Electric power time sequence representation learning method, system, equipment and medium
By introducing causal expansion convolution module and interactive convolution module into the power timing representation learning model, the problem that the existing technology is difficult to deal with complex power data is solved, and a more efficient medium- and long-term power consumption prediction effect is achieved.
Patent Information
- Application Number
- CN202411852845.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-16
- Publication Date
- 2025-03-25
AI Technical Summary
Existing power timing means that learning methods are difficult to effectively deal with complex time scales, variable input lengths and long-term dependencies in medium and long-term power consumption data, resulting in poor prediction results.
The power timing representation learning model of the causal expansion convolution module and the interactive convolution module is adopted. The receptive field is expanded by causal expansion convolution blocks and long-term dependence is captured, and the interactive convolution block enhances short-term dependence and feature interaction.
It improves the accuracy and expression ability of power consumption prediction, can more effectively capture complex dependencies in power data, and improves the accuracy of medium- and long-term power consumption prediction.
Smart Images
Figure CN119693188A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of electric power market, and relates to an electric power time series representation learning method, system, equipment and medium. Background Art
[0002] Electricity time series representation learning aims to improve the ability to model medium- and long-term electricity consumption patterns, which is crucial for the efficiency and accuracy of electricity consumption forecasting. Although traditional representation learning methods have been quite effective in fields such as natural language processing or video, they still face challenges in processing electricity time series. The main problems include: medium- and long-term electricity consumption forecasting requires processing data at different time scales, and the representation method needs to allow highly variable input lengths; different features in the electricity time series have complex interdependencies at different times, and a single convolutional layer may not be able to fully capture these interactions; short-term fluctuations in electricity consumption data and seasonal trends, as well as long- and short-term dependencies, need to be captured simultaneously, and traditional methods have difficulty integrating these complex changes. Therefore, it is necessary to design electricity time series representation learning methods to accurately process complex electricity data, thereby improving the performance of forecasting tasks.
[0003] Traditional power time series representation learning methods are mainly based on machine learning models and deep learning models. Machine learning models, such as random forests and K-nearest neighbor algorithms (KNN), are good at processing nonlinear relationships and multidimensional feature data, and perform well in short-term power consumption prediction tasks, but have limitations when dealing with long-term dependent data. Common deep learning models include convolutional neural networks (CNNs) and Transformer models: Convolutional neural networks (CNNs) are excellent at capturing short-term patterns in time series due to their local receptive fields, but ignore the dependencies between sequences; Transformer models can effectively capture long-term dependencies with their self-attention mechanism, but perform poorly when processing long sequences. In order to make up for the shortcomings of existing methods, more advanced medium- and long-term power time series representation learning methods are urgently needed to improve the effect of power consumption prediction. Summary of the invention
[0004] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a method, system, device and medium for learning power time series representation. The method, system, device and medium can effectively process long sequences and provide a data basis for the prediction of power consumption.
[0005] To achieve the above object, the present invention discloses a method for learning power time series representation, comprising:
[0006] Acquiring electric energy consumption data to be represented;
[0007] The electric energy consumption data to be represented is input into a trained electric power time series representation learning model to obtain an electric power time series representation, wherein the electric power time series representation learning model includes a causal dilated convolution module and an interactive convolution module.
[0008] The further improvement of the power time series representation learning method of the present invention is:
[0009] Furthermore, before inputting the electric energy consumption data to be represented into the trained electric power time series representation learning model, the method further includes:
[0010] Build a dataset;
[0011] Constructing a power time series representation learning model, wherein the power time series representation learning model includes a causal dilated convolution module and an interactive convolution module;
[0012] The power time series representation learning model is trained and tested based on the data set to obtain a trained power time series representation learning model.
[0013] Furthermore, the output result H of the i-th layer in the causal dilated convolution module is i for:
[0014]
[0015]
[0016] Where i represents the current layer number, φ represents the Leaky ReLU activation function, WN represents weight normalization, Conv(·) represents one-dimensional causal dilated convolution, and H i-1 represents the output result of the i-1th layer, k represents the size of the convolution kernel, and dilation(·) represents the expansion rate of the i-th layer.
[0017] Furthermore, the output of the IC module is IC for:
[0018] C1=σ(Conv(C input ,k=1))⊙Conv(C input ,k=3)(4)
[0019] C2=Conv(C inpnt ,k=1)⊙σ(Conv(C input ,k=3)) (5)
[0020] O IC =Conv(C1+C2,k=1)(6)
[0021] Among them, σ represents the Swish activation function, Conv(·) represents the one-dimensional causal dilated convolution, ⊙ represents element-wise multiplication, C1 is the calculation result of the first convolutional layer, and C2 is the calculation result of the second convolutional layer.
[0022] The present invention discloses a power time sequence representation learning system, comprising:
[0023] An acquisition module, used for acquiring the electric energy consumption data to be represented;
[0024] A representation module is used to input the electric energy consumption data to be represented into a trained power time series representation learning model to obtain a power time series representation, wherein the power time series representation learning model includes a causal dilated convolution module and an interactive convolution module.
[0025] The further improvement of the power time series representation learning system of the present invention is:
[0026] Furthermore, it also includes:
[0027] The first building module is used to build a data set;
[0028] A second framework module is used to construct a power time series representation learning model, wherein the power time series representation learning model includes a causal dilated convolution module and an interactive convolution module;
[0029] A training module is used to train and test the power time series representation learning model based on the data set to obtain a trained power time series representation learning model.
[0030] Furthermore, the output result H of the i-th layer in the causal dilated convolution module is i for:
[0031]
[0032] Where i represents the current layer number, φ represents the Leaky ReLU activation function, WN represents weight normalization, Conv(·) represents one-dimensional causal dilated convolution, and H i-1 represents the output result of the i-1th layer, k represents the size of the convolution kernel, and dilation(·) represents the expansion rate of the i-th layer.
[0033] Furthermore, the output of the IC module is IC for:
[0034] C1=σ(Conv(C input ,k=1))⊙Conv(C input ,k=3)(4)
[0035] C2=Conv(C inpnt,k=1)⊙σ(Conv(C input ,k=3)) (5)
[0036] O IC =Conv(C1+C2,k=1)(6)
[0037] Among them, σ represents the Swish activation function, Conv(·) represents the one-dimensional causal dilated convolution, ⊙ represents element-wise multiplication, C1 is the calculation result of the first convolutional layer, and C2 is the calculation result of the second convolutional layer.
[0038] The present invention discloses a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the power timing representation learning method are implemented.
[0039] The present invention discloses a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the power timing representation learning method are implemented.
[0040] The present invention has the following beneficial effects:
[0041] During specific operation, the electric power time series representation learning method, system, device and medium described in the present invention input the electric energy consumption data to be represented into the trained electric power time series representation learning model to obtain the electric power time series representation, wherein the electric power time series representation learning model includes a causal dilated convolution module and an interactive convolution module, wherein the causal dilated convolution block uses an exponentially expanded convolution kernel to expand the receptive field while keeping the parameter quantity unchanged, and by virtue of causality, makes the current output only depend on the past input, which meets the requirements of variable-length input of actual electric energy consumption; in addition, the interactive convolution block relies on the mechanism of multi-branch convolution parallel interaction to effectively capture the relationship between different time steps and features, and improve the model expression ability. At the same time, the CDC module and the IC module complement and cooperate, the CDC module is responsible for capturing long-term dependencies, and the IC module is responsible for enhancing short-term dependencies, jointly solving the long-term and short-term dependency modeling problems, and improving the prediction accuracy of downstream medium- and long-term electric energy consumption. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:
[0043] Figure 1 It is a framework diagram of the CDIC model in the present invention;
[0044] Figure 2 is a schematic diagram of a CDC module in the present invention;
[0045] Figure 3 is a schematic diagram of an IC module in the present invention;
[0046] Figure 4 It is a timing diagram of the power consumption data in the present invention. DETAILED DESCRIPTION
[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0048] In the description of the present invention, it should be understood that the terms “include” and “comprises” indicate the presence of described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or collections thereof.
[0049] It should also be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, the singular forms "a", "an" and "the" are intended to include plural forms unless the context clearly indicates otherwise.
[0050] It should be further understood that the term "and / or" used in the present specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present invention generally indicates that the associated objects are in an "or" relationship.
[0051] It should be understood that, although the terms first, second, third, etc. may be used to describe preset ranges, etc. in the embodiments of the present invention, these preset ranges should not be limited to these terms. These terms are only used to distinguish preset ranges from each other. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.
[0052] The word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.
[0053] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention described and shown in the drawings here can usually be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0054] Various structural schematic diagrams of the embodiments disclosed in the present invention are shown in the accompanying drawings. These figures are not drawn to scale, and some details are magnified and some details may be omitted for the purpose of clear expression. The shapes of various regions and layers shown in the figures and the relative sizes and positional relationships therebetween are only exemplary, and may deviate in practice due to manufacturing tolerances or technical limitations, and those skilled in the art may additionally design regions / layers with different shapes, sizes, and relative positions according to actual needs.
[0055] Embodiment 1
[0056] The power time series representation learning method of the present invention comprises the following steps:
[0057] 1) Build a dataset;
[0058] Specifically, historical power consumption data is collected, the historical power consumption data is preprocessed, a data set is constructed based on the preprocessed historical power consumption data, and the data set is divided into a training set and a test set.
[0059] In addition, the process of preprocessing the historical power consumption data is as follows:
[0060] Based on the historical power consumption data, an original data table including a timestamp and related features is constructed, the data in the original data table is normalized, and then the data in the normalized data table is divided into a training set and a test set.
[0061] 2) Constructing a power time series representation learning model, which includes a causal dilated convolution module (CDC) and an interactive convolution module (IC), referring to Figure 1 , Figure 2 and Figure 3 ;
[0062] The CDC module uses exponentially expanding convolution kernels to capture the dependencies of time intervals while maintaining temporal causality to effectively cope with input sequences of variable lengths. The IC module applies a multi-branch convolutional parallel interaction mechanism to extract features from different scales, enhancing the ability to capture the interactive relationships between multiple features and the ability to express temporal relationships.
[0063] Specifically, the operation process of the CDC module is:
[0064] The original sequence X∈R B×T×F After normalization, the input X of the CDC module is obtained input ∈R B×T×F , while maintaining causality, capturing long-term dependencies and fluctuation trends, the CDC module ultimately outputs O CDC ∈R B×T×F ;
[0065] Use formula (1) to obtain the calculation results of the first convolution module in each layer Use formula (2) to obtain the calculation results of the second convolution module in each layer: Use formula (3) to obtain the calculation result H of each convolution layer i And the output result of CDC module O CDC ;
[0066]
[0067] Where i represents the current layer number, φ represents the Leaky ReLU activation function, WN represents weight normalization, Conv(·) represents one-dimensional causal dilated convolution, and H i-1 represents the output result of the i-1th layer, k represents the size of the convolution kernel, dilation(·) represents the expansion rate of the i-th layer, when i=1, H0=X input ;
[0068]
[0069] When i=3, the final output result of the CDC module is O. CDC =H3.
[0070] The operation process of the IC module is:
[0071] The output of the CDC module is CDC After layer normalization, we get the input C of IC block.input ∈R B×T×F , relying on the mechanism of multi-branch convolution parallel interaction, effectively captures the relationship between different time steps and features, and enhances short-term dependencies. According to equations (4) to (6), the final representation of the IC block is obtained. IC ∈R B×T×F ;
[0072] Among them, the calculation result C1∈R of the first convolutional layer is obtained using formula (4): B×T×F′ , use formula (5) to obtain the calculation result of the second convolutional layer C2∈R B×T×F′ , use formula (6) to obtain the output O of the IC module IC ;
[0073] C1=σ(Conv(C input ,k=1))⊙Conv(C input ,k=3)(4)
[0074] Among them, σ represents the Swish activation function and ⊙ represents element-wise multiplication.
[0075] C2=Conv(C input ,k=1)⊙σ(Conv(C input ,k=3)) (5)
[0076] O IC =Conv(C1+C2,k=1)(6)
[0077] 3) Training and testing the power time series representation learning model based on the training set and the test set to obtain a trained power time series representation learning model;
[0078] 4) Obtain the electric energy consumption data to be represented, and input the electric energy consumption data to be represented into the trained electric power time series representation learning model to obtain the electric power time series representation. It should be noted that the electric power time series representation can be used in downstream medium- and long-term electric energy consumption prediction tasks.
[0079] Simulation experiment
[0080] Taking the IHEPC dataset as an example, the present invention is studied; the research data includes 2075259 time steps and 9 features. The representation methods adopt the CDIC method, TS2Vec, TCN, LSTnet and T-Loss method. The downstream tasks are all power consumption prediction, and the evaluation indicator is MSE.
[0081] The following steps are used to represent and predict the power consumption data:
[0082] Collect historical electricity consumption data, download the public dataset IHEPC, and store it in the cloud server. The dataset contains 2075259 time steps of personal household electricity consumption data with a time granularity of 1 minute, such as Figure 4 shown.
[0083] The historical data is preprocessed and then divided into training set and test set. The training set / test set is divided into 1 year / remaining respectively.
[0084] The power time series representation learning method using CDIC method, TS2Vec, TCN, LSTnet and T-Loss method is used to train the model using the training set, and the model performance is evaluated using the test set to obtain the optimized representation learning model.
[0085] Input the electric energy consumption data to be represented, and the model outputs the electric power time series representation.
[0086] The power time series representation is used for medium- and long-term power consumption forecasting tasks, and the evaluation index is MSE.
[0087] The MSE is calculated using formula (7):
[0088]
[0089] Among them, y represents the true value, Represents the predicted value.
[0090] The experimental results of this embodiment are shown in Table 1, which respectively test the prediction results after being expressed by the CDIC method, TS2Vec, TCN, LSTnet and T-Loss methods.
[0091] Table 1
[0092]
[0093]
[0094] Experiments show that the CDIC method (the present invention) has significantly improved accuracy on the IHEPC dataset compared to other benchmark methods. In order to verify the overall effectiveness, five experiments used different power time series representation learning methods to obtain representation sequences for prediction. From Table 1, it can be found that the CDIC method can more accurately capture the complex dependencies of power data, enhance the expression ability and prediction accuracy, and verify its superiority in power time series representation learning and power consumption prediction.
[0095] It can be found that the present invention enhances the expression ability of power time series and the prediction effect of power consumption to a certain extent. At the same time, the experiment of this embodiment shows that compared with the suboptimal results, the CDIC algorithm has an overall decrease of 4.77% in the MSE index. Compared with the other four methods, the CDIC method shows better performance in all experimental results. Compared with the suboptimal method TS2Vec, the representation sequence obtained by the CDIC method shows better performance as an input for medium- and long-term power consumption prediction, which further verifies the superiority of the CDIC method on power time series data.
[0096] It should be noted that the present invention has the following characteristics:
[0097] The causal dilated convolution block can expand the receptive field while keeping the number of parameters unchanged, and the current output only depends on the past input, which meets the requirements of variable-length input of actual power consumption.
[0098] The interactive convolution block can effectively capture the relationship between different time steps and features and improve the model's expressiveness.
[0099] The CDC module and IC module complement and work together to solve the problem of long-term and short-term dependency modeling and improve the accuracy of downstream medium- and long-term electricity consumption forecasts.
[0100] Embodiment 2
[0101] The power time sequence representation learning system of the present invention comprises:
[0102] An acquisition module, used for acquiring the electric energy consumption data to be represented;
[0103] A representation module is used to input the electric energy consumption data to be represented into a trained electric power time series representation learning model to obtain an electric power time series representation, wherein the electric power time series representation learning model includes a causal dilated convolution module and an interactive convolution module.
[0104] As an embodiment of the present invention, it also includes:
[0105] The first building module is used to build a data set;
[0106] A second framework module is used to construct a power time series representation learning model, wherein the power time series representation learning model includes a causal dilated convolution module and an interactive convolution module;
[0107] A training module is used to train and test the power time series representation learning model based on the data set to obtain a trained power time series representation learning model.
[0108] As an implementation mode of the present invention, the output result H of the i-th layer in the causal dilated convolution module is i for:
[0109]
[0110] Where i represents the current layer number, φ represents the Leaky ReLU activation function, WN represents weight normalization, Conv(·) represents one-dimensional causal dilated convolution, and H i-1 represents the output result of the i-1th layer, k represents the size of the convolution kernel, and dilation(·) represents the expansion rate of the i-th layer.
[0111] As an embodiment of the present invention, the output of the IC module is IC for:
[0112] C1=σ(Conv(C input ,k=1))⊙Conv(C input ,k=3)(4)
[0113] C2=Conv(C input ,k=1)⊙σ(Conv(C input ,k=3)) (5)
[0114] O IC =Conv(C1+C2,k=1)(6)
[0115] Among them, σ represents the Swish activation function, Conv(·) represents the one-dimensional causal dilated convolution, ⊙ represents element-wise multiplication, C1 is the calculation result of the first convolutional layer, and C2 is the calculation result of the second convolutional layer.
[0116] The division of modules in the embodiments of the present application is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present application may be integrated into a processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.
[0117] Embodiment 3
[0118] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps of the power time series representation learning method are implemented, for example, including: constructing a data set; constructing a power time series representation learning model, the power time series representation learning model includes a causal dilation convolution module and an interactive convolution module; training and testing the power time series representation learning model based on the data set to obtain a trained power time series representation learning model; obtaining power consumption data to be represented; inputting the power consumption data to be represented into the trained power time series representation learning model to obtain a power time series representation, wherein the power time series representation learning model includes a causal dilation convolution module and an interactive convolution module. The memory may include a memory, such as a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk memory, etc. The processor, the network interface, and the memory are interconnected through an internal bus, which may be an industrial standard architecture bus, a peripheral component interconnection standard bus, an extended industrial standard architecture bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. The memory is used to store programs, specifically, the programs may include program codes, and the program codes include computer operation instructions. The memory may include internal memory and non-volatile memory, and provide instructions and data to the processor.
[0119] Embodiment 4
[0120] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the power time series representation learning method are implemented, for example, including: constructing a data set; constructing a power time series representation learning model, the power time series representation learning model includes a causal dilation convolution module and an interactive convolution module; training and testing the power time series representation learning model based on the data set to obtain a trained power time series representation learning model; obtaining power consumption data to be represented; inputting the power consumption data to be represented into the trained power time series representation learning model to obtain a power time series representation, wherein the power time series representation learning model includes a causal dilation convolution module and an interactive convolution module. Specifically, the computer-readable storage medium includes, but is not limited to, for example, a volatile memory and / or a non-volatile memory. The volatile memory may include a random access memory (RAM) and / or a cache memory (cache), etc. The non-volatile memory may include a read-only memory (ROM), a hard disk, a flash memory, an optical disk, a magnetic disk, etc.
[0121] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.
[0122] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0123] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0124] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.
[0125] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and disclosure of the invention. This application is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed by the present invention. The specification and examples are to be considered exemplary only, and the true scope and spirit of the present invention are indicated by the following claims.
[0126] It should be understood that the present invention is not limited to the exact construction that has been described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.
[0127] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any way. Any simple modification, change and equivalent structural change made to the above embodiment based on the technical essence of the present invention still falls within the protection scope of the technical solution of the present invention.
Claims
1. A method for learning power time series representation, characterized in that: include: Acquiring electric energy consumption data to be represented; The electric energy consumption data to be represented is input into a trained electric power time series representation learning model to obtain an electric power time series representation, wherein the electric power time series representation learning model includes a causal dilated convolution module and an interactive convolution module.
2. The method for learning power time series representation according to claim 1, characterized in that: Before inputting the electric energy consumption data to be represented into the trained electric power time series representation learning model, the method further includes: Build a dataset; Constructing a power time series representation learning model, wherein the power time series representation learning model includes a causal dilated convolution module and an interactive convolution module; The power time series representation learning model is trained and tested based on the data set to obtain a trained power time series representation learning model.
3. The method for learning power time series representation according to claim 1, characterized in that: The output result H of the i-th layer in the causal dilated convolution module i for: Where i represents the current layer number, φ represents the Leaky ReLU activation function, WN represents weight normalization, Conv(·) represents one-dimensional causal dilated convolution, and H i-1 represents the output result of the i-1th layer, k represents the size of the convolution kernel, and dilation(·) represents the expansion rate of the i-th layer.
4. The method for learning power time series representation according to claim 1, characterized in that: The output of the IC module O IC for: C1=σ(Conv(C input ,k=1))⊙Conv(C input ,k=3)(4) C2=Conv(C input ,k=1)⊙σ(Conv(C input ,k=3)) (5) O IC =Conv(C1+C2,k=1)(6) Among them, σ represents the Swish activation function, Conv(·) represents the one-dimensional causal dilated convolution, ⊙ represents element-wise multiplication, C1 is the calculation result of the first convolutional layer, and C2 is the calculation result of the second convolutional layer.
5. A power time series representation learning system, characterized in that: include: An acquisition module, used for acquiring the electric energy consumption data to be represented; A representation module is used to input the electric energy consumption data to be represented into a trained power time series representation learning model to obtain a power time series representation, wherein the power time series representation learning model includes a causal dilated convolution module and an interactive convolution module.
6. The power time series representation learning system according to claim 5, characterized in that: Also includes: The first building module is used to build a data set; A second framework module is used to construct a power time series representation learning model, wherein the power time series representation learning model includes a causal dilated convolution module and an interactive convolution module; A training module is used to train and test the power time series representation learning model based on the data set to obtain a trained power time series representation learning model.
7. The power time series representation learning system according to claim 5, characterized in that: The output result H of the i-th layer in the causal dilated convolution module i for: Where i represents the current layer number, φ represents the Leaky ReLU activation function, WN represents weight normalization, Conv(·) represents one-dimensional causal dilated convolution, and H i-1 represents the output result of the i-1th layer, k represents the size of the convolution kernel, and dilation(·) represents the expansion rate of the i-th layer.
8. The power time series representation learning system according to claim 5, characterized in that: The output of the IC module O IC for: C1=σ(Conv(C input ,k=1))⊙Conv(C input ,k=3)(4) C2=Conv(C input ,k=1)⊙σ(Conv(C input ,k=3)) (5) O IC =Conv(C1+C2,k=1)(6) Among them, σ represents the Swish activation function, Conv(·) represents the one-dimensional causal dilated convolution, ⊙ represents element-wise multiplication, C1 is the calculation result of the first convolutional layer, and C2 is the calculation result of the second convolutional layer.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the power timing representation learning method as described in any one of claims 1 to 4 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 power sequence representation learning method as described in any one of claims 1 to 4 are implemented.