New energy system transient stability evaluation method and system based on machine learning and data driving

Abstract

The invention discloses a new energy system transient stability evaluation method and system based on machine learning and data driving. Comprising the following steps: performing time domain simulation on an electromechanical transient model and an electromagnetic transient model containing a new energy system to obtain a sample set; and constructing a system stability classification model based on a long short-term memory network, training the system stability classification model based on a data driving method, and realizing evaluation of electromagnetic transient stability through an electromechanical transient simulation monitoring curve. According to the method, the mapping relation between the electromechanical time scale and the electromagnetic time scale is established through data driving, the system stability under the electromagnetic transient scale is predicted only based on the system operation condition under the electromechanical transient scale, and the problems that the electromechanical transient modeling precision of a high-proportion new energy system is low, a real instability scene cannot be accurately reflected, and the modeling precision is low are solved. And electromagnetic transient simulation is slow in speed and low in efficiency, and transient stability evaluation of the new energy system is realized.

Description

Technical Field

[0001] The present invention belongs to the field of energy technology, and relates to a method and system for evaluating electromechanical-electromagnetic transient stability of an electric power system, and in particular to a method and system for evaluating transient stability of a new energy system based on machine learning and data-driven. Background Art

[0002] As the proportion of new energy such as wind power and photovoltaics in the power system continues to increase, the connection of a large number of power electronic equipment has greatly increased the scale and complexity of the new energy power system. The component composition and dynamic process of the system are becoming increasingly complex, which puts higher requirements on the modeling accuracy and computing power of the simulation components.

[0003] For power systems with a high proportion of new energy (referred to as new energy systems), traditional electromechanical transient simulation mainly reflects the operating conditions of the system near the power frequency, and cannot simulate the high-frequency characteristics of power electronic equipment and its control loop, and it is difficult to meet the needs of system transient stability analysis; although electromagnetic transient simulation can more accurately reflect the dynamic characteristics of power electronic equipment, the nonlinearity and high-order terms of the electromagnetic model greatly increase the amount of simulation calculations and calculation time, making it difficult to apply to real-time simulation of large-scale power systems. In order to solve the dual problems of low accuracy of electromechanical transient modeling of high-proportion new energy systems, which cannot accurately reflect the real instability scenario, and slow and inefficient electromagnetic transient simulation, the present invention discloses a transient stability assessment method and system for new energy systems based on machine learning and data-driven. Summary of the invention

[0004] To solve the above problems, the present invention proposes a method and system for transient stability assessment of new energy systems based on machine learning and data-driven to solve the problem of transient stability assessment of new energy systems with a high proportion.

[0005] The technical solution adopted by the present invention is as follows:

[0006] A method for transient stability assessment of a new energy system based on machine learning and data-driven includes the following steps:

[0007] Conduct time-domain simulations on the electromechanical transient model and electromagnetic transient model of the power system with a high proportion of new energy, design typical operating conditions and fault scenarios, and build a trusted sample set for electromechanical and electromagnetic transient time-domain simulations;

[0008] A system stability classification model based on long short-term memory network (LSTM) is constructed. The model input is the electromechanical transient simulation monitoring curve, and the model output is the electromagnetic transient simulation stability classification probability value.

[0009] The electromechanical and electromagnetic transient time-domain simulation credible sample set is preprocessed, and the system stability classification model is trained based on a data-driven method using the preprocessed sample set to achieve the evaluation of electromagnetic transient stability through the electromechanical transient simulation monitoring curve.

[0010] Furthermore, the electromagnetic transient model adopts a three-phase model, in which the computing element model is described by the abc three-phase instantaneous values, and the model includes detailed modeling of the phase-locked loop control structure and the power electronic converter; the electromechanical transient model adopts a phasor model, in which the computing element model is described by the fundamental phasor, and the model ignores the fast transient processes of the converter switching dynamics and the phase-locked loop dynamics.

[0011] Furthermore, the design of typical operating conditions and fault scenarios and the construction of a trusted sample set for time domain simulation specifically include: considering the system load rate and new energy output to design a typical operating condition set for the new energy system; considering the fault type and fault location to design a typical fault set for the new energy system; performing electromechanical transient and electromagnetic transient simulations on the new energy system respectively, traversing all feasible combinations of the typical operating condition set and the typical fault set, and obtaining an electromechanical transient simulation monitoring curve sample set and an electromagnetic transient simulation stability result sample set.

[0012] Furthermore, the system stability classification model includes an LSTM network formed by connecting multiple LSTM units through a chain structure.

[0013] Furthermore, the electromechanical and electromagnetic transient time domain simulation credible sample set is preprocessed, specifically:

[0014] The electromechanical transient simulation curve samples are cut, data for a period of time before and after the fault is retained, and normalization is performed; the electromagnetic transient simulation stability result samples are converted into Boolean type, in which the system stability result corresponds to 0 and the system instability result corresponds to 1.

[0015] Furthermore, the system stability classification model is trained based on a data-driven method using the preprocessed sample set, specifically:

[0016] During the forward propagation process, the system stability features in the input electromechanical transient simulation monitoring curve are extracted through the LSTM network, and the stability features are mapped into a one-dimensional classification probability value through the linear transformation of the fully connected layer. The fully connected layer uses a sigmoid activation function to make the classification probability value between 0 and 1;

[0017] During the back-propagation process, the Adam optimizer with weight decay is used to calculate the classification error according to the binary cross entropy (BCELoss) loss function. The classification error is fed back based on the gradient descent back-propagation algorithm to calculate the parameter gradient and update the model parameters.

[0018] Furthermore, the Adam optimizer that introduces weight decay specifically includes: using an adaptive Adam optimizer to update model parameters, adding a weight decay term when calculating gradients, adjusting the learning rate of each parameter by calculating the first-order moment estimate and the second-order moment estimate of the gradient, calculating the update amount of each parameter according to the learning rate, and applying the update amount to update the parameters.

[0019] Furthermore, the binary cross entropy loss function is:

[0020] L = -ylog(y p )-(1-y)log(1-y p )

[0021] Among them, L is the classification error of the model, y is the actual electromagnetic transient simulation stability result, and y p It is the stable classification probability value of the electromagnetic transient simulation output by the model.

[0022] Furthermore, the output of the system stability classification model is a classification probability value between 0 and 1, and the classification probability value is compared with a preset threshold. When the classification probability value is lower than the threshold, the evaluation result is that the system is stable, and when it is higher than the threshold, the evaluation result is that the system is unstable.

[0023] A transient stability assessment system for new energy systems based on machine learning and data-driven, comprising:

[0024] Data acquisition module: used to perform time domain simulation on the electromechanical transient model and electromagnetic transient model of the power system containing a high proportion of new energy, design typical operating conditions and fault scenarios, and build a trusted sample set for electromechanical and electromagnetic transient time domain simulation;

[0025] Model building module: used to build a system stability classification model based on long short-term memory network. The model input is the electromechanical transient simulation monitoring curve, and the model output is the electromagnetic transient simulation stability classification probability value;

[0026] Model training module: used to preprocess the trusted sample set of electromechanical and electromagnetic transient time domain simulation, and use the preprocessed sample set to train the system stability classification model based on the data-driven method, so as to realize the evaluation of electromagnetic transient stability through the electromechanical transient simulation monitoring curve.

[0027] The beneficial effects of the present invention are:

[0028] The present invention constructs a system stability classification model based on machine learning, establishes a mapping relationship between electromechanical and electromagnetic time scales through data-driven, and realizes the prediction of system stability under the electromagnetic transient scale based only on the system operation under the electromechanical transient scale, thereby solving the dual problems of low accuracy of electromechanical transient modeling of high-proportion new energy systems, which cannot accurately reflect the real instability scene, and slow and inefficient electromagnetic transient simulation, and realizes transient stability assessment of new energy systems. The present invention can provide a basis for the analysis of the physical mechanism of transient stability and dominant instability characteristics of high-proportion new energy systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a flow chart of system stability classification model training and prediction in an embodiment of the present invention.

[0030] Figure 2 LSTM unit structure diagram in an embodiment of the present invention.

[0031] Figure 3 This is a preprocessing flow chart in an embodiment of the present invention.

[0032] Figure 4 This is a topological diagram of a new energy system in an embodiment of the present invention.

[0033] Figure 5 It is an error curve diagram of the model training process in an embodiment of the present invention. DETAILED DESCRIPTION

[0034] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0035] A method for transient stability assessment of a new energy system based on machine learning and data-driven includes the following steps:

[0036] (1) Perform time domain simulation on the electromechanical transient model and electromagnetic transient model of the power system containing a high proportion of renewable energy, design typical operating conditions and fault scenarios, and construct a credible sample set for electromechanical and electromagnetic transient time domain simulation.

[0037] The electromagnetic transient model adopts a three-phase model, in which the calculation element model is described by the abc three-phase instantaneous values, and the model includes detailed modeling of control structures such as phase-locked loops and power electronic converters; the electromechanical transient model adopts a phasor model, in which the calculation element model is described by fundamental wave phasors, and the model ignores fast transient processes such as converter switch dynamics and phase-locked loop dynamics.

[0038] The typical working conditions and fault scenarios of the design are used to construct a trusted sample set for time domain simulation, specifically including:

[0039] 1) Considering factors such as system load rate and new energy output, design a typical operating condition set for the new energy system;

[0040] 2) Considering factors such as fault type and fault location, design a typical fault set for new energy systems;

[0041] 3) Perform electromechanical transient and electromagnetic transient simulations on the new energy system respectively, traverse all feasible combinations of typical operating conditions and typical fault sets, and obtain the electromechanical transient simulation monitoring curve sample set and the electromagnetic transient simulation stability result sample set.

[0042] The specific steps to obtain the electromechanical transient simulation monitoring curve include:

[0043] Based on the electromechanical transient model of the new energy system, run the electromechanical transient simulation; record the system operation data such as power angle, voltage, frequency, etc. at each moment as the monitoring curve. The electromechanical transient simulation monitoring curve includes but is not limited to: the maximum power angle difference of the generator, the minimum / maximum bus voltage of the system, the minimum / maximum bus frequency of the system, etc.

[0044] The specific steps to obtain stable results of electromagnetic transient simulation include:

[0045] Based on the electromagnetic transient model of the new energy system, the electromagnetic transient simulation is run to determine whether the new energy system can maintain stable operation after being disturbed by a fault. The stability criteria used include power angle, voltage and frequency criteria. If any instability criterion is triggered, the system is judged to be unstable, otherwise the system is stable.

[0046] Specifically, the power angle instability criterion refers to the maximum power angle difference of the generator exceeding a certain angle; the voltage instability criterion refers to the system voltage not recovering to the allowable range after the fault; the frequency instability criterion refers to the system frequency not recovering to the allowable range after the fault. In addition, the system stability criterion can also include the new energy disconnection or repeated low voltage ride-through criterion, commutation failure criterion, DC blocking criterion, etc.

[0047] (2) A system stability classification model based on the long short-term memory network (LSTM) is constructed. The model input is the electromechanical transient simulation monitoring curve X = [x(1),…,x(T)], and the model output is the electromagnetic transient simulation stability classification probability value. p ∈(0,1) represents.

[0048] The system stability classification model includes an LSTM network formed by connecting multiple LSTM units through a chain structure. The LSTM unit structure diagram is shown in Figure 2 As shown, by introducing the forget gate, input gate and output gate, the long-term dependency and short-term memory problems are effectively handled. It is suitable for processing time series data, and the weight parameters and bias of the LSTM network are determined through model training.

[0049] (3) Preprocess the credible sample set of electromechanical and electromagnetic transient time-domain simulation, and use the preprocessed sample set to train the system stability classification model based on the data-driven method to evaluate the electromagnetic transient stability through the electromechanical transient simulation monitoring curve. The model training and prediction flow chart is as follows: Figure 1 shown.

[0050] like Figure 3 As shown, the preprocessing of the electromechanical and electromagnetic transient time domain simulation credible sample set is specifically as follows:

[0051] 1) Cut the electromechanical transient simulation curve sample to retain only the data for a period of time before and after the fault; for multiple monitoring curves in the same sample, select one or several of them as input;

[0052] 2) Converting the electromagnetic transient simulation stability result sample into a Boolean type, wherein the system stability result in the electromagnetic transient simulation stability result corresponds to 0 (False), and the system instability result corresponds to 1 (True);

[0053] 3) Normalization: Use standardization or maximum and minimum value normalization methods to normalize the samples; electromagnetic transient samples do not need to be normalized because they have been converted into Boolean type;

[0054] 4) Sample set division: randomly shuffle the samples and divide them into training set, test set and validation set in proportion;

[0055] 5) Data type conversion: Depending on the machine learning environment (using CPU or GPU), the data type is converted to CPU tensor or GPU tensor.

[0056] The method of using the preprocessed sample set to train the system stability classification model based on a data-driven method is specifically as follows:

[0057] During the forward propagation process, the system stability characteristics in the input electromechanical transient simulation monitoring curve are extracted through the LSTM network. The stability feature is mapped into a one-dimensional classification probability value through the linear transformation of the fully connected layer, and the fully connected layer adopts the sigmoid activation function so that the classification probability value Y p Between 0 and 1;

[0058] During the back-propagation process, the Adam optimizer with weight decay is introduced. The classification error is calculated based on the binary cross entropy BCELoss (Binary Cross Entropy Loss) loss function. The classification error is fed back based on the gradient descent back-propagation algorithm to calculate the parameter gradient and update the model parameters.

[0059] The Adam (Adaptive momentum) optimizer with weight decay specifically includes: using an adaptive Adam optimizer to update model parameters, adding a weight decay term when calculating the gradient, adjusting the learning rate of each parameter by calculating the first-order moment estimate and the second-order moment estimate of the gradient, calculating the update amount of each parameter according to the learning rate, and applying the update amount to update the model parameters. The adaptive learning rate makes the Adam optimizer more flexible in handling different parameter updates; adding the weight decay term to the gradient of the Adam optimizer enhances the generalization ability of the model and solves the overfitting problem.

[0060] The binary cross entropy loss function is used to evaluate the difference between the model output and the actual electromagnetic simulation results. The calculation formula is:

[0061] L = -ylog(y p )-(1-y)log(1-y p )

[0062] Among them, L is the classification error of the model, y is the actual electromagnetic transient simulation stability result, and y p It is the stable classification probability value of the electromagnetic transient simulation output by the model.

[0063] After the model is trained, the trained system stability classification model is used to predict the stability results of the current system under the electromagnetic transient scale, including:

[0064] 1) Input the electromechanical transient simulation monitoring curve to be evaluated into the trained system stability classification model, extract features through the n-layer LSTM network, map the features into one-dimensional results through the fully connected layer, and ensure that the output is between 0 and 1 through the sigmoid activation function;

[0065] 2) Compare the model output with the preset threshold to determine the stability of the current system under the electromagnetic transient scale: when it is lower than the threshold, the evaluation result is that the system is stable; when it is higher than the threshold, the evaluation result is that the system is unstable.

[0066] The present invention discloses a transient stability assessment method and system for a new energy system based on machine learning and data-driven. The present invention constructs a system stability classification model based on machine learning, establishes a mapping relationship between electromechanical and electromagnetic time scales through data-driven, and realizes the prediction of system stability under the electromagnetic transient scale based only on the system operation under the electromechanical transient scale, thereby solving the dual problems of low accuracy of electromechanical transient modeling of high-proportion new energy systems, which cannot accurately reflect the real instability scene, and slow electromagnetic transient simulation speed and low efficiency, to a certain extent, and realizes transient stability assessment of new energy systems. The present invention can provide a basis for the analysis of the physical mechanism of transient stability and dominant instability characteristics of high-proportion new energy systems.

[0067] A specific embodiment of the present invention is as follows:

[0068] In such Figure 4 The effectiveness of the proposed method is verified in the high-proportion new energy AC / DC hybrid system shown in the figure. The system includes six wind farms and six photovoltaic power stations, with a total new energy output of 1800MW. Based on the load curve and the output curve of the new energy unit, a variety of working conditions are randomly generated to form a typical working condition set of the system. The output of each synchronous generator and new energy equipment is determined by power flow calculation; three-phase short-circuit faults are set at different lines to form a typical fault set of the system.

[0069] Electromechanical and electromagnetic transient simulations are performed on the system under various operating conditions and fault scenarios. A three-phase short circuit fault occurs at t=1s, and the circuit breaker is activated 0.1s later to cut off the faulty line. For electromechanical simulation, the maximum power angle difference, the lowest bus voltage and the highest bus frequency of the system at each moment are recorded, and the sampling step is ΔT=10ms. For electromagnetic simulation, the system stability is judged by the criteria shown in Table 1. If any instability criterion is triggered, the system is judged to be unstable, otherwise the system is stable. All feasible combinations of the system's typical operating condition set and typical fault set are traversed to form a trusted sample set for time domain simulation.

[0070] Table 1 System instability criterion

[0071]

[0072] The sample data is preprocessed, and the part of t=0~3s is intercepted from the electromechanical transient data; the electromagnetic transient simulation results are converted into Boolean type, with system stability corresponding to 0 (False) and system instability corresponding to 1 (True); the maximum and minimum value normalization method is used to process the data; the samples are randomly shuffled and divided into training set, test set and validation set according to 8:1:1; since the machine learning environment is GPU, the data type is converted to GPU tensor.

[0073] The model structure parameters and training parameters are shown in Table 2. During the forward propagation process, the LSTM network extracts the system stability features in the input data, and maps the stability features into a one-dimensional classification result through the linear transformation of the fully connected layer. The sigmoid activation function is used to ensure that the output result is between 0 and 1; during the back propagation process, the classification error is calculated using the binary cross entropy loss function, and the back propagation algorithm based on gradient descent is used to return the classification error to calculate the parameter gradient and update the model parameters. The model is trained based on the data-driven method. The training process uses the training set as input. After each round of training, the validation set is used to evaluate the current training effect. The optimizer uses the Adam optimizer, introduces weight decay to improve the generalization ability of the model, and the weight decay coefficient is set to 1e-5.

[0074] Table 2 LSTM model structure parameters and training parameters

[0075]

[0076]

[0077] The transient stability evaluation of the embodiment was performed using the method of the present invention, and the results are as follows:

[0078] Figure 5 The error curve of the system stability classification model training process is given. It can be seen from the results that the training set error and the validation set error both decrease with the number of iterations and eventually reach a low level, verifying the effectiveness of the model training. The trained model is evaluated, the training set accuracy is 99.18%, and the test set accuracy is 96.67%, where the accuracy refers to the percentage of the number of samples correctly classified by the model relative to the total number of samples. The accuracy of the model under different system load rates is shown in Table 3. It can be seen from the results that the method designed by the present invention can accurately evaluate the transient stability of the system at the electromagnetic scale, and has a certain generalization ability.

[0079] Table 3. Model accuracy at different system load rates

[0080]

[0081] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0082] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 Process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0083] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0084] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0085] The above specific implementation modes are used to explain the present invention rather than to limit the present invention. Any modification and change made to the present invention within the spirit of the present invention and the protection scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A transient stability assessment method for new energy systems based on machine learning and data-driven, characterized in that: The following steps are involved: Conduct time-domain simulations on the electromechanical transient model and electromagnetic transient model of the power system with a high proportion of new energy, design typical operating conditions and fault scenarios, and build a trusted sample set for electromechanical and electromagnetic transient time-domain simulations; A system stability classification model based on long short-term memory network is constructed. The model input is the electromechanical transient simulation monitoring curve, and the model output is the electromagnetic transient simulation stability classification probability value. The electromechanical and electromagnetic transient time-domain simulation credible sample set is preprocessed, and the system stability classification model is trained based on a data-driven method using the preprocessed sample set to achieve the evaluation of electromagnetic transient stability through the electromechanical transient simulation monitoring curve.

2. The method for transient stability assessment of new energy systems based on machine learning and data-driven according to claim 1, characterized in that: The electromagnetic transient model adopts a three-phase model, in which the calculation element model is described by the abc three-phase instantaneous values, and the model includes detailed modeling of the phase-locked loop control structure and the power electronic converter; the electromechanical transient model adopts a phasor model, in which the calculation element model is described by the fundamental wave phasor, and the model ignores the fast transient process of the converter switch dynamics and the phase-locked loop dynamics.

3. The method for transient stability assessment of new energy systems based on machine learning and data-driven according to claim 1, characterized in that: The typical operating conditions and fault scenarios are designed to construct a trusted sample set for time domain simulation, specifically including: considering the system load rate and new energy output to design a typical operating condition set for the new energy system; considering the fault type and fault location to design a typical fault set for the new energy system; performing electromechanical transient and electromagnetic transient simulations on the new energy system respectively, traversing all feasible combinations of the typical operating condition set and the typical fault set, and obtaining an electromechanical transient simulation monitoring curve sample set and an electromagnetic transient simulation stability result sample set.

4. The method for transient stability assessment of new energy systems based on machine learning and data-driven according to claim 1, characterized in that: The system stability classification model includes an LSTM network formed by connecting multiple LSTM units through a chain structure.

5. The method for transient stability assessment of new energy systems based on machine learning and data-driven according to claim 3 is characterized in that: The preprocessing of the electromechanical and electromagnetic transient time domain simulation credible sample set is specifically as follows: The electromechanical transient simulation curve samples are cut, data for a period of time before and after the fault is retained, and normalization is performed; the electromagnetic transient simulation stability result samples are converted into Boolean type, in which the system stability result corresponds to 0 and the system instability result corresponds to 1.

6. The method for transient stability assessment of new energy systems based on machine learning and data-driven according to claim 1, characterized in that: The method of using the preprocessed sample set to train the system stability classification model based on a data-driven method is specifically as follows: During the forward propagation process, the system stability features in the input electromechanical transient simulation monitoring curve are extracted through the LSTM network, and the stability features are mapped into a one-dimensional classification probability value through the linear transformation of the fully connected layer. The fully connected layer uses a sigmoid activation function to make the classification probability value between 0 and 1; During the back-propagation process, the Adam optimizer with weight decay is used to calculate the classification error according to the binary cross entropy loss function. The classification error is fed back based on the gradient descent back-propagation algorithm to calculate the parameter gradient and update the model parameters.

7. The method for transient stability assessment of new energy systems based on machine learning and data-driven according to claim 6, characterized in that: The Adam optimizer with weight decay specifically includes: using an adaptive Adam optimizer to update model parameters, adding a weight decay term when calculating gradients, adjusting the learning rate of each parameter by calculating the first-order moment estimate and the second-order moment estimate of the gradient, calculating the update amount of each parameter according to the learning rate, and applying the update amount to update the model parameters.

8. The method for transient stability assessment of new energy systems based on machine learning and data-driven according to claim 6, characterized in that: The binary cross entropy loss function is: L=-ylog(y p )-(1-y)log(1-y p ) Among them, L is the classification error of the model, y is the actual electromagnetic transient simulation stability result, and y p It is the stable classification probability value of the electromagnetic transient simulation output by the model.

9. The method for transient stability assessment of new energy systems based on machine learning and data-driven according to claim 1, characterized in that: The output of the system stability classification model is a classification probability value between 0 and 1. The classification probability value is compared with a preset threshold. When the classification probability value is lower than the threshold, the evaluation result is that the system is stable, and when it is higher than the threshold, the evaluation result is that the system is unstable.

10. A transient stability assessment system for new energy systems based on machine learning and data-driven, characterized in that: include: Data acquisition module: used to perform time domain simulation on the electromechanical transient model and electromagnetic transient model of the power system containing a high proportion of new energy, design typical operating conditions and fault scenarios, and build a trusted sample set for electromechanical and electromagnetic transient time domain simulation; Model building module: used to build a system stability classification model based on long short-term memory network. The model input is the electromechanical transient simulation monitoring curve, and the model output is the electromagnetic transient simulation stability classification probability value; Model training module: used to preprocess the trusted sample set of electromechanical and electromagnetic transient time domain simulation, and use the preprocessed sample set to train the system stability classification model based on the data-driven method, so as to realize the evaluation of electromagnetic transient stability through the electromechanical transient simulation monitoring curve.

Images