Source-load-price multi-task medium and long term prediction method and system based on small sample learning

Abstract

The invention relates to the technical field of comprehensive energy, and provides a source-load-price multi-task medium and long term prediction method and system based on small sample learning. The method comprises the steps that a target data set and a source domain big data set in a short period are acquired, and each of the target data set and the source domain big data set comprises transaction electric quantity, new energy generating capacity, electric power transaction price and weather data; and training a prediction model by using the source domain big data set, migrating the trained prediction model to the target data set for secondary training, and predicting the enhanced target data set by using the secondarily trained prediction model to obtain a medium and long term prediction result of the source load price. In order to solve the problem of small power transaction data volume, transfer learning and data enhancement are utilized, and in combination with transaction data of other power markets, medium and long term prediction of the source charge price of the target power market is realized.

Description

Technical Field

[0001] The present invention relates to the field of integrated energy technology, and in particular to a multi-task medium- and long-term prediction method and system for source-load price based on small sample learning. Background Art

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

[0003] The search found that there is a lack of relevant patents for the problem of electricity price prediction for small data sets. The Chinese patent with application number "CN2024104759895" discloses a medium- and long-term electricity price prediction method. It predicts wind power generation, hydropower generation, photovoltaic power generation, cross-regional power transmission generation and economic development through weather factors, and predicts electricity prices in the medium and long term. The Chinese patent with application number "CN2023105199415" discloses a multi-task joint prediction method and system for source-load price of integrated energy system, which combines multi-layer convolution with spatiotemporal attention mechanism, and uses the soft sharing and hard sharing mechanism of long-term and short-term memory network to perform source-load price joint prediction to obtain joint prediction results. The Chinese patent with application number "CN2023116591049" discloses a medium- and long-term power load multi-task learning prediction method with multi-factor fusion, which takes into account historical load, meteorological and economic characteristics, extracts data features through pyramid-type GCN, extracts time series features through TCN, and uses multi-task learning method to assist power load prediction and improve prediction accuracy.

[0004] The above methods use multiple features, multi-task prediction and frequency domain analysis to improve prediction accuracy. However, deep learning algorithms often require a large amount of historical data as support, otherwise it is very easy to fail to converge. At present, the market reform of electricity is still in its infancy, and the amount of electricity market transaction data is small, which cannot meet the high-precision multi-task medium- and long-term prediction of source-load prices. Summary of the invention

[0005] In order to solve the technical problems existing in the above-mentioned background technology, the present invention provides a multi-task medium- and long-term prediction method and system for source-load price based on small sample learning. Aiming at the problem of small amount of electricity trading data, the present invention utilizes transfer learning and data enhancement, combined with other electricity market trading data, to realize the medium- and long-term prediction of source-load price in the target electricity market.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] The first aspect of the present invention provides a multi-task medium- and long-term prediction method for source-load price based on small sample learning.

[0008] A multi-task medium- and long-term prediction method for source-load price based on small sample learning, including:

[0009] Obtain a short-term target data set and a source domain big data set, where both the target data set and the source domain big data set include transaction power, new energy power generation, power transaction prices, and weather data;

[0010] The prediction model is trained using a large data set in the source domain, and the trained prediction model is transferred to the target data set for secondary training. The secondary trained prediction model is used to predict the enhanced target data set to obtain the medium- and long-term prediction results of the source-load price.

[0011] The process of using the prediction model includes: based on the transaction power, new energy power generation, power transaction price and weather data, using the CDN network, respectively extracting the first local feature, the first feature trend, the second local feature, the second feature trend, the third local feature, the third feature trend, the fourth local feature and the fourth feature trend; inputting the result output by the CDN network into the GCN network to extract the time series graph structure data; based on the time series graph structure data, using the Informer network to obtain the short-term / medium-term and long-term prediction results of the source-load price.

[0012] Furthermore, the process of using the CDN network includes: the CDN network includes a convolution channel and a difference channel, the convolution channel fills the 0th and -1th moments in the input n-length sequence data with zeros, then performs a Conv1D operation on the data to obtain data features of n+2 length, and then deletes the features of the n+1th and n+2th time steps to obtain data features of n length; the difference channel gradually subtracts the features extracted by each channel according to the time sequence, outputs the 1st time step as 0, the tth time step as the t time step feature minus the t-1 time step feature, and outputs the n-length feature difference to express the change of the feature over time.

[0013] Furthermore, the process of using the Informer network includes: mapping the input time-series graph structure data to a high-dimensional space through an encoder, extracting high-dimensional features layer by layer using a multi-head attention mechanism and a multi-layer encoder, and outputting short-term / medium-term and long-term prediction results of source-charge price through a decoder.

[0014] Furthermore, the short-term / medium-term and long-term forecast results of the source-load price include: short-term / medium-term and long-term forecast data of new energy power generation, short-term / medium-term and long-term forecast data of power load and short-term / medium-term and long-term forecast data of electricity price.

[0015] Furthermore, the process of enhancing the target data set includes: in the process of performing secondary training on the prediction model, merging the prediction result with the original target data set to obtain an enhanced target data set.

[0016] Furthermore, after obtaining the target data set and the source domain large data set, the method further includes: preprocessing the data set, including denoising, missing value filling and standardization.

[0017] The second aspect of the present invention provides a multi-task medium- and long-term prediction method system for source-load price based on small sample learning.

[0018] A multi-task medium- and long-term prediction method system for source-load price based on small sample learning, comprising:

[0019] The data collection module is used to obtain the target data set and the source domain big data set in the short term, wherein the target data set and the source domain big data set both include the transaction power, the new energy power generation, the power transaction price and the weather data;

[0020] The training module is used to train the prediction model using a large data set in the source domain, migrate the trained prediction model to the target data set for secondary training, and use the secondary trained prediction model to predict the enhanced target data set to obtain the medium- and long-term prediction results of the source-load price;

[0021] The prediction model module is used to extract the first local feature, the first feature trend, the second local feature, the second feature trend, the third local feature, the third feature trend, the fourth local feature and the fourth feature trend based on the transaction power, the new energy power generation, the power transaction price and the weather data, respectively, using the CDN network; input the result output by the CDN network into the GCN network to extract the time series graph structure data; based on the time series graph structure data, the Informer network is used to obtain the short-term / medium-term and long-term prediction results of the source-load price.

[0022] A third aspect of the present invention provides a computer-readable storage medium.

[0023] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the multi-task medium- and long-term prediction method for source-charge price based on small sample learning as described in the first aspect above.

[0024] A fourth aspect of the present invention provides a computer device.

[0025] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps in the multi-task medium- and long-term prediction method for source-charge price based on small sample learning as described in the first aspect above are implemented.

[0026] A fifth aspect of the present invention provides a computer program product or a computer program.

[0027] The present invention provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in the multi-task medium- and long-term prediction method of source-charge price based on small sample learning as described in the first aspect above.

[0028] Compared with the prior art, the present invention has the following beneficial effects:

[0029] The present invention provides a multi-task medium- and long-term prediction method and system for source-load price based on small sample learning, which considers the local features of data and the changing trend separately, considers the correlation between the local features of data and the changing trend, uses a deep learning algorithm to learn the correlation between data features and trends, and builds two multi-task prediction models of renewable energy output, power load, and power trading price with short-term and medium- and long-term prediction lengths; by pre-training the prediction model, the model learns the data relationship between weather, renewable energy, load and electricity price, performs secondary training on the short-term prediction model using the data of the target area, and performs data enhancement on the data of the target area; and uses the enhanced target data set to retrain the medium- and long-term prediction model to adapt to the target area and avoid overfitting problems caused by too little data.

[0030] In view of the problem of lack of data in the initial stage of electricity marketization, the present invention uses a variety of deep learning algorithms combined with attention mechanisms to effectively capture the complex situation of multi-task changes in source-load prices, proposes a CDN model structure, uses data enhancement methods to expand small data set data samples, and introduces transfer learning methods to enable the model to achieve better prediction results even with less data. The present invention combines deep learning algorithms, data enhancement, and transfer learning, and can be adjusted and optimized according to different regions and different types of new energy. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The accompanying drawings in the specification, which constitute 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 improper limitations on the present invention.

[0032] Figure 1 It is a flow chart of a multi-task medium- and long-term prediction method of source-load price based on small sample learning shown in the present invention;

[0033] Figure 2 It is a schematic diagram of a CDN network shown in the present invention. DETAILED DESCRIPTION

[0034] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0035] It should be noted that the following detailed descriptions are all illustrative and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meanings as those commonly understood by those skilled in the art to which the present invention belongs.

[0036] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, it indicates the presence of features, steps, operations, devices, components and / or combinations thereof.

[0037] It should be noted that the flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the methods and systems according to various embodiments of the present disclosure. It should be noted that each box in the flowchart or block diagram can represent a module, a program segment, or a part of a code, and the module, program segment, or a part of a code may include one or more executable instructions for implementing the logical functions specified in each embodiment. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the flowchart and / or block diagram, and the combination of boxes in the flowchart and / or block diagram can be implemented using a dedicated hardware-based system that performs a specified function or operation, or can be implemented using a combination of dedicated hardware and computer instructions.

[0038] Starting from the two aspects of electricity cost and supply and demand relationship, the present invention combines deep learning algorithm to explore the prediction method of new energy, electricity load and electricity transaction price, and considers the problem of less available data in the initial stage of current electricity market construction, and proposes a multi-task medium- and long-term prediction method and system for source-load price based on small sample learning. The technical scheme of the present invention is described in detail through several embodiments below:

[0039] Embodiment 1

[0040] like Figure 1 As shown, the present invention adopts a multi-task medium- and long-term prediction method for source-charge price based on small sample learning, comprising the following steps:

[0041] Step 1: Data collection and processing

[0042] The power market transaction data with a step size of 1 hour are obtained from multiple power markets, including transaction volume, renewable energy power generation, power transaction price and local weather data. The target power market transaction data and other power market transaction data are divided into a target data set and a source domain big data set. The missing values ​​in the data set are filled by using the mean of the data at the previous and next time points, and the data set is denoised and standardized.

[0043] Step 2: Build the model

[0044] Traditional CNN uses information from future time steps when extracting features from time series data, which is very unreasonable in practical applications. To address this problem, the present invention proposes a convolutional difference network (CDN), such as Figure 2 As shown, based on the traditional CNN architecture, it is divided into a convolution channel and a difference channel. The convolution channel fills the past time steps with zeros according to different convolution kernels, and the design of information extraction is realized by deleting the last bit of the extracted features. For example, when the traditional CNN extracts time series features, when the convolution kernel size is 5, the features at the tth time step contain the information of 5 time steps (t-2, t-1, t, t+1, t+2), and contain the information of the two future time steps t+1 and t+2. The CDN proposed in the present invention inputs data of a sequence of length n, fills the data at the 0th and -1th time steps with zeros, and then performs Conv1D operation on the data to obtain data features of length n+2, deletes the features of the n+1th and n+2th time steps, and obtains data features of length n. In this way, the extracted data features only contain the information of the current time step and the past time step, and the problem of future information leakage is eliminated. In addition, the CDN difference channel gradually subtracts the features extracted by each channel in time sequence, outputs the first time step as 0, the tth time step as the tth time step feature minus the t-1th time step feature, and outputs the n-length feature difference to express the change of features over time.

[0045] The calculation process of a single CDN channel is as follows:

[0046] x=Conv1d k ([O k ,T n ]) (1)

[0047] x out =x[X1,X2...X n ] (2)

[0048] x D =x[X1,X2...X n ]-x[X1,X1,X2...X n-1 ] (3)

[0049] Among them, k is the convolution kernel size, O kis a 1×k zero matrix, T n is an input matrix of length n, Conv1d is a one-dimensional convolution operation, x is the convolution result, X i is the i-th element of x, x out is the feature output matrix, x D is the characteristic trend matrix.

[0050] Build the CDN_GCN_Informer prediction model, use CDN to extract data features and feature trends, use GCN to build the correlation between different features through the structure and nodes of the graph, and effectively learn the relationship between different features through multi-layer convolution and aggregation of neighbor features. The Informer model has excellent long-sequence prediction capabilities compared to other models. Learn the graph structure time series data extracted by GCN to achieve medium- and long-term prediction of electricity prices.

[0051] Set the output length of the CDN_GCN_Informer model. The model is divided into short-term prediction model and medium- and long-term prediction model according to the output length. The output length of the short-term prediction model is one week, and the output length of the medium- and long-term prediction model is one month.

[0052] CDN_GCN_Informer short-term prediction model inputs temperature, light, wind speed, renewable energy power generation, power load and historical electricity prices to predict renewable energy power generation, power load and electricity prices. CDN is used to extract data features and feature trends, GCN is used to obtain graph data of time series data, capture the correlation between data, and input the time series graph data into the Informer model. The encoder maps the input graph data to a high-dimensional space, and uses a multi-head attention mechanism and a multi-layer encoder to extract high-dimensional features layer by layer. The decoder outputs the forecast data of renewable energy power generation, power load and electricity prices.

[0053] Step 3: Model pre-training

[0054] The source domain big data set is divided into data sets according to the input requirements of each model. The learning rate of CDN_GCN_Informer short-term and medium-term prediction models is adjusted to 0.01. The source domain big data set is used for 100 pre-training, so that these models can learn the approximate data features and the correlation between features, ensuring the correct learning direction of the model on the small sample data set and preventing overfitting.

[0055] Step 4: Short-term prediction model migration

[0056] The pre-trained short-term prediction model is transferred to the target domain dataset, the learning rate is adjusted to 0.001, and the target domain dataset is used to train the short-term prediction model again to make it more suitable for the target area dataset.

[0057] Step 5: Data Augmentation

[0058] The target domain dataset data is input into the short-term prediction model to obtain the prediction results, and these prediction results are merged with the original target dataset, which can not only prevent overfitting due to too few data samples, but also expand reliable data.

[0059] Step 6: Electricity Price Forecast

[0060] Migrate the CDN_GCN_Informer medium- and long-term prediction model trained in step 3 to the expanded target dataset. Similarly, lower the learning rate to 0.001 and use the expanded target dataset for fine-tuning. This retains the learned data features and adapts to the data features of the target area to achieve medium- and long-term prediction of source-load prices.

[0061] The present invention provides a multi-task medium- and long-term prediction method and system for source-load price based on small sample learning, which considers the local features of data and the changing trend separately, considers the correlation between the local features of data and the changing trend, uses a deep learning algorithm to learn the correlation between data features and trends, and builds two multi-task prediction models of renewable energy output, power load, and power trading price with short-term and medium- and long-term prediction lengths; by pre-training the prediction model, the model learns the data relationship between weather, renewable energy, load and electricity price, performs secondary training on the short-term prediction model using the data of the target area, and performs data enhancement on the data of the target area; and uses the enhanced target data set to retrain the medium- and long-term prediction model to adapt to the target area and avoid overfitting problems caused by too little data.

[0062] Embodiment 2

[0063] This embodiment provides a multi-task medium- and long-term prediction method system for source-load price based on small sample learning, including:

[0064] The data collection and processing module is used to obtain power market transaction data with a step size of 1 hour from multiple power markets, including transaction volume, renewable energy power generation, power transaction price and local weather data, divide the target power market data and other power market data into a target data set and a source domain big data set, fill in the missing values ​​in the data set, and use the mean of the data before and after the time point to fill in the missing values.

[0065] The prediction model module is used to use CDN to extract data features and feature trends in weather, new energy, electricity load, and electricity price data. GCN is used to build the correlation between different features and feature trends through the structure and nodes of the graph. Through multi-layer convolution and aggregation of neighbor features, the correlation between different features can be effectively learned. The Informer model has excellent long-sequence prediction capabilities compared to other models. It learns the graph structure time series data extracted by GCN and completes the construction of multi-output prediction models for electricity price new energy, electricity load, and electricity price data.

[0066] The pre-training module is used to split the source domain large data set according to the input requirements of each model, and pre-train the constructed prediction model so that the model can learn the data features of weather, new energy, electricity load, electricity price data in multiple regions, as well as the correlation between the features.

[0067] The short-term prediction model migration module is used to migrate the trained short-term prediction model to the target dataset for fine-tuning so that it adapts to the data characteristics of the target dataset and obtains better performance on the dataset.

[0068] The target data set data enhancement module is used to input the data of the target data set into the short-term prediction model adapted to the target data set to obtain the prediction result, merge the prediction result with the target data set to expand the target data and complete the data enhancement.

[0069] The medium- and long-term prediction module is used to retrain the medium- and long-term prediction model using the expanded target data as training data, which not only ensures that the model will not be overfitted due to too few data samples, but also ensures the reliability of the data, enhances the generalization ability of the model, and improves the accuracy of the prediction results of the source-load price multi-task medium- and long-term prediction model.

[0070] Embodiment 3

[0071] This embodiment provides a computer-readable storage medium having a computer program stored thereon. When the program is executed by a processor, the steps in the multi-task medium- and long-term prediction method of source-charge price based on small sample learning as described in the first embodiment above are implemented.

[0072] Embodiment 4

[0073] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps in the multi-task medium- and long-term prediction method for source-charge price based on small sample learning as described in the first embodiment are implemented.

[0074] Embodiment 5

[0075] This embodiment provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in the multi-task medium- and long-term prediction method for source-charge price based on small sample learning described in the first embodiment.

[0076] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) containing computer-usable program code.

[0077] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 produce 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.

[0078] 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.

[0079] 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 The steps for the functions specified in one or more boxes.

[0080] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The storage medium can be a disk, an optical disk, a read-only memory (ROM) or a random access memory (RAM), etc.

[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A multi-task medium- and long-term prediction method for source-load price based on small sample learning, characterized by: include: Obtain a short-term target data set and a source domain big data set, where both the target data set and the source domain big data set include transaction power, new energy power generation, power transaction prices, and weather data; The prediction model is trained using a large data set in the source domain, and the trained prediction model is transferred to the target data set for secondary training. The secondary trained prediction model is used to predict the enhanced target data set to obtain the medium- and long-term prediction results of the source-load price. The process of using the prediction model includes: based on the transaction power, new energy power generation, power transaction price and weather data, using the CDN network, respectively extracting the first local feature, the first feature trend, the second local feature, the second feature trend, the third local feature, the third feature trend, the fourth local feature and the fourth feature trend; inputting the result output by the CDN network into the GCN network to extract the time series graph structure data; based on the time series graph structure data, using the Informer network to obtain the short-term / medium-term and long-term prediction results of the source-load price.

2. The multi-task medium- and long-term prediction method for source-charge price based on small sample learning according to claim 1 is characterized in that: The process of using the CDN network includes: the CDN network includes a convolution channel and a difference channel, the convolution channel fills the 0th and -1th moments in the input n-length sequence data with zeros, then performs Conv1D operation on the data to obtain data features of n+2 length, and then deletes the features of the n+1th and n+2th time steps to obtain data features of n length; the difference channel gradually subtracts the features extracted by each channel according to the time sequence, outputs the first time step as 0, the tth time step as the t time step feature minus the t-1 time step feature, and outputs the n-length feature difference to express the change of the feature over time.

3. The multi-task medium- and long-term prediction method for source-load price based on small sample learning according to claim 1 is characterized in that: The process of using the Informer network includes: mapping the input time-series graph structure data to a high-dimensional space through an encoder, extracting high-dimensional features layer by layer using a multi-head attention mechanism and a multi-layer encoder, and outputting short-term / medium-term and long-term prediction results of source-charge price through a decoder.

4. The multi-task medium- and long-term prediction method for source-charge price based on small sample learning according to any one of claims 1 to 3, characterized in that: The short-term / medium-term and long-term forecast results of source-load price include: short-term / medium-term and long-term forecast data of new energy power generation, short-term / medium-term and long-term forecast data of power load and short-term / medium-term and long-term forecast data of electricity price.

5. The multi-task medium- and long-term prediction method for source-charge price based on small sample learning according to claim 1 is characterized in that: The process of enhancing the target data set includes: in the process of performing secondary training on the prediction model, merging the prediction result with the original target data set to obtain an enhanced target data set.

6. The multi-task medium- and long-term prediction method for source-charge price based on small sample learning according to claim 1 is characterized in that: After obtaining the target data set and the source domain large data set, the method further includes: preprocessing the data set, including denoising, missing value filling and standardization.

7. A multi-task medium- and long-term prediction method system for source-load price based on small sample learning, characterized in that: include: The data collection module is used to obtain the target data set and the source domain big data set in the short term, wherein the target data set and the source domain big data set both include the transaction power, the new energy power generation, the power transaction price and the weather data; The training module is used to train the prediction model using a large data set in the source domain, migrate the trained prediction model to the target data set for secondary training, and use the secondary trained prediction model to predict the enhanced target data set to obtain the medium- and long-term prediction results of the source-load price; The prediction model module is used to extract the first local feature, the first feature trend, the second local feature, the second feature trend, the third local feature, the third feature trend, the fourth local feature and the fourth feature trend based on the transaction power, the new energy power generation, the power transaction price and the weather data, respectively, using the CDN network; input the result output by the CDN network into the GCN network to extract the time series graph structure data; based on the time series graph structure data, the Informer network is used to obtain the short-term / medium-term and long-term prediction results of the source-load price.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps in the multi-task medium- and long-term prediction method of source-charge price based on small sample learning as described in any one of claims 1-6 are implemented.

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 program, the steps in the multi-task medium- and long-term prediction method of source-charge price based on small sample learning as described in any one of claims 1-6 are implemented.

10. A computer program product, characterized in that The computer program product includes a computer program, and when the computer program is executed by a processor, the steps in the multi-task medium- and long-term prediction method of source-charge price based on small sample learning as described in any one of claims 1-6 are implemented.

Images