Temperature prediction method and system for automated computer room based on artificial intelligence

Abstract

The present invention provides a temperature prediction method and system for an automated computer room based on artificial intelligence, and relates to the technical field of temperature prediction. The method includes: obtaining historical key point temperature data and historical thermal image data of a data center; establishing a three-dimensional space model of a data center, and aligning the coordinates of the historical data; establishing a temperature contribution weight model, and obtaining the temperature contribution weight of the key point temperature data to the thermal image data; predicting the key point predicted temperature based on the historical key point temperature data; and obtaining the three-dimensional scene temperature prediction data of the data center based on the temperature contribution weight. This solution makes full use of the comprehensiveness of infrared thermal imagers and the accuracy of temperature sensors, realizes the temperature prediction of the three-dimensional scene of the data center, and corrects the temperature of the infrared thermal imager through the temperature sensor, maximizes the advantages of infrared thermal imagers and temperature sensors, and improves the accuracy and comprehensiveness of the temperature prediction of the computer room of the data center.

Description

Technical Field

[0001] The present invention relates to the technical field of temperature prediction, and in particular to a temperature prediction method and system for an automated computer room based on artificial intelligence. Background Art

[0002] An automated computer room refers to a facility that uses automation technology to manage and control a data center or computer room. Such a computer room is usually equipped with advanced monitoring systems, automated operation and maintenance tools, and environmental control systems to improve operational efficiency, reduce labor costs, and ensure high availability and security of the system.

[0003] Temperature control is a key link in the environmental control system, which can help data center managers better control environmental conditions, thereby ensuring stable operation of equipment and extending its service life. In the process of temperature control, if the temperature can be predicted in advance, the accuracy of temperature control can be further improved. Servers and other IT equipment in data centers generate a lot of heat during operation. If the heat is not dissipated in a timely and effective manner, it may cause hardware damage or performance degradation. By accurately predicting the temperature change trend, measures can be taken in advance to avoid the formation of local hot spots or the rise of the overall average temperature, such as adjusting the working mode of the cooling system or adding additional cooling devices. On the other hand, accurately predicting the temperature can also help better plan the energy consumption of the cooling system and provide support for improving the overall efficiency of the system.

[0004] Therefore, how to effectively ensure the accuracy and comprehensiveness of temperature prediction data is an urgent problem to be solved. Summary of the invention

[0005] In order to improve the above problems, the present invention provides a temperature prediction method and system for an automated computer room based on artificial intelligence.

[0006] A first aspect of an embodiment of the present invention provides a temperature prediction method for an automated computer room based on artificial intelligence, the method comprising:

[0007] The automation room obtains the historical key point temperature data and historical thermal image data of the data center, wherein the historical key point temperature data is collected by the temperature sensor set at the key point position, and the historical thermal image data is collected by the infrared thermal imager set at two or more different angles;

[0008] Establishing a three-dimensional spatial model of the data center, and aligning the historical key point temperature data and the historical thermal image data on the three-dimensional spatial model according to the key point positions;

[0009] A temperature contribution weight model is established, and training is performed based on the aligned historical key point temperature data and historical thermal image data to obtain the temperature contribution weight of the key point temperature data to the thermal image data in terms of time sequence and three-dimensional space distance;

[0010] Predicting the temperature of the key point according to the historical key point temperature data to obtain the key point predicted temperature;

[0011] The predicted temperature of the key point and the currently acquired current thermal image data are input into the temperature contribution weight model, and the three-dimensional scene temperature prediction data of the data center is obtained according to the temperature contribution weight.

[0012] Optionally, the step of aligning the coordinates of the historical key point temperature data and the historical thermal image data on the three-dimensional space model according to the key point positions specifically includes:

[0013] Determine the key point position coordinates of the key point position in the three-dimensional space model;

[0014] Binding the key point position coordinates with the historical key point temperature data collected at the key point position coordinates;

[0015] Obtain three-dimensional scene temperature data at each three-dimensional coordinate in the three-dimensional space model according to historical thermal image data;

[0016] According to the three-dimensional coordinates in the three-dimensional space model, the historical key point temperature data bound with the key point position coordinates are aligned with the three-dimensional scene temperature data.

[0017] Optionally, the input of the temperature contribution weight model is the coordinate-aligned key point temperature data and the three-dimensional scene temperature data, and the output is the temperature contribution weight of the key point temperature data to the thermal image data in the time sequence and the three-dimensional space distance. The step of training based on the aligned historical key point temperature data and the historical thermal image data to obtain the temperature contribution weight of the key point temperature data to the thermal image data in the time sequence and the three-dimensional space distance specifically includes:

[0018] Grouping the historical key point temperature data and the three-dimensional scene temperature data obtained according to the historical thermal image data according to a time sequence;

[0019] According to the data collection time interval, the three-dimensional scene temperature data at a later time point is used as a comparison object of the three-dimensional scene temperature data at a previous time point for training;

[0020] For the key point temperature data collected by each temperature sensor, calculate the temperature contribution weight of the temperature value of the three-dimensional scene temperature data that changes over time in the three-dimensional coordinates from near to far in all directions outward with the key point position coordinates as the center.

[0021] Optionally, the step of performing training based on the aligned historical key point temperature data and historical thermal image data to obtain the temperature contribution weights of the key point temperature data to the thermal image data in time sequence and three-dimensional space distance specifically includes:

[0022] When the temperature value of the three-dimensional scene temperature data at a certain three-dimensional coordinate is simultaneously affected by a plurality of different key point temperature data, the temperature contribution weight of each key point temperature data to the three-dimensional scene temperature data at the three-dimensional coordinate is calculated respectively.

[0023] Optionally, when performing temperature prediction, the input of the temperature contribution weight model is the currently acquired key point temperature data, the predicted key point predicted temperature, and the three-dimensional scene temperature data obtained based on the currently acquired current thermal image data, and the output is the three-dimensional scene temperature prediction data obtained according to the temperature contribution weight. The step of obtaining the three-dimensional scene temperature prediction data of the data center according to the temperature contribution weight specifically includes:

[0024] Determining, according to the temperature contribution weight, the three-dimensional coordinates affected by the key point temperature data in the three-dimensional scene temperature data;

[0025] Extract the temperature values ​​at the affected three-dimensional coordinates respectively;

[0026] Calculate the predicted temperature values ​​at these three-dimensional coordinates according to the time interval between the predicted temperature of the key point and the current time point and the temperature contribution weight value at each three-dimensional coordinate at the corresponding time interval;

[0027] The predicted temperature value at the affected three-dimensional coordinate is combined with other three-dimensional scene temperature data that is not affected by the key point temperature data to obtain the three-dimensional scene temperature prediction data.

[0028] Optionally, the method for calculating the temperature contribution weight includes:

[0029] Extract the temperature values ​​of the three-dimensional scene temperature data on the three-dimensional coordinates from near to far in all directions outward with the key point position coordinates as the center;

[0030] Calculate the difference between the temperature value of the three-dimensional scene temperature data at each three-dimensional coordinate and the key point temperature data at the key point position coordinate;

[0031] Determine whether the temperature values ​​of the three-dimensional scene temperature data at the three-dimensional coordinates where the differences exist change in a direction that reduces the differences within a subsequent preset time threshold interval;

[0032] If yes, then determine these three-dimensional coordinates as three-dimensional coordinates affected by the key point temperature data;

[0033] For each three-dimensional coordinate affected by the key point temperature data, the temperature contribution weight of the key point temperature data to the three-dimensional scene temperature data is obtained according to the unit time gradient size of the temperature value changing in the direction that reduces the difference.

[0034] Optionally, the method further comprises:

[0035] Performing learning and training on the historical thermal image data through an artificial intelligence model;

[0036] When performing temperature prediction, the predicted thermal image data at the same time point as the predicted temperature of the key point is obtained through the trained artificial model;

[0037] The three-dimensional scene temperature data is obtained by using the predicted thermal image data to replace the current thermal image data.

[0038] Optionally, the step of predicting the temperature of the key point according to the historical key point temperature data to obtain the key point predicted temperature specifically further includes:

[0039] The time series features of the temperature data at key historical points are extracted through the LSTM model;

[0040] The temperature of the key point is predicted according to the extracted time series features to obtain the predicted temperature of the key point at a specific time point.

[0041] Optionally, the step of obtaining the three-dimensional scene temperature data at each three-dimensional coordinate in the three-dimensional space model according to the historical thermal image data specifically includes:

[0042] Extracting the depth information of the data center from the 3D spatial model;

[0043] Based on the depth information and historical thermal image data acquired by the infrared thermal imager at two or more different angles, three-dimensional scene temperature data on three-dimensional coordinates that can be directly captured by the infrared thermal imager is calculated;

[0044] The interpolation method is used to estimate the temperature data of the three-dimensional scene at the three-dimensional coordinates that cannot be photographed;

[0045] The calculation results are combined to obtain the three-dimensional scene temperature data at each three-dimensional coordinate in the three-dimensional space model.

[0046] A second aspect of an embodiment of the present invention provides a temperature prediction system for an automated computer room based on artificial intelligence, comprising:

[0047] A historical data acquisition unit, used to acquire historical key point temperature data and historical thermal image data of the data center, wherein the historical key point temperature data is acquired by a temperature sensor set at a key point position, and the historical thermal image data is acquired by an infrared thermal imager set at two or more different angles;

[0048] A three-dimensional model building unit, used to build a three-dimensional space model of the data center, and align the historical key point temperature data and the historical thermal image data on the three-dimensional space model according to the key point positions;

[0049] A weight model building unit is used to build a temperature contribution weight model, and to perform training based on aligned historical key point temperature data and historical thermal image data to obtain the temperature contribution weight of the key point temperature data to the thermal image data in terms of time sequence and three-dimensional space distance;

[0050] A key point temperature prediction unit, used to predict the temperature of the key point according to the historical key point temperature data to obtain a key point predicted temperature;

[0051] The three-dimensional temperature prediction unit is used to input the key point predicted temperature and the current thermal image data currently acquired into the temperature contribution weight model, and obtain the three-dimensional scene temperature prediction data of the data center according to the temperature contribution weight.

[0052] In summary, the present invention provides a temperature prediction method and system for an automated computer room based on artificial intelligence. Based on the temperature data collected by an infrared thermal imager and a temperature sensor from two different angles, the comprehensiveness of the infrared thermal imager and the accuracy of the temperature sensor are fully utilized to realize the temperature prediction of the three-dimensional scene of the data center. The temperature of the infrared thermal imager is corrected by the temperature sensor, so as to maximize the advantages of the infrared thermal imager and the temperature sensor, and improve the accuracy and comprehensiveness of the temperature prediction of the computer room of the data center. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0054] Figure 1 A method flow chart of a temperature prediction method for an automated computer room based on artificial intelligence according to an embodiment of the present invention;

[0055] Figure 2 This is a functional module block diagram of an artificial intelligence-based automated computer room temperature prediction system according to an embodiment of the present invention.

[0056] Reference numerals:

[0057] Historical data acquisition unit 110; three-dimensional model building unit 120; weight model building unit 130; key point temperature prediction unit 140; three-dimensional temperature prediction unit 150. DETAILED DESCRIPTION

[0058] Temperature control is a key link in the environmental control system of the automated computer room. It can help data center managers better control environmental conditions, thereby ensuring stable operation of equipment and extending its service life. In the process of temperature control, if the temperature can be predicted in advance, the accuracy of temperature control can be further improved. Servers and other IT equipment in the data center generate a lot of heat during operation. If the heat is not dissipated in a timely and effective manner, it may cause hardware damage or performance degradation. By accurately predicting the temperature change trend, measures can be taken in advance to avoid the formation of local hot spots or the rise of the overall average temperature, such as adjusting the working mode of the cooling system or adding additional cooling devices. On the other hand, accurately predicting the temperature can also help better plan the energy consumption of the cooling system and provide support for improving the overall efficiency of the system.

[0059] Therefore, effective temperature monitoring is crucial to ensuring the stable operation of the data center. When conducting temperature monitoring, how to effectively ensure the accuracy and comprehensiveness of the temperature monitoring data is a problem that needs to be solved urgently.

[0060] In view of this, the designers of the present invention have designed a temperature prediction method and system for an automated computer room based on artificial intelligence.

[0061] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0062] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0064] In the description of the present invention, it should be noted that the terms "top", "bottom", "inside", "outside", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the invented product is usually placed when in use, which is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0065] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0066] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0067] In the data center, the collection of temperature data through infrared thermal imagers and temperature sensors are two common temperature monitoring methods. The data processing system of the automated computer room can respectively communicate with the infrared thermal imager and the temperature sensor to obtain the temperature data collected by them, and process and save the collected temperature data. Based on the processing results of the acquired temperature data, the cooling system in the environmental control system of the automated computer room can perform corresponding control actions to achieve effective temperature control of the data center.

[0068] Among them, infrared thermal imagers can achieve non-contact measurement, and temperature measurement can be performed without direct contact with the equipment. For some key equipment, the impact of temperature monitoring equipment on the equipment can be reduced. At the same time, infrared thermal imagers can generate temperature distribution images of the entire area, providing intuitive thermal images, which are convenient for understanding and analyzing temperature distribution and for subsequent implementation of countermeasures. The characteristic of temperature sensors is that they can provide very accurate temperature readings, especially when measuring at fixed points. Therefore, when predicting temperature, if the advantages of the two measurement methods can be combined, more accurate and comprehensive temperature prediction data can be obtained.

[0069] The following is a detailed description of a temperature prediction method for an automated computer room based on artificial intelligence provided in this embodiment.

[0070] See also Figure 1 This embodiment provides a temperature prediction method for an automated computer room based on artificial intelligence, the method comprising:

[0071] Step S101: The automation room obtains the historical key point temperature data and historical thermal image data of the data center. The historical key point temperature data is collected by the temperature sensor set at the key point position, and the historical thermal image data is collected by the infrared thermal imager set at two or more different angles.

[0072] Temperature sensors are relatively accurate in monitoring temperature data at a single point, and are usually set at key points that require special attention. Historical key point temperature data is a collection of key point temperature data collected by temperature sensors over a certain period of time in the past. By analyzing historical key point temperature data, temperature prediction for key point locations can be achieved to a certain extent.

[0073] Infrared thermal imagers can obtain thermal image data within the shooting range, but due to the shooting angle, a single infrared thermal imager can only obtain thermal image data from one perspective. Considering the different shapes of equipment in the data center computer room and the influence of the internal structure of the computer room, in order to obtain more complete thermal image data of the entire computer room through the infrared thermal imager, it is necessary to shoot the same scene from different angles. Therefore, it is necessary to obtain thermal image data from at least two angles at the same time to basically meet the actual needs.

[0074] Step S102 , establishing a three-dimensional space model of the data center, and aligning the coordinates of the historical key point temperature data and the historical thermal image data on the three-dimensional space model according to the key point positions.

[0075] Establishing a three-dimensional spatial model of the data center can standardize the internal spatial structure of the data center. By aligning the coordinates of the historical key point temperature data and the historical thermal image data on the three-dimensional spatial model, the temperature data of two different temperature monitoring devices can be expressed in the same coordinate system, facilitating unified data processing.

[0076] Specifically, as a preferred implementation, step S102 specifically includes:

[0077] Determine the key point position coordinates of the key point position in the three-dimensional space model;

[0078] Binding the key point position coordinates with the historical key point temperature data collected at the key point position coordinates;

[0079] Obtain three-dimensional scene temperature data at each three-dimensional coordinate in the three-dimensional space model according to historical thermal image data;

[0080] According to the three-dimensional coordinates in the three-dimensional space model, the historical key point temperature data bound with the key point position coordinates are aligned with the three-dimensional scene temperature data.

[0081] Based on the key point positions set by the temperature sensor in the real scene, the corresponding coordinates of the temperature sensor in the three-dimensional space model can be obtained. Based on this, the three-dimensional coordinates in the three-dimensional space can be associated with the temperature data at the coordinate position. On the other hand, according to the shooting content of the thermal image data, each pixel in the thermal image data can be mapped to the coordinates in the three-dimensional space, and then the temperature value obtained according to the pixel value can be associated with the three-dimensional coordinates. After associating the historical key point temperature data and the historical thermal image data with the unified three-dimensional coordinates respectively, the coordinate alignment of the two in the three-dimensional space model is achieved.

[0082] It should be noted that when obtaining the 3D scene temperature data at each 3D coordinate in the 3D space model based on the historical thermal image data, since the infrared thermal imager takes a flat image, it is necessary to merge the thermal image data from two or more different angles to obtain the 3D scene temperature data. The specific steps include:

[0083] Extracting the depth information of the data center from the 3D spatial model;

[0084] Based on the depth information and historical thermal image data acquired by the infrared thermal imager at two or more different angles, three-dimensional scene temperature data on three-dimensional coordinates that can be directly captured by the infrared thermal imager is calculated;

[0085] The interpolation method is used to estimate the temperature data of the three-dimensional scene at the three-dimensional coordinates that cannot be photographed;

[0086] The calculation results are combined to obtain the three-dimensional scene temperature data at each three-dimensional coordinate in the three-dimensional space model.

[0087] When obtaining the temperature data of the three-dimensional scene at each three-dimensional coordinate in the three-dimensional space model based on the infrared thermal imager, in addition to the thermal image data from multiple angles mentioned above, the depth information of the data center room is also required. The depth information can be directly extracted from the established three-dimensional space model. When merging,

[0088] First, the thermal images of different perspectives are registered to align them in the coordinate system of the established three-dimensional space model. Then the depth information is registered with the thermal image to ensure that each pixel corresponds to a coordinate point in the three-dimensional space. Finally, the temperature value of each pixel in the thermal image is associated with its corresponding three-dimensional coordinate point.

[0089] It should be noted that based on the temperature detection principle of infrared thermal imagers, only the surface temperature of the photographed object can be accurately obtained, that is, the temperature value of each pixel can be directly mapped to the point of its corresponding three-dimensional coordinates. For points that cannot be directly captured in the thermal image, they can only be estimated through the temperature values ​​of other points that can be directly captured (interpolation methods can be used, such as nearest neighbor interpolation, bilinear interpolation, cubic spline interpolation, etc.), which may cause the temperature values ​​obtained by difference calculation to be inaccurate.

[0090] Through the above process, the preprocessing of historical key point temperature data and historical thermal image data is completed.

[0091] Step S103, establishing a temperature contribution weight model, training based on the aligned historical key point temperature data and historical thermal image data, and obtaining the temperature contribution weights of the key point temperature data to the thermal image data in time sequence and three-dimensional space distance.

[0092] As mentioned in the above description, the temperature values ​​at some coordinates in the 3D scene temperature data are obtained by difference calculation, so there is an inaccuracy problem, which is caused by the working principle of the infrared thermal imager. Correspondingly, based on the working principle of the temperature sensor, the key point temperature data collected at the key point position is relatively accurate.

[0093] In actual scenes, the key point positions are usually selected at locations where temperature fluctuations or changes occur frequently. Therefore, if the temperature changes rapidly, the temperature sensor set at the key point position can monitor the temperature change more timely. For infrared thermal imagers, if the coordinates where the temperature changes cannot be directly captured, it is necessary to wait for the temperature to be transferred to the coordinate position that can be directly captured before it can be reflected in the thermal image data. However, there is a certain delay and attenuation in the conduction of temperature. After aligning the historical key point temperature data and the three-dimensional scene temperature data obtained based on the historical thermal image data, the differences in time and position of the temperature data can be found by comparing the continuous data changes of the two within the same period of time.

[0094] For example, when the temperature of a key point inside the device suddenly rises, the temperature sensor set at that position can immediately monitor the temperature change, while for the infrared thermal imager, it needs to wait for the temperature to be transmitted to the surface of the device before it can be photographed, which results in a time delay. At the same time, the temperature value reflected in the captured thermal image data may be lower than the temperature value monitored by the temperature sensor, which is caused by the attenuation of temperature over distance. For some coordinates that are more than a certain distance from the key point, since the temperature cannot be transmitted here, no matter how the temperature at the key point changes, it will not be affected. Let's take a specific example to illustrate. Assuming that when a device is working, its two different surfaces are both 25 degrees through thermal image data feedback. Based on the interpolation algorithm, it can be calculated that the coordinate temperature between the two surfaces is also 25 degrees. When the temperature of a certain point inside the device suddenly rises to 35 degrees, since the temperature transmission takes time, when the temperature sensor has already detected 35 degrees, the thermal image data captured by the infrared thermal imager still shows 25 degrees, and there is a 10-degree data error for the key point. As time goes by, the thermal image data captures the temperature rise. At this time, due to the attenuation of distance, the thermal image data captured by the infrared thermal imager still shows 30 degrees. Based on the interpolation algorithm, it can be calculated that the temperature of the key point position between the two surfaces is also 30 degrees. There is a 5-degree data error at the key point position. It can be seen that this error changes with the transfer of temperature, that is, the passage of time. If you need to get the accurate temperature, you need to compensate the thermal image data based on the key point temperature data obtained by the sensor, and the specific compensation ratio needs to be determined based on two factors: time sequence and three-dimensional space distance.

[0095] The temperature contribution weight is a parameter defined to compensate for the deviation of temperature data caused by temperature conduction. The larger the temperature contribution weight, the higher the proportion that needs to be compensated. This compensation is limited to the coordinates obtained by interpolation calculation, and no compensation is required for the coordinates of the temperature value directly captured by the infrared thermal imager.

[0096] By establishing a temperature contribution weight model, based on actual historical data, the weight that needs to be compensated for the three-dimensional coordinates around each key point position is calculated, which facilitates the corresponding compensation of the temperature values ​​on these three-dimensional space coordinates during subsequent temperature prediction.

[0097] Specifically, the input of the temperature contribution weight model is the coordinate-aligned key point temperature data and the three-dimensional scene temperature data, and the output is the temperature contribution weight of the key point temperature data to the thermal image data in time sequence and three-dimensional space distance.

[0098] A large amount of pre-processed coordinate-aligned key point temperature data and three-dimensional scene temperature data are input into the temperature contribution weight model for training to obtain accurate temperature contribution weight distribution.

[0099] Specifically, the method for model training includes:

[0100] Grouping the historical key point temperature data and the three-dimensional scene temperature data obtained according to the historical thermal image data according to a time sequence;

[0101] According to the data collection time interval, the three-dimensional scene temperature data at a later time point is used as a comparison object of the three-dimensional scene temperature data at a previous time point for training;

[0102] For the key point temperature data collected by each temperature sensor, calculate the temperature contribution weight of the temperature value of the three-dimensional scene temperature data that changes over time in the three-dimensional coordinates from near to far in all directions outward with the key point position coordinates as the center.

[0103] It should be noted that since the temperature contribution weight is affected by the time dimension, when performing group training, the historical key point temperature data and the historical thermal image data need to be aligned in time order according to the data collection time interval. By comparing the differences at different time points, the impact of time delay on the temperature contribution weight is determined. By judging whether the three-dimensional scene temperature data changes in the direction of the key point temperature data, the impact of distance on the temperature contribution weight is determined.

[0104] Specifically, the method for calculating the temperature contribution weight includes:

[0105] Extract the temperature values ​​of the three-dimensional scene temperature data on the three-dimensional coordinates from near to far in all directions outward with the key point position coordinates as the center;

[0106] Calculate the difference between the temperature value of the three-dimensional scene temperature data at each three-dimensional coordinate and the key point temperature data at the key point position coordinate;

[0107] Determine whether the temperature values ​​of the three-dimensional scene temperature data at the three-dimensional coordinates where the differences exist change in a direction that reduces the differences within a subsequent preset time threshold interval;

[0108] If yes, then determine these three-dimensional coordinates as three-dimensional coordinates affected by the key point temperature data;

[0109] For each three-dimensional coordinate affected by the key point temperature data, the temperature contribution weight of the key point temperature data to the three-dimensional scene temperature data is obtained according to the unit time gradient size of the temperature value changing in the direction that reduces the difference.

[0110] If the temperature value of the three-dimensional scene temperature data changes in a direction that reduces the difference, it means that these coordinates are coordinates that need to be compensated. For the coordinates that need to be compensated, the numerical value of the temperature contribution weight is determined according to the speed of the temperature change and the value of the change. When the temperature at the key point position changes, the key point temperature data has a greater temperature contribution weight for the three-dimensional scene temperature data at the coordinates closer to the key point position, and the smaller the time delay is. The closer and farther the key point position is, the smaller the temperature contribution weight is, and the greater the time delay is. For example, there are two points A and B, where the distance from A to the key point position is smaller than that from B to the key point position. When the temperature influence is transmitted from the key point position to A, the temperature contribution weight at A position gradually increases from 0 until it stabilizes to a fixed value a, which is related to the distance from A to the key point position. When the temperature influence is transmitted from the key point to A, since the distance of B is greater than A, the temperature contribution weight at B is still 0 until a certain period of time passes and the temperature influence is transmitted from the key point to B. At this time, the temperature contribution weight at B gradually increases from 0 until it stabilizes to another fixed value b, which is related to the distance from B to the key point, and b≤a.

[0111] It should be noted that in actual scenes, the density of temperature sensors is usually relatively high for areas that need to be focused on. Therefore, the temperature value of the three-dimensional scene temperature data at some three-dimensional coordinates will be affected by multiple different key point temperature data at the same time. As a preferred implementation of the embodiment of the present invention, when the temperature value of the three-dimensional scene temperature data at a certain three-dimensional coordinate is affected by multiple different key point temperature data at the same time, the temperature contribution weight of each key point temperature data to the three-dimensional scene temperature data at the three-dimensional coordinate is calculated separately.

[0112] In different data centers, due to the differences in the internal structures of the data centers, the differences in the materials and shapes of the electronic equipment used, and the different locations of the temperature sensors, the temperature contribution weights calculated after training with historical key point temperature data and historical thermal image data are also different.

[0113] Step S104: predicting the temperature of the key point according to the historical key point temperature data to obtain the key point predicted temperature.

[0114] There are many methods for presetting the data of a single point based on the historical temperature data. In this embodiment, as a preferred implementation, step S104 specifically includes:

[0115] The time series features of the temperature data at key historical points are extracted through the LSTM model;

[0116] The temperature of the key point is predicted according to the extracted time series features to obtain the predicted temperature of the key point at a specific time point.

[0117] The LSTM (Long Short-Term Memory) model is a special type of recurrent neural network (RNN) model. The characteristics of LSTM make it very suitable for processing and predicting important events in time series, even if the time intervals between these events are long.

[0118] As other implementation methods of the embodiments of the present invention, there are many ways to predict single-point temperature data based on historical data, which can be to extract the time series characteristics of historical data, or to predict based on other relevant parameters. In this solution, any publicly available single-point temperature prediction algorithm can be used to predict the single-point temperature, and no specific limitation is made here.

[0119] After obtaining the predicted temperature of the key points, the three-dimensional scene of the data center can be predicted based on the temperature contribution weight. It should be noted that the predicted temperature of the key points includes not only the specific temperature value, but also the predicted time point of the temperature value.

[0120] Step S105 , inputting the predicted temperature of the key point and the currently acquired current thermal image data into the temperature contribution weight model, and obtaining the three-dimensional scene temperature prediction data of the data center according to the temperature contribution weight.

[0121] When performing temperature prediction, the input of the temperature contribution weight model is the currently acquired key point temperature data, the predicted key point predicted temperature, and the three-dimensional scene temperature data obtained based on the currently acquired current thermal image data, and the output is the three-dimensional scene temperature prediction data obtained according to the temperature contribution weight.

[0122] Inside the data center, the environmental conditions are very fixed, and the main factors that cause temperature changes include two: one is the temperature increase of the equipment, and the other is the cooling effect of the refrigeration equipment. Therefore, the temperature change will not appear suddenly on a large range of coordinates, but is based on the temperature change of a small number of points, and then gradually spreads to a certain area. Therefore, when predicting the temperature of the entire three-dimensional scene, it is mainly based on the three-dimensional scene temperature data obtained from the current thermal image data currently acquired, and then the three-dimensional scene temperature data is temperature compensated on some three-dimensional coordinates according to the predicted key point predicted temperature, and the three-dimensional scene temperature prediction data is calculated. When performing compensation calculations, it is necessary to use the temperature contribution weight obtained by the temperature contribution weight model. The process of calculating the three-dimensional scene temperature prediction data is to bring the key point predicted temperature into the temperature contribution weight at the corresponding three-dimensional coordinates, and then combine it with the three-dimensional scene temperature data. Specifically, as a preferred implementation, step S105 specifically includes:

[0123] Determining, according to the temperature contribution weight, the three-dimensional coordinates affected by the key point temperature data in the three-dimensional scene temperature data;

[0124] Extract the temperature values ​​at the affected three-dimensional coordinates respectively;

[0125] Calculate the predicted temperature values ​​at these three-dimensional coordinates according to the time interval between the predicted temperature of the key point and the current time point and the temperature contribution weight value at each three-dimensional coordinate at the corresponding time interval;

[0126] The predicted temperature value at the affected three-dimensional coordinate is combined with other three-dimensional scene temperature data that is not affected by the key point temperature data to obtain the three-dimensional scene temperature prediction data.

[0127] When calculating the predicted temperature value for each three-dimensional coordinate, first calculate the difference between the value of the three-dimensional scene temperature data on the three-dimensional coordinate and the predicted temperature of the key point, and then, according to the time point corresponding to the predicted temperature of the key point, bring in the value of the temperature contribution weight at different time delays to obtain the predicted temperature value at each time delay point in the future for a certain period of time on the three-dimensional coordinate.

[0128] By combining the predicted temperature value calculation results of all three-dimensional coordinates that need to be calculated, the three-dimensional scene temperature prediction data of the data center at each time delay point is obtained.

[0129] Specifically, as a preferred implementation, step S105 specifically further includes:

[0130] Performing learning and training on the historical thermal image data through an artificial intelligence model;

[0131] When performing temperature prediction, the predicted thermal image data at the same time point as the predicted temperature of the key point is obtained through the trained artificial model;

[0132] The three-dimensional scene temperature data is obtained by using the predicted thermal image data to replace the current thermal image data.

[0133] In order to further improve the accuracy of the prediction, in addition to directly performing temperature compensation calculations on the three-dimensional scene temperature data obtained based on the current thermal image data currently acquired, it is also possible to first predict the thermal image data separately to obtain predicted thermal image data at subsequent time delay points, and then use the same calculation method to perform temperature compensation calculations based on the three-dimensional scene temperature data obtained by predicting the thermal image data to obtain three-dimensional scene temperature prediction data.

[0134] It should be noted that there are currently many public algorithms for predicting thermal image data at a certain point in the future based on historical thermal image data using artificial intelligence models. The embodiments of the present invention do not make specific limitations on this, and any public algorithm may be used.

[0135] The temperature prediction method provided by the embodiment of the present invention can not only predict the temperature of the key point position, but also realize the three-dimensional scene temperature prediction of the entire data center based on the temperature prediction result of the key point position, and realizes the prediction for the temperature of each three-dimensional coordinate, making the prediction more comprehensive. Since the predicted temperature value combines the characteristics of the temperature sensor and the infrared thermal imager, the prediction result is more accurate.

[0136] Through the above-mentioned whole method process, the three-dimensional scene temperature prediction data of the whole data center can be obtained. On this basis, for the convenience of viewing, the three-dimensional scene temperature prediction data can be visualized, and then the three-dimensional visualization software can be used to display the three-dimensional scene temperature prediction results of the data center. Color coding can be used to represent different temperature ranges to make the temperature distribution more intuitive.

[0137] In summary, the present invention provides an artificial intelligence-based temperature prediction method for an automated computer room. Based on the temperature data collected by an infrared thermal imager and a temperature sensor from two different angles, the method makes full use of the comprehensiveness of the infrared thermal imager and the accuracy of the temperature sensor, realizes the temperature prediction of the three-dimensional scene of the data center, and corrects the temperature of the infrared thermal imager through the temperature sensor, maximizes the advantages of the infrared thermal imager and the temperature sensor, and improves the accuracy and comprehensiveness of the temperature prediction of the data center computer room.

[0138] like Figure 2 As shown, the present invention implements an artificial intelligence-based temperature prediction system for an automated computer room, the system comprising:

[0139] A historical data acquisition unit 110 is used to acquire historical key point temperature data and historical thermal image data of the data center, wherein the historical key point temperature data is acquired by a temperature sensor disposed at a key point position, and the historical thermal image data is acquired by an infrared thermal imager disposed at two or more different angles;

[0140] A three-dimensional model building unit 120 is used to build a three-dimensional space model of the data center, and align the historical key point temperature data and the historical thermal image data on the three-dimensional space model according to the key point positions;

[0141] The weight model building unit 130 is used to build a temperature contribution weight model, and to perform training based on the aligned historical key point temperature data and historical thermal image data to obtain the temperature contribution weight of the key point temperature data to the thermal image data in terms of time sequence and three-dimensional space distance;

[0142] A key point temperature prediction unit 140, configured to predict the temperature of the key point according to the historical key point temperature data to obtain a key point predicted temperature;

[0143] The three-dimensional temperature prediction unit 150 is used to input the key point predicted temperature and the currently acquired current thermal image data into the temperature contribution weight model, and obtain the three-dimensional scene temperature prediction data of the data center according to the temperature contribution weight.

[0144] The temperature prediction system of the automated computer room based on artificial intelligence provided in the embodiment of the present invention is used to implement the temperature prediction method of the automated computer room based on artificial intelligence. Therefore, the specific implementation method is the same as the above method and will not be repeated here.

[0145] In summary, the present invention provides an artificial intelligence-based temperature prediction method and system for an automated computer room. Based on the temperature data collected by an infrared thermal imager and a temperature sensor from two different angles, the present invention makes full use of the comprehensiveness of the infrared thermal imager and the accuracy of the temperature sensor, realizes the temperature prediction of the three-dimensional scene of the data center, and corrects the temperature of the infrared thermal imager through the temperature sensor, maximizes the advantages of the infrared thermal imager and the temperature sensor, and improves the accuracy and comprehensiveness of the temperature prediction of the data center computer room.

[0146] In several embodiments disclosed in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and a module, a program segment or a part of the code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0147] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0148] If the functions are implemented in the form of software function modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

Claims

1. A temperature prediction method for an automated computer room based on artificial intelligence, characterized in that: The method comprises: The automation room obtains the historical key point temperature data and historical thermal image data of the data center, wherein the historical key point temperature data is collected by the temperature sensor set at the key point position, and the historical thermal image data is collected by the infrared thermal imager set at two or more different angles; Establishing a three-dimensional spatial model of the data center, and aligning the historical key point temperature data and the historical thermal image data on the three-dimensional spatial model according to the key point positions; A temperature contribution weight model is established, and training is performed based on the aligned historical key point temperature data and historical thermal image data to obtain the temperature contribution weight of the key point temperature data to the thermal image data in terms of time sequence and three-dimensional space distance; Predicting the temperature of the key point according to the historical key point temperature data to obtain the key point predicted temperature; Inputting the predicted temperature of the key point and the currently acquired current thermal image data into the temperature contribution weight model, and obtaining the three-dimensional scene temperature prediction data of the data center according to the temperature contribution weight; The input of the temperature contribution weight model is the coordinate-aligned key point temperature data and the three-dimensional scene temperature data, and the output is the temperature contribution weight of the key point temperature data to the thermal image data in the time sequence and the three-dimensional space distance. The step of training based on the aligned historical key point temperature data and the historical thermal image data to obtain the temperature contribution weight of the key point temperature data to the thermal image data in the time sequence and the three-dimensional space distance specifically includes: Grouping the historical key point temperature data and the three-dimensional scene temperature data obtained according to the historical thermal image data according to a time sequence; According to the data collection time interval, the three-dimensional scene temperature data at a later time point is used as a comparison object of the three-dimensional scene temperature data at a previous time point for training; For the key point temperature data collected by each temperature sensor, calculate the temperature contribution weight of the temperature value of the three-dimensional scene temperature data that changes over time in the three-dimensional coordinates from near to far in all directions outward with the key point position coordinate as the center; The step of training based on the aligned historical key point temperature data and historical thermal image data to obtain the temperature contribution weight of the key point temperature data to the thermal image data in time sequence and three-dimensional space distance specifically includes: When the temperature value of the three-dimensional scene temperature data at a certain three-dimensional coordinate is simultaneously affected by a plurality of different key point temperature data, the temperature contribution weight of each key point temperature data to the three-dimensional scene temperature data at the three-dimensional coordinate is calculated respectively; When performing temperature prediction, the input of the temperature contribution weight model is the currently acquired key point temperature data, the predicted key point predicted temperature, and the three-dimensional scene temperature data obtained based on the currently acquired current thermal image data, and the output is the three-dimensional scene temperature prediction data obtained according to the temperature contribution weight. The step of obtaining the three-dimensional scene temperature prediction data of the data center according to the temperature contribution weight specifically includes: Determining, according to the temperature contribution weight, the three-dimensional coordinates affected by the key point temperature data in the three-dimensional scene temperature data; Extract the temperature values ​​at the affected three-dimensional coordinates respectively; Calculate the predicted temperature values ​​at these three-dimensional coordinates according to the time interval between the predicted temperature of the key point and the current time point and the temperature contribution weight value at each three-dimensional coordinate at the corresponding time interval; The predicted temperature value at the affected three-dimensional coordinate is combined with other three-dimensional scene temperature data that is not affected by the key point temperature data to obtain the three-dimensional scene temperature prediction data; The calculation method of the temperature contribution weight includes: Extract the temperature values ​​of the three-dimensional scene temperature data on the three-dimensional coordinates from near to far in all directions outward with the key point position coordinates as the center; Calculate the difference between the temperature value of the three-dimensional scene temperature data at each three-dimensional coordinate and the key point temperature data at the key point position coordinate; Determine whether the temperature values ​​of the three-dimensional scene temperature data at the three-dimensional coordinates where the differences exist change in a direction that reduces the differences within a subsequent preset time threshold interval; If yes, then determine these three-dimensional coordinates as three-dimensional coordinates affected by the key point temperature data; For each three-dimensional coordinate affected by the key point temperature data, the temperature contribution weight of the key point temperature data to the three-dimensional scene temperature data is obtained according to the unit time gradient size of the temperature value changing in the direction that reduces the difference.

2. The temperature prediction method of an automated computer room based on artificial intelligence according to claim 1 is characterized in that: The step of aligning the coordinates of the historical key point temperature data and the historical thermal image data on the three-dimensional space model according to the key point positions specifically includes: Determine the key point position coordinates of the key point position in the three-dimensional space model; Binding the key point position coordinates with the historical key point temperature data collected at the key point position coordinates; According to the historical thermal image data, the three-dimensional scene temperature data at each three-dimensional coordinate in the three-dimensional space model is obtained; According to the three-dimensional coordinates in the three-dimensional space model, the historical key point temperature data bound with the key point position coordinates are aligned with the three-dimensional scene temperature data.

3. The temperature prediction method of an automated computer room based on artificial intelligence according to claim 2 is characterized in that: The method further comprises: Performing learning and training on the historical thermal image data through an artificial intelligence model; When performing temperature prediction, the predicted thermal image data at the same time point as the predicted temperature of the key point is obtained through the trained artificial intelligence model; The three-dimensional scene temperature data is obtained by using the predicted thermal image data to replace the current thermal image data.

4. The temperature prediction method of an automated computer room based on artificial intelligence according to claim 3 is characterized in that: The step of predicting the temperature of the key point according to the historical key point temperature data to obtain the key point predicted temperature specifically includes: The time series features of the temperature data at key historical points are extracted through the LSTM model; The temperature of the key point is predicted according to the extracted time series features to obtain the predicted temperature of the key point at a specific time point.

5. The temperature prediction method of an automated computer room based on artificial intelligence according to claim 2 is characterized in that: The step of obtaining the three-dimensional scene temperature data at each three-dimensional coordinate in the three-dimensional space model according to the historical thermal image data specifically includes: Extracting the depth information of the data center from the 3D spatial model; Based on the depth information and historical thermal image data acquired by the infrared thermal imager at two or more different angles, three-dimensional scene temperature data on three-dimensional coordinates that can be directly captured by the infrared thermal imager is calculated; The interpolation method is used to estimate the temperature data of the three-dimensional scene at the three-dimensional coordinates that cannot be photographed; The calculation results are combined to obtain the three-dimensional scene temperature data at each three-dimensional coordinate in the three-dimensional space model.

6. The temperature prediction system of the automated computer room based on artificial intelligence is characterized by: include: A historical data acquisition unit, used to acquire historical key point temperature data and historical thermal image data of the data center, wherein the historical key point temperature data is acquired by a temperature sensor set at a key point position, and the historical thermal image data is acquired by an infrared thermal imager set at two or more different angles; A three-dimensional model building unit, used to build a three-dimensional space model of the data center, and align the historical key point temperature data and the historical thermal image data on the three-dimensional space model according to the key point positions; A weight model building unit is used to build a temperature contribution weight model, and to perform training based on aligned historical key point temperature data and historical thermal image data to obtain the temperature contribution weight of the key point temperature data to the thermal image data in terms of time sequence and three-dimensional space distance; A key point temperature prediction unit, used to predict the temperature of the key point according to the historical key point temperature data to obtain a key point predicted temperature; A three-dimensional temperature prediction unit, used for inputting the predicted temperature of the key point and the current thermal image data currently acquired into the temperature contribution weight model, and obtaining the three-dimensional scene temperature prediction data of the data center according to the temperature contribution weight; The input of the temperature contribution weight model is the coordinate-aligned key point temperature data and the three-dimensional scene temperature data, and the output is the temperature contribution weight of the key point temperature data to the thermal image data in time sequence and three-dimensional space distance. The weight model establishment unit is specifically used for: Grouping the historical key point temperature data and the three-dimensional scene temperature data obtained according to the historical thermal image data according to a time sequence; According to the data collection time interval, the three-dimensional scene temperature data at a later time point is used as a comparison object of the three-dimensional scene temperature data at a previous time point for training; For the key point temperature data collected by each temperature sensor, calculate the temperature contribution weight of the temperature value of the three-dimensional scene temperature data that changes over time in the three-dimensional coordinates from near to far in all directions outward with the key point position coordinate as the center; The step of training based on the aligned historical key point temperature data and historical thermal image data to obtain the temperature contribution weight of the key point temperature data to the thermal image data in time sequence and three-dimensional space distance specifically includes: When the temperature value of the three-dimensional scene temperature data at a certain three-dimensional coordinate is simultaneously affected by a plurality of different key point temperature data, the temperature contribution weight of each key point temperature data to the three-dimensional scene temperature data at the three-dimensional coordinate is calculated respectively; When performing temperature prediction, the input of the temperature contribution weight model is the currently acquired key point temperature data, the predicted key point predicted temperature, and the three-dimensional scene temperature data obtained based on the currently acquired current thermal image data, and the output is the three-dimensional scene temperature prediction data obtained according to the temperature contribution weight. The three-dimensional temperature prediction unit is specifically used for: Determining, according to the temperature contribution weight, the three-dimensional coordinates affected by the key point temperature data in the three-dimensional scene temperature data; Extract the temperature values ​​at the affected three-dimensional coordinates respectively; Calculate the predicted temperature values ​​at these three-dimensional coordinates according to the time interval between the predicted temperature of the key point and the current time point and the temperature contribution weight value at each three-dimensional coordinate at the corresponding time interval; The predicted temperature value at the affected three-dimensional coordinate is combined with other three-dimensional scene temperature data that is not affected by the key point temperature data to obtain the three-dimensional scene temperature prediction data; The calculation method of the temperature contribution weight includes: Extract the temperature values ​​of the three-dimensional scene temperature data on the three-dimensional coordinates from near to far in all directions outward with the key point position coordinates as the center; Calculate the difference between the temperature value of the three-dimensional scene temperature data at each three-dimensional coordinate and the key point temperature data at the key point position coordinate; Determine whether the temperature values ​​of the three-dimensional scene temperature data at the three-dimensional coordinates where the differences exist change in a direction that reduces the differences within a subsequent preset time threshold interval; If yes, then determine these three-dimensional coordinates as three-dimensional coordinates affected by the key point temperature data; For each three-dimensional coordinate affected by the key point temperature data, the temperature contribution weight of the key point temperature data to the three-dimensional scene temperature data is obtained according to the unit time gradient size of the temperature value changing in the direction that reduces the difference.

Images