Wind turbine generators and compressed air energy storage energy management systems that dynamically match multiple wind conditions

Abstract

The present invention relates to the field of wind power generation system management, specifically to a wind turbine and compressed air energy storage energy management system that dynamically matches multiple wind conditions. The present application realizes dynamic response by setting up a main and a sub-storage unit. In order to improve efficiency, the sub-storage unit is generally in a dormant state. The present application obtains meteorological forecast data in real time and makes rolling predictions on abnormal wind conditions that may occur in future time periods. Once abnormal wind conditions such as gusts, turbulence, and abnormal decreases in wind power are predicted, the sub-storage unit is immediately started and the compressed air in the sub-storage unit is preheated. When abnormal wind conditions occur, the main and sub-storage units with opposite working states can be used to quickly respond to wind turbine power fluctuations caused by abnormal wind conditions, thereby avoiding fluctuations in grid-connected power.

Description

Technical Field

[0001] The present invention relates to the field of wind power generation system management, and in particular to a wind turbine and compressed air energy storage management system that dynamically matches multiple wind conditions. Background Art

[0002] Air energy storage units, commonly referred to as compressed air energy storage (CAES), use electricity to compress and store air, releasing it when needed to drive turbines for power generation. It is a type of large-scale energy storage technology suitable for balancing grid loads and improving the utilization of renewable energy.

[0003] Because wind power generation is affected by the environment, its power generation fluctuates. The amount of wind power connected to the grid is also affected by both the generation and consumption sides. In these situations, the introduction of air energy storage units can help shift peak loads and fill valleys. Therefore, compressed air storage units play a vital role in improving wind power utilization and ensuring stable power supply.

[0004] However, current air energy storage units experience a lag effect when implementing peak-shaving and valley-filling in power generation. This is because turbine startup requires time, and the compressed air temperature must be controlled to improve efficiency during both the storage and power generation phases. Turbulence, gusts, or sudden drops in wind speed in the wind power system area can easily cause dramatic changes in wind turbine generator power. If the air energy storage unit is unable to respond in a timely manner, this can easily cause grid fluctuations. Summary of the Invention

[0005] In view of this, an object of the present invention is to provide a wind turbine and compressed air energy storage management system that dynamically matches multiple wind conditions, which can solve the technical problems in the background technology.

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

[0007] an acquisition module, configured to acquire predicted atmospheric state data for a future time period, the operating state and operating power of the main storage unit at a current time point, and the real-time generated power of the wind turbine, wherein the future time period includes multiple future time points of a target duration after the current time point, and the predicted atmospheric state data includes temperature, wind speed, air pressure, and humidity;

[0008] A prediction module, configured to predict the wind conditions for a future time period based on a pre-built wind condition prediction model and the meteorological forecast data to obtain predicted wind conditions, wherein the wind condition prediction model characterizes the corresponding relationship between the atmospheric state data and the wind conditions;

[0009] a pre-start module, configured to start the secondary storage unit and pre-regulate the air temperature in the secondary storage unit when the predicted wind condition is an abnormal wind condition, wherein the primary storage unit is configured to respond to energy storage or power generation at the current time point, and the secondary storage unit is configured to respond to wind turbine power fluctuations that may occur in a future time period;

[0010] A management module is used to extract the change characteristics of the real-time power generation power, match the change characteristics with the abnormal wind conditions, and obtain a matching result; and dynamically adjust and manage the working status and working power of the main storage unit and the working status and working power of the secondary storage unit based on the real-time power generation power and the matching result.

[0011] Furthermore, the method for constructing the wind condition prediction model includes:

[0012] Acquire a plurality of historical sample data, wherein the historical sample data includes atmospheric state data, ground wind speed, and ground wind direction at a plurality of historical time points;

[0013] Screen out a variety of abnormal wind condition target sample data based on the ground wind speed and ground wind direction at multiple historical time points , , For the Atmospheric state data at historical time points The value of Indicates the type of abnormal wind conditions;

[0014] Classify the target sample data based on the type of abnormal wind conditions to obtain multiple data sets;

[0015] Extract each target sample data in each data set The value characteristics of each atmospheric state data , fluctuation characteristics And the changing characteristics , and based on the value characteristics , the fluctuation characteristics And the changing characteristics The constructed feature vector , , is the sample data sequence number;

[0016] Based on the feature vector A deep learning model and a statistical model are constructed respectively, and a wind condition prediction model is constructed based on the deep learning recognition model and the statistical model, wherein the output of the wind condition prediction model is obtained by weighting the output results of the deep learning recognition model and the output results of the statistical model, and the weight is determined based on the cumulative number of learning times of the deep learning recognition model.

[0017] Furthermore, based on the feature vector Build deep learning models and statistical models separately, including:

[0018] Density clustering is performed on the characteristic vectors of each atmospheric state data in each data set to obtain multiple clusters of each atmospheric state data. ;

[0019] Clusters with a data volume greater than a preset threshold As the target cluster, a reference feature vector is constructed based on multiple feature vectors in the target cluster. and reference range , , reference range , represents the average value of the parameters in brackets, Indicates the reference range of the parameter in brackets;

[0020] Calculate each eigenvector and the reference eigenvector of each abnormal wind condition The cosine similarity between , The mathematical expression is:

[0021] ,

[0022] The cosine similarity As a feature vector The label of the abnormal wind condition possibility is used to obtain training data, wherein when there are multiple labels of the same abnormal wind condition possibility, the label is the average value of multiple cosine similarities;

[0023] A training data set is constructed based on the training data, and an artificial neural network is trained based on the training data set to obtain a deep learning recognition model; a statistical model is constructed based on multiple reference ranges of multiple types of abnormal wind conditions.

[0024] Furthermore, target sample data with abnormal wind conditions are screened out based on the ground wind speed and ground wind direction at multiple historical time points, including:

[0025] Based on the pre-built sliding window, it slides along the time axis of the historical sample data, and calculates the wind speed change within the sliding window at each slide. and wind direction variance , where wind speed change , is the maximum wind speed, is the minimum wind speed;

[0026] The wind speed change Greater than the preset change threshold, or wind direction variance The data interval corresponding to the sliding window that is greater than the preset variance threshold is regarded as the abnormal interval, and the historical sample data containing the abnormal interval is regarded as the target sample data.

[0027] Furthermore, the wind conditions in the future time period are predicted based on the pre-built wind condition prediction model and the meteorological forecast data to obtain the predicted wind conditions, including:

[0028] Extract the value characteristics of each atmospheric state data from the meteorological forecast data , fluctuation characteristics And the changing characteristics ; and based on the value characteristics , fluctuation characteristics And the changing characteristics Constructing the input vector ;

[0029] The input vector Input into the deep learning recognition model to obtain the first probability of multiple abnormal wind conditions ; and the input vector Input into the statistical model to obtain the second probability of multiple abnormal wind conditions , wherein, when the input vector falls into the reference range of different wind conditions, a second probability is performed based on the number of data corresponding to the target cluster in the reference range;

[0030] The probability of each abnormal wind condition is weighted and summed to obtain the final probability of each abnormal wind condition. , where the final probability The mathematical expression is:

[0031] ,

[0032] Where, The amount of trained data for deep learning recognition models, The target amount of training data for deep learning recognition models, ;

[0033] When there is an abnormal wind condition with a final probability greater than the preset probability threshold, the abnormal wind condition with a final probability greater than the preset probability threshold is used as the prediction result; otherwise, it is determined that there is no abnormal wind condition risk.

[0034] Furthermore, the value feature , the fluctuation characteristics And the changing characteristics The extraction process includes:

[0035] Calculate the average value of atmospheric state data and obtain the value characteristics ; Calculate the standard deviation of atmospheric state data to obtain fluctuation characteristics ; and perform linear regression fitting on the atmospheric state data to obtain the slope of the atmospheric state data, and use the slope as a change feature .

[0036] Furthermore, starting the secondary storage unit and pre-adjusting the air temperature in the secondary storage unit includes:

[0037] Determining an operating state of the main storage unit, wherein the operating state includes an energy storage state and a power generation state;

[0038] When the main storage unit is in an energy storage state and the abnormal wind condition is an abnormal drop in wind force or turbulence, starting the turbine generator of the auxiliary storage unit and preheating the compressed air in the auxiliary storage unit to a first temperature value;

[0039] When the main storage unit is in a power generation state and the abnormal wind condition is a gust or turbulence, the motor and compressor unit of the secondary storage unit are started, and the compressed air in the secondary storage unit is cooled to a second temperature value, wherein the first temperature value is higher than the second temperature value.

[0040] Furthermore, extracting the variation characteristics of the real-time generated power and matching the variation characteristics with the abnormal wind conditions to obtain a matching result includes:

[0041] Sampling the current time segment of the real-time generated power to obtain the generated power at multiple time points, wherein the current time segment is a time period of unit length before the current time point;

[0042] Calculating the variance of the power generation at multiple time points, and performing linear regression fitting on the power generation at multiple time points to obtain a slope; and matching the abnormal wind condition based on the variance of the power generation at the multiple time points and the slope to obtain a matching result.

[0043] Furthermore, dynamically adjusting and managing the working state and working power of the primary storage unit and the working state and working power of the secondary storage unit based on the real-time generated power and the matching result includes:

[0044] When the change characteristics match the abnormal wind conditions, the grid-connected power requirement of wind power generation is obtained. ;

[0045] Calculate the grid-connected power requirement With the real-time power generation The difference ,exist When the storage power of the storage unit in the energy storage state in the main storage unit and the auxiliary storage unit is set to ;

[0046] exist When the storage power of the storage unit in the power generation state among the main storage unit and the auxiliary storage unit is set to ,in, is the minimum holding power of another storage unit.

[0047] Furthermore, the main storage unit and the secondary storage unit are connected via an electrically controlled valve.

[0048] The beneficial effects of the present invention are as follows: the wind turbine and compressed air energy storage energy management system of the present invention that dynamically matches multiple wind conditions, the present application realizes dynamic response by setting up main and sub-storage units. In order to improve efficiency, the sub-storage unit is generally in a dormant state. The present application obtains meteorological forecast data in real time and makes rolling predictions on abnormal wind conditions that may occur in future time periods. Once abnormal wind conditions such as gusts, turbulence, and abnormal decrease in wind force are predicted, the sub-storage unit is immediately started and the compressed air in the sub-storage unit is preheated. When abnormal wind conditions occur, the main and sub-storage units with opposite working states are used to quickly respond to power fluctuations of wind turbines caused by abnormal wind conditions, thereby avoiding fluctuations in grid-connected power. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0050] Figure 1 This is an application scenario diagram of a wind turbine and a compressed air energy storage management system that dynamically matches multiple wind conditions in this application;

[0051] Figure 2 This is a structural diagram of a wind turbine and a compressed air energy storage management system that dynamically matches multiple wind conditions, as shown in one embodiment of the present application;

[0052] Figure 3 This is a schematic diagram of the model construction process in one embodiment of the present application;

[0053] Figure 4 Schematic diagram of the wind condition prediction process in one embodiment of the present application. DETAILED DESCRIPTION

[0054] The following describes the embodiments of the present invention through specific examples. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0055] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention. Therefore, the drawings only show the layers related to the present invention and are not drawn according to the number, shape and size of the layers in actual implementation. In actual implementation, the type, quantity and proportion of each layer can be changed arbitrarily, and the layer layout type may also be more complicated.

[0056] In the following description, numerous details are set forth to provide a more thorough explanation of the embodiments of the present invention; however, it is apparent to one skilled in the art that the embodiments of the present invention may be practiced without these specific details.

[0057] One of the main problems with wind power generation at wind power stations is the volatility and instability of their output power. Due to the randomness and uncontrollability of wind energy itself, wind turbines cannot generate stable electricity, resulting in difficulties in grid connection and use. This phenomenon is called "wind abandonment". In order to solve this problem, the prior art has proposed a technology for converting wind energy into compressed air energy storage. Compressed air energy storage is a technology that converts excess wind energy into high-pressure air and stores it, and then releases it when needed to drive turbines to generate electricity. Specifically, wind turbines drive air compressors to generate high-pressure air, which is stored and then released when needed to drive turbines to generate electricity. This technology can not only balance the volatility of wind energy, but also achieve stable power generation around the clock.

[0058] Figure 1 This is an application scenario diagram of a wind turbine generator and a compressed air energy storage energy management system that dynamically matches multiple wind conditions in this application, such as Figure 1 As shown, the present application is implemented in a wind power station and includes a primary storage unit 110, a secondary storage unit 120, and a management control unit 130. The motor and compressor components of the primary and secondary storage units 110, 120 are connected to the wind turbine generators. During startup, they compress air and pump it into a container. The generators of the primary and secondary storage units 110, 120 are connected to the power grid via a coupling device. During periods of high demand, compressed air is released to drive the blades of the generators, generating electricity.

[0059] The management and control unit 130 is used to obtain real-time weather forecast data from the network and perform rolling forecast analysis to predict possible abnormal wind conditions. Abnormal wind conditions can have a significant impact on wind power generation, causing rapid changes or fluctuations. Therefore, this application uses the coordination of the primary storage unit 110 and the secondary storage unit 120 to achieve rapid response and avoid power grid fluctuations.

[0060] Figure 2 This is a structural diagram of a wind turbine generator and a compressed air energy storage energy management system that dynamically matches multiple wind conditions, as shown in an embodiment of the present application. Figure 2 As shown: The wind turbine generator and compressed air energy storage energy management system for dynamically matching multiple wind conditions in this embodiment includes:

[0061] an acquisition module 210 for acquiring predicted atmospheric state data for a future time period, acquiring the operating state and operating power of the main storage unit at a current time point, and acquiring the real-time generated power of the wind turbine, wherein the future time period includes multiple future time points of a target duration after the current time point, and the predicted atmospheric state data includes temperature, wind speed, air pressure, and humidity;

[0062] Meteorological data in the atmosphere are closely related to surface wind conditions. For example: (1) Temperature differences are one of the main factors driving wind formation. When the surface of a region warms faster than the surrounding areas, the air in that region expands and rises, forming a low-pressure area on the surface. The surrounding cold air flows toward this low-pressure area, generating wind. This phenomenon is common in coastal or desert areas, where the land warms faster than the ocean during the day, causing sea breezes to blow from the ocean to the land. (2) Wind speed directly reflects the strength of surface wind conditions. It is affected by many factors, including the pressure gradient force on the horizontal plane (i.e., the difference in air pressure between different locations) and the Coriolis effect caused by the rotation of the earth. Generally speaking, the greater the pressure difference, the faster the wind speed. (3) The distribution of air pressure has a direct impact on the direction and speed of wind. Air always flows from high-pressure areas to low-pressure areas in an attempt to balance the pressure difference. Therefore, understanding the air pressure conditions at different locations in an area can help predict the direction of wind. In addition, the contrast in strength between high and low pressure systems also affects wind speed; generally, the steeper the pressure gradient, the faster the wind speed. (4) Humidity itself does not directly cause wind formation, but it indirectly affects wind by affecting air density. Wet air is lighter than dry air because it contains more water vapor molecules, which can change the air's buoyancy and flow patterns. For example, in some cases, the rise of humid air can cause localized changes in airflow, which in turn affects the direction and speed of wind. In addition, humidity is crucial for storm development because precipitation is more likely to form under high humidity conditions.

[0063] Therefore, this application obtains atmospheric state forecast data for future time periods (such as the next hour) through meteorological public information channels.

[0064] In addition, under normal circumstances, only the main storage unit participates in the peak-shaving and valley-filling mechanism of the wind power system. The auxiliary storage unit is in a dormant state. However, if abnormal wind conditions occur, it may be impossible to respond to abnormal wind conditions due to the regulation lag of the main storage unit. For example, the wind force drops abnormally, causing the power of the wind turbine to drop sharply. At this time, if the compressor is working, it is necessary to shut down the compressor, restart the turbine generator, and heat the compressed air inside (the function is to expand the air and improve the power generation efficiency). There is a high probability that the generator power drop will not be responded to in time, resulting in power grid fluctuations. In order to avoid the above situation, this application needs to predict abnormal wind conditions, and the prediction process is described as follows.

[0065] A prediction module 220 is configured to predict the wind conditions in a future time period based on a pre-built wind condition prediction model and the meteorological forecast data to obtain predicted wind conditions, wherein the wind condition prediction model represents the corresponding relationship between the atmospheric state data and the wind conditions;

[0066] The prediction model in this application uses a combination of statistical models and AI models. This is because ground wind conditions are not only related to meteorological data, but also to terrain features. Obstacles such as mountains and buildings can cause wind channel effects and vortex phenomena, thereby changing the wind speed and direction in local areas. Different wind power stations are located in unique terrains, so it is difficult to find universal sample data to train AI models, and the amount of data required for AI models is much greater than that of statistical models.

[0067] Therefore, this application adopts a combination of statistical models and AI models, uses a small amount of sample data to construct statistical models and AI models, and as the amount of data continues to increase, increases the credibility weight of the AI ​​model output structure, and gradually transitions from a model that is mainly based on statistical models and supplemented by AI models to a model that is mainly based on AI models. Figure 3 This is a schematic diagram of the model construction process in an embodiment of the present application, such as Figure 3 As shown, the model is constructed as follows:

[0068] (1) Acquiring a plurality of historical sample data, wherein the historical sample data includes atmospheric state data, ground wind speed, and ground wind direction at a plurality of historical time points;

[0069] The multiple historical sample data are sample data collected at the current wind power station. Only by collecting the sample data of the current wind power station can the correspondence between the current regional atmospheric data and the ground wind conditions be accurately reflected.

[0070] (2) Screening out a variety of abnormal wind condition target sample data based on the ground wind speed and ground wind direction at multiple historical time points , , For the Atmospheric state data at historical time points The value of Indicates the type of abnormal wind conditions;

[0071] First, we need to extract target sample data from historical sample data that shows abnormal wind conditions. This application targets three types of abnormal wind conditions that can cause abnormal changes in wind power: a) gusts, where wind speed suddenly increases, leading to an abnormal increase in wind power. b) wind drops, where wind speed suddenly decreases, leading to an abnormal decrease in wind power. c) turbulence, where wind direction and wind speed fluctuate erratically, leading to abnormal fluctuations in wind power.

[0072] In view of the above three situations, this application uses the following method to extract target sample data The specific process includes:

[0073] (2-1) Based on the pre-built sliding window, slide along the time axis of the historical sample data, and calculate the wind speed change within the sliding window at each slide. and wind direction variance , where wind speed change , is the maximum wind speed, is the minimum wind speed;

[0074] In this application, the wind direction is the angle with the reference direction, such as the angle with the south direction. The mathematical expression is:

[0075] ,

[0076] Where, The first The wind direction in the historical sample data, is the number of historical sample data in the sliding window, is the average wind direction (i.e., the average angle) of the historical sample data within the sliding window.

[0077] Since wind direction is a cyclic data, in order to avoid calculation errors when calculating the average wind direction, each wind direction is converted into two components, with a certain wind direction angle For example, the two components are: , ; then calculate the average of all components and , and finally the average angle is obtained by inverse trigonometric function , .

[0078] In this embodiment, the time scale of historical sample data is relatively long, for example, 2-5 hours, while abnormal wind conditions often occur in a short period of time. Therefore, in order to extract abnormal wind events in a short period of time, this application uses a window width of 0.2-0.5 hours to extract features from historical samples.

[0079] The characteristics include wind speed variation and wind direction variance. Wind speed variation is the difference between the maximum wind speed and the minimum wind speed, reflecting the maximum change in wind speed during this period. Wind direction variance reflects the stability of wind direction within the time window. Fluctuations in wind direction will also cause fluctuations in wind speed. and wind direction variance It can determine whether there is a rapid increase or decrease in wind speed, as well as the fluctuation of wind speed or direction.

[0080] The wind direction variance in this embodiment can be adjusted by extracting wind direction samples during a period of stable wind conditions, calculating the variance of the wind direction samples, and combining experience to adjust the setting. For example, if the wind direction variance exceeds 0.5, it may be considered that the wind direction has changed significantly during this period.

[0081] (2-2) The wind speed change Greater than the preset change threshold, or wind direction variance The data interval corresponding to the sliding window that is greater than the preset variance threshold is regarded as the abnormal interval, and the historical sample data containing the abnormal interval is regarded as the target sample data.

[0082] Wind speed change Greater than the preset change threshold, or wind direction variance When it is greater than the preset variance threshold, it indicates that there is an abnormal change in the wind condition characteristics within the window, so the corresponding historical sample data is the target sample data.

[0083] In addition, in order to unify the time scale of the sample data, this application uniformly extracts the data of the abnormal time period and the period before the abnormal time period as the target sample data. The time scale of the target sample data is one hour.

[0084] (3) Classify the target sample data based on the type of abnormal wind conditions to obtain multiple data sets;

[0085] As mentioned above, when extracting wind speed variation and wind direction variance, the abnormal wind condition type of the target data sample can be directly labeled. For example, a positive wind speed variation indicates gusty winds; a negative wind speed variation indicates an abnormal decrease in wind speed; and a large wind direction variance indicates turbulence.

[0086] (4) Extract each target sample data in each data set The value characteristics of each atmospheric state data , fluctuation characteristics And the changing characteristics , and based on the value characteristics , the fluctuation characteristics And the changing characteristics The constructed feature vector , , is the sample data sequence number;

[0087] The wind on the surface is caused by the dynamic changes of air in the atmosphere, so by extracting the target sample data The dynamic characteristics of the wind are used to analyze the corresponding relationship with the types of abnormal wind conditions.

[0088] Specifically, the value feature is the average value, the fluctuation characteristics is the standard deviation, the variation characteristics is the overall slope, and the extraction process includes:

[0089] Calculate the average value of atmospheric state data and obtain the value characteristics ; Calculate the standard deviation of atmospheric state data to obtain fluctuation characteristics ; and perform linear regression fitting on the atmospheric state data to obtain the slope of the atmospheric state data, and use the slope as a change feature .

[0090] When fitting linear regression, first construct a linear relationship function ,in, is the value of atmospheric state data (such as the value of atmospheric temperature), For time, is the slope, is the error term.

[0091] Then, the data of multiple time points in the target sample data are brought into the linear relationship function, and the slope reflecting the overall change can be obtained by fitting it through the least squares method. Slope It can reflect the approximate rate of change of the atmospheric state data in the target sample data over time, so it is used as a change feature .

[0092] (5) Based on the feature vector A deep learning model and a statistical model are constructed respectively, and a wind condition prediction model is constructed based on the deep learning recognition model and the statistical model, wherein the output of the wind condition prediction model is obtained by weighting the output results of the deep learning recognition model and the output results of the statistical model, and the weight is determined based on the cumulative number of learning times of the deep learning recognition model.

[0093] Finally, based on the previous extraction Build statistical models and AI analysis models. The specific construction process is as follows:

[0094] (5-1) Density clustering is performed on the characteristic vectors of each atmospheric state data in each data set to obtain multiple clusters of each atmospheric state data. ;

[0095] In this embodiment, use Density clustering algorithm divides clusters according to the local density of feature vectors. When dividing, it is necessary to set the neighborhood radius and the minimum number of samples.

[0096] (5-2) Clusters with a data volume greater than a preset threshold As the target cluster, a reference feature vector is constructed based on multiple feature vectors in the target cluster. and reference range , , reference range , represents the average value of the parameters in brackets, Indicates the reference range of the parameter in brackets. The reference range is the range that complies with the principle of three times the standard deviation based on the mean and standard deviation of the parameter in brackets;

[0097] After clustering, if the amount of data within a cluster exceeds a preset threshold, clustering has identified the atmospheric state patterns that trigger certain abnormal wind conditions. When the values ​​of the atmospheric state data's value, fluctuation, and change characteristics fall within the target cluster's reference ranges for value, fluctuation, and change characteristics, respectively, there's a certain probability that the corresponding abnormal wind condition will occur.

[0098] For the target cluster, this application constructs the corresponding reference feature vector by extracting the average value of each atmospheric state parameter . And extract the average value of each atmospheric state parameter and standard deviation , to construct reference ranges based on each atmospheric state parameter , reference range Satisfy the three times standard deviation principle, that is, the reference range middle .

[0099] (5-3) Calculate each eigenvector and the reference eigenvector of each abnormal wind condition The cosine similarity between , The mathematical expression is:

[0100] ,

[0101] (5-4) The cosine similarity As a feature vector The label of the abnormal wind condition possibility is used to obtain training data, wherein when there are multiple labels of the same abnormal wind condition possibility, the label is the average value of multiple cosine similarities;

[0102] In this application, when building an AI model, it is necessary to label the sample data. If manual labeling is used, there will be problems such as subjective labeling and low labeling efficiency. Therefore, this application characterizes the typical characteristics of abnormal wind conditions by constructing one or more reference feature vectors corresponding to multiple abnormal wind condition labels. Then calculate the feature vector The cosine similarity with the reference feature vector is used to mark the possibility of its label.

[0103] For example, the feature vector The cosine similarities of the three typical features corresponding to the label "gust" are 25%, 60%, and 65%, respectively, and their average value is 50%, which is the feature vector The probability of the "gust" label. At the same time, the feature vector The cosine similarities of the two typical features corresponding to the label "turbulence" are 75% and 15%, respectively, with an average of 40%. Finally, the probabilities of each label are normalized, resulting in the following labeling results: "gusts": 55.5%; "turbulence": 45.5%.

[0104] The intermediate components of the statistical model are used to label the sample data, which has the advantages of high labeling efficiency and strong reference value.

[0105] (5-5) Constructing a training data set based on the training data, and training an artificial neural network based on the training data set to obtain a deep learning recognition model; and constructing a statistical model based on multiple reference ranges of multiple types of abnormal wind conditions.

[0106] In this application, a gradient descent method was used to train the artificial neural network. However, due to the small amount of data in the early stages, the fitting effect was poor. Therefore, a statistical model was constructed as a supplement. The statistical model in this application was constructed based on the multiple reference ranges of the various abnormal wind conditions extracted above.

[0107] Figure 4 This is a flow chart of wind condition prediction in one embodiment of the present application, as shown in FIG. Figure 4 As shown, the process of predicting wind conditions in the future time period based on the above model includes:

[0108] S1, extracting the value characteristics of each atmospheric state data from the meteorological forecast data , fluctuation characteristics And the changing characteristics ; and based on the value characteristics , fluctuation characteristics And the changing characteristics Constructing the input vector ;

[0109] Among them, the value feature , fluctuation characteristics And the changing characteristics Please refer to the previous text for the extraction process, which will not be repeated here.

[0110] S2, the input vector Input into the deep learning recognition model to obtain the first probability of multiple abnormal wind conditions ; and the input vector Input into the statistical model to obtain the second probability of multiple abnormal wind conditions , wherein, when the input vector falls into the reference range of different wind conditions, a second probability is performed based on the number of data corresponding to the target cluster in the reference range;

[0111] The AI ​​model can automatically output the corresponding probabilities of three abnormal wind conditions, while the statistical model requires further processing.

[0112] For example, the input vector When the air falls within only one or more reference ranges of the label “gust”, the corresponding second probability is: “gust”: 100%.

[0113] If the input vector If a data point falls within one of the reference ranges for the labels "Gust" and "Turbulence," the probability is expressed based on the amount of data in the target cluster corresponding to the two reference ranges. For example, if the reference range data for the label "Gust" is A and the reference range data for the label "Turbulence" is B, the second probability is: "Gust": A / (A+B); "Turbulence": B / (A+B).

[0114] S3, perform weighted summation on the probability of each abnormal wind condition to obtain the final probability of each abnormal wind condition , where the final probability The mathematical expression is:

[0115] ,

[0116] Where, The amount of trained data for deep learning recognition models, The target amount of training data for deep learning recognition models, .

[0117] Indicates the training progress, with a maximum value of 1. Finally, the outputs of the two models are weighted based on the training progress of the learning recognition model. The greater the training progress, the greater the weight of the AI ​​model. Therefore, through weighting, a model that relies primarily on statistical models and supplemented by AI models can be gradually transitioned to relying solely on the AI ​​model for recognition. Good prediction accuracy is maintained in the early, middle, and late stages of training.

[0118] S4. When there is an abnormal wind condition with a final probability greater than a preset probability threshold, the abnormal wind condition with a final probability greater than the preset probability threshold is used as a prediction result; otherwise, it is determined that there is no abnormal wind condition risk.

[0119] Finally, if there is a final probability If it is greater than 70%, the risk of abnormal wind conditions is predicted.

[0120] a pre-start module 230 for starting the secondary storage unit and pre-regulating the air temperature in the secondary storage unit when the predicted wind condition is an abnormal wind condition, wherein the primary storage unit is used to respond to energy storage or power generation at the current time point, and the secondary storage unit is used to respond to wind turbine power fluctuations that may occur in a future time period;

[0121] When there is a risk of abnormal wind conditions, in order to quickly respond to changes in wind turbine power, it is necessary to pre-start the secondary storage unit and adjust it. This needs to be started according to the situation. The current working state of the main storage unit may be energy storage state, power generation state or idle state. At this time:

[0122] (1) when the main storage unit is in an energy storage state and the abnormal wind condition is an abnormal decrease in wind force or turbulence, starting the turbine generator of the auxiliary storage unit and preheating the compressed air in the auxiliary storage unit to a first temperature value;

[0123] When the main storage unit is in the energy storage state, its motor and compressor unit are in operation. At this time, the turbine generator of the main storage unit cannot be pre-started. If there is a risk of an abnormal drop in wind speed or turbulence, the wind turbine generator may experience a sudden drop in power generation, and the main storage unit cannot respond quickly to generate power. Therefore, the turbine generator of the secondary storage unit is started, and the compressed air in the secondary storage unit is preheated to a first temperature value; at this time, the turbine generator of the secondary storage unit operates at minimum power and generates electricity and is connected to the grid. This is to cope with the possible drop in generator power.

[0124] Furthermore, if the primary storage unit is in a charging state but the abnormal wind condition is gusty, there is no need to activate the secondary storage unit. The primary storage unit can simply respond by increasing the power of the motor and compressor unit.

[0125] (2) When the main storage unit is in a power generation state and the abnormal wind condition is a gust or turbulence, the motor and the compressor unit of the secondary storage unit are started, and the compressed air in the secondary storage unit is cooled to a second temperature value, wherein the first temperature value is higher than the second temperature value.

[0126] When the primary storage unit is generating electricity, its turbine generator is in operation, so the motor and compressor unit cannot be pre-started. Therefore, if there are gusts or turbulence, the wind turbine power may suddenly increase. If the primary storage unit cannot respond in time, the wind power will be wasted. Therefore, the motor and compressor unit of the secondary storage unit are started at this time and maintained at the minimum energy storage power to cope with the possible sudden increase in wind turbine power.

[0127] In addition, if the main storage unit is in the power generation state and the abnormal wind condition is an abnormal decrease in wind speed, there is no need to start the auxiliary storage unit, and it is only necessary to increase the power generation power of the turbine generator to respond.

[0128] Another situation where pre-start management of the main storage unit is necessary is when the main storage unit is idle due to energy saturation and abnormal wind conditions include an abnormal drop in wind speed or turbulence. The main storage unit can be pre-started and preheated, and maintained at minimum power to trigger the necessary "valley filling" mechanism.

[0129] The management module 240 is used to extract the change characteristics of the real-time power generation power, match the change characteristics with the abnormal wind conditions, and obtain a matching result; and dynamically adjust and manage the working status and working power of the main storage unit and the working status and working power of the secondary storage unit based on the real-time power generation power and the matching result.

[0130] Finally, after pre-startup, it is necessary to keep an eye on the changing characteristics of real-time power generation to determine whether the previous prediction is accurate. If it is accurate, the main storage unit and the auxiliary storage unit are further dynamically adjusted to maintain the power supply stability and response speed of the wind power system.

[0131] The characteristics of real-time power generation are reflected by its volatility and its changing trend. If its volatility and changing trend match the abnormal wind conditions predicted above, it indicates that further response adjustment is needed. The extraction and matching process of the characteristics of real-time power generation includes:

[0132] (1) Sampling the current time segment of the real-time generated power to obtain the generated power at multiple time points, wherein the current time segment is a time period of unit length before the current time point;

[0133] Among them, the power generation power at multiple time points is collected on a rolling basis to ensure the timeliness of the data.

[0134] (2) calculating the variance of the power generation at multiple time points, performing linear regression fitting on the power generation at multiple time points to obtain a slope; and matching the abnormal wind condition based on the variance of the power generation at multiple time points and the slope to obtain a matching result.

[0135] Variance is used to reflect volatility, while slope reflects the characteristics of change. The linear regression fitting process can be referred to in the previous article and will not be repeated here.

[0136] A large variance in generated power indicates fluctuations in generator output power, which in turn indicates fluctuations in wind direction or speed, confirming abnormal turbulent wind conditions. A steep rise or fall in power, however, indicates gusty winds or an unusual drop in wind speed.

[0137] (3) When the change characteristics match the abnormal wind conditions, the grid-connected power demand of wind power generation is obtained. ;

[0138] If the real-time power generation of the generator is consistent with the predicted result, the accuracy of the predicted result is verified. Therefore, the pre-started auxiliary storage unit is used to perform dynamic response.

[0139] Grid-connected power demand It can be a set constant value or a power dispatch demand value issued by the dispatch center.

[0140] (4) Calculate the grid-connected power requirement With the real-time power generation The difference ,exist When the storage power of the storage unit in the energy storage state in the main storage unit and the auxiliary storage unit is set to ;exist When the storage power of the storage unit in the power generation state among the main storage unit and the auxiliary storage unit is set to ,in, is the minimum holding power of another storage unit.

[0141] Finally, the pre-started secondary storage unit is used to achieve dynamic response. , which means the real-time power generation If the required power is exceeded, the storage unit in the energy storage state will be used to respond to the peak shaving mechanism, and the other storage unit will maintain the minimum power in the regulation and wait for call. , which means the real-time power generation If the required power cannot be met, the storage unit in the power generation state is used to respond to the valley filling mechanism, and the other storage unit maintains the minimum holding power in the regulation and waits for call.

[0142] The above process can quickly respond to various abnormal fluctuations, and as a redundant device for the wind power generation system, it can ensure that the peak shaving and valley filling mechanism continues to be effective after the main storage unit is damaged, which has many advantages.

[0143] In addition, the containers of the main storage unit and the secondary storage unit can be connected / isolated through electronically controlled valves to achieve pressure balance, capacity expansion, functional backup and other functions.

[0144] The present invention dynamically matches the wind turbine and compressed air energy storage energy management system for multiple wind conditions, and the present application realizes dynamic response by setting up a main and a sub-storage unit. In order to improve efficiency, the sub-storage unit is generally in a dormant state. The present application obtains meteorological forecast data in real time and makes rolling predictions on abnormal wind conditions that may occur in future time periods. Once abnormal wind conditions such as gusts, turbulence, and abnormal decrease in wind force are predicted, the sub-storage unit is immediately started and the compressed air in the sub-storage unit is preheated. When abnormal wind conditions occur, the main and sub-storage units with opposite working states are used to quickly respond to the power fluctuations of the wind turbine caused by the abnormal wind conditions, thereby avoiding fluctuations in grid-connected power.

[0145] This embodiment further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, any one of the methods in this embodiment is implemented, wherein the method is the execution logic of this system.

[0146] This embodiment also provides an electronic terminal, including: a processor and a memory;

[0147] The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory, so that the terminal executes any one of the methods in this embodiment.

[0148] Regarding the computer-readable storage medium in this embodiment, those skilled in the art will appreciate that all or part of the steps in the aforementioned method embodiments can be implemented using hardware associated with the computer program. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps in the aforementioned method embodiments. The aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0149] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication with each other. The memory is used to store computer programs, the communication interface is used for communication, and the processor and the transceiver are used to run computer programs so that the electronic terminal executes the various steps of the above method.

[0150] In this embodiment, the memory may include a random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk storage.

[0151] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, and discrete hardware components.

[0152] In the above embodiments, although the present invention has been described in conjunction with specific embodiments of the present invention, many replacements, modifications and variations of these embodiments will be apparent to those skilled in the art based on the foregoing description. The embodiments of the present invention are intended to cover all such replacements, modifications and variations that fall within the broad scope of the appended claims.

[0153] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the present invention. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by one of ordinary skill in the art without departing from the spirit and technical principles disclosed herein are intended to be covered by the claims of the present invention.

Claims

1. A wind turbine generator and compressed air energy storage energy management system that dynamically matches multiple wind conditions, characterized by: include: an acquisition module, configured to acquire predicted atmospheric state data for a future time period, the operating state and operating power of the main storage unit at a current time point, and the real-time generated power of the wind turbine, wherein the future time period includes multiple future time points of a target duration after the current time point, and the predicted atmospheric state data includes temperature, wind speed, air pressure, and humidity; A prediction module, configured to predict the wind conditions for a future time period based on a pre-built wind condition prediction model and the meteorological forecast data to obtain predicted wind conditions, wherein the wind condition prediction model characterizes the corresponding relationship between the atmospheric state data and the wind conditions; a pre-start module, configured to start the secondary storage unit and pre-regulate the air temperature in the secondary storage unit when the predicted wind condition is an abnormal wind condition, wherein the primary storage unit is configured to store compressed air and be used to respond to energy storage or power generation at the current time point, and the secondary storage unit is configured to store compressed air and be used to respond to wind turbine power fluctuations that may occur in future time periods, wherein the motor and compressor assembly of the primary storage unit and the motor and compressor assembly of the secondary storage unit are connected to the wind turbine for compressing air and pumping it into a container; the generators of the primary storage unit and the secondary storage unit are connected to the power grid via a coupling device, and generate electricity by releasing compressed air to drive the blades of the generator; A management module is used to extract the change characteristics of the real-time power generation power, match the change characteristics with the abnormal wind conditions, and obtain a matching result; and dynamically adjust and manage the working status and working power of the main storage unit and the working status and working power of the secondary storage unit based on the real-time power generation power and the matching result.

2. The wind turbine generator and compressed air energy storage energy management system that dynamically matches multiple wind conditions according to claim 1 is characterized in that: The method for constructing the wind condition prediction model includes: Acquire a plurality of historical sample data, wherein the historical sample data includes atmospheric state data, ground wind speed, and ground wind direction at a plurality of historical time points; Screen out a variety of abnormal wind condition target sample data based on the ground wind speed and ground wind direction at multiple historical time points , , For the Atmospheric state data at historical time points The value of Indicates the type of abnormal wind conditions; Classify the target sample data based on the type of abnormal wind conditions to obtain multiple data sets; Extract each target sample data in each data set The value characteristics of each atmospheric state data , fluctuation characteristics And the changing characteristics , and based on the value characteristics , the fluctuation characteristics And the changing characteristics The constructed feature vector , , is the sample data sequence number; Based on the feature vector A deep learning model and a statistical model are constructed respectively, and a wind condition prediction model is constructed based on the deep learning recognition model and the statistical model, wherein the output of the wind condition prediction model is obtained by weighting the output results of the deep learning recognition model and the output results of the statistical model, and the weight is determined based on the cumulative number of learning times of the deep learning recognition model.

3. The wind turbine generator and compressed air energy storage energy management system capable of dynamically matching multiple wind conditions according to claim 2 is characterized in that: Based on the feature vector Build deep learning models and statistical models separately, including: Density clustering is performed on the characteristic vectors of each atmospheric state data in each data set to obtain multiple clusters of each atmospheric state data. ; Clusters with a data volume greater than a preset threshold As the target cluster, a reference feature vector is constructed based on multiple feature vectors in the target cluster. and reference range , , reference range , The parameters in brackets are the average values. Indicates the reference range of the parameter in brackets; Calculate each eigenvector and the reference eigenvector of each abnormal wind condition The cosine similarity between , The mathematical expression is: ， The cosine similarity As a feature vector The label of the abnormal wind condition possibility is used to obtain training data, wherein when there are multiple labels of the same abnormal wind condition possibility, the label is the average value of multiple cosine similarities; A training data set is constructed based on the training data, and an artificial neural network is trained based on the training data set to obtain a deep learning recognition model; a statistical model is constructed based on multiple reference ranges of multiple types of abnormal wind conditions.

4. The wind turbine generator and compressed air energy storage energy management system capable of dynamically matching multiple wind conditions according to claim 2, characterized in that: Target sample data with abnormal wind conditions is screened out based on the ground wind speed and ground wind direction at multiple historical time points, including: Based on the pre-built sliding window, it slides along the time axis of the historical sample data, and calculates the wind speed change within the sliding window at each slide. and wind direction variance , where wind speed change , is the maximum wind speed, is the minimum wind speed; The wind speed change Greater than the preset change threshold, or wind direction variance The data interval corresponding to the sliding window that is greater than the preset variance threshold is regarded as the abnormal interval, and the historical sample data containing the abnormal interval is regarded as the target sample data.

5. The wind turbine generator and compressed air energy storage energy management system capable of dynamically matching multiple wind conditions according to claim 3 is characterized in that: Predicting the wind conditions for a future time period based on the pre-built wind condition prediction model and the meteorological forecast data to obtain predicted wind conditions includes: Extract the value characteristics of each atmospheric state data from the meteorological forecast data , fluctuation characteristics And the changing characteristics ; and based on the value characteristics , fluctuation characteristics And the changing characteristics Constructing the input vector ; The input vector Input into the deep learning recognition model to obtain the first probability of multiple abnormal wind conditions ; and the input vector Input into the statistical model to obtain the second probability of multiple abnormal wind conditions , wherein, when the input vector falls into the reference range of different wind conditions, a second probability is performed based on the number of data corresponding to the target cluster in the reference range; The probability of each abnormal wind condition is weighted and summed to obtain the final probability of each abnormal wind condition. , where the final probability The mathematical expression is: ， Where, The amount of trained data for deep learning recognition models, The target training data volume of the deep learning recognition model, ; When there is an abnormal wind condition with a final probability greater than the preset probability threshold, the abnormal wind condition with a final probability greater than the preset probability threshold is used as the prediction result; otherwise, it is determined that there is no abnormal wind condition risk.

6. The wind turbine generator and compressed air energy storage energy management system capable of dynamically matching multiple wind conditions according to claim 2 or 5, characterized in that: The value feature , the fluctuation characteristics And the changing characteristics The extraction process includes: Calculate the average value of atmospheric state data and obtain the value characteristics ; Calculate the standard deviation of atmospheric state data to obtain fluctuation characteristics ; and perform linear regression fitting on the atmospheric state data to obtain the slope of the atmospheric state data, and use the slope as a change feature .

7. The wind turbine generator and compressed air energy storage energy management system capable of dynamically matching multiple wind conditions according to claim 1, characterized in that: Starting the secondary storage unit and pre-regulating the air temperature in the secondary storage unit includes: Determining an operating state of the main storage unit, wherein the operating state includes an energy storage state and a power generation state; When the main storage unit is in an energy storage state and the abnormal wind condition is an abnormal drop in wind force or turbulence, starting the turbine generator of the auxiliary storage unit and preheating the compressed air in the auxiliary storage unit to a first temperature value; When the main storage unit is in a power generation state and the abnormal wind condition is a gust or turbulence, the motor and compressor unit of the secondary storage unit are started, and the compressed air in the secondary storage unit is cooled to a second temperature value, wherein the first temperature value is higher than the second temperature value.

8. The wind turbine generator and compressed air energy storage energy management system capable of dynamically matching multiple wind conditions according to claim 1, characterized in that: Extracting a change feature of the real-time generated power and matching the change feature with the abnormal wind condition to obtain a matching result, including: Sampling the current time segment of the real-time generated power to obtain the generated power at multiple time points, wherein the current time segment is a time period of unit length before the current time point; Calculating the variance of the power generation at multiple time points, and performing linear regression fitting on the power generation at multiple time points to obtain a slope; and matching the abnormal wind condition based on the variance of the power generation at the multiple time points and the slope to obtain a matching result.

9. The wind turbine generator and compressed air energy storage energy management system capable of dynamically matching multiple wind conditions according to claim 1, characterized in that: Dynamically adjusting and managing the working state and working power of the primary storage unit and the working state and working power of the secondary storage unit based on the real-time generated power and the matching result, including: When the change characteristics match the abnormal wind conditions, the grid-connected power requirement of wind power generation is obtained. ; Calculate the grid-connected power requirement With the real-time power generation The difference ,exist When the storage power of the storage unit in the energy storage state in the main storage unit and the auxiliary storage unit is set to ; exist When the storage power of the storage unit in the power generation state among the main storage unit and the auxiliary storage unit is set to ,in, is the minimum holding power of another storage unit.

10. The wind turbine generator and compressed air energy storage energy management system capable of dynamically matching multiple wind conditions according to claim 1, characterized in that: The main storage unit and the secondary storage unit are communicated with each other through an electrically controlled valve.

Images