the impedance change rate more accurate and better reflecting the characteristics of load surges or other rapid changes.

[0077] The impedance change rate is converted into a characteristic value and a change rate characteristic function is constructed. This function can quantify the intensity of the change rate. A series of nonlinear transformations are performed to obtain the characteristic value reflecting the load fluctuation, thereby enhancing the sensitivity to the rate of change when the sudden load surges. The expression is as follows:

[0078]

[0079] Where C(t) is the characteristic value of the rate of change, which reflects the intensity of the change in the internal impedance of the converter at time t, and β is the expansion coefficient, which is used to adjust the amplification degree of the rate change;

[0080] This step increases the sensitivity to large changes (such as load surges) through an exponential function. When the load surges, the rate of impedance change increases significantly, resulting in a large increase in the value of C(t), which is a sign of the load surge.

[0081] The internal impedance change rate reference value is generated by taking a time-weighted average of the change rate characteristic value C(t) and combining it with the current load state. In order to accurately evaluate the load change of the current converter, this reference value not only reflects the intensity of the impedance change, but also takes into account the continuity and sharpness of the load change. This process combines the change pattern of the previous data to ensure the timeliness and accuracy of the reference value. The calculation expression is as follows:

[0082]

[0083] In the formula, R Z is the reference value of the internal impedance change rate, and γ is the weight coefficient, which is used to adjust the contribution of the change rate characteristic value C(t) to the total change.

[0084] The weighted summation formula can be used to comprehensively evaluate the impedance change intensity during the load surge. The reference value R Z The increase in indicates a dramatic change in the converter's internal impedance, suggesting that the load is surging.

[0085] The larger the internal impedance change rate reference value generated by analyzing the internal impedance change of the converter under the detection window, the greater the value means that the converter is in a load surge state. When the load surges, the power handling capacity of the internal electrical components of the converter (such as IGBT, MOSFET) will change rapidly, causing its internal impedance to rise sharply. The internal impedance change rate measures the rate at which the impedance changes over time. In the case of a load surge, the current demand increases sharply, and the change of the internal impedance usually presents a higher rate, so the impedance change rate value is larger. On the contrary, when the load is stable, the current change of the converter is relatively stable, and the change of the internal impedance is relatively slow, resulting in a smaller impedance change rate value. Through real-time data analysis under the monitoring window, the generated internal impedance change rate reference value can be used as an indicator of load fluctuation. When the reference value is large, it means that the load is increasing sharply and the converter needs to adjust quickly to cope with higher power demand and heat dissipation pressure; when the reference value is small, it means that the load is stable and the converter is operating normally.

[0086] A significant increase in the on-state maintenance time of the switch elements inside the converter usually indicates that the converter is in a load surge state. This is because when the load surges, the converter needs to convert more electrical energy into the power required by the load, which requires its internal switch elements (such as IGBT or MOSFET) to extend the on-time to ensure that sufficient current flows into the load. Under normal circumstances, the on-time of the switch elements is relatively short to optimize the conversion efficiency and reduce switching losses. However, in the case of a sharp increase in load, in order to cope with the sudden current demand, the converter will automatically adjust the control strategy to extend the on-time of the switch elements, thereby increasing the continuity of current transmission. This change is a direct manifestation of the load surge, because it reflects the converter's response to sudden load changes, and the extended on-time ensures that the current can be supplied to the load in a timely and stable manner (referring to the converter's need to adjust the working state of its internal switch elements to ensure that it can provide sufficient current supply when the load demand surges to meet the power requirements of the load equipment). If the on-time is too long, it may lead to increased heat loss of the switch elements, which in turn affects the efficiency and stability of the converter. Therefore, the increase in the on-state maintenance time is a typical dynamic change when the load surges, indicating that the converter is adjusting its operating parameters to cope with the changing load pressure.

[0087] The specific steps of analyzing the on-state maintenance time of the switch element inside the converter to generate the converter current on-time reference value are as follows:

[0088] First, relevant switch element status data is collected in real time from the converter control system, including switching frequency, switching duration, load current fluctuation and temperature change. The acquired data provides a basis for subsequent analysis and can accurately reflect the switching operation of the converter under different load conditions. By monitoring the precise moment of each switching operation, the on-time of the switch element is extracted, that is, the duration from on to off of each switch. The formula is as follows:

[0089] T conduct (i) = ∫ i i+1 State(t)dt

[0090] Where, T conduct (i) is the on-time of the i-th switch operation, State(t) is a time function, which indicates the on-state of the switch element at any time t. It is used to describe the on-state of the switch element. When State(t) = 1, it indicates that the switch element is in the on-state, and when State(t) = 0, it indicates that the switch element is in the off-state. i i+1State(t)dt is the calculation of the on-time of the switch element between the i-th to the i+1-th switching operations. By integrating the time interval, the on-time of each switching cycle of the converter can be obtained, providing a basis for subsequent load fluctuation and heat recovery evaluation;

[0091] The on-time of each switching cycle is calculated by integration to obtain the on-time of each operation cycle.

[0092] After obtaining the on-time, the on-time fluctuation is analyzed. When the load increases sharply, the on-time is usually significantly extended and fluctuates greatly; when the load is stable, the on-time is relatively constant and fluctuates less. The cumulative change of the on-time change is calculated to reflect the severity of the load fluctuation. The calculation expression is as follows:

[0093]

[0094] Where, T conduct (i-1) is the on-time of the i-1th switch operation, i.e., the on-time of the previous switch operation, m is the total number of switch operations within the detection window, and ω is an exponential coefficient, which is used to weight and amplify the impact of on-time differences (i.e., on-time fluctuations), especially in the case of a load surge, ΔT conduct is the on-time variation, which indicates the total variation of the on-time of the switch elements inside the converter within the detection window;

[0095] If ΔT conduct A significant increase in the value indicates a surge in the load, whereas a decrease in the value indicates a relatively stable load.

[0096] According to the on-time change ΔT conduct Identify the load surge, introduce a dynamic weight coefficient to adjust the importance of the on-time change, and automatically optimize the weight coefficient based on the historical load pattern. The adjustment of this weight coefficient is based on the load change trend of the converter to ensure that the system can respond more sensitively when the load surges. By recursively updating the weight coefficient, the response sensitivity to the on-time change can be enhanced when the load fluctuates greatly. The expression is as follows:

[0097]

[0098] Where W conduct is the on-time weight coefficient, which is used to measure the influence of the current on-time of the converter on the load fluctuation. conduct (i) is the change in on-time of the i-th switching operation, ΔT conduct (j) is the change in on-time of the jth switching operation, It is the exponential gain factor, which is used to adjust the exponential parameter of the sensitivity of the whole conduction time change;

[0099] Through this step, the system can dynamically adjust the weights based on historical load fluctuations to reflect the relative importance of on-time when the load surges.

[0100] Based on the on-time T conduct (i) and the on-time weight coefficient W conduct Generate the converter current conduction time reference value, the generation formula is as follows:

[0101]

[0102] Where γ is the exponential weighting factor, T ref It is the reference value of the converter current conduction time.

[0103] This step combines the fluctuation of the on-time and the weight coefficient to provide an accurate reference value for the load status of the converter, helping the system to respond promptly when the load surges.

[0104] The exponential weighting factor γ is a weight coefficient introduced when processing time series data, which aims to emphasize the importance of certain specific time points in the data in a nonlinear way. In the on-time analysis of the converter, the role of γ is to increase the response to longer on-times. When the load surges, the switching elements of the converter need to remain on for a longer time to adapt to the rapidly changing load demand. By using the exponential weighting factor γ, the system is able to give greater weight to longer on-times, which helps to capture signs of load surges more sensitively. Specifically, a larger γ value will make the time window with a longer on-time more significant to the final reference value T ref The contribution of γ is greater, thereby improving the system's ability to respond to severe load fluctuations. In short, the role of γ is to amplify the impact of load surges so that the system can adjust heat recovery and other control strategies in a timely manner.

[0105] The larger the performance value of the converter current on-time reference value generated by analyzing the on-state maintenance time of the switch elements inside the converter, the more it indicates that the converter is in a state of load surge. The on-time reference value reflects the on-time duration of the switch elements (such as IGBT and MOSFET) inside the converter in response to the load demand. When the load surges, the converter needs to provide more power to the load, and the current demand rises sharply, resulting in a corresponding extension of the on-time of the switch elements. Because a longer on-time means that the converter takes longer to complete stable power output during the current transmission process, which is usually a typical feature of a load surge. When the converter is in a stable load state, the current changes less, the load demand is relatively stable, the on-time of the switch elements is maintained within the normal range, and the reference value is smaller. Therefore, by monitoring the current on-time reference value of the converter, if its performance value is larger, it can be inferred that the load may have surged, and vice versa, if the performance value is smaller, it means that the converter is under stable load conditions.

[0106] The quantified feature data is input into a pre-trained deep learning model, and the load change of the converter is intelligently evaluated through the deep learning model;

[0107] The internal impedance change rate reference value and the converter current conduction time reference value generated after analysis are input into a pre-learned deep learning model, and a load fluctuation index is generated through the deep learning model. The load fluctuation index is used to perform an intelligent evaluation of the converter load change.

[0108] A pre-learned deep learning model is an artificial intelligence model that has been trained and optimized on a specific historical data set to identify, predict, and process load fluctuations and other dynamic behaviors of the converter under different operating conditions. In this model, a large amount of converter operation data is usually used, including various parameters such as temperature, load changes, current, voltage, internal impedance, and the on-time of the switching element. By using this historical data, the deep learning model can learn the complex operating mode of the converter under different load conditions, the characteristics of load fluctuations, and its response to various dynamic changes. The core advantage of the deep learning model is that it can handle highly complex and nonlinear relationships and continuously optimize the prediction ability through training. The training process includes selecting a suitable algorithm (such as convolutional neural network, recurrent neural network, etc.), setting an appropriate loss function, and adjusting the model parameters so that the model can predict the load changes under different operating conditions as accurately as possible. This training process is usually continued until the model's performance on the validation set and test set reaches the expected accuracy.

[0109] In the intelligent evaluation of load fluctuations applied to converters, the pre-learned deep learning model uses the input features such as the internal impedance change rate and the current conduction time to intelligently evaluate the load fluctuations of the converter. These input features are obtained by analyzing and extracting the data during the operation of the converter, and they can reflect the key dynamics of load surges or fluctuations. The deep learning model uses the input feature values ​​to generate a load fluctuation index based on the rules and patterns learned during its training process. This load fluctuation index is a numerical indicator that can reflect the degree and nature of the current load change of the converter, thereby helping the system to determine whether load regulation, optimization of heat recovery mode or other response strategies are needed. The deep learning model can identify the high-order correlations between various input features through a complex neural network structure, and dynamically adjust the evaluation results based on these features. As the model is continuously updated and optimized, it will be able to provide more and more accurate load fluctuation indexes, helping the intelligent control system to respond to load fluctuations quickly and accurately, and improving the overall stability and economic benefits of the system.

[0110] The deep learning model is not limited here, and can realize the internal impedance change rate reference value R Z and the converter current conduction time reference value T ref After comprehensive analysis, the load fluctuation index Load is generated flu In order to realize the technical solution of the present invention, the present invention provides a specific implementation method;

[0111] Load Fluctuation Index Load flu The generation formula is as follows:

[0112]

[0113] Where, q1 and q2 are the reference values ​​of internal impedance change rate R Z and the converter current conduction time reference value T ref The preset proportional coefficient, and q1 and q2 are both greater than 0.

[0114] It can be seen from the load fluctuation index that the larger the internal impedance change rate reference value generated after analyzing the internal impedance change of the converter under the detection window, the larger the converter current on-time reference value generated after analyzing the on-state maintenance time of the internal switching element of the converter, then the larger the load fluctuation index performance value generated when the converter load change is intelligently evaluated by the pre-learned deep learning model, indicating that the probability that the converter is in a load surge is greater, and vice versa, it indicates that the probability that the converter is in a load surge is smaller.

[0115] Based on the evaluation results of the deep learning model, the load change of the converter is divided into two states: load surge and load stability.

[0116] The load fluctuation index generated by the pre-learned deep learning model when intelligently evaluating the load change of the converter is compared and analyzed with the pre-set load fluctuation index reference threshold, and the load change of the converter is divided. The division steps are as follows:

[0117] If the load fluctuation index is greater than the load fluctuation index reference threshold, the load change of the converter is classified as a load surge;

[0118] If the load fluctuation index is less than or equal to the load fluctuation index reference threshold, the load change of the converter is classified as load stability.

[0119] Load surge refers to a sudden and dramatic increase in load demand during the operation of the converter, usually beyond the expected normal operating range. Load stability refers to the converter's ability to maintain a relatively constant load level over a period of time, with a small load change range and the system in normal operation.

[0120] For stable load, continue to adopt the time-driven periodic data collection strategy and maintain the existing monitoring frequency to ensure that the heat recovery system maintains efficient operation under stable load conditions;

[0121] For stable load conditions, the time-driven periodic data collection strategy is continued and the existing monitoring frequency is maintained to ensure that the system can maintain continuous and efficient heat recovery when the load changes smoothly. Stable load usually means that the working state of the converter does not fluctuate violently and the heat generation of the system is relatively stable, so there is no need to frequently adjust the data collection cycle. By maintaining the existing monitoring frequency, the intelligent control system can obtain the necessary operating data in real time and optimize the heat recovery process based on this data without causing unnecessary computing burden or waste of resources. This approach helps ensure that the heat recovery system can still operate efficiently during stable load periods, while avoiding excessive response and waste of resources, maximizing the heat recovery efficiency and ensuring system stability and economy.

[0122] In case of load surge, based on the evaluation results of deep learning model, the data collection cycle is dynamically adjusted, the collection frequency is increased, and the load fluctuation and heat change are responded to in time, so as to ensure the timely recovery of waste heat;

[0123] For load surges, based on the evaluation results of the deep learning model, the data collection cycle is dynamically adjusted, the collection frequency is increased, and the load fluctuations and heat changes are responded to in a timely manner, thereby ensuring the timely recovery of waste heat. The specific steps are as follows:

[0124] When the load fluctuation index Load fluWhen the load fluctuation index is greater than the reference threshold (i.e., entering high load fluctuation), the data collection cycle is dynamically adjusted according to the evaluation results. In order to timely capture the changes caused by the load surge, the data collection frequency needs to increase exponentially, thereby ensuring that the real-time monitoring system can quickly respond to load fluctuations and heat changes. The formula is as follows:

[0125]

[0126] In the formula, Load flu is the load fluctuation index, Load ref is the reference threshold of the load fluctuation index, e is the weaving base number, θ is the adjustment coefficient, which controls the rate of increase of the sampling frequency when the load fluctuation index exceeds the reference threshold, and f base is the basic sampling frequency, f sampling is the frequency of new data collection;

[0127] This step means that when the load fluctuation index exceeds the reference threshold, the frequency of data collection will increase exponentially to ensure that the system can collect more data in a shorter time, accurately evaluate and respond to the heat changes caused by the load surge, and ensure the efficient operation of the heat recovery system.

[0128] With the dynamic adjustment of data acquisition frequency, the heat recovery system also responds dynamically to load fluctuations and heat changes. The purpose of increasing the acquisition frequency is to ensure that the system can obtain key information such as temperature fluctuations and power output fluctuations caused by load surges in real time, so as to optimize the operation of the thermoelectric conversion module in time. The heat recovery efficiency adjustment formula is as follows:

[0129]

[0130] Where η recovery is the adjusted heat recovery efficiency, η nominal is the heat recovery efficiency under normal conditions, and μ is the heat recovery response coefficient.

[0131] With the load fluctuation index Load flu Exceeding the reference threshold Load ref , the heat recovery efficiency is improved exponentially, ensuring that waste heat can be recovered and converted into usable energy to the maximum extent when the load surges. This dynamic adjustment not only improves the efficiency of heat recovery, but also optimizes the stability and economy of the system, avoiding heat loss and waste of resources caused by delayed response.

[0132] For load surges, the steps of dynamically adjusting the data collection cycle based on the model evaluation results, increasing the collection frequency, and responding to load fluctuations and heat changes in a timely manner are aimed at improving the system's responsiveness to sudden load changes, so as to ensure that the heat recovery system can capture the necessary operating data in a timely manner and make adaptive adjustments when the load changes sharply. Load surges are usually accompanied by rapid changes in important parameters such as temperature, power, and current inside the converter, which not only affect the operating efficiency of the converter, but also may cause overheating or failure risks. Therefore, by dynamically adjusting the data collection cycle, higher frequency monitoring can be achieved, thereby obtaining more fine-grained system operation data. Increasing the collection frequency helps to timely identify abnormal conditions such as heat accumulation and power fluctuations caused by load surges, and avoid energy waste and equipment damage caused by delayed response. In particular, when the load fluctuation index is greater than the set reference threshold, the power conversion efficiency and heat dissipation requirements of the converter may fluctuate violently. Obtaining this information in a timely manner helps the system to quickly adjust the heat recovery mechanism, optimize the heat dissipation path, and improve energy efficiency and stability. Through this dynamic adjustment mechanism, the intelligent control system can ensure that the converter operates efficiently and stably during load surges, thereby reducing energy waste, improving the economy of the system, and reducing the impact of risks such as overheating on equipment life and operational reliability. Therefore, this step is crucial to ensuring the long-term stable operation of the entire system under complex working conditions.

[0133] The present invention dynamically adjusts the data collection cycle based on a deep learning model. This solution significantly improves the response speed and efficiency of the heat recovery system, while optimizing the economy of the system. In the case of a load surge, the system can detect load fluctuations in time and automatically increase the frequency of data collection, thereby quickly responding to temperature changes and power fluctuations, ensuring that waste heat can be recovered in time, and avoiding heat waste and overheating risks. When the load is stable, the system maintains the existing regular monitoring strategy to avoid resource waste and unnecessary energy consumption. Through this dynamic adjustment mechanism, the heat recovery system not only improves energy utilization efficiency when the load surges and reduces the potential risks caused by overheating, but also optimizes the use of resources and computing resources when the load is stable, thereby significantly improving the overall energy efficiency, economy and return on investment of the system.

[0134] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0135] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

[0136] It should be noted that, in this article, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0137] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0138] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0139] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0140] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0141] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0142] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0143] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for recovering heat energy from heat dissipation in a power electronic converter, characterized in that: The following steps are involved: Firstly, a time-driven strategy is used to regularly collect the working status data of the converter to ensure the continuity and controllability of the data; Preprocessing the acquired raw data, and after the data preprocessing is completed, extracting key features reflecting the surge in converter load; The extracted key features are analyzed in detail under the detection window, and the load changes of the converter are quantified by analyzing the key features; The quantified feature data is input into a pre-trained deep learning model, and the load change of the converter is intelligently evaluated through the deep learning model; Based on the evaluation results of the deep learning model, the load change of the converter is divided into two states: load surge and load stability. For stable load, continue to adopt the time-driven periodic data collection strategy and maintain the existing monitoring frequency to ensure that the heat recovery system maintains efficient operation under stable load conditions; In the case of load surges, the data collection cycle is dynamically adjusted based on the evaluation results of the deep learning model, the collection frequency is increased, and load fluctuations and heat changes are responded to in a timely manner, thereby ensuring that waste heat can be recovered in a timely manner.

2. The heat recovery method for power electronic converter according to claim 1, characterized in that: Key features reflecting the surge in converter load are extracted therefrom, the extracted features including the converter internal impedance change and the on-state maintenance time of the converter internal switching elements. The extracted converter internal impedance change and the on-state maintenance time of the converter internal switching elements are analyzed under the detection window, and an internal impedance change rate reference value and a converter current on-time reference value are generated respectively. The converter load change is quantified by the internal impedance change rate reference value and the converter current on-time reference value.

3. The heat recovery method for heat dissipation of a power electronic converter according to claim 2, characterized in that: The specific steps for analyzing the internal impedance change of the converter in the detection window to generate the reference value of the internal impedance change rate are as follows: First, the original working status data is collected from the sensors of the converter. In order to ensure the accuracy of the data, the original data is preprocessed and the instantaneous impedance of the converter is calculated by the acquired data. The calculation expression is as follows: Where V(t) is the instantaneous voltage, which represents the value of the voltage at the output of the converter at time t, I(t) is the instantaneous current, which represents the value of the current at the output of the converter at time t, and Z(t) is the instantaneous impedance; Calculate the change rate of the internal impedance of the converter. The change rate of the impedance is an important indicator to measure the converter's response to load fluctuations. The calculation expression is as follows: In the formula, is the impedance change rate, Z(t-1) is the instantaneous impedance at the moment before time t, α is the weighting coefficient, and n is the total number of time points; The impedance change rate is converted into a characteristic value, and a change rate characteristic function is constructed to obtain a characteristic value reflecting load fluctuations, thereby enhancing the sensitivity to the change rate when the sudden load surges. The expression is as follows: Where C(t) is the characteristic value of the rate of change, and β is the expansion coefficient; The internal impedance change rate reference value is generated by taking a time-weighted average of the change rate characteristic value C(t) and combining it with the current load state. The calculation expression is as follows: In the formula, R Z is the reference value of the internal impedance change rate, and γ is the weight coefficient, which is used to adjust the contribution of the change rate characteristic value C(t) to the total change.

4. The heat recovery method for power electronic converter according to claim 2, characterized in that: The specific steps of analyzing the on-state maintenance time of the switch element inside the converter to generate the converter current on-time reference value are as follows: First, relevant switch element status data is collected in real time from the converter control system. The acquired data provides a basis for subsequent analysis. By monitoring the precise moment of each switch operation, the on-time of the switch element is extracted, that is, the duration from on to off of each switch. The formula is as follows: T conduct (i)=∫ i i+1 State(t)dt Where, T conduct (i) is the conduction time of the i-th switch operation, State(t) is a time function, which represents the conduction state of the switch element at any time t, ∫ i i+1 State(t)dt is the calculation of the conduction time of the switch element between the i-th to the i+1-th switching operations; After obtaining the on-time, the on-time fluctuation analysis is performed to calculate the cumulative change of the on-time. The calculation expression is as follows: Where, T conduct (i-1) is the on-time of the i-1th switch operation, i.e., the on-time of the previous switch operation, m is the total number of switch operations within the detection window, ω is an exponential coefficient, ΔT conduct is the on-time variation; According to the on-time change ΔT conduct Identify the load surge, introduce a dynamic weight coefficient to adjust the importance of the on-time change, and automatically optimize the weight coefficient based on the historical load pattern. The expression is as follows: Where W conduct is the on-time weight coefficient, ΔT conduct (i) is the change in on-time of the i-th switching operation, ΔT conduct (j) is the change in on-time of the jth switching operation, is the exponential magnification factor; Based on the on-time T conduct (i) and the on-time weight coefficient W conduct Generate the converter current conduction time reference value, the generation formula is as follows: Where γ is the exponential weighting factor, T ref It is the reference value of the converter current conduction time.

5. The heat recovery method for power electronic converter according to claim 2, characterized in that: The internal impedance change rate reference value and the converter current conduction time reference value generated after analysis are input into a pre-learned deep learning model, and a load fluctuation index is generated through the deep learning model. The load fluctuation index is used to perform an intelligent evaluation of the converter load change.

6. The heat recovery method for power electronic converter according to claim 5, characterized in that: The load fluctuation index generated by the pre-learned deep learning model when intelligently evaluating the load change of the converter is compared and analyzed with the pre-set load fluctuation index reference threshold, and the load change of the converter is divided. The division steps are as follows: If the load fluctuation index is greater than the load fluctuation index reference threshold, the load change of the converter is classified as a load surge; If the load fluctuation index is less than or equal to the load fluctuation index reference threshold, the load change of the converter is classified as load stability.

7. The method for recovering heat energy from heat dissipation of a power electronic converter according to claim 6, characterized in that: For load surges, based on the evaluation results of the deep learning model, the data collection cycle is dynamically adjusted, the collection frequency is increased, and the load fluctuations and heat changes are responded to in a timely manner, thereby ensuring the timely recovery of waste heat. The specific steps are as follows: When the load fluctuation index Load flu When it is greater than the reference threshold of the load fluctuation index, the data collection cycle is dynamically adjusted according to the evaluation results to ensure that the real-time monitoring system can respond quickly to load fluctuations and heat changes. The formula is as follows: In the formula, Load flu is the load fluctuation index, Load ref is the reference threshold of the load fluctuation index, e is the weaving and dyeing base, θ is the adjustment coefficient, and f base is the basic sampling frequency, f sampling is the frequency of new data collection; With the dynamic adjustment of data acquisition frequency, the heat recovery system also dynamically responds to load fluctuations and heat changes. The heat recovery efficiency adjustment formula is as follows: Where η recovery is the adjusted heat recovery efficiency, η nominal is the heat recovery efficiency under normal conditions, and μ is the heat recovery response coefficient.

Images