Typhoon transit area wind power output prediction method and terminal equipment

Abstract

The invention provides a typhoon transit area wind power output prediction method and terminal equipment. The typhoon transit area wind power output prediction method comprises the following steps that the radius of a typhoon center and the radius of each stage of wind ring on four quadrants in the next day are predicted; defining a shutdown wind ring and a commissioning full-power wind ring according to the maximum shutdown wind speed and the commissioning full-power wind speed of each wind power plant in the region; according to the predicted typhoon center movement direction and movement speed, the wind ring range of a shutdown wind ring and a commissioning full-generation wind ring in each time period is obtained, and the start-stop time of each wind power plant in the area is set; and predicting the output condition according to the position relationship between the wind power plant and the shutdown wind ring and the operation full wind ring. According to the method, the wind power output of a plurality of wind power plants in different areas such as a tenth-level wind ring, a seventh-level wind ring and a peripheral region of a typhoon motion path is subjected to classified prediction, and the influence and impact on wind power output prediction errors are actively reduced by setting the start-stop time of the wind power plants in the tenth-level wind ring range on the typhoon motion path; the influence of a typhoon passing wind power plant output climbing event on a power system is reduced, and the operation economy of the power system is improved.

Description

Technical Field

[0001] The present invention relates to the field of power system safety control, and in particular to a method and terminal equipment for predicting wind power output in a typhoon-transit area. Background Art

[0002] With the development of new power systems, the power structure and output characteristics of the power grid have changed greatly. Among them, the high proportion of wind power access system is one of the significant features of this change. However, as the proportion of wind power in the power system increases, it is more seriously affected by the climate environment. The impact of these climate changes may trigger new failure mechanisms and have a negative impact on the normal operation and benefits of wind farms. In particular, extreme weather conditions such as typhoons may cause large fluctuations in the power output of regional power grids, causing regional power grid power imbalances and power grid safety accidents.

[0003] By evaluating the output power of multiple wind farms in a province (hereinafter referred to as "wind power output"), we can predict anomalies and take corresponding measures to avoid power outages and even safety accidents caused by imbalance between power generation output and load of power grid. Therefore, it is of great significance to accurately predict wind power output.

[0004] In traditional wind power output prediction methods, meteorological factors such as wind speed and wind direction are usually used to establish a neural network model to predict wind power output. Since meteorological factors such as wind speed and wind direction will fluctuate violently during extreme weather such as typhoons, which will affect the accuracy of wind power output prediction, traditional prediction methods are difficult to achieve accurate predictions. During typhoons, due to the strong winds and air pressure changes brought by them, the wind speed in a large area will rise significantly, and the output of wind farms will quickly rise to the maximum power, and then quickly drop to 0 due to the wind turbine cut-out. The "large ups and downs" process causes the wind power output of the provincial power grid to climb. Specifically, when the typhoon wind speed exceeds the designed cut-out wind speed of the wind turbine, in order to protect the safety of the wind turbine, the wind turbine will automatically cut out and shut down. Due to the large impact range of typhoons, this cut-out and shutdown behavior often occurs in multiple wind farms in a large area, resulting in a sudden drop in the output of a large number of wind farms, thereby causing violent fluctuations in the wind power output of the regional power grid. Due to the strong randomness of typhoons in time and space, although the meteorological field has made significant breakthroughs in typhoon path prediction, there are relatively few studies on wind power output prediction during typhoons.

[0005] CN111146800A provides a method, device and storage medium for predicting power loss in a wind farm, which obtains weather forecast data for a future specified time period of the wind farm; based on the weather forecast data, analyzes the expected number of wind turbines that will be shut down in the wind farm during the specified time period; and predicts the power loss of the wind farm during the specified time period based on the initial predicted power and the expected number of wind turbines that will be shut down. However, this method is designed for the wind turbines in the wind farm to be cut out and shut down, and only predicts the number of wind turbines that will be shut down during the specified time period. This method predicts the number of wind turbines that will be shut down during a specified time period in the wind farm based on numerical weather forecast data in the area where the wind farm is located, but the wind speed prediction under a typhoon has a strong uncertainty, and predicting the number of wind turbines that will be cut out and shut down during a specified time period based on the wind speed prediction within a small range of the wind farm is likely to cause a large deviation. Summary of the invention

[0006] The purpose of the present invention is to provide a method and terminal device for predicting wind power output in a typhoon-passing area in view of the deficiencies of the above-mentioned prior art, so as to reduce the impact of wind farm output ramp-up events in typhoon-passing areas on the power system and improve the reliability of power system operation.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0008] A method for predicting wind power output in a typhoon-passing area comprises the following steps:

[0009] S1. Obtain the typhoon center and the radius of each level of wind circle in the quadrant in each time period within the current regional typhoon setting time in the meteorological forecast, wherein the quadrants include the northeast quadrant, northwest quadrant, southwest quadrant, and southeast quadrant with the typhoon center as the origin;

[0010] S2. When the wind farm in the area is located within the shutdown wind circle during a certain period of time, the wind farm stops operating; otherwise, the wind farm starts operating; the shutdown wind circle is the area between the tenth-level wind circle and the typhoon center;

[0011] S3, predicting the output of the wind farm according to the positional relationship between the wind farm and the shutdown wind circle and the fully operational wind circle; the fully operational wind circle is the area between the tenth-level wind circle and the seventh-level wind circle;

[0012] S4. Accumulate the output of multiple wind farms in the region in each time period within a set time period to obtain a regional wind power output prediction result in each time period.

[0013] The present invention predicts the tenth-level wind circle and the seventh-level wind circle in combination with the moving trajectory of the typhoon center, and uses the tenth-level wind circle and the seventh-level wind circle to simplify the problem. The wind speed of the tenth-level wind circle is >24.5m / s, and most of the wind turbines will be cut out and shut down (cut-out wind speed 25m / s); the wind speed between the seventh-level wind circle and the tenth-level wind circle is above the rated wind speed of the wind turbine, and the wind turbines in this area will be fully powered; the wind speed outside the seventh-level wind circle is lower than the rated wind speed, and the output of the wind turbine will change with the wind speed, and it is necessary to combine the wind speed prediction to make a power generation output prediction. Since the prediction of the moving trajectory of the typhoon center and the size of the wind circle is relatively macroscopic and stable, the prediction difficulty is less than the numerical weather forecast of the grid by grid in the typhoon area, and the use of the tenth-level wind circle and the seventh-level wind circle simplifies the cut-out wind turbines and the fully powered wind turbines, the regional wind power output predicted by the method of the present invention can exclude the influence of the wind speed prediction deviation in the tenth-level wind circle and the seventh-level wind circle, and the result of the wind power output prediction is more stable and accurate.

[0014] The present invention predicts the output of dozens or hundreds of wind farms in a region. The typhoon region can be divided into seven-level wind circles, ten-level wind circles and other parts, and then predicted in pieces. Generally, numerical weather forecasts predict the weather according to a 25*25 km grid. For weather events such as typhoons, the grid-based wind speed prediction method is generally difficult to ensure the accuracy of the prediction. Therefore, compared with the existing method of predicting the power generation output of a single wind farm based on the grid wind speed prediction results, the present invention is based on the prediction of the typhoon structure movement trajectory (typhoon center, ten-level wind circle and seven-level wind circle radius) The result is more stable and reduces the probability of large deviations.

[0015] In step S1, based on the published typhoon center position, central air pressure, maximum wind speed, radius of wind circles in four quadrants and moving speed of wind circles in each quadrant of the current typhoon, the range of wind circles of the typhoon on the moving path in each time period within the set time is determined, and the typhoon center of the current typhoon within the set time and the radius of wind circles at each level in the quadrant are obtained.

[0016] In step S3, the wind speed in the level 7 wind circle is greater than 13.9m / s, and the wind speed in the level 10 wind circle is greater than 24.5m / s. When the wind farm in the region is located between the level 10 wind circle and the level 7 wind circle, the wind speed reaches the rated wind speed of the wind turbine, and the wind farm in this range is fully operational, and the output of the wind farm is predicted to be the rated power; when the wind farm is located in the level 10 wind circle (the cut-out wind speed of the wind turbine is generally 25m / s), if the wind farm performs preventive control, all wind turbines will be shut down after entering the level 10 wind circle; if preventive control is not performed, most of the wind turbines in the wind farm will also be shut down because the wind speed is greater than the cut-out wind speed. It can be judged that the range between the level 10 wind circle and the typhoon center is the shutdown wind circle, and the output of the wind farm is predicted to be 0.

[0017] In order to further accurately predict the output of the wind farm, in step S3, when the wind farm is located outside the full-power wind circle, the specific implementation process of predicting the output of the wind farm includes:

[0018] Acquire historical operation data of the wind farm, construct a data set using the historical operation data, and randomly divide the data set into a training set and a test set; wherein the data set includes a plurality of data pairs, each data pair includes wind speed, wind direction, and corresponding power;

[0019] The training set is used as an input of a neural network, and the neural network is trained to obtain a prediction model.

[0020] The neural network is a Tab-net model. The Tab-Net model is an interpretable table learning neural network based on a sequential attention mechanism. On the basis of retaining the end-to-end and representation learning of a deep neural network, it also has the advantages of interpretability and sparse feature selection of a tree model. The input of Tab-Net can directly use raw data without any preprocessing, and is trained using gradient descent-based optimization, and can be flexibly integrated into end-to-end learning. The present invention adopts the Tab-net model to further improve the prediction efficiency and the accuracy of the prediction results.

[0021] In the present invention, when the wind farm is located outside the fully operational wind circle (corresponding to the seventh-level wind circle), the wind farm operation data acquired in real time is used as the input of the prediction model to obtain the predicted wind farm output.

[0022] In the present invention, in order to ensure data accuracy and take into account prediction efficiency, the set time is 24 hours.

[0023] As an inventive concept, the present invention also provides a terminal device, comprising:

[0024] one or more processors;

[0025] A memory having one or more programs stored thereon, which, when executed by the one or more processors, enables the one or more processors to implement the steps of the above method of the present invention.

[0026] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention provides a method for predicting wind power output in a typhoon-transit area, which starts from the two aspects of suppressing volatility and improving accuracy, and classifies and predicts the wind power output in different areas such as the tenth-level wind circle, the seven-level wind circle and the peripheral areas along the typhoon's movement path. By setting the start and stop time of the wind farm within the tenth-level wind circle on the typhoon path, the influence and impact of the wind power output forecast error are actively reduced, the impact of the typhoon-transit wind power output climbing event on the power system is reduced, and the economy of the power system operation is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a schematic flow chart of a method for predicting wind power output in a typhoon-passing area according to an embodiment of the present invention;

[0028] Figure 2 This is an example diagram of the wind circle range of a typhoon on the moving path according to an embodiment of the present invention;

[0029] Figure 3 The Tab-Net encoder architecture diagram of an embodiment of the present invention;

[0030] Figure 4 This is a structural diagram of the attention mechanism converter according to an embodiment of the present invention;

[0031] Figure 5 This is a structural diagram of a feature converter according to an embodiment of the present invention;

[0032] Figure 6 4 is a decoder structure diagram of Tab-Net according to an embodiment of the present invention;

[0033] Figure 7 This is a diagram of Tab-Net prediction results according to an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The following will be combined with the accompanying drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. It should be understood by those skilled in the art that these implementations are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0035] Example 1

[0036] like Figure 1 As shown, the present invention provides a method for predicting power of a wind farm during a typhoon, comprising the following steps:

[0037] S1. Predict the typhoon center of the current typhoon and the radius of each level of wind circle in the quadrant in each period of the next day. The quadrants include the northeast quadrant, northwest quadrant, southwest quadrant, and southeast quadrant with the typhoon center as the origin.

[0038] Since the meteorological data involved in a typhoon include air pressure, the location of the typhoon center, the range of each level of wind circles, the wind direction of each typhoon area, etc., and the wind speed and direction inside each wind circle are also different. If all meteorological factors are considered, wind power prediction will be more complicated and more difficult. Therefore, in order to simplify the prediction process, this embodiment 1 only considers the quadrant radius of the typhoon center and the tenth level wind circle and the seventh level wind circle.

[0039] Specifically, Figure 2As shown, in the step S1, according to the typhoon center position, central air pressure, maximum wind speed, radius of wind circles in four quadrants and moving speed of wind circles in each quadrant of the current typhoon released by the meteorological station, the range of wind circles on the movement path of the typhoon in each period in the next 24 hours is determined, and the typhoon center of the current typhoon and the radius of wind circles at all levels in the quadrants in the next day are obtained.

[0040] Step S2: according to the maximum shutdown wind speed and the full-power wind speed of the wind farm, determine the range of the current typhoon level 10 wind circle and level 7 wind circle, and define the shutdown wind circle and the full-power wind circle.

[0041] According to the radial profile of wind speed, outside the maximum wind speed radius, when the azimuth average wind speed decays to a certain size, the corresponding radius is the size of the distance from the typhoon center. If you draw a circle with this radius at the center of the typhoon, you will get a wind circle, which represents the range covered by the wind field with a wind speed greater than a certain wind speed in the low-level wind field of the typhoon outside the maximum wind speed radius. The radius size of the typhoon level 10 wind circle and level 7 wind circle refers to the distance from the typhoon center when the near-ground wind speed decays to 24.5m / s and 13.9m / s outside the maximum wind speed radius. Generally speaking, the level 7 wind circle is the area most widely affected by the typhoon and is also one of the areas most prone to disasters. The level 10 wind circle refers to the area within this range where the average wind speed reaches level 10 or above. The wind speed range of level 10 wind is generally 24.5-28.4m / s, which is a significant sign of the typhoon's intensity. The level 10 wind circle usually means that the typhoon is already very strong and will cause serious disasters to the local area.

[0042] It should be noted that the ranges of the level 7 and level 10 wind circles of typhoons of different intensities will also be different.

[0043] Specifically, in this embodiment 1, the wind speed in the level 7 wind circle is greater than 13.9 m / s, and the wind speed in the level 10 wind circle is greater than 24.5 m / s. When the wind farm is located between the level 10 wind circle and the level 7 wind circle, the wind farm reaches the full-power wind speed for operation, and the wind farm in this range is fully operated, and the range between the level 10 wind circle and the level 7 wind circle is judged as the full-power wind circle for operation; when the wind farm is located in the level 10 wind circle, the wind farm frequently fluctuates within the cut-out wind speed, and the wind farm in this range is cut out and shut down, and the range between the level 10 wind circle and the typhoon center is judged as the shutdown wind circle.

[0044] The full-power wind speed for operation refers to the minimum wind speed at which the wind turbine can achieve full power efficiency during normal operation.

[0045] Step S3: According to the predicted moving direction and moving speed of the typhoon center, the wind circle ranges of the shutdown wind circle and the fully operational wind circle in each time period are obtained, and the start and stop time of each wind farm is set.

[0046] Specifically, in step S3, after predicting the time for each wind farm to enter and exit the level 10 wind circle, the start and stop time of the wind farm located on the typhoon path is set. When the wind farm is within the shutdown wind circle, the wind farm stops running; when the wind farm is not within the shutdown wind circle, the wind farm starts running.

[0047] Step S4: predicting the output of the wind farm according to the positional relationship between the wind farm and the shutdown wind circle and the fully operational wind circle.

[0048] Specifically, in step S4, when the wind farm is located in the fully operational wind circle, the output of each wind turbine in the wind farm is predicted to be its rated power; when the wind farm is located in the shutdown wind circle, the output of each wind turbine in the wind farm is predicted to be 0; when the wind farm is located outside the fully operational wind circle, the wind speed is not affected by the typhoon, and the Tab-Net wind farm output prediction model is used to predict the wind farm output.

[0049] The Tab-Net model is an interpretable tabular learning neural network based on the sequential attention mechanism. While retaining the end-to-end and representation learning of deep neural networks, it also has the advantages of interpretability and sparse feature selection of tree models.

[0050] Tab-Net can directly use raw data as input without any preprocessing, and is trained using gradient descent-based optimization, which can be flexibly integrated into end-to-end learning.

[0051] Tab-Net uses sequential attention to select features to be inferred in each decision step.

[0052] Tab-Net encoder architecture is as follows Figure 3 shown.

[0053] The Tab-Net encoder architecture mainly consists of feature transformers, attention transformers, and feature masks for each decision step. Tabular data includes categorical data and numerical data. TabNet uses raw numerical data and maps categorical features to numerical features using trainable embeddings. Each decision step inputs the same B×D feature matrix; B is the batch size and D is the dimension of the feature. The features of each decision step of Tab-Net are determined by the output of the previous decision step through the attention transformer, and the output processed feature representation is aggregated into the overall decision.

[0054] The structure of the attention mechanism converter is as follows Figure 4 As shown, the shielding layer performs feature selection through the output of the attention mechanism converter:

[0055] M[i]=Sparsemax(P[i-1]).h i (a[i-1])

[0056]

[0057] Where: a[i] represents the input feature; h i represents the combination of the fully connected layer and the batch normalization (BN) layer, h i The function of is to realize the linear combination of features; P[i] is the prior scale information, which summarizes the importance of the features in the previous decision step before the current decision step. Sparsemax (Martins and Astudillo 2016) is used to normalize the coefficients to achieve sparse selection of prominent features. The Sparsemax layer makes the feature selection more sparse; M[i] is the shielding layer matrix; L sparse represents the sparsity of M[i], L sparse The smaller it is, the sparser M[i] is.

[0058] The feature converter structure of Tab-Net is as follows Figure 5 shown.

[0059] The feature converter is mainly composed of a fully connected (FC) layer, a batch normalization (BN) layer, and a gated linear unit (GLU) layer. A feature conversion block has a 4-layer network, 2 of which are shared in all decision steps and trained together in all steps to achieve common feature calculations; the last 2 layers are related to the decision steps and are trained and calculated separately at each step.

[0060] The decoder structure of Tab-Net is as follows Figure 6 As shown, it consists of a feature transformer, an attention transformer, and a feature mask. The representation processed by a segmentation block is used by the attention transformer in the subsequent steps and also for the entire output. For, the feature selection mask provides interpretable information about the model function for each step, and the summary mask obtains the important attribution of the global feature.

[0061] Using the Tab-net model, the hourly operating data of a wind farm for one year is used as a training sample, and wind speed and wind direction are used as input to predict power. The first 75% of the data is selected as the training set, and the last 25% of the data is selected as the test set. The grid search algorithm is used to traverse the given parameter combination for parameter tuning. Finally, the hyperparameter settings that make the TabNet prediction results optimal are obtained. Tab-Net model parameters, prediction results, model mean absolute error MAE, mean square error MSE and determination coefficient r2 are shown in Tables 1 and 2. Figure 7 shown.

[0062] Table 1 Tab-Net model parameters

[0063]

[0064] Table 2 Evaluation indicators

[0065]

[0066]

[0067] Example 2

[0068] Embodiment 2 of the present invention provides a terminal device corresponding to the above-mentioned embodiment 1. The terminal device may be a processing device for a client, such as a mobile phone, a laptop computer, a tablet computer, a desktop computer, etc., to execute the method of the above-mentioned embodiment.

[0069] The terminal device of the second embodiment includes a memory, a processor, and a computer program stored in the memory; the processor executes the computer program in the memory to implement the steps of the method of the first embodiment above.

[0070] In some implementations, the memory may be a high-speed random access memory (RAM), and may also include a non-volatile memory (non volatile memory), such as at least one disk memory.

[0071] In some other implementations, the processor may be a central processing unit (CPU), a digital signal processor (DSP), or other general-purpose processors of various types, which are not limited herein.

[0072] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, which all belong to the protection scope of the present invention.

[0073] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0074] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A method for predicting wind power output in a typhoon-passing area, characterized in that: The following steps are involved: S1. Obtain the typhoon center and the radius of each level of wind circle in the quadrant in each time period within the current regional typhoon setting time in the meteorological forecast, wherein the quadrants include the northeast quadrant, northwest quadrant, southwest quadrant, and southeast quadrant with the typhoon center as the origin; S2. When the wind farm in the area is located within the shutdown wind circle during a certain period of time, the wind farm stops operating; otherwise, the wind farm starts operating; the shutdown wind circle is the area between the tenth-level wind circle and the typhoon center; S3, predicting the output of the wind farm according to the positional relationship between the wind farm and the shutdown wind circle and the fully operational wind circle; the fully operational wind circle is the area between the tenth-level wind circle and the seventh-level wind circle; S4. Accumulate the output of multiple wind farms in the region in each time period within a set time period to obtain a regional wind power output prediction result in each time period.

2. The method for predicting wind power output in a typhoon-passing area according to claim 1, characterized in that: In step S1, the range of the wind circle of the typhoon on the movement path of each time period of the typhoon within the set time is determined according to the published typhoon center position, central air pressure, maximum wind speed, radius of the wind circle on the four quadrants and moving speed of the wind circle in each quadrant within the set time, and the typhoon center of the current typhoon within the set time and the radius of the wind circle at each level in the quadrant are obtained.

3. The method for predicting wind power output in a typhoon-passing area according to claim 1, characterized in that: In the step S3, when the wind farm is located in the fully operational wind circle, the output of the wind farm is predicted to be rated power; when the wind farm is located in the shutdown wind circle, the output of the wind farm is predicted to be 0.

4. The method for predicting wind power output in a typhoon-passing area according to claim 1, characterized in that: In step S3, when the wind farm is located outside the full-power wind circle, the specific implementation process of predicting the output of the wind farm includes: Acquire historical operation data of the wind farm, construct a data set using the historical operation data, and randomly divide the data set into a training set and a test set; wherein the data set includes a plurality of data pairs, each data pair includes wind speed, wind direction, and corresponding power; The training set is used as an input of a neural network, and the neural network is trained to obtain a prediction model.

5. The method for predicting wind power output in a typhoon-passing area according to claim 4, characterized in that: The neural network is a Tab-net model.

6. The method for predicting wind power output in a typhoon-passing area according to claim 4, characterized in that: When the wind farm is located outside the full-power wind circle, the real-time wind farm operation data is used as the input of the prediction model to obtain the predicted wind farm output.

7. The method for predicting wind power output in a typhoon-passing area according to any one of claims 1 to 7, characterized in that: The set time is 24h.

8. A terminal device, characterized in that: include: one or more processors; A memory having one or more programs stored thereon, which, when the one or more programs are executed by the one or more processors, enables the one or more processors to implement the steps of the method according to any one of claims 1 to 7.

Images