Time series prediction method and apparatus for field of energy, and medium

Abstract

A time series prediction method and apparatus for the field of energy, and a medium, which relate to the technical field of energy prediction. The method comprises: collecting power generation time series data and a trend picture, and converting the trend picture into time series data; constructing and training a Lasso regression model, and saving difference time series data between a time series and a prediction result of the Lasso regression model; pre-processing the time series data and the difference time series data; inputting the pre-processed data into a time-series large model for model training; and using the trained time-series large model to perform power generation time series prediction. In the present application, preliminary prediction is performed by means of a regression model, and a time series model is used to perform difference correction on the prediction result, thereby effectively improving the accuracy of the prediction result and the interpretability of the model.

Description

A time series forecasting method, apparatus, and medium for the energy sector.

[0001] This application claims priority to Chinese Patent Application No. 202411205956.5, filed on August 30, 2024, entitled "A method, apparatus and medium for time series forecasting in the energy field", the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of energy forecasting technology, and to a time series forecasting method, apparatus, and medium for the energy field. Background Technology

[0003] Currently, with the popularization of renewable energy, especially the widespread application of solar power generation systems, accurate prediction of electricity consumption and solar power generation has become crucial for grid management and dispatch. Traditional prediction methods, such as ARIMA based on statistical models, can handle time series data to some extent, but their prediction performance is often unsatisfactory when faced with nonlinear, high-dimensional data and complex spatiotemporal dependencies. These methods struggle to capture complex patterns and dynamic changes in the data, leading to significant errors in the prediction results.

[0004] On the other hand, deep learning methods, such as recurrent neural networks like LSTM, while possessing powerful nonlinear modeling capabilities, face challenges in practical applications, including poor generalization and long training times. Whenever a new dataset or data pattern changes, deep learning models typically require retraining, which not only consumes significant time and computing resources but also limits their application in real-time prediction scenarios.

[0005] Furthermore, traditional forecasting methods often only accept time-series data as input. For data presented in the form of trend charts or line graphs, manual conversion is required, which undoubtedly increases the complexity and workload of data processing. At the same time, much energy-related data not only contains time-series information but also information on covariates closely related to the forecast results (such as sunlight, temperature, and pressure). How to effectively utilize this information to improve forecast accuracy is also a problem that traditional methods struggle to solve.

[0006] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] To provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not intended as a general commentary, nor is it intended to identify key / important components or describe the scope of protection of these embodiments, but rather as a prelude to the detailed description that follows.

[0008] This disclosure provides a time series forecasting method, apparatus, and medium for the energy sector, which increases the accuracy of forecasts and the interpretability of models.

[0009] In some embodiments, the method includes:

[0010] Collect time series data and trend images of power generation, and convert the trend images into time series data;

[0011] Construct and train a Lasso regression model, and save the time series data of the difference between the time series and the prediction results of the Lasso regression model;

[0012] Preprocessing of time series data and difference time series data includes data cleaning and standardization.

[0013] The preprocessed data is input into a large time series model for model training. The large time series model is an Encoder-Decoder architecture.

[0014] Power generation time series prediction is performed using a trained large time series model, including time series without covariates and time series with covariates.

[0015] Preferably, the time series forecasting process is as follows:

[0016] For time series without covariates, input the historical time series into the time series prediction model and output the predicted time series;

[0017] For time series data with covariates, input the historical time series and the corresponding covariates into the Lasso regression model for training to obtain prediction results for historical data and prediction results for future data; subtract the historical time series from the prediction results of the Lasso regression model for historical data, and feed the difference into the time series prediction model to obtain the predicted difference; add the prediction results of the Lasso regression model for future data to the predicted difference to output the final prediction result.

[0018] Preferably, the specific method for converting trend images into time series data is as follows:

[0019] The trend chart is grayscaled and binarized to generate a binarized image; OpenCV's Hough transform is used to detect straight lines in the binarized image and identify the x-axis and y-axis; OCR is used to identify the scales and markings of the coordinate axes in the binarized image; Cancy edge detection and connectivity analysis are used to trace the lines of the polyline chart; the intersection points of the lines with the x-axis and y-axis are identified, and the coordinate values ​​of each point are calculated; the extracted image is converted into a time series.

[0020] Preferably, the time series large model encoder part includes a multi-layer Transformer encoder block, each block containing a multi-head attention mechanism and a feedforward neural network;

[0021] The decoder section consists of multi-layer Transformer decoder blocks, each of which includes multi-head attention, contextual attention, and a feedforward neural network.

[0022] Preferably, the encoder further includes a position encoding module, which implements position encoding using a sine function, as shown in the following formula:

[0023] Where d represents the embedding dimension and i represents the dimension index.

[0024] Preferably, the feedforward neural network employs two linear transformations and the ReLU activation function.

[0025] Preferably, during the training of large time series models, a teacher-mandated method is used for training, with mean squared error (MAE) as the loss function. Reinforcement learning is also introduced, and model parameters are updated using the policy gradient method, as shown in the following formula:

[0026] Where, θ t+1 Let α represent the gradient at the next time step, and π represent the learning rate. θ Let a represent the policy function. t Let R represent the action and R represent the reward function.

[0027] Preferably, the covariates include light intensity, temperature, and pressure. For time series data containing covariates, a staggered correspondence method is used, where historical covariates are used as inputs for future prediction.

[0028] In some embodiments, the apparatus includes a processor and a memory storing program instructions, the processor being configured to execute the time series forecasting method for the energy sector when the program instructions are executed.

[0029] In some embodiments, the medium stores program instructions that, when executed, perform the time series forecasting method for the energy sector.

[0030] This disclosure provides a time series forecasting method, apparatus, and medium for the energy sector, which can achieve the following technical effects:

[0031] This method can process time series data and trend images simultaneously. By using image processing and OCR technology, it automatically converts trend images into time series data, significantly improving the efficiency and automation of data processing.

[0032] By constructing a large-scale time series model and training it with a large amount of energy-related data, the model can discover potential patterns in the data, thereby improving prediction accuracy. This method exhibits particularly better predictive performance for nonlinear, high-dimensional data and complex spatiotemporal dependencies.

[0033] Once trained, the large time series model of this method can be applied to multiple time series prediction tasks at once without retraining the model for each new task, which greatly improves the model's generalization ability and application scope.

[0034] This method innovatively combines traditional regression methods (such as LASSO regression) with time series forecasting. It makes preliminary predictions through a regression model and uses a time series model to correct the difference in the prediction results, effectively improving the accuracy of the prediction results and the interpretability of the model.

[0035] For time series data containing covariates, this method uses past covariates for oblique prediction, avoiding the problem of unknown future covariates and improving the effective application of covariates.

[0036] The above general description and the description below are exemplary and illustrative only and are not intended to limit this application. Attached Figure Description

[0037] One or more embodiments are illustrated by way of example with reference to the accompanying drawings. These illustrations and drawings do not constitute a limitation on the embodiments. Elements having the same reference numerals in the drawings are shown as similar elements. The drawings are not to be scaled. And wherein:

[0038] Figure 1 is a schematic diagram of the method flow;

[0039] Figure 2 is a schematic diagram of the method for converting trend images into time series data;

[0040] Figure 3 is a schematic diagram of the prediction process provided in an embodiment of this disclosure;

[0041] Figure 4 is a schematic diagram of the apparatus provided in an embodiment of this disclosure. Detailed Implementation

[0042] To provide a more detailed understanding of the features and technical content of the embodiments of this disclosure, the implementation of the embodiments of this disclosure will be described in detail below with reference to the accompanying drawings. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this disclosure. In the following technical description, for ease of explanation, several details are used to provide a full understanding of the disclosed embodiments. However, one or more embodiments may still be implemented without these details. In other cases, well-known structures and devices may be simplified in their depiction to simplify the drawings.

[0043] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of this disclosure described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion.

[0044] Unless otherwise stated, the term "multiple" means two or more.

[0045] Example 1

[0046] Referring to Figure 1, a time series forecasting method for the energy sector includes:

[0047] S1: Collect time series data and trend images of power generation, and convert the trend images into time series data.

[0048] S2: Construct and train a Lasso regression model, and save the time series data of the difference between the time series and the prediction results of the Lasso regression model.

[0049] S3: Preprocess time series data and difference time series data, including data cleaning and standardization.

[0050] S4: Input the preprocessed data into the time series large model for model training. The time series large model is an Encoder-Decoder architecture.

[0051] S5: Use the trained time series model to predict power generation time series, including time series without covariates and time series with covariates.

[0052] As a refinement of the above embodiment, the specific method for converting the trend image into time series data in step S1, as shown in Figure 2, is as follows:

[0053] The trend chart is grayscaled and binarized to generate a binarized image; OpenCV's Hough transform is used to detect straight lines in the binarized image and identify the x-axis and y-axis; OCR is used to identify the scales and markings of the coordinate axes in the binarized image; Cancy edge detection and connectivity analysis are used to trace the lines of the polyline chart; the intersection points of the lines with the x-axis and y-axis are identified, and the coordinate values ​​of each point are calculated; the extracted image is converted into a time series.

[0054] As a refinement of the above embodiment, in step S2, the model not only learns the time series data but also needs to learn the prediction of the difference between the time series and the linear regression data. Therefore, it is necessary to construct a Lasso regression model, train each time series dataset, and then subtract the predicted result of the training model from the actual value to generate the difference data. The difference data is also time series data, and the difference data is saved as part of the training data.

[0055] As a refinement of the above embodiments, the time series data preprocessing in step S3 includes:

[0056] First, the data is cleaned. Due to inaccurate identification or problems with the original data, there may be null or outlier values ​​in the data. Null and outlier values ​​are deleted.

[0057] Then, standardization is performed to reduce the impact of noise; the standardization formula is as follows:

[0058] Among them, z i These are standardized data, where μ is the mean, σ is the variance, and x is the standard deviation. i This represents the original value at time i.

[0059] As a refinement of the above embodiments, the structure of the large time series model in step S4 is an Encoder-Decoder architecture, where the encoder is responsible for capturing information from the input features, and the decoder is responsible for generating the predicted sequence. The Transformer architecture is used as a foundation, and component position encoding and cyclic residual connections specific to time series prediction are added.

[0060] The large time series model encoder part includes multi-layer Transformer encoder blocks, each containing a multi-head attention mechanism and a feedforward neural network.

[0061] The encoder also includes a position encoding module, which implements position encoding p using a sine function. t，i For time series data, location encoding is necessary because the raw input does not have explicit location information. The formula is as follows:

[0062] Where d represents the embedding dimension and i represents the dimension index.

[0063] The multi-head attention mechanism is as follows:

[0064] Multi-head attention allows the model to focus on different parts of the input sequence from different positional perspectives. For each head h, the query Q, key K, and value V are defined as follows:

[0065] Attention Score A ij The calculation is as follows:

[0066] Where, x t The input vector representing the attention mechanism Representing the Query matrix Represents the key matrix Let Q represent the Value matrix. i Represents the query after matrix transformation d represents the transpose of the Key after matrix transformation. k It is the dimension of the key vector, A ij It is the attention distribution obtained through the softmax function.

[0067] The feedforward neural network employs two linear transformations and the ReLU activation function, as shown in the following formula:

[0068] FFN(x) = max(0, xW1+b1)W2+b2

[0069] Where FFN(x) represents the feedforward neural network function, x represents the input information of the network, W1 represents the weight parameters of linear layer 1, b1 represents the bias parameters of linear layer 1, W2 represents the weight parameters of linear layer 2, and b2 represents the bias parameters of linear layer 2.

[0070] The decoder consists of multi-layer Transformer decoder blocks, each containing multi-head attention, contextual attention, and a feedforward neural network. In the first layer of the decoder, masked multi-head attention is used to ensure the model only sees past sequence values. The contextual attention layer uses the encoder's output as keys and values ​​to focus on information from the entire input sequence. Recurrent residual connections are introduced in the decoder, meaning that after each decoder layer, the output of the current time step is combined with the output of the previous time step. This helps maintain consistency of temporal information and helps mitigate the vanishing gradient problem.

[0071] Optionally, during the training of the large time series model, a teacher-forced training method is employed, where the actual sequence values ​​are used as the input for the next time step. The mean squared error (MAE) is used as the loss function, and reinforcement learning is introduced to update the model parameters through the policy gradient method, as shown in the following formula:

[0072] Where, θ t+1 Let α represent the gradient at the next time step, and π represent the learning rate. θ Let a represent the policy function. t Let R represent the action and R represent the reward function.

[0073] As a refinement of the above embodiments, the covariates include light intensity, temperature, and pressure. For time series data containing covariates, a staggered correspondence method is adopted, in which historical covariates are used as inputs for future prediction.

[0074] As a refinement of the above embodiment, the time series prediction process in step S5, as shown in Figure 3, is as follows:

[0075] For time series without covariates, input the historical time series into the time series prediction model and output the predicted time series;

[0076] For time series data with covariates, input the historical time series and the corresponding covariates into the Lasso regression model for training to obtain prediction results for historical data and prediction results for future data; subtract the historical time series from the prediction results of the Lasso regression model for historical data, and feed the difference into the time series prediction model to obtain the predicted difference; add the prediction results of the Lasso regression model for future data to the predicted difference to output the final prediction result.

[0077] It's important to note that covariates are variables used to predict outcomes. For example, in solar power generation, factors like sunlight, temperature, and pressure can be considered covariates. Appropriate use of covariates can significantly improve prediction accuracy. However, since future covariates are unknown, past covariates can be used as input. For instance, to predict power generation over the next ten days, data from the previous ten days can be considered as covariates for the next ten days. This method improves the application of covariates and avoids the problem of unknowable future covariates.

[0078] Example 2

[0079] The time series forecasting process is described below with reference to specific embodiments:

[0080] To predict solar power generation over the next 32 days, data on solar power generation, sunlight, and temperature from the previous year are available. 128 time points are selected and used as inputs to construct the entire prediction system.

[0081] The input data is X1 = [p] -127 ,p -127 [,…,p0] contains 128 data points, representing the power generation data of the past 128 days.

[0082] Input data 2 is X2 = [t] -159 , t -158 [t0, ..., t0] contains 160 data points, namely 128 historical data points plus 32 data points to be predicted, representing temperature, which are used as input covariates. A staggered correspondence is employed to avoid the problem of unavailable future covariates.

[0083] Input data 3 is X3 = [i -159 i -158 The data set [i, ..., i0] contains 160 data points, representing the sum of historical data and the number of data to be predicted, and is used as a covariate input. A staggered correspondence is employed to avoid the problem of unavailable future covariates.

[0084] The data was first trained using a LASSO linear regression model, which takes only a few tens of seconds to train. The LASSO model training used X2[0:128] as the first 128 data points of X2 and X3[0:128] as the first 128 data points of X3, with X1 as the training target. After training, a fitted model f(x) was generated. Inputting X2[0:128] and X3[0:128] into f(x) yielded y1, representing the predicted power generation over the past 128 days. The difference between y1 and the actual value X1 was calculated as d = y1 - X1. This d was then fed into a large time series model, with the number of output time points set to 32. The model output y2, a sequence of 32 numbers representing the prediction of the difference between the actual data and the regression model for the next 32 days. Then, X2[128:] and X3[128:] are input into the fitted model f(x), where the input consists of 32 data points and the output is y3, representing the power generation predicted by the regression model for the next 32 days using covariates. Since there is a difference between the regression model and the actual value, in order to correct the bias, the bias y2 predicted by the large time series model is added to y3, and then the result is output, i.e., y2+y3 is the final prediction result.

[0085] Example 3

[0086] Referring to Figure 4, this embodiment of the present disclosure provides a time series forecasting device 300 for the energy sector, including a processor 304 and a memory 301. Optionally, the device may further include a communication interface 302 and a bus 303. The processor 304, communication interface 302, and memory 301 can communicate with each other via the bus 303. The communication interface 302 can be used for information transmission. The processor 304 can call logical instructions in the memory 301 to execute the time series forecasting method for the energy sector described in the above embodiment.

[0087] Furthermore, the logic instructions in the aforementioned memory 301 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium.

[0088] The memory 301, as a computer-readable storage medium, can be used to store software programs and computer-executable programs, such as program instructions / modules corresponding to the methods in the embodiments of this disclosure. The processor 304 executes functional applications and data processing by running the program instructions / modules stored in the memory 301, thereby implementing the time series forecasting method for the energy field described in the above embodiments.

[0089] The memory 301 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the terminal device. Furthermore, the memory 301 may include high-speed random access memory and may also include non-volatile memory.

[0090] This disclosure also provides a computer-readable storage medium storing computer-executable instructions configured to perform the above-described time series forecasting method for the energy sector.

[0091] The aforementioned computer-readable storage medium may be a transient computer-readable storage medium or a non-transitory computer-readable storage medium.

[0092] The technical solutions of this disclosure can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in this disclosure. The aforementioned storage medium can be a non-transitory storage medium, including: a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, and other media capable of storing program code; it can also be a transient storage medium.

[0093] The foregoing description and accompanying drawings fully illustrate embodiments of this disclosure to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, procedural, and other changes. The embodiments represent only possible variations. Individual components and functions are optional unless explicitly required, and the order of operation may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the terminology used in this application is for describing embodiments only and is not intended to limit the claims. As used in the description of embodiments and claims, the singular forms “a,” “an,” and “the” are intended to equally include the plural forms unless the context clearly indicates otherwise. Similarly, the term “and / or” as used in this application means including one or more of the associated listed items and all possible combinations thereof. Additionally, when used in this application, the term "comprise" and its variations "comprises" and / or "comprising" refer to the presence of stated features, integrals, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof. Without further limitations, an element defined by the phrase "comprises a..." does not exclude the presence of other identical elements in the process, method, or apparatus that includes said element. In this document, each embodiment may focus on the differences from other embodiments, and similar or identical parts between embodiments can be referred to mutually. For methods, products, etc., disclosed in the embodiments, if they correspond to the method section disclosed in the embodiments, the relevant parts can be referred to the description of the method section.

[0094] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the embodiments of this disclosure. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0095] The methods and products (including but not limited to devices and equipment) disclosed in the embodiments herein can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For instance, the division of units may be merely a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the coupling or direct coupling or communication connection shown or discussed between each other may be through some interfaces, and the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to implement this embodiment according to actual needs. In addition, the functional units in the embodiments of this disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0096] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

Claims

1. A time series forecasting method for the energy sector, characterized in that, Includes the following steps: Collect time series data and trend images of power generation, and convert the trend images into time series data; Construct and train a Lasso regression model, and save the time series data of the difference between the time series and the prediction results of the Lasso regression model; Preprocessing of time series data and difference time series data includes data cleaning and standardization. The preprocessed data is input into a large time series model for model training. The large time series model is an Encoder-Decoder architecture. Power generation time series prediction is performed using a trained large time series model, including time series without covariates and time series with covariates.

2. The time series forecasting method for the energy sector according to claim 1, characterized in that, The time series forecasting process is as follows: For time series without covariates, input the historical time series into the time series prediction model and output the predicted time series; For time series data with covariates, input the historical time series and the corresponding covariates into the Lasso regression model for training to obtain prediction results for historical data and prediction results for future data; calculate the difference between the historical time series and the prediction results of the Lasso regression model for historical data, and feed the difference into the time series prediction model to obtain the prediction difference. The final prediction result is output by adding the difference between the Lasso regression model's prediction of future data and the prediction.

3. The time series forecasting method for the energy sector according to claim 1, characterized in that, The specific method for converting trend images into time series data is as follows: The trend chart is grayscaled and binarized to generate a binarized image; OpenCV's Hough transform is used to detect straight lines in the binarized image and identify the x-axis and y-axis; OCR is used to identify the scales and markings of the coordinate axes in the binarized image; and Cancy edge detection and connectivity analysis are used to trace the lines of the polyline chart. Identify the intersection points of the lines with the x-axis and y-axis directions, and calculate the coordinate value of each point accordingly; The extracted images are converted into time series.

4. The time series forecasting method for the energy sector according to claim 1, characterized in that, The time series large model encoder part includes a multi-layer Transformer encoder block, each block containing a multi-head attention mechanism and a feedforward neural network; The decoder section consists of multi-layer Transformer decoder blocks, each of which includes multi-head attention, contextual attention, and a feedforward neural network.

5. The time series forecasting method for the energy sector according to claim 4, characterized in that, The encoder also includes a position encoding module, which implements position encoding using a sine function, as shown in the following formula: Where d represents the embedding dimension and i represents the dimension index.

6. The time series forecasting method for the energy sector according to claim 4, characterized in that, The feedforward neural network employs two linear transformations and the ReLU activation function.

7. The time series forecasting method for the energy sector according to claim 1, characterized in that, During the training of the large time series model, a teacher-forced training method is used, with mean squared error (MAE) as the loss function. Reinforcement learning is also introduced, and the model parameters are updated using the policy gradient method, as shown in the following formula: Where, θ t+1 Let α represent the gradient at the next time step, and π represent the learning rate. θ Let a represent the policy function. t Let R represent the action and R represent the reward function.

8. The time series forecasting method for the energy sector according to claim 1, characterized in that, The covariates include light intensity, temperature, and pressure. For time series data containing covariates, a staggered correspondence method is used, where historical covariates are used as inputs for future predictions.

9. A time series forecasting device for the energy sector, comprising a processor and a memory storing program instructions, characterized in that, The processor is configured to execute, when running the program instructions, the time series forecasting method for the energy sector as described in any one of claims 1 to 8.

10. A storage medium storing program instructions, characterized in that, When the program instructions are executed, they perform the time series forecasting method for the energy sector as described in any one of claims 1 to 8.

Images