New energy power grid mechanical-electromagnetic simulation trajectory mapping method and system based on residual long short-term memory network

Abstract

The application discloses a new energy power grid electromechanical-electromagnetic simulation trajectory mapping method and system based on a residual long short-term memory network. The method comprises the following steps: respectively performing digital simulation on the electromechanical transient process and the electromagnetic transient process of a high-proportion new energy power grid system, obtaining electromechanical and electromagnetic transient time sequence trajectory samples and performing pretreatment; constructing a time sequence trajectory mapping model based on a residual long short-term memory network, the model comprising an input layer, a plurality of LSTM layers, a full connection layer and a regression layer, and a residual network being introduced between the LSTM layers; and training the model by using the pretreated electromechanical and electromagnetic transient time sequence trajectory samples, so as to realize the mapping from the electromechanical transient simulation time sequence trajectory to the electromagnetic transient simulation time sequence trajectory. The application can realize the prediction of the dynamic response process of the system at the electromagnetic transient scale based on only the existing electromechanical transient simulation trajectory data, and reflect the safety and stability of the system under the consideration of the high-frequency dynamics of power electronics.

Description

Technical Field

[0001] This invention belongs to the field of energy technology and relates to a method and system for electromechanical-electromagnetic simulation trajectory mapping of new energy power grids, and in particular to a method and system for electromechanical-electromagnetic simulation trajectory mapping of new energy power grids based on residual long short-term memory networks. Background Technology

[0002] With the increasing proportion of new energy sources such as wind power and photovoltaics in the power grid, the power grid's electronic characteristics are becoming more prominent, its operating characteristics are becoming more complex, and its stability state is undergoing profound changes. To address the new challenges faced by power grids with a high proportion of new energy, accurate and efficient power grid simulation technology can effectively assess the stability of new energy grid connection and the system's operating characteristics, and is one of the key means to ensure the safe and stable operation of the power grid. However, the large-scale integration of power electronic devices has significantly increased the scale and complexity of the power grid, and the system's component composition and dynamic processes are becoming increasingly complex, placing high demands on the accuracy of component modeling and computational capabilities in simulation.

[0003] Transient simulation of new energy power grids can generally be divided into electromechanical transient simulation and electromagnetic transient simulation. Electromechanical transient simulation mainly reflects the operating conditions of the system near the power frequency, such as the changes in rotor mechanical motion caused by changes in the electromagnetic torque of generators and motors. The calculation model usually uses fundamental phasors for description, which cannot simulate the high-frequency characteristics of power electronic equipment and its control loop. With the development of DC transmission and the continuous increase in the penetration rate of new energy, electromechanical transient simulation is difficult to meet the stability analysis requirements of high-proportion new energy power grids. In contrast, electromagnetic transient simulation can more accurately reflect the dynamic characteristics of power electronic equipment. However, electromagnetic models consider smaller time scales of dynamics, which greatly increases the computational load and time of the simulation. For systems of a certain scale, it is usually difficult to carry out electromagnetic transient simulation, and it cannot provide real-time basis for decision-making and scheduling of new energy power grids.

[0004] Considering the current situation where renewable energy grid dispatching mainly focuses on the dynamics of electromechanical time scales, this invention provides a method and system for mapping electromechanical-electromagnetic simulation trajectories in renewable energy grids based on residual long short-term memory networks. This invention uses data-driven learning to study the differences in the dynamic responses of the system at different time scales in electromechanical and electromagnetic systems, extracts the correlation features and mapping relationships of the simulated time-series trajectories at the two time scales, and constructs a time-series trajectory mapping model from electromechanical transients to electromagnetic transients. This achieves a method for reasonably predicting the transient processes of the system at the electromagnetic transient time scale based solely on the dynamic response characteristics of the system at the electromechanical time scale, thereby enabling a certain degree of safety and stability assessment of the system considering high-frequency dynamics of power electronics. Summary of the Invention

[0005] To address the aforementioned issues, this invention proposes a method and system for mapping electromechanical-electromagnetic simulation trajectories of new energy power grids based on residual long short-term memory networks, thus resolving the problem that existing electromechanical and electromagnetic simulations are difficult to apply to new energy power grids in the context of continuously increasing new energy penetration rates.

[0006] The technical solution adopted in this invention is as follows:

[0007] A method for mapping electromechanical-electromagnetic simulation trajectories of new energy power grids based on residual long short-term memory networks includes the following steps:

[0008] A method for mapping electromechanical-electromagnetic simulation trajectories of new energy power grids based on residual long short-term memory networks includes the following steps:

[0009] Digital simulations were performed on the electromechanical and electromagnetic transient processes of a high-proportion renewable energy power grid system to obtain electromechanical and electromagnetic transient time-series trajectory samples, and these samples were preprocessed.

[0010] A time-series trajectory mapping model based on residual long short-term memory network (R-LSTM) is constructed. The model input is electromechanical transient time-series trajectory, and the output is electromagnetic transient time-series trajectory. The time-series trajectory mapping model includes an input layer, several LSTM layers, a fully connected layer, and a regression layer, and a residual network is introduced between the input of the first LSTM layer and the output of the last LSTM layer.

[0011] The time-series trajectory mapping model is trained using preprocessed electromechanical and electromagnetic transient time-series trajectory samples to achieve the mapping from electromechanical transient simulation time-series trajectory to electromagnetic transient simulation time-series trajectory.

[0012] Furthermore, in the electromechanical transient simulation time-series trajectory of the high-proportion new energy power grid system, the computational element model is described by the fundamental wave vector, and in the electromagnetic transient simulation time-series trajectory, the computational element model is described by the abc three-phase instantaneous values.

[0013] Furthermore, the preprocessing of the electromechanical and electromagnetic transient time-series trajectory samples specifically involves: resampling and linear interpolation of the electromechanical and electromagnetic transient time-series trajectory samples based on the fault time scale and frequency domain response characteristics to obtain electromechanical and electromagnetic discrete time-series data with the same sequence length and corresponding key fault nodes.

[0014] Furthermore, the time-series trajectory mapping model incorporates several LSTM layers connected in series to abstract the input data features layer by layer, extracting the features required for the task.

[0015] Furthermore, the residual network uses an identity mapping to directly transmit the input signal of the first LSTM layer to the last LSTM layer, as part of the feature vector of the final output of several LSTM layers.

[0016] Furthermore, the feature vector includes the output signal of the LSTM layer and the input signal transmitted through the residual network, and its expression is:

[0017]

[0018] in, This is the output of the residual LSTM network, where G(X) is the feature vector output by several LSTM layers. It is the input signal for residual network transmission. and W r These represent the activation function and the projection matrix, respectively.

[0019] Furthermore, the training process of the time-series trajectory mapping model is as follows:

[0020] Supervised learning is used to predict electromagnetic transient time-series trajectory data based on input electromechanical transient time-series trajectory data. The mapping error between the predicted electromagnetic transient time-series trajectory data and the real electromagnetic transient time-series trajectory data is calculated using a loss function. The mapping error is backpropagated based on the gradient descent algorithm of machine learning to update the network parameters.

[0021] Furthermore, the regression layer in the time-series trajectory mapping model uses the HuberLoss loss function. When the prediction deviation is greater than the threshold parameter, the squared error is used, and when the prediction deviation is less than the threshold parameter, the linear error is used.

[0022] A new energy power grid electromechanical-electromagnetic simulation trajectory mapping system based on residual long short-term memory networks includes:

[0023] Data acquisition module: used to perform digital simulation of the electromechanical transient process and electromagnetic transient process of high proportion of new energy power grid system, obtain electromechanical and electromagnetic transient time series trajectory samples, and preprocess the electromechanical and electromagnetic transient time series trajectory samples;

[0024] Model building module: used to build a time-series trajectory mapping model based on residual long short-term memory network. The model input is electromechanical transient time-series trajectory, and the output is electromagnetic transient time-series trajectory. The time-series trajectory mapping model includes an input layer, several LSTM layers, a fully connected layer and a regression layer, as well as a residual network introduced between the input of the first LSTM layer and the output of the last LSTM layer.

[0025] Model training module: used to train the time-series trajectory mapping model using preprocessed electromechanical and electromagnetic transient time-series trajectory samples, so as to realize the mapping from electromechanical transient simulation time-series trajectory to electromagnetic transient simulation time-series trajectory.

[0026] A computer device, comprising:

[0027] One or more processors;

[0028] Memory, used to store one or more programs;

[0029] When the one or more programs are executed by the one or more processors, the one or more processors implement the above-mentioned electromechanical-electromagnetic simulation trajectory mapping method for new energy power grids based on residual long short-term memory networks.

[0030] The beneficial effects of this invention are:

[0031] This invention provides a method for mapping electromechanical-electromagnetic simulation trajectories in new energy power grids based on residual long short-term memory networks. It aims to identify the differences in the dynamic response of the system at different time scales in electromechanical and electromagnetic modes through a data-driven approach. By extracting the correlation features and mapping relationships of the simulated time-series trajectories at different time scales, a time-series trajectory mapping model from electromechanical transients to electromagnetic transients is constructed. This achieves reasonable prediction of the dynamic response process at the electromagnetic transient time scale based solely on the system's electromechanical time-scale response characteristics. This invention can provide a foundation for analyzing the transient stability physical mechanisms and dominant instability characteristics of high-proportion new energy power grids that consider power electronic dynamics. Attached Figure Description

[0032] Figure 1 This is a flowchart of the training and prediction process for the time-series trajectory mapping model in an embodiment of the present invention.

[0033] Figure 2 This is a structural diagram of the LSTM unit in an embodiment of the present invention.

[0034] Figure 3 This is a diagram of the R-LSTM model with residual network connections introduced in this embodiment of the invention.

[0035] Figure 4 This is a modeling diagram of a three-unit, nine-node wind turbine grid-connected system in an embodiment of the present invention.

[0036] Figure 5 This is a graph showing the model training error and verification error in an embodiment of the present invention.

[0037] Figure 6 The diagram shows the simulation waveform and the prediction result of the mapping model in the embodiment of the present invention. Detailed Implementation

[0038] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] A method for mapping electromechanical-electromagnetic simulation trajectories of new energy power grids based on residual long short-term memory networks includes the following steps:

[0040] S1. Digital simulations are performed on the electromechanical and electromagnetic transient processes of the high-proportion renewable energy power grid system to obtain time-series trajectory samples of electromechanical and electromagnetic transients at different time scales. These time-series trajectory samples are then preprocessed. Specific steps include: For electromechanical transient modeling of the high-proportion renewable energy power grid system, the computational element model is described using a fundamental phasor. The amplitude V of voltage and current at any given time at each renewable energy grid connection point is measured. s rms (t), I rms (t). Among them, the magnitude of the controlled current source current for the doubly-fed wind turbine connected to the grid. The grid-connected power S(t) is calculated using the asynchronous motor equation and the grid-side impedance equation.

[0041] Electromagnetic transient modeling is performed on a high-proportion renewable energy power grid. The computational component model uses a three-phase instantaneous value description (abc). The three-phase instantaneous values ​​of voltage and current (v) at any time are measured at each renewable energy grid connection point. sabc and i abc and effective value I emt (t).

[0042] Assume Δt emt and Δt rms Let T be the simulation step size for electromagnetic transients and electromechanical transients, respectively. The total simulation duration is T. The electromechanical transient time trajectory is expressed by the following formula:

[0043]

[0044] The electromagnetic transient time trajectory is represented by the following formula:

[0045]

[0046] in, These represent the electromechanical transient time-series trajectories of voltage and current at the new energy grid-connected node, respectively. These represent the electromagnetic transient time-series trajectories of the voltage and current at the new energy grid-connected node, respectively. (k×Δt) rms (k×Δt) emt ) represent the sampling time on the electromechanical and electromagnetic trajectories corresponding to the k-th sampling point, respectively.

[0047] To address the issue of inconsistent parameter settings such as step size and fault time in electromechanical and electromagnetic simulations, the electromechanical-electromagnetic time-series trajectory data of new energy systems are preprocessed to obtain discrete time-series data X=[x(1),…,x(K)] and Y=[y(1),…,y(K)] corresponding to key fault nodes with the same sequence length. Here, X and Y are the discrete time-series data of electromechanical and electromagnetic systems, respectively, and K is the sequence length.

[0048] S2. Construct a time-series trajectory mapping model based on a residual-based Long Short-Term Memory (R-LSTM) network. The model input is the electromechanical transient time-series trajectory, and the output is the electromagnetic transient time-series trajectory. The time-series trajectory mapping model includes an input layer, several LSTM layers, a fully connected layer, and a regression layer, as well as a residual network introduced between the input of the first LSTM layer and the output of the last LSTM layer. The specific steps are as follows:

[0049] (1) Construct a time series mapping model based on a Long Short-Term Memory (LSTM) neural network. Through an n-layer cascaded LSTM neural network, the features of the input data are abstracted layer by layer to extract time-series feature representations. The basic structure of the LSTM unit is as follows: Figure 2 As shown;

[0050] (2) Introducing a branch to connect the input of the first-layer LSTM and the output of the nth-layer LSTM, this connection signal branch is called residual connection, and its structure is as follows: Figure 3 A residual network is composed of a series of residual blocks. The general representation of a residual block is as follows:

[0051]

[0052] In the formula, G(X) is the feature vector output by several LSTM layers. It is the input signal for residual network transmission. It is the activation function of the residual connection. This branch allows the identity mapping of the input signal to propagate directly to the final layer of the LSTM as a feature vector. This also helps deeper networks avoid overfitting and network degradation problems, and makes them easier to optimize.

[0053] (3) R-LSTM network output feature vector In electromechanical transient simulation, the dimensionality reduction representation of the original physical quantity simulation sequence is used. The feature vector retains the main information and modes in the original data, while removing redundancy and noise, making subsequent processing and analysis more efficient and accurate. The feature vector is passed through a fully connected network to obtain the final prediction result.

[0054] Furthermore, the HuberLoss loss function is introduced into the time series mapping model to calculate the model regression error, specifically including:

[0055] Instead of the traditional RMSE loss function, the HuberLoss loss function with hyperparameter δ is introduced into the regression layer of the model:

[0056] L(y,f(x))=RMSE+L δ(y,f(x))

[0057]

[0058] In the formula, y is the true value of the electromagnetic transient simulation trajectory, f(x) is the predicted value of the electromagnetic transient simulation trajectory output by the model, and δ determines the sensitivity of the Huber Loss loss function to outliers. When the prediction deviation is greater than δ, the squared error is used to increase the sensitivity to peak fitting error. When the prediction deviation is less than δ, the linear error is used to make the training more robust.

[0059] Based on the characteristics of short duration and large variation of key extreme data such as impact current or voltage drop depth in power grid transient processes, the model's extreme point fitting accuracy is improved by extracting peak features from the time-series trajectory and adaptively adjusting the loss weight.

[0060] S3. The time-series trajectory mapping model is trained using the preprocessed electromechanical and electromagnetic transient time-series trajectory samples to achieve the mapping from electromechanical transient simulation time-series trajectory to electromagnetic transient simulation time-series trajectory. The specific steps are as follows:

[0061] Based on the electromechanical transient time-series trajectories X=[x(1),…,x(K)] and Y=[y(1),…,y(K)] of the aforementioned high-proportion renewable energy power grid system, the time-series trajectory mapping model is trained using supervised learning. The model predicts the electromagnetic transient time-series trajectories Y for each electrical quantity based on the input electromechanical transient time-series trajectory X. The prediction results are represented by Y. p =[y p (1),…,y p [K] indicates that the mapping error of the model is calculated based on the above loss function, and the network parameters are updated by backpropagating the mapping error using the gradient descent algorithm based on machine learning. The flowchart of the time-series trajectory mapping model training and prediction process is as follows: Figure 1 As shown.

[0062] This invention provides a method for mapping electromechanical-electromagnetic simulation trajectories in new energy power grids based on residual long short-term memory networks. It aims to identify the differences in dynamic response of new energy power grid systems at different time scales using a data-driven approach. By extracting the correlation features and mapping relationships of simulated time-series trajectories at different time scales, a time-series trajectory mapping model from electromechanical transients to electromagnetic transients is constructed. This achieves reasonable prediction of the dynamic response process at the electromagnetic transient time scale based solely on the system's electromechanical time-scale response characteristics. This invention can provide a foundation for analyzing the transient stability physical mechanisms and dominant instability characteristics of high-proportion new energy power grids that consider power electronic dynamics.

[0063] A specific embodiment of the present invention is as follows:

[0064] The effectiveness of the proposed method was verified in an IEEE-39 doubly-fed wind turbine grid-connected system. Figure 4 As shown, IEEE 9-Bus system models under electromechanical and electromagnetic transient conditions are established respectively. A doubly-fed induction generator (DFIG) is added at Bus 3 to replace part of the output of the synchronous generator at Bus 3. Multiple operating conditions are randomly generated based on the load curve and the turbine output curve. Power flow calculations are performed for each operating condition to determine the active power output P of each synchronous generator. Gi With no effort Q Gi .

[0065] Electromechanical and electromagnetic transient simulations were performed on a doubly-fed induction generator (DFIG) grid-connected system under low-voltage conditions. A three-phase short-circuit fault was randomly selected at different locations on different lines, and the fault was cleared after 0.2 seconds. The voltage Vs, current Is, and active power output PDFIG and reactive power output QDFIG at the turbine's grid connection point were observed before and after the fault. The simulation duration was T = 2 seconds. The step size for both electromagnetic and electromechanical simulations was set to Δt. emt =25us and Δt rms =5ms, obtaining the electromechanical simulation timing trajectory under different operating conditions and faults. (f) s Sampling was performed at 0.5ms intervals to obtain electromechanical and electromagnetic transient time series samples (X,Y) with a sequence length of K=4000, which were then stored as data samples for subsequent analysis. The model and training parameter settings are shown in Table 1. 9000 data samples containing various physical quantities were obtained from Simulink simulation. After randomizing the dataset, 30% of the samples were used as the training set and 70% as the test set. The input sequence length of the model is the length of the electromechanical transient simulation sequence, and the output sequence length is the length of the electromagnetic transient simulation sequence that the model needs to predict. The parameter settings of the LSTM network referenced the settings of classic LSTM models and were ultimately determined through multiple experiments. The number of hidden units N in the LSTM network was... h Set to 200, which is the feature vector output by the R-LSTM network. The dimension of the LSTM is set to 20, the dimensionality reduction ratio is 20, the number of fully connected layer units is set to 50, the random inactivation rate is set to 20%, and the ReLU activation function is used for the LSTM output.

[0066] The simulation calculations of the embodiments were performed using the method of the present invention, and the results are as follows:

[0067] The prediction performance of R-LSTM neural network, commonly used LSTM neural network, and convolutional neural network (CNN) was compared and studied on a PC platform with an i5-3.4GHz CPU and 8GB RAM. The results are listed in Table 2.

[0068] Table 1 Comparison of calculation results from different models

[0069]

[0070] Comparing different algorithms, CNN has a certain advantage in training and computation time compared to LSTM. However, when the training time is similar, the prediction accuracy of the residual-connected LSTM is significantly improved. For different physical quantities, the residual-connected LSTM neural network achieves the minimum RMSE error. Among these, the model has the highest prediction accuracy for the EMT dynamic trajectory of the grid connection point voltage and current. The training and validation errors of the model are as follows: Figure 5 As shown, the R-LSTM model converges faster and performs better on the validation set, demonstrating its resistance to overfitting. The electromechanical / electromagnetic simulation waveforms and prediction results of the wind turbine grid-connected current and voltage are shown in Figure 6. As can be seen from the figure, the time-series trajectory mapping model of this invention has good identification ability and fitting accuracy for different physical quantities. Therefore, the introduction of a residual network in the model effectively improves the model's generalization ability when predicting electromagnetic simulation sequences of other physical quantities. Simultaneously, the model can effectively extract sequence peak features and adjust the loss weights, improving the fitting accuracy of the mapping model in the peak region of the EMT simulation trajectory, enabling the model to better predict the maximum current and minimum voltage that can only be reflected in the dynamics of the electromagnetic model.

[0071] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied 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.

[0072] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0073] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0074] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0075] The above specific embodiments are used to explain and illustrate the present invention, but not to limit the present invention. Any modifications and changes made to the present invention within the spirit and scope of the claims shall fall within the protection scope of the present invention.

Claims

1. A method for mapping electromechanical-electromagnetic simulation trajectories of new energy power grids based on residual long short-term memory networks, characterized in that, Includes the following steps: Digital simulations were performed on the electromechanical and electromagnetic transient processes of a high-proportion renewable energy power grid system to obtain electromechanical and electromagnetic transient time-series trajectory samples, and these samples were preprocessed. A time-series trajectory mapping model based on residual long short-term memory network is constructed. The model input is electromechanical transient time-series trajectory, and the output is electromagnetic transient time-series trajectory. The time-series trajectory mapping model includes an input layer, several LSTM layers, a fully connected layer, and a regression layer, as well as a residual network introduced between the input of the first LSTM layer and the output of the last LSTM layer. The time-series trajectory mapping model is trained using preprocessed electromechanical and electromagnetic transient time-series trajectory samples to achieve the mapping from electromechanical transient simulation time-series trajectory to electromagnetic transient simulation time-series trajectory.

2. The method for mapping electromechanical-electromagnetic simulation trajectories of new energy power grids based on residual long short-term memory networks according to claim 1, characterized in that, In the electromechanical transient simulation time-series trajectory of the high-proportion new energy power grid system, the computational element model is described by the fundamental wave vector, and in the electromagnetic transient simulation time-series trajectory, the computational element model is described by the abc three-phase instantaneous values.

3. The method for mapping electromechanical-electromagnetic simulation trajectories of new energy power grids based on residual long short-term memory networks according to claim 1, characterized in that, The preprocessing of the electromechanical and electromagnetic transient time-series trajectory samples specifically involves: resampling and linear interpolation of the electromechanical and electromagnetic transient time-series trajectory samples based on the fault time scale and frequency domain response characteristics to obtain electromechanical and electromagnetic discrete time-series data with the same sequence length and corresponding key fault nodes.

4. The method for mapping electromechanical-electromagnetic simulation trajectories of new energy power grids based on residual long short-term memory networks according to claim 1, characterized in that, The time-series trajectory mapping model consists of several LSTM layers connected in series, which abstract the features of the input data layer by layer to extract the features required for the task.

5. The method for mapping electromechanical-electromagnetic simulation trajectories of new energy power grids based on residual long short-term memory networks according to claim 1, characterized in that, The residual network uses an identity mapping to directly transmit the input signal of the first LSTM layer to the last LSTM layer, as part of the feature vector of the final output of several LSTM layers.

6. The method for mapping electromechanical-electromagnetic simulation trajectories of a new energy power grid based on residual long short-term memory networks according to claim 5, characterized in that, The feature vector includes the output signal of the LSTM layer and the input signal transmitted through the residual network, and its expression is: in, This is the output of the residual LSTM network, where G(X) is the feature vector output by several LSTM layers. It is the input signal for residual network transmission. and W r These represent the activation function and the projection matrix, respectively.

7. The method for mapping electromechanical-electromagnetic simulation trajectories of new energy power grids based on residual long short-term memory networks according to claim 1, characterized in that, The training process of the time-series trajectory mapping model is as follows: Supervised learning is used to predict electromagnetic transient time-series trajectory data based on input electromechanical transient time-series trajectory data. The mapping error between the predicted electromagnetic transient time-series trajectory data and the real electromagnetic transient time-series trajectory data is calculated using a loss function. The mapping error is backpropagated based on the gradient descent algorithm of machine learning to update the network parameters.

8. The method for electromechanical-electromagnetic simulation trajectory mapping of new energy power grids based on residual long short-term memory networks according to claim 1, characterized in that, The regression layer in the time-series trajectory mapping model uses the HuberLoss loss function. When the prediction deviation is greater than the threshold parameter, the squared error is used, and when the prediction deviation is less than the threshold parameter, the linear error is used.

9. A new energy power grid electromechanical-electromagnetic simulation trajectory mapping system based on residual long short-term memory networks, characterized in that, include: Data acquisition module: used to perform digital simulation of the electromechanical transient process and electromagnetic transient process of high proportion of new energy power grid system, obtain electromechanical and electromagnetic transient time series trajectory samples, and preprocess the electromechanical and electromagnetic transient time series trajectory samples; Model building module: used to build a time-series trajectory mapping model based on residual long short-term memory network. The model input is electromechanical transient time-series trajectory, and the output is electromagnetic transient time-series trajectory. The time-series trajectory mapping model includes an input layer, several LSTM layers, a fully connected layer and a regression layer, as well as a residual network introduced between the input of the first LSTM layer and the output of the last LSTM layer. Model training module: used to train the time-series trajectory mapping model using preprocessed electromechanical and electromagnetic transient time-series trajectory samples, so as to realize the mapping from electromechanical transient simulation time-series trajectory to electromagnetic transient simulation time-series trajectory.

10. A computer device, characterized in that, The computer device includes: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the electromechanical-electromagnetic simulation trajectory mapping method for new energy power grids based on residual long short-term memory networks as described in any one of claims 1-8.

Images