Simulation parameter optimization method and device for electromagnetic transient simulation

Abstract

The invention provides a simulation parameter optimization method and device for electromagnetic transient simulation, the electromagnetic transient simulation is used for network access evaluation and / or model evaluation of a wind turbine generator, the simulation parameter optimization method comprises the following steps: obtaining simulation data and actual measurement data of electromagnetic transient simulation of a to-be-evaluated wind turbine generator, the simulation data comprises input data, output data and an intermediate variable; calculating a first updated simulation parameter corresponding to a simulation parameter of the electromagnetic transient simulation based on the input data, the output data and the intermediate variable; calculating a second updated simulation parameter corresponding to the simulation parameter based on the input data, the output data, the intermediate variable and the simulation parameter; and calculating a final optimized simulation parameter based on the first updated simulation parameter and the second updated simulation parameter.

Description

Technical Field

[0001] The present disclosure generally relates to the fields of wind power and data processing, and more particularly, to a method and apparatus for optimizing simulation parameters of electromagnetic transient simulation. Background Art

[0002] Currently, in the grid connection simulation scenario for a wind farm, due to the large number of simulation conditions for all relevant wind turbines in the wind farm and the large amount of data involved in the implementation process of the simulation, it is becoming increasingly difficult to process and analyze these large amounts of data.

[0003] Regarding the current domestic grid connection simulation status, the main solutions adopted when performing the above simulations and related data processing are as follows: Step 1, the simulation test engineer performs simulations on all conditions one by one on a local computer; Step 2, the simulation test engineer then performs data processing and comparison on the simulation data one by one. If it is determined that the comparison result does not meet the predetermined requirements, manual simulation deviation analysis is performed and the simulation model is debugged to optimize the simulation parameters; Step 3, repeat the above two steps until all condition comparison results can meet the predetermined requirements. The main problems existing in the current grid connection simulation related processing are as follows: it requires a large amount of time and energy input from a large number of professionals; it takes a lot of time. Summary of the Invention

[0004] An exemplary embodiment of the present disclosure is to provide a method and apparatus for optimizing simulation parameters of electromagnetic transient simulation, which can improve the verification accuracy and calculation accuracy of electromagnetic transient simulation, and reduce the difficulty of error analysis and debugging of simulation model parameters.

[0005] According to an aspect of an embodiment of the present disclosure, there is provided a method for optimizing simulation parameters of electromagnetic transient simulation, where the electromagnetic transient simulation is used for grid connection assessment and / or model assessment of a wind turbine, and the simulation parameter optimization method includes: obtaining simulation data and measured data of the electromagnetic transient simulation of the wind turbine to be evaluated, where the simulation data includes input data, output data, and intermediate variables; calculating a first updated simulation parameter corresponding to the simulation parameter of the electromagnetic transient simulation based on the input data, the output data, and the intermediate variables; calculating a second updated simulation parameter corresponding to the simulation parameter based on the input data, the output data, the intermediate variables, and the simulation parameter; calculating a final optimized simulation parameter based on the first updated simulation parameter and the second updated simulation parameter.

[0006] Optionally, the step of calculating the first updated simulation parameters corresponding to the simulation parameters of the electromagnetic transient simulation may include: calculating deviation data for preset parameters based on the output data and the measured data; calculating the first updated simulation parameters by inputting the input data, the deviation data, and the intermediate variables into an LSTM model, where the LSTM model is constructed based on reference simulation data and reference measured data of a reference wind turbine with the same or similar model type as the wind turbine to be evaluated.

[0007] Optionally, the step of calculating the second updated simulation parameters corresponding to the simulation parameters may include: calculating the second updated simulation parameters by inputting the input data, the output data, the intermediate variables, and the simulation parameters into an XGBoost model, where the XGBoost model is constructed based on reference simulation data and reference measured data of a reference wind turbine with the same or similar model type as the wind turbine to be evaluated.

[0008] Optionally, the step of calculating the second updated simulation parameters corresponding to the simulation parameters may include: calculating the second updated simulation parameters by inputting the input data, the output data, the intermediate variables, and the simulation parameters into a RandomForest model, where the RandomForest model is constructed based on reference simulation data and reference measured data of a reference wind turbine with the same or similar model type as the wind turbine to be evaluated.

[0009] Optionally, the step of calculating the final optimized simulation parameters may include: calculating a first weight associated with the first updated simulation parameters, a second weight, and a third weight associated with the second updated simulation parameters by inputting the first updated simulation parameters and the second updated simulation parameters into a fully connected neural network model based on an attention mechanism, where the first weight corresponds to the first updated simulation parameters calculated by the LSTM model, the second weight corresponds to the second updated simulation parameters calculated by the XGBoost model, and the third weight corresponds to the second updated simulation parameters calculated by the RandomForest model; calculating the final optimized simulation parameters based on the first updated simulation parameters, the second updated simulation parameters, the first weight, the second weight, and the third weight.

[0010] Optionally, the electromagnetic transient simulation may include high and low voltage ride-through simulation and / or impedance evaluation simulation, and the electromagnetic transient simulation is capable of performing parallel simulation calculations for multiple operating conditions.

[0011] Optionally, each of the simulation parameters, the first updated simulation parameters, the second updated simulation parameters, and the optimized simulation parameters may include at least one of the following parameters: voltage loop proportional coefficient, voltage loop integral coefficient, current loop proportional coefficient, current loop integral coefficient, phase-locked loop proportional coefficient, phase-locked loop integral coefficient, phase-locked loop voltage locking parameter, phase-locked loop voltage unlocking parameter, full-wave voltage feedforward coefficient, fundamental wave voltage feedforward coefficient, low-ride-through active current coefficient, low-ride-through reactive current coefficient, and active power recovery speed.

[0012] Optionally, the intermediate variable may include at least one of the following variables: active current set value, reactive current set value, positive-sequence voltage, negative-sequence voltage, phase-locked loop output frequency, phase-locked loop output phase, DC bus voltage, low voltage ride-through flag, high voltage ride-through flag, and fault flag.

[0013] According to another aspect of the embodiments of the present disclosure, there is provided a simulation parameter optimization device for electromagnetic transient simulation, where the electromagnetic transient simulation is used for grid connection assessment and / or model assessment of a wind turbine, and the simulation parameter optimization device includes: an acquisition unit configured to acquire simulation data and measured data of the electromagnetic transient simulation of the wind turbine to be evaluated, where the simulation data includes input data, output data, and intermediate variables; a first parameter optimization unit configured to calculate first updated simulation parameters corresponding to the simulation parameters of the electromagnetic transient simulation based on the input data, the output data, and the intermediate variables; a second parameter optimization unit configured to calculate second updated simulation parameters corresponding to the simulation parameters based on the input data, the output data, the intermediate variables, and the simulation parameters; and a third parameter optimization unit configured to calculate final optimized simulation parameters based on the first updated simulation parameters and the second updated simulation parameters.

[0014] Optionally, the operation of the first parameter optimization unit to calculate the first updated simulation parameters corresponding to the simulation parameters of the electromagnetic transient simulation may include: calculating deviation data for a preset parameter based on the output data and the measured data; and calculating the first updated simulation parameters by inputting the input data, the deviation data, and the intermediate variables into an LSTM model, where the LSTM model is constructed based on reference simulation data and reference measured data of a reference wind turbine that is the same as or similar to the model of the wind turbine to be evaluated.

[0015] Optionally, the operation of the second parameter optimization unit to calculate the second updated simulation parameter corresponding to the simulation parameter may include: calculating the second updated simulation parameter by inputting the input data, the output data, the intermediate variable, and the simulation parameter into an XGBoost model, where the XGBoost model is constructed based on reference simulation data and reference measured data of a reference wind turbine whose model is the same as or similar to the model of the wind turbine to be evaluated.

[0016] Optionally, the operation of the second parameter optimization unit to calculate the second updated simulation parameter corresponding to the simulation parameter may include: calculating the second updated simulation parameter by inputting the input data, the output data, the intermediate variable, and the simulation parameter into a RandomForest model, where the RandomForest model is constructed based on reference simulation data and reference measured data of a reference wind turbine whose model is the same as or similar to the model of the wind turbine to be evaluated.

[0017] Optionally, the operation of the third parameter optimization unit to calculate the final optimized simulation parameter may include: calculating a first weight associated with the first updated simulation parameter, a second weight, and a third weight associated with the second updated simulation parameter by inputting the first updated simulation parameter and the second updated simulation parameter into a fully connected neural network model based on an attention mechanism, where the first weight corresponds to the first updated simulation parameter calculated by an LSTM model, the second weight corresponds to the second updated simulation parameter calculated by an XGBoost model, and the third weight corresponds to the second updated simulation parameter calculated by a RandomForest model; calculating the final optimized simulation parameter based on the first updated simulation parameter, the second updated simulation parameter, the first weight, the second weight, and the third weight.

[0018] Optionally, the electromagnetic transient simulation may include high and low voltage ride-through simulation and / or impedance evaluation simulation, and the electromagnetic transient simulation can perform parallel simulation calculations for multiple working conditions.

[0019] Optionally, each of the simulation parameter, the first updated simulation parameter, the second updated simulation parameter, and the optimized simulation parameter may include at least one of the following parameters: voltage loop proportional coefficient, voltage loop integral coefficient, current loop proportional coefficient, current loop integral coefficient, phase-locked loop proportional coefficient, phase-locked loop integral coefficient, phase-locked loop voltage locking parameter, phase-locked loop voltage unlocking parameter, full-wave voltage feed-forward coefficient, fundamental wave voltage feed-forward coefficient, low-ride-through active current coefficient, low-ride-through reactive current coefficient, and active power recovery speed.

[0020] Optionally, the intermediate variable may include at least one of the following variables: active current set value, reactive current set value, positive sequence voltage, negative sequence voltage, PLL output frequency, PLL output phase, DC bus voltage, low voltage ride-through flag, high voltage ride-through flag, and fault flag.

[0021] According to another aspect of the embodiments of the present disclosure, there is provided a computer-readable storage medium, when the instructions in the computer-readable storage medium are run by at least one processor, causing the at least one processor to execute the simulation parameter optimization method as described above.

[0022] According to another aspect of the embodiments of the present disclosure, there is provided a computer device, the computer device includes: at least one processor; at least one memory storing computer-executable instructions, wherein, when the computer-executable instructions are run by the at least one processor, causing the at least one processor to execute the simulation parameter optimization method as described above.

[0023] According to the simulation parameter optimization method and device for electromagnetic transient simulation of the exemplary embodiments of the present disclosure, it is proposed to optimize the simulation parameters of the electromagnetic transient simulation model by using a variety of simulation parameter optimization models to obtain more accurate output results of the electromagnetic transient simulation model, improving the verification accuracy and calculation accuracy of the electromagnetic transient simulation for wind turbines, and reducing the difficulty of error analysis and debugging of the simulation model parameters.

[0024] In addition, through the simulation parameter optimization method and device for electromagnetic transient simulation of the present disclosure, it is also possible to effectively shorten the time for simulation and data processing using the electromagnetic transient simulation model (for example, comparison graph generation time, error calculation time, report generation time, etc.), effectively improving the simulation efficiency. In addition, through the simulation parameter optimization method and device for electromagnetic transient simulation of the present disclosure, it is also possible to further improve the simulation efficiency by performing parallel simulation processing for multiple working conditions.

[0025] Additional aspects and / or advantages of the general concept of the present disclosure will be set forth in part in the description that follows, and in part will be obvious from the description, or may be learned by practice of the general concept of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Through the following description with reference to the drawings of the exemplary embodiments shown, the above and other objects and features of the exemplary embodiments of the present disclosure will become more apparent, wherein:

[0027] Figure 1 is a flowchart showing the simulation parameter optimization method for electromagnetic transient simulation according to the exemplary embodiments of the present disclosure;

[0028] Figure 2is an overall flowchart showing an electromagnetic transient simulation according to an exemplary embodiment of the present disclosure;

[0029] Figure 3 is a simulation flowchart for multiple working conditions according to a comparative example;

[0030] Figure 4 is a parallel simulation flowchart for multiple working conditions according to an exemplary embodiment of the present disclosure;

[0031] Figure 5 is a data processing flowchart for multiple working conditions according to a comparative example;

[0032] Figure 6 is a data processing flowchart for multiple working conditions according to an exemplary embodiment of the present disclosure;

[0033] Figure 7 is an overall flowchart showing an example simulation parameter optimization method according to an exemplary embodiment of the present disclosure;

[0034] Figure 8 is a flowchart showing an example simulation parameter optimization method according to an exemplary embodiment of the present disclosure;

[0035] Figures 9A to 9E is a simulation result diagram of a comparative example that does not use the simulation parameter optimization method of the present disclosure;

[0036] Figures 10A to 10E is a simulation result diagram of an example of the present disclosure that uses the simulation parameter optimization method of the present disclosure;

[0037] Figure 11 is a block diagram showing a simulation parameter optimization device for electromagnetic transient simulation according to an exemplary embodiment of the present disclosure;

[0038] Figure 12 is a block diagram showing a computer device according to an exemplary embodiment of the present disclosure. Detailed Embodiments

[0039] Reference will now be made in detail to the embodiments of the present disclosure, examples of which are illustrated in the accompanying drawings, wherein like reference numerals always refer to like components. The embodiments will be described below with reference to the accompanying drawings to explain the present disclosure.

[0040] At present, there are three main deficiencies in the application scenarios of wind farm grid access assessment and new model certification: (1) There are many working conditions involved in the simulation of digital models for electromagnetic transient simulation and the testing of wind farms / wind turbines to be evaluated (for example, semi-physical testing and / or on-site measurements), and the simulation process requires a lot of manpower and computing time; (2) When verifying the simulation results of the digital model and the test results of the wind farm / wind turbine to be evaluated, the workload of data preprocessing and analysis calculation, comparison of the above simulation results and the above test results, and error calculation for a large amount of simulation result data is very large; (3) In the above verification process, it is difficult to analyze the cause of the error. At the same time, the parameters of the digital model need to be adjusted repeatedly. The parameter optimization of the digital model requires strict capabilities of the simulation test personnel and is also very time-consuming due to a large amount of repetitive work.

[0041] In view of the related problems in the prior art, the present disclosure proposes a simulation parameter optimization method and device for electromagnetic transient simulation, which can solve the problem of high resource consumption for wind power electromagnetic transient simulation, data processing and simulation error verification, specifically including the following three aspects:

[0042] (1) In view of the problem of numerous working conditions during simulation testing, the present disclosure uses a cloud platform to implement an electromagnetic transient simulation method, and implements multi-condition parallel simulation calculation and distributed data storage based on a big data framework on the cloud platform to improve simulation efficiency;

[0043] (ii) In order to solve the problem of huge amount of digital model simulation data and test data, multi-threaded computing is implemented on the cloud platform based on the big data framework to quickly realize data processing, simulation comparison chart generation, error calculation, and simulation comparison report generation, thereby improving data processing efficiency;

[0044] (III) Based on the large amount of digital model simulation and test data currently available, the simulation parameter optimization model is implemented in the cloud platform in combination with machine learning and deep learning algorithms. Through historical data training and iteration, the parameter optimization model can automatically optimize and predict, and the optimized and predicted parameters are re-entered into the digital model to perform simulation, so that the calibration error between the digital simulation and the test data meets the preset requirements, reducing the difficulty of error analysis.

[0045] Refer to the following Figures 1 to 12 The simulation parameter optimization method and device of electromagnetic transient simulation according to the present disclosure are specifically described.

[0046] Before describing in detail the simulation parameter optimization method 100 for electromagnetic transient simulation according to an exemplary embodiment of the present disclosure (hereinafter referred to as “simulation parameter optimization method 100 ”), reference is first made to Figures 2 to 7 The overall process of electromagnetic transient simulation applied by the simulation parameter optimization method 100 according to the present disclosure is described.

[0047] Specifically, Figure 2 is an overall flowchart showing the electromagnetic transient simulation according to an exemplary embodiment of the present disclosure, and Figure 7 is an overall flowchart showing an exemplary simulation parameter optimization method according to an exemplary embodiment of the present disclosure.

[0048] Referring to Figure 2 , the functional modules involved in the overall process of the electromagnetic transient simulation according to the present disclosure mainly may include a parallel simulation calculation module, a simulation data processing module, and a simulation parameter optimization module.

[0049] Specifically, for the parallel simulation calculation module, the parallel simulation calculation module may receive various operating condition data involved in each operating condition to be simulated of the wind turbine to be evaluated, as well as the unit-related parameters and data of the wind turbine to be evaluated (as its input data), perform parallel simulation processing on these input data, and output data including at least both the three-phase voltage instantaneous value and the three-phase current instantaneous value (as its output data).

[0050] For example, referring to Figure 3 and Figure 4 , Figure 3 is a simulation flowchart for multiple operating conditions according to a comparative example, and Figure 4 is a parallel simulation flowchart for multiple operating conditions according to an exemplary embodiment of the present disclosure.

[0051] As Figure 3 shown, according to the previous working mode, the simulations are carried out one by one for each operating condition on a single computer, and finally the data is stored on the local computer.

[0052] As Figure 4 shown, the parallel simulation calculation module according to the present disclosure can be implemented by using various cloud platforms. On the cloud platform, based on the big data computing framework cluster, multiple simulation machines can perform multi-operating-condition parallel simulation calculations simultaneously, greatly improving the simulation efficiency and effectively solving the problem that the serial working mode of a single simulation machine consumes a large amount of time.

[0053] Specifically, the parallel simulation process may include the following steps: First, receive the list of all operating conditions to be simulated, and pair the operating condition name + number for each operating condition; perform multi-node deployment on the cloud platform. Assuming the number of deployed nodes is N, it means that there are N simulation machines on the cloud platform performing simulation calculations; evenly distribute the simulation operating condition numbers for each node, and the simulation machines corresponding to the nodes perform parallel simulations according to the corresponding operating conditions in the configuration file. The multiple simulation machines do not affect each other and execute the simulations independently.

[0054] Through the above parallel simulation calculation, the working efficiency is increased by, for example, N times compared with serial simulation. And on the premise of sufficient cloud platform resources, the more nodes are deployed, the higher the simulation efficiency is improved.

[0055] Referring again to Figure 2 , for the simulation data processing module, the simulation data processing module can receive the output data of the parallel simulation calculation module and the measured data of the electromagnetic transient simulation of the wind turbine to be evaluated (for example, it can be the semi-physical test data and / or on-site measured data as described above) (as its input data), perform parallel processing on these input data (for example, error calculation, generation of simulation comparison graphs, generation of simulation reports), and output data including at least the error calculation results (as its output data).

[0056] For example, referring to Figure 5 and Figure 6 , Figure 5 is a flowchart showing data processing for multiple working conditions according to a comparative example, and Figure 6 is a flowchart showing data processing for multiple working conditions according to an exemplary embodiment of the present disclosure.

[0057] As Figure 5 shown, the amount of simulation data and measured data is huge. According to the previous single-simulation machine working mode, the simulation test data for each working condition is processed and calculated one by one on a single computer, simulation comparison graphs are generated, error calculations are performed, and simulation reports are generated. However, due to a large number of simulation working conditions and a large amount of data obtained from the simulation of each working condition, this data processing method is extremely time-consuming and laborious for data processing and error calculation for each working condition.

[0058] As Figure 6 shown, in order to improve the working efficiency, the function of the simulation data processing module is implemented using a big data framework in the cloud platform, so as to achieve multi-threaded calculation and quickly realize parallel data processing, perform simulation data calculation, generation of simulation comparison graphs, error calculation, and generation of simulation comparison reports for each working condition. Through parallel data processing, the efficiency of data processing and various calculations is greatly improved.

[0059] Specifically, similar to the parallel simulation calculation module, the simulation data processing process may include the following steps: First, within the allocated N nodes, perform parallel simulation data calculation, error calculation, and generation of comparison graphs for the corresponding working conditions according to the working condition numbers; then, after all the processes of simulation data calculation, error calculation, and generation of comparison graphs for all working conditions are completed, save all data (including data and comparison graphs, etc.) in the database and sort them according to the working condition numbers; secondly, generate a simulation report according to the data calculation results.

[0060] Through the above parallel data processing method, the working efficiency is also improved by, for example, N times compared with serial data processing.

[0061] According to an embodiment of the present disclosure, the parameters involved in simulation data processing mainly include at least one of the following items: positive-sequence voltage effective value, active power effective value, reactive power effective value, active current effective value, and reactive current effective value. The specific calculation processes of these parameters are as follows:

[0062] Step a), according to the instantaneous values of the three-phase voltages (Ua, Ub, Uc) and three-phase currents (Ia, Ib, Ic) obtained by simulation, calculate the voltage effective value and current effective value. The specific calculation formulas can be referred to as shown in the following formulas (1) to (12):

[0063]

[0064] Among them, Uacos, Uasin, Ubcos, Ubsin, Uccos, and Ucsin respectively represent the cosine components and sine components (i.e., voltage effective values) of the three-phase (i.e., ABC three-phase) voltages, and Iacos, Iasin, Ibcos, Ibsin, Iccos, and Icsin respectively represent the cosine components and sine components (i.e., current effective values) of the three-phase currents. T represents the fundamental wave period, and f represents the fundamental wave frequency.

[0065] Step b), calculate the sine value and cosine value of the positive-sequence voltage and the sine value and cosine value of the positive-sequence current. The specific calculation formulas can be referred to as shown in the following formulas (13) to (16):

[0066]

[0067] Among them, Ucos and Usin respectively represent the cosine component and sine component of the positive-sequence voltage, and Icos and Isin respectively represent the cosine component and sine component of the positive-sequence current.

[0068] Step c), calculate the positive-sequence voltage effective value, active power effective value, and reactive power effective value. The specific calculation formulas can be referred to as shown in the following formulas (17) to (19):

[0069]

[0070] Among them, U represents the positive-sequence voltage effective value, P represents the active power effective value, and Q represents the reactive power effective value.

[0071] Step d), calculate the active current effective value Ip and reactive current effective value Iq. The specific calculation formulas can be referred to as shown in the following formulas (20) to (21):

[0072]

[0073] The RMS positive-sequence voltage, RMS active power, RMS reactive power, RMS active current, and RMS reactive current calculated through the above steps a) to d) are parameters considered for error calculation, and the deviation between the simulated values and the measured values of the above parameters can be used to confirm the accuracy of the simulation model.

[0074] As an example, the above error calculation according to the present disclosure mainly involves calculating the deviation values for the above five parameters (i.e., RMS positive-sequence voltage, RMS active power, RMS reactive power, RMS active current, and RMS reactive current). For example, it may include the steady-state interval average error F1, transient interval average error F2, steady-state interval average absolute error F3, transient interval average absolute error F4, and steady-state interval maximum error for each operating condition. The specific calculation formulas can refer to those shown in the following formulas (22) to (25):

[0075]

[0076] Wherein, Xs and Xm respectively represent the simulation data and measured data of the above five parameters, and Kstart and Kend respectively represent the sequence numbers of the first and last simulation data and measured data during error calculation.

[0077] For example, for the error results represented by formulas (22) to (25), Figure 2 the predetermined criteria shown may be that the above error results for each operating condition need to satisfy: based on the national standard requirements, the steady-state interval average absolute error for all operating conditions does not exceed 5%; and the transient interval average error for all operating conditions does not exceed 20%.

[0078] Referring again to Figure 2 , for the simulation parameter optimization module, in the case where it is determined that the data verification result of the output data output from the simulation data processing module does not meet the predetermined criteria, the simulation parameter optimization module is used to perform data optimization processing on the simulation parameters.

[0079] Specifically, if the error results for all operating conditions meet a predetermined standard (e.g., the predetermined standard described above), there is no need to use the parameter optimization function, and this simulation test work ends; if there are error results for one or more operating conditions that do not meet the predetermined standard, the simulation parameter optimization module is used to optimize the simulation parameters. Specifically, a simulation parameter optimization model is constructed based on machine learning algorithms and deep learning algorithms, and through a large amount of existing historical data training and iteration, the simulation parameter optimization model can automatically optimize and predict. The optimized simulation parameters are re-input into the parallel simulation calculation module and the simulation is performed again, and then the simulation output data is processed by the simulation data processing module until the error results meet the predetermined standard and the simulation test work ends.

[0080] The specific optimization process can refer to the following description of the simulation parameter optimization method 100.

[0081] Figure 1 FIG. is a flowchart showing a simulation parameter optimization method 100 for electromagnetic transient simulation according to an exemplary embodiment of the present disclosure. As an example, a wind farm may include multiple wind turbines (in the present disclosure, may be simply referred to as wind turbines or turbines). The turbines / wind turbines of the wind farm are connected to the power grid in an appropriate manner.

[0082] Refer to Figure 1 , in step S101, simulation data and measured data of the electromagnetic transient simulation of the wind turbine to be evaluated are obtained.

[0083] Here, the simulation data includes input data, output data, and intermediate variables.

[0084] According to an embodiment of the present disclosure, the input data is the input data for the simulation model of the electromagnetic transient simulation, and includes various operating condition data involved in each operating condition to be simulated of the wind turbine to be evaluated, as well as the turbine-related parameters and data of the wind turbine to be evaluated.

[0085] According to an embodiment of the present disclosure, the output data is the output data for the simulation model (hereinafter may be simply referred to as the simulation model) of the electromagnetic transient simulation, and at least includes data of both the three-phase voltage instantaneous value and the three-phase current instantaneous value.

[0086] According to an embodiment of the present disclosure, the intermediate variables may include, for example, at least one of the following variables generated during the simulation of the simulation model: active current set value, reactive current set value, positive sequence voltage, negative sequence voltage, PLL output frequency, PLL output phase, DC bus voltage, low voltage ride-through flag, high voltage ride-through flag, and fault flag.

[0087] For the application scenarios of the present disclosure, the electromagnetic transient simulation according to the embodiments of the present disclosure is used for grid connection assessment and / or model assessment of wind turbines. The electromagnetic transient simulation according to the embodiments of the present disclosure may include high and low voltage ride-through simulation and / or impedance assessment simulation. In addition, the electromagnetic transient simulation according to the embodiments of the present disclosure is capable of performing parallel simulation calculations for multiple working conditions.

[0088] For example, in the scenarios of grid connection assessment of wind farms and new model certification, there are many simulation working conditions. The grid connection simulation tests mainly include high and low voltage ride-through simulation and impedance assessment simulation.

[0089] Specifically, regarding the high and low voltage ride-through simulation, according to the national standard requirements, 40 working conditions need to be simulated for the wind turbine converter, and 81 working conditions need to be simulated for the energy storage inverter. In addition, additional simulation working conditions may also be simulated according to the actual grid conditions. For example, in the case of additionally performing adaptability simulation for different short-circuit ratios, the number of working conditions to be simulated doubles accordingly. The number of simulation working conditions is: the number of short-circuit ratios multiplied by the standard number of working conditions. For example, in the case of needing to perform high and low voltage ride-through adaptability simulation under five short-circuit ratios, the total number of simulation working conditions for the wind turbine converter is 200 (5 multiplied by 40), and the total number of simulation working conditions for the energy storage inverter is 405 (5 multiplied by 81).

[0090] In addition, regarding the impedance assessment simulation, according to the national standard requirements, 30 working conditions need to be simulated for both the wind turbine converter and the energy storage inverter. In addition, additional simulation working conditions may also be simulated according to the actual grid conditions.

[0091] In step S102, based on the input data, output data, and intermediate variables, calculate a first updated simulation parameter corresponding to the simulation parameter of the electromagnetic transient simulation.

[0092] As an example, each of the above-mentioned simulation parameters, first updated simulation parameters, and second updated simulation parameters and optimized simulation parameters to be described later may include at least one of the following parameters for the simulation model: voltage loop proportional coefficient, voltage loop integral coefficient, current loop proportional coefficient, current loop integral coefficient, phase-locked loop proportional coefficient, phase-locked loop integral coefficient, phase-locked loop voltage locking parameter, phase-locked loop voltage unlocking parameter, full-wave voltage feedforward coefficient, fundamental wave voltage feedforward coefficient, low-ride active current coefficient, low-ride reactive current coefficient, and active recovery speed.

[0093] According to an embodiment of the present disclosure, the step of calculating a first updated simulation parameter corresponding to a simulation parameter of an electromagnetic transient simulation may include: step S1021 (not shown), calculating deviation data for a preset parameter based on output data and measured data; step S1022 (not shown), calculating the first updated simulation parameter by inputting input data, deviation data, and an intermediate variable into an LSTM (i.e., Long Short-Term Memory) model.

[0094] Here, the LSTM model is constructed based on reference simulation data and reference measured data of a reference wind turbine unit of the same or similar type as the wind turbine unit to be evaluated.

[0095] For example, the data used to train the LSTM model may be a historical data set, including a data set that has been simulated for a simulation model of the same type and whose error meets a predetermined standard after comparison with measured data, and a simulation parameter table corresponding to the data set.

[0096] As an example, the input data input into the LSTM model may include first feature input data, second feature input data, and key intermediate variable data.

[0097] Specifically, the first feature input data may include: error results for voltage, active power, reactive power, active current, and reactive current (for example, results of the above four errors F1 to F4 for these parameters (i.e., a total of 20 error values)).

[0098] The second feature input data may include grid and transformer parameters, such as grid impedance, grid voltage, transformer short-circuit voltage percentage, short-circuit loss, no-load loss, no-load current percentage, grid voltage dip / raise amplitude.

[0099] The key intermediate variable data is a key intermediate variable related to the control strategy obtained during the calculation process of the simulation model, and includes, for example, active current set value, reactive current set value, positive-sequence voltage, negative-sequence voltage, PLL output frequency, PLL output phase, DC bus voltage, low voltage ride-through flag, high voltage ride-through flag, fault flag.

[0100] In addition, feature engineering can be performed on the first feature input data, the second feature input data, and the key intermediate variable data. Specifically, first, the features involved in the first feature input data, the second feature input data, and the key intermediate variable data are continuous features except for the flag bit-related features. These continuous features can be normalized to obtain the data distribution of each continuous feature, and based on the data distribution, the probability distribution diagram of the feature values in each continuous feature can be obtained. Secondly, for discrete features such as flag bits, one-hot encoding can be performed to obtain the one-hot feature vector of each feature. Finally, based on the above probability distribution and one-hot vector, feature crossing and fusion are performed on all features to obtain new features.

[0101] As an example, the first feature input data, the key intermediate variable data, and the new features (which reflect the second feature input data and its association with other data) can be used as the input vector x input to the LSTM model.

[0102] As an example, the core modeling concept of the LSTM model is the "cell state". The cell state runs through the entire LSTM sequence. The internal structure of each cell state is the same, and inside, the structure of "gates" (including three gates, namely, the forget gate, the input gate, and the output gate) is used to add and delete the information brought by the previous cell state to the cell state at each moment. Specifically, the forget gate is used to determine which information passed from the previous moment to be discarded by the current cell state, the input gate is used to determine which new information to add to the current cell state, and the output gate is used to determine which information to output to the next moment according to the current moment information. The overall processing process of the specific model (steps S1 to S5) can refer to the following equations (26) to (32):

[0103] In step S1, update the forget gate output:

[0104] f (t) = δ(W f * h (t-1) + U f * x t + b f ) (26)

[0105] In step S2, update the two parts of the input gate output:

[0106] i (t) = δ(W i * h (t-1) + U i * x t + b i ) (27)

[0107]

[0108] In step S3, update the cell state:

[0109]

[0110] In step S4, update the output of the output gate:

[0111] O (t) = δ(W0 * h (t-1) + U0 * x t + b0) (30)

[0112] h (t) = O (t) ⊙ tanh(C t ) (31)

[0113] In step S5, update the simulation parameter output predicted by the previous model:

[0114] y = g(V * h (t) + b) (32)

[0115] where x t is the input vector at the current moment, h (t) is the cell state information output at the current moment, h (t-1) is the cell state information input at the previous moment, W f 、U f 、b f 、W i 、U i 、b i 、W a 、U a 、b a 、W0, U0, b0, V, b are the LSTM model parameters obtained through training iterations. y is the output value of the LSTM model finally predicted by the LSTM model.

[0116] As an example, the parameters of the LSTM model can be trained using the backpropagation algorithm. The specific training process is as follows: First, calculate the output value of each neuron through forward calculation, and calculate respectively according to the above method in the calculation process of the LSTM model; then, determine the optimization objective function, calculate the error threshold of each neuron, and construct the loss function; finally, update all model parameters (i.e., W f 、U f 、b f 、W i 、U i 、b i 、W a 、U a 、b a 、W0, U0, b0, V, b) according to the gradient of the loss function.

[0117] The output value of the LSTM model according to the present disclosure is a simulation parameter vector, specifically including the key simulation parameters of the predicted simulation model. For example, the voltage loop proportional and integral coefficients, the current loop proportional and integral coefficients, the phase-locked loop proportional and integral coefficients, the phase-locked loop voltage locking and unlocking parameters, the full-wave and fundamental voltage feed-forward coefficients, the low-ride-through active and reactive current coefficients, the active power recovery speed, etc. The above parameters are all key parameters affecting the simulation results.

[0118] By adopting the LSTM model, since the simulation data as the input is sequential data corresponding to time and values, the LSTM model, as a special RNN (i.e., Recurrent Neural Network, recurrent neural network) model, can effectively alleviate the "long-term dependence" characteristic of sequential data, thereby effectively ensuring the information integrity of the simulation data.

[0119] In step S103, based on the input data, output data, intermediate variables, and simulation parameters, calculate the second updated simulation parameters corresponding to the simulation parameters;

[0120] According to the embodiments of the present disclosure, the step of calculating the second updated simulation parameters corresponding to the simulation parameters may include: inputting the input data, output data, intermediate variables, and simulation parameters into an XGBoost (i.e., eXtreme Gradient Boosting, extreme gradient boosting tree) model to calculate the second updated simulation parameters.

[0121] Here, the XGBoost model is constructed based on the reference simulation data and reference measured data of a reference wind turbine with the same or similar model as the wind turbine to be evaluated.

[0122] For example, the data used to train the XGBoost model can be a dataset similar to the historical dataset used to train the LSTM model above, which will not be elaborated here.

[0123] As an example, the input data input into the XGBoost model may include the third feature input data, the fourth feature input data, the key intermediate variable data, and the simulation parameters used by the current simulation model.

[0124] Specifically, the third feature input data may include: the voltage, active power, reactive power, active current, and reactive current calculated based on the output data of the simulation model.

[0125] The fourth feature input data may be the same as the second feature input data, which will not be elaborated here. In addition, the key intermediate variable data here is the same as the key intermediate variable data described for the LSTM model, which will not be elaborated here.

[0126] The simulation parameters used in the current simulation model can be the reference key simulation parameters included in the historical dataset, and may include, for example: voltage loop proportional and integral coefficients, current loop proportional and integral coefficients, phase-locked loop proportional and integral coefficients, phase-locked loop voltage locking and unlocking parameters, full-wave and fundamental voltage feedforward coefficients, low-ride active and reactive current coefficients, active recovery speed, etc.

[0127] In addition, feature engineering processing similar to that involved in the feature engineering described in the LSTM model can also be performed on the third feature input data, the fourth feature input data, the key intermediate variable data, and the simulation parameters used in the current simulation model to obtain new features. Specifically, except for also processing the third feature input data and the simulation parameters used in the current simulation model as continuous features, other processing is similar to the corresponding processing in the feature engineering involved in the LSTM model, which will not be elaborated here.

[0128] As an example, the third feature input data (and / or its error result), the key intermediate variable data, and the new features can be used as the input vector x of each sample input to the XGBoost model i .

[0129] As an example, the prediction process of the XGBoost model is achieved through the following objective function:

[0130]

[0131] where, f k (x i ) represents the result of the kth tree in the model predicting x i , T represents the number of leaf nodes in the tree model, w j represents the value on the leaf node, g i represents the first derivative value of the loss function of the difference between the actual simulation value yi and the predicted values of the previous k - 1 trees, hi is the second derivative value, and γ and λ are regularization parameters.

[0132] In addition, Obj is the model objective function, and represents the difference between the simulation data and the measured data. The smaller this objective function, the more accurate the comparison result between the simulation data and the measured data, and the better the performance of the XGBoost model.

[0133] The output value of the XGBoost model according to the present disclosure is the simulation parameter vector y, and the specific parameters included are the same as those included in the output value described for the LSTM model, which will not be elaborated here.

[0134] The XGBoost model adopted in this disclosure is a large-scale parallel boosting tree algorithm and a model ensemble based on decision trees. Through the parallel computing method adopted in the training of this model, high training efficiency can be achieved, and by flexibly adjusting the controllable parameters in this model, a better actual prediction effect can be obtained.

[0135] According to an embodiment of the present disclosure, the step of calculating the second updated simulation parameter corresponding to the simulation parameter may include: calculating the second updated simulation parameter by inputting the input data, output data, intermediate variables, and simulation parameter into a RandomForest (i.e., random forest) model, where the RandomForest model is constructed based on reference simulation data and reference measured data of a reference wind turbine whose model is the same as or similar to the model of the wind turbine to be evaluated.

[0136] For example, the data used to train the RandomForest model can be a dataset similar to the historical dataset used to train the LSTM model described above, which will not be elaborated here. In addition, the input and output data of the RandomForest model are the same as the input and output data of the XGBoost model respectively, which will not be elaborated here.

[0137] Regarding the prediction process of the RandomForest model, RandomForest is also an ensemble model based on decision trees. The biggest difference from the XGBoost model is that: when training the model, both the training samples and feature selection are randomized. Multiple trees are trained simultaneously by randomizing the sample and feature selection, and the average value of the training results of multiple trees is taken to obtain the final predicted simulation model parameters.

[0138] As an example, the updated simulation parameters obtained according to the RandomForest model and the XGBoost model can be combined in a predetermined manner as the second updated simulation parameter (which will be specifically elaborated later).

[0139] In addition, when calculating the first updated simulation parameter and / or the second updated simulation parameter, other machine learning algorithms or deep learning algorithms can be used to replace or be added as new models to the simulation parameter optimization method.

[0140] For example, other machine learning algorithms and deep learning algorithms (e.g., LightGBM (i.e., Light Gradient Boosting Machine) model) can also be used to replace or be added as a new model to the simulation parameter optimization method. For example, the updated simulation parameters obtained from the RandomForest model, XGBoost model, and LightGBM model combined in the same way as the above-mentioned predetermined way can be used as the second updated simulation parameters.

[0141] In step S104, based on the first updated simulation parameters and the second updated simulation parameters, the final optimized simulation parameters are calculated.

[0142] As an example, the step of calculating the final optimized simulation parameters may include: inputting the first updated simulation parameters and the second updated simulation parameters into a fully connected neural network model based on the Attention mechanism to calculate the first weight associated with the first updated simulation parameters, and the second weight and the third weight associated with the second updated simulation parameters; calculating the final optimized simulation parameters based on the first updated simulation parameters, the second updated simulation parameters, the first weight, the second weight, and the third weight.

[0143] For example, the first weight corresponds to the first updated simulation parameters calculated by the LSTM model, the second weight corresponds to the second updated simulation parameters calculated by the XGBoost model, and the third weight corresponds to the second updated simulation parameters calculated by the RandomForest model.

[0144] Specifically, the above-mentioned fully connected neural network can be a two-layer fully connected neural network. The input of this neural network is the parameter input vector [yl, yx, yr] synthesized based on the prediction results of the above three models. After being calculated by the neural network, the output of the output layer is the vector [al, ax, ar], and the elements in this vector are values within the range of 0 to 1, that is, the importance (e.g., weight) of the prediction results of the three simulation parameter prediction models this time.

[0145] The final output result Y of the above-mentioned fully connected neural network is shown in the following formula (34):

[0146] Y = yl * al + yx * ax + yr * ar (34)

[0147] Among them, yl, yx, and yr respectively represent the simulation parameter results calculated by the LSTM, XGBoost, and RandomForest models, and al, ax, and ar respectively represent the weights corresponding to the LSTM, XGBoost, and RandomForest models calculated by the Attention mechanism.

[0148] Figure 8 is a flowchart showing an example simulation parameter optimization method according to an exemplary embodiment of the present disclosure. Referring to Figure 8 , the example simulation parameter optimization method predicts simulation parameters based on three models: LSTM, XGBoost, and RandomForest, and combines the Attention mechanism to perform weighted average processing on the prediction results of the three models to obtain the final simulation parameters predicted by the simulation parameter optimization method. The final simulation parameters are re-input into the simulation model for new simulation processing. Here, the flow of the example simulation parameter optimization method is as Figure 8 shown and will not be elaborated here.

[0149] Next, referring to Figures 9A to 10E the beneficial effects of the simulation parameter optimization method of the present disclosure will be described. Figures 9A to 9E is a simulation result graph of a comparative example that does not use the simulation parameter optimization method of the present disclosure, and Figures 10A to 10E is a simulation result graph of an example of the present disclosure that uses the simulation parameter optimization method of the present disclosure.

[0150] From Figures 9A to 10E the simulation result graph, it can be seen that the verification results for various parameters are more accurate. That is, the simulation results of the example of the present disclosure using the simulation parameter optimization method of the present disclosure are significantly better than the simulation results of the comparative example that does not use the simulation parameter optimization method of the present disclosure.

[0151] By adopting the simulation parameter optimization method and device for electromagnetic transient simulation of the present disclosure, the verification accuracy between the simulation model and the measured data can be improved. For example, the verification accuracy can reach that the average absolute error in the steady-state interval under all conditions of high and low voltage crossing is within 1%, and the average error in the transient interval under all conditions is within 5%. Such errors are respectively much lower than the standards of the average absolute error within 5% in the steady-state interval under all conditions and the average error within 20% in the transient interval under all conditions based on national standards.

[0152] In addition, by adopting the simulation parameter optimization method and device for electromagnetic transient simulation of the present disclosure, the verification accuracy and calculation accuracy of the electromagnetic transient simulation for wind turbines are improved, and the difficulty of error analysis and debugging the simulation model parameters is reduced.

[0153] In addition, through the simulation parameter optimization method and device for electromagnetic transient simulation of the present disclosure, it is also possible to effectively shorten the time for simulation and data processing using the electromagnetic transient simulation model (for example, the time for generating comparison graphs, calculating errors, generating reports, etc.), and effectively improve the simulation efficiency. In addition, through the simulation parameter optimization method and device for electromagnetic transient simulation of the present disclosure, it is also possible to perform parallel simulation processing for multiple working conditions, further improving the simulation efficiency, thereby greatly reducing the hardware and human resource requirements for wind power electromagnetic transient simulation and evaluation.

[0154] In addition, through the simulation parameter optimization method and device for electromagnetic transient simulation of the present disclosure, it is also possible to quickly support the grid connection of wind farms and the certification of new machine models, enabling wind farms to be connected to the grid for power generation in a timely manner, helping to avoid losses in relevant revenues, and also helping new machine model units to pass the certification and be put into subsequent use as soon as possible.

[0155] Figure 11 FIG. is a block diagram showing a simulation parameter optimization device 1100 for electromagnetic transient simulation according to an exemplary embodiment of the present disclosure.

[0156] The electromagnetic transient simulation according to an embodiment of the present disclosure is used for grid connection assessment and / or model assessment of wind turbine units. As an example, the electromagnetic transient simulation may include high and low voltage ride-through simulation and / or impedance assessment simulation, and the electromagnetic transient simulation is capable of performing parallel simulation calculations for multiple working conditions.

[0157] Referring to Figure 11 , the simulation parameter optimization device 1100 for electromagnetic transient simulation includes: an acquisition unit 1110, a first parameter optimization unit 1120, a second parameter optimization unit 1130, and a third parameter optimization unit 1140.

[0158] According to an embodiment of the present disclosure, the acquisition unit 1110 is configured to: acquire simulation data and measured data of the electromagnetic transient simulation of the wind turbine unit to be evaluated.

[0159] Here, the simulation data includes input data, output data, and intermediate variables. As an example, the intermediate variables here may include at least one of the following variables: active current set value, reactive current set value, positive sequence voltage, negative sequence voltage, PLL output frequency, PLL output phase, DC bus voltage, low voltage ride-through flag, high voltage ride-through flag, and fault flag.

[0160] According to an embodiment of the present disclosure, the first parameter optimization unit 1120 is configured to: calculate first updated simulation parameters corresponding to the simulation parameters of the electromagnetic transient simulation based on the input data, output data, and intermediate variables.

[0161] As an example, the operation of the first parameter optimization unit 1120 to calculate the first updated simulation parameter corresponding to the simulation parameter of the electromagnetic transient simulation may include: calculating deviation data for a preset parameter based on the output data and the measured data; calculating the first updated simulation parameter by inputting the input data, the deviation data, and the intermediate variable into the LSTM model. Here, the LSTM model is constructed based on the reference simulation data and the reference measured data of a reference wind turbine unit that is the same as or similar to the model of the wind turbine unit to be evaluated.

[0162] According to an embodiment of the present disclosure, the second parameter optimization unit 1130 is configured to: calculate a second updated simulation parameter corresponding to the simulation parameter based on the input data, the output data, the intermediate variable, and the simulation parameter.

[0163] As an example, the operation of the second parameter optimization unit 1130 to calculate the second updated simulation parameter corresponding to the simulation parameter may include: calculating the second updated simulation parameter by inputting the input data, the output data, the intermediate variable, and the simulation parameter into the XGBoost model. Here, the XGBoost model is constructed based on the reference simulation data and the reference measured data of a reference wind turbine unit that is the same as or similar to the model of the wind turbine unit to be evaluated.

[0164] As an example, the operation of the second parameter optimization unit 1130 to calculate the second updated simulation parameter corresponding to the simulation parameter may include: calculating the second updated simulation parameter by inputting the input data, the output data, the intermediate variable, and the simulation parameter into the RandomForest model. Here, the RandomForest model is constructed based on the reference simulation data and the reference measured data of a reference wind turbine unit that is the same as or similar to the model of the wind turbine unit to be evaluated.

[0165] According to an embodiment of the present disclosure, the third parameter optimization unit 1140 is configured to: calculate the final optimized simulation parameter based on the first updated simulation parameter and the second updated simulation parameter.

[0166] As an example, the operation of the third parameter optimization unit 1140 to calculate the final optimized simulation parameter may include: calculating a first weight associated with the first updated simulation parameter, a second weight, and a third weight associated with the second updated simulation parameter by inputting the first updated simulation parameter and the second updated simulation parameter into a fully connected neural network model based on an attention mechanism; calculating the final optimized simulation parameter based on the first updated simulation parameter, the second updated simulation parameter, the first weight, the second weight, and the third weight.

[0167] Here, the first weight corresponds to the first updated simulation parameter calculated by the LSTM model, the second weight corresponds to the second updated simulation parameter calculated by the XGBoost model, and the third weight corresponds to the second updated simulation parameter calculated by the RandomForest model.

[0168] As an example, each of the above-mentioned simulation parameter, first updated simulation parameter, second updated simulation parameter, and optimized simulation parameter may include at least one of the following parameters: voltage loop proportional coefficient, voltage loop integral coefficient, current loop proportional coefficient, current loop integral coefficient, phase-locked loop proportional coefficient, phase-locked loop integral coefficient, phase-locked loop voltage locking parameter, phase-locked loop voltage unlocking parameter, full-wave voltage feedforward coefficient, fundamental voltage feedforward coefficient, low-ride-through active current coefficient, low-ride-through reactive current coefficient, and active power recovery speed.

[0169] It should be understood that the specific processing performed by the simulation parameter optimization device for electromagnetic transient simulation according to the exemplary embodiments of the present disclosure has been described in detail with reference to Figures 1 to 1 0 and will not be elaborated here.

[0170] It should be understood that each unit in the simulation parameter optimization device for electromagnetic transient simulation according to the exemplary embodiments of the present disclosure can be implemented as a hardware component and / or a software component. Those skilled in the art can implement each unit using, for example, a field-programmable gate array (FPGA) or an application-specific integrated circuit (ASIC) according to the processing performed by the defined units.

[0171] The exemplary embodiments of the present disclosure provide a computer-readable storage medium. When the instructions in the computer-readable storage medium are run by at least one processor, at least one processor is caused to execute the simulation parameter optimization method as described above. The computer-readable storage medium is any data storage device that can store data read by a computer system. Examples of computer-readable storage media include: read-only memory, random access memory, compact disc read-only memory, magnetic tape, floppy disk, optical data storage device, and carrier waves (such as data transmission via the Internet through a wired or wireless transmission path).

[0172] Figure 12 is a block diagram showing a computer device 1200 according to an exemplary embodiment of the present disclosure.

[0173] The computer device 1200 according to the exemplary embodiments of the present disclosure includes: at least one processor 1210 and at least one memory 1220. The memory 1220 stores computer-executable instructions. When the computer-executable instructions are run by at least one processor 1210, at least one processor 1210 is caused to execute the simulation parameter optimization method as described above.

[0174] Although some exemplary embodiments of the present disclosure have been shown and described, those skilled in the art should understand that these embodiments can be modified without departing from the principles and spirit of the present disclosure, which is defined by the claims and their equivalents.

Claims

1. A method for optimizing simulation parameters of electromagnetic transient simulation, characterized in that, The electromagnetic transient simulation is used for grid connection assessment and / or model assessment of wind turbines. The simulation parameter optimization method includes: Obtain the simulation data and measured data of the electromagnetic transient simulation of the wind turbine to be evaluated. Among them, the simulation data includes input data, output data, and intermediate variables. Based on the input data, the output data, and the intermediate variables, calculate the first updated simulation parameters corresponding to the simulation parameters of the electromagnetic transient simulation. Based on the input data, the output data, the intermediate variables, and the simulation parameters, calculate the second updated simulation parameters corresponding to the simulation parameters. Based on the first updated simulation parameters and the second updated simulation parameters, calculate the final optimized simulation parameters.

2. The simulation parameter optimization method according to claim 1, wherein The step of calculating the first updated simulation parameters corresponding to the simulation parameters of the electromagnetic transient simulation includes: Based on the output data and the measured data, calculate the deviation data for the preset parameters. By inputting the input data, the deviation data, and the intermediate variables into the LSTM model, calculate the first updated simulation parameters. Among them, the LSTM model is constructed based on the reference simulation data and reference measured data of the reference wind turbine with the same or similar model as the wind turbine to be evaluated.

3. The simulation parameter optimization method according to claim 1, wherein The step of calculating the second updated simulation parameters corresponding to the simulation parameters includes: By inputting the input data, the output data, the intermediate variables, and the simulation parameters into the XGBoost model, calculate the second updated simulation parameters. Among them, the XGBoost model is constructed based on the reference simulation data and reference measured data of the reference wind turbine with the same or similar model as the wind turbine to be evaluated.

4. The simulation parameter optimization method according to claim 1, characterized in that The step of calculating the second updated simulation parameters corresponding to the simulation parameters includes: By inputting the input data, the output data, the intermediate variables, and the simulation parameters into the RandomForest model, calculate the second updated simulation parameters. Among them, the RandomForest model is constructed based on the reference simulation data and reference measured data of the reference wind turbine with the same or similar model as the wind turbine to be evaluated.

5. The simulation parameter optimization method according to claim 1, characterized in that The step of calculating the final optimized simulation parameters includes: By inputting the first updated simulation parameters and the second updated simulation parameters into the fully connected neural network model based on the attention mechanism, calculate the first weight associated with the first updated simulation parameters, the second weight and the third weight associated with the second updated simulation parameters. Among them, the first weight corresponds to the first updated simulation parameters calculated by the LSTM model, the second weight corresponds to the second updated simulation parameters calculated by the XGBoost model, and the third weight corresponds to the second updated simulation parameters calculated by the RandomForest model. Based on the first updated simulation parameters, the second updated simulation parameters, the first weight, the second weight, and the third weight, calculate the final optimized simulation parameters.

6. The simulation parameter optimization method according to claim 1, wherein The electromagnetic transient simulation includes a low-high voltage ride-through simulation and / or an impedance evaluation simulation, and the electromagnetic transient simulation is capable of performing parallel simulation calculations for multiple operating conditions.

7. The simulation parameter optimization method according to claim 1, characterized in that The simulation parameters, the first updated simulation parameters, the second updated simulation parameters, and the optimized simulation parameters each include at least one of the following parameters: voltage loop proportional coefficient, voltage loop integral coefficient, current loop proportional coefficient, current loop integral coefficient, phase-locked loop proportional coefficient, phase-locked loop integral coefficient, phase-locked loop voltage locking parameter, phase-locked loop voltage unlocking parameter, full-wave voltage feedforward coefficient, fundamental voltage feedforward coefficient, low-ride-through active current coefficient, low-ride-through reactive current coefficient, and active power recovery speed.

8. The simulation parameter optimization method according to claim 1, characterized in that, The intermediate variables include at least one of the following variables: active current set value, reactive current set value, positive-sequence voltage, negative-sequence voltage, phase-locked loop output frequency, phase-locked loop output phase, DC bus voltage, low voltage ride-through flag, high voltage ride-through flag, and fault flag.

9. A simulation parameter optimization device for electromagnetic transient simulation, characterized in that The electromagnetic transient simulation is used for grid connection evaluation and / or model evaluation of a wind turbine generator set, and the simulation parameter optimization device includes: An acquisition unit, configured to: acquire simulation data and measured data of the electromagnetic transient simulation of the wind turbine generator set to be evaluated, where the simulation data includes input data, output data, and intermediate variables; A first parameter optimization unit, configured to: calculate first updated simulation parameters corresponding to the simulation parameters of the electromagnetic transient simulation based on the input data, the output data, and the intermediate variables; A second parameter optimization unit, configured to: calculate second updated simulation parameters corresponding to the simulation parameters based on the input data, the output data, the intermediate variables, and the simulation parameters; A third parameter optimization unit, configured to: calculate final optimized simulation parameters based on the first updated simulation parameters and the second updated simulation parameters.

10. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are run by at least one processor, the at least one processor is caused to execute the simulation parameter optimization method according to any one of claims 1 to 8.

11. A computer device, characterized in that, Comprising: At least one processor; At least one memory storing computer-executable instructions, wherein, when the computer-executable instructions are run by the at least one processor, the at least one processor is caused to execute the simulation parameter optimization method according to any one of claims 1 to 8.

Images