Extreme weather load prediction method and system

Abstract

The invention belongs to the technical field of power load prediction, and provides an extreme weather load prediction method and system, and the method comprises the steps: obtaining historical power load data and to-be-predicted day meteorological data; performing random sparse mode decomposition on the acquired historical power load data and the weather data of the day to be predicted, extracting a mode component, and constructing a load prediction feature data set; calculating the feature comprehensive similarity of the constructed load prediction feature data set, and constructing a load prediction comparison learning data set; and carrying out load prediction on the to-be-predicted day based on the constructed load prediction comparison learning data set to obtain a load prediction result of the extreme weather.

Description

Technical Field

[0001] The present invention belongs to the technical field of electric load forecasting, and particularly relates to an extreme weather load forecasting method and system. Background Art

[0002] The statements in this part merely provide background technical information related to the present invention and do not necessarily constitute prior art.

[0003] Load forecasting is the core link in the operation and dispatching of power systems, and its accuracy directly affects the stability, economy and reliability of power systems; by forecasting the power load within a certain period in the future, power systems can reasonably arrange power generation plans, optimize grid dispatching, prevent power accidents and reduce operating costs. With the gradual opening of the power market and the development of smart grid technologies, the demand and challenges of load forecasting are increasing day by day. Especially in the face of complex and changing external environmental conditions, it is particularly important to achieve high-precision load forecasting.

[0004] Under extreme weather conditions, such as high temperature, heavy rain, typhoon, cold wave, etc., the load of power systems often fluctuates violently. These weather conditions not only affect power demand, but may also damage power infrastructure, thus affecting the stability of power supply. The changes in the power load under extreme weather conditions have the following several significant characteristics:

[0005] (1) The non-linearity and complexity are enhanced. The power load fluctuations caused by extreme weather often present non-linear and complex change patterns, and traditional linear models are difficult to effectively capture their internal laws;

[0006] (2) The data volatility increases. During extreme weather events, the volatility of meteorological data and load data increases significantly, bringing greater uncertainty and challenges to the prediction model;

[0007] (3) The lack of historical data. Extreme weather events are relatively rare, resulting in limited relevant historical data. Traditional data-driven methods may be difficult to make effective predictions by fully utilizing these limited data. Fourth, the time series dependence is strong. The load data may show stronger time series dependence under extreme weather conditions, and the prediction model needs to have strong time series modeling capabilities.

[0008] Current load forecasting methods mainly include traditional statistical methods, such as autoregressive integrated moving average model, seasonal decomposition method; machine learning methods, such as support vector machine, random forest; deep learning methods, such as long short-term memory network, convolutional neural network, Transformer and its variants. Although there are already various load forecasting methods currently, under extreme weather conditions, there are still the following deficiencies:

[0009] (1)Insufficient ability to model complex non - linear relationships:

[0010] Traditional methods and some machine learning methods are difficult to comprehensively capture the complex non - linear load fluctuations caused by extreme weather.

[0011] (2)Limitations in feature extraction

[0012] Existing methods are difficult to fully utilize the multi - dimensional information of load data and meteorological data in the feature extraction stage, resulting in insufficient expression of key information by the model.

[0013] (3)Insufficient utilization of historical similarity

[0014] Existing methods usually fail to fully consider the similarity relationship between the day to be predicted and historical days, and cannot effectively use the similar patterns in historical data for prediction.

[0015] (4)Lack of effective representation learning mechanism

[0016] Although deep learning methods have advantages in feature learning, they lack a highly targeted representation learning mechanism and are difficult to achieve high - precision prediction under complex weather conditions.

[0017] Therefore, there is an urgent need for a method that can comprehensively utilize various models, fully explore the multi - dimensional information in load data and meteorological data, and improve the accuracy and robustness of load prediction under extreme weather conditions. Summary of the Invention

[0018] To solve the above problems, the present invention proposes an extreme - weather load prediction method and system, which effectively combines random sparse modal decomposition and contrast learning. By random sparse modal decomposition, it fully explores the multi - dimensional information in load data and meteorological data, and uses contrast learning technology to learn potential feature representations, overcoming the deficiencies of existing methods in complex environments and improving the accuracy and robustness of load prediction under extreme weather conditions.

[0019] According to some embodiments, the first solution of the present invention provides an extreme - weather load prediction method, adopting the following technical solutions:

[0020] An extreme - weather load prediction method, comprising:

[0021] Obtain historical power load data and meteorological data of the day to be predicted;

[0022] Perform random sparse modal decomposition on the obtained historical power load data and meteorological data of the day to be predicted, extract modal components, and construct a load prediction feature dataset;

[0023] Calculate the feature comprehensive similarity of the constructed load prediction feature dataset, and construct a load prediction contrast learning dataset;

[0024] Based on the constructed load prediction contrast learning dataset, perform load prediction for the day to be predicted to obtain the load prediction results for extreme weather.

[0025] As a further technical limitation, the obtained historical power load data includes at least power load, temperature, wind speed, humidity, and precipitation; the obtained meteorological data for the day to be predicted includes at least temperature, wind speed, humidity, and precipitation.

[0026] Furthermore, perform signal decomposition of power load and temperature through the random sparse modal decomposition method based on particle swarm optimization, and extract the periodic mode, trend mode, and residual mode of power load and temperature respectively; construct a feature dataset with the extracted periodic mode, trend mode, and residual mode of power load, and the periodic mode, trend mode, and residual mode of temperature, that is, complete the construction of the load prediction feature dataset.

[0027] As a further technical limitation, use the dynamic time warping method to calculate the similarity of modal features between the day to be predicted and historical days in sequence, sum up the obtained similarities, and obtain the comprehensive feature similarity of the constructed load prediction feature dataset.

[0028] As a further technical limitation, during the process of constructing the load prediction contrast learning dataset, sort the obtained comprehensive feature similarities to determine the most similar days, obtain the set of similar days and the set of dissimilar days; construct a contrast learning task based on the obtained set of similar days and the set of dissimilar days, obtain positive and negative sample representations and construct a contrast learning task reconstruction loss function, obtain the overall contrast learning loss function, complete the contrast learning training, and obtain the load prediction contrast learning dataset.

[0029] As a further technical limitation, perform contrast learning feature encoding on the obtained load prediction contrast learning dataset, and perform load prediction according to the encoded dataset and the time series prediction model TSMixer to obtain the load prediction results for extreme weather.

[0030] According to some embodiments, the second solution of the present invention provides an extreme weather load prediction system, which adopts the following technical solutions:

[0031] An extreme weather load prediction system, comprising:

[0032] An acquisition module configured to acquire historical power load data and meteorological data for the day to be predicted;

[0033] An extraction module configured to perform random sparse modal decomposition on the obtained historical power load data and meteorological data for the day to be predicted, extract modal components, and construct a load prediction feature dataset;

[0034] A construction module, configured to calculate the feature comprehensive similarity of the constructed load prediction feature dataset and construct a load prediction contrast learning dataset;

[0035] A prediction module, configured to perform load prediction for the day to be predicted based on the constructed load prediction contrast learning dataset, and obtain the load prediction result for extreme weather.

[0036] According to some embodiments, the third solution of the present invention provides a computer-readable storage medium, adopting the following technical solution:

[0037] A computer-readable storage medium, on which a program is stored, and when the program is executed by a processor, it implements the steps in an extreme weather load prediction method as described in the first solution of the present invention.

[0038] According to some embodiments, the fourth solution of the present invention provides an electronic device, adopting the following technical solution:

[0039] An electronic device, including a memory, a processor, and a program stored on the memory and running on the processor. When the processor executes the program, it implements the steps in an extreme weather load prediction method as described in the first solution of the present invention.

[0040] According to some embodiments, the fifth solution of the present invention provides a computer program product, adopting the following technical solution:

[0041] A computer program product, including software code, and the program in the software code executes the steps in an extreme weather load prediction method as described in the first solution of the present invention.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] The present invention effectively combines random sparse modal decomposition and contrast learning, fully excavates multi-dimensional information in load data and meteorological data through random sparse modal decomposition, and uses contrast learning technology to learn potential feature representations, overcoming the deficiencies of existing methods in complex environments and improving the accuracy and robustness of load prediction under extreme weather conditions. Description of the Drawings

[0044] The specification drawings constituting a part of this embodiment are used to provide a further understanding of this embodiment. The schematic embodiments and descriptions of this embodiment are used to explain this embodiment and do not constitute an improper limitation of this embodiment.

[0045] Figure 1 It is a flowchart of the extreme weather load prediction method in the first embodiment of the present invention;

[0046] Figure 2It is the architecture diagram of the extreme weather load prediction method in Embodiment 1 of the present invention;

[0047] Figure 3 It is the schematic diagram of the load data of a regional power grid on a certain day in Embodiment 1 of the present invention and the decomposed periodic mode, trend mode and residual mode;

[0048] Figure 4 It is the schematic diagram of the temperature data of a regional power grid on a certain day in Embodiment 1 of the present invention and the decomposed periodic mode, trend mode and residual mode;

[0049] Figure 5 It is the schematic diagram of the principle of contrast learning in Embodiment 1 of the present invention;

[0050] Figure 6 It is the framework diagram of contrast learning and prediction model in Embodiment 1 of the present invention;

[0051] Figure 7 It is the comparison diagram of predicted load and actual load in Embodiment 1 of the present invention;

[0052] Figure 8 It is the structural block diagram of an extreme weather load prediction system in Embodiment 2 of the present invention. Specific Embodiments

[0053] The present invention will be further described below in conjunction with the drawings and embodiments.

[0054] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0055] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components and / or their combinations.

[0056] In the present invention, terms such as "upper", "lower", "left", "right", "front", "rear", "vertical", "horizontal", "side", "bottom", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only relationship terms determined for the convenience of describing the structural relationship of each component or element of the present invention, and do not specifically refer to any component or element of the present invention, and should not be construed as a limitation to the present invention.

[0057] In the present invention, terms such as "fixed connection", "connected", "connected to" should be understood in a broad sense, which may mean a fixed connection, an integral connection or a detachable connection; it may be directly connected or indirectly connected through an intermediate medium. For relevant scientific research or technical personnel in this field, the specific meanings of the above terms in the present invention can be determined according to specific circumstances, and it should not be construed as a limitation to the present invention.

[0058] Without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0059] Embodiment 1

[0060] Embodiment 1 of the present invention introduces an extreme weather load prediction method.

[0061] As Figure 1 shown, an extreme weather load prediction method includes:

[0062] Obtain historical power load data and meteorological data of the day to be predicted;

[0063] Perform random sparse modal decomposition on the obtained historical power load data and meteorological data of the day to be predicted, extract modal components, and construct a load prediction feature data set;

[0064] Calculate the feature comprehensive similarity of the constructed load prediction feature data set, and construct a load prediction contrast learning data set;

[0065] Based on the constructed load prediction contrast learning data set, perform load prediction for the day to be predicted to obtain the load prediction result of extreme weather.

[0066] This embodiment adopts the framework as Figure 2 shown. By performing random sparse modal decomposition on load data and meteorological data, extract periodic components, trend components and residual components; calculate similarity based on the decomposed components to construct positive and negative sample pairs for contrast learning; use contrast learning technology to learn latent feature representations and input them into the TSMixer model for load prediction; given the data of the day to be measured, the load prediction value of this day can be obtained through the prediction model.

[0067] Next, this embodiment will elaborate on the extreme weather load prediction method.

[0068] This embodiment obtains data such as load, temperature, wind speed, humidity, precipitation, etc. of the regional power grid within a certain period of time, with a time granularity of 15 minutes, and constructs a data set x(t) = [y(t), x1(t), x2(t), x3(t), x4(t)]; where y(t), x1(t), x2(t), x3(t), x4(t) represent load, temperature, wind speed, humidity and precipitation respectively, and t represents time.

[0069] In this embodiment, it is assumed that the prediction date is the T-th day, the day to be predicted is the (T + 1)-th day, the historical load is the data before the (T - 1)-th day, and the characteristics such as temperature, wind speed, humidity, and precipitation include the data on the (T + 1)-th day and before. The load y(t) and temperature x1(t) data are decomposed using the particle swarm optimization-based stochastic sparse modal decomposition method. The number of modes of the stochastic sparse modal decomposition method is set to 3, which respectively correspond to the periodic mode y s (t), the trend mode y t (t), and the residual mode y r (t), and the periodic mode of temperature trend mode and residual mode The optimization ranges of the hyperparameters of the stochastic sparse modal decomposition method, namely the highest frequency ω max , window length Δ, and the number of random features S are respectively [0.01, 100], [10L, 100L], where L is the data length; the fitness function of the particle swarm algorithm is the sum of the products of the reciprocals of the permutation entropy of each mode obtained by decomposing each data and the correlation coefficient, that is where x(t) is the signal to be decomposed, m i (t) is the i-th mode obtained by decomposition, PeEn(m i (t)) is the permutation entropy of m i (t), corr(x(t), m i (r)) is the Pearson correlation coefficient between the decomposed signal and the decomposed mode, and the calculation formula of the correlation coefficient is where Cov(·, ·) represents covariance; the parameters with the highest fitness are taken as the optimal hyperparameters of the stochastic sparse modal decomposition for decomposition. The modal component results extracted for the load and temperature are respectively as Figure 3 and Figure 4 shown

[0070] In this embodiment, the modes obtained by decomposing the load and temperature, as well as data such as wind speed, humidity, and precipitation, are used to form a feature dataset The dataset is segmented along the time axis at 96 points per day; the similarity between the 96 points of the day to be predicted (i.e., the (T + 1)-th day) in the feature dataset and the 96 points of each day in the nearest month's historical days (i.e., from the (T - 31)-th day to the (T - 1)-th day) is calculated. The dynamic time warping method is used for the calculation method of similarity. The similarity between the day to be predicted and each feature of the historical days is calculated according to the cosine similarity in turn, and then the similarities of each feature are summed to obtain the comprehensive similarity between the day to be predicted and each historical day. The comprehensive similarities are sorted, and the day with the highest comprehensive similarity is taken as the most similar day, and the k days with the smallest comprehensive similarity are taken as the dissimilar days

[0071] In this embodiment, it is assumed that the training set contains N days, and the training set is constructed. h i =(x i , y i ). Repeat the previous steps for N days in the training set D to obtain the set D of similar days in the training set pos and the set D of dissimilar days in the training set neg . Construct a contrastive learning task. The contrastive learning network includes a feature encoder a ground-truth projector and a feature projector A contrastive learning training sample is where contains k negative samples. The principle of contrastive learning is as Figure 5 shown.

[0072] Obtain the positive and negative sample representations. Given Define Construct the reconstruction loss function of the contrastive learning task. For a pair of positive samples, the loss function of this contrastive learning is

[0073]

[0074] where

[0075] Construct the information bottleneck loss function, that is Therefore, the overall loss function of contrastive learning is expressed as

[0076] In this embodiment, after contrastive learning training, the feature encoder Encode all x in the training data set D Train the encoded data set through the time series prediction model TSMixer to obtain the prediction network f as Figure 6 shown. ω .

[0077] It should be noted that this embodiment uses the mutation time memory hybrid Transformer network (M-DualFormer) to train the training data set D. The M-DualFormer used includes:

[0078] (1) Feature extraction layer, that is, it includes a feature projector g that projects the encoded data x into the model hidden space.

[0079] (2) Temporal Mixing Layer, which includes a temporal mixer TT (a PatchTST) for mixing the input representations along the temporal dimension, a feature mixer FT (a PatchTST) for mixing the input data along the feature dimension, a temporal feature mixer MLP for final feature fusion, and a mutation temporal memory module M for recording various data patterns encountered during training.

[0080] (3) Output Layer, which includes an output projector f for projecting the latent variables after temporal mixing to the final predicted values.

[0081] For each sample encoded by contrastive learning Combined with the feature projector g, the latent representation x is obtained i = g(x i ); The obtained latent representation first passes through a temporal mixer for capturing relationships in the temporal dimension, a feature mixer for capturing relationships in the feature dimension. The captured dimensional relationships are recorded in a memory module M, and m contains two types of memory units m time , m feature respectively used to save The capacity of the memory module M is L. When the capacity reaches the upper limit, the data stored in it will be overwritten by the latest data according to the time proximity; will undergo final fusion through a temporal feature mixer The final will be projected through the output projector to obtain the initial predicted value At the same time, according to and M time , M feature the cosine similarities of different samples saved in M are used to select the top-K most similar samples and these K samples pass through the temporal feature mixer and the output projector to obtain K predicted values The final result is where are learnable parameters.

[0082] Given the data x T of the day to be predicted, a trained feature encoder a trained prediction network f ω , encode through the feature encoder obtain the load prediction value y of the day to be predicted through the prediction network T = f ω (z T ).

[0083] In this embodiment, the load data from March 1, 2022 to August 1, 2024, as well as the temperature, humidity, wind speed, and precipitation data of the weather forecast are selected. Using the random sparse modal decomposition method, the load data and temperature data are respectively decomposed into three modes, namely, period, trend, and residual; among them, the load period mode captures the daily cycle and weekly cycle of the load to form the typical load of each week; the trend mode shows whether the current load is in an upward or downward trend; the residual part captures the sudden changes in the load. The three modes of the load, the three modes of the temperature, humidity, wind speed, and precipitation data are segmented, with one segment per day. Then, for the k-th day in the dataset, 30 days are traced back to find 3 similar days and 3 dissimilar days for the k-th day.

[0084] The feature encoder, feature projector, and ground truth encoder are trained using the found similar days and dissimilar days to learn a hidden representation for the feature data (i.e., temperature, humidity, wind speed, precipitation) of each day. Then, this hidden representation is combined with the date data, including what day of the week, month, and year the current day is, as the input to the TSMixer to predict the load, and the TSMixer is trained according to the prediction error.

[0085] This embodiment aims to predict the load on August 5, 2024. Based on the acquired temperature data from July 5 to August 5, 2024, the temperature data is subjected to modal decomposition. The decomposed modes of August 5 are taken, and combined with the temperature, humidity, wind speed, and precipitation of that day, and fed into the feature encoder to obtain the hidden encoding of the meteorological features. This hidden encoding and the date elements are fed into the trained TSMixer to obtain Figure 7 the load prediction value for August 5 as shown.

[0086] This embodiment effectively combines random sparse modal decomposition and contrast learning. By using random sparse modal decomposition, multi-dimensional information in the load data and meteorological data is fully mined, and potential feature representations are learned using contrast learning technology, overcoming the deficiencies of existing methods in complex environments. The accuracy rate is as high as 98.1%, improving the load prediction accuracy and robustness under extreme weather conditions, and having high practical value and application prospects.

[0087] Embodiment 2

[0088] Embodiment 2 of the present invention introduces an extreme weather load prediction system.

[0089] As Figure 8 shown, an extreme weather load prediction system includes:

[0090] An acquisition module configured to acquire historical power load data and meteorological data of the day to be predicted;

[0091] An extraction module, configured to perform random sparse modal decomposition on the acquired historical power load data and the meteorological data of the day to be predicted, extract modal components, and construct a load prediction feature data set;

[0092] A construction module, configured to calculate the feature comprehensive similarity of the constructed load prediction feature data set and construct a load prediction contrast learning data set;

[0093] A prediction module, configured to perform load prediction for the day to be predicted based on the constructed load prediction contrast learning data set, and obtain the load prediction result for extreme weather.

[0094] The detailed steps are the same as those of an extreme weather load prediction method provided in Embodiment 1, and will not be elaborated here.

[0095] Embodiment 3

[0096] Embodiment 3 of the present invention provides a computer-readable storage medium.

[0097] A computer-readable storage medium, on which a program is stored, and when the program is executed by a processor, it implements the steps in an extreme weather load prediction method as described in Embodiment 1 of the present invention.

[0098] The detailed steps are the same as those of an extreme weather load prediction method provided in Embodiment 1, and will not be elaborated here.

[0099] Embodiment 4

[0100] Embodiment 4 of the present invention provides an electronic device.

[0101] An electronic device, including a memory, a processor, and a program stored on the memory and running on the processor, and when the processor executes the program, it implements the steps in an extreme weather load prediction method as described in Embodiment 1 of the present invention.

[0102] The detailed steps are the same as those of an extreme weather load prediction method provided in Embodiment 1, and will not be elaborated here.

[0103] Embodiment 5

[0104] Embodiment 5 of the present invention provides a computer program product.

[0105] A computer program product, including software code, and the program in the software code executes the steps in an extreme weather load prediction method as described in Embodiment 1 of the present invention.

[0106] The detailed steps are the same as those of an extreme weather load prediction method provided in Embodiment 1, and will not be elaborated here.

[0107] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented in various computer languages. For example, object-oriented programming languages such as Java and interpreted scripting languages such as JavaScript, etc.

[0108] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present invention. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 block or multiple blocks.

[0109] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means, and the instruction means realizes the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 block or multiple blocks.

[0110] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable device provide steps for realizing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 block or multiple blocks.

[0111] Although the preferred embodiments of the present invention have been described, those skilled in the art can make additional changes and modifications once they know the basic creative concepts. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the present invention.

[0112] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

[0113] The above are only the preferred embodiments of this example and are not used to limit this example. For those skilled in the art, this example can have various changes and modifications. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of this example shall be included within the protection scope of this example.

Claims

1. An extreme weather load prediction method, characterized in that, Including: Obtain historical power load data and meteorological data of the day to be predicted; Perform random sparse modal decomposition on the obtained historical power load data and meteorological data of the day to be predicted, extract modal components, and construct a load prediction feature dataset; Calculate the feature comprehensive similarity of the constructed load prediction feature dataset, and construct a load prediction contrastive learning dataset; Based on the constructed load prediction contrastive learning dataset, perform load prediction for the day to be predicted to obtain the load prediction result for extreme weather.

2. The extreme weather load prediction method according to claim 1, wherein The obtained historical power load data includes at least power load, temperature, wind speed, humidity, and precipitation; the obtained meteorological data of the day to be predicted includes at least temperature, wind speed, humidity, and precipitation.

3. The extreme weather load prediction method according to claim 2, wherein Perform signal decomposition of power load and temperature through the random sparse modal decomposition method based on particle swarm optimization, and extract the periodic mode, trend mode, and residual mode of power load and temperature respectively; the periodic mode, trend mode, and residual mode of the extracted power load, as well as the periodic mode, trend mode, and residual mode of temperature, constitute a feature dataset, thus completing the construction of the load prediction feature dataset.

4. The extreme weather load prediction method according to claim 1, wherein Adopt the dynamic time warping method to calculate the similarity of modal features between the day to be predicted and historical days in sequence, and accumulate and sum the obtained similarities to obtain the feature comprehensive similarity of the constructed load prediction feature dataset.

5. The extreme weather load prediction method according to claim 1, wherein During the process of constructing the load prediction contrastive learning dataset, sort the obtained feature comprehensive similarity to determine the most similar days, obtain the set of similar days and the set of dissimilar days; construct a contrastive learning task according to the obtained set of similar days and set of dissimilar days, obtain positive and negative sample representations and construct a contrastive learning task reconstruction loss function, obtain the overall contrastive learning loss function, complete the contrastive learning training, and obtain the load prediction contrastive learning dataset.

6. The extreme weather load prediction method according to claim 1, wherein Perform contrastive learning feature encoding on the obtained load prediction contrastive learning dataset, and perform load prediction according to the encoded dataset and the time series prediction model to obtain the load prediction result for extreme weather.

7. An extreme weather load prediction system, characterized in that, Including: An acquisition module configured to obtain historical power load data and meteorological data of the day to be predicted; An extraction module configured to perform random sparse modal decomposition on the obtained historical power load data and meteorological data of the day to be predicted, extract modal components, and construct a load prediction feature dataset; A construction module configured to calculate the feature comprehensive similarity of the constructed load prediction feature dataset and construct a load prediction contrastive learning dataset; A prediction module configured to perform load prediction for the day to be predicted based on the constructed load prediction contrastive learning dataset to obtain the load prediction result for extreme weather.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by a processor, it implements the steps of an extreme weather load prediction method as described in any one of claims 1-6.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, When the processor executes the program, it implements the steps of an extreme weather load prediction method as described in any one of claims 1-6.

10. A computer program product, comprising software code, characterized in that, The program in the software code executes the steps of an extreme weather load prediction method as described in any one of claims 1-6.

Images