Cooling capacity prediction method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of computer models, and provides a cooling capacity prediction method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining cooling capacity associated data of a target area; performing multiple times of global feature extraction on the cooling capacity associated data to obtain multiple cooling capacity associated features, and fusing the multiple cooling capacity associated features to obtain a cooling capacity fusion feature; according to a local dependency relationship of the cooling capacity fusion features, carrying out local feature extraction on the cooling capacity fusion features to obtain a cooling capacity feature sequence; and performing cooling capacity prediction on the target area based on the cooling capacity feature sequence, namely, capturing the global dependency characteristics in the input cooling capacity associated data sequence, then modeling the local dependency structure of the cooling capacity characteristics, and forming complementary fusion between the global modeling module and the local modeling module. The prediction performance of the model under a non-linear and non-stationary time sequence is effectively improved, and then an accurate cooling capacity prediction result is obtained according to the output cooling capacity feature sequence.

Description

Technical Field

[0001] The present application relates to the technical field of cooling capacity prediction, and in particular to a cooling capacity prediction method and device, electronic equipment, and storage medium. Background Art

[0002] Currently, commercial buildings in my country primarily use centralized cooling to create comfortable indoor temperature and humidity environments. Common centralized cooling stations primarily include chillers, water pumps, cooling towers, and other equipment. Ice storage centralized cooling systems (referred to as "ice storage systems") also incorporate ice storage equipment, storing the cooling energy produced by the chillers for use when needed. Consequently, cooling capacity prediction has been increasingly adopted in the intelligent operation of commercial building cooling stations in recent years.

[0003] Based on the predicted hourly cooling demand for buildings over the next day, cooling station managers can select appropriate equipment operation strategies to achieve energy conservation and economic efficiency. Ice storage systems facilitate cross-temporal cooling capacity scheduling and respond to building electricity demand. Their basic principle is to leverage the time-of-use (TOU) electricity pricing for commercial electricity in cities. Ice is stored in storage tanks during nighttime, when electricity prices are lower, and during daytime, when electricity prices are higher, ice is released to provide cooling, replacing chillers and electrical cooling. This reduces and shifts peak electricity demand during peak hours, thereby regulating building electricity consumption. Due to cost and space constraints, the capacity of ice storage tanks often cannot meet the full-day cooling demand of buildings during the cooling season. Therefore, it is necessary to allocate ice melting cooling periods based on TOU electricity pricing, allowing limited ice to be used during high-price periods when electric cooling is uneconomical. To allocate ice melting cooling capacity for different periods of the day ahead, it is necessary to predict building cooling demand for each period. Therefore, cooling capacity prediction for commercial building ice storage systems is of great engineering significance. However, existing technologies cannot accurately predict cooling capacity. Summary of the Invention

[0004] In view of this, embodiments of the present application provide a cooling capacity prediction method and device, an electronic device, and a storage medium to solve the problem in the prior art that cooling capacity cannot be accurately predicted.

[0005] In a first aspect of an embodiment of the present application, a cooling capacity prediction method is provided, which includes: obtaining cooling capacity related data of a target area; performing multiple global feature extractions on the cooling capacity related data to obtain multiple cooling capacity related features, and fusing the multiple cooling capacity related features to obtain cooling capacity fusion features; performing local feature extraction on the cooling capacity fusion features according to the local dependency of the cooling capacity fusion features to obtain a cooling capacity feature sequence; a prediction module is used to predict the cooling capacity of the target area based on the cooling capacity feature sequence.

[0006] According to a second aspect of an embodiment of the present application, a cooling capacity prediction device is provided, which includes: an acquisition module for acquiring cooling capacity related data of a target area; a feature module for performing multiple global feature extractions on the cooling capacity related data to obtain multiple cooling capacity related features, and fusing the multiple cooling capacity related features to obtain cooling capacity fusion features; performing local feature extraction on the cooling capacity fusion features according to the local dependency of the cooling capacity fusion features to obtain a cooling capacity feature sequence; and a prediction module for performing cooling capacity predictions on the target area based on the cooling capacity feature sequence.

[0007] According to a third aspect of an embodiment of the present application, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above method when executing the computer program.

[0008] According to a fourth aspect of the embodiments of the present application, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps of the above method are implemented.

[0009] Compared with the prior art, the beneficial effects of the embodiments of the present application are as follows: in the embodiments of the present application, by obtaining the cold-related data of the target area; performing multiple global feature extractions on the cold-related data to obtain multiple cold-related features, and fusing the multiple cold-related features to obtain a cold fusion feature; performing local feature extraction on the cold fusion feature according to the local dependency of the cold fusion feature to obtain a cold feature sequence; performing cold prediction on the target area based on the cold feature sequence, wherein the present application improves the modeling capability of the trend of changes in remote time series factors by capturing the global dependency characteristics in the input cold-related data sequence; and then models the local dependency structure of the cold feature to enhance the model's response capability to short-term dynamic changes. The above-mentioned global and local modeling modules form a complementary fusion, which effectively improves the prediction performance of the model under nonlinear and non-stationary time series, and then obtains accurate cold prediction results based on the output cold feature sequence, avoiding the problem of inability to accurately predict cold in related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0011] Figure 1 This is a flow chart of a cooling capacity prediction method provided in an embodiment of the application;

[0012] Figure 2 This is a flow chart of another cooling capacity prediction method provided in an embodiment of the present application;

[0013] Figure 3 This is a flow chart of another cooling capacity prediction method provided in an embodiment of the present application;

[0014] Figure 4 This is a schematic structural diagram of a cooling capacity prediction device provided in an embodiment of the present application;

[0015] Figure 5 This is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0016] In the following description, specific details such as specific system structures and techniques are provided for purposes of illustration rather than limitation to facilitate a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obscuring the description of the present application with unnecessary detail.

[0017] A cooling capacity prediction method and device according to an embodiment of the present application will be described in detail below with reference to the accompanying drawings.

[0018] Figure 1 This is a flow chart of a cooling capacity prediction method provided by an embodiment of the present application. Figure 1 As shown, the cooling capacity prediction method includes:

[0019] S101, obtaining cooling capacity related data of a target area;

[0020] S102, performing multiple global feature extractions on the cooling capacity correlation data to obtain multiple cooling capacity correlation features, and fusing the multiple cooling capacity correlation features to obtain a cooling capacity fusion feature;

[0021] S103, performing local feature extraction on the cold quantity fusion feature according to the local dependency of the cold quantity fusion feature to obtain a cold quantity feature sequence;

[0022] S104: Predicting the cooling capacity of the target area based on the cooling capacity feature sequence.

[0023] It should be noted that the cooling capacity prediction method provided in the embodiments of the present application can be executed by part or all of an electronic device, wherein the electronic device can be a server or a terminal. The server in the embodiments of the present application can be a single server or a server cluster composed of multiple servers. The terminal in the embodiments of the present application can be a smart phone, a personal computer, a tablet computer, a wearable device, an intelligent robot, or other intelligent hardware devices. In the following method embodiments, the execution subject is an electronic device as an example for explanation.

[0024] It can be understood that the above-mentioned target areas are areas with cooling needs, such as commercial buildings and cold stations. This application will obtain cooling capacity related data of the target areas. The cooling capacity related data are data that will affect the cooling capacity demand of the target areas. The cooling capacity related data include but are not limited to at least one of the flow of people in the target area, the indoor temperature of the target area, the outdoor temperature of the target area, the humidity of the target area, and the area of ​​the target area.

[0025] After obtaining the cold energy correlation data, this application will perform multiple global feature extractions on the cold energy correlation data to obtain multiple cold energy correlation features; it can be understood that this application performs global feature extraction on the cold energy correlation data through the Transformer network; specifically, in the Transformer network, the multi-head self-attention mechanism is used to calculate the global weight relationship between different time steps in the cold energy correlation data, and obtain multiple global attention features representing the cold energy correlation. Among them, the self-attention mechanism in the Transformer network can directly model the dependency relationship between any time points in the cold energy correlation data sequence, and perform multiple global feature extractions to achieve multiple rounds of superimposed deep global semantic fusion, thereby improving the perception of nonlinear cold energy driving factors (such as weather and user behavior mutations), and thus obtaining cold energy correlation features with rich semantic representation.

[0026] The multiple sets of features output by the above-mentioned Transformer layers are subjected to feature splicing, weighting or channel compression fusion to generate cold fusion features of unified dimension for the subsequent local feature modeling process. It can be understood that the present application reorganizes the cold fusion features obtained in step S102 into a sequence structure in chronological order, and inputs them into a bidirectional long short-term memory network (LSTM network). In the LSTM network, the contextual dependencies of the cold fusion features in the short-term time range are captured through the forward and backward cyclic units respectively; the forward and backward output results are fused to generate a cold feature sequence with temporal local characteristics and contextual continuity; the cold feature sequence is subjected to feature pooling or linear transformation in the time dimension to facilitate subsequent prediction tasks. This step can fully explore the changing trend of cold data at the micro scale (such as minute level, hour level), and improve the model's response to short-term disturbances and peak offsets, thereby making up for the Transformer's relatively weak ability to model local details.

[0027] Finally, based on the cooling feature sequence, the regression prediction module is used to predict the cooling demand of the target area in the future period and obtain the cooling prediction results.

[0028] According to the technical solution provided in the embodiment of the present application, the cold capacity correlation data of the target area is obtained; multiple global feature extractions are performed on the cold capacity correlation data to obtain multiple cold capacity correlation features, and the multiple cold capacity correlation features are fused to obtain cold capacity fusion features; based on the local dependency relationship of the cold capacity fusion features, local feature extraction is performed on the cold capacity fusion features to obtain a cold capacity feature sequence; based on the cold capacity feature sequence, the cold capacity of the target area is predicted, wherein the present application improves the modeling capability of the trend of changes in remote time series factors by capturing the global dependency characteristics in the input cold capacity correlation data sequence; then the local dependency structure of the cold capacity features is modeled to enhance the model's response capability to short-term dynamic changes. The above-mentioned global and local modeling modules form a complementary fusion, which effectively improves the prediction performance of the model under nonlinear and non-stationary time series, and then obtains accurate cold capacity prediction results based on the output cold capacity feature sequence, avoiding the problem of inability to accurately predict cold capacity in related technologies.

[0029] In some embodiments, as Figure 2 As shown in the figure, multiple global feature extractions are performed on the cooling capacity correlation data to obtain multiple cooling capacity correlation features, including:

[0030] S201, performing sequence filling processing on the cooling quantity associated data to obtain cooling quantity sequence data, and dividing the cooling quantity sequence data to obtain multiple sub-cooling quantity sequence data;

[0031] S202, performing feature fusion on multiple sub-cooling sequence data to obtain fused cooling sequence features;

[0032] S203. Feature extraction is performed on the fused cold sequence features through multiple attention heads to obtain cold correlation features corresponding to each attention head. Different attention heads focus on different time series.

[0033] It can be understood that since cooling-related data usually contains multiple physical quantities, such as temperature (°C), humidity (%), wind speed (m / s), load (kW), and equipment operating status (0 / 1), the numerical ranges of these features vary greatly. In order to avoid large-value features dominating subsequent use and masking the influence of small-value features, thereby causing bias in subsequent predictions, this application performs sequence filling processing on the cooling-related data. In the step of obtaining cooling sequence data, the cooling-related data will first be normalized. It can be understood that the above-mentioned normalization processing includes but is not limited to scaling all values ​​in the cooling-related data to the range of [0,1], or converting all values ​​in the cooling-related data into a distribution with a mean of 0 and a standard deviation of 1, thereby avoiding gradient explosion or disappearance, ensuring that the parameters in the network can be updated in a balanced manner in different dimensions, improving the stability and convergence speed of networks such as Transformer and LSTM during training, and improving the final prediction accuracy and generalization ability.

[0034] After normalizing the cold-related data, the original time series data obtained after normalization is directly used as the cold sequence data. This application also splits the cold sequence data to obtain multiple sub-cold sequence data; specifically, the cold sequence data is divided according to the set time step window size to construct multiple time segments with fixed length (sub-cold sequence data). Each patch contains multi-dimensional features of a fixed time step to reduce computational complexity and provide an efficient input format for subsequent networks.

[0035] In some cases, to ensure that the sequence length of the cold series data is divisible by the block size, the original time series data needs to be zero-padded. For example, if the original time series data length is 3000 and the block size is 3888, it needs to be padded to 3888. The padded sequence is split into a series of small blocks of the specified size (patch_size), each containing multi-dimensional features. The original shape (batch_size, sequence_length, num_features) is transformed into (batch_size, num_patches, patch_size * num_features). The specific method involves reshaping the original time series according to the patch size and feature dimensions, converting the original shape from (sequence_length, num_features) to (num_patches, patch_size, num_features), where sequence_length = num_patches * patch_size. The features within each patch are then flattened to form (num_patches, patch_size * num_features), which serves as the model input format.

[0036] In some examples, this application also standardizes each patch (mean-standard deviation normalization) to unify the input dimension of the subsequent network; it can be understood that by adopting the Patch splitting strategy, the original high-dimensional, long-sequence cold sequence data can be encoded into a structurally unified, length-controlled input format, significantly reducing the computational burden of the Transformer in long sequence modeling, while retaining the cold feature patterns of the key time intervals, providing efficient input representation for the subsequent attention mechanism modeling.

[0037] In some examples, the present application performs feature fusion on multiple sub-cold sequence data to obtain fused cold sequence features; specifically, after encoding each patch into a fixed-length vector, the patches are fused using splicing, weighting, averaging or other aggregation methods to generate fused cold sequence features as a unified representation of the subsequent Transformer input.

[0038] The fused cold sequence features are fed into the Transformer model, where a multi-head self-attention mechanism is used to extract global dependency features from the fused cold sequence features. Each attention head independently calculates its focused time region and feature pattern, generating multiple cold correlation features. Different attention heads focus on different time segments or feature dimensions within the cold sequence. These multiple cold correlation features can be further fused to generate a globally aware cold fusion feature representation.

[0039] This multi-stage processing flow can effectively improve the Transformer model's ability to model long-range dependencies in cold data, while reducing computational overhead through Patch encoding, improving the adaptability of the input structure and parallel processing efficiency.

[0040] According to the technical solution provided in the embodiment of the present application, sequence filling processing is performed on the cold quantity associated data to obtain cold quantity sequence data, and the cold quantity sequence data is segmented to obtain multiple sub-cold quantity sequence data; feature fusion is performed on the multiple sub-cold quantity sequence data to obtain fused cold quantity sequence features; feature extraction is performed on the fused cold quantity sequence features through multiple attention heads to obtain cold quantity associated features corresponding to each attention head. Different attention heads focus on different time series. This multi-stage processing flow can effectively improve the Transformer model's ability to model long-range dependencies in cold quantity data, and at the same time reduce computational overhead through Patch encoding, thereby improving the adaptability of the input structure and parallel processing efficiency.

[0041] In some embodiments, multiple cold quantity correlation features are fused to obtain cold quantity fusion features, including: performing residual connection on the cold quantity correlation features corresponding to each attention head and the fused cold quantity sequence features to obtain the cold quantity fusion features. Among them, the cold quantity correlation features output by each attention head are residually connected with the aforementioned fused cold quantity sequence features, that is, the features output by each attention head and the input features are weighted element by element or directly added; the result of the residual connection is used to retain the original input information while introducing the global dependency information extracted by the attention module, thereby generating a stable and more information-rich cold quantity fusion feature. It can be understood that by introducing the residual connection structure, the problem of feature degradation in multiple rounds of attention calculations is avoided, the stability and expression ability of the model in the deep feature fusion process are effectively improved, and the modeling effect of the cold quantity feature on key time points and their changing trends is enhanced.

[0042] Preferably, if Figure 3 As shown in the figure, after the residual connection is performed, the fused cold fusion features are subjected to layer normalization to further improve the stability of feature distribution and accelerate model convergence.

[0043] In some examples, such as Figure 3 As shown, this application also introduces two fully connected layers to further extract the cold fusion features, and uses Dropout to prevent overfitting.

[0044] In some examples, the cooling capacity of the target area is predicted based on the cooling capacity feature sequence, including: extracting the cooling capacity feature of the cooling capacity feature sequence to obtain the predicted cooling capacity feature; performing cooling capacity prediction based on the predicted cooling capacity feature to obtain the cooling capacity prediction result corresponding to the target area. It can be understood that the present application designs a multi-task learning mechanism for classification tasks (such as event detection) and regression tasks (such as load forecasting). The Adam optimizer and the adaptive learning rate scheduling strategy are used to enhance the convergence ability of the model constructed by the Transformer network and the LSTM network in the early stage of training.

[0045] In order to better understand the present application, the present application provides a more specific embodiment for illustration. The present application provides a cooling capacity prediction method, which is applied to an electronic device. The electronic device is provided with a cooling capacity prediction model. The cooling capacity prediction model includes: a data processing module, a Transformer input module, a Transformer encoder layer module, a bidirectional LSTM module, and an output module. Figure 3 As shown. The functions of each module are as follows:

[0046] The data processing module is used to preprocess the cold quantity correlation data, including normalization and sequence padding, to ensure that the processed cold quantity correlation data can adapt to the input format of the subsequent model. Specifically, the cold quantity correlation data is preprocessed into an input format suitable for the cold quantity prediction model, including normalization and sequence padding operations. Normalization is to calculate the mean and standard deviation of each feature column in the original data, and standardize the data so that its values ​​are distributed in the range of [-1,1]. This operation reduces the impact of scale differences between different feature dimensions on model training. Sequence padding is specifically to ensure that the length of the input sequence can be divided by the fixed block size, and zero padding is performed at the end of the time series data. This ensures the stability and consistency of subsequent block operations.

[0047] The specific steps include: normalizing the cold quantity associated data, then performing sequence filling processing on the original data obtained after the normalization processing to obtain cold quantity sequence data, and dividing the cold quantity sequence data to obtain multiple sub-cold quantity sequence data.

[0048] Transformer input module: used to extract features from multiple sub-cooling sequence data, add position encoding, and finally perform dropout to avoid overfitting.

[0049] Transformer encoder layer module: extracts the global characteristics of the input fused cold sequence features (multiple sub-cold sequence features are fused to obtain the fused cold sequence features) through the multi-head attention mechanism and the fully connected layer.

[0050] The goal of the Transformer encoder layer module is to model the dependencies between different time steps in a time series through a global attention mechanism.

[0051] The Transformer encoder layer module uses a multi-head attention mechanism (MHA). This multi-head attention mechanism calculates attention weights between each time step, capturing global contextual information. Implementation: For the input time series features, query, key, and value matrices are constructed and attention weights are calculated using dot products. The multi-head mechanism uses multiple attention heads simultaneously for feature extraction, with each head focusing on a different aspect of the time series. Ultimately, the outputs of all heads are concatenated.

[0052] Residual connection: Function: Directly add the output of the multi-head attention mechanism to the input to form a residual connection, which avoids gradient disappearance and enhances deep feature propagation.

[0053] Layer normalization: Function: Normalization is performed after the residual connection to stabilize model training and improve convergence speed.

[0054] Feedforward network: It includes two fully connected layers, which respectively extract characteristic patterns in the time series and cooperate with Dropout to prevent overfitting.

[0055] Then normalization is performed after the residual connection to reduce internal covariate transfer and improve training efficiency.

[0056] Bidirectional LSTM: Combines the outputs of the forward and backward LSTMs to obtain contextual information and enhance the model's understanding of sequence dependencies.

[0057] Output layer: Converts the cooling feature sequence into the target prediction value.

[0058] It can be understood that the Transformer module of this application extracts the dependencies of global time steps through a multi-head attention mechanism, which is suitable for global modeling of long time series. The bidirectional LSTM module extracts local context information between adjacent time steps and captures the dynamic changes of time series through bidirectional propagation. The features output by the Transformer are used as the input of the LSTM to achieve unified modeling of global and local characteristics. In addition, this application designs a multi-task learning mechanism for classification tasks (such as event detection) and regression tasks (such as load forecasting). The Adam optimizer and the adaptive learning rate scheduling strategy are used to enhance the convergence ability of the model in the early stages of training.

[0059] In some examples, after the present application predicts the cooling capacity of the target area based on the cooling capacity feature sequence, the method further includes: displaying the cooling capacity prediction result through an interactive interface so that relevant personnel can know the cooling capacity prediction result.

[0060] All of the above optional technical solutions can be combined in any way to form optional embodiments of the present application, and will not be described in detail here.

[0061] The following are device embodiments of the present application, which can be used to implement the method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.

[0062] This embodiment also provides a cooling capacity prediction device, such as Figure 4 As shown, the device includes:

[0063] An acquisition module 401 is used to acquire cooling capacity related data of a target area;

[0064] The feature module 402 is used to perform multiple global feature extractions on the cooling capacity correlation data to obtain multiple cooling capacity correlation features, and fuse the multiple cooling capacity correlation features to obtain a cooling capacity fusion feature;

[0065] According to the local dependency of the cold quantity fusion feature, local feature extraction is performed on the cold quantity fusion feature to obtain the cold quantity feature sequence;

[0066] The prediction module 403 is used to predict the cooling capacity of the target area based on the cooling capacity feature sequence.

[0067] In some examples, the feature module 402 is also used to perform sequence filling processing on the cold quantity associated data to obtain cold quantity sequence data, segment the cold quantity sequence data to obtain multiple sub-cold quantity sequence data; perform feature fusion on the multiple sub-cold quantity sequence data to obtain fused cold quantity sequence features; perform feature extraction on the fused cold quantity sequence features through multiple attention heads to obtain cold quantity associated features corresponding to each attention head, and different attention heads focus on different time series.

[0068] The cold quantity association features corresponding to each attention head and the fused cold quantity sequence features are residually connected to obtain the cold quantity fusion features.

[0069] In some examples, the feature module 402 is further configured to obtain local dependencies of the cold fusion features through a bidirectional long short-term memory network to perform local feature extraction on the cold fusion features to obtain a cold feature sequence.

[0070] In some examples, the feature module 402 is further configured to extract cooling features from the cooling feature sequence to obtain predicted cooling features; perform cooling prediction based on the predicted cooling features to obtain cooling prediction results corresponding to the target area.

[0071] In some examples, the prediction module 403 is further configured to display the cooling capacity prediction result through an interactive interface.

[0072] According to the technical solution provided by the embodiment of the present application, the cooling capacity prediction device obtains the cooling capacity related data of the target area; performs multiple global feature extractions on the cooling capacity related data to obtain multiple cooling capacity related features, and fuses the multiple cooling capacity related features to obtain cooling capacity fusion features; performs local feature extraction on the cooling capacity fusion features according to the local dependency of the cooling capacity fusion features to obtain a cooling capacity feature sequence; performs cooling capacity prediction on the target area based on the cooling capacity feature sequence, wherein the present application improves the modeling capability of the changing trend of remote time series factors by capturing the global dependency characteristics in the input cooling capacity related data sequence; and then models the local dependency structure of the cooling capacity features to enhance the model's response capability to short-term dynamic changes. The above-mentioned global and local modeling modules form a complementary fusion, which effectively improves the prediction performance of the model under nonlinear and non-stationary time series, and then obtains accurate cooling capacity prediction results based on the output cooling capacity feature sequence, thereby avoiding the problem of inability to accurately perform cooling capacity prediction in related technologies.

[0073] Figure 5 Schematic diagram of the electronic device 5 provided in the embodiment of the present application. Figure 5 As shown, the electronic device 5 of this embodiment includes: a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable by the processor 501. When the processor 501 executes the computer program 503, the steps of the above-mentioned method embodiments are implemented. Alternatively, when the processor 501 executes the computer program 503, the functions of the modules / units in the above-mentioned device embodiments are implemented.

[0074] The electronic device 5 may be a desktop computer, a notebook, a PDA, a cloud server, or other electronic device. The electronic device 5 may include but is not limited to a processor 501 and a memory 502. Those skilled in the art will appreciate that Figure 5 This is merely an example of the electronic device 5 and does not limit the electronic device 5 . The electronic device 5 may include more or fewer components than shown in the figure, or different components.

[0075] The processor 501 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0076] The memory 502 can be an internal storage unit of the electronic device 5, such as a hard disk or memory of the electronic device 5. The memory 502 can also be an external storage device of the electronic device 5, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. The memory 502 can also include both an internal storage unit of the electronic device 5 and an external storage device. The memory 502 is used to store computer programs and other programs and data required by the electronic device.

[0077] Those skilled in the art will clearly understand that for the sake of convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0078] If the integrated module / 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 present application implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, it can implement the steps of the above-mentioned various method embodiments. The computer program may include computer program code, which may be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0079] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A cooling capacity prediction method, characterized in that: The method comprises: Obtain cooling capacity related data of the target area; Performing multiple global feature extractions on the cooling capacity correlation data to obtain multiple cooling capacity correlation features, and fusing the multiple cooling capacity correlation features to obtain a cooling capacity fusion feature; According to the local dependency relationship of the cooling fusion feature, local feature extraction is performed on the cooling fusion feature to obtain a cooling feature sequence; The cooling capacity of the target area is predicted based on the cooling capacity characteristic sequence.

2. The method according to claim 1, characterized in that Perform multiple global feature extractions on the cooling capacity correlation data to obtain multiple cooling capacity correlation features, including: Performing sequence filling processing on the cooling quantity associated data to obtain cooling quantity sequence data, and segmenting the cooling quantity sequence data to obtain a plurality of sub-cooling quantity sequence data; Perform feature fusion on multiple sub-cooling sequence data to obtain fused cooling sequence features; The fused cooling sequence features are extracted using multiple attention heads to obtain cooling correlation features corresponding to each attention head. Different attention heads focus on different time series.

3. The method according to claim 2, characterized in that A plurality of the cooling capacity correlation features are fused to obtain a cooling capacity fusion feature, including: The cold quantity association feature corresponding to each attention head is residually connected with the fused cold quantity sequence feature to obtain the cold quantity fusion feature.

4. The method according to claim 2, characterized in that According to the local dependency of the cold quantity fusion feature, local feature extraction is performed on the cold quantity fusion feature to obtain a cold quantity feature sequence, including: The local dependency of the cold quantity fusion feature is obtained through a bidirectional long short-term memory network to perform local feature extraction on the cold quantity fusion feature to obtain the cold quantity feature sequence.

5. The method according to claim 4, characterized in that Performing cooling capacity prediction on the target area based on the cooling capacity characteristic sequence includes: Extracting cooling features from the cooling feature sequence to obtain predicted cooling features; A cooling capacity prediction is performed based on the predicted cooling capacity characteristics to obtain a cooling capacity prediction result corresponding to the target area.

6. The method according to claim 5, characterized in that After performing cooling capacity prediction on the target area based on the cooling capacity feature sequence, the method further includes: displaying the cooling capacity prediction result through an interactive interface.

7. A cooling capacity prediction device, characterized in that: The device comprises: An acquisition module is used to obtain cooling capacity related data of a target area; a feature module configured to perform multiple global feature extractions on the cooling capacity correlation data to obtain multiple cooling capacity correlation features, fuse the multiple cooling capacity correlation features to obtain cooling capacity fusion features, and perform local feature extraction on the cooling capacity fusion features based on local dependencies of the cooling capacity fusion features to obtain a cooling capacity feature sequence; A prediction module is used to predict the cooling capacity of the target area based on the cooling capacity feature sequence.

8. The device according to claim 7, characterized in that The feature module is also used to perform sequence filling processing on the cold quantity associated data to obtain cold quantity sequence data, segment the cold quantity sequence data to obtain multiple sub-cold quantity sequence data; perform feature fusion on the multiple sub-cold quantity sequence data to obtain fused cold quantity sequence features; perform feature extraction on the fused cold quantity sequence features through multiple attention heads to obtain cold quantity associated features corresponding to each attention head, and different attention heads focus on different time series.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images