Cabinet temperature control method and device of distributed control system

Abstract

The invention discloses a cabinet temperature control method and device for a distributed control system, and the method comprises the steps: reconstructing a three-dimensional temperature distribution model in a cabinet through the combination of real-time temperature data and a spatial interpolation technology, so as to obtain the current temperature distribution data; and based on a neural network, in combination with the historical thermal load data, the operation state data of the system, the external environment data and the temperature distribution data, determining thermal load prediction data and temperature distribution prediction data in a future preset time period. And based on the real-time temperature data, the thermal load prediction data and the temperature distribution prediction data, a multi-objective optimization algorithm is adopted to optimize the control instruction of the temperature adjustment equipment in each partition in the cabinet. According to the optimization result of the control instruction, the temperature adjusting device of each partition in the cabinet is controlled, partition temperature control of the cabinet is realized, the thermal load and the temperature are predicted by using a spatial interpolation technology in combination with a neural network, partition control instruction optimization is carried out on the cabinet, and the precision and the response speed of the temperature control method are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of temperature control, and particularly relates to a cabinet temperature control method and device of a distributed control system. BACKGROUND

[0002] With the continuous improvement of industrial automation, the distributed control system is increasingly widely applied in production, control and monitoring. The cabinet of the distributed control system is highly integrated and has a large heat generation, and temperature abnormalities thereof will directly affect the working performance and service life of the system.

[0003] In the prior art, the temperature in the cabinet is controlled based on sensor temperature measurement and combined with PID control. This control method cannot accurately obtain the complex three-dimensional temperature distribution in the cabinet, and cannot quickly respond to local hot spots and dynamic heat load changes, which results in slow response and low precision of temperature control, and it is difficult to meet the current production requirements. SUMMARY

[0004] The present application provides a cabinet temperature control method and device of a distributed control system to improve the precision and response speed of the temperature control method.

[0005] According to an aspect of the present application, a cabinet temperature control method of a distributed control system is provided, which comprises the following steps.

[0006] The real-time temperature data in the cabinet is used in combination with a spatial interpolation technique to reconstruct a three-dimensional temperature distribution model in the cabinet to obtain temperature distribution data in the current cabinet.

[0007] Based on a neural network, historical heat load data, system operating state data, external environment data and the temperature distribution data are combined to determine heat load prediction data and temperature distribution prediction data in a future preset time period.

[0008] Based on the real-time temperature data, the heat load prediction data and the temperature distribution prediction data, a multi-objective optimization algorithm is used to optimize the control instructions of the temperature control equipment in each partition of the cabinet.

[0009] According to the optimization results of the control instructions, the temperature control equipment in each partition of the cabinet is controlled.

[0010] Optionally, the neural network comprises a recurrent neural network and a deep neural network.

[0011] The neural network is used in combination with historical heat load data, system operating state data, external environment data and the temperature distribution data to determine the heat load prediction data and the temperature distribution prediction data in the future preset time period, which comprises the following steps.

[0012] determining, based on the recurrent neural network, the thermal load prediction data in a future preset time period in combination with the historical thermal load data, the operating state data of the system and the external environment data;

[0013] determining, based on the deep neural network, the temperature distribution prediction data in the future preset time period in combination with the thermal load prediction data and the temperature distribution data in the current cabinet.

[0014] Optionally, the recurrent neural network comprises a long short-term memory network model, which is trained offline by training data, wherein the training data comprises historical thermal load data, historical external environment data and historical temperature distribution data, and the historical temperature distribution data is determined by temperature field reconstruction using the three-dimensional temperature distribution model and historical temperature data.

[0015] Optionally, the spatial interpolation technique comprises radial basis function interpolation.

[0016] reconstructing a three-dimensional temperature distribution model in the cabinet by using real-time temperature data in the cabinet to obtain temperature distribution data in the current cabinet, comprises:

[0017] using a radial basis function to represent a spatial correlation function φ (‖P-P i ‖) of the sampling data of each temperature sensor to the target point temperature in the cabinet, wherein ‖P-P i ‖ is the Euclidean distance between the target point P and the i-th temperature sensor, and φ () is the radial basis function.

[0018] determining a correlation accumulation function F (x, y, z) based on each spatial correlation function and the corresponding weight coefficient wherein w i is the weight coefficient associated with the target point temperature of the i-th temperature sensor, and N is the total number of temperature sensors.

[0019] representing a correlation function F (x, y, z) of the target point temperature in the cabinet and its three-dimensional position coordinates in combination with the correlation accumulation function and a linear trend item as the three-dimensional temperature distribution model, wherein x, y and z are coordinate values on each coordinate axis in a three-dimensional coordinate system, and β0, β1, β2 and β3 are coefficients of the linear trend item.

[0020] inputting the real-time temperature data into the correlation function to reconstruct temperature data of other position points in the cabinet to obtain the temperature distribution data in the current cabinet.

[0021] Optionally, the system functions of each of the sub-zones of the cabinet are different, and at least one set of the temperature control devices and a plurality of temperature sensors are arranged in a single sub-zone.

[0022] Optionally, the multi-objective optimization algorithm comprises a multi-objective model predictive control algorithm, and the temperature control devices comprise fans.

[0023] The multi-objective optimization algorithm is used to optimize the control instructions of the temperature control devices in each sub-zone of the cabinet based on the real-time temperature data, the thermal load prediction data, and the temperature distribution prediction data, comprising:

[0024] Minimizing the temperature deviation, minimizing the total energy consumption, and minimizing the noise are optimization objectives, and a target function of the multi-objective optimization algorithm is determined. V j is the optimal rotation speed of the jth fan, J is a cost function, T i is the real-time temperature of the ith sub-zone, T target,i is the target temperature of the ith sub-zone, P fan,j (V j ) is the operating power of the jth fan, Noise fan,k (V k ) is the noise level of the kth fan, and a1, a2, and a3 are weight coefficients.

[0025] The optimal rotation speed of the fan is optimized by using the multi-objective model predictive control algorithm based on the target function, the constraint condition, the real-time temperature data, the thermal load prediction data, and the temperature distribution prediction data.

[0026] According to the optimal rotation speed obtained by optimization, a control signal for each fan is determined.

[0027] Optionally, the constraint condition comprises a rotation speed constraint condition of the fan, a temperature constraint condition of each sub-zone, a total power consumption constraint condition, and a coordination constraint condition, wherein the coordination constraint condition comprises a temperature difference constraint between adjacent sub-zones, a rotation speed relative relationship constraint, and a heat transfer balance constraint.

[0028] Optionally, after the three-dimensional temperature distribution model in the cabinet is reconstructed by using the real-time temperature data in the cabinet in combination with a spatial interpolation technique to obtain the temperature distribution data in the current cabinet, the method further comprises:

[0029] According to the temperature distribution data in the current cabinet, real-time thermal risk judgment is performed.

[0030] According to the result of the real-time thermal risk judgment, audible and visual alarm and remote communication alarm are performed.

[0031] In the neural network-based method, after the thermal load prediction data and the temperature distribution prediction data in the future preset period are determined by combining historical thermal load data, system running state data, external environment data, and the temperature distribution data, the method further comprises:

[0032] According to the temperature distribution prediction data, a thermal risk pre-judgment is performed.

[0033] According to the result of the thermal risk pre-judgment, the audible and light alarm and the remote communication alarm are performed.

[0034] According to another aspect of the present application, a cabinet temperature control device of a distributed control system is provided, which comprises a plurality of temperature sensors, a plurality of temperature adjusting devices, a plurality of environment sensors, and a microcontroller, wherein the microcontroller is connected with the temperature sensors, the temperature adjusting devices, and the environment sensors, respectively, and is configured to implement the cabinet temperature control method of the distributed control system according to any one of claims 1-8.

[0035] Optionally, the cabinet temperature control device of the distributed control system further comprises an audible and light alarm module and a communication module, wherein the microcontroller is connected with the audible and light alarm module and the communication module, respectively; the audible and light alarm module is configured to perform audible and light alarm according to the result of real-time thermal risk judgment and the result of thermal risk pre-judgment; and the communication module is configured to perform communication alarm according to the result of thermal risk judgment and the result of thermal risk pre-judgment.

[0036] The cabinet temperature control method and device of the distributed control system according to the embodiments of the present application reconstruct a three-dimensional temperature distribution model in the cabinet by using real-time temperature data in combination with a spatial interpolation technique to obtain current temperature distribution data. Based on a neural network, thermal load prediction data and temperature distribution prediction data in a future preset period are determined by combining historical thermal load data, system running state data, external environment data, and temperature distribution data. Based on real-time temperature data, thermal load prediction data, and temperature distribution prediction data, a multi-objective optimization algorithm is used to optimize control instructions of the temperature adjusting devices in each partition in the cabinet. According to the optimization result of the control instructions, the temperature adjusting devices in each partition in the cabinet are controlled to achieve partition temperature control of the cabinet. The spatial interpolation technique in combination with the neural network is used to predict thermal load and temperature, and the control instructions of the cabinet are optimized in a partition manner, thereby improving the accuracy and response speed of the temperature control method.

[0037] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present application, nor is it used to limit the scope of the present application. Other features of the present application will become apparent through the following description. BRIEF DESCRIPTION OF DRAWINGS

[0038] In order to make the technical solutions in the embodiments of the present application clearer, the accompanying drawings needed in the embodiment description will be briefly introduced. Obviously, the accompanying drawings in the following description only show some embodiments of the present application, and all other embodiments obtained by those of ordinary skill in the art without any creative effort based on the accompanying drawings should belong to the protection scope of the present application.

[0039] Figure 1 A flowchart of a cabinet temperature control method of a distributed control system according to an embodiment of the present application is shown.

[0040] Figure 2 A zoning diagram of a cabinet according to an embodiment of the present application is shown.

[0041] Figure 3 A flowchart of another cabinet temperature control method of a distributed control system according to an embodiment of the present application is shown.

[0042] Figure 4 A flowchart of still another cabinet temperature control method of a distributed control system according to an embodiment of the present application is shown.

[0043] Figure 5 A composition diagram of a cabinet temperature control device of a distributed control system according to an embodiment of the present application is shown.

[0044] Figure 6 A composition diagram of another cabinet temperature control device of a distributed control system according to an embodiment of the present application is shown.

[0045] Figure 7 A composition diagram of still another cabinet temperature control device of a distributed control system according to an embodiment of the present application is shown.

[0046] Figure 8 A structural diagram of an electronic device that can be used to implement an embodiment of the present application is shown. DETAILED DESCRIPTION

[0047] In order to make the technical solutions in the embodiments of the present application clearer, the accompanying drawings needed in the embodiment description will be briefly introduced. Obviously, the accompanying drawings in the following description only show some embodiments of the present application, and all other embodiments obtained by those of ordinary skill in the art without any creative effort based on the accompanying drawings should belong to the protection scope of the present application.

[0048] It should be noted that the terms "first", "second", and the like in the description and in the claims of the present application and the above-described drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0049] To solve the problems in the background art, the present application proposes a cabinet temperature control method of a distributed control system. Figure 1 A flowchart of a cabinet temperature control method of a distributed control system is proposed for embodiments of the present application, referring to Figure 1 The cabinet temperature control method of the distributed control system comprises:

[0050] S101, using real-time temperature data in the cabinet and combining spatial interpolation technology, reconstructing a three-dimensional temperature distribution model in the cabinet to obtain temperature distribution data in the current cabinet.

[0051] Specifically, the real-time temperature data in the cabinet includes temperature values and corresponding position information, which can be obtained from temperature sensors distributed at various positions in the cabinet as input data for the spatial interpolation technology. The three-dimensional temperature distribution model refers to a mathematical model representing the temperature at any position in the cabinet, which can be constructed using real-time temperature data combined with spatial interpolation technology, wherein the spatial interpolation technology can include Kriging interpolation or radial basis function interpolation. The temperature distribution data includes temperature data of multiple position points distributed in the cabinet, which includes both measured values collected by temperature sensors and interpolation data obtained by spatial interpolation technology. After reconstructing the three-dimensional temperature distribution model, the coordinates of any position in the cabinet can be input to obtain the temperature data of that point, achieving the obtaining of temperature distribution data. The user can determine the number of interpolation data in the temperature distribution data according to actual needs.

[0052] S102, based on a neural network, combining historical heat load data, system operating state data, external environment data and temperature distribution data, determining heat load prediction data and temperature distribution prediction data in a future preset period.

[0053] Specifically, the neural network is trained offline by combining historical data with corresponding interpolation data, wherein the historical data can include historical heat load data, historical external environment data and historical temperature sensor data. The heat load data refers to the power consumption data of the key heat generating devices in the distributed control system, which plays a major driving role in the temperature change of the cabinet. The heat load data can include the power consumption data of CPU, communication equipment and power supply. The external environment data refers to the environmental state parameters of the space outside the cabinet. The external environment data can include environmental temperature, environmental humidity and environmental air pressure, and other environmental influence parameters outside the cabinet. Exemplarily, the neural network includes a recurrent neural network and a deep neural network. The historical heat load data, the running state data of the system and the external environment data are substituted into the corresponding recurrent neural network model to predict the heat load prediction data in the future preset time period. Then, the heat load prediction data is substituted into the corresponding deep neural network model to predict the temperature distribution prediction data in the future preset time period.

[0054] S103, based on the real-time temperature data, the heat load prediction data and the temperature distribution prediction data, a multi-objective optimization algorithm is used to optimize the control instructions of the temperature adjusting equipment in each partition of the cabinet.

[0055] Specifically, Figure 2 A partitioning schematic diagram of a cabinet is provided for the embodiments of the present application, which combines Figure 1 and Figure 2 The partitions in the cabinet can be determined according to at least one of the function distribution, heat source distribution and cooling demand of the distributed control system. Exemplarily, the partitions in the cabinet can include a CPU area, an input / output interface area, a power supply area and an environment area. Each partition is provided with at least one set of temperature adjusting equipment and at least one set of temperature sensors, which can avoid the extensive and inefficient control mode caused by global unified control, and realize independent and coordinated detection and control of the temperature in the cabinet. The temperature adjusting equipment refers to the equipment for adjusting the temperature in each partition. Exemplarily, the CPU area can be deployed with 4 to 6 temperature sensors, the input / output interface area can be deployed with 2 to 4 temperature sensors, the power supply area can be deployed with 2 temperature sensors, and the environment area can be deployed with 2 temperature sensors, totaling 10 to 20 temperature sensors. The temperature adjusting equipment can include at least one of an air conditioner, a fan or an air duct.

[0056] Multi-objective optimization algorithms are algorithms used to solve optimization problems with multiple conflicting objectives. For example, multi-objective optimization algorithms may include non-dominated sorting genetic algorithms and multi-objective particle swarm optimization algorithms. A multi-objective optimization algorithm aims to minimize temperature deviation, total energy consumption, and noise by optimizing the control commands of various temperature-regulating devices. For example, the optimization variable in a multi-objective optimization algorithm can be the target speed of a fan. After determining the target speed, the corresponding control command is determined based on the difference between the target speed and the current speed. For instance, if the target speed is less than the current speed, the duty cycle of the fan control command can be increased accordingly based on the speed difference.

[0057] S104. Based on the optimization results of the control commands, control the temperature control equipment in each zone of the cabinet.

[0058] In the cabinet temperature control method of the distributed control system provided in this embodiment, real-time temperature data within the cabinet is combined with spatial interpolation technology to reconstruct a three-dimensional temperature distribution model within the cabinet, thereby obtaining the current temperature distribution data within the cabinet. Based on a neural network, and combining historical heat load data, system operating status data, external environmental data, and temperature distribution data, predicted heat load data and predicted temperature distribution data for a preset future time period are determined. Based on the real-time temperature data, predicted heat load data, and predicted temperature distribution data, a multi-objective optimization algorithm is used to optimize the control commands for temperature-regulating devices in each zone of the cabinet. According to the optimization results of the control commands, the temperature-regulating devices in each zone of the cabinet are controlled, realizing zoned temperature control of the cabinet. By using spatial interpolation technology combined with a neural network to predict heat load and temperature, and then optimizing the control commands for each zone of the cabinet, the accuracy and response speed of the temperature control method are greatly improved.

[0059] Figure 3 This is a flowchart illustrating another cabinet temperature control method for a distributed control system provided in an embodiment of the present invention. Based on the foregoing embodiments, refer to... Figure 3 Cabinet temperature control methods for distributed control systems include:

[0060] S301. Using radial basis functions, express the spatial correlation function φ(‖PP) of the sampling data of each temperature sensor with the temperature of the target point inside the cabinet. i ||).

[0061] Specifically, ‖PP i ‖ represents the Euclidean distance between the target point P and the i-th temperature sensor. This distance quantifies the spatial proximity of the two points and is fundamental for evaluating the correlation (or influence) of the temperature sensor's sampling data on the temperature of surrounding points. φ() is the radial basis function, serving as the spatial correlation function in the three-dimensional temperature distribution model. φ(‖PP) i‖) for measuring the local influence or spatial correlation of the sampling data of the ith temperature sensor to the temperature of the target point P. The function takes the distance ‖P-P i ‖ as the independent variable, and is usually designed to decay with the increase of the distance, exemplarily, the radial basis function adopts a quadratic function , where r is the distance between two points; c is a smoothing parameter for adjusting the sensitivity of the function response to the distance, and the setting of this spatial correlation function can ensure that the closer the sensor to the target point, the greater the contribution of the sampling data of the sensor to the temperature data of the target point.

[0062] S302, based on each spatial correlation function and the corresponding weight coefficient, determining a correlation accumulation function

[0063] Specifically, w i is the weight coefficient associated with the temperature of the target point by the ith temperature sensor, and the value of w i reflects the relative contribution intensity and direction of each temperature sensor in the global temperature field reconstruction. These coefficients are not preset constants, but are obtained by optimization and solution through least squares method combined with regularization term (such as ridge regression or Lasso regression). The optimization goal of the weight coefficient is to make the weighted combination of all temperature sensors in the cabinet best fit the known measured temperature values of the temperature sensors, and at the same time enhance the stability and generalization ability of the model. N is the total number of temperature sensors.

[0064] represents the summation of the weighted influence of all N sensors in the cabinet. Through this accumulation, the algorithm fuses the sampling data from all known temperature sensors and comprehensively evaluates the contribution of the sampling data of these temperature sensors to the temperature of the target point.

[0065] The correlation accumulation function This item technically realizes the local influence weighted superposition based on spatial correlation. It quantifies the physical influence of each temperature sensor in the cabinet on the temperature of the unknown point through the radial basis function, and combines the weight coefficient obtained by optimization to accurately fuse the sampling data of discrete temperature sensors, so as to accurately infer the continuous and smooth temperature distribution in the three-dimensional space of the cabinet from the limited temperature sensor data. This is a key technical link to realize local hot spot identification and overall temperature field monitoring.

[0066] S303, combining the correlation accumulation function and the linear trend item, representing the correlation function of the temperature of the target point in the cabinet and its three-dimensional position coordinates as a three-dimensional temperature distribution model.

[0067] Specifically, T(x, y, z) is the temperature of any unknown point P(x, y, z) in the cabinet, in units of ℃ or K, x, y, and z are the coordinate values on each coordinate axis in the three-dimensional coordinate system, and β0, β1, β2, and β3 are the coefficients of the linear trend term, which is used to capture the overall linear change trend of the temperature field in the cabinet. Through this step, the system can obtain more comprehensive temperature distribution information than discrete temperature sampling data points, and accurately identify local hot spots. The number N and position P of the sensors i The number N and position P of the sensors need to be optimized and deployed according to the cabinet size, heat density, and required reconstruction accuracy to ensure the accuracy and efficiency of the temperature field reconstruction.

[0068] S304, input the real-time temperature data into the correlation function to reconstruct the temperature data of other position points in the cabinet to obtain the temperature distribution data in the current cabinet.

[0069] Specifically, the real-time temperature data includes temperature values and corresponding position information, which can be obtained from temperature sensors distributed at various positions in the cabinet as input data for spatial interpolation techniques. The real-time temperature data is substituted into the correlation function determined by the interpolation technique to reconstruct the temperature data of other position points in the cabinet to obtain the temperature distribution data in the current cabinet. The other position points are randomly distributed interpolation points, and the number of interpolation points can be set according to the required reconstruction accuracy of the user, or can be adapted to the setting density of the temperature control equipment in the cabinet.

[0070] S305, based on the recurrent neural network, combine the historical heat load data, system running state data, and external environment data to determine the heat load prediction data in the future preset time period.

[0071] Specifically, the recurrent neural network can include a long short-term memory network (LSTM), and the long short-term memory network model can be represented as Q load (t+Δt)=LSTM(Q hist ,S system ,T ambient )。Q hist is the historical heat load data of the distributed control system, which includes the power consumption data of the distributed control system in the past period of time. For example, the historical heat load data can include the historical power consumption data of key heat generating devices such as CPU, communication equipment, and power supply. S system refers to system running state data, such as real-time sampling values of key performance indicators such as CPU utilization, memory usage, network traffic, and power output, which can directly reflect the current heat generation of the system. T ambientis referred to as the external environment data of the cabinet, for example, the trend of the external temperature, which has an impact on the overall thermal load in the cabinet. Although the internal temperature of the cabinet is mainly affected by the thermal load, the change of the external environment temperature also affects the heat exchange between the cabinet and the outside world, thereby affecting the overall drift of the internal temperature field and the heat dissipation efficiency. Including it in the prediction helps the long short-term memory network model to capture the characteristics of the thermal dynamics in the cabinet more comprehensively, and the prediction result will be an important input for the zoning control strategy to achieve predictive control.

[0072] The long short-term memory network model is trained offline by training data, wherein the training data includes historical thermal load data, historical external environment data and historical temperature distribution data, and the historical temperature distribution data is determined by reconstructing the temperature field using a three-dimensional temperature distribution model and historical temperature data. The long short-term memory network can also update the model parameters by combining online transfer learning or regular retraining mechanism to adapt to the changes in thermal characteristics of the cabinet in the long-term operation of the distributed control system. The input features of the long short-term memory network model include historical thermal load data Q hist , system state data S system and the trend of external environment data T ambient in the past preset time period. The long short-term memory network structure can contain multiple layers of long short-term memory network units and combine a fully connected layer to output thermal load prediction data Q load (t+Δt) in the future preset time period Δt. The selection of the future preset time period Δt needs to balance the prediction accuracy and system response time to determine, for example, the future preset time period Δt can be set to 5 to 15 minutes.

[0073] S306, based on the deep neural network, combining the thermal load prediction data and the temperature distribution data in the current cabinet, determining the temperature distribution prediction data in the future preset period.

[0074] Specifically, the core mathematical model of the embodiment is a dynamic temperature field prediction model, which is a deep neural network model, represented as T(x, y, z, t+Δt) = F(T hist , Q load , V fan , P ambient} + ∈(t), where T(x, y, z, t + Δt) is the predicted temperature distribution data of the cabinet within a preset time period, which can be represented by a three-dimensional temperature distribution model, which represents the temperature value at any spatial coordinate (x, y, z) inside the cabinet of the distributed control system at the current time in the future preset time period. It should be particularly noted that the conventional temperature control system usually only focuses on the temperature of a few temperature sensor points, while the model of the present application can predict and obtain the continuous three-dimensional temperature distribution of the entire cabinet interior, which is crucial for identifying and managing local hot spots, and also provides comprehensive temperature information for fine zoning control.

[0075] F is a nonlinear mapping function realized by a deep neural network, which is the core computing engine of the model, represents a complex function relationship, and can be realized by one or more deep neural networks. The deep neural network learns a large amount of historical temperature data, historical heat load data, historical operation data of temperature control equipment, and historical external environment data, etc. historical data, establishes a nonlinear mapping relationship between the input variables of the current reconstructed temperature distribution data, heat load prediction data, current temperature control equipment operation data and external environment data and the future temperature field distribution, which can capture the dynamic and nonlinear physical laws between heat conduction, heat convection, heat radiation, various equipment heating and cooling in the cabinet. For example, the deep neural network can be a multi-layer fully connected network, or for more complex spatio-temporal correlation, the deep neural network may, for example, use a convolutional neural network to process spatial features, and can also use a graph neural network to process the topological relationship between temperature sensors, thereby better modeling the changes in the temperature field. F can also use auxiliary features other than the main variables listed, such as time information, to enhance the function's ability to predict future temperatures.

[0076] T hist is a matrix composed of historical and current temperature distribution data in the cabinet, with dimensions n x m x k, n sensors, m time points, and k features, for example, features can include temperature values and change rates, and the temperature distribution data matrix can be used to learn the dynamic evolution law of the temperature field.

[0077] Q load is heat load prediction data, including real-time or predicted power consumption data such as CPU load, communication load, and power supply load, which is the main driving factor of temperature change. Specifically, Q load is a multi-dimensional vector, and its components represent the predicted values of each heat load, for example, Q load = [Q CPU , Q Comm , Q Power ] T , where Q CPU represents the heat generated by CPU power consumption or load, QComm Q represents the heat generated by the operation of the communication module. Power Q represents the heat generated by the operation of the power module.

[0078] V fan is the fan speed control data corresponding to the PWM control signal of each fan, which is the execution variable of the system's active temperature regulation.

[0079] P ambient is external environmental data, including environmental temperature, humidity, air pressure and other external influencing factors.

[0080] is the error compensation term of the model, which is updated and optimized in real time through online error feedback mechanisms such as Kalman filtering based on prediction residuals or adaptive learning rate adjustment, to improve the online adaptability and prediction accuracy of the model, and ensure the robustness of the model when the thermal characteristics of the cabinet change over a long period of operation.

[0081] S307, minimizing temperature deviation, minimizing total energy consumption and minimizing noise as optimization objectives, determining the objective function of the multi-objective optimization algorithm

[0082] Specifically, V j is the optimal speed of the jth fan. In practical applications, according to V j PWM control signal of the fan can be further determined, which directly determines the actual speed of the fan, the air volume generated thereby, the power consumption and the noise level. The optimization algorithm determines the optimal V j combination that minimizes the objective function J through iterative search method. j It is worth noting that V k and V j in the right side of the optimization objective function are the same concept, both representing the optimal fan speed that the algorithm is seeking.

[0083] J is a comprehensive, dimensionless cost function that integrates the three mutually restrictive performance indicators through weighted summation, and the unity of dimension is guaranteed by the appropriate dimension of the weight coefficient in front of each term.

[0084] α1∑ i (T i -T target,i ) 2T is the temperature deviation control term in the objective function, which is designed to minimize the deviation between the actual temperature of the core partition in the cabinet and its target temperature, so as to ensure that the core electronic components run in a safe, stable and high-performance temperature range. The core partition can be the most system safety related partitions in the cabinet, for example, including the CPU area, the input / output interface area, and the power supply area. i T is the real-time temperature of the i-th core partition. target,i T is the target temperature of the i-th partition, which can be determined according to at least one of the system manufacturer's recommendations, reliability standards and system performance requirements. i -T target,i ) 2 is a penalty function in the form of squared deviation, which produces positive punishment regardless of whether the temperature deviates from the target value or not, guiding the optimization direction. Compared with the general linear deviation, the squared deviation can amplify the large deviation, prompting the optimization algorithm to preferentially process the area where the temperature deviates seriously from the target value, improving the responsiveness and accuracy of the control. i represents the sum of the squared values of the temperature deviations of all monitored partitions, ensuring the temperature balance and compliance rate of the entire cabinet partition. α1 is the weight coefficient of the temperature deviation control term, also known as the priority adjustment coefficient of this term, and its value directly reflects the importance of the system to the temperature control accuracy. The larger the value of α1, the more the algorithm tends to sacrifice energy consumption or noise to strictly maintain the target temperature.

[0085] α2∑ j P fan,j (V j ) is the energy consumption control term in the objective function, which is designed to minimize the total energy consumption of all temperature control devices (mainly fans) in the cabinet, to achieve green operation of the cabinet and reduce operating costs. fan,j (V j ) is the running power of the j-th fan, which is a nonlinear function of its speed V j The power consumption of the fan and the speed are nonlinearly related (such as cubic relationship), and this energy consumption control term needs to accurately reflect this characteristic. j represents the sum of the power consumptions of all participating fans, representing the total energy consumption of all fans in the cabinet under the current running state. α2 is the weight coefficient of the energy consumption control term, also known as the priority adjustment coefficient of this term. The larger the value of α2, the more the algorithm prioritizes energy saving and consumption reduction, which may choose a lower fan speed under the premise of meeting the temperature constraints.

[0086] α3∑ k Noise fan,k (V k) is a noise control term in the objective function, which is designed to reduce the noise level generated during the operation of the cabinet to improve the working environment or meet the noise limit requirements of specific application scenarios. For example, the application scenario can be a laboratory or office environment. Noise fan,k (V k ) is the noise level of the kth fan, which is also a nonlinear function of the fan speed V k . For example, the noise increases significantly with the increase of the speed. The noise control term can be a normalized noise indicator. ∑ k represents the sum of the noise contributions of all fans, representing the overall noise level generated by all fans in the cabinet. α3 is the weight coefficient of the noise control term, also known as the priority adjustment coefficient of this term. The larger the value of α3, the more the multi-objective optimization algorithm values the noise control, and it may make trade-offs in energy consumption and temperature to achieve lower noise output.

[0087] The setting of the three weight coefficients α1, α2 and α3 is the key to achieving multi-objective balance. These coefficients are not fixed and can be dynamically configured and optimized according to actual working conditions, operation strategies and user preferences. For example, the weight coefficient determination method includes but is not limited to: offline optimization, based on a large amount of historical running data, through multi-objective optimization algorithm (such as NSGA-II or MOPSO) for offline training and optimization, to find the Pareto optimal frontier, and select the appropriate weight combination; expert experience, combined with the field knowledge of cabinet design experts and thermal management engineers for initialization setting; online adaptive adjustment, combined with machine learning or reinforcement learning technology, to make real-time or periodic fine-tuning during system operation according to external environmental changes, user feedback or higher-level system goals; multi-objective decision-making method, using the analytic hierarchy process to systematically quantify the importance of different objectives to guide weight allocation.

[0088] S308, based on the objective function, the constraint condition, the real-time temperature data, the thermal load prediction data and the temperature distribution prediction data, the optimal speed of the fan is optimized by using the multi-objective model predictive control algorithm.

[0089] Specifically, the multi-objective optimization algorithm can include multi-objective genetic algorithm, particle swarm algorithm or model predictive control algorithm (MPC for short). Although it does not appear directly in the objective function, the optimization implementation of the multi-objective optimization algorithm must strictly comply with a series of constraint conditions to ensure the safe and stable operation of the system. These constraint conditions can include fan speed constraint conditions, temperature constraint conditions in each partition, total power consumption constraint conditions and coordination constraint conditions, wherein the coordination constraint conditions include temperature difference constraints between adjacent partitions, speed relative relationship constraints and heat transfer balance constraints. For example, the temperature constraint condition in each partition is T min,i ≤ T i≤T max,i Minimum temperature T min,i and maximum temperature T max,i The settings ensure that the temperature in each zone remains within the safe operating range of the equipment; the fan speed constraint is V. min,j ≤V j ≤V max,j Minimum rotational speed V min,j and maximum rotational speed V max,j The settings ensure that the fan operates efficiently within its operational range; the total power consumption constraint is P. total ≤P limit Power consumption limit P limit The settings can limit the overall system energy consumption to a set threshold; cooperative constraints refer to constraints set considering the airflow interaction and thermal coupling effects between zones, to achieve coordinated operation of fans in each zone, avoiding the impact of local overcooling or overheating on other zones. For example, restrictions on the relative change of fan speed between adjacent zones to avoid airflow short-circuiting or local negative pressure, and constraints on heat transfer balance across zones. These constraints are incorporated into the optimization model in the form of linear or nonlinear inequalities, becoming the limiting conditions in the optimization solution process.

[0090] This optimization algorithm is triggered and executed in each control cycle to obtain the latest prediction results of temperature distribution and heat load, as well as the real-time temperature data acquired by the current temperature sensor. Based on these inputs, the objective function, and constraints, the algorithm calculates the optimal rotational speed for the next time step.

[0091] S309. Based on the optimal speed obtained through optimization, determine the control signal for each fan.

[0092] Specifically, based on the optimal speed corresponding to the next time step determined in step S308, the corresponding fan control signal can be determined according to the optimal speed and sent to the corresponding fan drive mechanism to realize the thermal management of the cabinet.

[0093] S310. Based on the optimization results of the control commands, control the temperature control equipment in each zone of the cabinet.

[0094] The cabinet temperature control method of the distributed control system provided by the embodiment is based on a recurrent neural network to determine thermal load prediction data in a future preset period. The temperature distribution prediction data in the future preset period is determined based on a deep neural network. The optimal rotating speed of the fan is optimized by using a multi-objective model predictive control algorithm, and then the control signals of the respective fans corresponding to the optimal rotating speed are determined, so that the fan control signals are accurately determined, the temperature control precision is improved, the temperature fluctuation is effectively inhibited, and the long-term stable operation of the key equipment in the cabinet is ensured. The partition control can finely dissipate heat for the corresponding partitions of different heat generating components, effectively eliminate local overheating points, reduce the chip junction temperature, thereby significantly prolong the service life of the system, and on the other hand, make the fan power distribution more reasonable, supply cooling on demand, and reduce the power consumption of the entire system. The neural network has self-learning and self-adaptive capabilities, and can efficiently and predictively control temperature changes, avoiding excessive overheating caused by lagging regulation.

[0095] Figure 4 The flowchart of another cabinet temperature control method of a distributed control system provided by the embodiment is based on the foregoing embodiment, and the cabinet temperature control method of the distributed control system comprises Figure 4 , the cabinet temperature control method of the distributed control system comprises:

[0096] S401, reconstruct a three-dimensional temperature distribution model in the cabinet by using real-time temperature data in the cabinet in combination with a spatial interpolation technique, to obtain temperature distribution data in the current cabinet.

[0097] Specifically, the real-time temperature data in the cabinet includes temperature values and corresponding position information, which can be obtained from temperature sensors distributed at various positions in the cabinet as input data for the spatial interpolation technique. The three-dimensional temperature distribution model refers to a mathematical model representing the temperature at any position in the cabinet, which can be constructed by using real-time temperature data in combination with a spatial interpolation technique. The spatial interpolation technique can include Kriging interpolation or radial basis function interpolation. The temperature distribution data includes temperature data of multiple position points distributed in the cabinet, which includes not only measured values collected by temperature sensors but also interpolation data obtained by the spatial interpolation technique. After reconstructing the three-dimensional temperature distribution model, the coordinates of any position in the cabinet can be input to obtain the temperature data of the point, thereby achieving the obtaining of the temperature distribution data. The user can determine the number of interpolation data in the temperature distribution data according to actual needs.

[0098] S402, perform real-time thermal risk judgment according to the temperature distribution data in the current cabinet.

[0099] Specifically, according to the temperature distribution data in the current cabinet, it is determined whether there is a position exceeding the temperature threshold. If there is, it is determined that the point has a thermal runaway risk and needs to be alarmed in the next step, otherwise it is determined that there is no thermal runaway risk in the cabinet and the next step of alarm is not needed.

[0100] S403, performing sound and light alarm and remote communication alarm according to the result of real-time thermal risk judgment.

[0101] Specifically, the sound and light alarm can be implemented by using a buzzer and a warning light, and the remote communication alarm is to transmit an alarm signal to the on-duty personnel or user through a communication device to remind the on-duty personnel or user to respond in time. In the case where it is determined in step S402 that the cabinet exists thermal runaway risk, sound and light alarm and remote communication alarm are performed, and the alarm signal can include the position information of the thermal runaway point, otherwise, sound and light alarm and remote communication alarm are not performed.

[0102] S404, determining thermal load prediction data and temperature distribution prediction data in a future preset period based on a neural network, in combination with historical thermal load data, system running state data, external environment data and temperature distribution data.

[0103] Specifically, the neural network is trained offline by combining historical data with corresponding interpolation data, wherein the historical data can include historical thermal load data, historical external environment data and historical temperature sensing data. The thermal load data refers to the power consumption data of the key heat generating devices in the distributed control system, which mainly drives the temperature change of the cabinet. The thermal load data can include the power consumption data of CPU, communication equipment and power supply. The external environment data refers to the environmental state parameters of the space outside the cabinet, which can include environmental temperature, environmental humidity and environmental air pressure. Exemplarily, the neural network includes a recurrent neural network and a deep neural network. The historical thermal load data, system running state data and external environment data are substituted into the corresponding recurrent neural network model to predict the thermal load prediction data in the future preset period. Then, the thermal load prediction data is substituted into the corresponding deep neural network model to predict the temperature distribution prediction data in the cabinet in the future preset period.

[0104] S405, performing thermal risk prediction according to the temperature distribution prediction data.

[0105] Specifically, similar to the real-time thermal risk judgment in step S402, according to the temperature distribution prediction data, it is determined whether there is a position exceeding the temperature threshold value, if there is, it is determined that the point exists thermal runaway risk in the future preset period, and the next step of alarm needs to be performed, otherwise, it is determined that there is no thermal runaway risk in the cabinet, and the next step of alarm does not need to be performed.

[0106] S406, performing sound and light alarm and remote communication alarm according to the result of thermal risk prediction.

[0107] Specifically, the sound and light alarm can be implemented by using a buzzer and a warning light, and the remote communication alarm is to transmit an alarm signal to the on-duty personnel or user through a communication device to remind the on-duty personnel or user to respond in time. In the case where it is determined in step S405 that the cabinet has a thermal runaway risk in the future preset time period, the sound and light alarm and the remote communication alarm are performed, and the alarm signal can include the position information of the thermal runaway point, otherwise the sound and light alarm and the remote communication alarm are not performed.

[0108] S407, based on the real-time temperature data, the thermal load prediction data and the temperature distribution prediction data, a multi-objective optimization algorithm is used to optimize the control instructions of the temperature adjusting devices in each partition of the cabinet.

[0109] Specifically, the partitions in the cabinet can be determined according to at least one of the functional distribution of the distributed control system, the heat source distribution and the cooling demand. Exemplarily, the partitions in the cabinet can include a CPU area, an input / output interface area, a power supply area and an environment area. Each partition is provided with at least one set of temperature adjusting devices and at least one set of temperature sensors, which can avoid the extensive and inefficient control mode caused by global unified control, and realize independent and coordinated detection and control of the temperature in the cabinet. The temperature adjusting device refers to a device for adjusting the temperature in each partition, which can include at least one of an air conditioner, a fan or an air duct.

[0110] The multi-objective optimization algorithm is an algorithm for solving optimization problems with multiple conflicting objectives. Exemplarily, the multi-objective optimization algorithm can include a non-dominated sorting genetic algorithm and a multi-objective particle swarm algorithm. The multi-objective optimization algorithm aims to minimize the temperature deviation, total energy consumption and noise, and optimizes the control instructions of each temperature adjusting device. Exemplarily, the optimization variable of the multi-objective optimization algorithm can be the target speed of the fan. After the target speed of the fan is determined, the corresponding control instruction is determined according to the difference between the target speed and the current speed of the fan. For example, if the target speed is less than the current speed, the duty cycle of the fan control instruction can be increased according to the speed difference.

[0111] S408, according to the optimization result of the control instructions, the temperature adjusting devices in each partition of the cabinet are controlled.

[0112] The cabinet control method of the distributed control system provided in this embodiment, after reconstructing a three-dimensional temperature distribution model within the cabinet using real-time temperature data and spatial interpolation technology to obtain the current temperature distribution data within the cabinet, also includes real-time thermal risk assessment based on the current temperature distribution data; and audible and visual alarms and remote communication alarms based on the results of the real-time thermal risk assessment. Furthermore, after determining the predicted thermal load data and temperature distribution data for a preset future period based on neural networks, combined with historical heat load data, system operating status data, external environment data, and temperature distribution data, the method also includes thermal risk prediction based on the temperature distribution prediction data; and audible and visual alarms and remote communication alarms based on the results of the thermal risk prediction. This achieves dual assessment of thermal risk and prediction, resulting in precise temperature control that reduces system failure rate. Combined with multi-mode alarms and remote monitoring management, it significantly reduces cabinet maintenance costs and risks.

[0113] The present invention also provides a cabinet temperature control device for a distributed control system. Figure 5 This is a schematic diagram of the composition of a cabinet temperature control device for a distributed control system provided in an embodiment of the present invention. Based on the foregoing embodiments, refer to... Figure 5 The cabinet temperature control device 500 of the distributed control system includes multiple temperature sensors 501, multiple temperature regulating devices 503, multiple environmental sensors 502 and a microcontroller 504. The microcontroller 504 is connected to the temperature sensors 501, the temperature regulating devices 503 and the environmental sensors 502 respectively. The microcontroller 504 is used to implement the cabinet temperature control method of the distributed control system in any of the above embodiments.

[0114] Optionally, Figure 6 This is a schematic diagram of the cabinet temperature control device of another distributed control system provided in an embodiment of the present invention. Based on the foregoing embodiments, refer to... Figure 6 The cabinet temperature control device 500 of the distributed control system also includes an audible and visual alarm module 601 and a communication module 602. The microcontroller 504 is connected to both the audible and visual alarm module 601 and the communication module 602. The audible and visual alarm module 601 is used to issue audible and visual alarms based on the results of real-time thermal risk assessment and thermal risk prediction. The communication module 602 is used to issue communication alarms based on the results of thermal risk assessment and thermal risk prediction.

[0115] For example, Figure 7 This is a schematic diagram illustrating the composition of a cabinet temperature control device in another distributed control system provided in an embodiment of the present invention. Figure 7This invention presents the macroscopic architecture of the cabinet temperature control device for a distributed control system proposed in this embodiment, detailing the hierarchical relationships, data flow, and control logic among the functional modules. The aim is to achieve accurate sensing, predictive control, and adaptive optimization of the temperature field within the cabinet of the distributed control system. (Refer to...) Figure 7 The cabinet temperature control device of the distributed control system can be divided into a sensing layer 701, a control and processing layer 702, an execution layer 703, and an external interaction layer 704.

[0116] The perception layer 701 is the system's data acquisition front end, responsible for acquiring the operating environment inside and outside the cabinet and the system's key operating parameters. A multi-point temperature sensor network 705 is deployed inside the cabinet in key hot zones and potential hot spots, enabling real-time, high-precision acquisition of multi-dimensional temperature data to assist the microcontroller in constructing the original temperature field information inside the cabinet. The device interface 706 provides a communication interface with the main control device or other functional modules of the distributed control system to obtain real-time operating loads of core heat-generating components, such as CPU utilization, network throughput, and power output, serving as key input data for assessing current and future system heat dissipation. The environmental sensor 502 monitors external environmental data such as temperature, humidity, and air pressure outside the cabinet or in the operating environment, providing external thermal boundary conditions for the system. The control and processing layer 702 serves as the system's core intelligent hub, responsible for in-depth data processing, intelligent decision-making, and the generation of control commands. The microcontroller 504 assumes the system's main control function, coordinating data exchange between modules and performing data acquisition and preprocessing tasks (including data filtering, calibration, and formatting). The core function of microcontroller 504 is to run the cabinet temperature control method of the distributed control system. This cabinet temperature control method is the core technology of this invention, integrating advanced machine learning techniques. It is responsible for comprehensively analyzing the data input from the sensing layer, performing complex calculations such as temperature field reconstruction, heat load prediction, temperature prediction, and multi-zone coordinated control. Based on the algorithm results, it generates precise fan control command outputs and triggers alarm signals to deal with abnormal situations. The fan drive module 708 can receive the PWM signal generated by the microcontroller and convert it into the electrical signal required to drive the DC cooling fan. The communication module 602 is responsible for the data link between the local temperature control system and the remote monitoring platform. Through supported communication protocols, such as Wi-Fi, Ethernet, or Modbus TCP, it realizes the uploading of real-time temperature data, fan operating status, alarm information, etc., and the reception of remote configuration commands. When the system detects an abnormal temperature or potential risk, the audible and visual alarm module 601 is immediately triggered to issue a local warning. At the same time, the communication module 602 pushes the alarm information to the remote monitoring platform to ensure that the management personnel respond in a timely manner. The alarm information can be notified to the management personnel through at least one of the following methods: SMS, email, and App push. The alarm information may include information such as the location of the abnormal area, the current temperature value, the expected temperature value, and the duration of the over-temperature.

[0117] The execution layer 703 is the key link in realizing physical intervention, translating control commands into actual heat dissipation actions. By precisely adjusting the speed of the cooling fan 713, differentiated and refined heat dissipation management of different logical zones within the cabinet can be achieved.

[0118] The external interaction layer 704 provides an interface between the system and external management systems and operators, enabling data visualization, remote management, and persistent storage. The remote monitoring platform 711 provides a graphical user interface, intuitively displaying the real-time temperature field, historical data, prediction curves, and fan operating status inside the cabinet. The remote monitoring platform 711 supports remote configuration parameters and policy adjustments to achieve remote and intelligent management of the cabinet temperature control devices in the distributed control system. Remote configuration parameters can include the target temperature for temperature adjustment and the preset temperature for alarms. The data storage module 712 is used to persistently store various types of data generated during system operation, including historical temperature data, historical heat load data, fan operation records, and algorithm model parameters. This data is crucial for system performance evaluation, fault diagnosis, and continuous optimization training of machine learning models.

[0119] Figure 8 A schematic diagram of an electronic device that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0120] like Figure 8 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0121] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0122] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as the cabinet control method of a distributed control system.

[0123] In some embodiments, the cabinet control method of the distributed control system can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the cabinet control method of the distributed control system described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the cabinet control method of the distributed control system by any other suitable means (e.g., by means of firmware).

[0124] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0125] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0126] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0127] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0128] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0129] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0130] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0131] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A cabinet temperature control method for a distributed control system, characterized in that, include: By combining real-time temperature data inside the rack with spatial interpolation technology, a three-dimensional temperature distribution model inside the rack is reconstructed to obtain the current temperature distribution data inside the rack. Based on neural networks, and combining historical heat load data, system operating status data, external environment data, and temperature distribution data, the predicted heat load data and temperature distribution data for a future preset time period are determined. Based on the real-time temperature data, the heat load prediction data, and the temperature distribution prediction data, a multi-objective optimization algorithm is used to optimize the control commands of the temperature control equipment in each zone of the cabinet. Based on the optimization results of the control commands, the temperature control equipment in each zone of the cabinet is controlled.

2. The cabinet temperature control method for a distributed control system according to claim 1, characterized in that, The neural network includes recurrent neural networks and deep neural networks; The process of determining the predicted heat load and temperature distribution data for the future preset time period based on neural networks, combined with historical heat load data, system operating status data, external environment data, and the temperature distribution data, includes: Based on the recurrent neural network, combined with the historical heat load data, the system's operating status data, and the external environment data, the predicted heat load data for a future preset time period is determined. Based on the deep neural network, and combined with the heat load prediction data and the current temperature distribution data in the cabinet, the temperature distribution prediction data for the future preset time period is determined.

3. The cabinet temperature control method for a distributed control system according to claim 2, characterized in that, The recurrent neural network includes a long short-term memory network model, which is trained offline using training data. The training data includes historical heat load data, historical external environment data, and historical temperature distribution data. The historical temperature distribution data is determined by reconstructing the temperature field using the three-dimensional temperature distribution model and the historical temperature data.

4. The cabinet temperature control method for a distributed control system according to claim 1, characterized in that, The spatial interpolation technique includes radial basis function interpolation; The process of reconstructing a three-dimensional temperature distribution model within the server rack using real-time temperature data to obtain current temperature distribution data within the rack includes: Using radial basis functions, the spatial correlation function φ(‖PP) of the sampling data of each temperature sensor with the temperature of the target point inside the cabinet is expressed. i ‖), where ‖PP i ‖ is the Euclidean distance between the target point P and the i-th temperature sensor, and φ() is the radial basis function; Based on each of the aforementioned spatial correlation functions and their corresponding weight coefficients, the correlation accumulation function is determined. φ(‖PP i ‖), where w i is the weighting coefficient associated with the i-th temperature sensor for the target point temperature, and N is the total number of temperature sensors; Combining the aforementioned correlation accumulation function and linear trend term, the correlation function between the temperature of the target point inside the cabinet and its three-dimensional position coordinates is expressed. As the three-dimensional temperature distribution model, x, y, and z are the coordinate values ​​on each coordinate axis in the three-dimensional spatial coordinate system, and β0, β1, β2, and β3 are the coefficients of the linear trend term; The real-time temperature data is input into the relevant function to reconstruct the temperature data of other locations within the cabinet, thereby obtaining the current temperature distribution data within the cabinet.

5. The cabinet temperature control method for a distributed control system according to claim 1, characterized in that, The system functions corresponding to each partition of the cabinet are different, and each partition is equipped with at least one set of temperature control devices and multiple temperature sensors.

6. The cabinet temperature control method of the distributed control system according to claim 5, characterized in that, The multi-objective optimization algorithm includes a multi-objective model predictive control algorithm; the temperature control device includes a fan; Based on the real-time temperature data, the predicted heat load data, and the predicted temperature distribution data, a multi-objective optimization algorithm is used to optimize the control commands for temperature control devices in each zone of the cabinet, including: The objective function of the multi-objective optimization algorithm is determined by setting the optimization objectives as minimizing temperature deviation, total energy consumption, and noise. Among them, V j Let J be the optimal speed of the j-th fan, J be the cost function, and T be the speed of the j-th fan. i T is the real-time temperature of the i-th partition. target,i For the target temperature of the i-th partition, P fan,j (V j ) represents the operating power of the j-th fan, Noise fan,k (V k ) represents the noise level of the k-th fan, and α1, α2, and α3 are the weighting coefficients for each item; Based on the objective function, constraints, real-time temperature data, heat load prediction data, and temperature distribution prediction data, the optimal speed of the fan is optimized using the multi-objective model predictive control algorithm. Based on the optimal speed obtained through optimization, control signals for each of the fans are determined.

7. The cabinet temperature control method for a distributed control system according to claim 6, characterized in that, The constraints include the fan speed constraints, the temperature constraints of each zone, the total power consumption constraints, and the collaborative constraints, wherein the collaborative constraints include the temperature difference constraints between adjacent zones, the relative speed constraints, and the heat transfer balance constraints.

8. The cabinet temperature control method for a distributed control system according to claim 1, characterized in that, After reconstructing a three-dimensional temperature distribution model within the server rack using real-time temperature data and spatial interpolation techniques to obtain the current temperature distribution data within the server rack, the method further includes: Real-time thermal risk assessment is performed based on the current temperature distribution data within the server rack; Based on the results of the real-time thermal risk assessment, audible and visual alarms and remote communication alarms are generated; After determining the predicted heat load data and temperature distribution data for the future preset time period based on neural networks, combined with historical heat load data, system operating status data, external environment data, and the temperature distribution data, the method further includes: Based on the temperature distribution prediction data, a thermal risk assessment is performed; Based on the results of the thermal risk prediction, the audible and visual alarms and the remote communication alarms are triggered.

9. A cabinet temperature control device for a distributed control system, characterized in that, include: The system comprises multiple temperature sensors, multiple temperature control devices, multiple environmental sensors, and a microcontroller, wherein the microcontroller is connected to the temperature sensors, the temperature control devices, and the environmental sensors respectively, and the microcontroller is used to implement the cabinet temperature control method of the distributed control system according to any one of claims 1-8.

10. The cabinet temperature control device of the distributed control system according to claim 9, characterized in that, Also includes: The system includes an audible and visual alarm module and a communication module, with the microcontroller connected to both the audible and visual alarm module and the communication module. The audible and visual alarm module is used to issue audible and visual alarms based on the results of real-time thermal risk assessment and thermal risk prediction. The communication module is used to issue communication alarms based on the results of thermal risk assessment and thermal risk prediction.

Images