Wind power prediction method and device based on combined generative adversarial neural network

By combining the generation of adversarial neural networks, combining the historical laws of wind power and numerical weather forecasts, a GAN model is formed, which solves the problem of poor prediction effect in wind power prediction and achieves higher prediction accuracy.

CN118920452AInactive Publication Date: 2024-11-08WUHAN UNIV

Patent Information

Application Number
CN202410947015.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-08
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, in the prediction of wind power, prediction is only made from one angle, resulting in poor prediction effects and it is difficult to accurately predict the randomness and intermittentity of wind power.

Method used

Using a method based on combination generation of adversarial neural networks (GANs), a generator is constructed based on the historical laws of wind power and a discriminator based on numerical weather forecasts, forming a GAN model, combining the trained GAN model to predict wind power, improving prediction accuracy .

Benefits of technology

Through the confrontation between the generator and the discriminator, the accuracy of the prediction results is gradually improved, so that the statistical prediction method results can be more consistent with the physical model prediction results, and the accuracy of wind power prediction is improved.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118920452A_ABST
    Figure CN118920452A_ABST
Patent Text Reader

Abstract

The invention discloses a wind power prediction method and device based on a combined generative adversarial neural network. According to the method, a generative adversarial neural network (GAN) is composed of a generator (G) and an adversarial device (D), the generator is constructed based on wind power historical law learning, and a discriminator is constructed based on a physical model of numerical weather prediction (NWP), so that in the training process of the GAN, the generator and the discriminator confront continuously, the performance of the generator and the performance of the discriminator are improved gradually, and the training efficiency of the GAN is improved. According to the invention, the result generated by the generator can be generated with false and messy true and has similar characteristics with the input data of the discriminator, so that the statistical prediction method result output in the generator can better fit the physical model prediction result to a certain extent, and the accuracy of the prediction result is further improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the technical field of rock power data analysis, and in particular to a wind power prediction method and device based on a combined generative adversarial neural network. Background Art

[0002] Wind power generation is one of the clean energy sources that my country has vigorously developed in recent years. However, due to the randomness, intermittency and volatility of its output caused by the weather, it will affect the stable operation of the power system. Accurate prediction of wind power generation is one of the important means to ensure the stable operation of the power system. The use of combined generative adversarial neural networks can improve the accuracy of wind power prediction, which has important theoretical significance and practical value.

[0003] In the related technology, the prediction methods of wind power can be divided into two categories: physical model prediction method and statistical prediction method, which respectively predict wind power output from the main physical factors affecting wind power output and the learning of historical laws of wind power output. However, if wind power is predicted from only one perspective, the prediction effect is poor. Summary of the invention

[0004] In view of this, the present application provides a wind power prediction method and device based on a combined generative adversarial neural network, which can improve the accuracy of wind power prediction.

[0005] In a first aspect, the present application provides a wind power prediction method based on a combined generative adversarial neural network, comprising: A learning-based generator based on historical patterns of wind power; Construct a discriminator based on the wind power prediction model of numerical weather forecast; The generator and the discriminator are combined to form a GAN model, and wind power is predicted based on the trained GAN model.

[0006] Optionally, a generator is constructed based on historical wind power law learning, including: Decomposing the wind power sequence based on wavelet decomposition to obtain signals including detail signals and approximate signals; Predict each signal based on LSTM to obtain the prediction results of each signal; The prediction results of each signal are weighted to obtain the generated data results of the generator.

[0007] Optionally, the detail signal and the approximate signal are expressed by the following formula: ; ; In the above formula, represents an approximate signal, Represents the detail signal, represents the scale of the wavelet basis function, represents the position of the wavelet basis function, represents the wavelet basis function, Represented in scale and the location is The wavelet coefficients under the wavelet basis function are represents the normalized wind power history series, Indicates time.

[0008] Optionally, a discriminator is constructed based on a wind power prediction model of numerical weather forecast, specifically: Based on BP neural network, the wind power is predicted with meteorological data to obtain a discriminator.

[0009] Optionally, the meteorological data are meridional wind speed, latitudinal wind speed and air density in numerical weather forecasts.

[0010] Optionally, in the GAN model: the loss function of the generator It is expressed in the following formula: ; Loss function of the discriminator It is expressed in the following formula: ; Objective function of the GAN model It is expressed in the following formula: ; Where: represents the probability that the discriminator judges that the input comes from a real sample, Represents the data that the generator will generate, represents the real sample data, For the generated samples, It means that when the random variable Obey the original real sample data Generator generates data when distributed The expected value after taking the logarithm is It means that when the random variable Obey the original generated sample data Distribution The expected value of is the objective function of the generator G and the discriminator D, which represents the maximum and minimum game between the generator and the discriminator.

[0011] Optionally, before building a generator based on wind power historical law learning or building a discriminator based on a wind power prediction model of numerical weather forecast, the method further includes: Standardize the original wind power data.

[0012] In a second aspect, the present application provides a wind power prediction method and device based on a combined generative adversarial neural network, comprising: The first building module is used to learn and build a generator based on the historical law of wind power; The second building module is used to build a discriminator based on the wind power prediction model of numerical weather forecast; A prediction module is used to combine the generator and the discriminator to form a GAN model, and predict wind power based on the trained GAN model.

[0013] In a third aspect, the present application provides a computer-readable storage medium, comprising a program, which, when executed on a computer, enables the computer to execute the method as described above.

[0014] In a fourth aspect, the present application provides an execution device, comprising a processor and a memory, wherein the processor is coupled to the memory; The memory is used to store programs; The processor is used to execute the program in the memory, so that the execution device executes the above method.

[0015] In the method disclosed in the present application, a generative adversarial neural network (GAN) is composed of two parts: a generator (G) and an adversary (D), and the generator is constructed based on the learning of historical laws of wind power, and the discriminator is constructed based on the physical model of numerical weather forecasting (NWP). In this way, during the training process of GAN, the generator and the discriminator continuously confront each other and gradually improve their performance, so that the results generated by the generator can be indistinguishable from the real ones and have similar characteristics to the discriminator input data, so that the statistical prediction method results output in the generator can be more consistent with the physical model prediction results to a certain extent, thereby improving the accuracy of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The technical solution and other beneficial effects of the present application will be made apparent by describing in detail the specific implementation methods of the present application in conjunction with the accompanying drawings.

[0017] Figure 1 An operational flow chart of a wind power prediction method based on a combined generative adversarial neural network provided by an exemplary embodiment is shown.

[0018] Figure 2 A schematic diagram of a GAN model provided by an exemplary embodiment is shown.

[0019] Figure 3A block diagram of a wind power prediction device based on a combined generative adversarial neural network provided by an exemplary embodiment is shown.

[0020] Figure 4 A block diagram of the execution device provided by an exemplary embodiment is shown. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present application, so the present application is not limited by the specific embodiments disclosed below. The terms "first", "second", etc. in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or modules is not necessarily limited to those steps or modules clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products or devices. The naming or numbering of steps in this application does not mean that the steps in the method flow must be executed in the time / logical sequence indicated by the naming or numbering. The process steps that have been named or numbered can change the execution order according to the technical purpose to be achieved, as long as the same or similar technical effects can be achieved. The division of units in this application is a logical division. There may be other division methods when it is implemented in actual applications. For example, multiple units can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection between units can be electrical or other similar forms, which are not limited in this application. In addition, the units or sub-units described as separate components may or may not be physically separated, may or may not be physical units, or may be distributed in multiple circuit units, and some or all of the units may be selected according to actual needs to achieve the purpose of the present application.

[0022] Please refer to Figure 1 ,in Figure 1A flowchart of a wind power prediction method based on a combined generative adversarial neural network provided by an exemplary embodiment is shown. The method is implemented through steps 102-108.

[0023] In step 102, a generator is constructed based on the learning of historical laws of wind power.

[0024] It can be understood that the "historical law of wind power" here refers to the known data that discloses the corresponding relationship between wind power and its influencing factors (weather factors). The known data can be measured data in historical records or open source data, etc., but is not limited to this.

[0025] The learning construction generator based on the historical laws of wind power can be implemented in actual operation by conventional methods such as machine learning and deep learning.

[0026] In a specific demonstration with a wide range of applications, a generator is constructed based on historical wind power learning, which can be achieved by combining wavelet decomposition and long short-term memory model (LSTM).

[0027] Please refer to Figure 2 Specifically, the generator constructed based on wind power historical law learning includes: Step 1021 : Decompose the wind power sequence based on wavelet decomposition to obtain signals including detail signals and approximate signals.

[0028] The wavelet coefficients can be realized by performing inner product operation on the signal and the wavelet basis function. The formula is: (1); Where: is the normalized wind power history series, Indicates time, represents the scale of the wavelet basis function, represents the position of the wavelet basis function, represents the conjugate complex number of the wavelet basis function, Represented in scale and the location is The wavelet coefficients of the original signal under the wavelet basis function.

[0029] Reconstructing the calculated wavelet coefficients can decompose the original signal into approximate signals With detail signal , and the calculation formulas are as follows: (2); (3); In the above formula, represents the wavelet basis function.

[0030] If you need to decompose the signal into multiple layers, you can use the approximate signal Continue to decompose according to the above method until the required number of layers is reached.

[0031] Step 1022: predict each signal based on LSTM to obtain a prediction result for each signal.

[0032] The long short-term memory model (LSTM) is a variant of the recurrent neural network (RNN). It introduces a memory unit and three gating units (forget gate, input gate, and output gate). The calculation formulas of each gating unit are as follows.

[0033] The calculation formula of the forget gate is: (4); Where: is the output of the forget gate, is the Sigmoid function, and is the weight matrix and bias of the forget gate, is the hidden state connected to the previous time step and Vector of inputs for the current time step.

[0034] The input gate calculation formula is: (5); Where: is the output of the input gate, and are the weight matrices and biases of the input gate.

[0035] The output gate calculation formula is: (6) ; Where: is the output of the output gate, and is the weight matrix and bias of the output gate.

[0036] The update of the memory unit is completed through the combination of the forget gate, the input gate and the candidate memory unit. The candidate memory unit contains the new information of the current time step. The update calculation formula of the memory unit is: (7); Where: is the updated memory unit, is the output of the forget gate, is the value of the memory cell at the previous time step, is the output of the input gate, is the value of the candidate memory cell.

[0037] Finally, the hidden state of the current time step is calculated through the tanh activation function of the output gate and the memory unit. The hidden state calculation formula is: (8); in, is the hidden state at the current time step, is the output of the output gate, tanh is the hyperbolic tangent function, is the value of the memory cell at the current time step.

[0038] Step 1023: weight the prediction results of each signal to obtain the generated data results of the generator.

[0039] Here, the weighted coefficient can be selected according to actual needs.

[0040] In step 104, a discriminator is constructed based on the wind power prediction model of numerical weather forecast.

[0041] Reference again Figure 2 In actual operation, the normalized numerical weather forecast data is used as the input layer input, here, the longitudinal wind speed, zonal wind speed and air density, etc. The wind power is used as the output layer output to construct the BP neural network, the activation function is the Sigmoid function, the loss function is the MSE function, and the constructed BP neural network is trained.

[0042] data The outputs flowing from the input layer into the hidden layer are: Output of the hidden layer: (9); (10); Where: is the hidden layer k The output of the node, The weighted sum of the input data, For Node k Corresponding to the weight of each data input node, For Node k The corresponding neuron bias.

[0043] Any hidden layer The output is: (11); Where: is the total number of hidden layers, For hidden layer The output, is the hidden layer The number of nodes.

[0044] The final output of the neural network is the last hidden layer. The output value is: (12); Assume that the data used for training is , the loss function is expressed as: (13); According to the gradient descent method, the parameter update method is: (14); Where: is a given learning rate.

[0045] The gradient of each hidden layer parameter can be obtained in reverse order, and the BP network can be trained using the gradient descent algorithm.

[0046] In step 106, the generator and the discriminator are combined to form a GAN model, and wind power is predicted based on the trained GAN model.

[0047] Reference again Figure 2 ,The generator has been constructed in the previous steps. The forecast results of the wind power prediction model based on numerical weather forecast are used as the discriminant basis of the discriminator D. The discriminant model uses a convolutional neural network (CNN) to construct and train the combined GAN model.

[0048] The loss functions of the generator are: (15); The loss functions of the discriminator are: (16); The objective function is: (17); Where: represents the probability that the discriminator judges that the input comes from a real sample, Represents the data that the generator will generate, represents the real sample data, For the generated samples, It means that when the random variable Obey the original real sample data Generator generates data when distributed The expected value after taking the logarithm is It means that when the random variable Obey the original generated sample data Distribution The expected value of is the objective function of the generator G and the discriminator D, which represents the maximum and minimum game between the generator and the discriminator.

[0049] In a preferred specific example, before building a generator based on wind power historical law learning or building a discriminator based on a wind power prediction model of numerical weather forecast, the method further includes: Standardize the original wind power data.

[0050] In actual operation, the standardization process is as follows: First, collect the data needed for wind power prediction, including historical wind power operation data, numerical weather forecast related data, such as longitudinal wind speed, latitudinal wind speed, and air density, and process the abnormal values ​​and null values ​​in the collected original data.

[0051] Secondly, the data after preliminary processing is sorted and the sorted data is standardized. The standardized calculation formula is as follows: (18); Where: for t After time standardization, , is the length of the data; for t The original data at the moment; is the mean of the original data; is the standard deviation of the original data.

[0052] Finally, according to the length of the data, the entire period is divided into the rate period and the test period.

[0053] Please refer to Figure 3 , which shows a block diagram of a wind power prediction device based on a combined generative adversarial neural network provided by the present application. The device 200 includes: A first construction module 202 is used to construct a generator based on the learning of historical laws of wind power; A second construction module 204 is used to construct a discriminator based on a wind power prediction model of numerical weather forecast; The prediction module 206 is used to combine the generator and the discriminator to form a GAN model, and predict wind power based on the trained GAN model.

[0054] Since the above methods have been discussed in detail, the specific implementation of the above modules will not be repeated here.

[0055] Next, an execution device provided by an embodiment of the present application is introduced. Figure 4 , Figure 4This is a schematic diagram of a structure of an execution device provided in an embodiment of the present application. The execution device 300 can be specifically manifested as an autonomous driving vehicle, a mobile phone, a tablet, a laptop computer, a desktop computer, a monitoring data processing device, etc., which is not limited here. Figure 1 The functions of the execution device in the corresponding embodiment. Specifically, the execution device 300 includes: a receiver 301, a transmitter 302, a processor 303 and a memory 304 (wherein the number of the processor 303 in the execution device 300 can be one or more, Figure 4 In the example of FIG. 3 , the processor 303 may include an application processor 3031 and a communication processor 3032. In some embodiments of the present application, the receiver 301, the transmitter 302, the processor 303 and the memory 304 may be connected via a bus or other means.

[0056] The memory 304 may include a read-only memory and a random access memory, and provides instructions and data to the processor 303. A portion of the memory 304 may also include a non-volatile random access memory (NVRAM). The memory 304 stores processor and operation instructions, executable modules or data structures, or subsets thereof, or extended sets thereof, wherein the operation instructions may include various operation instructions for implementing various operations.

[0057] The processor 303 controls the operation of the execution device. In a specific application, the various components of the execution device are coupled together through a bus system, wherein the bus system includes not only a data bus but also a power bus, a control bus, and a status signal bus, etc. However, for the sake of clarity, various buses are referred to as bus systems in the figure.

[0058] The method disclosed in the above embodiment of the present application can be applied to the processor 303, or implemented by the processor 303. The processor 303 can be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 303. The above processor 303 can be a general processor, a digital signal processor (digital signal processing, DSP), a microprocessor or a microcontroller, and can further include an application specific integrated circuit (application specific integrated circuit, ASIC), a field programmable gate array (field-programmable gate array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The processor 303 can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present application. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to execute, or it can be executed by a combination of hardware and software modules in the decoding processor. The software module may be located in a storage medium mature in the art, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 304, and the processor 303 reads the information in the memory 304 and completes the steps of the above method in combination with its hardware.

[0059] The receiver 301 can be used to receive input digital or character information and generate signal input related to the relevant settings and function control of the execution device. The transmitter 302 can be used to output digital or character information through the first interface; the transmitter 302 can also be used to send instructions to the disk group through the first interface to modify the data in the disk group; the transmitter 302 can also include a display device such as a display screen.

[0060] In the embodiment of the present application, the processor 303 is used to execute Figure 1 The specific manner in which the application processor 3031 in the processor 303 performs the above steps is the same as that in the present application. Figure 1 The corresponding method embodiments are based on the same concept, and the technical effects they bring are the same as those in this application. Figure 1 The corresponding method embodiments are the same. For specific contents, please refer to the description in the method embodiments shown above in this application, which will not be repeated here.

[0061] The above is only a preferred specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed in the present application should be covered within the protection scope of the present application.

Claims

1. A wind power prediction method based on combined generative adversarial neural network, characterized in that: include: A learning-based generator based on historical patterns of wind power; Construct a discriminator based on the wind power prediction model of numerical weather forecast; The generator and the discriminator are combined to form a GAN model, and wind power is predicted based on the trained GAN model.

2. The method according to claim 1, characterized in that: The generator is built based on historical wind power law learning, including: Decomposing the wind power sequence based on wavelet decomposition to obtain various signals including detail signals and approximate signals; Predict each signal based on LSTM to obtain the prediction results of each signal; The prediction results of each signal are weighted to obtain the generated data results of the generator.

3. The method according to claim 1, characterized in that The detail signal and the approximate signal are expressed by the following formula: ; ; In the above formula, represents an approximate signal, Represents the detail signal, represents the scale of the wavelet basis function, represents the position of the wavelet basis function, represents the wavelet basis function, Represented in scale and the location is The wavelet coefficients under the wavelet basis function are represents the normalized wind power history series, Indicates time.

4. The method according to claim 1, characterized in that: The discriminator is constructed based on the wind power prediction model of numerical weather forecast, specifically: Based on BP neural network, the wind power is predicted with meteorological data to obtain a discriminator.

5. The method according to claim 4, characterized in that The meteorological data are the meridional wind speed, zonal wind speed and air density in the numerical weather forecast.

6. The method according to claim 1, characterized in that In the GAN model: the loss function of the generator It is expressed in the following formula: ; Loss function of the discriminator It is expressed in the following formula: ; Objective function of the GAN model It is expressed in the following formula: ; Where: represents the probability that the discriminator judges that the input comes from a real sample, Represents the data that the generator will generate, represents the real sample data, For the generated samples, It means that when the random variable Obey the original real sample data Generator generates data when distributed The expected value after taking the logarithm is It means that when the random variable Obey the original generated sample data Distribution The expected value of is the objective function of the generator G and the discriminator D, which represents the maximum and minimum game between the generator and the discriminator.

7. The method according to claim 1, characterized in that Before building a generator based on wind power historical law learning or building a discriminator based on a wind power prediction model of numerical weather forecast, the method further includes: Standardize the original wind power data.

8. A wind power prediction method and device based on combined generative adversarial neural network, characterized in that: include: The first building module is used to build a generator based on the learning of historical laws of wind power; The second building module is used to build a discriminator based on the wind power prediction model of numerical weather forecast; A prediction module is used to combine the generator and the discriminator to form a GAN model, and predict wind power based on the trained GAN model.

9. A computer-readable storage medium, characterized in that: The invention comprises a program, which, when being executed on a computer, causes the computer to execute the method according to any one of claims 1 to 7.

10. An execution device, characterized in that: comprising a processor and a memory, wherein the processor is coupled to the memory; The memory is used to store programs; The processor is configured to execute the program in the memory so that the execution device executes the method according to any one of claims 1 to 7.

Citation Information

Patent Citations

  • Distributed regional wind power prediction method

    CN112598180A

  • Multi-feature fusion generative adversarial network ultra-short-term wind power prediction method

    CN115526376A

Cited By

  • Microgrid short-term load prediction method and device, electronic equipment and storage medium

    CN121328859A

  • Unified space-time wind power prediction method based on Fourier generative adversarial network

    CN121461293A