Wind and light load sequential sequence generation method and device, computer program product and storage medium

Abstract

The invention discloses a meteorological factor-based multi-time scale wind and light load sequential sequence generation method, which comprises the following steps of: firstly, constructing a wind and light load sequential sequence generation model, combining BiLSTM, MTL and Attention mechanisms by the model, and respectively mapping seasons, months, date types and meteorological factors as input characteristics into vector forms, so as to obtain a wind and light load sequential sequence generation model; on the basis of historical wind power, photovoltaic and load data, corresponding input characteristics are obtained, and a wind and light load sequential sequence is generated by using a wind and light load sequential sequence generation model; according to the method, the influence of meteorological factors on sequence generation and the internal relation of the wind-solar charge sequence are fully considered, and the accuracy of sequential sequence generation is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power system prediction, and in particular to a method, device, storage medium and computer program for generating a wind-solar-load timing sequence. Background Art

[0002] With the advancement of the "dual carbon" goals, my country's installed capacity of renewable energy has grown on a large scale and is gradually replacing traditional fossil energy. Among the many renewable energy sources, wind power and photovoltaics have developed particularly rapidly, but compared with traditional fossil energy, the output of renewable energy has significant randomness and volatility. With the grid connection of a large number of flexible loads represented by electric vehicles, the uncertainty on the load side of the power system is also increasing. Faced with the strong uncertainty of both source and load, it is necessary to establish a wind-light-load time series generation method that is sufficiently representative and can accurately restore historical output characteristics as the basis for the operation and planning of the power system. This is of great significance for the construction of a new power system.

[0003] At present, the relevant research on time series generation at home and abroad can be divided into two categories: model-based generation methods and data-driven generation methods. Model-based methods usually assume that wind, solar output and load demand meet a certain probability distribution. With the changes in geographical location, unit type, etc., the output / power consumption curves of wind, solar and load may not conform to the assumed general distribution, which will lead to a large gap between the generation results of the time series and the actual situation. Data-driven generation methods usually train neural networks based on relevant theories such as machine learning and deep learning, input historical data of wind, solar and load into the neural network, and obtain the future distribution of wind, solar and load.

[0004] Most of the existing wind-solar-load time series generation methods focus on one of the three types: wind, solar, and load, and few studies are conducted on wind, solar, and load at the same time. In addition, in the process of time series generation, the influencing factors are not considered comprehensively, and the accuracy of time series generation is low. Summary of the invention

[0005] In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention is: how to provide a method for generating a wind-solar-load timing sequence by combining multiple factors for research and improving the accuracy of timing sequence generation.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] A method for generating a wind-solar-load timing sequence, comprising:

[0008] Construct a wind-solar-load time series generation model, which includes three BiLSTM neural networks, which correspond to wind power sequence, photovoltaic sequence and load sequence respectively. The three BiLSTM neural networks are connected to the information sharing layer of the MTL neural network to realize information sharing interaction, and three Attention neural networks corresponding to wind power sequence, photovoltaic sequence and load sequence are established respectively. The Attention neural network is used to realize differentiated attention to shared information.

[0009] Season, month, date type and meteorological factors are used as input features and mapped into vector form input features respectively;

[0010] Based on the historical wind power, photovoltaic and load data, and obtaining the corresponding vector form input features, the wind-solar-load time series generation model is used to generate the wind-solar-load time series.

[0011] BiLSTM is an improvement on the long short-term neural network which only includes the forward transmission process. The network includes not only the forward transmission but also the back-propagation process. The specific structure is as follows: Figure 2 shown.

[0012] The formula for BiLSTM data processing is as follows:

[0013] f (t) =σ(W fy y (t-1) +W fx x (t) +b f ) (2)

[0014] i (t) =σ(W iy y (t-1) +W ix x (t) +b i ) (3)

[0015] u (t) =tanh(W uy y (t-1) +W ux x (t) +b u ) (4)

[0016] c (t) =i (t) ·u (t) +f t ·c (t-1) (5)

[0017] o (t) =σ(W oy y(t-1) +W ox x (t) +b o ) (6)

[0018] y (t) =o (t) ·tanh(c (t) ) (7)

[0019] In the formula, i (t) ,u (t) ,c (t) ,o (t) Respectively represent the input gate, forget gate, cell state and output gate; W is the variable weight factor, b is the bias matrix, σ and tanh represent the sigmoid and tanh activation functions respectively; x(t) and y(t) represent the input and output variables respectively; through the above calculations, the final input result of the BiLSTM model can be expressed by the following function:

[0020]

[0021] In the formula, and Represent the output results of forward LSTM and reverse LSTM respectively, W y′f and W y′b is the weight matrix, b y′ is the output deviation matrix; g() is the activation function of the output layer. Since the output sequence generation must take into account the speed, the present invention selects ReLu, a function with a small amount of calculation, as the model activation function.

[0022] In the feature selection part, the present invention selects multiple feature quantities such as season, date and meteorological factors to generate wind-solar-load sequence. However, these feature variables have different effects on the final result. For example, wind speed and air pressure have a greater impact on wind power output, while radiation intensity and temperature have a greater impact on photovoltaic output. In view of this, in order to fully consider the differences in the degree of influence of different factors, in the present invention, a multi-task learning (MTL) structure based on the attention mechanism is used to improve the generation quality of wind-solar-load sequence.

[0023] Multi-task learning is a type of transfer learning. It uses a unique sharing theory to train multiple tasks, namely wind power output, photovoltaic output and load demand in this invention. During the training process, the correlation of different tasks is fully involved to obtain hidden correlation information, thereby improving the generalization ability of the model.

[0024] Wind-solar-load sequence generation includes multiple subtasks, each of which predicts different categories of time series. Therefore, each subtask should have different focuses on shared features. As an effective screening mechanism for shared features, the attention mechanism can provide differentiated feature pools for different subtasks.

[0025] The attention mechanism, by imitating the human attention mechanism, assigns different weight parameters to each input feature, thereby achieving the effect of focusing on valid information and ignoring invalid information. The specific structure is as follows Figure 3 As shown, the calculation formula is shown in formula (9).

[0026]

[0027] Among them, V represents the input feature, Qi represents the i-th feature sequence of the input feature; Softmax(Q i V T ) is the weight sequence of the input features; in the actual training process, the higher the calculated attention weight is, the greater the contribution of the feature to the output node is.

[0028] By combining BiLSTM, MTL and Attention mechanisms, the present invention establishes a BiLSTM-MTL-Attention model. Through the layer-by-layer progressive model framework structure of BiLSTM-MTL-Attention, the similarities between different sequences can be more reasonably and efficiently utilized. Compared with the separate training of wind power, photovoltaic and load subtasks, the training performance can be improved and the sequence generation quality can be improved.

[0029] As an optimization, the season includes four factors: spring, summer, autumn and winter; the month includes three factors: first month, second month and third month; the date type includes two factors: working day and non-working day; the meteorological factors include six factors: temperature, humidity, air pressure, wind speed, dew point and weather type;

[0030] Months are represented by 3D 0-1 vectors, with the first month being [1,0,0], the second month being [0,1,0], and the third month being [0,0,1]; date types are represented by 2D 0-1 vectors, with working days being [1,0] and non-working days being [0,1]; weather types are represented by 4D 0-1 vectors, with sunny days being [1,0,0,0], partially cloudy days being [0,1,0,0], mostly cloudy days being [0,0,1,0], and rainy and snowy days being [0,0,0,1];

[0031] Temperature, humidity, air pressure, wind speed and dew point are continuous variables. Formula (1) is used to project the values ​​of these variables into the interval [0,1], where x represents the value of the variable before projection, x′ represents the value after projection, and xmin and x max They represent the minimum and maximum values ​​that the variable can take, respectively. Before projection, the variable values ​​are all selected as the average value of the day.

[0032]

[0033] As an optimization, wind power, photovoltaic and load data from a period of time in history are selected, and combined with the corresponding input features, they are sampled at intervals as a data set. In the data set, a part is selected as the test set according to the time length, and the rest is used as the training set.

[0034] The present invention also discloses a wind-solar-load timing sequence generating device, comprising:

[0035] A model building module is used to build a wind-solar-load time series generation model. The wind-solar-load time series generation model includes at least one BiLSTM neural network. Each BiLSTM neural network corresponds to a wind power sequence, a photovoltaic sequence, and a load sequence. The BiLSTM neural network is connected to the information sharing layer of the MTL neural network to realize information sharing interaction, and establishes an Attention neural network corresponding to the wind power sequence, the photovoltaic sequence, and the load sequence. The Attention neural network is used to realize differentiated attention to shared information.

[0036] The input feature acquisition module is used to take season, month, date type and meteorological factors as input features and map them into vector input features respectively;

[0037] The execution module is used to obtain the input features in the form of corresponding vectors based on the historical wind power, photovoltaic and load data, and generate a wind-solar-load time series sequence using a wind-solar-load time series sequence generation model.

[0038] As an optimization, the season includes four factors: spring, summer, autumn and winter; the month includes three factors: first month, second month and third month; the date type includes two factors: working day and non-working day; the meteorological factors include six factors: temperature, humidity, air pressure, wind speed, dew point and weather type;

[0039] Months are represented by 3D 0-1 vectors, with the first month being [1,0,0], the second month being [0,1,0], and the third month being [0,0,1]; date types are represented by 2D 0-1 vectors, with working days being [1,0] and non-working days being [0,1]; weather types are represented by 4D 0-1 vectors, with sunny days being [1,0,0,0], partially cloudy days being [0,1,0,0], mostly cloudy days being [0,0,1,0], and rainy and snowy days being [0,0,0,1];

[0040] Temperature, humidity, air pressure, wind speed and dew point are continuous variables. Formula (1) is used to project the values ​​of these variables into the interval [0,1], where x represents the value of the variable before projection, x′ represents the value after projection, and x min and x max They represent the minimum and maximum values ​​that the variable can take, respectively. Before projection, the variable values ​​are all selected as the average value of the day.

[0041]

[0042] As an optimization, wind power, photovoltaic and load data from a period of time in history are selected, and combined with the corresponding input features, they are sampled at intervals as a data set. In the data set, a part is selected as the test set according to the time length, and the rest is used as the training set.

[0043] A computer program product comprises a computer program, and when the computer program is executed by a computer, the method described above is implemented.

[0044] A computer-readable storage medium stores a computer program, which, when executed by a computer, implements the method described above.

[0045] Compared with the prior art, the present invention has the following advantages: The present invention constructs a three-layer neural network model based on "bidirectional long-term and short-term neural network-multi-task learning-attention mechanism" to achieve wind-solar-load time series generation. The method fully considers the impact of meteorological factors on sequence generation and the inherent connection between wind-solar-load sequences, greatly improving the accuracy of time series generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a framework diagram of the wind-solar-load timing sequence generation model in the present invention;

[0047] Figure 2 It is a structural diagram of the BiLSTM network in the present invention;

[0048] Figure 3 It is a structural diagram of the Attention network in the present invention;

[0049] Figure 4 It is the wind power sequence distribution diagram under different models in the present invention;

[0050] Figure 5 It is the photovoltaic sequence distribution diagram under different models in the present invention;

[0051] Figure 6 It is the load sequence distribution diagram under different models in the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical scheme and advantages of the embodiments of the present invention clearer, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work belong to the scope of protection of the present invention.

[0053] The method for generating a wind-solar load timing sequence in this specific implementation manner includes:

[0054] Construct a wind-solar-load time series generation model, which includes at least one BiLSTM neural network. Each BiLSTM neural network corresponds to a wind power sequence, a photovoltaic sequence, and a load sequence. The BiLSTM neural network is connected to the information sharing layer of the MTL neural network to realize information sharing interaction. In addition, an Attention neural network corresponding to the wind power sequence, the photovoltaic sequence, and the load sequence is established. The Attention neural network is used to realize differentiated attention to shared information.

[0055] Season, month, date type and meteorological factors are used as input features and mapped into vector form input features respectively;

[0056] Based on the historical wind power, photovoltaic and load data, and obtaining the corresponding vector form input features, the wind-solar-load time series generation model is used to generate the wind-solar-load time series.

[0057] In this specific implementation, the season includes four factors: spring, summer, autumn and winter; the month includes three factors: first month, second month and third month; the date type includes two factors: working day and non-working day; the meteorological factor includes six factors: temperature, humidity, air pressure, wind speed, dew point and weather type;

[0058] Months are represented by 3D 0-1 vectors, with the first month being [1,0,0], the second month being [0,1,0], and the third month being [0,0,1]; date types are represented by 2D 0-1 vectors, with working days being [1,0] and non-working days being [0,1]; weather types are represented by 4D 0-1 vectors, with sunny days being [1,0,0,0], partially cloudy days being [0,1,0,0], mostly cloudy days being [0,0,1,0], and rainy and snowy days being [0,0,0,1];

[0059] Temperature, humidity, air pressure, wind speed and dew point are continuous variables. Formula (1) is used to project the values ​​of these variables into the interval [0,1], where x represents the value of the variable before projection, x′ represents the value after projection, and x min and x max They represent the minimum and maximum values ​​that the variable can take, respectively. Before projection, the variable values ​​are all selected as the average value of the day.

[0060]

[0061] In this specific implementation, wind power, photovoltaic and load data of a period of time in history are selected, and combined with corresponding input features, sampling is performed at intervals as a data set. A part of the data set is selected as a test set according to the time length, and the rest is used as a training set.

[0062] During the specific implementation process, based on the open source meteorological factors, wind power, photovoltaic and load measured data, all historical data from 2018 to 2019 were selected as the data set with a step size of 15 minutes. Two months of data were extracted from the data set as the test set, and the rest were used as the training set.

[0063] In order to fully evaluate the effect of the present invention, the following five models are set as comparative examples:

[0064] Model 1: Traditional probability sampling model;

[0065] Model 2: LSTM model;

[0066] Model 3: BiLSTM model;

[0067] Model 4: BiLSTM-MTL model;

[0068] Model 5: Wind-solar-load time series generation model (BiLSTM-MTL-Attention);

[0069] Model 1 uses the traditional probability sampling method to generate sequences, where the wind speed uses Weibull distribution, the solar radiation intensity uses Beta distribution, and the load uses normal distribution. The parameter settings of each distribution are shown in Table 1:

[0070]

[0071] Table 1

[0072] The sequences generated by the five models are compared with the historical real output sequences. The results are as follows: Figures 4 to 6 As shown, comparative analysis Figure 4 , 5, 6 found that: the sequence generated by using model 1 has the largest deviation from the true sequence, the sequence generated by using model 5, i.e. the model proposed in the present invention, is closest to the true sequence distribution, and the sequence generation effects of models 2, 3 and 4 are gradually improved, which reflects the great role of bidirectional long-term and short-term neural networks, multi-task learning and attention mechanism in improving the quality of wind-solar-load sequence generation.

[0073] Table 2 compares the sequence error values ​​generated by the five models, considering two error indicators, namely MAPE (mean absolute percentage error) and RMSE (root mean square error). The calculation formulas of each indicator are shown in formula (10) and formula (11). i and pi represent the real sequence output and the generated sequence output respectively.

[0074]

[0075]

[0076] As can be seen from Table 2, the training error of model 5 is greatly reduced compared with models 1, 2, 3, and 4, which reflects that the present invention can greatly improve the accuracy of wind-solar-load time series generation.

[0077]

[0078] Table 2

[0079] A multi-time scale wind-solar-load time series generation device based on meteorological factors, comprising:

[0080] A model building module is used to build a wind-solar-load time series generation model. The wind-solar-load time series generation model includes at least one BiLSTM neural network. Each BiLSTM neural network corresponds to a wind power sequence, a photovoltaic sequence, and a load sequence. The BiLSTM neural network is connected to the information sharing layer of the MTL neural network to realize information sharing interaction, and establishes an Attention neural network corresponding to the wind power sequence, the photovoltaic sequence, and the load sequence. The Attention neural network is used to realize differentiated attention to shared information.

[0081] The input feature acquisition module is used to take season, month, date type and meteorological factors as input features and map them into vector input features respectively;

[0082] The execution module is used to obtain the input features in the form of corresponding vectors based on the historical wind power, photovoltaic and load data, and generate a wind-solar-load time series sequence using a wind-solar-load time series sequence generation model.

[0083] The seasons include four factors: spring, summer, autumn and winter; the months include three factors: first month, second month and third month; the date types include two factors: working day and non-working day; the meteorological factors include six factors: temperature, humidity, air pressure, wind speed, dew point and weather type;

[0084] Months are represented by 3D 0-1 vectors, with the first month being [1,0,0], the second month being [0,1,0], and the third month being [0,0,1]; date types are represented by 2D 0-1 vectors, with working days being [1,0] and non-working days being [0,1]; weather types are represented by 4D 0-1 vectors, with sunny days being [1,0,0,0], partially cloudy days being [0,1,0,0], mostly cloudy days being [0,0,1,0], and rainy and snowy days being [0,0,0,1];

[0085] Temperature, humidity, air pressure, wind speed and dew point are continuous variables. Formula (1) is used to project the values ​​of these variables into the interval [0,1], where x represents the value of the variable before projection, x′ represents the value after projection, and x min and x max They represent the minimum and maximum values ​​that the variable can take, respectively. Before projection, the variable values ​​are all selected as the average value of the day.

[0086]

[0087] Wind power, photovoltaic and load data of a period of time in history are selected, and combined with the corresponding input features, sampling is performed at intervals as a data set. A part of the data set is selected as a test set according to the time length, and the rest is used as a training set.

[0088] A computer program product comprises a computer program, and when the computer program is executed by a computer, the method described above is implemented.

[0089] A computer-readable storage medium stores a computer program, which, when executed by a computer, implements the method described above.

[0090] Those skilled in the art will appreciate that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the present application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code. The scheme in the embodiments of the present application can be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0091] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0092] 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 including 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.

[0093] 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 in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0094] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0095] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for generating a wind-solar load timing sequence, characterized in that: include: Construct a wind-solar-load time series generation model, which includes at least one BiLSTM neural network. Each BiLSTM neural network corresponds to a wind power sequence, a photovoltaic sequence, and a load sequence. The BiLSTM neural network is connected to the information sharing layer of the MTL neural network to realize information sharing interaction. In addition, an Attention neural network corresponding to the wind power sequence, the photovoltaic sequence, and the load sequence is established. The Attention neural network is used to realize differentiated attention to shared information. Season, month, date type and meteorological factors are used as input features and mapped into vector form input features respectively; Based on the historical wind power, photovoltaic and load data, and obtaining the corresponding vector form input features, the wind-solar-load time series generation model is used to generate the wind-solar-load time series.

2. The method for generating a wind-solar-load timing sequence according to claim 1, characterized in that: The seasons include four factors: spring, summer, autumn and winter; the months include three factors: first month, second month and third month; the date types include two factors: working day and non-working day; the meteorological factors include six factors: temperature, humidity, air pressure, wind speed, dew point and weather type; Months are represented by 3D 0-1 vectors, with the first month being [1,0,0], the second month being [0,1,0], and the third month being [0,0,1]. Date types are represented by 2D 0-1 vectors, with working days being [1,0] and non-working days being [0,1]. Weather types are represented by 4-dimensional 0-1 vectors, with sunny days represented as [1,0,0,0], partially cloudy days represented as [0,1,0,0], mostly cloudy days represented as [0,0,1,0], and rainy and snowy days represented as [0,0,0,1]; Temperature, humidity, air pressure, wind speed and dew point are continuous variables. Formula (1) is used to project the values ​​of these variables into the interval [0,1], where x represents the value of the variable before projection, x′ represents the value after projection, and x min and x max They represent the minimum and maximum values ​​that the variable can take, respectively. Before projection, the variable values ​​are all selected as the average value of the day.

3. The method for generating a wind-solar-load timing sequence according to claim 1, characterized in that: Wind power, photovoltaic and load data of a period of time in history are selected, and combined with the corresponding input features, sampling is performed at intervals as a data set. A part of the data set is selected as a test set according to the time length, and the rest is used as a training set.

4. A wind-solar load timing sequence generation device, characterized in that: include: A model building module is used to build a wind-solar-load time series generation model. The wind-solar-load time series generation model includes at least one BiLSTM neural network. Each BiLSTM neural network corresponds to a wind power sequence, a photovoltaic sequence, and a load sequence. The BiLSTM neural network is connected to the information sharing layer of the MTL neural network to realize information sharing interaction, and establishes an Attention neural network corresponding to the wind power sequence, the photovoltaic sequence, and the load sequence. The Attention neural network is used to realize differentiated attention to shared information. The input feature acquisition module is used to take season, month, date type and meteorological factors as input features and map them into vector input features respectively; The execution module is used to obtain the input features in the form of corresponding vectors based on the historical wind power, photovoltaic and load data, and generate a wind-solar-load time series sequence using a wind-solar-load time series sequence generation model.

5. The wind-solar-load timing sequence generation device according to claim 4 is characterized in that: The seasons include four factors: spring, summer, autumn and winter; the months include three factors: first month, second month and third month; the date types include two factors: working day and non-working day; the meteorological factors include six factors: temperature, humidity, air pressure, wind speed, dew point and weather type; Months are represented by 3D 0-1 vectors, with the first month being [1,0,0], the second month being [0,1,0], and the third month being [0,0,1]. Date types are represented by 2D 0-1 vectors, with working days being [1,0] and non-working days being [0,1]. Weather types are represented by 4-dimensional 0-1 vectors, with sunny days represented as [1,0,0,0], partially cloudy days represented as [0,1,0,0], mostly cloudy days represented as [0,0,1,0], and rainy and snowy days represented as [0,0,0,1]; Temperature, humidity, air pressure, wind speed and dew point are continuous variables. Formula (1) is used to project the values ​​of these variables into the interval [0,1], where x represents the value of the variable before projection, represents the value after projection, and represents the minimum and maximum values ​​of the variable, respectively. The values ​​of the variables before projection are all selected as the average value of the day.

6. The wind-solar-load timing sequence generation device according to claim 4 is characterized in that: Wind power, photovoltaic and load data of a period of time in history are selected, and combined with the corresponding input features, sampling is performed at intervals as a data set. A part of the data set is selected as a test set according to the time length, and the rest is used as a training set.

7. A computer program product, characterized in that: The method comprises a computer program, which, when executed by a computer, implements the method according to any one of claims 1 to 3.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a computer, the method according to any one of claims 1 to 3 is implemented.

Images