Large-space air-conditioning environment multi-point temperature and humidity prediction method
By constructing private and shared feature vectors in large industrial space air-conditioned environments and combining attention and gating weights for temperature and humidity prediction, the problem of inaccurate multi-point prediction in existing technologies is solved, achieving higher accuracy and robustness in temperature and humidity prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SOUTHWEST JIAOTONG UNIV
- Filing Date
- 2025-12-31
- Publication Date
- 2026-07-14
AI Technical Summary
Existing technologies cannot accurately predict temperature and humidity at multiple points in large industrial air-conditioned environments, especially due to the complex temperature and humidity coupling relationship in the space, making it difficult to achieve precise control.
By acquiring real-time temperature and humidity data and index data from multiple measuring points, the correlation coefficient is determined, a common feature set is selected, private and shared feature vectors are constructed, and the prediction results are linearly fused using attention weights and gating weights to achieve temperature and humidity prediction for each measuring point.
It improves the accuracy and robustness of multi-point temperature and humidity prediction in large-space air-conditioned environments, and has better precision and generalization performance.
Smart Images
Figure CN121594464B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of computer technology, specifically relating to a method for predicting temperature and humidity at multiple points in a large-space air-conditioned environment. Background Technology
[0002] With the development of industrial automation and intelligent manufacturing, high-end industries such as pharmaceuticals, semiconductors, and precision instrument manufacturing have put forward increasingly stringent requirements for the temperature, humidity, and cleanliness of the production environment. Cleanrooms, precision laboratories, and other large-space air-conditioning environments are usually characterized by large volume, high ceilings, high equipment heating power, and complex supply and return air systems. They require precise control of large spaces to solve problems such as long-term deviations from the temperature and humidity range required by the process and high energy consumption. Existing technologies are usually designed for civil public buildings, such as bedrooms, for temperature and humidity prediction. However, in industrial and other large-space environments, measuring points are often arranged around key process equipment, and the spatial temperature and humidity coupling relationship is more complex, making it difficult to predict the temperature and humidity of multiple measuring points in a large space.
[0003] Therefore, how to accurately predict the temperature and humidity at multiple points in large-space air-conditioned environments such as industrial spaces is a technical problem that needs to be solved by those skilled in the art. Summary of the Invention
[0004] The purpose of this invention is to solve the technical problem that existing technologies cannot accurately predict the temperature and humidity at multiple points in a large-space air-conditioned environment.
[0005] To achieve the above technical objectives, this invention provides a method for predicting temperature and humidity at multiple points in a large-space air-conditioned environment, the method comprising:
[0006] Acquire real-time temperature and humidity data from multiple measuring points and multiple index data within the space;
[0007] The correlation coefficient between the real-time temperature and humidity data of each measuring point and the data of each index is determined, and the data of each index are filtered based on the correlation coefficient to obtain the common feature set corresponding to all measuring points.
[0008] Based on the public feature set, the private feature vector of each measuring point is determined, and the private feature vectors of all measuring points are concatenated into a shared feature vector.
[0009] The attention weight of each measurement point is determined, specifically the attention weight of other measurement points relative to the current measurement point, and the shared vector of each measurement point is determined based on the attention weight and the shared feature vector;
[0010] The shared vector and private feature vector of each measurement point are concatenated to obtain the fused input vector, and the gate weights of the private channel and the shared channel are determined based on the fused input vector.
[0011] The first prediction result and the second prediction result are obtained by predicting the private feature vector and the shared vector of the measurement point respectively. Then, the first prediction result and the second prediction result are linearly fused based on the gate weight of the private channel and the gate weight of the shared channel to obtain the final prediction result of the measurement point.
[0012] Furthermore, the specific data indicators include: air temperature, air humidity, static pressure, ambient temperature, ambient humidity, supply air temperature, supply air humidity, supply air volume, water valve opening, air valve opening, and local cooling coil cooling capacity.
[0013] Furthermore, the correlation coefficient is specifically determined using the following formula:
[0014] ;
[0015] In the formula, The correlation coefficient between the temperature or humidity X at the measuring point and the index data Y. For covariance, Let X be the standard deviation. Let Y be the standard deviation. For expectation mathematical operators, Let X be the population mean. This represents the population average of Y.
[0016] Furthermore, the private feature vector is determined using the following formula:
[0017] ;
[0018] ;
[0019] ;
[0020] ;
[0021] ;
[0022] In the formula, This represents the historical temperature and humidity data for the m-th measuring point, specifically a T-row. A real matrix of columns, Let be a matrix space over the real number field. For the private features of the m-th measurement point, Let m be the weight matrix of the linear layer specific to the m-th measurement point. Let m be the bias vector of the linear layer specific to the m-th measurement point. To The encoded sequence obtained by encoding. This is the private encoder mapping function for the m-th measurement point. The decoded sequence is obtained by decoding the encoded sequence. This is the private decoder mapping function for the m-th measurement point. This is the private feature vector of the m-th measurement point, which is also the private feature vector obtained by time pooling the decoded sequence. This is for time pooling.
[0023] Furthermore, the shared feature vector is specifically determined by the following formula:
[0024] ;
[0025] ;
[0026] ;
[0027] ;
[0028] ;
[0029] ;
[0030] ;
[0031] In the formula, The global input matrix is obtained by concatenating the private features of all measurement points. , L rows over the real number field Column matrix space, This is the output after the first layer of convolutional pooling. It is a non-linear activation function. This is the relative offset index of the convolution kernel in the time dimension. These are the convolution weight parameters at offset k in the first convolutional layer. for In time index The value at that location, This is the bias term for the first convolutional layer. for The result after max pooling For max pooling, This is the output of the second convolutional pooling layer. These are the convolution weight parameters for the second convolutional layer at offset k. for In time index The feature vector at that location, This is the bias term for the second convolutional layer. for The result after max pooling This is the output feature sequence of the second convolutional layer after ReLU activation. For shared encoder functions, , used to Encode to obtain the encoded sequence , To The globally shared decoded sequence obtained by decoding , To slice the globally shared decoding sequence according to the task index. This is the shared decoding sequence corresponding to the m-th measurement point. This is the shared feature vector of the m-th measurement point.
[0032] Furthermore, the attention weight is specifically determined using the following formula:
[0033] ;
[0034] ;
[0035] In the formula, Score the attention weight of the k-th measurement point relative to the m-th measurement point. For the gating weights of private channels, Assign a weight matrix to the attention score. This is the shared feature vector of the k-th measurement point. Let k be the attention weight relative to m. The total number of measurement points. The index vector for softmax normalized summation.
[0036] Furthermore, the first and second weights are determined specifically according to the following formula:
[0037] ;
[0038] ;
[0039] ;
[0040] ;
[0041] In the formula, To score for gate control, These are the gate weights after Softmax normalization. This is the weight matrix of the gated fully connected layer. This is the fusion input vector for the m-th measurement point. For gated bias. For the gating weights of private channels, For the gating weight of the shared channel, This is the shared vector for the m-th measurement point.
[0042] Furthermore, the final prediction result is determined using the following formula:
[0043] ;
[0044] ;
[0045] ;
[0046] In the formula, This is the first prediction result for the m-th measurement point. This is the second prediction result for the m-th measurement point. The linear mapping weight matrix for the private prediction branch. This is the private feature vector of the m-th measurement point. This is the bias vector for the output layer of the private branch. To share the linear mapping weight matrix of the prediction branch, The bias vector for the shared branch output layer. This is the final prediction result for the m-th measurement point.
[0047] This invention provides a multi-point temperature and humidity prediction method for large-space air-conditioned environments. Compared with existing technologies, this method includes: acquiring real-time temperature and humidity data from multiple measuring points and multiple indicator data within the space; determining the correlation coefficient between the real-time temperature and humidity data of each measuring point and the indicator data, and filtering the indicator data based on the correlation coefficient to obtain a common feature set corresponding to all measuring points; determining the private feature vector of each measuring point based on the common feature set, and concatenating the private feature vectors of all measuring points into a shared feature vector; determining the attention weight of each measuring point, wherein the attention weight is specifically the attention weight of other measuring points relative to the current measuring point. The shared vector for each measurement point is determined based on the attention weights and shared feature vectors. The shared vector and private feature vector of each measurement point are concatenated to obtain a fusion input vector. The gate weights of the private channel and the shared channel are determined based on the fusion input vector. A first prediction result and a second prediction result are obtained by predicting based on the private feature vector and the shared vector of the measurement point, respectively. Then, the first prediction result and the second prediction result are linearly fused based on the gate weights of the private channel and the shared channel to obtain the final prediction result of the measurement point. This method can accurately predict the temperature and humidity of multiple points in a large space in an industrial air-conditioning environment. Attached Figure Description
[0048] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0049] Figure 1 The diagram shown is a flowchart illustrating the multi-point temperature and humidity prediction method for large-space air-conditioned environments provided in the embodiments of this specification.
[0050] Figure 2 The diagram shown is a schematic diagram of the measuring point arrangement in an embodiment of this specification;
[0051] Figure 3 The diagram shown is a schematic representation of the overall framework for multi-point temperature and humidity prediction in the embodiments of this specification. Detailed Implementation
[0052] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0053] like Figure 1 The diagram illustrates a flowchart of a multi-point temperature and humidity prediction method for large-space air-conditioned environments provided in this specification. While this specification provides the method operation steps shown in the following embodiments or accompanying drawings, based on conventional methods or without creative effort, the method may include more or fewer operation steps, or steps that are logically not causally related. The execution order of these steps is not limited to the execution order shown in the embodiments or accompanying drawings of this specification. When the method is applied in actual devices, servers, or terminal products, it can be executed sequentially or in parallel according to the methods shown in the embodiments or accompanying drawings (e.g., in a parallel processor or multi-threaded processing environment, or even in a distributed processing or server cluster implementation environment).
[0054] The multi-point temperature and humidity prediction method for large-space air-conditioned environments provided in the embodiments of this specification can be applied to terminal devices such as client and server devices, such as... Figure 1 As shown, the method specifically includes the following steps:
[0055] Step S101: Obtain real-time temperature and humidity data from multiple measuring points and multiple index data within the space.
[0056] Specifically, such as Figure 2 The diagram shows the arrangement of measuring points in a large space. Figure 2 The CCP set up 9 measuring points, including 6 temperature measuring points and 3 relative humidity measuring points. The specific index data include: air temperature, air humidity, space static pressure, external ambient temperature, external ambient humidity, supply air temperature, supply air humidity, supply air volume, water valve opening, air valve opening, and local cooling coil cooling capacity.
[0057] After obtaining the real-time temperature and humidity data and indicator data, they were all normalized. The processing procedure is as follows:
[0058] ;
[0059] In the formula, For normalization value, For data to be normalized, The minimum value of all data. This represents the maximum value of all data.
[0060] Step S102: Determine the correlation coefficient between the real-time temperature and humidity data of each measuring point and the data of each indicator, and filter the data of each indicator based on the correlation coefficient to obtain the common feature set corresponding to all measuring points.
[0061] In this embodiment of the application, the correlation coefficient is specifically determined by the following formula:
[0062] ;
[0063] In the formula, The correlation coefficient between the temperature or humidity X at the measuring point and the index data Y. For covariance, Let X be the standard deviation. Let Y be the standard deviation. For expectation mathematical operators, Let X be the population mean. This represents the population average of Y.
[0064] In the above formula, , A value of 0 indicates that the two variables are uncorrelated, a positive value indicates a positive correlation, a negative value indicates a negative correlation, and the larger the absolute value, the stronger the correlation. This invention selects the features that have the greatest impact on the results from all features to construct the model. Therefore, the absolute value of the correlation coefficient for all temperature and humidity measurement data points being greater than 0.2 is used as the threshold for feature selection.
[0065] After processing the common feature set using a sliding window approach, the following processing is performed. The sliding window approach involves extracting a sample data point from the first data point to the (N+1)th data point of the common feature set with a step size of 1, and then continuing the sliding process until the end of the common feature set.
[0066] Step S103: Determine the private feature vector of each measurement point based on the public feature set, and concatenate the private feature vectors of all measurement points into a shared feature vector.
[0067] Specifically, the private feature vector is determined using the following formula:
[0068] ;
[0069] ;
[0070] ;
[0071] ;
[0072] ;
[0073] In the formula, This represents the historical temperature and humidity data for the m-th measuring point, specifically a T-row. A real matrix of columns, Let be a matrix space over the real number field. For the private features of the m-th measurement point, Let m be the weight matrix of the linear layer specific to the m-th measurement point. Let m be the bias vector of the linear layer specific to the m-th measurement point. To The encoded sequence obtained by encoding. This is the private encoder mapping function for the m-th measurement point. To The decoded sequence obtained by decoding, This is the private decoder mapping function for the m-th measurement point. This is the private feature vector of the m-th measurement point, which is also the private feature vector obtained by time pooling the decoded sequence. This is for time pooling.
[0074] like Figure 3 The diagram shows the overall framework for multi-point temperature and humidity prediction. Figure 3 In this context, Linear m is a dedicated linear layer corresponding to the m-th measurement point. It is used to determine the private feature Task m feature of the m-th measurement point and output it to the private encoder Encoder for encoding. Then the input is fed into the private decoder to obtain the result. And then After performing time pooling to obtain the private feature vector, it is input into the gating unit G. The private encoder mapping function and the private decoder mapping function can be implemented by stacking sub-layers such as multi-head attention, feedforward network, residual connection, and layer normalization, or can be implemented in other ways by those skilled in the art according to the actual situation.
[0075] The shared feature vector is specifically determined by the following formula:
[0076] ;
[0077] ;
[0078] ;
[0079] ;
[0080] ;
[0081] ;
[0082] ;
[0083] In the formula, The global input matrix is obtained by concatenating the private features of all measurement points. , L rows over the real number field Column matrix space, This is the output after the first layer of convolutional pooling. It is a non-linear activation function. This is the relative offset index of the convolution kernel in the time dimension. These are the convolution weight parameters at offset k in the first convolutional layer. for In time index The value at that location, This is the bias term for the first convolutional layer. for The result after max pooling For max pooling, This is the output of the second convolutional pooling layer. These are the convolution weight parameters for the second convolutional layer at offset k. for In time index The feature vector at that location, This is the bias term for the second convolutional layer. for The result after max pooling This is the output feature sequence of the second convolutional layer after ReLU activation. For shared encoder functions, , used to Encode to obtain the encoded sequence , To The globally shared decoded sequence obtained by decoding , To slice the globally shared decoding sequence according to the task index. This is the shared decoding sequence corresponding to the m-th measurement point. This is the shared feature vector of the m-th measurement point.
[0084] Specifically, in Figure 3 In the middle, the concatenated global input matrix The input is fed into a CNN, which consists of two convolutional and pooling layers. The output of the CNN... After encoding and decoding, a shared feature vector is obtained, where... The encoding process is by Figure 3 The shared Encoder module in the middle handles the decoding process, which is performed by... Figure 3 The shared Decoder module in the library handles the processing.
[0085] Step S104: Determine the attention weight of each measurement point. Specifically, the attention weight is the attention weight of other measurement points relative to the current measurement point, and determine the shared vector of each measurement point based on the attention weight and the shared feature vector.
[0086] The attention weights are determined using the following formula:
[0087] ;
[0088] ;
[0089] In the formula, Score the attention weight of the k-th measurement point relative to the m-th measurement point. For the gating weights of private channels, Assign a weight matrix to the attention score. This is the shared feature vector of the k-th measurement point. Let k be the attention weight relative to m. The total number of measurement points. The index vector for softmax normalized summation.
[0090] Step S105: After concatenating the shared vector and private feature vector of each measurement point, a fused input vector is obtained, and the gate weight of the private channel and the gate weight of the shared channel are determined based on the fused input vector.
[0091] In this embodiment, the gate weight of the private channel and the gate weight of the shared channel are determined according to the following formula:
[0092] ;
[0093] ;
[0094] ;
[0095] ;
[0096] In the formula, To score for gate control, These are the gate weights after Softmax normalization. This is the weight matrix of the gated fully connected layer. This is the fusion input vector for the m-th measurement point. For gated bias. For the gating weights of private channels, For the gating weight of the shared channel, This is the shared vector for the m-th measurement point.
[0097] Specifically, such as Figure 3 As shown, in the gating unit G, the input of G is the fused input vector, and then two gating weights are obtained through the above processing, that is... Figure 3 Output in the middle.
[0098] Step S106: Based on the private feature vector and shared vector of the measurement point, make predictions to obtain the first prediction result and the second prediction result respectively. Then, based on the gate weight of the private channel and the gate weight of the shared channel, linearly fuse the first prediction result and the second prediction result to obtain the final prediction result of the measurement point.
[0099] In this embodiment of the application, the final prediction result is determined using the following formula:
[0100] ;
[0101] ;
[0102] ;
[0103] In the formula, This is the first prediction result for the m-th measurement point. This is the second prediction result for the m-th measurement point. The linear mapping weight matrix for the private prediction branch. This is the private feature vector of the m-th measurement point. This is the bias vector for the output layer of the private branch. To share the linear mapping weight matrix of the prediction branch, The bias vector for the shared branch output layer. This is the final prediction result for the m-th measurement point.
[0104] Compared with other temperature and humidity prediction models, this scheme has better accuracy, robustness, and generalization performance, as shown in Table 1 below:
[0105] Table 1
[0106]
[0107] Based on the above-described method for predicting temperature and humidity at multiple points in a large-space air-conditioned environment, one or more embodiments of this specification also provide a platform or terminal for predicting temperature and humidity at multiple points in a large-space air-conditioned environment. This platform or terminal may include devices, software, modules, plug-ins, servers, clients, etc., using the methods described in the embodiments of this specification, combined with necessary hardware implementation. Based on the same innovative concept, the systems in one or more embodiments provided in this specification are as described in the following embodiments. Since the implementation schemes and methods for solving the system problem are similar, the specific system implementation in the embodiments of this specification can refer to the implementation of the aforementioned methods. Repeated details will not be repeated. The terms "unit" or "module" used below can refer to a combination of software and / or hardware that achieves a predetermined function. Although the systems described in the following embodiments are preferably implemented in software, hardware implementation, and a combination of software and hardware, are also possible and contemplated.
[0108] It should be noted that the system described above may include other implementation methods based on the description of the corresponding method embodiments. The specific implementation methods can be referred to the description of the corresponding method embodiments above, and will not be elaborated here.
[0109] This application also provides an electronic device, including:
[0110] processor;
[0111] Memory used to store the processor's executable instructions;
[0112] The processor is configured to perform the methods provided in the embodiments described above.
[0113] The electronic device provided in this application embodiment stores executable instructions of a processor in a memory. When the processor executes the executable instructions, it can acquire real-time temperature and humidity data from multiple measuring points and multiple index data within a space; determine the correlation coefficient between the real-time temperature and humidity data of each measuring point and each index data, and filter the index data based on the correlation coefficient to obtain a common feature set corresponding to all measuring points; determine the private feature vector of each measuring point based on the common feature set, and concatenate the private feature vectors of all measuring points into a shared feature vector; determine the attention weight of each measuring point, wherein the attention weight is specifically the attention of other measuring points relative to the current measuring point. The system calculates the weights and determines the shared vector for each measurement point based on the attention weights and shared feature vectors. It then concatenates the shared vector and private feature vector of each measurement point to obtain a fused input vector, and determines the gate weights for the private channel and the shared channel based on this fused input vector. Predictions are made using the private feature vector and shared vector of the measurement point to obtain a first prediction result and a second prediction result. Finally, the first and second prediction results are linearly fused based on the gate weights for the private channel and the shared channel to obtain the final prediction result for the measurement point. This system enables accurate temperature and humidity prediction for large spaces and multiple locations in industrial air-conditioning environments.
[0114] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.
[0115] The methods or apparatus described in the embodiments provided in this specification can implement business logic through a computer program and record it on a storage medium. The storage medium can be read and executed by a computer to achieve the effects of the solutions described in the embodiments of this specification, such as:
[0116] Acquire real-time temperature and humidity data from multiple measuring points and multiple index data within the space;
[0117] The correlation coefficient between the real-time temperature and humidity data of each measuring point and the data of each index is determined, and the data of each index are filtered based on the correlation coefficient to obtain the common feature set corresponding to all measuring points.
[0118] Based on the public feature set, the private feature vector of each measuring point is determined, and the private feature vectors of all measuring points are concatenated into a shared feature vector.
[0119] The attention weight of each measurement point is determined, specifically the attention weight of other measurement points relative to the current measurement point, and the shared vector of each measurement point is determined based on the attention weight and the shared feature vector;
[0120] The shared vector and private feature vector of each measurement point are concatenated to obtain the fused input vector, and the gate weights of the private channel and the shared channel are determined based on the fused input vector.
[0121] The first prediction result and the second prediction result are obtained by predicting the private feature vector and the shared vector of the measurement point respectively. Then, the first prediction result and the second prediction result are linearly fused based on the gate weight of the private channel and the gate weight of the shared channel to obtain the final prediction result of the measurement point.
[0122] The storage medium can include physical devices for storing information, typically digitizing the information and then storing it using electrical, magnetic, or optical methods. The storage medium can include: devices that store information using electrical energy, such as various types of memory, like RAM and ROM; devices that store information using magnetic energy, such as hard disks, floppy disks, magnetic tapes, magnetic core memory, bubble memory, and USB flash drives; and devices that store information using optical methods, such as CDs or DVDs. Of course, there are other readable storage media, such as quantum memories and graphene memories.
[0123] The embodiments in this specification are not limited to conforming to industry communication standards, standard computer resource data update and data storage rules, or the situations described in one or more embodiments of this specification. Slightly modified implementations based on certain industry standards or custom methods or embodiments can also achieve the same, equivalent, or similar, or predictable, implementation effects as described above. Embodiments that utilize these modified or modified methods for data acquisition, storage, judgment, and processing still fall within the scope of optional implementations of the embodiments in this specification.
[0124] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, ASICs, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.
[0125] The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or plug-ins may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces, indirect coupling or communication connection between devices or units, and may be electrical, mechanical, or other forms.
[0126] These computer program instructions can also be loaded onto a computer or other programmable resource data updating device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0127] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, system embodiments are basically similar to method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. In the description of this specification, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this specification. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described can be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification and the features of different embodiments or examples.
[0128] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.
Claims
1. A method for predicting temperature and humidity at multiple points in a large-space air-conditioned environment, characterized in that, The method includes: Acquire real-time temperature and humidity data from multiple measuring points and multiple index data within the space; The correlation coefficient between the real-time temperature and humidity data of each measuring point and the data of each index is determined, and the data of each index are filtered based on the correlation coefficient to obtain the common feature set corresponding to all measuring points. Based on the public feature set, the private feature vector of each measuring point is determined, and the private feature vectors of all measuring points are concatenated into a shared feature vector. The attention weight of each measurement point is determined, specifically the attention weight of other measurement points relative to the current measurement point, and the shared vector of each measurement point is determined based on the attention weight and the shared feature vector; The shared vector and private feature vector of each measurement point are concatenated to obtain the fused input vector, and the gate weights of the private channel and the shared channel are determined based on the fused input vector. The first prediction result and the second prediction result are obtained by predicting the private feature vector and the shared vector of the measurement point respectively. Then, the first prediction result and the second prediction result are linearly fused based on the gate weight of the private channel and the gate weight of the shared channel to obtain the final prediction result of the measurement point.
2. The method for predicting temperature and humidity at multiple points in a large-space air-conditioned environment as described in claim 1, characterized in that, The specific data indicators include: air temperature, air humidity, static pressure, ambient temperature, ambient humidity, supply air temperature, supply air humidity, supply air volume, water valve opening, air valve opening, and local cooling coil cooling capacity.
3. The method for predicting temperature and humidity at multiple points in a large-space air-conditioned environment as described in claim 1, characterized in that, The correlation coefficient is determined using the following formula: ； In the formula, The correlation coefficient between the temperature or humidity X at the measuring point and the index data Y. For covariance, Let X be the standard deviation. Let Y be the standard deviation. For expectation mathematical operators, Let X be the population mean. This represents the population average of Y.
4. The method for predicting temperature and humidity at multiple points in a large-space air-conditioned environment as described in claim 1, characterized in that, The private feature vector is determined using the following formula: ； ； ； ； ； In the formula, This represents the historical temperature and humidity data for the m-th measuring point, specifically a T-row. A real matrix of columns, Let be a matrix space over the real number field. For the private features of the m-th measurement point, Let m be the weight matrix of the linear layer specific to the m-th measurement point. Let m be the bias vector of the linear layer specific to the m-th measurement point. To The encoded sequence obtained by encoding. This is the private encoder mapping function for the m-th measurement point. The decoded sequence is obtained by decoding the encoded sequence. This is the private decoder mapping function for the m-th measurement point. This is the private feature vector of the m-th measurement point, which is also the private feature vector obtained by time pooling the decoded sequence. This is for time pooling.
5. The method for predicting temperature and humidity at multiple points in a large-space air-conditioned environment as described in claim 4, characterized in that, The shared feature vector is specifically determined by the following formula: ； ； ； ； ； ； ； In the formula, The global input matrix is obtained by concatenating the private features of all measurement points. , L rows over the real number field Column matrix space, This is the output after the first layer of convolutional pooling. It is a non-linear activation function. This is the relative offset index of the convolution kernel in the time dimension. These are the convolution weight parameters at offset k in the first convolutional layer. for In time index The value at that location, This is the bias term for the first convolutional layer. for The result after max pooling For max pooling, This is the output of the second layer of convolutional pooling. These are the convolution weight parameters for the second convolutional layer at offset k. for In time index The feature vector at that location, This is the bias term for the second convolutional layer. for The result after max pooling This is the output feature sequence of the second convolutional layer after ReLU activation. For shared encoder functions, , used to Encode to obtain the encoded sequence , To The globally shared decoded sequence obtained by decoding , To slice the globally shared decoding sequence according to the task index. This is the shared decoding sequence corresponding to the m-th measurement point. This is the shared feature vector of the m-th measurement point.
6. The method for predicting temperature and humidity at multiple points in a large-space air-conditioned environment as described in claim 1, characterized in that, The attention weights are determined using the following formula: ； ； In the formula, Score the attention weight of the k-th measurement point relative to the m-th measurement point. For the gate weights of private channels, Assign a weight matrix to the attention score. This is the shared feature vector of the k-th measurement point. Let k be the attention weight relative to m. The total number of measurement points. The index vector for softmax normalized summation.
7. The method for predicting temperature and humidity at multiple points in a large-space air-conditioned environment as described in claim 6, characterized in that, The gate weights for private channels and shared channels are determined using the following formulas: ； ； ； ； In the formula, To score for gate control, These are the gate weights after Softmax normalization. This is the weight matrix of the gated fully connected layer. This is the fusion input vector for the m-th measurement point. For gated bias. For the gate weights of private channels, For the gating weight of the shared channel, This is the shared vector for the m-th measurement point.
8. The method for predicting temperature and humidity at multiple points in a large-space air-conditioned environment as described in claim 7, characterized in that, The final prediction result is determined using the following formula: ； ； ； In the formula, This is the first prediction result for the m-th measurement point. This is the second prediction result for the m-th measurement point. The linear mapping weight matrix for the private prediction branch. This is the private feature vector of the m-th measurement point. This is the bias vector for the output layer of the private branch. To share the linear mapping weight matrix of the prediction branch, The bias vector for the shared branch output layer. This represents the final prediction result for the m-th measurement point.