Electrical control cabinet cooling device and control method thereof

Abstract

The invention relates to the technical field of semiconductor electrical equipment cooling, and discloses an electrical control cabinet cooling device and a control method thereof. The thermal load trend, the maximum temperature rise expected value and the temperature rise rate change interval in the electrical control cabinet can be predicted in real time based on historical operation data training and the current operation state. The method comprises the following steps: carrying out prospective evaluation on a cooling demand by combining a feature extraction mechanism and a time sequence modeling capability so as to output a more refined predictive control parameter, and finally generating a multi-dimensional cooling control instruction according to the predictive control parameter and a current operation state so as to realize dynamic scheduling of a cooling assembly. And the response timeliness and the energy efficiency matching degree are obviously improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cooling semiconductor electrical equipment, and in particular to a cooling device and a control method for an electrical control cabinet. Background Art

[0002] Electrical control cabinets, as an essential component of industrial automation systems, are widely used in power distribution, control, and communication systems. They integrate numerous high-power electrical components and control modules, which can easily generate significant heat during extended operation. Inadequate heat dissipation or improper temperature control can lead to component damage, decreased system performance, and even safety hazards.

[0003] Traditional cooling solutions for electrical control cabinets typically employ fixed fan forced ventilation or mechanical cooling triggers based on set temperature limits. These strategies rely on single-point temperature detection and static thresholds, making them incapable of accurately responding to dynamic changes in the thermal load within the cabinet. When equipment loads fluctuate rapidly or the external environment fluctuates significantly, fixed strategies are prone to cooling lag, excessive cooling, and energy waste, making them incapable of meeting the dual requirements of heat dissipation efficiency and energy utilization under complex operating conditions. Summary of the Invention

[0004] In view of this, the object of the present invention is to provide an electrical control cabinet cooling device and a control method thereof, so as to solve the technical problem of the lack of accurate evaluation and dynamic response control of the temperature rise process in the existing control process of the electrical control cabinet cooling.

[0005] The first aspect of the present invention discloses a cooling device for an electrical control cabinet, which includes a data acquisition module, a cooling load prediction module, a control module and an execution module; wherein,

[0006] The data acquisition module is used to collect relevant operating data of the electrical control cabinet, including cabinet temperature data, current and voltage load data, cabinet component operating status parameters and ambient temperature and humidity information;

[0007] The cooling load prediction module is used to train a cooling load prediction model based on historical operating data. The cooling load prediction model predicts the thermal load trend, expected maximum temperature rise, and temperature rise rate change range based on current operating data as the model prediction results. The prediction control parameters are calculated based on the model prediction results. The prediction control parameters include the expected cooling power, the cooling component preload response time, and the wind speed classification scheduling scheme.

[0008] The control module is used to generate cooling control instructions based on the predicted control parameters and operating data;

[0009] The execution module is used to receive the cooling control instruction and execute a corresponding cooling operation according to the cooling control instruction.

[0010] Furthermore, the cooling load prediction module includes a feature extraction unit, a time trend modeling unit and a prediction output unit; wherein,

[0011] The feature extraction unit is used to perform data preprocessing and timing analysis operations based on the operating data to obtain multidimensional key factors; the multidimensional key factors include the temperature difference between the inside and outside of the cabinet, the intensity of the operating load fluctuation, the current fluctuation cycle, the start and stop status of the cabinet components, and the distribution of local high-heat areas.

[0012] Furthermore, the feature extraction unit is further configured to extract time series features that are sensitive to power mutations from the operating data; the time series features include short-term slope changes of current and / or voltage, acceleration factors of load power, and abnormal temperature rise rates of heat-sensitive components;

[0013] After extracting the time series features, the fluctuation concentration of the time series features is analyzed, and the time period with abnormal concentration of features is regarded as the high fever risk triggering area.

[0014] Furthermore, the time trend modeling unit is used to take the extracted multidimensional key factors, time series features and high heat risk trigger areas as input features, and construct a cooling load prediction model based on the fusion of the gated recurrent unit network and the residual attention mechanism.

[0015] Furthermore, the cooling load prediction model constructed based on the fusion of the gated recurrent unit network and the residual attention mechanism specifically includes:

[0016] Based on the input characteristics of the time trend modeling unit, a nonlinear temperature rise trend curve within a preset length sliding time window is constructed, and the exponentially weighted moving average method is used to separate short-term temperature rise fluctuations and medium- and long-term temperature rise trends;

[0017] The separated medium- and long-term temperature rise trend is used as the benchmark curve, and the residual sequence of the actual temperature rise sequence is constructed in combination with the short-term temperature rise fluctuation. The temperature rise trend is divided into multiple fitting stages based on the benchmark curve. Local fitting and residual analysis are performed on each stage based on the residual sequence to extract the residual change characteristics.

[0018] According to the residual variation characteristics and the preset mutation identification threshold, the load mutation boundary and abnormal temperature rise segment are identified to form the mutation focus section;

[0019] A sequence modeling structure is constructed based on the gated recurrent unit network. The residual change features are integrated into the sequence modeling structure to form a multi-head attention weight distribution, and the mutation attention segment is focused modeled to finally obtain a cooling load prediction model.

[0020] Furthermore, the prediction output unit is used to calculate the prediction control parameters based on the model prediction results and the preset control parameter mapping rules, and transmit the prediction control parameters to the control module.

[0021] Furthermore, the control parameter mapping rules specifically include:

[0022] Generate the expected cooling power value based on the joint judgment rule mapping of the temperature rise rate change range and the current heat load slope in the cabinet;

[0023] The cooling component preload response time is calculated based on the advance warning time of the maximum temperature rise expected value relative to the temperature over-limit threshold and the thermal inertia response parameters of the cabinet components;

[0024] Based on the slope change section of the heat load trend curve within the target time period and the short-term fluctuation frequency matching graded wind speed adjustment curve, a wind speed graded scheduling plan is formed.

[0025] Furthermore, the control module includes a temperature rise dynamic evaluation unit and a cooling strategy control unit; wherein,

[0026] The temperature rise dynamic assessment unit is used to determine the current cabinet temperature change rate and power input trend based on operating data, and to assess the current temperature rise risk level and risk level change trend by combining the maximum temperature rise expected value and temperature rise rate change range in the predictive control parameters and the preset cabinet enclosure level parameters;

[0027] The cooling strategy control unit is used to determine the use strategy of the prediction control parameters according to the temperature rise risk level and the risk level change trend, and generate corresponding cooling control instructions based on the use strategy of the prediction control parameters.

[0028] Furthermore, the cooling strategy control unit also sets an advance cooling strategy. When it is identified that the temperature rise risk level continues to rise and it is determined according to the model prediction results that the high heat load stage is about to arrive, the pre-cooling start operation is preferentially performed based on the advance cooling strategy.

[0029] A second aspect of the present invention discloses a cooling control method for an electrical control cabinet, which is applied to the device disclosed in the first aspect. The method comprises:

[0030] Collect relevant operating data of the electrical control cabinet, including cabinet temperature data, current and voltage load data, cabinet component operating status parameters, and ambient temperature and humidity information;

[0031] A cooling load prediction model is trained based on historical operating data. The model then uses the current operating data to predict the thermal load trend, expected maximum temperature rise, and temperature rise rate variation range as the model prediction results. Predictive control parameters are calculated based on the model prediction results. These parameters include the expected cooling power, cooling component preload response time, and wind speed grading scheduling scheme.

[0032] Generate cooling control instructions based on predictive control parameters and operating data;

[0033] Execute corresponding cooling operations according to the cooling control instruction.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] By constructing a cooling load prediction model driven by operational data, this invention can predict the internal thermal load trend, expected maximum temperature rise, and temperature rise rate variation range of electrical control cabinets in real time based on historical operational data training and current operational status. Furthermore, by combining feature extraction mechanisms with time series modeling capabilities, this model performs a forward-looking assessment of cooling demand, outputting more refined predictive control parameters. Finally, based on these predictive control parameters and current operational status, multi-dimensional cooling control instructions are generated, enabling dynamic scheduling of cooling components, significantly improving responsiveness and energy efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0037] Figure 1 This is a schematic structural diagram of a cooling device for an electrical control cabinet disclosed in Embodiment 1 of the present invention;

[0038] Figure 2 The present invention is a flowchart of a method for controlling the temperature drop and cooling of an electrical control cabinet disclosed in another embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0040] Example 1

[0041] The first aspect of the present invention discloses a cooling device for an electrical control cabinet. Figure 1 , Figure 1This is a schematic diagram of the structure of a cooling device for an electrical control cabinet disclosed in an embodiment of the present invention, which includes a data acquisition module, a cooling load prediction module, a control module, and an execution module; wherein,

[0042] The data acquisition module is used to collect relevant operating data of the electrical control cabinet, including cabinet temperature data, current and voltage load data, cabinet component operating status parameters and ambient temperature and humidity information;

[0043] The cooling load prediction module is used to train a cooling load prediction model based on historical operating data. The cooling load prediction model predicts the thermal load trend, expected maximum temperature rise, and temperature rise rate change range based on current operating data as the model prediction results. The prediction control parameters are calculated based on the model prediction results. The prediction control parameters include the expected cooling power, the cooling component preload response time, and the wind speed classification scheduling scheme.

[0044] The control module is used to generate cooling control instructions based on the predicted control parameters and operating data;

[0045] The execution module is used to receive the cooling control instruction and perform the corresponding cooling operation according to the cooling control instruction.

[0046] Furthermore, the cooling load prediction module includes a feature extraction unit, a time trend modeling unit and a prediction output unit; wherein,

[0047] The feature extraction unit is used to perform data preprocessing and timing analysis operations based on the operating data to obtain multidimensional key factors; the multidimensional key factors include the temperature difference between the inside and outside of the cabinet, the intensity of the operating load fluctuation, the current fluctuation cycle, the start and stop status of the cabinet components, and the distribution of local high-heat areas.

[0048] Furthermore, the feature extraction unit is further configured to extract time series features that are sensitive to power mutations from the operating data; the time series features include short-term slope changes of current and / or voltage, acceleration factors of load power, and abnormal temperature rise rates of heat-sensitive components;

[0049] After extracting the time series features, the fluctuation concentration of the time series features is analyzed, and the time period with abnormal concentration of features is regarded as the high fever risk triggering area.

[0050] In an embodiment of the present invention, the purpose of the feature extraction unit is to conduct an in-depth analysis of the operating data of the electrical control cabinet to extract multidimensional key factors and time series characteristics that reflect the trend of thermal load changes and mutation behavior, thereby providing highly representative and highly sensitive input variables for subsequent modeling links.

[0051] Specifically, the feature extraction unit first preprocesses the collected operational data, including outlier removal, data interpolation and completion, normalization, and timestamp alignment, to ensure the continuity, stability, and timeliness of the data for subsequent analysis. After preprocessing, it further performs time series analysis, analyzing trends, identifying fluctuations, and extracting cycles across the time dimension of various operational data sequences to generate multidimensional key factors.

[0052] Multidimensional key factors include but are not limited to:

[0053] Temperature difference between inside and outside the cabinet: calculated by comparing the temperature inside the cabinet with the current ambient temperature, used to measure the heat dissipation gradient;

[0054] Operating load fluctuation intensity: calculated based on current and / or voltage time series, reflecting load stability;

[0055] Current fluctuation cycle: Analyze the periodic components of the current curve to identify periodic high-load operation characteristics;

[0056] Cabinet component start and stop status: count the number of times each component starts or stops within a certain time window to obtain the operation switching frequency;

[0057] Distribution of local high-temperature areas: Combine multi-point temperature sensing data to construct a heat distribution map inside the cabinet, and use spatial thermal clustering methods to delineate high-temperature hotspot areas.

[0058] In addition, the feature extraction unit is also used to identify timing features that are sensitive to sudden power changes, including but not limited to:

[0059] Short-term slope changes of current and / or voltage: Calculate the first-order difference of current and / or voltage through a sliding window to identify fast-changing segments;

[0060] Load power acceleration factor: Extract the load power change rate trend based on the second-order difference;

[0061] Abnormal temperature rise rate of heat-sensitive components: Calculate the temperature change slope of a specific sensing point and combine it with the heat capacity property to determine whether it is abnormal.

[0062] After completing the above feature extraction, we further calculate the concentration of fluctuations of each time series feature. This means analyzing whether the fluctuation amplitude of a certain type of feature within the sliding window is significantly concentrated within a short period of time. For example, the concentration degree is evaluated by the local variance value. When the fluctuation is significantly concentrated, the corresponding time period is marked as a high fever risk trigger area, indicating the potential risk of short-term thermal mutation.

[0063] Through the aforementioned feature extraction, the embodiments of the present invention not only capture thermal trends under normal operating conditions but also identify high-frequency, large-fluctuation, and critical operating anomalies. The extracted multidimensional key factors, temporal characteristics, and high-heat risk trigger areas serve as important input features for the subsequent cooling load prediction model. This helps improve the model's ability to identify chronic heat accumulation and short-term temperature rise mutations, enhances the foresight and accuracy of cooling strategies, and ultimately enables dynamic perception and intelligent control of the temperature rise status of electrical control cabinets.

[0064] Furthermore, the time trend modeling unit is used to take the extracted multidimensional key factors, time series features and high-heat risk trigger areas as input features, and construct a cooling load prediction model based on the fusion of the gated recurrent unit network and the residual attention mechanism.

[0065] Furthermore, the cooling load prediction model is constructed based on the fusion of the gated recurrent unit network and the residual attention mechanism, which specifically includes:

[0066] Based on the input characteristics of the time trend modeling unit, a nonlinear temperature rise trend curve within a preset length sliding time window is constructed, and the exponentially weighted moving average method is used to separate short-term temperature rise fluctuations and medium- and long-term temperature rise trends;

[0067] The separated medium- and long-term temperature rise trend is used as the benchmark curve, and the residual sequence of the actual temperature rise sequence is constructed in combination with the short-term temperature rise fluctuation. The temperature rise trend is divided into multiple fitting stages based on the benchmark curve. Local fitting and residual analysis are performed on each stage based on the residual sequence to extract the residual change characteristics.

[0068] According to the residual variation characteristics and the preset mutation identification threshold, the load mutation boundary and abnormal temperature rise segment are identified to form the mutation focus section;

[0069] A sequence modeling structure is constructed based on the gated recurrent unit network. The residual change features are integrated into the sequence modeling structure to form a multi-head attention weight distribution, and the mutation attention segment is focused modeled to finally obtain a cooling load prediction model.

[0070] In this implementation, the time trend modeling unit is designed to perform deep time series modeling on the extracted multidimensional key factors, temporal features, and high-heat risk trigger areas. This allows the construction of a cooling load prediction model that can dynamically perceive complex thermal trends. This model, which integrates a gated recurrent unit (GRU) network with a residual attention mechanism, possesses bidirectional recognition capabilities for nonlinear heat accumulation and sudden change characteristics.

[0071] Specifically, a nonlinear temperature rise trend curve within a sliding time window is first constructed based on the input features. This curve uses the sampling time series as the horizontal axis and key thermal characteristics (such as the temperature difference between the inside and outside of the cabinet and the intensity of operating load fluctuations) as the vertical axis, forming an original time series that reflects temperature rise changes. In addition, to remove random fluctuations and improve modeling stability, the exponentially weighted moving average (EWMA) method is used to smooth the curve, separating short-term temperature rise fluctuations from medium- and long-term temperature rise trends. Short-term fluctuations are mainly used to capture the thermal response caused by transient load disturbances, while medium- and long-term trends are used to assess the risk of sustained heat accumulation.

[0072] After trend separation, the medium- and long-term temperature rise trend is used as a baseline curve. Short-term fluctuations are then differentiated from the actual measured temperature series to generate a residual sequence, revealing the deviations from the baseline trend at each stage. Furthermore, the entire temperature rise trend is divided into multiple fitting stages based on this baseline curve. Within each stage, local polynomial fitting and rate of change analysis are performed on the residual sequence to extract residual variation characteristics that reflect abnormal temperature rise behavior, such as slope spikes and fluctuation expansion points.

[0073] After obtaining the residual variation characteristics, we identify load mutation boundaries and abnormal temperature rise segments based on preset mutation identification thresholds (such as residual sudden increase thresholds and fitting slope inflection points). These abnormal segments are marked as mutation focus segments. Understandably, these segments place higher timeliness and strength requirements on the cooling response and require special attention.

[0074] Subsequently, a deep sequence modeling structure is constructed using a gated recurrent unit (GRU) network within the temporal trend modeling unit to learn the temporal dependencies and nonlinear associations of the overall temperature rise trend. A multi-head residual attention mechanism is introduced into the feature sequence output by the GRU, constructing an attention weight distribution matrix guided by residual variation characteristics. This mechanism automatically enhances the ability to identify and model sudden attention segments, resulting in higher sensitivity and prediction accuracy for segments with dramatic temperature changes.

[0075] The model structure that integrates GRU sequence modeling and attention focusing mechanism can output multiple prediction results within the target time period, including the thermal load trend curve in the cabinet, the expected maximum temperature rise value, and the temperature rise rate change range, which serve as the basic input for the subsequent prediction control parameter generation.

[0076] By introducing a dual modeling strategy of residual decomposition and attention mechanism, this embodiment of the present invention effectively improves the adaptability of the cooling load prediction model in the face of sudden load changes, short-term disturbances, and nonlinear heat accumulation. This not only enhances the model's perception of microscopic thermal risks but also ensures hierarchical and refined control of subsequent cooling responses, thereby improving the energy efficiency and fault prevention capabilities of the entire electrical control cabinet cooling system.

[0077] As a preferred implementation of Example 1 of the present invention, a residual-driven multi-head attention mechanism is proposed to enhance the cooling load prediction model's ability to detect sudden temperature rise events and abnormal thermal fluctuations. This mechanism is embedded within a deep time series modeling framework built using a gated recurrent unit (GRU) network. This mechanism, guided by the changing characteristics of temperature rise residuals, adaptively adjusts the attention distribution weights, thereby enhancing the model's ability to model sudden temperature rise events and critical response time points.

[0078] Specifically, after the feature extraction and residual analysis stages, the time trend modeling unit obtains the baseline temperature rise trend curve and residual sequence. To construct the residual-driven attention weight, the residual value at each moment is first standardized within the sliding window to construct the following residual driving factor:

[0079]

[0080] in, is the residual driving factor at time t; is the residual between the actual temperature rise at time t and the fitted reference temperature rise; is the mean of the residual sequence in the sliding window; is the residual standard deviation adjustment coefficient, which is used to control the sensitivity of the residual driving factor.

[0081] The residual driving factor is used to reflect the relative abnormality of the residual at each moment. The closer it is to a local mutation, the higher the weight. Subsequently, this factor is combined with the GRU network output sequence to perform a weighted correction on the original attention mechanism to form a residual-driven attention distribution:

[0082]

[0083] in, is the attention distribution weight at the tth moment; is the query vector matrix, the parameter matrix involved in the attention score calculation; is the hidden state vector output by the GRU network at the current moment t; is one of the set of hidden state vectors for all relevant time steps; is the total number of time steps, i.e. the sequence length; ( ) is the transpose operation.

[0084] Through this attention distribution mechanism, the model automatically focuses on moments with high residual drive values ​​within the multi-head attention layer, thereby enhancing the model's ability to identify critical periods of "rapid heat accumulation" or "warning cooling failure." Furthermore, this residual-driven attention mechanism not only retains the temporal modeling capabilities of the gated structure but also utilizes small but significant thermal fluctuations as the core basis for attention scheduling, enhancing the sensitivity and stability of cooling load prediction in complex temperature rise scenarios.

[0085] Furthermore, the prediction output unit is used to calculate the prediction control parameters based on the model prediction results and the preset control parameter mapping rules, and transmit the prediction control parameters to the control module.

[0086] Furthermore, the control parameter mapping rules specifically include:

[0087] Generate the expected cooling power value based on the joint judgment rule mapping of the temperature rise rate change range and the current heat load slope in the cabinet;

[0088] The cooling component preload response time is calculated based on the advance warning time of the maximum temperature rise expected value relative to the temperature over-limit threshold and the thermal inertia response parameters of the cabinet components;

[0089] Based on the slope change section of the heat load trend curve within the target time period and the short-term fluctuation frequency matching graded wind speed adjustment curve, a wind speed graded scheduling plan is formed.

[0090] Specifically, in an embodiment of the present invention, the purpose of providing a prediction output unit is to dynamically calculate and generate prediction control parameters that are compatible with the cooling control task based on the output results of the cooling load prediction model and preset control parameter mapping rules, and transmit them to the control module. The main output results of the cooling load prediction model include the thermal load trend curve within the target time period, the expected maximum temperature rise value, the temperature rise rate change range, etc. The prediction output unit completes the control quantity conversion under multiple parameter dimensions by performing feature extraction, rule matching, and threshold calculation operations on the model output results, thereby achieving timely response to cooling demand in future time periods and strategic linkage.

[0091] Specifically, the control parameter mapping rules include the following three core mapping mechanisms:

[0092] Generation mechanism of expected cooling power value:

[0093] The prediction output unit generates the expected cooling power value based on a combined judgment rule based on the "temperature rise rate change range" and the "current cabinet heat load slope." The temperature rise rate change range represents the upper and lower limits of the cabinet temperature change rate within the prediction time window, reflecting the trend of heat accumulation in the short term; the heat load slope quantifies the rate of change of the current heat input. By cross-matching the two, scenarios with the potential for high temperature rise can be identified. Based on this, the expected cooling power value is output based on a preset interpolation mapping table. This operation avoids response delays caused by temperature lags and improves control efficiency.

[0094] Calculation mechanism for cooling component preload response time:

[0095] The prediction output unit determines the corresponding over-limit lead time based on the difference between the "maximum expected temperature rise" and the "temperature over-limit threshold." This calculation, combined with the cabinet components' thermal inertia response parameters (such as heat capacity and thermal conduction delay), calculates the required preheating time for the cooling components. This calculation ensures that the cooling equipment activates promptly before the heat load enters the high-temperature range, overcoming the delayed cooling response and inability to cover the early stages of a heat surge in traditional solutions.

[0096] Generation mechanism of wind speed classification scheduling scheme:

[0097] For cooling fan control, the predictive output unit matches a set of multi-level wind speed setting curves based on the slope variation of the thermal load trend curve (indicating the speed of load increase or decrease) and the frequency component of short-term thermal load fluctuations. Based on the load intensity and frequency of fluctuations during the predicted period, the appropriate wind speed level is selected. The overall air volume distribution and noise control strategy are optimized by adjusting the timing and span of wind speed switching.

[0098] The synergistic effect of these three mapping mechanisms eliminates the need for static threshold judgments in the cooling control system. Instead, it dynamically adjusts based on trends in the model output, enhancing the intelligence of the control device. The predictive output unit design allows for rapid adaptation to temperature fluctuations under varying load scenarios, achieving more energy-efficient and efficient cooling control.

[0099] Furthermore, the control module includes a temperature rise dynamic evaluation unit and a cooling strategy control unit; wherein,

[0100] The temperature rise dynamic assessment unit is used to determine the current cabinet temperature change rate and power input trend based on operating data, and to assess the current temperature rise risk level and risk level change trend by combining the maximum temperature rise expected value and temperature rise rate change range in the predictive control parameters and the preset cabinet enclosure level parameters;

[0101] The cooling strategy control unit is used to determine the use strategy of the prediction control parameters according to the temperature rise risk level and the risk level change trend, and generate corresponding cooling control instructions based on the use strategy of the prediction control parameters.

[0102] Furthermore, the cooling strategy control unit also sets an advance cooling strategy. When it is identified that the temperature rise risk level continues to rise and it is determined according to the model prediction results that the high heat load stage is about to arrive, the pre-cooling start operation is preferentially performed based on the advance cooling strategy.

[0103] Specifically, in this embodiment of the present invention, the control module includes a dynamic temperature rise assessment unit and a cooling strategy control unit, which together form the intelligent core module for cooling scheduling and risk management in electrical control cabinets. This module dynamically senses temperature rise trends, assesses thermal risk levels, and adaptively adjusts the predictive control parameter invocation strategy accordingly, achieving flexible cooling control tailored to dynamic load evolution.

[0104] The temperature rate of change is calculated by taking the temperature difference between consecutive time samples to measure the current rate of heat accumulation. The power input trend is calculated based on the slope of the historical power time series to predict future short-term power consumption changes. The cabinet enclosure level is a factory-configured parameter that reflects the control cabinet's airtightness and heat dissipation capabilities. The temperature rise dynamic assessment unit integrates these factors and uses a multi-rule fusion algorithm to generate the current temperature rise risk level (e.g., low, medium, high, or very high) and its trend (e.g., increasing, decreasing, or stable), providing status reference for downstream control strategies.

[0105] The cooling strategy control unit is used to dynamically determine the use strategy of the predictive control parameters based on the above evaluation results and generate the final cooling control instructions accordingly. Its strategy optimization logic mainly includes:

[0106] When the temperature rise risk level is high and / or shows an upward trend, the cooling strategy control unit will increase the expected cooling power, for example, by multiplying it by a dynamic risk coefficient for enhanced compensation, to quickly respond to potential thermal shocks;

[0107] Regarding the preload response time of the cooling component, if the predicted temperature rise exceeds the limit time point in advance, the cooling strategy control unit will prioritize triggering the cooling equipment preheating to ensure that it is effectively started before the high heat load stage;

[0108] The application strategy of the wind speed graded scheduling scheme is also adjusted synchronously to optimize the frequency and duration of wind speed gear calls according to trend changes, improve energy efficiency and control side effects such as vibration or noise.

[0109] If the risk level continues to rise and the high heat load stage is determined to be approaching based on the model prediction results, the cooling strategy control unit can trigger the advance cooling strategy. Even if the current temperature has not exceeded the default threshold, it can actively enter the pre-cooling state based on the model prediction results to provide redundant cooling capacity.

[0110] Based on the above regulation, the generated cooling control instructions include but are not limited to cooling component start and stop control instructions, cooling power control instructions, wind speed gear and speed regulation instructions, preload (preheating) response control instructions, minimum cooling duration locking instructions and dead zone adjustment instructions.

[0111] Through the above design, the control module in the present invention not only generates cooling instructions based on static data, but also can perform predictive control and strategy adaptive optimization according to the dynamic temperature rise trend, thereby significantly improving the response agility, energy efficiency utilization and operation safety of the control cabinet in a nonlinear thermal load environment.

[0112] Example 2

[0113] The second aspect of the present invention discloses a cooling control method for an electrical control cabinet. Figure 2 , Figure 2 1 is a flow chart of a method for controlling the temperature drop of an electrical control cabinet according to another embodiment of the present invention. The method includes:

[0114] Collect relevant operating data of the electrical control cabinet, including cabinet temperature data, current and voltage load data, cabinet component operating status parameters, and ambient temperature and humidity information;

[0115] A cooling load prediction model is trained based on historical operating data. The model then uses the current operating data to predict the thermal load trend, expected maximum temperature rise, and temperature rise rate variation range as the model prediction results. Predictive control parameters are calculated based on the model prediction results. These parameters include the expected cooling power, cooling component preload response time, and wind speed grading scheduling scheme.

[0116] Generate cooling control instructions based on predictive control parameters and operating data;

[0117] Execute corresponding cooling operations according to the cooling control instruction.

[0118] Furthermore, the construction and training process of the cooling load prediction model includes feature extraction operations, time trend modeling operations, and prediction output operations.

[0119] Feature extraction involves preprocessing and time-series analysis based on operational data to identify multidimensional key factors. These factors include the temperature difference between the cabinet interior and exterior, load fluctuation intensity, current fluctuation cycle, cabinet component start / stop status, and the distribution of localized high-heat areas.

[0120] Extract time series features that are sensitive to power surges from the operating data. These features include short-term slope changes in current and / or voltage, acceleration factors of load power, and abnormal temperature rise rates of heat-sensitive components.

[0121] After extracting the time series features, the fluctuation concentration of the time series features is analyzed, and the time period with abnormal concentration of features is regarded as the high fever risk triggering area.

[0122] The time trend modeling operation includes taking the extracted multidimensional key factors, time series features and high-heat risk trigger areas as input features, and building a cooling load prediction model based on the fusion of gated recurrent unit network and residual attention mechanism.

[0123] Furthermore, the cooling load prediction model is constructed based on the fusion of the gated recurrent unit network and the residual attention mechanism, specifically including:

[0124] Based on the input characteristics of the time trend modeling unit, a nonlinear temperature rise trend curve within a preset length sliding time window is constructed, and the exponentially weighted moving average method is used to separate short-term temperature rise fluctuations and medium- and long-term temperature rise trends;

[0125] The separated medium- and long-term temperature rise trend is used as the benchmark curve, and the residual sequence of the actual temperature rise sequence is constructed in combination with the short-term temperature rise fluctuation. The temperature rise trend is divided into multiple fitting stages based on the benchmark curve. Local fitting and residual analysis are performed on each stage based on the residual sequence to extract the residual change characteristics.

[0126] According to the residual variation characteristics and the preset mutation identification threshold, the load mutation boundary and abnormal temperature rise segment are identified to form the mutation focus section;

[0127] A sequence modeling structure is constructed based on the gated recurrent unit network. The residual change features are integrated into the sequence modeling structure to form a multi-head attention weight distribution, and the mutation attention segment is focused modeled to finally obtain a cooling load prediction model.

[0128] Furthermore, the prediction output operation includes calculating the predicted control parameters based on the model prediction results and preset control parameter mapping rules.

[0129] The control parameter mapping rules specifically include:

[0130] Generate the expected cooling power value based on the joint judgment rule mapping of the temperature rise rate change range and the current heat load slope in the cabinet;

[0131] The cooling component preload response time is calculated based on the advance warning time of the maximum temperature rise expected value relative to the temperature over-limit threshold and the thermal inertia response parameters of the cabinet components;

[0132] Based on the slope change section of the heat load trend curve within the target time period and the short-term fluctuation frequency matching graded wind speed adjustment curve, a wind speed graded scheduling plan is formed.

[0133] Furthermore, generating the cooling control instruction according to the predicted control parameters and the operating data specifically includes:

[0134] Determine the current cabinet temperature change rate and power input trend based on operating data, and evaluate the current temperature rise risk level and risk level change trend by combining the maximum expected temperature rise value and temperature rise rate change range in the predictive control parameters and the preset cabinet enclosure level parameters;

[0135] The use strategy of the predictive control parameters is determined according to the temperature rise risk level and the risk level change trend, and the corresponding cooling control instructions are generated based on the use strategy of the predictive control parameters.

[0136] Furthermore, the method also includes setting an advance cooling strategy. When it is identified that the temperature rise risk level continues to rise and it is determined based on the model prediction result that a high heat load stage is about to arrive, a pre-cooling start operation is preferentially performed based on the advance cooling strategy.

[0137] It should be noted that the specific implementation process of Example 2 is similar to that of Example 1 and will not be repeated in this embodiment.

[0138] Finally, it should be noted that the cooling device for an electrical control cabinet and the control method thereof disclosed in the embodiment of the present invention are only preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, rather than to limit them. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A cooling device for an electrical control cabinet, characterized in that: The device includes a data acquisition module, a cooling load prediction module, a control module and an execution module; wherein, The data acquisition module is used to collect relevant operating data of the electrical control cabinet, including cabinet temperature data, current and voltage load data, cabinet component operating status parameters and ambient temperature and humidity information; The cooling load prediction module is used to train a cooling load prediction model based on historical operating data. The cooling load prediction model predicts the thermal load trend, expected maximum temperature rise, and temperature rise rate change range based on current operating data as the model prediction results. The prediction control parameters are calculated based on the model prediction results. The prediction control parameters include the expected cooling power, the cooling component preload response time, and the wind speed classification scheduling scheme. The control module is used to generate cooling control instructions based on the predicted control parameters and operating data; The execution module is used to receive the cooling control instruction and execute a corresponding cooling operation according to the cooling control instruction.

2. The cooling device for an electrical control cabinet according to claim 1, characterized in that: The cooling load prediction module includes a feature extraction unit, a time trend modeling unit and a prediction output unit; wherein, The feature extraction unit is used to perform data preprocessing and timing analysis operations based on the operating data to obtain multidimensional key factors; the multidimensional key factors include the temperature difference between the inside and outside of the cabinet, the intensity of the operating load fluctuation, the current fluctuation cycle, the start and stop status of the cabinet components, and the distribution of local high-heat areas.

3. The cooling device for an electrical control cabinet according to claim 2, characterized in that: The feature extraction unit is further configured to extract time series features that are sensitive to power mutations from the operating data; the time series features include short-term slope changes of current and / or voltage, acceleration factors of load power, and abnormal temperature rise rates of heat-sensitive components; After extracting the time series features, the fluctuation concentration of the time series features is analyzed, and the time period with abnormal concentration of features is regarded as the high fever risk triggering area.

4. The cooling device for an electrical control cabinet according to claim 3, characterized in that: The time trend modeling unit is used to take the extracted multidimensional key factors, time series features and high heat risk trigger areas as input features, and build a cooling load prediction model based on the fusion of the gated recurrent unit network and the residual attention mechanism.

5. The cooling device for an electrical control cabinet according to claim 4, characterized in that: The cooling load prediction model constructed based on the fusion of the gated recurrent unit network and the residual attention mechanism specifically includes: Based on the input characteristics of the time trend modeling unit, a nonlinear temperature rise trend curve within a preset length sliding time window is constructed, and the exponentially weighted moving average method is used to separate short-term temperature rise fluctuations and medium- and long-term temperature rise trends; The separated medium- and long-term temperature rise trend is used as the benchmark curve, and the residual sequence of the actual temperature rise sequence is constructed in combination with the short-term temperature rise fluctuation. The temperature rise trend is divided into multiple fitting stages based on the benchmark curve. Local fitting and residual analysis are performed on each stage based on the residual sequence to extract the residual change characteristics. According to the residual variation characteristics and the preset mutation identification threshold, the load mutation boundary and abnormal temperature rise segment are identified to form the mutation focus section; A sequence modeling structure is constructed based on the gated recurrent unit network. The residual change features are integrated into the sequence modeling structure to form a multi-head attention weight distribution, and the mutation attention segment is focused modeled to finally obtain a cooling load prediction model.

6. The cooling device for an electrical control cabinet according to claim 2, characterized in that: The prediction output unit is used to calculate the prediction control parameters based on the model prediction results and the preset control parameter mapping rules, and transmit the prediction control parameters to the control module.

7. The cooling device for an electrical control cabinet according to claim 6, characterized in that: The control parameter mapping rules specifically include: Generate the expected cooling power value based on the joint judgment rule mapping of the temperature rise rate change range and the current heat load slope in the cabinet; The cooling component preload response time is calculated based on the advance warning time of the maximum temperature rise expected value relative to the temperature over-limit threshold and the thermal inertia response parameters of the cabinet components; Based on the slope change section of the heat load trend curve within the target time period and the short-term fluctuation frequency matching graded wind speed adjustment curve, a wind speed graded scheduling plan is formed.

8. The cooling device for an electrical control cabinet according to any one of claims 1 to 7, characterized in that: The control module includes a temperature rise dynamic evaluation unit and a cooling strategy control unit; wherein, The temperature rise dynamic assessment unit is used to determine the current cabinet temperature change rate and power input trend based on operating data, and to assess the current temperature rise risk level and risk level change trend by combining the maximum temperature rise expected value and temperature rise rate change range in the predictive control parameters and the preset cabinet enclosure level parameters; The cooling strategy control unit is used to determine the use strategy of the prediction control parameters according to the temperature rise risk level and the risk level change trend, and generate corresponding cooling control instructions based on the use strategy of the prediction control parameters.

9. The cooling device for an electrical control cabinet according to claim 8, characterized in that: The cooling strategy control unit further sets an advance cooling strategy. When it is identified that the temperature rise risk level continues to rise and it is determined according to the model prediction result that the high heat load stage is about to arrive, the pre-cooling start operation is preferentially performed based on the advance cooling strategy.

10. A cooling control method for an electrical control cabinet, the cooling control method being applied to the cooling control system according to any one of claims 1 to 9, characterized in that: The method comprises: Collect relevant operating data of the electrical control cabinet, including cabinet temperature data, current and voltage load data, cabinet component operating status parameters, and ambient temperature and humidity information; A cooling load prediction model is trained based on historical operating data. The model then uses the current operating data to predict the thermal load trend, expected maximum temperature rise, and temperature rise rate variation range as the model prediction results. Predictive control parameters are calculated based on the model prediction results. These parameters include the expected cooling power, cooling component preload response time, and wind speed grading scheduling scheme. Generate cooling control instructions based on predictive control parameters and operating data; Execute corresponding cooling operations according to the cooling control instruction.

Images