A method for determining transient stability of power systems based on hybrid drive of multi-source data

Abstract

The present invention discloses a method for determining transient stability of an electric power system based on hybrid drive of multi-source data, and belongs to the field of deep learning and stability analysis of electric power systems. The present invention first determines the initial operating point of the target electric power system under different load levels based on Latin hypercube sampling for the target electric power system, performs transient stability simulation in combination with different system fault information to obtain transient stability margin, and forms a training sample set through data processing; then, constructs and uses the training sample set to train to obtain a transient stability judgment model for an electric power system with system fault information, system steady-state operation information, and generator power angle trajectory change information after a fault as input, and transient stability margin as output, and finally is used to realize rapid determination of transient stability of the electric power system. This method can effectively replace the complex and time-consuming electromechanical transient simulation process, and meet the needs of rapid response in actual power system stability control.

Description

Technical Field

[0001] The present invention belongs to the field of deep learning and power system stability analysis, and specifically relates to a method for determining transient stability of a power system based on hybrid drive of multi-source data. Background Art

[0002] Transient stability control of large-scale power systems is a core technology for improving grid efficiency and ensuring energy security. However, in practical applications, this field faces multiple challenges such as algorithm design, computational efficiency, and adaptability. In particular, the optimization calculation of emergency control strategies is huge, especially those involving transient stability constraints, which has become the main bottleneck restricting further development of this field.

[0003] At present, in order to meet the accuracy requirements and reduce the computing pressure, the power system transient stability emergency control mainly relies on the strategy of "offline pre-decision, real-time matching". Although this method has certain effects in practical applications, its potential risks cannot be ignored due to the mismatch phenomenon that may exist in the operating conditions. To this end, researchers have proposed an improved scheme of "online pre-decision, real-time matching" in order to improve the real-time and accuracy of control. However, this scheme still faces severe challenges in terms of computing efficiency, which limits its promotion in practical applications. Although some scholars have tried to use high-performance computing technology to improve computing efficiency, the actual effect is not ideal. Therefore, it is urgent to develop a new method for determining transient stability of power systems to efficiently replace the complex and time-consuming transient stability time domain simulation process in the actual solution process, fundamentally reduce the calculation scale, and significantly improve the efficiency of emergency control strategy solution. Summary of the invention

[0004] The purpose of the present invention is to solve the problems of large calculation scale and low efficiency in transient stability time domain simulation in the prior art, and to provide a method for determining transient stability of a power system based on hybrid drive of multi-source data, so as to quickly determine the transient stability status of the system during the emergency control of transient stability of the power system.

[0005] In order to achieve the above-mentioned invention object, the technical solution specifically adopted by the present invention is as follows:

[0006] A method for determining transient stability of a power system based on hybrid drive of multi-source data, the steps of which are as follows:

[0007] S1. For the target power system, the initial operation point of the target power system under different load levels is determined based on Latin hypercube sampling, transient stability simulation is performed in combination with different system fault information, and the transient stability margin is calculated according to the simulation results, thereby generating simulation data samples consisting of system fault information, system steady-state operation information, post-fault generator power angle trajectory change information and transient stability margin, and each simulation data sample is processed to form a training sample set that meets the input of the neural network model;

[0008] S2. Constructing a convolutional neural network model with system fault information, steady-state operation characteristics and trajectory change characteristics as input and transient stability margin as output, and training the convolutional neural network model using the training sample set to obtain a transient stability judgment model for the power system;

[0009] S3. Obtain system fault information, system steady-state operation information, and post-fault generator power angle trajectory change information of the target power system at the operating point to be predicted, and input them into the power system transient stability judgment model to obtain a transient stability margin prediction value of the target power system.

[0010] Preferably, in step S1, the method for generating simulation data samples is as follows:

[0011] S11. Uniformly sample multiple central load levels k within the load level range of the target power system; for each central load level k, obtain the fault line fl, fault duration ft, fault location fp, and fault type required for simulation through random sampling without replacement, and randomly sample load level values ​​ki within the neighborhood of the central load level k through Latin hypercube sampling, so that each sampled value forms system fault information;

[0012] S12. For each group of system fault information composed of sampling values, the load level value ki is used as the initial operating point of the target power system, and the corresponding fault line fl, fault duration ft, fault location fp and fault type are set in the simulated target power system, while retaining the system steady-state operation information of each generator in the target power system, the system steady-state operation information includes the generator active power PG, the generator reactive power QG, the machine-end bus voltage amplitude V, the machine-end bus voltage phase angle θ, the generator power angle δ, the generator online status Ol, the machine-end line active load PL and the machine-end line reactive load QL, and then the target power system is subjected to transient stability simulation to obtain the trajectory change information of the generator power angle δ of each generator after the fault occurs and the maximum power angle difference between all generators, and finally the maximum power angle difference between all generators is converted into the transient stability margin TSI as the sample label value.

[0013] Furthermore, the load level range is 80%~130%, the sampling range of the fault duration is 0.1~0.2s, the sampling range of the fault position is 20%~80%, the fault type is only sampled as a three-phase short circuit fault, the neighborhood range of each central load level k is ±2.5% centered on itself, the trajectory change information is the trajectory change information of the generator power angle δ 1s after the fault occurs; the maximum power angle difference is the maximum power angle difference between all generators within 10s after the start of the simulation.

[0014] Preferably, the transient stability margin TSI is based on the maximum power angle difference Δ between all generators. max The conversion formula is TSI = 100* (360-Δ max ) / (360+Δ max ).

[0015] Preferably, in S1, when processing the simulation data samples, Z-Score normalization and sample balancing processing need to be performed on the data.

[0016] Preferably, the convolutional neural network model consists of three feature extraction branches and a fully connected network. The system fault information, the system steady-state operation information, and the generator power angle trajectory change information after the fault are each extracted through a feature extraction branch, and each feature extraction branch is composed of alternating convolutional layers and pooling layers cascaded together; after the feature information extracted by the three feature extraction branches passes through the fully connected network, the transient stability margin prediction value of the target power system is output.

[0017] Furthermore, each of the feature extraction branches includes three convolution layers with ReLU functions and three pooling layers, and the convolution layers and the pooling layers are alternately arranged.

[0018] Furthermore, the fully connected network is formed by cascading three fully connected layers, wherein the first two fully connected layers output through the ReLU activation function, and the last fully connected layer outputs through the tanh activation function.

[0019] Preferably, during the training process of the convolutional neural network model, the Adam optimization algorithm is used to optimize the model parameters, and Smooth L1 Loss is used as the loss function.

[0020] Preferably, the transient stability simulation is implemented by PSAT transient stability simulation software.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] The present invention obtains samples that retain system fault information, steady-state operation characteristics, trajectory change characteristics and transient stability margin results through transient stability simulation, and uses these samples to train a convolutional neural network model, so that it can quickly predict the transient stability margin corresponding to the power system by inputting the system's fault information, steady-state characteristics and trajectory characteristics, meeting the need for rapid response in actual power system stability control. The results show that the convolutional neural network model constructed by this method has excellent fitting ability and has a high accuracy rate under different fault scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic diagram of the steps of the method for determining transient stability of a power system based on hybrid drive of multi-source data of the present invention;

[0024] Figure 2 is a flow chart of building a convolutional neural network model in an embodiment of the present invention;

[0025] Figure 3 is a network structure diagram of a convolutional neural network model in an embodiment of the present invention;

[0026] Figure 4 is a convolutional neural network model training flow chart in an embodiment of the present invention;

[0027] Figure 5 It is a curve showing how the accuracy and loss value of the convolutional neural network model change with the number of training rounds in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings.

[0029] On the contrary, the present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention as defined by the claims. Further, in order to make the public have a better understanding of the present invention, some specific details are described in detail in the detailed description of the present invention below. Those skilled in the art can fully understand the present invention without the description of these details.

[0030] In a preferred embodiment of the present invention, a method for determining transient stability of an electric power system based on a hybrid drive of multi-source data is provided. The method first generates transient stability simulation samples to obtain samples that retain system fault information, steady-state operation characteristics, trajectory change characteristics, and transient stability margin results. On this basis, key features are extracted and the input and output data structures are determined, and the sample data is then preprocessed. Then, according to the scale and structure of the electric power system to be evaluated, a convolutional neural network model structure for transient stability margin evaluation is constructed based on deep learning technology. Finally, the convolutional neural network model is trained and tested based on simulation samples, the network structure is continuously optimized, and hyperparameters are adjusted to improve the model's predictive performance. Based on the trained model, rapid determination of transient stability of an electric power system based on a hybrid drive of multi-source data can be achieved. In the following, combined with Figure 1 As shown, the specific steps of the method for determining transient stability of a power system based on hybrid drive of multi-source data are described in detail.

[0031] S1. For the target power system, the initial operating point of the target power system under different load levels is determined based on Latin hypercube sampling. Transient stability simulation is performed in combination with different system fault information, and the transient stability margin is calculated based on the simulation results, thereby generating simulation data samples consisting of system fault information, system steady-state operation information, generator power angle trajectory change information after fault, and transient stability margin. Each simulation data sample is processed to form a training sample set that meets the input of the neural network model.

[0032] It should be noted that the target power system in the present invention can be any power system that needs to perform transient stability judgment. The number of generators contained in each power system and the corresponding system steady-state operation information are different. In the subsequent simulation process, the parameters of each system itself can be determined according to the actual power system conditions.

[0033] In an embodiment of the present invention, the simulation data sample is realized by performing transient stability simulation on the target power system, and the specific simulation process can be realized by PSAT transient stability simulation software. However, transient stability simulation requires a series of input parameters, including system fault information and system steady-state operation information, which need to be determined by sampling or according to the actual system conditions. The method for generating simulation data samples is specifically introduced below, and its process is shown in S11 and S12:

[0034] S11. Uniformly sample multiple central load levels k within the load level range of the target power system; for each central load level k, obtain the fault line fl, fault duration ft, fault location fp, and fault type required for simulation through random sampling without replacement, and randomly sample load level values ​​ki within the neighborhood of the central load level k through Latin hypercube sampling, so that the sampled values ​​​​compose the system fault information.

[0035] It should be noted that in the process of the above-mentioned sampling to generate system fault information, the sampling range of each parameter is determined in advance according to the situation of the target power system itself. In the present invention, it is first necessary to uniformly sample multiple central load levels k within the load level range of the target power system. The central load level k is not used directly as the system load level during transient stability simulation, but it is necessary to construct a neighborhood range with the central load level k as the center, and then further randomly sample the load level value ki within the neighborhood range through Latin hypercube sampling. Thus, for each central load level k, a series of samples that solve the load level but are not completely consistent can be formed, thereby improving the sample diversity of subsequent simulations, and then improving the robustness of subsequent neural networks.

[0036] In an embodiment of the present invention, the load level range of the sampled central load level k is 80% to 130%, and the central load level k is preferably sampled at equal intervals with a step size of 5%. In addition, for several other parameters, the sampling range of the fault duration is preferably 0.1 to 0.2s, the sampling range of the fault position is preferably 20% to 80%, the fault type is only sampled as a three-phase short circuit fault, the neighborhood range of each central load level k is preferably a ±2.5% range centered on itself, and the trajectory change information is preferably the trajectory change information of the generator power angle δ 1s after the fault occurs; the maximum power angle difference is the maximum power angle difference between all generators within 10s after the start of the simulation. Under such a setting, only the trajectory change information 1s after the fault occurs, combined with other system fault information and system steady-state operation information, can be used to predict the TSI 10s after the start of the simulation, and whether the transient stability constraint is satisfied is directly determined based on this, so as to avoid the complex and time-consuming simulation of the subsequent transient stability simulator within 1 to 10s. Of course, in other embodiments, the above two parameters of 1s and 10s can be adjusted according to actual conditions, and are not fixed values.

[0037] S12. For each group of system fault information composed of sampling values, the load level value ki is used as the initial operating point of the target power system, and the corresponding fault line fl, fault duration ft, fault location fp and fault type are set in the simulated target power system, and the system steady-state operation information of each generator in the target power system is retained. The above-mentioned system steady-state operation information includes the generator active power PG, the generator reactive power QG, the machine-end bus voltage amplitude V, the machine-end bus voltage phase angle θ, the generator power angle δ, the generator online status Ol, the machine-end line active load PL and the machine-end line reactive load QL, and then the target power system is subjected to transient stability simulation to obtain the trajectory change information of the generator power angle δ of each generator after the fault occurs and the maximum power angle difference between all generators, and finally the maximum power angle difference between all generators is converted into the transient stability margin TSI as the sample label value.

[0038] In an embodiment of the present invention, the transient stability margin TSI is based on the maximum power angle difference Δ between all generators. max The conversion formula is TSI = 100* (360-Δ max ) / (360+Δ max ). The TSI value range is from -100 to 100, representing the system stability from severe instability to complete stability. When TSI is 0, it means that the system is in a critical stability state.

[0039] After completing the transient stability simulation of all groups of system fault information, a simulation data sample corresponding to each group of system fault information can be obtained. When all simulation data samples are constructed as corresponding neural network training data sets, it is also necessary to perform data processing on the simulation data samples so that they meet the input requirements and training requirements of the subsequent neural network. Generally speaking, when performing data processing on the simulation data samples, the data needs to be Z-Score normalized and sample balanced. Z-Score normalization belongs to the conventional technology of neural network input data processing, which will not be described in detail. The sample balance processing is to solve the problem of imbalance between positive and negative samples, because when the number of samples of one class is much larger than that of other classes, the model may be biased towards the majority class, resulting in a decrease in the recognition performance of the minority class. In an embodiment of the present invention, the samples can be classified into stable samples and unstable samples according to the transient stability margin TSI, so as to perform proportional control so that the number of stable samples and the number of unstable samples reach a reasonable level. After all simulation data samples are processed, the training sample set that meets the input of the neural network model can be used for training the neural network.

[0040] S2. Construct a convolutional neural network model with system fault information, steady-state operation characteristics and trajectory change characteristics as input and transient stability margin as output, and use the training sample set to train the convolutional neural network model to obtain a transient stability judgment model for the power system.

[0041] It should be noted that the above-mentioned convolutional neural network model can adopt a conventional convolutional neural network (CNN) architecture built based on deep learning technology, that is, feature extraction and activation output are realized through a combination of convolutional layers, pooling layers, fully connected layers and activation layers. In an embodiment of the present invention, since the convolutional neural network model is used for transient stability margin assessment and is driven by a mixture of multi-source data, in order to adapt to three types of input data, the convolutional neural network model is composed of three feature extraction branches and a fully connected network. The system fault information, the system steady-state operation information, and the post-fault generator power angle trajectory change information are each extracted through a feature extraction branch. Feature information, each feature extraction branch is composed of alternating convolutional layers and pooling layers cascaded; after the feature information extracted by the three feature extraction branches passes through the fully connected network, the transient stability margin prediction value of the target power system is output.

[0042] In an embodiment of the present invention, each feature extraction branch includes three convolutional layers with ReLU function activation output and three pooling layers, and the convolutional layers and the pooling layers are alternately arranged. The fully connected network is formed by cascading three fully connected layers, wherein the first two fully connected layers output through the ReLU activation function, and the last fully connected layer outputs through the tanh activation function, and the output of the tanh activation function layer is the predicted value of the transient stability margin. In addition, the ReLU function can be used for output after each convolutional layer, and the pooling layer can use maximum pooling.

[0043] It should be noted that the training of the convolutional neural network model belongs to the prior art, and the network parameters can be iteratively optimized by alternating forward propagation and backward propagation, and a reasonable loss function can be set according to the output parameters. In an embodiment of the present invention, during the training process of the convolutional neural network model, the Adam optimization algorithm is used to optimize the model parameters, and Smooth L1 Loss is used as the loss function. After the training is completed, the model that achieves the prediction accuracy can be used as the final power system transient stability judgment model and put into the reasoning application of actual scenarios.

[0044] S3. Obtain system fault information, system steady-state operation information, and post-fault generator power angle trajectory change information of the target power system at the operating point to be predicted, and input them into the power system transient stability judgment model to obtain a transient stability margin prediction value of the target power system.

[0045] It should be noted that the processing requirements for input data in the above-mentioned power system transient stability judgment model are completely consistent with the aforementioned training samples, and will not be elaborated. Among them, the operating point to be predicted is the operating point in the power system where transient stability judgment is required. Based on the system conditions of this operating point, three input data can be determined (among which, the information on the change of the generator power angle trajectory after the fault needs to be obtained through transient stability simulation, but only 1s of simulation is required, and there is no need to perform complex and time-consuming simulation within 1 to 10s), and then the transient stability margin prediction value is output through the model, thereby realizing fast and efficient transient stability margin calculation, and providing technical support for emergency control of transient stability in power systems.

[0046] In order to better demonstrate the specific implementation method and technical effect of the present invention, a specific embodiment is used below to demonstrate the application of the power system transient stability determination method based on multi-source data hybrid drive shown in S1~S3 in actual scenarios, and to demonstrate the achievable prediction performance.

[0047] Example

[0048] In this embodiment, a method for determining transient stability of a power system based on hybrid drive of multi-source data is provided, which includes the following three stages:

[0049] Phase 1: Data generation and sample construction

[0050] The power system of this embodiment includes 34 lines and a total of 10 generators. Therefore, in order to obtain a transient stability data sample data set covering most of the key disturbance scenarios of the power system, transient stability time domain simulation is used to generate sample data. By setting key fault scenarios, the PSAT transient stability simulation software is used to simulate the change trajectory of the system's disturbed state variables, and the transient stability coefficient TSI corresponding to the label is calculated. After generating a sufficient amount of original sample data with a reasonable distribution, the input features of the network are selected and extracted, and the input features that best reflect the transient stability characteristics and the corresponding stable characteristic label values ​​are selected, and the input and output data structures are determined. Finally, the sample data is preprocessed, and the data is normalized while controlling the ratio of stable and unstable samples to be within a reasonable range to ensure the validity of the sample data. The process of generating samples at this stage is as follows: Figure 2 As shown, the specific description is as follows.

[0051] Step 1: Key fault scenario setting: Set the system load level k to increase gradually in steps of 5% from 80% to 130%. A total of 11 central load levels are sampled, namely k = 80%, 85%, 90%, 95%, 100%, 105%, 110%, 115%, 120%, 125%, and 130%. Around each central load level k, the fault line fl, fault duration ft, fault location fp, and fault type required for simulation are obtained by random sampling without replacement, where: the fault line fl is randomly selected from 34 lines; the fault duration ft is randomly selected from 0.1s, 0.15s, and 0.2s; the fault location fp is randomly selected from 20%, 50%, and 80%; the fault type is uniformly set to a three-phase short circuit fault, and the load level value ki is randomly sampled in the neighborhood range (k-2.5%, k+2.5%) of the central load level k by Latin hypercube sampling, so that each group of sampled values ​​is composed of a system fault information, which can be used as a steady-state operation sample for subsequent simulation. Therefore, each central load level k can generate (34*3*3) steady-state operation samples of all fault combinations. For 11 central load levels k, a total of (11*34*3*3) steady-state operation samples can be obtained.

[0052] Step 2: Extract key features and determine input and output data structure: Before transient stability simulation, first select the fault removal time ft, fault location percentage fp, fault line fl and load level ki from the system fault information. These key parameters together constitute the fault information matrix; then focus on the system steady-state operation information, extract the core parameters such as generator active power PG, generator reactive power QG, machine-end bus voltage amplitude V, machine-end bus voltage phase angle θ, generator power angle δ and generator online status Ol, and integrate the active and reactive loads PL and QL of the machine-end line to construct a steady-state feature matrix. During the transient stability simulation process, each sample is set up in the PSAT transient stability simulation software according to the above system fault information and system steady-state operation information. The target power system is simulated with load level ki as the initial operating point, and the dynamic changes of the generator power angle δ are focused on during the simulation. By extracting the power angle trajectory change information within 1 second after the fault occurs, the trajectory feature matrix is ​​constructed. This matrix records the change trajectory of the generator power angle in detail during the transient process, which is of great significance for analyzing the transient stability of the system. After the transient stability simulation, the system transient stability margin result TSI is retained as the sample label information for convolutional neural network model training. TSI reflects the maximum power angle difference Δ between all generators in the power system during the transient process. max , then TSI = (360-Δ max ) / (360+Δ max)*100, where the maximum power angle difference is the maximum power angle difference between all generators within 10s after the start of the simulation. After extracting the above key features, the system fault information, system steady-state operation information, post-fault generator power angle trajectory change information and transient stability margin corresponding to each set of simulation results constitute a simulation data sample. In this embodiment, the input and output structure of the subsequent neural network is determined as shown in Table 1:

[0053] Table 1 Input and output data structure

[0054]

[0055] It should be noted that the above-mentioned system fault information, system steady-state operation information, and generator power angle trajectory change information after the fault all need to be constructed into a matrix form before they can be input into the neural network, that is, they need to be converted into a fault information matrix, a steady-state feature matrix, and a trajectory feature matrix.

[0056] Step 3: Sample data preprocessing: Each simulation data sample is processed to form a training sample set that meets the neural network model input.

[0057] On the one hand, it is necessary to extract key feature data including generator active power PG, generator power angle δ, etc., and perform Z-Score normalization:

[0058]

[0059] Among them, μ and σ are the mean and standard deviation of the sample data respectively. This normalization method will make the processed data load standard normal distribution (mean is 0, standard deviation is 1), which can maintain the valuable information in the original data and reduce the sensitivity to outliers.

[0060] On the other hand, it is necessary to classify the samples into stable samples and unstable samples according to the stability result TSI, and then perform sample balancing to control the proportion of the training sample set so that the number of stable samples and the number of unstable samples reach a reasonable level. Therefore, after completing the above two aspects of processing, a training sample set that meets the input of the neural network model is obtained.

[0061] Phase 2: Convolutional Neural Network Model Structure Construction

[0062] According to the scale and structure of the power system to be evaluated, a multi-source data hybrid-driven convolutional neural network model structure for transient stability margin assessment is constructed based on deep learning technology. The input layer of the network directly sets the convolution layer as the first layer of the network according to the preprocessed training set data format. Then, the pooling layer and the convolution layer are arranged alternately to gradually extract the dynamic features of the input data; the convolution layer is responsible for feature extraction, and the pooling layer is used to retain the most representative features. At the same time, a suitable activation function is added between the convolution layer and the pooling layer to enhance the nonlinear feature expression ability of the model. In the fully connected layer part, a structure with multiple layers of fully connected layers and activation functions alternating is adopted to learn more abstract and higher-level features. Finally, a layer of Tanh activation function is added to limit the output value to a specific range, while further improving the nonlinear expression ability of the model.

[0063] like Figure 3 As shown in FIG. 1 , the specific network structure of the CNN convolutional neural network model in this embodiment is shown, which is composed of three feature extraction branches and a fully connected network. The system fault information, the system steady-state operation information, and the generator power angle trajectory change information after the fault are each extracted through a feature extraction branch. Feature information, each feature extraction branch is composed of alternating convolution layers and pooling layers cascaded together. After each convolution layer, a ReLU function can be used for output, and the pooling layer can use the maximum pooling MaxPool, that is, each feature extraction branch uses convolution layer 1, ReLU function layer 1, pooling layer 1, and convolution layer 2. The cascade structure of layer 2, ReLU function layer 2, pooling layer 2, convolution layer 3, ReLU function layer 3, and pooling layer 3; the feature information extracted by the three feature extraction branches is fully connected and activated through a fully connected network. The fully connected network is composed of three cascaded fully connected layers, in which the first two fully connected layers are output through the ReLU activation function, and the last fully connected layer is output through the tanh activation function, that is, the fully connected layer 1, ReLU function layer 4, fully connected layer 2, ReLU function layer 5, fully connected layer 3 and tanh activation function layer are cascaded. The tanh activation function layer finally outputs the transient stability margin prediction value of the target power system. In this network model, after preprocessing, the fault information and steady-state feature information can be directly spliced ​​with the trajectory feature information after the convolution and pooling operation, and jointly input into the fully connected layer and fused in the fully connected layer to effectively combine the system state information before and after the fault occurs and the dynamic trajectory information after the fault occurs, so that it can more accurately capture and analyze the complex characteristics of the transient stability process of the power system.

[0064] In this embodiment, the network parameters of the convolution layer, pooling layer, fully connected layer and tanh activation function layer in the CNN convolutional neural network structure are shown in Table 2.

[0065] Table 2 CNN convolutional neural network structure

[0066]

[0067] Phase 3: Convolutional Neural Network Model Training and Optimization

[0068] Before model training, we first construct training sets and test sets based on the preprocessed training sample data, and then randomly shuffle them and load them in batches according to the batch_size to prepare for subsequent training and testing. Then, we select the Smooth L1 Loss described in the following formula as the loss function to evaluate the model performance, so as to effectively balance the loss calculation of outliers and general situations, and use the Adam optimizer to update the model parameters to minimize the loss function and converge to the optimal solution quickly and stably:

[0069]

[0070] in, is the label value of the transient stability margin, is the model estimate of the transient stability margin. When the difference between the two is small, the loss function is equivalent to using L2Loss; when the difference is large, the loss function is the translation of L1Loss.

[0071] The specific model training and testing process in this embodiment is as follows Figure 4 As shown. The model is continuously monitored and updated through the test set and training set, so that the model can gradually adapt to different data distributions and scene changes, thereby ensuring the effectiveness and generalization ability of the model. During the training process, strategies such as Early Stopping are also used to avoid overfitting and further improve the performance of the model. The final model test results are shown in Figure 5 As shown in the figure, the convolutional neural network model predicts the transient stability coefficient TSI of the system 10s later based on multi-source information input within 1s, with an accuracy of 92.7%, indicating that the proposed power system transient stability judgment method based on multi-source data hybrid drive can effectively replace the complex and time-consuming electromechanical transient simulation process and meet the needs of rapid response in actual power system stability control.

[0072] The above-described embodiment is only a preferred solution of the present invention, but it is not intended to limit the present invention. A person skilled in the relevant technical field may make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, any technical solution obtained by equivalent replacement or equivalent transformation falls within the protection scope of the present invention.

Claims

1. A method for determining transient stability of a power system based on hybrid drive of multi-source data, characterized in that: Here are the steps: S1. For the target power system, the initial operation point of the target power system under different load levels is determined based on Latin hypercube sampling, transient stability simulation is performed in combination with different system fault information, and the transient stability margin is calculated according to the simulation results, thereby generating simulation data samples consisting of system fault information, system steady-state operation information, post-fault generator power angle trajectory change information and transient stability margin, and each simulation data sample is processed to form a training sample set that meets the input of the neural network model; S2. Constructing a convolutional neural network model with system fault information, steady-state operation characteristics and trajectory change characteristics as input and transient stability margin as output, and training the convolutional neural network model using the training sample set to obtain a transient stability judgment model for the power system; S3, obtaining system fault information, system steady-state operation information and post-fault generator power angle trajectory change information of the target power system at the operation point to be predicted, and inputting them into the power system transient stability judgment model to obtain a transient stability margin prediction value of the target power system; In step S1, the method for generating simulation data samples is as follows: S11. uniformly sample multiple central load levels within the load level range of the target power system; for each central load level, obtain the fault line, fault duration, fault location, and fault type required for the simulation through random sampling without replacement, and randomly sample load level values ​​within the neighborhood of the central load level through Latin hypercube sampling, so that each sampled value forms system fault information; S12. For each group of system fault information composed of sampling values, the load level value is used as the initial operating point of the target power system, and the corresponding fault line, fault duration, fault location and fault type are set in the simulated target power system, while retaining the system steady-state operation information of each generator in the target power system, the system steady-state operation information includes the active power of the generator, the reactive power of the generator, the amplitude of the bus voltage at the machine end, the phase angle of the bus voltage at the machine end, the power angle of the generator, the online status of the generator, the active load of the machine end line and the reactive load of the machine end line, and then the transient stability simulation of the target power system is performed to obtain the trajectory change information of the generator power angle δ of each generator after the fault occurs and the maximum power angle difference between all generators, and finally convert the maximum power angle difference between all generators into the transient stability margin as the sample label value; The load level range is 80%~130%, the sampling range of the fault duration is 0.1~0.2s, the sampling range of the fault position is 20%~80%, the fault type is only sampled as a three-phase short circuit fault, the neighborhood range of each central load level is ±2.5% centered on itself, the trajectory change information is the trajectory change information of the generator power angle 1s after the fault occurs; the maximum power angle difference is the maximum power angle difference between all generators within 10s after the start of the simulation; The transient stability margin is based on the maximum power angle difference Δ between all generators. max The conversion formula is 100*(360-Δ max ) / (360+Δ max ); The convolutional neural network model consists of three feature extraction branches and a fully connected network. The system fault information, the system steady-state operation information, and the generator power angle trajectory change information after the fault are each extracted through a feature extraction branch. Each feature extraction branch is composed of alternating convolutional layers and pooling layers cascaded. The feature information extracted by the three feature extraction branches passes through the fully connected network and outputs the transient stability margin prediction value of the target power system; Each of the feature extraction branches includes three convolution layers with ReLU functions and three pooling layers, and the convolution layers and the pooling layers are alternately arranged; The fully connected network is composed of three cascaded fully connected layers, wherein the first two fully connected layers output through the ReLU activation function, and the last fully connected layer outputs through the tanh activation function.

2. The method for determining transient stability of a power system based on multi-source data hybrid drive according to claim 1, characterized in that: In S1, when processing the simulation data samples, it is necessary to perform Z-Score normalization and sample balancing processing on the data.

3. The method for determining transient stability of a power system based on multi-source data hybrid drive according to claim 1, characterized in that: During the training process of the convolutional neural network model, the Adam optimization algorithm is used to optimize the model parameters, and Smooth L1 Loss is used as the loss function.

4. The method for determining transient stability of a power system based on multi-source data hybrid drive according to claim 1, characterized in that: The transient stability simulation is implemented by PSAT transient stability simulation software.

Images