Impedance model identification method and system for new energy power generation system based on neural network

Abstract

The present invention provides a method and system for identifying an impedance model of a new energy power generation system based on a neural network. The training data set and the test data set of the new energy power generation system under different steady-state operating points are obtained by a frequency sweep method and normalized; the neural network is trained by the training data set to obtain a neural network with impedance characteristics of the new energy power generation system; the input data of the test data set is input into the neural network to obtain an impedance identification result; the mean square error is obtained in combination with the output data of the test data set, and the number of hidden layers of the neural network and the number of neurons in each hidden layer are adjusted so that the mean square error (MSE) is less than a set threshold; the input data under any steady-state operating point is input into the obtained neural network to obtain the impedance output under the corresponding operating point. The present invention can obtain the impedance characteristics of the power generation system under the corresponding operating point by only measuring the steady-state operating point data, and has high accuracy.

Description

Technical Field

[0001] The present invention relates to the field of renewable energy power generation technology, and in particular to a method and system for identifying an impedance model of a renewable energy power generation system based on a neural network, and also provides a corresponding terminal and a computer-readable storage medium. Background Art

[0002] In recent years, with the continuous increase in the penetration rate of new energy power generation such as wind power and photovoltaic power generation, the impact of new energy power generation systems on the stability of the power grid has become increasingly prominent. When analyzing the stability of the grid-connected system of the new energy power generation system, the impedance analysis method divides the new energy power generation system and the power grid into two subsystems: power supply and load. According to their respective control structures and parameters, impedance models are established separately, which reduces the difficulty of system modeling and analysis. It has become one of the mainstream methods for studying the stability of the grid-connected interactive system. However, theoretical impedance modeling requires the control structure and parameters of the new energy power generation system to be known in advance, and the derivation process is relatively complicated, which is not conducive to the application of black / gray box models and complex systems in actual engineering. Therefore, it is very necessary to establish a model identification method that can obtain black box impedance.

[0003] Impedance modeling is usually performed using theoretical methods. In the theoretical modeling method, the internal control structure and control parameters of the model need to be known in advance. According to the provided controller structure and parameters, a detailed impedance model is established using the small signal linearization theory. This process often requires a large amount of calculation, the derivation process is relatively complex, and the controller results and parameters need to be available in advance, which is not conducive to the impedance modeling of black / gray box models in actual engineering. The neural network-based modeling method inputs the measured model input and output data into the neural network, continuously trains the neural network, and ultimately makes the neural network as equivalent to the black box model as possible. However, this method will lead to problems such as long training time and large errors in training results due to the inconsistent value ranges of different physical quantity data.

[0004] After searching, we found:

[0005] The Chinese invention patent application "A method for online identification of power grid impedance based on neural network" with publication number CN112946363A includes: determining the relevant technical index requirements for power grid impedance identification; using the harmonic injection method to collect the required corresponding data samples, that is, injecting disturbance signals that meet the spectrum width and accuracy requirements into the power grid, and then sampling the voltage and current at the grid common point to obtain disturbance and response signals; establishing an impedance identification algorithm model based on a neural network, and determining the neural network algorithm and the initial assignment of related parameters based on the data sample characteristics obtained in step 2, and initializing the neural network; using the obtained data samples to train the established neural network model and select a training method; checking whether the fitting accuracy of the trained power grid impedance model under the required spectrum width meets the corresponding technical index requirements; and using the neural network algorithm for data processing. This impedance identification method is an online identification method, and the following problems still exist:

[0006] The impedance identification method directly inputs the collected data into the neural network for training. Due to the different value ranges between different data sets, the training time is too long and too many training times are required to achieve a more accurate identification effect. In the embodiment of the method, the number of learning times is set to 1,000,000 times and then the iteration is terminated, which consumes a lot of training time and is not conducive to the application of online identification. Summary of the invention

[0007] In view of the problems existing in the above-mentioned prior art, the present invention proposes a method and system for identifying an impedance model of a new energy power generation system based on a neural network, and also provides a corresponding terminal and a computer-readable storage medium.

[0008] According to one aspect of the present invention, a method for identifying an impedance model of a new energy power generation system based on a neural network is provided, comprising:

[0009] The frequency sweep method is used to obtain the training data set and test data set of the renewable energy power generation system at different steady-state operating points;

[0010] Performing normalization preprocessing on the training data set and the test data set respectively, and converting the training data set and the test data set to the same value interval;

[0011] Using the normalized training data set to train a neural network to obtain a neural network model that characterizes the impedance characteristics of a new energy power generation system;

[0012] Using the normalized test data set to test the neural network model, and outputting an impedance identification result based on the neural network;

[0013] Performing error analysis on the impedance identification result based on the normalized test data set to obtain the mean square error (MSE); if the mean square error (MSE) is greater than or equal to a set threshold, adjusting the number of hidden layers and the number of neurons in each hidden layer of the neural network model until the mean square error (MSE) is less than the set threshold, thereby obtaining an impedance model;

[0014] Input data of any steady-state operating point of the new energy power generation system is input into the impedance model to obtain the impedance output at the corresponding operating point, thereby completing the impedance identification of the new energy power generation system.

[0015] Optionally, the method of obtaining a training data set and a test data set of a new energy power generation system at different steady-state operating points by using a frequency sweep method includes:

[0016] Build a time domain simulation model of the renewable energy power generation system based on Matlab / Simulink, or build a hardware-in-the-loop simulation model of the renewable energy power generation system based on a real-time simulator based on the actual renewable energy power generation system controller to obtain a simulation model of the renewable energy power generation system;

[0017] The active power output by the simulation model of the new energy power generation system is changed at equal intervals to obtain the voltage V of the dq axis at each working point. d 、V q and current I d ,I q Data, frequency f req And each operating point at frequency f req The sequence impedance Z obtained under pp , Z pn , Z np and Z nn The training data set and the test data set are obtained respectively.

[0018] Optionally, the new energy power generation system simulation model includes: a photovoltaic power generation system simulation model, a doubly-fed wind power generation system simulation model or a full-power wind power generation system simulation model.

[0019] Optionally, obtaining a training data set and a test data set respectively includes:

[0020] The active power output of the simulation model of the new energy power generation system is changed at equal intervals, and the dq axis voltage V at the grid connection point in the steady state is collected. d 、V q and current I d ,I q data, and with frequency f req Together they are constructed into an input data set;

[0021] In steady state, the injection frequency at the grid connection point is f reqThe small signal voltage disturbance is measured, the voltage and current response at the corresponding frequency is measured, and the sequence impedance Z of the simulation model of the new energy power generation system is calculated at each operating point. pp , Z pn , Z np and Z nn , taking the magnitude and phase angle of the sequence impedance as output data sets;

[0022] Using the input data set and the output data set as training data sets;

[0023] The interval of the active power output by the simulation model of the new energy power generation system is reduced, and the input data set and the output data set under multiple working points are obtained in the same way as the test data set.

[0024] Optionally, the calculation of the sequence impedance Z of the simulation model of the new energy power generation system at each operating point is pp , Z pn , Z np and Z nn ,include:

[0025] At frequency f req The positive and negative sequence voltages V under the disturbance are obtained by injecting a three-phase positive sequence voltage disturbance at 1 Hz. p1 、V n1 and positive and negative sequence current I p1 ,I n1 ; at frequency f req The three-phase negative sequence voltage disturbance is injected at 1Hz to obtain the positive and negative sequence voltage V under the disturbance. p2 、V n2 and current I p2 ,I n2 ; Repeat this process to collect positive and negative sequence voltage and current data;

[0026] According to the positive and negative sequence voltage and current data, the input impedance Z of the simulation model of the new energy power generation system in the sequence coordinate system is calculated:

[0027]

[0028] The input impedance Z includes the following four elements: Z pp , Z pn , Z np and Z nn , with Z pp , Z pn , Z np and Z nn The amplitude and phase angle data of are used as the impedance data set at a specific working point;

[0029] The active power output of the simulation model of the new energy power generation system is set to change from 0.1pu to 1pu at an interval of 0.1pu per unit, and the dq axis voltage V at the grid connection point in the steady state is collected. d 、V q and current I d ,I q data, and with frequency f req together construct an input data set to form a training data set; repeat the above impedance acquisition process to obtain multiple impedance data sets at different working points, and obtain an output data set of the training data set;

[0030] The active power output of the simulation model of the new energy power generation system is set to change from 0.1pu to 1pu at an interval of 0.05pu per unit, and the dq axis voltage V at the grid connection point in the steady state is collected. d 、V q and current I d ,I q data, and with frequency f req Together they are constructed into an input data set for the test data set; the above impedance acquisition process is repeated to obtain multiple groups of impedance data sets at different working points, and an output data set for the test data set is obtained.

[0031] Optionally, the using the normalized training data set to train a neural network to obtain a neural network model characterizing the impedance characteristics of a new energy power generation system includes:

[0032] Initialize the weights and biases of the neural network, set the initial number of hidden layers and neurons in the hidden layers, and set the initial learning rate;

[0033] The training data set is normalized according to the following formula and then input into the neural network, and the weights and biases on each neuron are continuously adjusted to obtain a neural network model that can characterize the impedance characteristics of the new energy power generation system:

[0034]

[0035] Among them, x max With x min are the maximum and minimum values ​​in the same physical quantity data set, respectively. max With y min are the upper and lower limits of the value range to be converted, respectively, i is the data to be normalized, y i For x i The normalized result of .

[0036] Optionally, the using the normalized test data set to test the neural network model and outputting an impedance identification result based on the neural network includes:

[0037] The input data set in the test data set is normalized and then input into the neural network model, and the output result is denormalized to obtain the impedance identification result based on the neural network.

[0038] Optionally, performing error analysis on the impedance identification result based on the normalized test data set includes:

[0039] The output data set in the test data set is used as reference data, and the impedance identification result is subjected to the following error analysis to obtain the mean square error (MSE):

[0040]

[0041] Where N is the number of samples, M is the number of neural network outputs, and y i m is the mth output data of the ith sample of the neural network, Y i m is the mth output data of the measured i-th sample;

[0042] If the obtained mean square error MSE is less than the set threshold, it is considered that the trained neural network can better fit the impedance characteristics of the new energy power generation system;

[0043] If the obtained mean square error MSE is greater than or equal to the set threshold, the number of hidden layers of the neural network and the number of neurons in each hidden layer are adjusted, and the neural network training is performed again.

[0044] Optionally, the set threshold is 2%.

[0045] Optionally, inputting input data of the new energy power generation system at any steady-state operating point into the impedance model to obtain the impedance output at the corresponding operating point to complete the impedance identification of the new energy power generation system includes:

[0046] Based on the impedance model, input data at any steady-state operating point is input to obtain the impedance curve of the new energy power generation system at the operating point.

[0047] According to another aspect of the present invention, a new energy power generation system impedance model identification system based on a neural network is provided, comprising:

[0048] A data set construction module, which uses a frequency sweep method to obtain a training data set and a test data set of a new energy power generation system at different steady-state operating points, and normalizes the training data set and the test data set respectively, so that the training data set and the test data set are converted to the same value interval;

[0049] A neural network training module, which uses the normalized training data set to train a neural network to obtain a neural network model that characterizes the impedance characteristics of a new energy power generation system;

[0050] A neural network testing module, which uses the normalized test data set to test the neural network model and outputs an impedance identification result based on the neural network;

[0051] An impedance model construction module, which performs error analysis on the impedance identification result based on the normalized test data set to obtain a mean square error (MSE); if the mean square error (MSE) is greater than or equal to a set threshold, the number of hidden layers and the number of neurons in each hidden layer of the neural network model are adjusted until the mean square error (MSE) is less than the set threshold, thereby obtaining an impedance model;

[0052] The impedance identification module inputs the input data of the new energy power generation system at any steady-state operating point into the impedance model, obtains the impedance output at the corresponding operating point, and completes the impedance identification of the new energy power generation system.

[0053] According to a third aspect of the present invention, there is provided a terminal comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor can be used to execute any of the above-mentioned methods or to run the above-mentioned system when executing the program.

[0054] According to a fourth aspect of the present invention, there is provided a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can be used to execute any of the above methods, or to run the above system.

[0055] Due to the adoption of the above technical solution, the present invention has the following beneficial effects:

[0056] The neural network-based impedance model identification method and system for a new energy power generation system provided by the present invention only need to input the measured model input and output data into the neural network, and continuously train the neural network to make the neural network as equivalent to the black box model as possible. Based on the obtained neural network, the steady-state operating point of the system is input into it to obtain the impedance model under the steady-state operating point, which solves the problem that the impedance model of the black box system is difficult to obtain when the operating point changes in actual engineering.

[0057] The neural network-based impedance model identification method and system provided by the present invention trains the neural network based on the measured voltage, current and impedance data to establish an impedance identification model of the new energy power generation system related to the working point. Based on the model, the steady-state working point of the system is input to obtain the impedance characteristics of the new energy power generation system at the working point, thereby solving the problem that the impedance model of the actual black box system is difficult to obtain and the impedance sweep takes a long time.

[0058] The impedance model identification method and system of the new energy power generation system based on neural network provided by the present invention normalizes the collected data, and the processed data are all within the same value range, eliminating the influence caused by inconsistent data value ranges, and can greatly speed up the training speed of the neural network when inputting the neural network training. Based on the obtained neural network, it is only necessary to input the steady-state operating point of the system into it to obtain the impedance model under the steady-state operating point.

[0059] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0061] Figure 1 The following is a flowchart of a method for identifying an impedance model of a new energy power generation system based on a neural network in one embodiment of the present invention;

[0062] Figure 2 A schematic diagram of the topology and control structure of a doubly-fed wind power generation system in a preferred embodiment of the present invention;

[0063] Figure 3 A neural network structure diagram in a preferred embodiment of the present invention;

[0064] Figure 4 (a) to (h) are impedance data graphs for training neural networks collected by a frequency sweep method in a preferred embodiment of the present invention;

[0065] Figure 5 (a) to (h) are respectively based on a preferred embodiment of the present invention Figure 4 Impedance data diagram of the neural network output obtained from each impedance data;

[0066] Figure 6 (a) to (h) are respectively error result diagrams between test data and impedance obtained by neural network training in a preferred embodiment of the present invention;

[0067] Figure 7The figure is a schematic diagram of the component modules of the impedance model identification system of a new energy power generation system based on a neural network in one embodiment of the present invention. DETAILED DESCRIPTION

[0068] The following is a detailed description of the embodiments of the present invention: This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given. It should be pointed out that for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention.

[0069] Figure 1 A flowchart of a method for identifying an impedance model of a new energy power generation system based on a neural network according to an embodiment of the present invention.

[0070] like Figure 1 As shown, the neural network-based impedance model identification method for a new energy power generation system provided in this embodiment may include the following steps:

[0071] S100, using a frequency sweep method to obtain a training data set and a test data set of a new energy power generation system at different steady-state operating points; performing normalization preprocessing on the training data set and the test data set, respectively, and converting the training data set and the test data set to the same value range;

[0072] S200, training a neural network using the normalized training data set to obtain a neural network model that characterizes impedance characteristics of a new energy power generation system;

[0073] S300, using the normalized test data set to test the neural network model, and outputting the impedance identification result based on the neural network;

[0074] S400, performing error analysis on the impedance identification result based on the normalized test data set to obtain the mean square error (MSE); if the mean square error (MSE) is greater than or equal to a set threshold, adjusting the number of hidden layers and the number of neurons in each hidden layer of the neural network model until the mean square error (MSE) is less than the set threshold, thereby obtaining an impedance model;

[0075] S500, inputting input data of the new energy power generation system at any steady-state operating point into the impedance model, obtaining impedance output at the corresponding operating point, and completing impedance identification of the new energy power generation system.

[0076] As a preferred embodiment, each step of the above method provided by this embodiment is further described in detail below.

[0077] The above S100 may further include the following steps:

[0078] S11, based on an actual renewable energy power generation system, or based on Matlab / Simulink to build a time domain simulation model of the renewable energy power generation system, or based on an actual renewable energy power generation system controller to build a new energy power generation system hardware-in-the-loop simulation model based on a real-time simulator, to obtain a renewable energy power generation system simulation model;

[0079] S12, change the active power output of the simulation model of the new energy power generation system at equal intervals, and obtain the voltage V of the dq axis at each working point d 、V q and current I d ,I q Data, frequency f req And each operating point at frequency f req The sequence impedance Z obtained under pp , Z pn , Z np and Z nn In a preferred embodiment, S12 may further include the following steps:

[0080] S121, change the active power output of the simulation model of the new energy power generation system at equal intervals, that is, change the working point, and collect the dq axis voltage V at the grid connection point in the steady state d 、V q and current I d ,I q data, and with frequency f req Together they are constructed into an input data set;

[0081] S122, the injection frequency at the grid connection point in steady state is f req The small signal voltage disturbance is measured, the voltage and current response at the corresponding frequency is measured, and the sequence impedance Z of the simulation model of the new energy power generation system is calculated at each operating point. pp , Z pn , Z np and Z nn , the magnitude and phase angle of the sequence impedance are taken as the output data set;

[0082] S123, using the input data set and the output data set as training data sets;

[0083] S124, reducing the interval of the active power output by the new energy power generation system simulation model, and obtaining input data sets and output data sets under multiple working points in the same way as test data sets.

[0084] In a specific application example, the new energy power generation system simulation model in the above S11 may include: a photovoltaic power generation system simulation model, a doubly-fed wind power generation system simulation model or a full-power wind power generation system simulation model.

[0085] In a preferred embodiment, the above calculation of the sequence impedance Z of the simulation model of the new energy power generation system at each operating point pp , Z pn , Z np and Z nn , may further include the following steps:

[0086] S12i, based on the simulation model of the renewable energy power generation system built in S11, at the frequency f req The positive and negative sequence voltages V under the disturbance are obtained by injecting a three-phase positive sequence voltage disturbance at 1 Hz. p1 、V n1 and positive and negative sequence current I p1 ,I n1 , inject three-phase negative sequence voltage disturbance to obtain the positive and negative sequence voltage V under the disturbance p2 、V n2 and current I p2 ,I n2 ; This process can be repeated to collect positive and negative sequence voltage and current data;

[0087] S12ii: Based on the small disturbance voltage and current data obtained in S12i, the input impedance Z of the new energy power generation system in the sequence coordinate system is calculated by the following formula:

[0088]

[0089] The input impedance Z consists of four elements: Z pp , Z pn , Z np , Z nn , with Z pp , Z pn , Z np , Z nn The amplitude and phase angle data are used as impedance data sets to collect impedance data sets at specific working points;

[0090] S12iii: setting the output power of the new energy power generation system to change from 0.1pu to 1pu at an interval of 0.1pu per unit, repeating the process in S12 to obtain multiple groups of impedance data sets (for example, ten groups of impedance data sets) at different working points as output data sets in the training data set;

[0091] S12iv: Set the output power of the new energy power generation system to change from 0.1pu to 1pu at an interval of 0.05pu per unit, and repeat the process in S12 to obtain multiple groups of impedance data sets under different working points as the output data sets in the test data set.

[0092] In a preferred embodiment, the above 200 may further include the following steps:

[0093] S201: Initialize the neural network; initialize the weights and biases of the neural network, set the initial number of hidden layers and the number of neurons in the hidden layers, and set the initial learning rate;

[0094] In a specific application example, the relationship between the input and output of the impedance model under multiple working points is continuous, and the BP neural network is selected for model identification. The weights and biases of the neural network are initialized, and the number of hidden layers is initially set to 3 based on experience, and the number of neurons in the hidden layers is 8, 12, and 16 respectively, and the learning rate is initially set to 0.01.

[0095] S202: training the neural network; normalizing the input data set and the output data set in the training data set to the interval (0, 1) and the interval (-1, 1) respectively according to the following formula and inputting them into the neural network. The neural network continuously adjusts the weights and biases on each neuron through training and learning, and obtains a neural network model that can characterize the impedance characteristics of the new energy power generation system:

[0096]

[0097] Among them, x max With x min are the maximum and minimum values ​​in the same physical quantity data set, respectively. max With y min are the upper and lower limits of the value range to be converted, respectively, i is the data to be normalized, y i For x i The normalized result of .

[0098] In a preferred embodiment, the above 300 may further include the following steps:

[0099] Based on the neural network model obtained in S13, the input data set in the test data set obtained in S12 is normalized to the interval (0, 1) and then input into the neural network model, and the output is denormalized to obtain the output impedance identification result based on the neural network.

[0100] In a preferred embodiment, the above S400 may further include the following steps:

[0101] Taking the output data set in the test data set obtained in S12 as the reference data, the obtained impedance identification result based on the neural network is subjected to error analysis based on the following formula to verify the effectiveness of the proposed neural network impedance model identification method:

[0102]

[0103] Where N is the number of samples, M is the number of neural network outputs, and y im is the mth output data of the ith sample of the neural network, Y i m is the mth output data of the measured i-th sample;

[0104] In a preferred embodiment, if the MSE is less than a set threshold, it is considered that the trained neural network can better fit the impedance characteristics of the new energy power generation system; if the MSE is greater than the set threshold, the number of hidden layers of the neural network and the number of neurons in each hidden layer are adjusted, and the neural network training is performed again.

[0105] In a specific application example, the threshold value is set to 2%. The smaller the threshold value is, the better. Selecting "2%" will hardly affect the results of subsequent studies using impedance.

[0106] In a preferred embodiment, the above S500 may further include the following steps:

[0107] Based on the impedance model, the impedance curve of the renewable energy power generation system under any steady-state operating point can be obtained by inputting the operating point.

[0108] The technical solution provided by the above embodiment of the present invention is further described in detail below in conjunction with the accompanying drawings and a specific application example.

[0109] Figure 2 The main circuit and control structure schematic diagram of a doubly-fed wind power generation system of a specific application example of the present invention. This specific application example takes a doubly-fed wind power generation system as an example, and builds a steady-state simulation model of the doubly-fed wind power generation system based on Matlab. On this basis, the transmission power of the doubly-fed wind power generation system is changed at equal intervals, and data sets for neural network training and testing are collected at different working points. The neural network is trained using the data set for training to obtain a neural network with the impedance characteristics of the doubly-fed wind power generation system. The impedance model input data set for testing is input into the neural network to obtain an impedance output based on the neural network. The output is subjected to error analysis with the output data set for testing to obtain the mean square error MSE. If MSE ≥ 2%, it is necessary to adjust the number of hidden layers of the neural network and the number of neurons in each hidden layer until MSE < 2%. Inputting input data at any steady-state working point into the obtained neural network can obtain the impedance output at the corresponding working point. This method can obtain impedance data under working point changes within a large range through a small amount of impedance data.

[0110] The specific workflow of the impedance model identification method can be found in Figure 1 The workflow diagram shown includes the following steps:

[0111] The first step is to build a steady-state simulation model of the doubly-fed wind power generation system based on Matlab;

[0112] The second step is to change the active power output of the doubly-fed wind power generation system at equal intervals and collect the dq axis voltage V at each working point. d 、V q and current I d ,I q Data, with frequency f req together as the input data set; based on the simulation model of the doubly-fed wind power generation system established in the first step, the frequency of the injection at the grid connection point is f req The small signal voltage disturbance is measured, and the voltage and current response at the corresponding frequency is measured to calculate the sequence impedance Z of the doubly fed wind power generation system at each operating point. pp , Z pn , Z np , Z nn , the amplitude and phase angle of the impedance are used as the output data set; the obtained input data set and output data set are used as the training set of the neural network; the output power interval is reduced, and the input and output data sets under more working points are obtained by the same method as the test set of the neural network; the training set and the test set are normalized to obtain the corresponding data set;

[0113] The third step is to train a neural network based on the impedance model input and output data set used for training in the second step, and obtain a neural network with impedance characteristics of a doubly-fed wind power generation system;

[0114] The fourth step is to input the impedance model input data set used for testing into the neural network to obtain the impedance output of the doubly-fed wind power generation system based on the neural network;

[0115] Step 5: Using the impedance model output data used for testing in step 2 as a reference, perform error analysis on the impedance output data based on the neural network obtained in step 4 to obtain the mean square error (MSE). If MSE is ≥ 2%, the number of hidden layers and the number of neurons in each hidden layer of the neural network need to be adjusted until MSE is < 2%;

[0116] In the sixth step, input data at any steady-state operating point is input into the obtained neural network to obtain the impedance characteristics of the doubly-fed wind power generation system at the corresponding operating point.

[0117] In a preferred embodiment, the first step specifically includes:

[0118] Based on Matlab, a simulation model of a doubly-fed wind power generation system under a specific transmission power is built. The model controller is divided into a machine-side controller and a grid-side controller. The machine-side controller adopts a power outer loop and a current inner loop control, and the grid-side controller adopts V dc / Q outer loop control and current inner loop control, the controller parameters adopt the parameters used in a certain actual project, and the doubly fed wind power generation system model is adjusted to make it operate stably.

[0119] In a preferred embodiment, the second step specifically includes:

[0120] Step 1: Based on the doubly-fed wind power generation system model built in the first step, collect the dq axis voltage V at the grid connection point at 1-100Hz in steady state d 、V q and current I d ,I q Data, with frequency f req Combined together into a 5×100 input data set;

[0121] Step 2, inject disturbance to obtain voltage and current under small disturbance;

[0122] Based on the doubly-fed wind power generation system model built in the first step, at the frequency f req The positive and negative sequence voltages V under the disturbance are obtained by injecting a three-phase positive sequence voltage disturbance at 1 Hz. p1 、V n1 and positive and negative sequence current I p1 ,I n1 , inject three-phase negative sequence voltage disturbance to obtain the positive and negative sequence voltage V under the disturbance p2 、V n2 and current I p2 ,I n2 ; Repeat this process to collect positive and negative sequence voltage and current data;

[0123] Step 3: Based on the small disturbance voltage and current data obtained in step 2, the input impedance Z of the new energy power generation system in the sequence coordinate system is calculated by the following formula:

[0124]

[0125] The output data set collected at a specific working point is an 8×100 matrix.

[0126] Step 4, setting the transmission power of the doubly fed wind power generation system to change from 0.1pu to 1pu at an interval of 0.1pu per unit, repeating the process in the second step to obtain ten groups of impedance data sets at different working points for training the neural network; setting the transmission power of the doubly fed wind power generation system to change from 0.1pu to 1pu at an interval of 0.05pu per unit, repeating the process in the second step to obtain multiple groups of impedance data sets at different working points for testing the trained neural network.

[0127] In a preferred embodiment, the third step further comprises:

[0128] Step i, initialize the neural network;

[0129] The relationship between the input and output of the impedance model under multiple working points is continuous, and the BP neural network is selected for model identification. The weights and biases of the neural network are initialized, and the number of hidden layers is initially set to 3 based on experience, and the number of neurons in the hidden layers is 8, 12, and 16 respectively, and the learning rate is initially set to 0.01.

[0130] Step ii, training the neural network;

[0131] The input and output data sets in the training data set obtained in the second step are normalized according to the following formula, and then input into the neural network. The neural network continuously adjusts the weights and biases on each neuron through training and learning, and finally obtains a neural network with the corresponding impedance characteristics of the doubly-fed wind power generation system:

[0132]

[0133] Among them, x max With x min are the maximum and minimum values ​​in the same physical quantity data set, respectively. max With y min are the upper and lower limits of the value range to be converted, respectively, i is the data to be normalized, y i For x i The normalized result of .

[0134] In a preferred embodiment, the fourth step further comprises:

[0135] Based on the neural network obtained in the third step, the input data set in the test data set obtained in the second step is normalized and then input into the neural network, and the output is denormalized to obtain the output impedance based on the neural network.

[0136] In a preferred embodiment, the fifth step specifically includes:

[0137] Taking the output data set in the test data set obtained in the second step as the reference data, the error analysis of the output impedance based on the neural network is performed based on the following formula to verify the effectiveness of the proposed neural network impedance model identification method:

[0138]

[0139] Where N is the number of samples, M is the number of neural network outputs, and y i m is the mth output data of the ith sample of the neural network, Y i mis the mth output data of the measured ith sample; if the MSE is less than 2%, it is considered that the trained neural network can better fit the impedance characteristics of the doubly fed wind power generation system; if the MSE is greater than 2%, the number of hidden layers of the neural network, the number of neurons in each hidden layer and the learning rate are adjusted, and the neural network training is performed again.

[0140] In a preferred embodiment, the sixth step specifically includes:

[0141] Based on the neural network model (i.e., impedance model) that meets the requirements in the fifth step, the impedance curve of the doubly-fed wind power generation system under the working point can be obtained by inputting any steady-state working point.

[0142] In a specific application example, the topology and control structure diagram of the doubly-fed wind power generation system is shown in Figure 2 In this example, the double-fed wind power generation system uses a power outer loop and a current inner loop for the machine-side controller, and a V dc / Q outer loop control and current inner loop control.

[0143] It should be noted that the proposed neural network-based impedance model identification method for renewable energy power generation systems is also applicable to doubly-fed wind power generation systems under other control structures and other types of renewable energy power generation systems, which will not be elaborated here.

[0144] In a specific application example, the structural diagram of the neural network is as follows Figure 3 As shown. The input of the neural network is the dq axis voltage V d 、V q ,dq axis current I d ,I q and frequency f req The data set consists of five quantities; the output of the neural network is four impedances Z pp , Z pn , Z np , Z nn The data set consists of eight quantities: amplitude and phase angle. The neural network initializes each neuron with initial weight w and bias b. The neural network continuously adjusts the weight w and bias b of each neuron during each training process, ultimately achieving the effect of accurately fitting the impedance of the doubly fed wind power generation system.

[0145] In a specific application example, the impedance data collected by the frequency sweep method is as follows: Figure 4 As shown in (a) to (h) in the figure, the frequency f req The three-dimensional diagram constructed by taking the per-unit value of input active power, impedance amplitude and phase angle as coordinates can more intuitively reflect the impedance of the doubly fed wind power generation system under different working points.

[0146] In a specific application example, the input data of the training data set is input into the trained neural network, and the obtained neural network predicts the impedance as follows Figure 5 It is easy to find that this figure is similar to Figure 4 The two data sets are extremely similar, indicating that the neural network can achieve good prediction results.

[0147] In a specific application example, the output of the test data set is subtracted from the output predicted by the neural network, and the impedance error graph obtained is as follows: Figure 6 As shown in (a) to (h), the error between the predicted impedance and the actual impedance is very small, with the maximum error being only 2%, and the impedance error at most operating points is less than 1%, indicating that the neural network can well simulate the impedance characteristics of the doubly fed wind power generation system at different operating points.

[0148] It can be seen from the above embodiments that the impedance model identification method provided in the above embodiments of the present invention can obtain the impedance of the doubly-fed wind power generation system at different operating points when the internal structure and parameters are unknown, which is convenient for stability analysis when the system operating point changes, and has the advantages of modularity, simplicity and accuracy.

[0149] Figure 7 A schematic diagram of the component modules of a neural network-based impedance model identification system for a new energy power generation system provided in one embodiment of the present invention.

[0150] like Figure 7 As shown, the neural network-based impedance model identification system for a new energy power generation system provided in this embodiment may include the following modules:

[0151] A data set construction module, which uses a frequency sweep method to obtain a training data set and a test data set of a new energy power generation system under different steady-state operating points, and normalizes the training data set and the test data set respectively, so that the training data set and the test data set are converted to the same value interval;

[0152] A neural network training module, which uses a normalized training data set to train a neural network to obtain a neural network model that characterizes the impedance characteristics of a new energy power generation system;

[0153] A neural network test module, which uses a normalized test data set to test the neural network model and outputs impedance identification results based on the neural network;

[0154] Impedance model building module, which performs error analysis on the impedance identification results based on the normalized test data set to obtain the mean square error (MSE); if the mean square error (MSE) is greater than or equal to the set threshold, the number of hidden layers and the number of neurons in each hidden layer of the neural network model are adjusted until the mean square error (MSE) is less than the set threshold, and the impedance model is obtained;

[0155] The impedance identification module inputs the input data of the renewable energy power generation system at any steady-state operating point into the impedance model, obtains the impedance output at the corresponding operating point, and completes the impedance identification of the renewable energy power generation system.

[0156] It should be noted that the steps in the method provided by the present invention can be implemented using corresponding modules in the system, and those skilled in the art can refer to the technical solution of the method to implement the composition of the system, that is, the embodiments in the method can be understood as preferred examples for constructing the system, which will not be elaborated here.

[0157] An embodiment of the present invention provides a terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it can be used to execute any one of the methods in the above embodiments of the present invention, or to run any one of the systems in the above embodiments of the present invention.

[0158] An embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, can be used to execute any method in the above embodiments of the present invention, or to run any system in the above embodiments of the present invention.

[0159] Optionally, the memory is used to store programs; the memory may include volatile memory (English: volatile memory), such as random-access memory (English: random-access memory, abbreviated: RAM), such as static random-access memory (English: static random-access memory, abbreviated: SRAM), double data rate synchronous dynamic random access memory (English: Double Data Rate Synchronous Dynamic Random Access Memory, abbreviated: DDR SDRAM), etc.; the memory may also include non-volatile memory (English: non-volatile memory), such as flash memory (English: flash memory). The memory is used to store computer programs (such as applications, functional modules, etc. that implement the above method), computer instructions, etc., and the above computer programs, computer instructions, etc. can be partitioned and stored in one or more memories. And the above computer programs, computer instructions, data, etc. can be called by the processor.

[0160] The above-mentioned computer programs, computer instructions, etc. may be stored in one or more memories in partitions, and the above-mentioned computer programs, computer instructions, data, etc. may be called by a processor.

[0161] The processor is used to execute the computer program stored in the memory to implement the various steps in the method involved in the above embodiment. For details, please refer to the relevant description in the above method embodiment.

[0162] The processor and the memory may be independent structures or integrated structures. When the processor and the memory are independent structures, the memory and the processor may be coupled and connected via a bus.

[0163] The impedance model identification method and system of the new energy power generation system based on the neural network provided by the above embodiment of the present invention only need to input the measured model input and output data into the neural network, and continuously train the neural network to make the neural network as equivalent as possible to the black box model. Based on the obtained neural network, the steady-state operating point of the system is input into it, and the impedance model under the steady-state operating point can be obtained, which solves the problem that the impedance model of the black box system is difficult to obtain when the operating point changes in the actual project; based on the measured voltage, current and impedance data, the neural network is trained to establish the impedance identification model of the new energy power generation system related to the operating point. Based on the model, the steady-state operating point of the system is input, and the impedance characteristics of the new energy power generation system under the operating point can be obtained, thereby solving the problem that the impedance model of the actual black box system is difficult to obtain and the impedance sweep frequency takes a long time; the impedance characteristics of the new energy power generation system under the corresponding operating point can be obtained only by measuring the steady-state operating point data, which is suitable for the impedance identification of the new energy power generation system or other power electronic equipment with a black box control model (i.e., the new energy power generation system model), and has high accuracy.

[0164] All matters not covered in the above embodiments of the present invention are well known in the art.

[0165] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art may make various modifications or variations within the scope of the claims, which do not affect the essence of the present invention.

Claims

1. A method for identifying impedance model of new energy power generation system based on neural network, characterized in that: include: The frequency sweep method is used to obtain the training data set and test data set of the renewable energy power generation system at different steady-state operating points; Performing normalization preprocessing on the training data set and the test data set respectively, and converting the training data set and the test data set to the same value interval; Using the normalized training data set to train a neural network to obtain a neural network model that characterizes the impedance characteristics of a new energy power generation system; Using the normalized test data set to test the neural network model, and outputting an impedance identification result based on the neural network; Performing error analysis on the impedance identification result based on the normalized test data set to obtain the mean square error (MSE); If the mean square error MSE is greater than or equal to a set threshold, the number of hidden layers and the number of neurons in each hidden layer of the neural network model are adjusted until the mean square error MSE is less than the set threshold, thereby obtaining an impedance model; Input data of any steady-state operating point of the new energy power generation system is input into the impedance model to obtain the impedance output at the corresponding operating point, thereby completing the impedance identification of the new energy power generation system.

2. The method for identifying impedance model of new energy power generation system based on neural network according to claim 1 is characterized in that: The method of using the frequency sweep method to obtain a training data set and a test data set for a new energy power generation system at different steady-state operating points includes: Build a time domain simulation model of the renewable energy power generation system based on Matlab / Simulink, or build a hardware-in-the-loop simulation model of the renewable energy power generation system based on a real-time simulator based on the actual renewable energy power generation system controller to obtain a simulation model of the renewable energy power generation system; The active power output by the simulation model of the new energy power generation system is changed at equal intervals to obtain the voltage V of the dq axis at each working point. d 、V q and current I d ,I q Data, frequency f req And each operating point at frequency f req The sequence impedance Z obtained under pp , Z pn , Z np and Z nn The training data set and the test data set are obtained respectively.

3. The method for identifying impedance model of new energy power generation system based on neural network according to claim 2 is characterized in that: The training data set and the test data set are obtained respectively, including: The active power output of the simulation model of the new energy power generation system is changed at equal intervals, and the dq axis voltage V at the grid connection point in the steady state is collected. d 、V q and current I d ,I q data, and with frequency f req Together they are constructed into an input data set; In steady state, the injection frequency at the grid connection point is f req The small signal voltage disturbance is measured, the voltage and current response at the corresponding frequency is measured, and the sequence impedance Z of the simulation model of the new energy power generation system is calculated at each operating point. pp , Z pn , Z np and Z nn , taking the magnitude and phase angle of the sequence impedance as output data sets; Using the input data set and the output data set as training data sets; The interval of the active power output by the simulation model of the new energy power generation system is reduced, and the input data set and the output data set under multiple working points are obtained in the same way as the test data set.

4. The method for identifying impedance model of new energy power generation system based on neural network according to claim 3 is characterized in that: The sequence impedance Z of the simulation model of the new energy power generation system is calculated at each operating point. pp , Z pn , Z np and Z nn , including: at frequency f req The positive and negative sequence voltages V under the disturbance are obtained by injecting a three-phase positive sequence voltage disturbance at 1 Hz. p1 、V n1 and positive and negative sequence current I p1 ,I n1 ; at frequency f req The three-phase negative sequence voltage disturbance is injected at 1Hz to obtain the positive and negative sequence voltage V under the disturbance. p2 、V n2 and current I p2 ,I n2 ; According to the positive and negative sequence voltage and current data, the input impedance Z of the simulation model of the new energy power generation system in the sequence coordinate system is calculated: The input impedance Z includes the following four elements: Z pp , Z pn , Z np and Z nn , with Z pp , Z pn , Z np and Z nn The amplitude and phase angle data of are used as the impedance data set at a specific working point; The active power output of the simulation model of the new energy power generation system is set to change from 0.1pu to 1pu at an interval of 0.1pu per unit, and the dq axis voltage V at the grid connection point in the steady state is collected. d 、V q and current I d ,I q data, and with frequency f req together construct an input data set to form a training data set; repeat the above impedance acquisition process to obtain multiple impedance data sets at different working points, and obtain an output data set of the training data set; The active power output of the simulation model of the new energy power generation system is set to change from 0.1pu to 1pu at an interval of 0.05pu per unit, and the dq axis voltage V at the grid connection point in the steady state is collected. d 、V q and current I d ,I q data, and with frequency f req Together they are constructed into an input data set for the test data set; the above impedance acquisition process is repeated to obtain multiple groups of impedance data sets at different working points, and an output data set for the test data set is obtained.

5. The method for identifying impedance model of new energy power generation system based on neural network according to claim 1, characterized in that: The method of training a neural network using the normalized training data set to obtain a neural network model that characterizes the impedance characteristics of a new energy power generation system includes: Initialize the weights and biases of the neural network, set the initial number of hidden layers and neurons in the hidden layers, and set the initial learning rate; The training data set is normalized according to the following formula and then input into the neural network. The neural network automatically adjusts the weights and biases on each neuron, and finally obtains a neural network model that can characterize the impedance characteristics of the new energy power generation system: Among them, x max With x min are the maximum and minimum values ​​in the same physical quantity data set, respectively. max With y min are the upper and lower limits of the value range to be converted, respectively, i is the data to be normalized, y i For x i The normalized result of .

6. The method for identifying impedance model of new energy power generation system based on neural network according to claim 1, characterized in that: The method of testing the neural network model by using the normalized test data set and outputting the impedance identification result based on the neural network includes: The input data set in the test data set is normalized and then input into the neural network model, and the output result is denormalized to obtain the impedance identification result based on the neural network.

7. The method for identifying impedance model of new energy power generation system based on neural network according to claim 1, characterized in that: The performing error analysis on the impedance identification result based on the normalized test data set includes: The output data set in the test data set is used as reference data, and the impedance identification result is subjected to the following error analysis to obtain the mean square error (MSE): Where N is the number of samples, M is the number of neural network outputs, and y i m is the mth output data of the ith sample of the neural network, Y i m is the mth output data of the measured i-th sample; If the obtained mean square error MSE is less than the set threshold, it is considered that the trained neural network can better fit the impedance characteristics of the new energy power generation system; If the obtained mean square error MSE is greater than or equal to the set threshold, the number of hidden layers of the neural network and the number of neurons in each hidden layer are adjusted, and the neural network training is performed again.

8. A new energy power generation system impedance model identification system based on neural network, characterized in that: include: A data set construction module, which uses a frequency sweep method to obtain a training data set and a test data set of a new energy power generation system at different steady-state operating points, and normalizes the training data set and the test data set respectively, so that the training data set and the test data set are converted to the same value interval; A neural network training module, which uses the normalized training data set to train a neural network to obtain a neural network model that characterizes the impedance characteristics of a new energy power generation system; A neural network testing module, which uses the normalized test data set to test the neural network model and outputs an impedance identification result based on the neural network; An impedance model construction module, which performs error analysis on the impedance identification result based on the normalized test data set to obtain a mean square error (MSE); if the mean square error (MSE) is greater than or equal to a set threshold, the number of hidden layers and the number of neurons in each hidden layer of the neural network model are adjusted until the mean square error (MSE) is less than the set threshold, thereby obtaining an impedance model; The impedance identification module inputs the input data of the new energy power generation system at any steady-state operating point into the impedance model, obtains the impedance output at the corresponding operating point, and completes the impedance identification of the new energy power generation system.

9. A terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, it can be used to execute the method described in any one of claims 1 to 7, or to run the system described in claim 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it can be used to execute the method described in any one of claims 1 to 7, or to run the system described in claim 8.

Images