Machine room environment detection method and device and electronic equipment

Abstract

The invention discloses a machine room environment detection method and device and electronic equipment. The method comprises the steps that multi-dimensional environment data of a plurality of collection points in a machine room are acquired, and the multi-dimensional environment data comprise temperature data and humidity data; a prediction model is adopted to analyze the multi-dimensional environment data, a prediction result is obtained, and the prediction result is used for reflecting the temperature and humidity environment state of the machine room; and determining the temperature and humidity environment state of the machine room corresponding to the prediction result. According to the invention, the technical problem that the overall temperature and humidity condition of the machine room is difficult to comprehensively control because the machine room temperature and humidity detection means adopted in the related technology emphasizes the temperature or humidity monitoring of a specific cabinet is solved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and in particular, to a method, device, and electronic device for detecting the computer room environment. Background Art

[0002] In environments such as data centers and computer rooms, temperature and humidity are important factors affecting the operating state of equipment. When the temperature and humidity deviate from the standard level, it may have an adverse impact on the equipment. The existing computer room temperature and humidity detection methods mainly focus on monitoring the temperature or humidity of a specific cabinet, and it is difficult to comprehensively control the overall temperature and humidity conditions of the computer room.

[0003] In view of the above problems, no effective solution has been proposed yet. Summary of the Invention

[0004] Embodiments of this application provide a method, device, and electronic device for detecting the computer room environment, so as to at least solve the technical problem that the existing computer room temperature and humidity detection methods mainly focus on monitoring the temperature or humidity of a specific cabinet, and it is difficult to comprehensively control the overall temperature and humidity conditions of the computer room.

[0005] According to one aspect of the embodiments of this application, a method for detecting the computer room environment is provided, including: obtaining multi-dimensional environment data of multiple collection points in the computer room, where the multi-dimensional environment data includes temperature data and humidity data; analyzing the multi-dimensional environment data by using a prediction model to obtain a prediction result, where the prediction result is used to reflect the temperature and humidity environment state of the computer room; and determining the temperature and humidity environment state of the computer room corresponding to the prediction result.

[0006] In some embodiments of this application, the prediction model is trained in the following manner: obtaining historical multi-dimensional environment data of the computer room and the true category corresponding to the historical multi-dimensional environment data, where the true category is used to describe the temperature and humidity environment state of the computer room; inputting the historical multi-dimensional environment data into an initial prediction model for training to obtain a predicted category; determining the loss value corresponding to the predicted category and the true category; determining the gradient of the model parameters of the initial prediction model corresponding to the loss value, and updating the model parameters according to the gradient; and stopping training the initial prediction model when the loss value meets a preset condition to obtain the prediction model.

[0007] In some embodiments of this application, the historical multi-dimensional environment data includes historical temperature data and historical humidity data. Before inputting the historical multi-dimensional environment data into the initial prediction model for training, it further includes: generating a temperature and humidity map based on the historical temperature data and the historical humidity data, where the abscissa of each temperature and humidity point in the temperature and humidity map represents temperature, and the ordinate of each temperature and humidity point represents humidity; and determining the true category corresponding to each temperature and humidity point in the temperature and humidity map.

[0008] In some embodiments of the present application, after generating a temperature-humidity map based on historical temperature data and historical humidity data, it further includes: obtaining first data corresponding to each pixel in the temperature-humidity map, where the first data includes the RGB color values of each pixel; determining second data corresponding to the first data, where the second data is used to represent the representation of the first data in the HSV color space, and the HSV color space includes a hue channel, a saturation channel, and a value channel; adjusting the target channel value corresponding to the first temperature-humidity point in the temperature-humidity map to obtain third data, where the target channel value includes the value of any one channel in the HSV color space; determining a first target temperature-humidity map corresponding to the third data, where the first target temperature-humidity map is used to represent the map obtained after adjusting the color values of each first temperature-humidity point in the temperature-humidity map.

[0009] In some embodiments of the present application, adjusting the target channel value corresponding to the first temperature-humidity point in the temperature-humidity map includes: determining the data corresponding to the first temperature-humidity point from the second data to obtain target data; obtaining the first target channel value corresponding to the target data; determining the offset corresponding to the target channel according to the first target channel value; adjusting the first target channel value according to the offset to obtain a second target channel value.

[0010] In some embodiments of the present application, after generating a temperature-humidity map based on historical temperature data and historical humidity data, it further includes: obtaining the coordinate information corresponding to the second temperature-humidity point in the temperature-humidity map; determining the rotation parameters corresponding to the temperature-humidity map, where the rotation parameters include the rotation center and the rotation angle; generating a first temperature-humidity map corresponding to the rotation parameters; adjusting the coordinate information according to the rotation parameters to obtain target coordinate information; projecting the target coordinate information onto the first temperature-humidity map to obtain a second target temperature-humidity map.

[0011] In some embodiments of the present application, the true category is determined in the following manner: when the temperature in the computer room is less than the first temperature threshold and the humidity in the computer room is less than the first humidity threshold, it is determined that the computer room corresponds to the first category label; when the temperature in the computer room is less than the second temperature threshold, it is determined that the computer room corresponds to the second category label, where the second temperature threshold is less than the first temperature threshold; when the temperature in the computer room is greater than the third temperature threshold and the humidity in the computer room is greater than the second humidity threshold, it is determined that the computer room corresponds to the third category label; when the temperature in the computer room is greater than the fourth temperature threshold, it is determined that the computer room corresponds to the fourth category label, where the fourth temperature is greater than the third temperature threshold.

[0012] According to another aspect of the embodiments of the present application, there is also provided a detection device for a computer room environment, including: an acquisition module, configured to acquire multi-dimensional environment data of multiple collection points in the computer room, where the multi-dimensional environment data includes temperature data and humidity data; a prediction module, configured to analyze the multi-dimensional environment data by using a prediction model to obtain a prediction result, where the prediction result is used to reflect the temperature and humidity environment state of the computer room; a determination module, configured to determine the temperature and humidity environment state of the computer room corresponding to the prediction result.

[0013] According to yet another aspect of the embodiments of the present application, there is also provided an electronic device, including: a memory and a processor, where the memory is used to store program instructions; the processor is connected to the memory and is configured to execute to implement the above-mentioned detection method for the computer room environment.

[0014] According to yet another aspect of the embodiments of the present application, there is also provided a non-volatile storage medium, where the non-volatile storage medium includes a stored computer program, and the device where the non-volatile storage medium is located executes the above-mentioned detection method for the computer room environment by running the computer program.

[0015] According to yet another aspect of the embodiments of the present application, there is also provided a computer program product, including computer instructions, where the computer instructions implement the above-mentioned detection method for the computer room environment when executed by a processor.

[0016] In the embodiments of the present application, by analyzing the multi-dimensional environment data of multiple collection points in the computer room by using a prediction model, the purpose of accurately reflecting the temperature and humidity environment state of the overall computer room is achieved, thereby realizing the technical effect of real-time monitoring of the computer room environment and ensuring that the equipment operates under the best temperature and humidity conditions, and further solving the technical problem that the existing computer room temperature and humidity detection means mainly focus on monitoring the temperature or humidity of a specific cabinet and it is difficult to comprehensively control the overall temperature and humidity condition of the computer room. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation to the present application. In the drawings:

[0018] Figure 1 is a hardware structure block diagram of a computer terminal for a detection method of a computer room environment according to an embodiment of the present application;

[0019] Figure 2 is a flowchart of a detection method of a computer room environment according to an embodiment of the present application;

[0020] Figure 3 is a schematic structural diagram of a prediction model for a detection method of a computer room environment according to an embodiment of the present application;

[0021] Figure 4 It is the prediction model training feature map a of a detection method for a computer room environment according to an embodiment of the present application;

[0022] Figure 5 It is the prediction model training feature map b of a detection method for a computer room environment according to an embodiment of the present application;

[0023] Figure 6 It is a schematic diagram of the prediction model training loss curve of a detection method for a computer room environment according to an embodiment of the present application;

[0024] Figure 7 It is a schematic diagram of the prediction model verification loss curve of a detection method for a computer room environment according to an embodiment of the present application;

[0025] Figure 8 It is the overall flowchart of a detection method for a computer room environment according to an embodiment of the present application;

[0026] Figure 9 It is a schematic structural diagram of a detection device for a computer room environment according to an embodiment of the present application. Detailed implementation manners

[0027] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0028] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not necessarily need to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the embodiments of the present application described herein can be implemented in an order other than those 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 may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0029] In order to better understand the embodiments of the present application, the technical terms involved in the embodiments of the present application are explained as follows:

[0030] Residual Network (ResNet): An innovative architecture in deep convolutional neural networks. By introducing residual blocks, ResNet allows features in the network to be directly passed from the input to the output, effectively alleviating the vanishing gradient or exploding gradient phenomenon during model training, enabling the network to be successfully trained to deeper levels while maintaining good performance and stability.

[0031] With the continuous increase in the installation capacity of communication computer rooms and the power density of equipment, the local hot spot problem of cabinet rows in the computer room has become increasingly prominent. Effectively monitoring the temperature and humidity conditions in the data center can ensure the stable operation of IT equipment in the data center in a relatively suitable environment. The monitoring methods for the computer room environment mainly rely on manual inspections, temperature or humidity sensors, and alarm systems. Among them, manual inspections are inefficient and costly, while temperature or humidity sensors can only respond passively and cannot predict the temperature and humidity status of the computer room in a timely manner. Moreover, the relevant monitoring methods mainly focus on using temperature or humidity sensors for local or single-index monitoring, making it difficult to comprehensively control the overall temperature and humidity field conditions of the computer room. Although the alarm system realizes timely warning, it is difficult to set reasonable alarm thresholds, and the alarm system cannot provide specific results of the temperature and humidity status, making it impossible to effectively judge the severity of the temperature and humidity status.

[0032] To solve the above technical problems, the embodiments of the present application provide corresponding solutions, which are described in detail below.

[0033] The embodiment of the method for detecting the computer room environment provided by the embodiments of the present application can be executed in a mobile terminal, a computer terminal, or a similar computing device. Figure 1 The hardware structure block diagram of a computer terminal for implementing the method for detecting the computer room environment is shown. As Figure 1 shown, the computer terminal 10 may include one or more processors (the processors may include, but are not limited to, processing devices such as microprocessor MCUs or programmable logic devices FPGAs, shown as 102a, 102b,..., 102n in the figure), a memory 104 for storing data, and a transmission module 106 for communication functions connected by wired and / or wireless networks. In addition, it may further include: a display, a keyboard, a cursor control device, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, and a BUS bus. Those of ordinary skill in the art can understand that Figure 1 the structure shown is only schematic and does not limit the structure of the above electronic device. For example, the computer terminal 10 may further include more or fewer components than Figure 1 shown, or have a different configuration from Figure 1 shown.

[0034] It should be noted that one or more of the above-mentioned processors and / or other data processing circuits can generally be referred to as "data processing circuits" herein. This data processing circuit can be embodied in software, hardware, firmware, or any combination thereof, in whole or in part. In addition, the data processing circuit can be a single independent processing module, or incorporated in whole or in part into any one of the other elements in the computer terminal 10. As involved in the embodiments of the present application, this data processing circuit is a kind of processor control (such as the selection of a variable resistance terminal path connected to an interface).

[0035] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage devices corresponding to the detection method of the computer room environment in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implements the above-mentioned detection method of the computer room environment. The memory 104 can include high-speed random access memory, and can also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memories. In some instances, the memory 104 can further include a memory remotely located relative to the processor, and these remote memories can be connected to the computer terminal 10 through a network. Examples of the above-mentioned network include but are not limited to the Internet, enterprise intranet, local area network, mobile communication network, and combinations thereof.

[0036] The transmission module 106 is used to receive or send data via a network. Specific examples of the above-mentioned network can include the wireless network provided by the communication provider of the computer terminal 10. In one instance, the transmission module 106 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station and thus communicate with the Internet. In one instance, the transmission module 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0037] The display can be, for example, a touch-screen liquid crystal display (LCD), which enables users to interact with the user interface of the computer terminal 10.

[0038] It should be noted here that in some alternative embodiments, the above-mentioned Figure 1 shown computer terminal can include hardware elements (including circuits), software elements (including computer code stored on a computer-readable medium), or a combination of both hardware elements and software elements. It should be pointed out that Figure 1 is only an example of a specific specific instance and is intended to show the types of components that can exist in the above-mentioned computer terminal.

[0039] Under the above operating environment, an embodiment of a method for detecting a computer room environment is provided in an embodiment of the present application. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0040] Figure 2 It is a flowchart of a method for detecting a computer room environment according to an embodiment of the present application. As Figure 2 shown, the method includes the following steps:

[0041] Step S202, obtain multi-dimensional environment data of multiple collection points in the computer room, where the multi-dimensional environment data includes temperature data and humidity data.

[0042] In the above step S202, the collection point refers to the position point of the temperature and humidity sensor set in the computer room, which is used to monitor the temperature and humidity environment at that point in real time. The distribution of the collection points covers the key areas of the computer room, such as server cabinets, ventilation openings, near air conditioners, etc. The number and layout of the collection points can be determined according to the size, structure of the computer room and the distribution of IT equipment.

[0043] In some embodiments of the present application, in order to ensure that the sensor network can fully cover the key areas of the computer room, the collection points can be determined in the following way: divide the computer room into multiple uniform grids, for example, divide it according to the rows, columns of the cabinets or functional areas. The size of each grid can be determined according to the size of the computer room and the air return characteristics; deploy temperature and humidity sensors in each grid, for example, deploy them at the center of the grid or key positions (such as near the equipment heat dissipation openings) to ensure that there is at least one collection point in each area of the computer room, so as to maintain the uniformity and comprehensiveness of monitoring even in a large computer room. In addition, considering that the temperature and humidity inside the computer room may vary with height, sensors can be deployed at the key heights (such as the bottom, middle and top of the cabinet) in each grid to form a hierarchical monitoring network for capturing the vertical temperature and humidity gradient inside the computer room.

[0044] Multi-dimensional environmental data includes, but is not limited to, temperature data and humidity data, and may also include environmental parameters such as air flow velocity and carbon dioxide concentration in the computer room, which are used to reflect the working environment of the computer room from multiple dimensions. In some embodiments of the present application, the temperature and humidity sensors transmit the collected data to the data processing and storage center in real time through the network. For example, the sensors can be configured to send data once every minute or every 5 minutes, and the adjustment is based on the environmental change frequency of the computer room and the requirement for real-time performance. Specifically, the temperature sensor measures the temperature at different positions in the computer room, converts the temperature into an electrical signal, and then converts it into a digital signal by an analog-to-digital converter (ADC), and finally sends it to the data processing center through the network; the humidity sensor measures the humidity at different positions in the computer room, and the data is also output in digital format.

[0045] Integrate the collected temperature data and humidity data to form a multi-dimensional environmental data set. For example, the data at different collection points can be aligned according to the time series to form a time-space matrix, where each row represents the temperature and humidity data of a collection point at different times, and each column represents the data of all collection points at a certain moment (for example, any specified moment). During the data integration process, data cleaning can also be performed to remove outliers and missing data to ensure the quality of the data. For example, reasonable temperature and humidity thresholds can be set to filter out abnormal data points caused by sensor failures or environmental interference.

[0046] Step S204, use a prediction model to analyze the multi-dimensional environmental data to obtain a prediction result, where the prediction result is used to reflect the temperature and humidity environmental state of the computer room.

[0047] In the above step S204, the prediction model includes a Residual Network (abbreviated as ResNet) model. Figure 3 It is a schematic diagram of the prediction model structure of a detection method for a computer room environment according to an embodiment of the present application. As Figure 3 shown, the present application can adopt the architecture of the ResNet-34 model, which can effectively alleviate the problem of gradient disappearance in deep networks and improve the learning and prediction performance of the model.

[0048] This model (34-layer residual, referring to a residual network with 34 layers of depth) includes the following structure:

[0049] (1) Input layer: Since the input data received by the ResNet model is usually an image, for example, the size of the image can be 224x224x3, that is, the input is a color image of 224x224 pixels, where 3 represents the three color channels of RGB. Therefore, before analyzing the above temperature and humidity data, preprocessing is required:

[0050] After collecting temperature and humidity data from multiple collection points in the computer room, the original temperature and humidity data are converted into an image format, that is, a temperature-humidity map is constructed. Among them, the color of each pixel represents the temperature and humidity combination at a specific collection point. For example, the temperature and humidity values can be mapped to the RGB or HSV color space, and different temperature and humidity states are represented by different colors. Specifically:

[0051] First, determine a color space, such as RGB (Red, Green, Blue) or HSV (Hue, Saturation, Value). The RGB space is based on the combination of the three colors red, green, and blue, while the HSV space focuses on the attributes of colors.

[0052] Then set the mapping rules: For the RGB space, the temperature data can be mapped to the value of the red channel, the humidity data can be mapped to the value of the green channel, and the blue channel can remain unchanged or be mapped to another environmental parameter, such as air flow speed, etc., or determined according to the combination ratio of temperature and humidity. For example, data with high temperature is mapped to a red value close to 255, and data with low temperature is mapped to a red value close to 0; data with high humidity is mapped to a green value close to 255, and data with low humidity is mapped to a green value close to 0.

[0053] For the HSV space, temperature can be mapped to the Hue (H) value, humidity is mapped to the Saturation (S) value, and the Value (V) can remain unchanged or be used to represent other parameters. For example, high-temperature data may be mapped to a high Hue value (such as yellow), and low-temperature data is mapped to a low Hue value (such as blue); high humidity is mapped to high saturation (bright colors), and low humidity is mapped to low saturation (lighter colors).

[0054] Data standardization: Before mapping, the temperature and humidity data can be standardized to ensure that they are mapped to the effective range of the color space. For example, if the temperature data range is from 0°C to 50°C and the humidity data range is from 0% to 100%, they can be mapped to the range from 0 to 255 respectively to match the channel value range of the RGB or HSV space.

[0055] Pixel value calculation: According to the above mapping rules, the temperature and humidity data of each collection point are converted into the values of color channels.

[0056] For example, for temperature t and humidity h, the following formula can be used to map to the red and green channels of the RGB space:

[0057] R = 255 * (t - T_min) / (T_max - T_min);

[0058] G = 255 * (h - H_min) / (H_max - H_min);

[0059] B = 0 # or remain unchanged.

[0060] Where T_min and T_max are the minimum and maximum values of the temperature data, and H_min and H_max are the minimum and maximum values of the humidity data.

[0061] For another example, for the HSV color space, the mapping process can be:

[0062] H = 255 * (t - T_min) / (T_max - T_min);

[0063] S = 255 * (h - H_min) / (H_max - H_min);

[0064] V = 255 # or remain unchanged.

[0065] Where T_min and T_max are the minimum and maximum values of the temperature data, and H_min and H_max are the minimum and maximum values of the humidity data.

[0066] Construct an image: Use these color values to construct an image. Each acquisition point corresponds to a pixel in the image. Assign the calculated RGB or HSV values to the corresponding pixels, and finally form an image reflecting the temperature and humidity distribution in the computer room.

[0067] To ensure the accuracy of the input data, it is also necessary to clean the above input data. For example, use the IQR method to remove outliers:

[0068] Sort all the temperature and humidity data in the computer room separately. For the temperature data and humidity data, sort them independently. This is done to calculate the quartiles of temperature and humidity respectively. In some embodiments of the present application, the quartiles can be calculated in the following manner:

[0069] 1) Calculate the first quartile (Q1) and the third quartile (Q3) of the temperature data, that is, find the values at the 25% and 75% positions in all the collected temperature data. In actual calculation, if the number of data points is not a multiple of 4, Q1 and Q3 may be between two data points, and interpolation calculation is required at this time. For the temperature data, denoted as T, then Q1 and Q3 can be calculated using np.percentile(T, 25) and np.percentile(T, 75): np.percentile(T, 25) will return a value such that 25% of the values in the data set are less than or equal to it, and the remaining 75% of the values are greater than it, and this value is Q1; np.percentile(T, 75) will return another value such that 75% of the values in the data set are less than or equal to it, and the remaining 25% of the values are greater than it, and this value is Q3.

[0070] 2) Similarly, for the humidity data H, calculate np.percentile(H, 25) and np.percentile(H, 75) respectively to obtain Q1 and Q3 of the humidity data.

[0071] 3) Calculate the IQR: Next, calculate the interquartile range (IQR) of the temperature and humidity data. The formula is: IQR = Q3 - Q1. This IQR can be used to indicate the fluctuation range of the middle 50% of the data points in the data set, thus providing a benchmark for the detection of outliers.

[0072] 4) Define the outlier boundaries: Use the IQR to define the lower and upper limits of the outliers for the temperature and humidity data. The calculation formulas for the lower limit (LowerBound) and the upper limit (Upper Bound) are Lower Bound = Q1 - 1.5 * IQR and Upper Bound = Q3 + 1.5 * IQR respectively. The selection of this range is based on the 1.5 - times rule of the IQR, which is an empirical rule used to identify possible outliers. Any value below the Lower Bound or above the Upper Bound will be marked as an outlier.

[0073] Through the above calculation process, the boundaries of the outliers can be automatically determined. At the same time, since it is calculated based on the actual environmental data in the computer room, the calculated threshold is more in line with the actual state of the computer room.

[0074] (2) Conv1: It refers to the first convolutional layer of the model, which serves as the starting part (stem) of the network. The size of its convolutional kernel is 7x7 pixels (7x7 conv), and the depth or number of channels of the output feature map is 64. That is, the input RGB image (3 channels) will be converted into 64 feature maps after the convolutional operation. Each feature map represents the features of a different aspect of the input data. The stride of Conv1 is 2 ( / 2). When the size of the input image is 224x224x3 (i.e., width 224 pixels, height 224 pixels, and 3 color channels), after being processed by the conv1 layer (using a 7x7 convolutional kernel with a stride of 2), the size of the output feature map becomes 112x112x64. The role of the Conv1 layer is to initially extract and transform the features of the input image (such as the temperature and humidity map). The convolutional operation can capture the local features of the image, while the stride and padding settings will affect the size of the feature map and the processing of edge information.

[0075] (3) Pool: Immediately following Conv1 is the max pooling layer, which is used to reduce the spatial dimension of the feature map, reduce the computational amount, and at the same time abstract the local features of the input to enhance the translational invariance of the model. Figure 3 The size of the pooling window in it can be 3x3, and the stride is 2. The max pooling layer downsamples the extracted feature map to extract the key features of the temperature and humidity map, which helps the model understand the macroscopic patterns of the temperature and humidity distribution, such as the overall trends of temperature and humidity, rather than just local details.

[0076] (4) Conv2, Conv3, Conv4, Conv5: Represent the main part of ResNet, that is, the convolutional layers composed of multiple residual blocks. The structure of the residual blocks allows the model to maintain stable propagation of gradients during the training of deep networks, making it possible to train deep networks. Among them, the Conv2 layer has a convolutional kernel size of 3x3 pixels (3x3 conv), and the depth or number of channels of the output feature map is 64. When the size of the input image is 112x112x64, after being processed by the Conv2 layer, the size of the output feature map becomes 56x56x64; the Conv3 layer has a convolutional kernel size of 3x3 pixels (3x3 conv), and the depth or number of channels of the output feature map is 128, with a stride of 2 ( / 2). When the size of the input image is 56x56x64, after being processed by the Conv3 layer, the size of the output feature map becomes 28x28x128; the Conv4 layer has a convolutional kernel size of 3x3 pixels (3x3 conv), and the depth or number of channels of the output feature map is 256, with a stride of 2 ( / 2). When the size of the input image is 28x28x128, after being processed by the Conv4 layer, the size of the output feature map becomes 14x14x256; the Conv5 layer has a convolutional kernel size of 3x3 pixels (3x3 conv), and the depth or number of channels of the output feature map is 512, with a stride of 2 ( / 2). When the size of the input image is 14x14x256, after being processed by the Conv5 layer, the size of the output feature map becomes 7x7x512.

[0077] In some embodiments of the present application, feature extraction is performed on the input temperature and humidity image through multiple convolutional layers and residual blocks. The convolutional layer can capture local features of temperature and humidity in the image, such as edges, textures, and local change patterns of temperature and humidity. The design of the residual blocks enables the model to handle deep network structures, avoid the problem of gradient disappearance, and be able to learn more complex temperature and humidity distribution patterns. For example, through the residual blocks, the model can learn the temperature and humidity differences between different regions, as well as the relationship between temperature and humidity changes and time, so as to extract comprehensive features of the computer room environment.

[0078] (5) GAP (avg pool): Represents the global average pooling layer (Global Average Pooling). The role of the average pooling layer is to compress the dimension of the feature map to 1 (i.e., 1x1x512), while retaining the channel information of the feature map. In the final stage of feature extraction, the feature map can be converted into a vector of a fixed size, which is convenient for subsequent fully connected layer processing. After multiple layers of feature extraction and downsampling, the model compresses the dimensions of all feature maps into a vector of a fixed size through the global average pooling layer, which can capture the overall state of the temperature and humidity environment in the computer room. Subsequently, the vector is sent to the fully connected layer for feature fusion to form a dense vector containing all key information, ready for classification.

[0079] (6) FC (fc 1000): Represents the fully connected layer, i.e., the logit layer. In ResNet, the fully connected layer is used to further transform the feature vector after global average pooling into a prediction vector. The output size of the fully connected layer is the same as the number of classes in the classification task. For example, in the ImageNet task, the output size is 1000, corresponding to 1000 different classifications. Through this layer, the model can finally output a class label representing the temperature and humidity environment state of the computer room. For example, if the output prediction probabilities are {"low humidity in winter": 0.1, "overcooling": 0.05, "high temperature in computer room": 0.01, "local hot spot": 0.75, "high humidity in summer": 0.05, "normal": 0.04}, then the prediction result is "local hot spot", indicating that the model believes that the current computer room environment state is most likely to have a local overheating phenomenon.

[0080] To accurately predict the temperature and humidity environment state of the computer room and provide data support for the early warning and maintenance of the computer room environment, the prediction model can be trained in the following way: Obtain the historical multi-dimensional environment data of the computer room and the corresponding true classes of the historical multi-dimensional environment data, where the true classes are used to describe the temperature and humidity environment state of the computer room; Input the historical multi-dimensional environment data into the initial prediction model for training to obtain the predicted classes; Determine the loss value corresponding to the predicted classes and the true classes; Determine the gradient of the model parameters of the initial prediction model corresponding to the loss value and update the model parameters according to the gradient; Stop training the initial prediction model when the loss value meets the preset conditions to obtain the prediction model.

[0081] The historical multi-dimensional environment data refers to the computer room environment data at different positions collected at different time points, including but not limited to multiple dimensions such as temperature, humidity, air pressure, wind speed, and air quality. The true classes of the historical multi-dimensional environment data refer to the classification labels of the computer room environment state marked by experts or historical records, which are used to evaluate the accuracy of the model prediction, guide the training process of the model, and ensure that the prediction results match the actual situation.

[0082] In some embodiments of the present application, the historical multi-dimensional environment data includes historical temperature data and historical humidity data. Before inputting the historical multi-dimensional environment data into the initial prediction model for training, the following steps can also be performed: Generate a temperature-humidity map based on the historical temperature data and historical humidity data, where the abscissa of each temperature-humidity point in the temperature-humidity map represents the temperature and the ordinate represents the humidity; Determine the true class corresponding to each temperature-humidity point in the temperature-humidity map.

[0083] Historical temperature data refers to the sequence of temperature data recorded at each collection point over a period of time. Historical humidity data refers to the sequence of humidity data recorded at each collection point over a period of time. A temperature-humidity graph refers to an image used to display the historical temperature and humidity data of multiple collection points, where each point represents the temperature value and humidity value of a specific collection point at a certain time point. In some embodiments of the present application, the historical temperature and humidity data of multiple collection points can be sorted according to the time sequence and the location of the collection points to form a two-dimensional data set containing temperature and humidity. With temperature as the abscissa and humidity as the ordinate, the data points of different collection points are plotted on the temperature-humidity graph, and each point represents the temperature and humidity state of a collection point at a certain time point.

[0084] For the true category corresponding to each temperature-humidity point in the temperature-humidity graph, threshold rules can be set for different temperature and humidity states according to the operation requirements of the computer room and the characteristics of the equipment, and the computer room environment state can be divided into multiple categories, such as "low humidity in winter", "over-cooling", "high temperature in the computer room", "local hot spot", etc. Traverse each point on the temperature-humidity graph, and assign a true category to each point according to the temperature and humidity values of each collection point and the preset threshold rules. For example, for a point with a temperature lower than 15°C and a humidity lower than 40%RH, it is marked as the "low humidity in winter" category.

[0085] In order to be able to finely divide the temperature and humidity environment state of the computer room, the true category of the historical multi-dimensional environment data can be determined in the following way: when the temperature of the computer room is less than the first temperature threshold and the humidity of the computer room is less than the first humidity threshold, determine that the computer room corresponds to the first category label; when the temperature of the computer room is less than the second temperature threshold, determine that the computer room corresponds to the second category label, where the second temperature threshold is less than the first temperature threshold; when the temperature of the computer room is greater than the third temperature threshold and the humidity of the computer room is greater than the second humidity threshold, determine that the computer room corresponds to the third category label; when the temperature of the computer room is greater than the fourth temperature threshold, determine that the computer room corresponds to the fourth category label, where the fourth temperature is greater than the third temperature threshold.

[0086] Through the above steps, the historical environment data points can be assigned to different category labels according to the preset temperature and humidity thresholds. The category labels can be further used to train a model (such as ResNet) to help the model learn the characteristics of each temperature and humidity state, and then accurately predict the temperature and humidity environment state where the new environment data points are located. The setting of the thresholds can be based on the specific environmental conditions of the computer room, the operation requirements of the equipment, and maintenance experience to ensure the rationality and effectiveness of the category labels. For example, a time series model can be trained using historical data to predict the mean and standard deviation of temperature and humidity at different seasons and time points, and the normal range can be set as the mean ± a certain multiple of the standard deviation.

[0087] By adjusting the color values of the temperature and humidity map, the recognition ability of the model can be enhanced. Specifically: Obtain the first data corresponding to each pixel in the temperature and humidity map, where the first data includes the RGB color values of each pixel; Determine the second data corresponding to the first data, where the second data is used to represent the representation of the first data in the HSV color space, and the HSV color space includes a hue channel, a saturation channel, and a value channel; Adjust the target channel value corresponding to the first temperature and humidity point in the temperature and humidity map to obtain the third data, where the target channel value includes the value of any one channel in the HSV color space; Determine the first target temperature and humidity map corresponding to the third data, where the first target temperature and humidity map is used to represent the map obtained by adjusting the color values of each first temperature and humidity point in the temperature and humidity map.

[0088] The first data refers to the RGB color values of each pixel in the temperature and humidity map, which are the original image data. The second data refers to the representation of the first data after being converted to the HSV color space, including the values of the three channels of hue, saturation, and value. In some embodiments of the present application, a color conversion function (such as matplotlib.colors.rgb_to_hsv) can be used to convert the first data (RGB color values) into the second data in the HSV color space, including the values of the three channels of hue, saturation, and value.

[0089] The target channel value refers to the channel value that needs to be adjusted in the HSV color space, which can be any one of hue, saturation, or value. By adjusting the target channel value, the visual characteristics of the color points in the temperature and humidity map can be changed, helping the model to better identify and classify the temperature and humidity points. In some embodiments of the present application, the distribution frequency of the temperature and humidity points in the HSV color space can be analyzed to determine a target channel that needs to be adjusted, such as the saturation channel. For the target channel (such as saturation), increase the dynamic range of its value so that the color points in different temperature and humidity states have higher contrast in the HSV color space, facilitating the model to distinguish.

[0090] In some embodiments of the present application, the target channel can be determined by the following method:

[0091] (1) Analyze the temperature and humidity map to identify the characteristics of pixel points in the HSV color space under different temperature and humidity states. For example, the distribution histograms of pixel points in the three channels of HSV under each temperature and humidity state can be plotted to observe the distribution patterns under different states, and the three channels of HSV can be adjusted respectively through experiments to observe the impact on the ability to distinguish temperature and humidity states, such as whether increasing saturation or value can more clearly distinguish the colors in the case of low humidity in winter and high humidity in summer.

[0092] (2) According to the results of the temperature and humidity distribution analysis, select the HSV channel that has the greatest impact on the discrimination of temperature and humidity states as the target channel. For example, the entropy of the three HSV channels can be calculated. The larger the entropy value, the greater the amount of information, indicating that this channel makes a greater contribution to distinguishing the temperature and humidity states. Select the channel with the largest entropy value as the target channel. It is also possible to use statistical methods (such as the Pearson correlation coefficient) to analyze the correlation between the three HSV channels and the temperature and humidity states (such as low humidity in winter, high temperature in the computer room, etc.), and select the channel with the highest correlation (or the absolute value closest to 1).

[0093] The third data refers to the HSV color space data after adjusting the target channel values, that is, the adjusted hue, saturation, and brightness values. The first target temperature and humidity map is obtained by adjusting the target channel values of each first temperature and humidity point in the temperature and humidity map, and then converting the data in the HSV color space back to the RGB color space. Since the first target temperature and humidity map undergoes color enhancement processing, the contrast between temperature and humidity points is increased, enabling the model to more accurately distinguish temperature and humidity states. In some embodiments of the present application, the adjusted HSV color space data (third data) can be inversely converted back to the RGB color space to obtain the RGB data after color adjustment, and the adjusted RGB data can be redrawn as an image to generate the first target temperature and humidity map.

[0094] To more precisely control the color change of the temperature and humidity map, the target channel value corresponding to the first temperature and humidity point in the temperature and humidity map can be adjusted in the following manner: Determine the data corresponding to the first temperature and humidity point from the second data to obtain the target data; Obtain the first target channel value corresponding to the target data; Determine the offset corresponding to the target channel based on the first target channel value; Adjust the first target channel value according to the offset to obtain the second target channel value.

[0095] The target data refers to the HSV values corresponding to specific pixel points (i.e., the first temperature and humidity points) in the HSV color space. Through the target data, the values of the corresponding channels can be precisely adjusted in the HSV space. In some embodiments of the present application, in the HSV space, specific temperature and humidity points (the first temperature and humidity points) can be found through coordinate positioning, and the HSV values of these points can be extracted from the second data (HSV color space data) as the target data.

[0096] The first target channel value refers to the current value of a selected channel of the target data in the HSV color space, which is used to determine the specific color attribute to be adjusted. For example, if the target channel is saturation, then the first target channel value is the original saturation value of this pixel point.

[0097] The offset refers to a numerical value used to adjust the first target channel value, which can be a positive or negative number and is used to increase or decrease the value of the target channel. In some embodiments of the present application, the determination of the offset can be based on a preset rule or experimental results, or a machine learning method can be used for prediction. For example, the preset rule can be that if the first target channel value (such as saturation) is within a preset interval, a fixed offset a is defined, and if it exceeds the preset interval, another offset b is defined to ensure an obvious color change.

[0098] The second target channel value refers to the value of the target channel in the adjusted HSV color space, which is the result obtained after applying the offset. In some embodiments of the present application, after the offset is determined, an addition or subtraction operation can be performed on the first target channel value to obtain the second target channel value. For example, if the offset is ΔS and the first target channel value is S, a non-linear function f(S, ΔS) is used to adjust the first target channel value, and its calculation formula is:

[0099] F(S, ΔS) = S * exp(ΔS)

[0100] Where exp(ΔS) is an exponential function that can have an exponential impact on the input value ΔS. When ΔS is a positive number, exp(ΔS) will be greater than 1, meaning that the first target channel value S will be amplified; when ΔS is a negative number, exp(ΔS) will be less than 1, meaning that the first target channel value S will be reduced.

[0101] Figure 4 It is the prediction model training feature map a of a detection method for a computer room environment according to an embodiment of the present application, as Figure 4As shown (this figure is a schematic diagram), where the abscissa of the Original Image (the left figure, corresponding to the original temperature and humidity map) is humidity, showing the humidity values at different positions in the computer room, and the ordinate is temperature, showing the temperature values at different positions in the computer room. The data points in the Original Image are represented by points of different colors in the figure to record the temperature and humidity in the computer room. Each color represents a different environmental state category (such as low humidity in winter, high temperature in the computer room, etc.). The distribution of temperature and humidity points in the original image reflects the actual environmental conditions in the computer room, including the distribution of cold and hot channels, the natural fluctuations of temperature and humidity, etc.; the abscissa of the Transformed Image (the right figure, corresponding to the first target temperature and humidity map) is humidity, which is the same as the original image, and the ordinate is temperature, which is the same as the original image. The colors of the temperature and humidity points in the Original Image are modified to enhance the generalization ability of the model. The purpose of color modification is to eliminate the possible interference to the model caused by a single color and ensure that the model pays more attention to the distribution characteristics of temperature and humidity points. In the Original Image, the distribution pattern of data points remains the same as the original image, but the colors are adjusted so that points with the same distribution pattern present different visual effects in different images, which helps the model learn the distribution pattern of temperature and humidity points without being affected by specific colors, thereby improving the adaptability and prediction accuracy of the model to new data. By adjusting the colors, the spatial relationship of temperature and humidity points is not changed, but the model is made to focus on learning the distribution pattern, so that it can still maintain high prediction accuracy and stability when facing variable environmental data.

[0102] To enhance the generalization ability of the model, after generating the temperature and humidity map based on historical temperature data and historical humidity data, the following steps can also be performed: obtain the coordinate information corresponding to the second temperature and humidity point in the temperature and humidity map; determine the rotation parameters corresponding to the temperature and humidity map, where the rotation parameters include the rotation center and the rotation angle; generate the first temperature and humidity map corresponding to the rotation parameters; adjust the coordinate information according to the rotation parameters to obtain the target coordinate information; project the target coordinate information onto the first temperature and humidity map to obtain the second target temperature and humidity map.

[0103] The rotation parameters include the rotation center and the rotation angle, which are used to define how to perform rotation processing on the image in the HSV color space. In some embodiments of the present application, the rotation center can be the center point of the image or the position of a specific temperature and humidity point, and the rotation angle can be a randomly generated angle or an angle set based on the temperature and humidity distribution law. For example, an angle range (such as 0° to 30°) can be set, a random angle can be generated as the rotation angle, and the center of the image can be selected as the rotation center. In addition, the distribution trend of the temperature and humidity points in the temperature and humidity map can be analyzed. If the temperature and humidity points tend to be distributed in a certain direction (for example, the humidity decreases as the temperature increases), an angle opposite to this trend can be set for rotation to balance the data distribution.

[0104] The first temperature and humidity map is a map obtained by rotating the original temperature and humidity map according to the rotation parameters.

[0105] The target coordinate information refers to the new coordinate information of the second temperature and humidity point after being adjusted by the rotation parameters, which is used to ensure the correct position of the temperature and humidity point after rotation and avoid data distortion caused by image rotation. In some embodiments of the present application, coordinate transformation formulas, such as translation and rotation matrices, can be used to transform the original coordinate information of the second temperature and humidity point according to the rotation angle and the rotation center to obtain the rotated coordinate information as the target coordinate information.

[0106] The second target temperature and humidity map is a processed image obtained by adjusting the coordinate information and projecting it onto the first temperature and humidity map. It has enhanced visual features and diversity and can be provided to the ResNet model as an optimized input, enabling the model to learn the relative relationship between the temperature and humidity points and the coordinate axes, rather than the relative positions of the temperature and humidity points in the map. In some embodiments of the present application, the adjusted target coordinate information is mapped onto the rotated first temperature and humidity map to generate a new temperature and humidity point distribution map, that is, the second target temperature and humidity map.

[0107] Figure 5 It is the prediction model training feature map b of a detection method for the computer room environment according to an embodiment of the present application, as Figure 5 shown (this figure is a schematic diagram), where, Original Image (the left figure, corresponding to the original temperature and humidity map) and Figure 3is the same as the Original Image in , which will not be elaborated here. The Transformed Image (the right figure, corresponding to the second target temperature and humidity map obtained by randomly rotating 0-30°) has the same data points as the original temperature and humidity map, but their positions have changed due to rotation. Although the positions of the data points have changed, the transformed image retains the basic characteristics of the temperature and humidity distribution pattern in the original image. Through rotation, the same data point distribution pattern is presented at different angles. The transformed images generated by random rotation increase the diversity of the training dataset, enabling the model to learn the performance of the temperature and humidity distribution pattern at different angles, that is, to learn the relative relationship between the temperature and humidity points and the coordinate axes, rather than the relative positions of the temperature and humidity points in the figure.

[0108] In some embodiments of the present application, the original temperature and humidity map, the first target temperature and humidity map, and the second target temperature and humidity map obtained in the above steps can be used as historical multi-dimensional environmental data for the training of the prediction model at the same time. Specifically:

[0109] (1) Data preprocessing: Convert the historical multi-dimensional environmental data (original temperature and humidity map, first target temperature and humidity map, second target temperature and humidity map) into a format suitable for input to the ResNet model, such as image data after standardization and normalization processing, to ensure that all images have the same size and color space. In addition, the preprocessed and enhanced image data can be sliced into a training set, a validation set, and a test set, and the ratio can be 6:2:2. The specific preprocessing method can refer to the preprocessing process corresponding to the input layer above, which will not be elaborated here. Figure 3 The preprocessing process corresponding to the input layer above will not be elaborated here.

[0110] (2) Model input: Use the preprocessed image data and the corresponding label data (true category) as inputs, and input them into the convolutional layer in the ResNet model (such as the model shown above) for feature extraction. The ResNet model extracts complex features in the input image through multiple convolutional layers, pooling layers, and residual blocks. Figure 3 The ResNet model extracts complex features in the input image through multiple convolutional layers, pooling layers, and residual blocks.

[0111] (3) Feature extraction and classification: The main part of the ResNet model (multiple residual blocks) performs feature extraction at multiple levels, and finally converts the extracted features into the classification probability of the computer room environment state through the fully connected layer and the Softmax layer, and outputs the predicted category. The predicted category can be defined as: 0 (low humidity in winter), 1 (over-cooling), 2 (high temperature in the computer room), 3 (local hot spot), 4 (high humidity in summer), 5 (normal). Specifically, the convolutional layer can detect and learn the local features in the image, and the combination of these local features constitutes the global pattern of the image. Even if the data points are discrete in the image, the model can understand and identify the environmental information carried by these points in the following ways:

[0112] 1) Local feature detection: The convolutional layer can operate on the image through a sliding window to detect and extract local features of temperature and humidity data points, such as edges, textures, and local color changes. For example, adjacent points with low temperature and high humidity may exhibit a specific pattern of local features.

[0113] 2) Global pattern understanding: The deep structure of the model, including multiple residual blocks, can combine these local features to form a more advanced global pattern understanding ability. For example, if data points with high humidity and low temperature in the image cluster into a specific pattern, the model can recognize this pattern and associate it with the environmental state of low humidity in winter.

[0114] 3) Spatial relationship analysis: When analyzing spatial characteristics such as the adjacent relationship, aggregation degree, and distribution shape of points, the model can understand the laws of temperature and humidity changes in the computer room. This ability enables the model to not only identify the information of individual points but also gain insights into the spatial relationships between points, such as the formation of hot and cold channels, temperature gradients, and humidity distribution patterns.

[0115] 4) Feature fusion and classification: After feature extraction, feature fusion and further feature abstraction are performed through the fully connected layer. Finally, the fused features are converted into a probability distribution of various environmental states through the Softmax layer. The predicted category output is the environmental state corresponding to the highest probability, such as "low humidity in winter" and "high temperature in the computer room".

[0116] (4) Calculate the loss: Compare the predicted category output by the ResNet model with the true category, and calculate the loss value using the defined loss function (such as Cross-Entropy Loss).

[0117] Figure 6 and Figure 7 are respectively the schematic diagram of the training loss curve and the validation loss curve of a detection method for a computer room environment according to an embodiment of the present application (the number of test set pictures is 1440), Figure 6 showing the change curve of the loss value during the training process of the ResNet model, describing the evolution trend of the training loss (TrainingLoss) over time during the model training iteration process. Among them, the abscissa is Step (training step), the ordinate is Loss (loss value), and the curve represents the change trend of the training loss. Figure 6 The curve in

[0118] Figure 7 The figure shows the change in the loss value of the ResNet model during the validation process. The generalization ability of the model and overfitting avoidance are evaluated through the validation set data. Here, the abscissa is Step (training step), the ordinate is Loss (loss value), and the curve represents the change trend of the validation loss. Figure 7 The curve in it reveals the change in the prediction ability of the model for the validation data during the training process. The decrease in the validation loss value indicates that the model not only performs well on the training set but also maintains a high classification accuracy on the unseen validation data.

[0119] (5) Backpropagation: Based on the calculated loss value, use the backpropagation algorithm to calculate the gradients of the model parameters (the rate of change of the model parameters with respect to the loss value, which reflects the direction and magnitude of parameter adjustment). The backpropagation algorithm propagates the partial derivatives of the loss function with respect to the model parameters in the opposite direction of the network. The gradient of each layer is determined by the gradient of the next layer and the output of this layer, and finally, the parameter gradients of all layers are obtained (such as the learning rate, the number of training epochs, etc.).

[0120] (6) Update parameters using gradient descent: According to the calculated gradients, use the gradient descent method or other optimization algorithms (such as Stochastic Gradient Descent SGD, Momentum Gradient Descent, Adaptive Learning Rate Algorithm Adam, etc.) to update the model parameters. The goal of the update is to adjust the parameters in the opposite direction of the gradient to minimize the loss function, that is, to improve the prediction accuracy of the model.

[0121] (7) Determine whether the preset conditions for stopping training are met: The preset conditions refer to the conditions used to decide when to stop training, including the loss value reaching a certain threshold, the number of training epochs reaching a predetermined value, etc. When the preset conditions are met, stop training to obtain the prediction model.

[0122] Step S206: Determine the temperature and humidity environmental state of the computer room corresponding to the prediction result.

[0123] In the above step S206, the prediction result refers to the class label output after the ResNet model classifies and predicts the temperature and humidity image, which reflects the model's judgment on the current temperature and humidity environmental state. In some embodiments of the present application, the prediction result can be a numerical label, and a pre-defined mapping table can be used. This table maps the digital label to the environmental state description one by one. For example, {0: low humidity in winter, 1: overcooling, 2: high temperature in the computer room, 3: local hot spot, 4: high humidity in summer, 5: normal}. According to the class label output by the model, look up the corresponding environmental state description in the mapping table to determine the temperature and humidity environmental state of the computer room.

[0124] Through the above steps S202 to S206, by using a prediction model to analyze the multi-dimensional environmental data at multiple collection points in the computer room, the purpose of accurately reflecting the temperature and humidity environmental state of the overall computer room is achieved, thereby realizing the technical effect of real-time monitoring of the computer room environment and ensuring that the equipment operates under the best temperature and humidity conditions. Furthermore, it solves the technical problem that the temperature and humidity detection means adopted in the related technology mainly focus on the temperature or humidity monitoring of a specific cabinet, and it is difficult to comprehensively control the overall temperature and humidity condition of the computer room.

[0125] Figure 8 It is the overall flowchart of a method for detecting the computer room environment according to an embodiment of the present application. As Figure 8 shown, in some embodiments of the present application, the method for detecting the computer room environment may include the following steps:

[0126] S802: Obtain temperature and humidity sensor data. Collect the temperature and humidity data at each point in the computer room through Internet of Things devices (such as temperature and humidity sensors), interact with the communication interface of the sensors, read real-time or historical data from the sensors, and store this data in a suitable data structure, such as a data table or an array, for subsequent processing. Specifically, the serial communication library in Python such as pySerial, or the network communication library such as socket, can be used to communicate with the sensors, read the data and store it.

[0127] S804: Data preprocessing. It includes operations such as data cleaning, standardization, and normalization. For example, clean the collected temperature and humidity data to remove invalid or abnormal records; standardize the data, such as using Z-score standardization to make the data have zero mean and unit variance, or use normalization to scale the data to the range of 0-1 to eliminate the influence of dimension and improve the model training efficiency. In addition, data interpolation can also be performed to fill in missing values, or feature engineering can be carried out to create new feature variables, such as the temperature and humidity difference, etc.

[0128] S806: Dataset splitting. It includes splitting into three parts: a training set, a validation set, and a test set. First, randomly shuffle the preprocessed data, and then split it into a training set, a validation set, and a test set according to a predetermined ratio (such as 6:2:2).

[0129] S8061: Perform model training on the training set segmented by S806. Use the segmented training set data to train the ResNet model. Specifically, input the training image data and the corresponding class labels into the model, and adjust the model parameters through backpropagation and optimizers (such as SGD, Adam, etc.) to minimize the loss function (such as cross-entropy loss). The training process can include multiple epochs, and each epoch traverses the entire training set once. In each epoch, the model will learn the features to identify the environmental state patterns from the temperature and humidity maps, and gradually improve the classification accuracy by adjusting the weights and biases.

[0130] S8062: Perform model validation on the validation set segmented by S806. Use the validation set data to evaluate the generalization ability of the model on unseen data. At the end of each epoch, use the validation set data for prediction and calculate the loss value between the prediction result and the actual label to monitor the overfitting situation of the model. If the validation loss starts to increase, it indicates that the model may be overfitting on the training set, and measures such as early stopping of training, regularization, etc. need to be taken to prevent overfitting.

[0131] S8063: Perform model testing on the test set segmented by S806. After the model training and validation are completed, use independent test set data to evaluate the final performance of the model. The test set data has not been used in the training and validation processes, so it can provide an unbiased performance evaluation. By predicting the environmental state of the test set data with the model and comparing it with the actual label, calculate metrics such as accuracy, recall, F1 score, etc. to determine the classification performance of the model.

[0132] S808: Obtain the optimal model. Determine the optimal model according to the performance on the validation set. For example, select the model with the lowest validation loss as the optimal model. The model checkpoint mechanism can be used to save the model parameters when the validation loss is the lowest, so that the optimal model can be used for inference after the training is completed.

[0133] S810: Use the optimal model for result inference. Use the optimal model obtained by S808 for real-time or offline environmental state classification. New temperature and humidity map data can be input into the model, and the model will output an environmental state classification label based on the learned feature patterns. This label can be used to monitor the computer room environment, timely detect potential anomalies such as local hotspots, overcooling, etc., and help the operation and maintenance personnel take measures to ensure the stable operation of the computer room.

[0134] Figure 9 is a structural diagram of a detection device for a computer room environment according to an embodiment of the present application, as Figure 9 shown, the device includes:

[0135] An acquisition module 902, configured to acquire multi-dimensional environmental data of multiple collection points in a computer room, where the multi-dimensional environmental data includes temperature data and humidity data;

[0136] A prediction module 904, configured to analyze the multi-dimensional environmental data by using a prediction model to obtain a prediction result, where the prediction result is used to reflect the temperature and humidity environmental state of the computer room;

[0137] A determination module 906, configured to determine the temperature and humidity environmental state of the computer room corresponding to the prediction result.

[0138] It should be noted that Figure 9 the detection device for the computer room environment shown is used to execute Figure 2 the detection method for the computer room environment shown, so Figure 2 the relevant explanations in the detection method for the computer room environment in Figure 9 also apply to the detection device for the computer room environment shown, which will not be elaborated here.

[0139] An embodiment of the present application further provides an electronic device, which includes a memory and a processor. The memory is used to store program instructions; the processor is connected to the memory and is configured to execute the steps of implementing the detection method for the computer room environment in various embodiments of the present application.

[0140] For example, the processor executes the following functions by executing the program instructions stored in the memory:

[0141] Acquire multi-dimensional environmental data of multiple collection points in a computer room, where the multi-dimensional environmental data includes temperature data and humidity data; analyze the multi-dimensional environmental data by using a prediction model to obtain a prediction result, where the prediction result is used to reflect the temperature and humidity environmental state of the computer room; determine the temperature and humidity environmental state of the computer room corresponding to the prediction result.

[0142] An embodiment of the present application further provides a non-volatile storage medium, which includes a stored computer program. The device where the non-volatile storage medium is located executes the steps of the detection method for the computer room environment in various embodiments of the present application by running the computer program.

[0143] An embodiment of the present application further provides a computer program product, including computer instructions, where the computer instructions, when executed by a processor, implement the steps of the detection method for the computer room environment in various embodiments of the present application.

[0144] An embodiment of the present application further provides a computer program, which, when executed by a processor, implements the steps of the detection method for the computer room environment in various embodiments of the present application.

[0145] The serial numbers of the above embodiments of the present application are only for description and do not represent the advantages and disadvantages of the embodiments.

[0146] In the above embodiments of the present application, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0147] In several embodiments provided by the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are merely illustrative. For example, the division of the units can be a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of units or modules can be in an electrical or other form.

[0148] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0149] In addition, the functional units in the various embodiments of the present application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0150] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage medium includes: USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks or optical discs and other various media that can store program codes.

[0151] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for detecting a computer room environment, characterized in that: include: Acquire multi-dimensional environmental data of multiple collection points in the computer room, wherein the multi-dimensional environmental data includes temperature data and humidity data; The multi-dimensional environmental data is analyzed using a prediction model to obtain a prediction result, wherein the prediction result is used to reflect the temperature and humidity environmental status of the computer room; Determine the temperature and humidity environment state of the computer room corresponding to the prediction result.

2. The method according to claim 1, characterized in that The prediction model is trained in the following way: Acquire historical multi-dimensional environmental data of the computer room and real categories corresponding to the historical multi-dimensional environmental data, wherein the real categories are used to describe the temperature and humidity environmental status of the computer room; Inputting the historical multi-dimensional environmental data into an initial prediction model for training to obtain a prediction category; Determine the loss value corresponding to the predicted category and the true category; Determine the gradient of the model parameters of the initial prediction model corresponding to the loss value, and update the model parameters according to the gradient; When the loss value meets a preset condition, the training of the initial prediction model is stopped to obtain the prediction model.

3. The method according to claim 2, characterized in that The historical multi-dimensional environmental data includes historical temperature data and historical humidity data. Before inputting the historical multi-dimensional environmental data into the initial prediction model for training, the method further includes: Generate a temperature and humidity diagram according to the historical temperature data and the historical humidity data, wherein the abscissa of each temperature and humidity point in the temperature and humidity diagram represents temperature, and the ordinate of each temperature and humidity point represents humidity; Determine the true category corresponding to each temperature and humidity point in the temperature and humidity map.

4. The method according to claim 3, characterized in that After generating a temperature and humidity graph according to the historical temperature data and the historical humidity data, the method further includes: Acquire first data corresponding to each pixel in the temperature and humidity map, wherein the first data includes the RGB color value of each pixel; Determine second data corresponding to the first data, wherein the second data is used to represent the first data in an HSV color space, and the HSV color space includes a hue channel, a saturation channel, and a lightness channel; Adjusting the target channel value corresponding to the first temperature and humidity point in the temperature and humidity map to obtain third data, wherein the target channel value includes the value of any channel in the HSV color space; A first target temperature and humidity diagram corresponding to the third data is determined, wherein the first target temperature and humidity diagram is used to represent a diagram obtained by adjusting the color value of each first temperature and humidity point in the temperature and humidity diagram.

5. The method according to claim 4, characterized in that Adjusting the target channel value corresponding to the first temperature and humidity point in the temperature and humidity map includes: Determine the data corresponding to the first temperature and humidity point from the second data to obtain target data; Obtaining a first target channel value corresponding to the target data; Determining an offset corresponding to the target channel according to the first target channel value; The first target channel value is adjusted according to the offset to obtain a second target channel value.

6. The method according to claim 3, characterized in that After generating a temperature and humidity graph according to the historical temperature data and the historical humidity data, the method further includes: Obtaining coordinate information corresponding to a second temperature and humidity point in the temperature and humidity map; Determine a rotation parameter corresponding to the temperature and humidity diagram, wherein the rotation parameter includes a rotation center and a rotation angle; generating a first temperature and humidity diagram corresponding to the rotation parameter; Adjusting the coordinate information according to the rotation parameter to obtain target coordinate information; The target coordinate information is projected onto the first temperature and humidity map to obtain a second target temperature and humidity map.

7. The method according to claim 2, characterized in that: The true class is determined in the following way: When the temperature of the computer room is lower than a first temperature threshold and the humidity of the computer room is lower than a first humidity threshold, determining that the computer room corresponds to a first category label; When the temperature of the computer room is lower than a second temperature threshold, determining that the computer room corresponds to a second category label, wherein the second temperature threshold is lower than the first temperature threshold; When the temperature of the computer room is greater than a third temperature threshold and the humidity of the computer room is greater than a second humidity threshold, determining that the computer room corresponds to a third category label; When the temperature of the equipment room is greater than a fourth temperature threshold, it is determined that the equipment room corresponds to a fourth category label, wherein the fourth temperature is greater than the third temperature threshold.

8. A device for detecting a computer room environment, characterized in that: include: An acquisition module, used to acquire multi-dimensional environmental data of multiple collection points in the computer room, wherein the multi-dimensional environmental data includes temperature data and humidity data; A prediction module, used to analyze the multi-dimensional environmental data using a prediction model to obtain a prediction result, wherein the prediction result is used to reflect the temperature and humidity environmental status of the computer room; A determination module is used to determine the temperature and humidity environment state of the computer room corresponding to the prediction result.

9. An electronic device, characterized in that: include: A memory and a processor, wherein the memory is used to store program instructions; the processor is connected to the memory and is used to execute the method for detecting the computer room environment as described in any one of claims 1 to 7.

10. A non-volatile storage medium, characterized in that: The non-volatile storage medium includes a stored computer program, wherein the device where the non-volatile storage medium is located executes the method for detecting a computer room environment as described in any one of claims 1 to 7 by running the computer program.

11. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by the processor, the method for detecting the computer room environment described in any one of claims 1 to 7 is implemented.

Images