Neural network model generation method, apparatus and device, and readable storage medium

Abstract

The invention provides a neural network model generation method, device and equipment and a readable storage medium, and the method comprises the steps: obtaining the power consumption information of a preset power system at a plurality of dates, carrying out the following processing of each power consumption information, obtaining the characteristics of each power consumption information, creating a deep echo state network, and generating a neural network model; and training the deep echo state network based on the features in each piece of power consumption information, wherein the deep echo state network is the neural network model. The deep echo state network can accurately predict the real-time electricity price.

Description

Technical Field

[0001] The present invention relates to the technical field of electricity price prediction, and in particular to a method, device, equipment and readable storage medium for generating a neural network model. Background Art

[0002] At present, electricity prices are the fulcrum of the electricity market. All participants in the electricity market are based on electricity prices. One of the core tasks of electricity marketization is to build a spot market, which mainly includes the day-ahead market, the intraday market and the real-time market. Accurately predicting the day-ahead electricity price one day in advance is crucial for formulating the power generation plan of the power generation system and reasonably adjusting the power generation system. In addition, with the rapid development of the power market and the large-scale access of renewable energy, the real-time electricity price in the power market has become increasingly volatile and uncertain, which has brought great challenges to the operation of the power system and the decision-making of market participants. Traditional electricity price forecasting methods often fail to fully capture the complexity and dynamics of electricity price data, resulting in low accuracy of the forecast results.

[0003] Therefore, developing an efficient model that can accurately predict real-time electricity prices has become an urgent problem to be solved. Summary of the invention

[0004] The object of the present invention is to provide a method, device, equipment and readable storage medium for generating a neural network model.

[0005] In order to achieve one of the above-mentioned invention purposes, an embodiment of the present invention provides a method for generating a neural network model for real-time electricity price prediction, comprising the following steps: obtaining power consumption information of a preset power system on multiple dates, each power consumption information comprising: loads L1, L2, ..., L 24 and the actual electricity prices Y1, Y2, ..., Y 24 , where L t represents the load at the tth hour, Y t represents the actual electricity price at hour t, 1≤t≤24; for any t, calculate Among them, the actual electricity prices are Y1, Y2, ..., Y 24 The average value is μ Y , standard deviation is σ Y , load L1, L2, ..., L 24 The average value is μ L , standard deviation is σ L ; Each power consumption information is processed as follows: Calculate

[0006] The minimum k value, where the actual electricity prices are Y1, Y2, ..., Y 24 The average value of k is a natural number; create a preset two-dimensional space and a grid located in the two-dimensional space,

[0007] ″″″″

[0008] The actual electricity prices Y1, Y2, ..., Y 24 and loads L1, L2, ..., L 24 Distribute in the grid; calculate Y t and L t Mutual information where p(Y′,L′) is Y t ′ and L′ t The joint probability of appearing in the same grid, p(Y′) and p(L′) are marginal probabilities; by adjusting the size and segmentation of the grid, MIC(Y t ,L t ) is maximized, Select one actual electricity price from each grid to get Num1 actual electricity prices Among them, Num1 is a natural number; create a deep echo state network based on the features in each power consumption information The deep echo state network is trained, and the deep echo state network is the neural network model.

[0009] As a further improvement of one embodiment of the present invention, the average value Standard Deviation average value Standard Deviation

[0010] As a further improvement of an embodiment of the present invention, the state update formula of the P-th layer reserve pool in the deep echo state network is: Among them, W in,P is the input weight matrix of the P-th layer reservoir, W (P) is the internal weight of the P-th layer reserve pool, X (P-1) (t) is the state of the P-1th layer reservoir at time t, X (P) (t) is the state of the P-th layer reserve pool at time t, and the number of layers of the reserve pool in the deep echo state network is Num2, Num2≤k.

[0011] As a further improvement of an embodiment of the present invention, the output of the deep echo state network is obtained by passing the output weight matrix W through the last layer of the reserve pool state. out The electricity price forecast Y″(t)=W is calculated out X (Num2) (t), the training objective of the deep echo state network is to minimize the difference between the predicted output Y″(t) and the actual electricity price Y tThe mean square error between

[0012] The embodiment of the present invention also provides a device for generating a neural network model for real-time electricity price prediction, comprising the following modules: an information acquisition module, for acquiring power consumption information of a preset power system on multiple dates, each power consumption information including: loads L1, L2, ..., L 24 and the actual electricity prices Y1, Y2, ..., Y 24 , where L t represents the load at the tth hour, Y t represents the actual electricity price at hour t, 1≤t≤24; for any t, calculate Among them, the actual electricity prices are Y1, Y2, ..., Y 24 The average value is μ Y , standard deviation is σ Y , load L1, L2, ..., L 24 The average value is μ L , standard deviation is σ L ; Information processing module, used to perform the following processing on each power consumption information: calculation The minimum k value, where the actual electricity price is Y1′, Y2′, ..., Y′ 24 The average value of k is a natural number; create a preset two-dimensional space and a grid located in the two-dimensional space, and in the preset two-dimensional space, the actual electricity prices Y1′, Y2′, ..., Y′ 24 and loads L1′, L2′, ..., L′ 24 Distribute in the grid; calculate Y t ′ and L′ t Mutual information where p(Y′,L′) is Y t ′ and L′ t The joint probability of appearing in the same grid, p(Y′) and p(L′) are marginal probabilities; by adjusting the size and segmentation of the grid, MIC(Y′ t ,L′ t ) is maximized, Select one actual electricity price from each grid to get Num1 actual electricity prices Among them, Num1 is a natural number; the model creation module is used to create a deep echo state network based on the features in each power consumption information The deep echo state network is trained, and the deep echo state network is the neural network model.

[0013] As a further improvement of one embodiment of the present invention, the average value Standard Deviation average value Standard Deviation

[0014] As a further improvement of an embodiment of the present invention, the state update formula of the P-th layer reserve pool in the deep echo state network is: Among them, W in,P is the input weight matrix of the P-th layer reservoir, W (P) is the internal weight of the P-th layer reserve pool, X (P-1) (t) is the state of the P-1th layer reservoir at time t, X (P) (t) is the state of the P-th layer reserve pool at time t, and the number of layers of the reserve pool in the deep echo state network is Num2, Num2≤k.

[0015] As a further improvement of an embodiment of the present invention, the output of the deep echo state network is obtained by passing the output weight matrix W through the last layer of the reserve pool state. out The electricity price forecast Y″(t)=W is calculated out X (Num2) (t), the training objective of the deep echo state network is to minimize the difference between the predicted output Y″(t) and the actual electricity price Y t The mean square error between

[0016] An embodiment of the present invention further provides a device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the above-mentioned generation method.

[0017] An embodiment of the present invention further provides a readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned generation method is implemented.

[0018] Compared with the prior art, the technical effect of the present invention is that: the embodiment of the present invention provides a method, device, equipment and readable storage medium for generating a neural network model, the generating method includes: obtaining power consumption information of multiple dates of a preset power system, performing the following processing on each power consumption information, and obtaining the characteristics of each power consumption information, creating a deep echo state network, and training the deep echo state network based on the characteristics in each power consumption information, the deep echo state network is the neural network model. The deep echo state network can accurately predict the real-time electricity price. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flowchart of a method for generating a neural network model in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The present invention will be described in detail below in conjunction with the various embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional changes made by a person skilled in the art based on these embodiments are all within the scope of protection of the present invention.

[0021] As used herein, terms indicating spatial relative positions such as "upper," "above," "lower," and "below" are used for the purpose of convenience to describe the relationship of one unit or feature relative to another unit or feature as shown in the accompanying drawings. Terms of spatial relative position may be intended to include different orientations of the device in use or operation in addition to the orientation shown in the drawings. For example, if the device in the figure is turned over, units described as being "below" or "beneath" other units or features will be located "above" the other units or features. Thus, the exemplary term "below" can encompass both above and below orientations. The device may be oriented in other ways (rotated 90 degrees or in other orientations), and the spatially relative descriptors used herein interpreted accordingly.

[0022] Furthermore, it should be understood that although the terms first, second, etc. may be used herein to describe various elements or structures, these described objects should not be limited by these terms. These terms are only used to distinguish these described objects from each other. For example, a first end may be referred to as a second end, and similarly a second end may also be referred to as a first end, without departing from the scope of protection of this application.

[0023] Embodiment 1 of the present invention provides a method for generating a neural network model for real-time electricity price prediction, such as Figure 1 As shown, the following steps are included:

[0024] Step 101: Obtain power consumption information of a preset power system on multiple dates, each of which includes: loads L1, L2, ..., L 24 and the actual electricity prices Y1, Y2, ..., Y 24 , where L t represents the load at the tth hour, Y t represents the actual electricity price at hour t, 1≤t≤24; for any t, calculate Among them, the actual electricity prices are Y1, Y2, ..., Y 24 The average value is μ Y , standard deviation is σ Y , load L1, L2, ..., L 24 The average value is μ L , standard deviation is σ LHere, the first hour is from 0:00 to 1:00, the second hour is from 1:00 to 2:00, ..., and the 23rd hour is from 23:00 to 24:00. Since the units of electricity price and load are different, in order to facilitate subsequent processing, both electricity price and load are normalized.

[0025] Step 102: Perform the following processing on each power consumption information: Calculate The minimum k value, where the actual electricity price is Y1′, Y2′, ..., Y′ 24 The average value of k is a natural number; create a preset two-dimensional space and a grid located in the two-dimensional space, and in the preset two-dimensional space, the actual electricity prices Y1′, Y2′, ..., Y′ 24 and loads L′1, L′2, ..., L′ 24 Distribute in the grid; calculate Y t ′ and L′ t Mutual information where p(Y′,L′) is Y t ′ and L′ t The joint probability of appearing in the same grid, p(Y′) and p(L′) are marginal probabilities; by adjusting the size and segmentation of the grid, MIC(Y t ′,L t ′) is maximized, Select one actual electricity price from each grid to get Num1 actual electricity prices Where Num1 is a natural number; here, the range of k is usually selected from 1 to several hours (i.e. 1≤k<10) to find autocorrelation in the short term. ACF (Auto Correlation Function) is used to identify the autocorrelation characteristics in the electricity price time series, especially the lag effect.

[0026] Step 103: Create a deep echo state network based on the features in each power consumption information The deep echo state network is trained, and the deep echo state network is the neural network model.

[0027] It can be understood as some standardized load data L1, L2, ..., L selected in the feature extraction process 24 These characteristics are usually load data that have an important impact on electricity price prediction.

[0028] This generation method has the following advantages: (1) Improving prediction accuracy: The deep echo state network is used to learn the deep features of the electricity price time series and capture the inherent laws of electricity price changes, thereby improving the prediction accuracy; (2) Enhancing the generalization ability of the model: Using ACF and MIC for feature extraction, the most relevant factors for electricity price prediction are automatically identified, and the model's prediction ability for unknown data is enhanced; (3) Providing real-time prediction support: The designed system can process real-time data and quickly generate electricity price prediction results, providing a scientific basis for real-time decision-making in the power market.

[0029] In this embodiment, the average Standard Deviation average value Standard Deviation

[0030] In this embodiment, the state update formula of the P-th layer reservoir in the deep echo state network is: Among them, W in,P is the input weight matrix of the P-th layer reservoir, W (P) is the internal weight of the P-th layer reserve pool, X (P-1) (t) is the state of the P-1th layer reservoir at time t, X (P) (t) is the state of the P-th layer reserve pool at time t, and the number of layers of the reserve pool in the deep echo state network is Num2, Num2≤k.

[0031] Here, W in,P is the input weight matrix of the P-th layer reserve pool. It maps the input data to the state space of the P-th layer reserve pool. The weight matrix contains two parts of input: the standardized electricity price data Y′ t+k-1 and selected load characteristics This is an input vector containing the normalized electricity price data Y′ at the previous time step t+k-1 and the load characteristics selected for the current time step

[0032] In this embodiment, the output of the deep echo state network is the output weight matrix W of the last layer of the reservoir state. out The electricity price forecast Y″(t)=W is calculated out X (Num2) (t), the training objective of the deep echo state network is to minimize the difference between the predicted output Y″(t) and the actual electricity price Y t The mean square error between

[0033] Through ACF(k) and MIC(Y t ′,L t′), as well as the construction and training of the deep echo state network model, this process provides a comprehensive solution for real-time electricity price prediction in the power market.

[0034] In the inventor's experiment, there is an electricity market price data and load data set, including the electricity price Y for 24 hours a day t and load L t Data. The specific data are as follows (assuming the data is before standardization):

[0035] Electricity price data Y t 45, 47, 50, 52, 49, 48, 53, 55, 58, 60, 62, 65, 67, 70, 68, 66, 64, 62, 61, 59, 57, 54, 52, 50; Load data L t It is: 200, 205, 210, 215, 220, 225, 230, 235, 240, 245, 250, 255, 260, 265, 270, 275, 280, 285, 290, 295, 300, 305, 310, 315.

[0036] Calculate the mean and standard deviation of electricity price and load, and then perform standardization: Mean electricity price Standard Deviation Load average and standard deviation Standardized data

[0037] Afterwards, the Deep Echo State Network (DeepESN) model training and prediction, a two-layer deep echo state network was constructed, and the standardized data was used for training and prediction.

[0038] Experimental results,After the model was trained, the electricity price of the last two hours was predicted. The prediction results are as follows: the actual value of the electricity price for 23 hours: 52, the actual value of the electricity price for 24 hours: 50. Prediction results: The predicted electricity price for 23 hours: 53.5, the predicted electricity price for 24 hours: 51.2.

[0039] Embodiment 2 of the present invention provides a device for generating a neural network model for real-time electricity price prediction, comprising the following modules:

[0040] The information acquisition module is used to acquire power consumption information of a preset power system on multiple dates, each of which includes: loads L1, L2, ..., L 24 and the actual electricity prices Y1, Y2, ..., Y 24 , where L t represents the load at the tth hour, Y trepresents the actual electricity price at hour t, 1≤t≤24; for any t, calculate Among them, the actual electricity prices are Y1, Y2, ..., Y 24 The average value is μ Y , standard deviation is σ Y , load L1, L2, ..., L 24 The average value is μ L , standard deviation is σ L ;

[0041] The information processing module is used to perform the following processing on each power consumption information: calculate The minimum k value, where the actual electricity price is Y1′, Y2′, ..., Y′ 24 The average value of k is a natural number; create a preset two-dimensional space and a grid located in the two-dimensional space, and in the preset two-dimensional space, the actual electricity prices Y1′, Y2′, ..., Y′ 24 and loads L1′, L2′, ..., L′ 24 Distribute in the grid; calculate Y t ′ and L′ t Mutual information where p(Y′,L′) is Y t ′ and L′ t The joint probability of appearing in the same grid, p(Y′) and p(L′) are marginal probabilities; by adjusting the size and segmentation of the grid, MIC(Y t ′,L t ′) is maximized, Select one actual electricity price from each grid to get Num1 actual electricity prices Among them, Num1 is a natural number;

[0042] Model creation module, used to create a deep echo state network based on the features in each power consumption information The deep echo state network is trained, and the deep echo state network is the neural network model.

[0043] In this embodiment, the average Standard Deviation average value Standard Deviation

[0044] In this embodiment, the state update formula of the P-th layer reservoir in the deep echo state network is: Among them, W in,P is the input weight matrix of the P-th layer reservoir, W (P) is the internal weight of the P-th layer reserve pool, X(P-1) (t) is the state of the P-1th layer reservoir at time t, X (P) (t) is the state of the P-th layer reserve pool at time t, and the number of layers of the reserve pool in the deep echo state network is Num2, Num2≤k.

[0045] In this embodiment, the output of the deep echo state network is the output weight matrix W of the last layer of the reservoir state. out The electricity price forecast Y″(t)=W is calculated out X (Num2) (t), the training objective of the deep echo state network is to minimize the difference between the predicted output Y″(t) and the actual electricity price Y t The mean square error between

[0046] Embodiment 3 of the present invention provides a device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the generation method in embodiment 1.

[0047] Embodiment 4 of the present invention provides a readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the generating method in embodiment 1 is implemented.

[0048] 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.

[0049] 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.

[0050] 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.

[0051] 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.

[0052] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention.

Claims

1. A method for generating a neural network model for real-time electricity price prediction, characterized in that: The following steps are involved: Obtain power consumption information of a preset power system on multiple dates, each power consumption information including: loads L1, L2, ..., L of the power system on the corresponding date 24 and the actual electricity prices Y1, Y2, ..., Y 24 , where L t represents the load at the tth hour, Y t represents the actual electricity price at hour t, 1≤t≤24; for any t, calculate Among them, the actual electricity prices are Y1, Y2, ..., Y 24 The average value is μ Y , standard deviation is σ Y , load L1, L2, ..., L 24 The average value is μ L , standard deviation is σ L ; The following processing is performed on each power consumption information: calculation The minimum k value, where the actual electricity price is Y′1, Y′2, ..., Y′ 24 The average value of k is a natural number; create a preset two-dimensional space and a grid located in the two-dimensional space, and in the preset two-dimensional space, the actual electricity prices Y′1, Y′2, ..., Y′ 24 and loads L′1, L′2, ..., L′ 24 Distribute in the grid; calculate Y′ t and L′ t Mutual information where p(Y′,L′) is Y′ t and L′ t The joint probability of appearing in the same grid, p(Y′) and p(L′) are marginal probabilities; by adjusting the size and segmentation of the grid, MIC(Y′ t ,L′ t ) is maximized, Select one actual electricity price from each grid to get Num1 actual electricity prices Among them, Num1 is a natural number; Create a deep echo state network based on the features in each power consumption information The deep echo state network is trained, and the deep echo state network is the neural network model.

2. The generation method according to claim 1, characterized in that: average value Standard Deviation average value Standard Deviation 3. The generation method according to claim 1, characterized in that: The state update formula of the Pth layer reservoir in the deep echo state network is: Among them, W in,P is the input weight matrix of the P-th layer reservoir, W (P) is the internal weight of the P-th layer reserve pool, X (P-1) (t) is the state of the P-1th layer reservoir at time t, X (P) (t) is the state of the P-th layer reserve pool at time t, and the number of layers of the reserve pool in the deep echo state network is Num2, Num2≤k.

4. The generation method according to claim 1, characterized in that: The output of the deep echo state network is composed of the last layer of reservoir state and the output weight matrix W out The electricity price forecast Y″(t)=W is calculated out X (Num2) (t), the training objective of the deep echo state network is to minimize the difference between the predicted output Y″(t) and the actual electricity price Y t The mean square error between 5. A device for generating a neural network model for real-time electricity price prediction, characterized in that: Includes the following modules: The information acquisition module is used to acquire power consumption information of a preset power system on multiple dates, each of which includes: loads L1, L2, ..., L 24 and the actual electricity prices Y1, Y2, ..., Y 24 , where L t represents the load at the tth hour, Y t represents the actual electricity price at hour t, 1≤t≤24; for any t, calculate Among them, the actual electricity prices are Y1, Y2, ..., Y 24 The average value is μ Y , standard deviation is σ Y , load L1, L2, ..., L 24 The average value is μ L , standard deviation is σ L ; The information processing module is used to perform the following processing on each power consumption information: calculate The minimum k value, where the actual electricity price is Y′1, Y′2, ..., Y′ 24 The average value of k is a natural number; create a preset two-dimensional space and a grid located in the two-dimensional space, and in the preset two-dimensional space, the actual electricity prices Y′1, Y′2, ..., Y′ 24 and loads L′1, L′2, ..., L′ 24 Distribute in the grid; calculate Y′ t and L′ t Mutual information where p(Y′,L′) is Y′ t and L′ t The joint probability of appearing in the same grid, p(Y′) and p(L′) are marginal probabilities; by adjusting the size and segmentation of the grid, MIC(Y′ t ,L′ t ) is maximized, Select one actual electricity price from each grid to get Num1 actual electricity prices Among them, Num1 is a natural number; Model creation module, used to create a deep echo state network based on the features in each power consumption information The deep echo state network is trained, and the deep echo state network is the neural network model.

6. The generating device according to claim 5, characterized in that average value Standard Deviation average value Standard Deviation 7. The generating device according to claim 5, characterized in that The state update formula of the Pth layer reservoir in the deep echo state network is: Among them, W in,P is the input weight matrix of the P-th layer reservoir, W (P) is the internal weight of the P-th layer reserve pool, X (P-1) (t) is the state of the P-1th layer reservoir at time t, X (P) (t) is the state of the P-th layer reserve pool at time t, and the number of layers of the reserve pool in the deep echo state network is Num2, Num2≤k.

8. The generating device according to claim 5, characterized in that The output of the deep echo state network is composed of the last layer of reservoir state and the output weight matrix W out The electricity price forecast Y″(t)=W is calculated out X (Num2) (t), the training objective of the deep echo state network is to minimize the difference between the predicted output Y″(t) and the actual electricity price Y t The mean square error between 9. A device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the generation method according to any one of claims 1 to 4.

10. A readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the generation method according to any one of claims 1 to 4 is realized.

Images