Method and system for generating wind power extreme scene caused by medium time scale weather process

The wind electrode extreme scenarios are generated through the conditional probability distribution model and convolutional layer structure, which solves the problems of generated scene effectiveness and stability in the prior art, and realizes accurate prediction of wind electrode extreme scenarios on the medium time scale and optimization of wind power system.

CN119006218APending Publication Date: 2024-11-22NORTH CHINA ELECTRIC POWER UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411105825.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-13
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The prior art has problems in the controllable generation of renewable energy output scenarios, the instability of GAN model training affects accuracy, and the lack of definition and analysis of wind electrode extreme scenarios on medium time scales.

Method used

Using the conditional probability distribution model, by collecting meteorological and wind power data, using full-connection layers and convolutional layers to build encoder and decoder, introduce noise addition and denoising processes, build M_X2Z and M_Z2Y models, generate wind electrode extreme scenarios and define extreme metrics , realize the controllable generation of wind electrode extreme scenarios caused by medium-time-scale weather processes.

Benefits of technology

Accurate prediction of wind electrode extreme scenarios is achieved, comprehensiveness and accuracy of predictions are improved, real-time operation strategies are provided, economic losses are reduced, and renewable energy reliability and confidence in utilization are promoted.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119006218A_ABST
    Figure CN119006218A_ABST
Patent Text Reader

Abstract

The invention discloses a method and system for generating a wind power extreme scene caused by a medium time scale weather process, and the method comprises the steps: data collection and preprocessing: training and generating daily output sequence features and a wind power multi-day extreme low output scene through constructing two conditional probability distribution models. And defining and identifying wind power extreme scenes by using the scene data, wherein the wind power extreme scenes comprise extreme low output and climbing scenes. Advanced fully connected layer and convolutional layer structures, noise adding and de-noising processes in a diffusion model, and technologies such as residual connection and the like are utilized. The method has the advantages that a wind power extreme scene with controllable output characteristics can be generated, and a reliable data analysis basis is provided for system optimization setting of an acceptable electric power and electric quantity balance gap and a maximum daily output fluctuation limit value.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of meteorological data analysis and wind power system optimization, and in particular to a method and system for generating wind power extreme scenarios triggered by medium-time-scale weather processes. Background Art

[0002] As global warming accelerates, rapidly intensifying tropical cyclones and excess greenhouse gases have brought about frequent weather events, which have had a significant impact on the output of renewable energy. The strong intermittent, random and volatile output characteristics of renewable energy, which are directly affected by weather conditions, have aggravated the difficulty of the power balance task of the new power system and brought huge challenges to the safe and stable operation of the power system. The controllable generation of time-scale wind power sequence scenarios in the study can help understand the cumulative impact and continuous changes of weather processes on wind power operations, and provide support for the power system to formulate relevant flexibility resource scheduling decision-making plans and conduct power system security assessments.

[0003] At present, domestic and foreign scholars have carried out a lot of research on renewable energy output scenario generation, gradually moving from random scenario generation to controllable scenario generation. Most models use deep learning generative models guided by conditional values. Such conditions can be explicit conditions, such as conditional generative adversarial networks and their variants, or implicit conditions obtained from latent spaces, such as diffusion models. There are also studies that apply explicit and implicit conditions at the same time to better obtain the effect of specified generation.

[0004] However, existing research has the following shortcomings in the controllable generation of renewable energy output scenarios:

[0005] 1. It is difficult to prove the effectiveness of the generated scenario by directly using meteorological characteristics as control indicators for wind power timing scenarios;

[0006] 2. The basic model of GAN and its variants are mostly used, and the instability of GAN model training affects the accuracy of controllable generation;

[0007] 3. Most of the consideration is given to the extreme conditions of daily output, but there is a lack of definition and analysis of extreme wind power scenarios on medium time scales. Summary of the invention

[0008] In view of the defects of the prior art, the present invention provides a method and system for generating wind power extreme scenarios triggered by medium-time-scale weather processes.

[0009] In order to achieve the above invention object, the technical solution adopted by the present invention is as follows:

[0010] A method for generating wind power extreme scenarios caused by mesoscale weather processes comprises the following steps:

[0011] Step 1, Data Collection:

[0012] Meteorological data including humidity, temperature, zonal wind speed, and radial wind speed were collected.

[0013] Collect wind power output data, including daily output data and daily ramp data.

[0014] Step 2, data preprocessing:

[0015] Clean and standardize meteorological data and wind power output data.

[0016] Use transfer learning techniques to generate an expanded sample set for training the model.

[0017] Step 3: Build the conditional probability distribution model M_X2Z and train it:

[0018] The input feature encoder is constructed using the fully connected layer FCN and the 2D convolution layer Conv2d to map the meteorological feature X into the latent space. During the model training process, the denoising and denoising processes are introduced, and the encoded input matrix and latent space intermediate variables are processed using the stacked 2D convolution layer and the dilated Conv2d convolution layer. The decoder is constructed through the residual stack layer skip stack and two layers of Conv2d to decode the latent variables into the daily power sequence feature Z.

[0019] Step 4: Generate the characteristic scenario of the daily force sequence and define the extreme measure:

[0020] The trained M_X2Z is ​​used to generate the characteristic scene of the daily power sequence under given meteorological conditions. The generated daily power sequence features include daily power, daily maximum upper fluctuation and daily maximum lower fluctuation. The probability analysis of the generated daily power sequence characteristic scene is carried out to define the daily power sequence characteristic threshold of the low-probability wind power extreme output event.

[0021] Step 5: Build the conditional probability distribution model M_Z2Y and train it:

[0022] Model M_Z2Y is constructed using the same network structure as model M_X2Z.

[0023] The conditional probability distribution between the daily output sequence feature Z and the wind power output scenario Y is fitted using the model M_Z2Y The daily output sequence feature Z is used as a condition to guide the generation of multi-day extreme low output scenarios of wind power, and the multi-day extreme low output scenario of wind power is generated.

[0024] Step 6: Generate extremely low wind power output scenario:

[0025] Using the trained conditional probability distribution model M_Z2Y, according to The setting generates an extremely low wind power output scenario.

[0026] The model M_Z2Y is used to generate extreme wind power ramping scenarios.

[0027] Step 7, defining wind power extreme events, including: defining a continuous multi-day extreme low output event where the average daily output is lower than a given threshold and defining a wind power extreme ramping scenario where the daily maximum upper fluctuation and the daily maximum lower fluctuation are higher than a given threshold.

[0028] Step 8: Identify wind power extreme events through the generated daily power sequence characteristics, and generate wind power extreme event scenarios with controllable daily power sequence characteristics.

[0029] Furthermore, the definition of the extreme low output event for consecutive days where the average daily output is lower than a given threshold is as follows:

[0030]

[0031] In the formula is the daily power generation sequence feature i starting from day d with a length of m, i = 1, ε is the wind power low output threshold, d is the starting day, m is the number of days, M e is the extreme low output days threshold. Further, the wind power extreme ramping scenario where the daily maximum upper fluctuation and the daily maximum lower fluctuation are higher than a given threshold is defined as follows:.

[0032]

[0033]

[0034] In the formula is the characteristic value of the daily force sequence characteristic i on day d, y t is the wind power output value at time t, t is any time within the day, Δt is the time interval within the day, is the wind power output fluctuation threshold.

[0035] Furthermore, the network loss functions used in model M_X2Z and model M_Z2Y are as follows:

[0036]

[0037] Where n is the number of diffusion steps, and s is n is the target sequence that gradually becomes clear after n denoising processes, ∈ is the noise injected at each step, and the commonly used normal distribution is selected as the prior distribution for the convenience of sampling and operation. c is the input condition, ∈ θ Denoising noise mapped for the denoising network.

[0038] The present invention also discloses a system for generating wind power extreme scenarios triggered by medium-time-scale weather processes. The system can be used to implement the above-mentioned method for generating wind power extreme scenarios, and specifically includes:

[0039] Data collection module: collects meteorological data and wind power output data;

[0040] Data preprocessing module: cleans and standardizes the collected meteorological data and wind power output data to ensure data quality and consistency; generates an expanded sample set for training the model;

[0041] M_X2Z model construction and training module:

[0042] Use the fully connected layer FCN and the 2D convolution layer Conv2d to build the input feature encoder;

[0043] The denoising and denoising processes are introduced to process the input matrix and latent space intermediate variables by stacking 2D convolutional layers and dilated convolutional layers;

[0044] Use the residual stack layer skip stack and two layers of Conv2d to build a decoder to decode the hidden variables into the sunrise force sequence feature Z;

[0045] Sunrise force sequence feature scene generation and extreme metric definition module:

[0046] The trained M_X2Z model is used to generate the characteristic scene of the daily output sequence under given meteorological conditions. The generated time series features include the daily output, the maximum daily upward fluctuation and the maximum daily downward fluctuation.

[0047] Perform probability analysis on the generated daily power sequence characteristic scenarios and define the daily power sequence characteristic thresholds for low-probability wind power extreme power output events;

[0048] M_Z2Y model construction and training module: Use the same network structure as M_X2Z to construct the conditional probability distribution model M_Z2Y; use M_Z2Y to fit the conditional probability distribution P(Y|Z) between the daily power sequence feature Z and the wind power output scenario Y; use the daily power sequence feature Z as a condition to guide the generation of wind power multi-day extreme low output scenarios, and generate wind power multi-day extreme low output scenarios Y;

[0049] Wind power extreme scenario generation module:

[0050] Using the trained model M_Z2Y, according to the set P(Y|Z), the wind power extreme low output scenario and wind power extreme climbing scenario are generated;

[0051] Wind power extreme event definition and identification module: defines wind power extreme events, including extreme low output events for consecutive days where the average daily output is lower than a given threshold; defines wind power extreme ramping scenarios where the daily maximum upper fluctuation and the daily maximum lower fluctuation are higher than a given threshold;

[0052] Extreme event identification and application module:

[0053] Using the generated daily output sequence characteristics and the multi-day extremely low wind power output scenario, the defined wind power extreme events are applied to identify extreme events;

[0054] Including early warning systems, risk assessment tools, etc., for wind power scenario management and optimized decision-making.

[0055] The present invention also discloses a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned wind power extreme scenario generation method when executing the program.

[0056] The present invention also discloses a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the above-mentioned wind power extreme scenario generation method is implemented.

[0057] Compared with the prior art, the advantages of the present invention are:

[0058] 1. Accurate prediction capability: Using the conditional probability distribution model, it can accurately predict extreme wind power scenarios that may occur under specific meteorological conditions, such as extremely low wind power output and extreme wind power ramp-up scenarios.

[0059] 2. Multi-dimensional data processing: It can effectively process a variety of meteorological characteristics and wind power output data, comprehensively consider the impact of multiple factors on the performance of wind power systems, and improve the comprehensiveness and accuracy of predictions.

[0060] 3. Model flexibility: The encoder and decoder built with fully connected layers and convolutional layers, as well as the denoising and denoising processes, enable the model to flexibly adapt to different data characteristics and changing weather conditions.

[0061] 4. Practicality: It provides real-time or near real-time wind power system operation strategies, allowing wind farm managers to adjust power generation plans and emergency measures according to predicted extreme events, thereby improving the operating efficiency and stability of the system.

[0062] 5. Reduce economic losses: By predicting the occurrence of extreme events in advance, economic losses caused by sudden weather changes can be effectively reduced, including reducing additional costs caused by failure to meet electricity demand.

[0063] 6. Promote the development of renewable energy: Improving the reliability and predictability of wind power systems will help increase confidence in the use of wind energy resources and promote the wider use of renewable energy in the energy structure. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 It is a flow chart of a method for generating a wind power extreme scenario according to an embodiment of the present invention.

[0065] Figure 2 It is a network structure diagram of a diffusion model according to an embodiment of the present invention.

[0066] Figure 3 The embodiment of the present invention is a wind power output scenario generation result diagram, where (a) is a normal output scenario and (b) is an extremely low output scenario.

[0067] Figure 4 This is a controllable generated result diagram of a single-day scenario of an extreme wind power ramp event in an embodiment of the present invention. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and examples.

[0069] The present invention transforms the wind power output conditional probability distribution modeling based on the medium-time scale meteorological characteristics into the conditional probability distribution of the daily output sequence characteristics Z under the influence of the medium-scale meteorological conditions and the conditional probability distribution of the wind power output scenario guided by the daily output sequence characteristics. First, a conditional probability distribution is fitted using the model M_X2Z. To describe the mapping relationship between meteorological and daily power sequence characteristics, the weather process is used as the driving force to construct the daily power sequence characteristic probability distribution that integrates weather process information. Then, based on the daily power sequence characteristic scenario generated by M_X2Z, the extreme metric of the wind power scenario is defined, and the boundary value of the small probability wind power extreme metric is given. Secondly, the model M_Z2Y is used to fit another conditional probability distribution To solve the problem of generating controllable wind power scenarios under weather processes, use daily output data and daily ramp data to drive the generation of multi-day extreme low wind power scenarios, and fit the probability distribution of wind power extreme output conditions that integrates weather process information and daily output sequence feature probability information. Finally, within the extreme measurement range of wind power scenarios, extreme scenarios are given to be controllable and generate wind power extreme output scenarios with a given probability of occurrence.

[0070] Definition of mesoscale wind power extreme events

[0071] Extreme wind power scenarios on a medium time scale can be defined into two categories: one is low wind for many days, resulting in low wind power output for many consecutive days; the other is extreme wind power ramping scenarios caused by excessive wind speed fluctuations.

[0072] The first type of event is a description of the total amount of wind power output over multiple days. In the event of no wind for multiple days, the average daily power output of wind power is relatively small. Therefore, the average daily power output is defined as the first feature in the daily power output sequence feature, and the sequence length is T. d , continuous low output for many days is an extreme output scenario:

[0073]

[0074] In the formula is the daily power generation sequence feature i starting from day d with a length of m, i = 1, ε is the wind power low output threshold, d is the starting day, m is the number of days, M e is the threshold for extremely low output days. In the daily average power sequence, if the average power value for m consecutive days is lower than the given threshold, it is a medium-scale wind power extreme output event. The second type of event is a description of the volatility of wind power output over multiple days. The daily maximum climbing rate and the daily maximum climbing rate are defined as the second and third characteristics of the daily power sequence characteristics, respectively. The wind power output fluctuation within an hour in a day reaches the fluctuation threshold or above, which means that an extreme wind power climbing scenario has occurred on that day:

[0075]

[0076] In the formula is the characteristic value of the daily force sequence characteristic i on day d, y t is the wind power output value at time t, t is any time within the day, Δt is the time interval within the day, is the wind power output fluctuation threshold.

[0077] Model structure

[0078] The diffusion model of M_X2Z and the diffusion model of M_Z2Y model have the same model structure, but differ in the dimension of neurons. Figure 2 The network structure of the diffusion model proposed in the present invention is shown. Each diffusion model includes three parts: input feature encoder, denoising process and decoder.

[0079] (1) Input feature encoder: It contains one fully connected network (FCN) layer and one 2D convolutional layer (Conv2d). In M_X2Z, meteorological features are mapped into the latent space as the conditions for guiding the generation of daily output sequence features; in M_Z2Y, the daily output sequence features are used as the conditions for guiding the generation of multi-day extreme low output scenarios of wind power.

[0080] (2) Denoising process: The diffusion model includes two processes: denoising and denoising. The denoising process only exists in the model training process, while the denoising process is the forward process of generating the target sequence. The model uses the pattern of Conv2d superimposed with a 2D dilated convolution layer (DilatedConv2d) to process the encoded input matrix and the labeled latent space intermediate variables respectively. Conv2d processes the input data through the local receptive field, which can efficiently capture the local patterns in long time series. The filter (convolution kernel) greatly reduces the amount of learnable parameters, which can improve the training efficiency and generalization ability of the model. Dilated Conv2d can expand the receptive field of neurons without changing the size of the matrix, and does not produce the precision loss related to the dimensionality change, thereby improving the generation quality of the target sequence by the diffusion model.

[0081] (3) Decoder: It contains one layer of residual stack (skip stack) and two layers of Conv2d, where the purpose of residual skip connection is to prevent network degradation. After multiple layers of denoising, the latent variables in M_X2Z output the daily output sequence feature Z through the decoder, and the latent variables in M_Z2Y output the wind power multi-day extreme low output scenario Y through the decoder.

[0082] The network loss function is as follows:

[0083]

[0084] Where n is the number of diffusion steps, and s is n is the target sequence that gradually becomes clear after n denoising processes, ∈ is the noise injected at each step, and the commonly used normal distribution is selected as the prior distribution for the convenience of sampling and operation. c is the input condition, ∈ θ Denoising noise mapped for the denoising network.

[0085] The present invention first uses transfer learning to expand the meteorological-wind output sample set. The expanded sample set is used to train M_X2Z and M_Z2Y. Then, the trained M_X2Z is ​​used to generate the power and climbing scenes under given meteorological conditions. The meteorological feature set I X The length is 4, including humidity, temperature, latitudinal wind speed and radial wind speed. Generate m∈{1,2,3,...,15} days of daily average power, daily maximum upper fluctuation and daily maximum lower fluctuation, etc., a total of 3 daily power generation time series feature scenarios, and perform probability analysis on the generated daily average power data and daily upper and lower fluctuation data to obtain the probability density distribution of daily average power, daily maximum upper fluctuation and daily maximum lower fluctuation. Assume that the probability of a certain wind power extreme output event is p, the wind power low output threshold ε or wind power output fluctuation threshold It can be obtained by the inverse cumulative distribution function. The probability of a very small probability event is defined as 0.01, the probability of a small probability extreme event is defined as 0.1, and the daily electricity consumption is a per-unit value, ranging from [0,1].

[0086] Obtaining the wind power low output threshold ε or wind power output fluctuation threshold under a given probability Finally, the characteristic sequence of the daily power sequence is given within the characteristic threshold range of the extreme daily power sequence, and the wind power extreme scenario is generated through the model M_Z2Y.

[0087] Taking the eight-year meteorological forecast data of a certain region in the Global Climate Model (GCM) as an example, the M_X2Z model is used to generate the daily average power series and daily climbing series under the above eight-year meteorological conditions, and the extreme wind power scenarios are identified based on their distribution. The extreme power thresholds of the wind power low power scenario and the extreme fluctuation thresholds of the wind power climbing scenario are given as shown in Table 1 and Table 2 respectively.

[0088] Table 1 Power thresholds for multi-day extreme low output events

[0089]

[0090] Taking m = 10 as an example, the maximum value of the ten-day average daily electricity is 0.53, and the average is 0.24. The ten-day wind power low output threshold ε = 0.122 under the condition of 0.1 probability, and the ten-day wind power low output threshold under the condition of 0.01 probability is only ε = 0.045. Among the 11,680 generated samples, there are 103 samples with daily average electricity less than the threshold, which can provide learning samples for M_Z2Y to generate extreme low wind power output scenarios.

[0091] Table 2 Fluctuation thresholds for single-day climbing scenarios

[0092]

[0093] The probability of a daily ramp-up rate greater than 24% of the rated capacity is 0.1, and the probability of an extreme ramp-up rate greater than 34% of the rated capacity is 0.01. The ramp-up scenario with a probability of 0.01 is used to construct an extreme ramp-up scenario label for M_Z2Y.

[0094] In order to evaluate the error between the generated scene sequence and the control index and analyze the effectiveness and accuracy of the controllable generation of M_Z2Y, the mean absolute error (MAE), root mean square error (RMSE) and continuous ranked probability score (CRPS) are used to measure the controllable accuracy of the generated mode, as shown in the following formula.

[0095]

[0096]

[0097] In the formula is the normalized symbol, To generate the characteristic value of the daily power sequence calculated on the i-th day in the multi-day power scenario, z i is the eigenvalue of the actual daily output sequence on the i-th day.

[0098]

[0099] Where N is the number of generated scenes, To generate the output value at the i-th moment in the output scenario, y i is the actual output value at the i-th moment.

[0100] It can be seen from Table 3 that M_Z2Y based on the diffusion model is significantly better than CGAN and CVAE in both deterministic and probabilistic indicators. The lower CRPS value indicates that the probability distribution predicted by M_Z2Y is closer to the actual situation, reflecting that the M_Z2Y model in the present invention has better uncertainty capture ability.

[0101] Table 3 Model accuracy indicators

[0102]

[0103] The generation results of conventional wind power scenarios and extreme wind power scenarios with a time scale of m = 10 days and a time series length of 240 are given. Figure 3 (a) is the result when the average daily electricity consumption for ten days is 5.83; Figure 3 (b) is the result of scenario generation when the average daily power consumption for ten days is 0.77. Figure 3 As shown, it can be seen that the controllable scene generated by the present invention can better wrap the actual scene and has good controllable mapping capabilities under both conventional scene and extreme scene metrics.

[0104] For the generation of extreme climbing scenarios, an upward fluctuation of 34% is selected for the upward climbing and a downward fluctuation of 32% is selected for the downward climbing. In the two extreme modes, 10 groups of curves with fluctuation intensity higher than the given index are selected for observation.

[0105] like Figure 4 As shown, it can be seen that since the daily power generation sequence characteristics generated by controlling the single-day power generation sequence scenario are the highest up and down fluctuations and power consumption of a single day, while satisfying the overall trend of controllable conditions, each group of scenarios retains uncertainty differences in the generation of hourly details, proving that the present invention has excellent controllable generation capabilities and uncertainty expression capabilities.

[0106] In another embodiment of the present invention, a system for generating wind power extreme scenarios caused by a mesoscale weather process is provided. The system can be used to implement the above-mentioned method for generating wind power extreme scenarios, and specifically includes:

[0107] Data collection module: collects meteorological data and wind power output data.

[0108] Data preprocessing module: cleans and standardizes the collected meteorological data and wind power output data to ensure data quality and consistency. Generates an expanded sample set for model training.

[0109] M_X2Z model construction and training module:

[0110] The input feature encoder is constructed using a fully connected layer (FCN) and a 2D convolutional layer (Conv2d).

[0111] The denoising and denoising processes are introduced to process the input matrix and latent space intermediate variables by stacking 2D convolutional layers and dilated convolutional layers (Dilated Conv2d).

[0112] The decoder is constructed using a skip stack and two layers of Conv2d to decode the latent variables into the output sequence feature Z.

[0113] Sunrise force sequence feature scene generation and extreme metric definition module:

[0114] The trained M_X2Z model is used to generate the daily power sequence feature scenarios under given meteorological conditions. The generated daily power sequence features include the daily average power, the daily maximum upper fluctuation and the daily maximum lower fluctuation.

[0115] A probability analysis is performed on the generated daily power sequence characteristic scenarios to define the daily power sequence characteristic thresholds for low-probability wind power extreme power output events.

[0116] M_Z2Y model construction and training module: Use the same network structure as M_X2Z to build the conditional probability distribution model M_Z2Y. Use M_Z2Y to fit the conditional probability distribution P(Y|Z) between the daily power sequence feature Z and the wind power output scenario Y. Use the daily power sequence feature Z as a condition to guide the generation of wind power multi-day extreme low output scenario, and generate wind power multi-day extreme low output scenario Y.

[0117] Wind power extreme scenario generation module:

[0118] Using the trained model M_Z2Y, according to the set P(Y|Z), the extremely low wind power output scenario and the extreme wind power climbing scenario are generated.

[0119] Wind power extreme event definition and identification module: Define wind power extreme events, including continuous multi-day extreme low output events where the daily average output is lower than a given threshold. Define wind power extreme ramping scenarios where the daily maximum upper fluctuation and daily maximum lower fluctuation are higher than a given threshold.

[0120] Extreme event identification and application module:

[0121] The generated daily output sequence features are used to identify wind power extreme events and generate multi-day wind power extreme low output scenarios with controllable characteristics.

[0122] Including early warning systems, risk assessment tools, etc., for wind power scenario management and optimized decision-making.

[0123] In another embodiment of the present invention, a terminal device is provided, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions to implement the corresponding method flow or corresponding functions; the processor described in the embodiment of the present invention can be used for the operation of the wind power extreme scene generation method.

[0124] In another embodiment of the present invention, the present invention further provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as at least one disk storage.

[0125] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the method for generating extreme wind power scenarios in the above-mentioned embodiment; one or more instructions in the computer-readable storage medium are loaded and executed by a processor.

[0126] 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 a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, 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, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0127] 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.

[0128] 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.

[0129] 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.

[0130] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the implementation methods of the present invention, and should be understood that the protection scope of the present invention is not limited to such special statements and embodiments. Those skilled in the art can make various other specific variations and combinations that do not deviate from the essence of the present invention based on the technical revelations disclosed by the present invention, and these variations and combinations are still within the protection scope of the present invention.

Claims

1. A method for generating extreme wind power scenarios caused by mesoscale weather processes, It is characterized in that The following steps are involved: Step 1, data collection: Collect meteorological data, including humidity, temperature, zonal wind speed, and radial wind speed; Collect wind power output data, including daily output data and daily ramp data; Step 2, data preprocessing: Clean and standardize meteorological data and wind power output data; Use transfer learning techniques to generate an expanded sample set for training the model; Step 3: Build the conditional probability distribution model M_X2Z and train it: The input feature encoder is constructed using the fully connected layer FCN and the 2D convolution layer Conv2d to map the meteorological feature X into the latent space. During the model training process, the denoising and denoising processes are introduced, and the encoded input matrix and latent space intermediate variables are processed using the stacked 2D convolution layer and the dilated convolution layer DilatedConv2d. The decoder is constructed using the residual stack layer skip stack and two layers of Conv2d to decode the latent variables into the daily power sequence feature Z. Step 4: Generate the characteristic scenario of the daily power sequence and define the extreme measurement: Use the trained M_X2Z to generate the characteristic scene of sunrise sequence under given meteorological conditions; Perform probability analysis on the generated daily power sequence characteristic scenarios and define the daily power sequence characteristic thresholds for low-probability wind power extreme power output scenarios; Step 5: Construct the conditional probability distribution model M_Z2Y and train it: Build model M_Z2Y using the same network structure as model M_X2Z; The conditional probability distribution between the daily output sequence feature Z and the wind power output scenario Y is fitted using the model M_Z2Y The daily output sequence feature Z is used as a condition to guide the generation of multi-day extreme low output scenarios, and the wind power extreme low output scenario Y is generated; Step 6: Generate extreme wind power scenarios: Using the trained conditional probability distribution model M_Z2Y, according to The setting generates extremely low wind power output scenarios; Use the model M_Z2Y to generate extreme wind power ramping scenarios; Step 7, defining extreme wind power scenarios, including: defining a continuous multi-day extreme low wind power output scenario where the daily average power is lower than a given threshold and defining an extreme wind power ramp scenario where the daily maximum upper fluctuation and the daily maximum lower fluctuation are higher than a given threshold; Step 8: Using the generated daily output sequence features and the extremely low wind power output scenario, the defined wind power extreme scenario is applied to identify the extreme scenario.

2. A method for generating wind power extreme scenarios according to claim 1, Features: The definition of the extremely low output scenario where the average daily power consumption is lower than a given threshold for multiple consecutive days is as follows: In the formula is the daily power generation sequence feature i starting from day d with a length of m, i = 1, ε is the wind power low output threshold, d is the starting day, m is the number of days, M e The threshold for extremely low output days.

3. A method for generating wind power extreme scenarios according to claim 1, Features: The wind power extreme ramping scenario where the daily maximum upper fluctuation and the daily maximum lower fluctuation are higher than a given threshold is defined as follows: In the formula is the characteristic value of the daily force sequence characteristic i on day d, y t is the wind power output value at time t, t is any time within the day, Δt is the time interval within the day, is the wind power output fluctuation threshold.

4. A method for generating wind power extreme scenarios according to claim 1, Features: The network loss functions used in model M_X2Z and model M_Z2Y are as follows: Where n is the number of diffusion steps, and s is n is the target sequence that gradually becomes clear after n denoising processes, ∈ is the noise injected at each step, and the commonly used normal distribution is selected as the prior distribution for the convenience of sampling and operation. c is the input condition, ∈θ is the denoising noise mapped by the denoising network.

5. A system for generating extreme wind power scenarios triggered by mesoscale weather processes. Features: The system can be used to implement the method for generating wind power extreme scenarios according to any one of claims 1 to 4, specifically comprising: Data collection module: collects meteorological data and wind power output data; Data preprocessing module: cleans and standardizes the collected meteorological data and wind power output data to ensure data quality and consistency; generates an expanded sample set for training the model; M_X2Z model building and training module: Use the fully connected layer FCN and the 2D convolution layer Conv2d to build the input feature encoder; The denoising and denoising processes are introduced to process the input matrix and latent space intermediate variables by stacking 2D convolutional layers and dilated convolutional layers; Use the residual stack layer skip stack and two layers of Conv2d to build a decoder to decode the hidden variables into the sunrise force sequence feature Z; Sunrise force sequence feature scene generation and extreme metric definition module: The trained M_X2Z model is used to generate the daily power sequence feature scenario under given meteorological conditions. The generated time series features include the daily average power, the daily maximum uphill climb and the daily maximum downhill climb. Perform probability analysis on the generated daily output sequence characteristic scenarios and define the daily output sequence characteristic thresholds for extremely low wind power output with low probability; M_Z2Y model construction and training module: Use the same network structure as M_X2Z to construct the conditional probability distribution model M_Z2Y; use M_Z2Y to fit the conditional probability distribution P(Y|Z) between the daily power sequence feature Z and the wind power output scenario Y; use the daily power sequence feature Z as a condition to guide the generation of wind power multi-day extreme low output scenarios, and generate wind power multi-day extreme low output scenarios Y; Wind power extreme scenario generation module: Using the trained model M_Z2Y, according to the set P(Y|Z), the wind power extreme low output scenario and wind power extreme climbing scenario are generated; Wind power extreme scenario definition and identification module: define wind power extreme scenarios, including scenarios of extremely low output for consecutive days where the daily average power is lower than a given threshold; define wind power extreme ramping scenarios where the daily maximum upper fluctuation and the daily maximum lower fluctuation are higher than a given threshold; Extreme scenario recognition and application module: Using the generated daily output sequence characteristics and the multi-day extreme low output scenario of wind power, the defined wind power extreme scenario is applied to identify the extreme scenario; Including early warning systems, risk assessment tools, etc., for wind power scenario management and optimized decision-making.

6. A computer device, Features: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for generating a wind power extreme scenario as claimed in any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium, Features: A computer program is stored thereon, and when the program is executed by a processor, the method for generating a wind power extreme scenario as described in one of claims 1 to 4 is implemented.