Method for acquiring impedance data of wind turbine in wind farm and method for modeling impedance of wind farm

Abstract

The application discloses a wind turbine impedance data acquisition method and a wind farm impedance modeling method, comprising: establishing a first database; training the first database to obtain a first neural network model; establishing a second database; training the second database to obtain a second neural network model; inputting climate data of the wind farm into the first neural network model, and taking the output of the first neural network model as the input of the second neural network model, wherein the output of the second neural network model comprises wind turbine impedance data in the wind farm. The application trains neural networks based on historical data and measured data, obtains accurate wind turbine impedance data of different types in the wind farm under typical climate data of four seasons, and establishes an accurate wind farm impedance model through the typical climate data of four seasons, the wind turbine impedance data, line impedance data and transformer impedance data.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of wind power generation, in particular to a wind turbine impedance data acquisition method and a wind farm impedance modeling method. BACKGROUND

[0002] To solve the power supply reliability and power system safety and stability operation problems caused by new energy grid-connected broadband oscillation, impedance-based modeling and stability analysis has been verified by engineering, which is considered as an effective method to identify and solve the actual oscillation problems in new energy stations, high-voltage direct current transmission systems, etc. This method has the advantages of clear physical meaning, simple application, easy to explain mechanism, and can quantify stability indicators, etc., and gradually becomes the mainstream method of broadband oscillation and stability analysis of new energy grid-connected system. Due to the existence of time delay effect, wake effect, and also affected by topography, when the wind speed fluctuates or the climate data changes, wind turbines with different operating states will have different working conditions. It is difficult to accurately depict the impedance characteristics of the wind farm under complex and variable operating conditions by using single machine equivalent or multiplication method for modeling. Therefore, it is of great significance to establish a high-precision wind farm impedance model under typical operating conditions for studying the grid-connected stability of wind farms.

[0003] The prior art discloses a data-model hybrid driven wind farm aggregation equivalent modeling method. The principle of this method is to obtain the wind speed of all units from the wind speed of the wind farm, and then to the single machine power from the unit wind speed, which can represent the detailed dynamic behavior of the wind farm with at most two equivalent machines. However, in the wind farm, there are multiple, even dozens or hundreds of wind turbines. Different numbered wind turbines are located in different positions of the wind farm, and there is a complex relationship between the working condition indicators of each numbered wind turbine under different climate data. The same climate data will have different effects on wind turbines at different positions. Because there is an uncertain dynamic process from the wind speed and other working condition conditions of the wind farm to the power of all units, this method cannot verify the power accuracy of each wind turbine under different weather conditions of the station, and this method does not consider the impedance of the wind turbine at different frequencies. In addition, in the process of modeling the wind farm, this method does not fully consider the transformer and line impedance, nor does it consider the influence of climate change on these impedances. Due to the above-mentioned deficiencies, the existing method cannot accurately obtain the impedance data of the wind turbines in the wind farm, and cannot accurately establish the impedance model of the wind farm. SUMMARY

[0004] The present application aims to solve the above-mentioned problems of the prior art, and provides a wind turbine impedance data acquisition method and a wind farm impedance modeling method, which can at least solve one of the above-mentioned problems of the prior art, and effectively realize the accurate acquisition of wind turbine impedance data in the wind farm and the accurate establishment of the impedance model of the wind farm.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present application is:

[0006] A method for obtaining impedance data of wind turbines in a wind farm, characterized in that it comprises the following steps:

[0007] Step S1, establishing a first database, the first database: historical climate data of the wind farm at each time, and wind turbine operating condition data of each numbered wind turbine at the corresponding time;

[0008] Step S2, according to the first database, constructing and training a first neural network model with climate data parameters as input and wind turbine operating condition data of each numbered wind turbine as output;

[0009] Step S3, establishing a second database, the second database: including climate data of the wind farm at each time, wind turbine operating condition data of the target wind turbine, and wind turbine sequence impedance data of each target wind turbine at the corresponding time;

[0010] Step S4, according to the second database, constructing and training a second neural network model with wind turbine operating condition parameters as input and wind turbine sequence impedance data as output;

[0011] Step S5, inputting the current climate data of the wind farm into the first neural network model, and inputting the output of the first neural network model into the second neural network model, and the output of the second neural network model is the impedance data of the wind turbine in the wind farm.

[0012] As a preferred embodiment, the climate data includes any one or more of temperature, humidity, wind speed, and wind direction; the wind turbine operating condition data includes any one or more of active power, reactive power, and voltage;

[0013] As a preferred embodiment, in step S1, the first database further includes position parameters of each numbered wind turbine; in step S2, the input of the first neural network model further includes the position parameters of each numbered wind turbine.

[0014] As a preferred embodiment, the position parameters include body position parameters and relative position parameters;

[0015] The body position parameters include any one or more of plane coordinates, plane coordinates, longitude, latitude, elevation, and cabin orientation angle;

[0016] The relative position parameters include wind turbine volume rate and adjacent wind turbine distance.

[0017] As a preferred embodiment, the first neural network model adopts a four-layer BP neural network structure; the transfer function between the input layer and the hidden layer of the four-layer BP neural network structure is a tanh function, the transfer function between the hidden layers is a sigmoid function, and the transfer function between the hidden layer and the output layer is a softplus function.

[0018] As a preferred embodiment, step S3 comprises:

[0019] S31: According to the differences of each wind turbine in the wind farm, each numbered wind turbine in the wind farm is classified, and target wind turbines in each class are determined.

[0020] S32: Under each climate data, the target wind turbine is subjected to semi-physical simulation test, impedance scanning is performed through the semi-physical simulation platform, and climate data, wind turbine operating condition data of the target wind turbine, and wind turbine serial impedance data of each target wind turbine at the corresponding time are obtained.

[0021] As a preferred embodiment, the output of the second neural network is the serial impedance of the wind turbine at different frequencies.

[0022] The second neural network model adopts a six-layer BP neural network structure, the transfer function between the input layer and the hidden layer of the six-layer BP neural network structure is a ReLU function, the transfer function between the hidden layers is a ReLU function, and the transfer function between the hidden layer and the output layer is a softplus function.

[0023] Based on the same inventive concept, the application also provides a wind farm impedance modeling method, which is characterized by comprising:

[0024] The wind turbine impedance data is obtained by the wind turbine impedance data acquisition method in the wind farm.

[0025] Based on the wind turbine impedance data, the line impedance data, and the transformer impedance data, a wind farm impedance model is generated.

[0026] As a preferred mode, the line impedance data acquisition method comprises:

[0027] Under the climate data of four seasons, the line impedance scanning is performed on each type of line in the wind farm to obtain line serial impedance data, and then the impedance data of each line in the field station is obtained based on the type and length of each line.

[0028] As a preferred mode, the transformer impedance data acquisition method comprises:

[0029] Under the climate data of four seasons, the transformer impedance scanning is performed on each type of transformer in the wind farm to obtain transformer serial impedance data, and then the impedance data of each transformer is obtained based on the type of each transformer.

[0030] Compared with existing technologies, this invention trains neural networks based on historical data and measured data respectively to obtain accurate wind turbine impedance data for different types of wind farms under typical seasonal climate data. By using typical seasonal climate data, wind turbine impedance data, line impedance data, and transformer impedance data, an accurate wind farm impedance model is established. An accurate wind farm impedance model can better predict the volatility and uncertainty of wind energy, thereby improving the stability of the power system. It also helps researchers studying wind farm impedance to reduce a lot of work and solves the problem of accurate impedance modeling of wind turbines and power stations in wind farms. Attached Figure Description

[0031] Figure 1 This is a case study diagram of wind farm impedance modeling in this invention.

[0032] Figure 2 This is a flowchart illustrating the overall implementation of the wind farm impedance modeling method of the present invention.

[0033] Figure 3 This is a detailed flowchart of the wind farm impedance modeling method of the present invention.

[0034] Figure 4 This is a schematic diagram of neural network training between weather data and wind turbine operating data at a wind farm.

[0035] Figure 5 This is a schematic diagram of neural network training for the relationship between wind turbine output and wind turbine impedance. Detailed Implementation

[0036] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments are 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0037] The application scenario of this invention is wind farm impedance modeling. For example... Figure 1 As shown, to establish an accurate impedance model for a wind farm with n wind turbines, it is necessary to obtain the impedance data of each wind turbine, transformer, and line. This invention starts from the conditions affecting wind power generation and obtains an accurate wind farm impedance model under typical four-season climate data (temperature, humidity, wind speed, and wind direction).

[0038] The first aspect of this invention provides a wind farm impedance modeling method, the overall implementation process of which is as follows: Figure 2As shown, according to the present application, first, typical climate data of a wind farm and actually measured line length data are input, then impedance data of each line are calculated, impedance data of each type of transformer are calculated from the climate data, wind turbine power and voltage of the current climate data are calculated based on a first neural network model, then impedance data of each wind turbine are calculated based on a second neural network model, and a wind farm impedance model is generated according to Figure 1 the topology and impedance data of each device (wind turbine, transformer, line) of the wind farm.

[0039] A detailed flowchart of an embodiment of the wind farm impedance modeling method described in the present application is shown in Figure 3 As shown, Figure 3 the impedance data generation process of each device (wind turbine, transformer, line) in the wind farm is shown in the dashed box, and by inputting typical climate data and through the trained neural network model and the actually measured data, the impedance data of each device of the wind farm can be obtained.

[0040] A second aspect of an embodiment of the present application provides a wind turbine impedance data acquisition method in a wind farm, and the implementation process is shown in Figure 3 The method includes the following steps:

[0041] Step S1, a first database is established, and the first database includes: historical climate data of each time point of the wind farm, and wind turbine operating condition data of each numbered wind turbine at the corresponding time point.

[0042] Specifically, the historical climate data of the wind farm can be recorded, which can be data under typical climate conditions of the four seasons; and the wind turbine operating condition data at the same time point can be recorded. More specifically, the wind turbine operating condition data of each wind turbine in the wind farm under the corresponding historical typical climate data of the four seasons is preferably obtained by the SCADA system (Supervisory Control And Data Acquisition, data acquisition and monitoring control system), but is not limited thereto.

[0043] In some preferred embodiments, the step S1 further includes time synchronization verification of the historical climate data and the corresponding wind turbine operating condition data, so as to ensure time synchronization (the error accuracy can be millisecond level).

[0044] Step S2, according to the first database, a first neural network model is constructed and trained with climate data parameters as input and wind turbine operating condition data of each numbered wind turbine as output. Preferably, the climate data parameters can include, but are not limited to, any one or more of: temperature, humidity, wind speed, and wind direction; and the wind turbine operating condition data can include, but are not limited to, any one or more of: active power, reactive power, and voltage.

[0045] Specifically, as shown in Figure 4As shown, the first neural network model can be trained on the first database using a multi-layer neural network. The inputs are temperature T, humidity RH, wind speed WS, and wind direction WD, and the outputs are the active power P1-Pn, reactive power Q1-Qn, and voltage U1-Un of each numbered wind turbine. The neural network structure and parameters are adjusted multiple times, and after training, the first neural network model related to the wind farm climate data and the operating conditions of each wind turbine is obtained.

[0046] The main purpose of constructing the first neural network model is:

[0047] In a wind farm, there are multiple, even dozens or hundreds of wind turbines. Wind turbines with different numbers are located in different positions in the wind farm. By constructing a first neural network model, it is possible to establish the intricate relationship between the operating data of each wind turbine with different numbers under various climate data, and reflect the different impacts of the same climate data on wind turbines at different locations, especially the different impacts of wind direction and wind speed on wind turbines at different locations.

[0048] exist Figure 4 In the preferred embodiment shown, the first neural network model adopts a four-layer BP neural network structure. The transfer function between the input layer and the hidden layer of this four-layer BP neural network structure is the tanh function, the transfer function between the hidden layers is the sigmoid function, and the transfer function between the hidden layer and the output layer is the softplus function. The number of neurons in each layer of this four-layer BP neural network structure can be set as needed. In a preferred embodiment, the topology of this four-layer BP neural network is 4×64×128×3n, where n is the number of wind turbines in the wind farm. The input data of the first neural network model is data from the wind farm meteorological tower, including temperature T, humidity RH, wind speed WS, and wind direction WD. The input data is preprocessed using a normalization method. The output data of the first neural network model are the active power P1-Pn, reactive power Q1-Qn, and voltage U1-Un of each numbered wind turbine.

[0049] Furthermore, steps S1-S2 above only consider the complex relationships between the operating data of each wind turbine under different climate conditions, assuming the turbine locations and corresponding numbers remain unchanged for a fixed wind farm. Since wind farm construction is very complex, and the scale of wind farms varies in different application scenarios, the number of wind turbines and their layout also differ. Therefore, to adapt to different wind farm structural layouts, the preferred approach is:

[0050] In step S1, the first database further comprises: position parameters of each numbered wind turbine; in step S2, the input of the first neural network model further comprises the position parameters of each numbered wind turbine. Preferably, the position parameters can be selected from, but are not limited to, including: body position parameters and relative position parameters; as shown in Table 1, the preferred embodiment of the body position parameters is given, which can be selected from, but is not limited to, including any one or more of plane coordinates X, plane coordinates Y, longitude, latitude, elevation, and cabin orientation angle; the relative position parameters can be selected from, but are not limited to, including one or more of wind turbine volume rate (number of wind turbines in a set range), adjacent wind turbine distance (average value of adjacent wind turbine distance in each direction, if the wind turbine is at the edge, there is no other wind turbine in a certain direction; or the distance in a certain direction exceeds the set threshold, then the distance in this direction is 0).

[0051] Table 1: Body position parameters of wind turbine

[0052]

[0053] Step S3, a second database is established, and the second database comprises: climate data of each time point of the wind farm, wind turbine working condition data of the target wind turbine, and wind turbine sequence impedance data of the grid connection point of each target wind turbine at the corresponding time point.

[0054] Specifically, in step S3, the related data in the second database is preferably obtained by semi-physical simulation test, including:

[0055] S31: According to the differences of each wind turbine in the wind farm, each numbered wind turbine in the wind farm is classified, and the target wind turbine in each type is determined.

[0056] Specifically, the differences of each wind turbine in the wind farm can be selected, such as finding the different wind turbine models in the station, classifying according to different models, power and the like, and selecting at least one as the target wind turbine in each type; or according to the working condition data of each numbered wind turbine in step S2, the differences of each wind turbine are determined, the different influences of the same climate data on different position numbered wind turbines are reflected, and each numbered wind turbine in the wind farm is classified, and at least one is selected as the target wind turbine in each type. Preferably, each wind turbine in the wind farm is accurately classified according to the differences, and then a representative of each type is randomly extracted for functional test. After the test result is that the function is complete, it is determined as the target wind turbine.

[0057] S32: Under each climate data, the target wind turbine is subjected to semi-physical simulation test, impedance scanning is performed through a semi-physical simulation platform, and climate data of each time point, wind turbine working condition data of the target wind turbine, and wind turbine sequence impedance data of the grid connection point of each target wind turbine at the corresponding time point are obtained.

[0058] Specifically, under typical four-season climate data, the current wind farm climate data, active power, reactive power, and voltage values ​​of the wind turbine output are recorded. The sequence impedance of the wind turbine grid connection point is obtained by impedance scanning through a hardware-in-the-loop simulation platform. The above operation is repeated to establish wind turbine sequence impedance data corresponding to active power, reactive power, and voltage under typical four-season climate data for each type of wind turbine, thus constructing a second database.

[0059] Step S4: Based on the second database, construct and train a second neural network model with wind turbine operating parameters as input and wind turbine sequence impedance data as output.

[0060] like Figure 5 As shown, the input values ​​for the active power, reactive power, and voltage of the fan are displayed, and the output values ​​for the fan are displayed in f1-f. m Sequence impedance at different frequencies (Hz). A second database established from various wind turbine models yields neural network models of the output and sequence impedance of each model, enabling impedance expansion from typical operating conditions to all operating conditions.

[0061] exist Figure 5 In the preferred embodiment shown, the second neural network model adopts a six-layer BP neural network structure. The transfer function between the input layer and the hidden layer of this six-layer BP neural network structure is the ReLU function, the transfer function between the hidden layers is the ReLU function, and the transfer function between the hidden layer and the output layer is the softplus function. The number of neurons in each layer of this six-layer BP neural network structure can be set as needed. In a preferred embodiment, the topology of this six-layer BP neural network is 4×40×56×128×256×m, where m is the number of frequencies of interest. The input data of the second neural network model are the active power P, reactive power Q, and voltage U of each wind turbine, and the output data of the second neural network model are the values ​​of each wind turbine at f1-f... m Sequence impedance at different frequencies (Hz).

[0062] Step S5: Input the current climate data of the wind farm into the first neural network model, and use the output of the first neural network model as the input of the second neural network model. The output of the second neural network model is the wind turbine impedance data in the wind farm.

[0063] After obtaining the wind turbine impedance data using the aforementioned method, a wind farm impedance model, i.e., an equivalent network, can be generated based on the wind turbine impedance data, line impedance data, transformer impedance data, and topology.

[0064] In this invention, the method for obtaining line impedance data preferably includes, but is not limited to, the following:

[0065] Under seasonal climate data, impedance scanning is performed on the lines of various types in the wind farm to obtain the line sequence impedance data. Then, based on the type and length of each line, the impedance data of each line in the station is obtained.

[0066] In this invention, the method for obtaining transformer impedance data preferably includes, but is not limited to, the following:

[0067] Under seasonal climate data, impedance scanning is performed on transformers of various types in the wind farm to obtain transformer sequence impedance data, and then the impedance data of each transformer is obtained based on the model of each transformer.

[0068] like Figure 3 As shown, typical climate data of the wind farm is input, and then a first neural network model trained based on historical data is used to calculate the active power, reactive power, and voltage of each wind turbine. The active power, reactive power, and voltage of each wind turbine are then input into a trained second neural network model to output the wind turbine sequence impedance. The middle and right branches respectively obtain the sequence impedance of lines and transformers by impedance scanning under typical climate data. The impedance of each line is obtained by inputting its model and length, and the impedance of each transformer is obtained by inputting its model. By inputting the impedance data of all equipment into the equivalent topology of the wind farm, the wind farm impedance model can be constructed.

[0069] In summary, addressing the shortcomings of existing methods in accurately acquiring wind turbine impedance data and accurately establishing wind farm impedance models, this invention provides a method for acquiring wind turbine impedance data and modeling wind farm impedance based on the fusion of measured data and neural networks. It establishes a neural network relating typical wind farm climate data to the power of each wind turbine using historical wind farm data. It measures the sequence impedance data of different types of wind turbines, transformers, and power lines under typical operating conditions using measured methods. Based on this data and utilizing the wind farm topology, an accurate impedance model of the wind farm is established. The neural network is trained using historical weather records from the wind farm and historical data stored in the SCADA system, ensuring reliable data sources. Inputting climate data and a specific wind turbine type is sufficient to identify sequence impedance data, making it convenient to use and significantly reducing the workload for researchers studying wind farm impedance, thus solving the problem of accurate impedance modeling of wind turbines and wind farms.

[0070] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0071] Those skilled in the art can understand that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or in combination of computer software and electronic hardware. Whether the functions are realized in hardware or software manner depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0072] In the embodiments provided by the present application, it should be understood that the method embodiments described above are only illustrative, and the disclosed methods can be implemented in other ways.

[0073] The above-described embodiments are merely used to illustrate the technical solutions of the present application, rather than limit the present application; even though the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalent replacements; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for obtaining impedance data of wind turbines in a wind farm, characterized in that, The method comprises the following steps: Step S1, establishing a first database, the first database comprising: historical climate data of each time of the wind farm, and wind turbine operating condition data of each numbered wind turbine at the corresponding time; the climate data comprising any one or more of temperature, humidity, wind speed, and wind direction; the wind turbine operating condition data comprising any one or more of active power, reactive power, and voltage; Step S2, constructing and training a first neural network model with climate data parameters as input and wind turbine operating condition data of each numbered wind turbine as output according to the first database; the first neural network model adopting a four-layer BP neural network structure; the transfer function between the input layer and the hidden layer of the four-layer BP neural network structure being a tanh function, the transfer function between the hidden layers being a sigmoid function, and the transfer function between the hidden layer and the output layer being a softplus function; Step S3, establishing a second database, the second database comprising: climate data of each time of the wind farm, wind turbine operating condition data of the target wind turbine, and wind turbine sequence impedance data of each target wind turbine at the corresponding time; Step S4, constructing and training a second neural network model with wind turbine operating condition parameters as input and wind turbine sequence impedance data as output according to the second database; the output of the second neural network being the sequence impedance of the wind turbine at different frequencies; the second neural network model adopting a six-layer BP neural network structure, the transfer function between the input layer and the hidden layer of the six-layer BP neural network structure being a ReLU function, the transfer function between the hidden layers being a ReLU function, and the transfer function between the hidden layer and the output layer being a softplus function; Step S5, inputting the current climate data of the wind farm into the first neural network model, and inputting the output of the first neural network model into the second neural network model, the output of the second neural network model being the wind turbine impedance data in the wind farm.

2. The method of claim 1, wherein, In step S1, the first database further comprises: position parameters of each numbered wind turbine; in step S2, the input of the first neural network model further comprises the position parameters of each numbered wind turbine.

3. The method of claim 2, wherein, The position parameters comprise: any one or more of body position parameters and relative position parameters; The body position parameters comprise: any one or more of plane coordinates, plane coordinates, longitude, latitude, elevation, and cabin orientation angle; The relative position parameters comprise: wind turbine volume rate, and distance from adjacent wind turbines.

4. The method of claim 1 to 3, wherein Step S3 comprises: S31: classifying each numbered wind turbine in the wind farm according to the differences between the wind turbines in the wind farm, and determining the target wind turbine in each class; S32: performing semi-physical simulation testing on the target wind turbine under each climate data, and obtaining the climate data of each time, the wind turbine operating condition data of the target wind turbine, and the wind turbine sequence impedance data of each target wind turbine at the corresponding time through impedance scanning on a semi-physical simulation platform.

5. A wind farm impedance modeling method, characterized by, The method comprises: obtaining the wind turbine impedance data by the wind turbine impedance data acquisition method in any one of claims 1 to 4; generating a wind farm impedance model based on the wind turbine impedance data, line impedance data, transformer impedance data, and according to a topological structure.

6. The wind farm impedance modeling method of claim 5, wherein, The line impedance data acquisition method comprises: Under the four seasons climate data, the line impedance data of each type of line in the wind farm is obtained by impedance scanning, and then the impedance data of each line in the station is obtained based on the type and length of each line.

7. The wind farm impedance modeling method of claim 5, wherein, The transformer impedance data acquisition method comprises: Under the four seasons climate data, the transformer impedance data of each type of transformer in the wind farm is obtained by impedance scanning, and then the transformer impedance data of each transformer is obtained based on the type of each transformer.

Images