Wind power prediction method and system under plateau meteorological condition

Abstract

The invention relates to the technical field of wind power prediction, in particular to a wind power prediction method and system under the plateau meteorological condition, and the method comprises the following steps: obtaining the air density and effective wind speed time series data of a target plateau wind power plant, and extracting the multi-scale fluctuation characteristics of the time series data through super empirical mode decomposition; by taking the multi-scale fluctuation characteristics as nodes, constructing a dynamic graph structure of a static correlation edge fusing a wind energy conversion physical law and a dynamic correlation edge based on time-varying correlation; constructing a model based on the graph attention network containing the edge convolution module, and combining physical consistency constraint training to obtain a trained power prediction model; and inputting the dynamic graph structure at the current moment into the model to obtain a preliminary predicted value, and correcting according to the running state of the fan to obtain a final result. The method is adaptive to complex characteristics of plateau meteorology, prediction errors are effectively reduced, accuracy and reliability of wind power prediction are improved, and powerful support is provided for power grid dispatching and wind power consumption.

Description

Technical Field

[0001] This invention relates to the field of wind power prediction technology, and more specifically, to a method and system for predicting wind power under plateau meteorological conditions. Background Technology

[0002] As the global energy structure transitions towards clean and low-carbon energy, wind energy, as a core component of renewable energy, is experiencing continuous expansion in its development and utilization. my country's plateau regions possess abundant wind energy resources; however, the unique geographical and meteorological conditions of these regions present numerous challenges to wind power forecasting. Existing forecasting methods lack the accuracy and reliability to meet practical application requirements, severely restricting the efficient integration of plateau wind power and the safe and stable operation of the power grid.

[0003] The core of wind power forecasting is to accurately predict wind power output over a future period based on meteorological parameters, wind turbine characteristics, and other information, providing a basis for decision-making in grid dispatching and electricity market transactions. For plain areas, existing forecasting methods have formed a relatively mature technical system, mainly using statistical models or conventional deep learning models combined with conventional meteorological data such as wind speed and temperature to achieve forecasting. However, if these forecasting schemes are directly applied to plateau environments, their prediction errors are often higher than the industry average in plain areas.

[0004] Therefore, how to accurately predict wind power under plateau meteorological conditions has become a technical problem that urgently needs to be solved in the development of the current wind power industry. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for predicting wind power under plateau meteorological conditions, so as to solve the technical problem of how to accurately predict wind power under plateau meteorological conditions.

[0006] This invention is achieved through the following technical solution: a method for predicting wind power under plateau meteorological conditions, comprising the following steps:

[0007] Acquire time-series data of air density, effective wind speed, and other power-related parameters of the target plateau wind farm, wherein the effective wind speed is the equivalent wind speed of the wind force on the swept surface of the wind turbine;

[0008] The air density time series data, effective wind speed time series data, and other power-related parameter time series data are respectively decomposed into time series to obtain component sequences containing fluctuation characteristics at different time scales for each time series data.

[0009] Using the component sequences of each time series data containing fluctuation characteristics at different time scales as node features, a dynamic graph structure is constructed. The dynamic graph structure includes static physical association edges established between air density-related nodes, effective wind speed-related nodes, and power reference values ​​based on the physical laws of wind energy conversion, as well as dynamic data association edges established based on the time-varying correlation between each component sequence.

[0010] A power prediction model is constructed and trained using the dynamic graph structure and its corresponding actual power data to obtain the trained power prediction model.

[0011] The dynamic graph structure, constructed based on the current air density time series data, effective wind speed time series data, and other power-related parameters, is used as the input to the trained power prediction model to obtain preliminary power prediction values.

[0012] The initial power prediction value is corrected based on the current operating status of the wind turbine to obtain the final power prediction value.

[0013] According to a preferred embodiment, the air density time-series data is calculated using the temperature, specific humidity, and atmospheric pressure of the target plateau wind farm through the ideal gas law, as shown in the following expression:

[0014]

[0015] In the above formula, Indicates time air density, This represents the gas constant for dry air. Indicates time temperature, Indicates time The specific moisture content, Indicates time Atmospheric pressure;

[0016] The effective wind speed time series data is obtained by correcting the wind speed at hub height using the vertical wind shear index, as shown in the following expression:

[0017]

[0018] In the above formula, Indicates time The effective wind speed, Indicates time Wheel hub height wind speed, Indicates the span of the swept surface height of the wind turbine blades. Indicates the height of the wind turbine hub. Indicates wind direction The corresponding vertical wind shear index.

[0019] According to a preferred embodiment, the time series decomposition adopts the hyperempirical mode decomposition method, which decomposes the air density time series data, effective wind speed time series data and other power-related parameter time series data into multiple intrinsic mode function components and a residual trend term.

[0020] According to a preferred embodiment, the eigenvalues ​​of the static physical association edges are expressed by the expression Define, where, This represents the learnable scaling factor. Indicates time The power reference value.

[0021] According to a preferred embodiment, the establishment of the dynamic data association edge specifically includes:

[0022] The mutual information of the component sequences between nodes within the sliding time window is determined by the following expression:

[0023]

[0024] In the above formula, Represents a node With nodes At any moment mutual information, Represents nodes estimated based on time window data. eigenvalue marginal probability distribution Represents nodes estimated based on time window data. eigenvalue marginal probability distribution Represents nodes estimated based on data from the same time window. With nodes The joint probability distribution of ;

[0025] The time-varying correlation between nodes is determined based on the mutual information, and dynamic data association edges are established based on the time-varying correlation.

[0026] According to a preferred embodiment, the power prediction model is based on a graph attention network that includes an edge convolution module, which is used to simultaneously learn the feature information of static physical association edges and dynamic data association edges and perform information aggregation.

[0027] According to a preferred embodiment, the information aggregation process further includes calculating the attention coefficient of nodes using edge features, as expressed below:

[0028]

[0029] In the above formula, Represents a node The set of neighboring nodes, This represents the activation function. This represents the learnable weight matrix used for node features. Indicates time node eigenvectors, Indicates time node eigenvectors, This represents the learnable weight matrix used for edge features. Indicates time Connecting nodes and The eigenvectors of the edges.

[0030] According to a preferred embodiment, the power prediction model is trained using a loss function that includes physical consistency constraints, expressed as follows:

[0031]

[0032] In the above formula, This represents the total number of training samples. Indicates the first The true power value of each sample Indicates the first The final power prediction value for each sample. Indicates the first The feature values ​​of the static physical association edges corresponding to each sample The hyperparameter represents the weights that balance the two loss terms. It is a very small positive number.

[0033] According to a preferred embodiment, the correction to the preliminary power prediction value is expressed as follows:

[0034]

[0035] In the above formula, Indicates time The final power prediction value, Indicates time Preliminary power forecast values, Indicates the number of abnormal operating conditions. Indicates the first Various abnormal operating conditions Indicates abnormal operating conditions The corresponding power reduction factor, Indicates an indicator function, when It is 1 if it is true, otherwise it is 0.

[0036] This invention also provides a wind power prediction system under plateau meteorological conditions, which is applied to the wind power prediction method under plateau meteorological conditions as described above. The system includes:

[0037] The data acquisition module is used to acquire time-series data of air density, effective wind speed and other power-related parameters of the target plateau wind farm, wherein the effective wind speed is the equivalent wind speed of the wind force on the swept surface of the wind turbine.

[0038] The time series decomposition module is used to perform time series decomposition on the air density time series data, effective wind speed time series data and other power-related parameter time series data respectively, to obtain component sequences of each time series data containing fluctuation characteristics at different time scales;

[0039] The graph construction module is used to construct a dynamic graph structure by using the component sequences of each time series data containing different time scale fluctuation characteristics as node features. The dynamic graph structure includes static physical association edges established between air density-related nodes, effective wind speed-related nodes and power reference values ​​based on the physical laws of wind energy conversion, as well as dynamic data association edges established based on the time-varying correlation between each component sequence.

[0040] The model building and training module is used to build a power prediction model and train it using the dynamic graph structure and its corresponding actual power data to obtain the power prediction model after training.

[0041] The prediction module is used to take the dynamic graph structure constructed based on the current air density time series data, effective wind speed time series data and other power-related parameters as input to the trained power prediction model to obtain the preliminary power prediction value.

[0042] The correction module is used to correct the preliminary power prediction value based on the current operating status of the wind turbine to obtain the final power prediction value.

[0043] The technical solution of the wind power prediction method and system under plateau meteorological conditions provided by the present invention has at least the following advantages and beneficial effects: The present invention constructs a dynamic graph structure that integrates the physical laws of wind energy conversion and the time-varying correlation of component sequences to adapt to the complex characteristics of plateau meteorology, and can provide accurate data correlation support for power prediction; by combining the real-time operating status of wind turbines to correct the preliminary prediction value, the prediction error in plateau environment can be effectively reduced, and the accuracy and reliability of wind power prediction can be improved. Attached Figure Description

[0044] Figure 1 This is a flowchart illustrating the wind power prediction method under plateau meteorological conditions provided in Embodiment 1 of the present invention.

[0045] Figure 2 The structural block diagram of the wind power prediction system under plateau meteorological conditions provided in Embodiment 2 of the present invention. Detailed Implementation

[0046] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0047] Example 1

[0048] This invention provides a method for predicting wind power under plateau meteorological conditions. Figure 1 This is a flowchart illustrating the wind power prediction method under these high-altitude meteorological conditions. (See attached diagram) Figure 1 As shown, the wind power prediction method under plateau meteorological conditions includes the following steps:

[0049] Step S1: Data Acquisition;

[0050] Time-series data of air density, effective wind speed, and other power-related parameters of the target plateau wind farm are obtained, wherein the effective wind speed is the equivalent wind speed of the wind force on the swept surface of the wind turbine.

[0051] In some implementations, plateau regions are characterized by high altitude, low atmospheric pressure, drastic temperature variations, and significant wind shear effects. Conventional methods for calculating air density and wind speed in plains areas are not directly applicable. Therefore, the air density time-series data in this embodiment is obtained by calculating the temperature, specific humidity, and atmospheric pressure of the target plateau wind farm using the ideal gas law, to accurately reflect the characteristics of thin air in plateau regions. The expression is as follows:

[0052]

[0053] In the above formula, Indicates time air density, This represents the gas constant for dry air. Indicates time temperature, Indicates time The specific moisture content, Indicates time Atmospheric pressure.

[0054] The effective wind speed time series data is obtained by correcting the wind speed at hub height using the vertical wind shear index, in order to eliminate the influence of uneven vertical wind speed distribution caused by the complex terrain of the plateau, and to better reflect the actual wind conditions of the wind turbine. The expression is as follows:

[0055]

[0056] In the above formula, Indicates time The effective wind speed, Indicates time Wheel hub height wind speed, Indicates the span of the swept surface height of the wind turbine blades. Indicates the height of the wind turbine hub. Indicates wind direction The corresponding vertical wind shear index.

[0057] It should be noted that this embodiment, through parameter calculation, can effectively solve the problem of accurately obtaining air density and wind speed in plateau areas, providing high-quality basic data for subsequent forecasting. Compared with the method of directly collecting air density and wind speed, it can effectively reduce parameter errors.

[0058] Step S2, timing decomposition;

[0059] The air density time series data, effective wind speed time series data, and other power-related parameter time series data are respectively subjected to time series decomposition to obtain component sequences containing fluctuation characteristics at different time scales for each time series data.

[0060] Given that plateau meteorological data exhibits multi-scale fluctuations, conventional decomposition methods struggle to extract effective features. Therefore, in some implementations, the time-series decomposition employs an empirical mode decomposition method, which decomposes air density time-series data, effective wind speed time-series data, and other power-related parameter time-series data into multiple intrinsic mode function components and a residual trend term. The intrinsic mode function components reflect short-term and medium-term fluctuations, while the residual trend term reflects long-term variation patterns.

[0061] By performing hyper-empirical mode decomposition on time-series data of air density, effective wind speed, and other power-related parameters, this embodiment can achieve accurate decomposition of multi-scale features of plateau meteorological data. This allows subsequent models to specifically learn the impact of different fluctuation patterns on power, avoid feature loss caused by single-scale analysis, and improve feature utilization.

[0062] Step S3: Graph construction;

[0063] Using the component sequences of each time series data containing fluctuation characteristics at different time scales as node features, a dynamic graph structure is constructed. The dynamic graph structure includes static physical association edges established between air density-related nodes, effective wind speed-related nodes, and power reference values ​​based on the physical laws of wind energy conversion, as well as dynamic data association edges established based on the time-varying correlation between each component sequence.

[0064] In some implementations, the eigenvalues ​​of the static physical association edges are expressed by the expression Define, where, This represents the learnable scaling factor. Indicates time The power reference value; this embodiment constructs static physical association edges to ensure that the association relationship conforms to the physical nature.

[0065] The establishment of the dynamic data association edge specifically includes:

[0066] The mutual information of the component sequences between nodes within the sliding time window is determined by the following expression:

[0067]

[0068] In the above formula, Represents a node With nodes At any moment mutual information, Represents nodes estimated based on time window data. eigenvalue marginal probability distribution Represents nodes estimated based on time window data. eigenvalue marginal probability distribution Represents nodes estimated based on data from the same time window. With nodes The joint probability distribution of ;

[0069] Furthermore, the time-varying correlation between nodes is determined based on the mutual information, and dynamic data association edges are established according to the time-varying correlation to capture the correlation between plateau meteorological parameters over time in real time.

[0070] Step S4: Model building and training;

[0071] A power prediction model is constructed and trained using the dynamic graph structure and its corresponding actual power data to obtain the trained power prediction model.

[0072] To address the limitations of conventional graph models in simultaneously learning the features of static physical edges and dynamic data edges, and their insufficient utilization of edge information, in some implementations, the power prediction model is based on a graph attention network containing edge convolution modules. These edge convolution modules are used to simultaneously learn the feature information of static physical associated edges and dynamic data associated edges and perform information aggregation.

[0073] The information aggregation process also includes calculating the attention coefficient of nodes using edge features, allowing the model to focus on related edges that have a greater impact on power. For example, edges related to wind speed have a higher attention weight during periods of strong winds. The expression for the attention coefficient is as follows:

[0074]

[0075] In the above formula, Represents a node The set of neighboring nodes, This represents the activation function. This represents the learnable weight matrix used for node features. Indicates time node eigenvectors, Indicates time node eigenvectors, This represents the learnable weight matrix used for edge features. Indicates time Connecting nodes and The eigenvectors of the edges.

[0076] To avoid the model's prediction results violating the physical laws of wind energy conversion, such as a decrease in air density but an increase in power, this embodiment uses a loss function that includes physical consistency constraints, as expressed below:

[0077]

[0078] In the above formula, This represents the total number of training samples. Indicates the first The true power value of each sample Indicates the first The final power prediction value for each sample. Indicates the first The feature values ​​of the static physical association edges corresponding to each sample The hyperparameter represents the weights that balance the two loss terms. It is a very small positive number.

[0079] It should be noted that this embodiment, by simultaneously learning physical correlation and dynamic data correlation features and ensuring the rationality of prediction through physical consistency constraints, can effectively improve the model convergence speed and the credibility of prediction results.

[0080] Step S5: Power prediction application;

[0081] A dynamic graph structure, constructed based on current time-series air density data, effective wind speed data, and other power-related parameters, is used as input to the trained power prediction model to obtain preliminary power predictions. It should be noted that by utilizing the correlation features of the dynamic graph structure, the model can quickly capture the mapping relationship between current plateau meteorological conditions and power, thereby reducing the error of the preliminary prediction.

[0082] Step S6: Power prediction correction;

[0083] In response to the fact that high-altitude wind turbines are susceptible to abnormal operating conditions due to environmental factors such as low temperatures and dust storms, such as blade icing and gearbox failures, predictions based solely on meteorological parameters cannot account for the impact of the equipment's own condition. Therefore, this embodiment corrects the preliminary power prediction value based on the wind turbine's current operating status to obtain the final power prediction value.

[0084] In some implementations, the correction to the initial power prediction value is expressed as:

[0085]

[0086] In the above formula, Indicates time The final power prediction value, Indicates time Preliminary power forecast values, Indicates the number of abnormal operating conditions. Indicates the first Various abnormal operating conditions Indicates abnormal operating conditions The corresponding power reduction factor, Indicates an indicator function, when It is 1 if it is true, otherwise it is 0.

[0087] It should be noted that this embodiment can further eliminate prediction deviations caused by abnormal operating conditions by correcting the equipment's operating status, and the prediction reliability is improved more significantly under extreme weather conditions.

[0088] Example 2

[0089] This embodiment, based on the technical solution provided in Embodiment 1, provides a wind power prediction system under plateau meteorological conditions. This system applies the wind power prediction method under plateau meteorological conditions described in Embodiment 1. (See also...) Figure 2 As shown, the system includes:

[0090] The data acquisition module is used to acquire time-series data of air density, effective wind speed and other power-related parameters of the target plateau wind farm, wherein the effective wind speed is the equivalent wind speed of the wind force on the swept surface of the wind turbine.

[0091] The time series decomposition module is used to perform time series decomposition on the air density time series data, effective wind speed time series data and other power-related parameter time series data respectively, to obtain component sequences of each time series data containing fluctuation characteristics at different time scales;

[0092] The graph construction module is used to construct a dynamic graph structure by using the component sequences of each time series data containing different time scale fluctuation characteristics as node features. The dynamic graph structure includes static physical association edges established between air density-related nodes, effective wind speed-related nodes and power reference values ​​based on the physical laws of wind energy conversion, as well as dynamic data association edges established based on the time-varying correlation between each component sequence.

[0093] The model building and training module is used to build a power prediction model and train it using the dynamic graph structure and its corresponding actual power data to obtain the power prediction model after training.

[0094] The prediction module is used to take the dynamic graph structure constructed based on the current air density time series data, effective wind speed time series data and other power-related parameters as input to the trained power prediction model to obtain the preliminary power prediction value.

[0095] The correction module is used to correct the preliminary power prediction value based on the current operating status of the wind turbine to obtain the final power prediction value.

[0096] It should be noted that the functions of each module of the wind power prediction system under plateau meteorological conditions in this embodiment are the same as those of the embodiment of the wind power prediction method under plateau meteorological conditions, and the technical effects are the same, so they will not be repeated here.

[0097] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting wind power under plateau meteorological conditions, characterized in that, Includes the following steps: Acquire time-series data of air density, effective wind speed, and other power-related parameters of the target plateau wind farm, wherein the effective wind speed is the equivalent wind speed of the wind force on the swept surface of the wind turbine; The air density time series data, effective wind speed time series data, and other power-related parameter time series data are respectively decomposed into time series to obtain component sequences containing fluctuation characteristics at different time scales for each time series data. Using the component sequences of each time series data containing fluctuation characteristics at different time scales as node features, a dynamic graph structure is constructed. The dynamic graph structure includes static physical association edges established between air density-related nodes, effective wind speed-related nodes, and power reference values ​​based on the physical laws of wind energy conversion, as well as dynamic data association edges established based on the time-varying correlation between each component sequence. A power prediction model is constructed and trained using the dynamic graph structure and its corresponding actual power data to obtain the trained power prediction model. The dynamic graph structure, constructed based on the current air density time series data, effective wind speed time series data, and other power-related parameters, is used as the input to the trained power prediction model to obtain preliminary power prediction values. The initial power prediction value is corrected based on the current operating status of the wind turbine to obtain the final power prediction value.

2. The wind power prediction method under plateau meteorological conditions as described in claim 1, characterized in that, The air density time-series data was obtained by calculating the temperature, specific humidity, and atmospheric pressure of the target plateau wind farm using the ideal gas law, as shown in the following expression: ； In the above formula, Indicates time air density, This represents the gas constant for dry air. Indicates time temperature, Indicates time The specific moisture content, Indicates time Atmospheric pressure; The effective wind speed time series data is obtained by correcting the wind speed at hub height using the vertical wind shear index, as shown in the following expression: ； In the above formula, Indicates time The effective wind speed, Indicates time Wheel hub height wind speed, Indicates the span of the swept surface height of the wind turbine blades. Indicates the height of the wind turbine hub. Indicates wind direction The corresponding vertical wind shear index.

3. The wind power prediction method under plateau meteorological conditions as described in claim 1, characterized in that, The time series decomposition adopts the hyperempirical mode decomposition method, which decomposes the air density time series data, effective wind speed time series data and other power-related parameter time series data into multiple intrinsic mode function components and a residual trend term.

4. The wind power prediction method under plateau meteorological conditions as described in claim 2, characterized in that, The eigenvalues ​​of the static physical association edges are expressed by the expression Define, where, This represents the learnable scaling factor. Indicates time The power reference value.

5. The wind power prediction method under plateau meteorological conditions as described in claim 1, characterized in that, The establishment of the dynamic data association edge specifically includes: The mutual information of the component sequences between nodes within the sliding time window is determined by the following expression: ； In the above formula, Represents a node With nodes At any moment mutual information, Represents nodes estimated based on time window data. eigenvalue marginal probability distribution Represents nodes estimated based on time window data. eigenvalue marginal probability distribution Represents nodes estimated based on data from the same time window. With nodes The joint probability distribution of ; The time-varying correlation between nodes is determined based on the mutual information, and dynamic data association edges are established based on the time-varying correlation.

6. The wind power prediction method under plateau meteorological conditions as described in claim 1, characterized in that, The power prediction model is based on a graph attention network that includes an edge convolution module, which is used to simultaneously learn the feature information of static physical association edges and dynamic data association edges and perform information aggregation.

7. The wind power prediction method under plateau meteorological conditions as described in claim 6, characterized in that, The information aggregation process also includes calculating the attention coefficient of nodes using edge features, as shown in the following expression: ； In the above formula, Represents a node The set of neighboring nodes, This represents the activation function. This represents the learnable weight matrix used for node features. Indicates time node eigenvectors, Indicates time node eigenvectors, This represents the learnable weight matrix used for edge features. Indicates time Connecting nodes and The eigenvectors of the edges.

8. The wind power prediction method under plateau meteorological conditions as described in claim 4, characterized in that, The power prediction model is trained using a loss function that includes physical consistency constraints, expressed as follows: ； In the above formula, This represents the total number of training samples. Indicates the first The true power value of each sample Indicates the first The final power prediction value for each sample. Indicates the first The feature values ​​of the static physical association edges corresponding to each sample The hyperparameter represents the weights that balance the two loss terms. It should be a very small positive number to avoid the denominator being zero.

9. The wind power prediction method under plateau meteorological conditions as described in claim 1, characterized in that, The correction to the initial power prediction value is expressed as follows: ； In the above formula, Indicates time The final power prediction value, Indicates time Preliminary power forecast values, Indicates the number of abnormal operating conditions. Indicates the first Various abnormal operating conditions Indicates abnormal operating conditions The corresponding power reduction factor, Indicates an indicator function, when It is 1 if it is true, otherwise it is 0.

10. A wind power prediction system under plateau meteorological conditions, characterized in that, The system, applied to the wind power prediction method under plateau meteorological conditions as described in any one of claims 1 to 9, comprises: The data acquisition module is used to acquire time-series data of air density, effective wind speed and other power-related parameters of the target plateau wind farm, wherein the effective wind speed is the equivalent wind speed of the wind force on the swept surface of the wind turbine. The time series decomposition module is used to perform time series decomposition on the air density time series data, effective wind speed time series data and other power-related parameter time series data respectively, to obtain component sequences of each time series data containing fluctuation characteristics at different time scales; The graph construction module is used to construct a dynamic graph structure by using the component sequences of each time series data containing different time scale fluctuation characteristics as node features. The dynamic graph structure includes static physical association edges established between air density-related nodes, effective wind speed-related nodes and power reference values ​​based on the physical laws of wind energy conversion, as well as dynamic data association edges established based on the time-varying correlation between each component sequence. The model building and training module is used to build a power prediction model and train it using the dynamic graph structure and its corresponding actual power data to obtain the power prediction model after training. The prediction module is used to take the dynamic graph structure constructed based on the current air density time series data, effective wind speed time series data and other power-related parameters as input to the trained power prediction model to obtain the preliminary power prediction value. The correction module is used to correct the preliminary power prediction value based on the current operating status of the wind turbine to obtain the final power prediction value.

Images