Transient voltage stability evaluation method based on deep belief network

Abstract

A transient voltage stability evaluation method based on a deep belief network comprises the following steps: step 1, acquiring system operation data based on historical operation data of a power system and analog simulation of a fault, and forming an initial sample set; 2, based on the initial sample set obtained in the step 1, performing normalization processing on the initial sample set by adopting a Z-Score standardization method, performing dimension reduction on data, generating an efficient sample set, and randomly dividing the efficient sample set into a training set and a test set according to a certain proportion; 3, constructing a deep belief network model based on a transient voltage stability margin index so as to represent a mapping relation between an input variable and an output variable, performing optimization by adopting an improved particle swarm optimization algorithm, and training the DBN model based on the preprocessing sample set obtained in the step 2 so as to find an optimal weight and an optimal threshold; and 4, performing parameter setting on the DBN model based on the optimal weight and the threshold obtained in the step 3, then testing the DBN by using the preprocessed test set, and judging the transient voltage stability of the power system.

Description

Technical Field

[0001] The present invention relates to the fields of transient voltage stability and new - energy voltage correlation, and particularly relates to a transient voltage stability assessment method based on a deep belief network. Technical Background

[0002] New energy has entered a rapid development stage, and the impact of new - energy access on the stability of the power grid must be considered. The access of new energy such as wind energy and solar energy poses higher challenges to the transient voltage stability of the power system. The transient voltage stability problem is one of the important considerations in the planning, operation, and design of the power system. Continuous disturbances in the power system will cause transient voltage instability and sometimes even lead to the failure of the entire system. In order to ensure the safe operation of the system, it is of great significance to evaluate the transient voltage stability of the system.

[0003] The patent document with the application publication number CN106208052A discloses a method for identifying weak points of power - grid transient voltage stability based on transient voltage stability limit tests. The specific steps are as follows: when simulating the power grid, a three - phase permanent short - circuit fault is applied at transient voltage sensitive points, and the fault - clearing time is continuously increased until the transient voltage of the system changes from stable to unstable; the document with the application publication number CN114626757A discloses a method and system for discriminating the voltage stability of a receiving - end system. The specific steps are as follows: obtaining the transient voltage stability data of the bus, segmenting the voltage curve using a difference equation. Combining with the existing practical criteria for transient voltage, different weights are assigned according to the different degrees of voltage drop. The cumulative amount of voltage drop is quantified by using the method of weighted integration. An evaluation index is constructed to form a complete discrimination method; the document with the application publication number CN110909795A discloses a method for determining the transient voltage stability of a power grid. The specific steps are as follows: constructing a sample library, classifying and labeling the samples, calculating the morphological similarity distance of each electrical quantity. Using a decision - tree algorithm to obtain an initial transient voltage determination model, and optimizing the initial transient voltage determination model to obtain the final transient voltage determination model. Real - time monitoring of the electrical data volume in the power grid, and when the power grid is disturbed or fails, using the final transient voltage determination model to determine the transient voltage stability of the power grid.

[0004] In summary, although the existing methods can evaluate the transient voltage stability of the power system, there are still problems in aspects such as solution time and solution accuracy. To solve these problems, the present invention proposes a transient voltage stability assessment method based on a deep belief network. Summary of the Invention

[0005] The object of the present invention is to solve the technical problems existing in the prior art pointed out in the background art, namely, complex numerical simulation and a large amount of calculations in the prior art, which result in a long time-consuming and inaccurate transient voltage stability assessment, and it is difficult to meet the requirements of real-time assessment. A transient voltage stability assessment method based on a deep belief network is proposed, which significantly improves the efficiency and accuracy of the assessment by combining deep learning and optimization algorithms.

[0006] To achieve the above-mentioned invention object, the technical solution proposed by the present invention is as follows:

[0007] A transient voltage stability assessment method based on a deep belief network, comprising the following steps:

[0008] Step 1: Based on the historical operation data of the power system and the simulation of faults, obtain the system operation data to form an initial sample set;

[0009] Step 2: Based on the initial sample set obtained in Step 1, use the Z-Score normalization method to normalize it, reduce the dimension of the data, generate an efficient sample set, and randomly divide it into a training set and a test set according to a certain proportion;

[0010] Step 3: Based on the transient voltage stability margin (TVSM) index, construct a deep belief network (DBN) model to represent the mapping relationship between input variables and output variables, and use an improved particle swarm optimization (PSO) algorithm for optimization. Based on the preprocessed sample set obtained in Step 2, train the DBN model to find the optimal weights and thresholds;

[0011] Step 4: Based on the optimal weights and thresholds obtained in Step 3, set the parameters of the DBN model, and then use the preprocessed test set to test the DBN to judge the transient voltage stability of the power system.

[0012] In Step 1, based on the historical operation data of the power system and the simulation of faults, consider the impact of the access of new energy on the transient voltage stability of the power system, obtain the system operation data, and form an initial sample set.

[0013] In Step 2, based on the initial sample set obtained in Step 1, use the Z-Score normalization method to normalize it, reduce the dimension of the data, generate an efficient sample set, and randomly divide it into a training set and a test set according to a certain proportion. The specific steps are as follows:

[0014] Step 2-1: For each feature in the sample set and σ, use the Z-Score formula for standardization.

[0015] Among them, the Z-Score normalization is defined as:

[0016]

[0017] In the formula: is the average value; σ is the standard deviation; n is the number of initial sample sets.

[0018] Step 2-2: Use the principal component analysis method to map the high-dimensional data of the initial sample set into a low-dimensional space, so as to reduce the dimension of the standardized data, and then obtain an efficient sample set;

[0019] Step 2-3: Randomly divide the efficient sample set into a training set and a test set, with 80% of the data as the training set and 20% of the data as the test set.

[0020] In Step 3, based on the Transient Voltage Stability Margin (TVSM) index, construct a Deep Belief Network (DBN) model to represent the mapping relationship between input variables and output variables, and use an improved Particle Swarm Optimization (PSO) algorithm for optimization. Based on the preprocessed sample set obtained in Step 2, train the DBN model to find the optimal weights and thresholds. The specific steps are as follows:

[0021] Step 3-1: The transient voltage stability index can be used to study the transient voltage stability of the power system. Use TVSM to detect the key buses in the power system to obtain the transient voltage stability margin of the power system. The specific steps are as follows:

[0022] Step 3-1-1: Use the π model in the power system to determine the key buses that have a greater impact on the system voltage stability;

[0023] Step 3-1-2: Build a dynamic model and perform time-domain simulation, and record the voltage change curve of the key buses over time;

[0024] Step 3-1-3: Calculate the TVSM index of the key buses, and then judge the transient voltage stability of the system through the calculated TVSM index.

[0025] Among them, the transient voltage stability margin index is as follows:

[0026]

[0027] In the formula: X sr is the line reactance from the sending end to the receiving end; Q ris the reactive power at the receiving end; U s is the voltage at the sending end; θ is the line impedance angle; δ is the voltage angle difference between the sending end and the receiving end.

[0028] Step 3-2: Based on the TVSM index obtained in Step 3-1, build a DBN model with an output layer of 2. The input layer of this model receives the operating variables of the power system, and the output layer judges the transient voltage stability of the system according to the calculated TVSM value. The closer the TVSM value is to 0, the more stable the system is.

[0029] Step 3-3: Based on the DBN model obtained in Step 3-3, use the improved PSO algorithm to optimize it, complete the training process and find the optimal weights and thresholds. The specific steps are as follows:

[0030] Step 3-3-1: The improved PSO algorithm mainly uses the tangent function to transform the constant inertia weight into a non-linear inertia weight function:

[0031]

[0032] In the formula: ω s is the initial inertia weight; ω e is the termination inertia weight; t is the current iteration number; t max is the maximum iteration number; k is the control factor.

[0033] Step 3-3-2: The weights and thresholds of the DBN model are represented as the position vectors of the particles. Through the training set obtained above, use the improved PSO algorithm to optimize the DBN model;

[0034] Step 3-3-3: Solve the fitness value of each particle, continuously update the position of the particle, find the global optimal solution, and then continuously iterate until the maximum iteration number is reached. The final global optimal position is the optimal weights and thresholds of the DBN model.

[0035] In Step 4, based on the optimal weights and thresholds obtained in Step 3, set the parameters of the DBN model, and then use the preprocessed test set to test the DBN to judge the transient voltage stability of the power system. The specific steps are as follows:

[0036] Step 4-1: Assign the optimal weights and threshold parameters of the DBN model represented by the position vector of the global optimal particle obtained in Step 3 to the DBN model;

[0037] Step 4-2: Use the test set obtained above that contains the operating variables of the power system and the corresponding TVSM values, input its features into the DBN model, and obtain the classification of the transient voltage stability of the system;

[0038] Step 4-3: Use four metrics, namely accuracy (A), recall (R), precision (P), and F1-score (F1), to measure the performance of the DBN model.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0040] 1. To address the impact of new energy on the transient voltage stability of the power system, the present invention adopts a DBN model and proposes a simple, fast, and computationally feasible evaluation method for the transient voltage stability of the power system.

[0041] 2. When optimizing the DBN model, the present invention uses an improved particle swarm optimization algorithm. Starting from a single particle, it iteratively searches for the optimal value by changing the direction, speed, and position of the particle, without the need to use the gradient information of the objective function. It has a wide range of applications, is simple and easy to implement, and has high computational efficiency.

[0042] 3. The Z-Score normalization method is used to normalize the initial sample set, generating an efficient sample set, standardizing the data, improving the data comparability, saving the training time, and enhancing the training efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The present invention will be further described below in conjunction with the drawings and embodiments:

[0044] Figure 1 is the flowchart of the method of the present invention;

[0045] Figure 2 is the deep belief network model diagram of the present invention;

[0046] Figure 3 is the restricted Boltzmann machine structure diagram of the present invention;

[0047] Figure 4 is the schematic diagram of the IEEE 30-node system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0048] The present invention provides a transient voltage stability evaluation method based on a deep belief network. The method includes the following steps, as Figure 1 shown:

[0049] Step 1: Based on the historical operation data of the power system and the simulation of faults, obtain the system operation data and form an initial sample set.

[0050] Step 2: Based on the initial sample set obtained in Step 1, use the Z-Score normalization method to normalize it, reduce the dimension of the data, generate an efficient sample set, and randomly divide it into a training set and a test set according to a certain ratio.

[0051] Step 3: Construct a deep belief network (DBN) model based on the transient voltage stability margin (TVSM) index to represent the mapping relationship between input variables and output variables. Use an improved particle swarm optimization (PSO) algorithm for optimization. Based on the preprocessed sample set obtained in Step 2, train the DBN model to find the optimal weights and thresholds.

[0052] Step 4: Set the parameters of the DBN model based on the optimal weights and thresholds obtained in Step 3, and then use the preprocessed test set to test the DBN to judge the transient voltage stability of the power system.

[0053] In Step 1, based on the historical operation data of the power system and the simulation of faults, considering the impact of new energy access on the transient voltage stability of the power system, obtain the system operation data to form an initial sample set.

[0054] In Step 2, based on the initial sample set obtained in Step 1, use the Z-Score normalization method to normalize it, reduce the dimension of the data, generate an efficient sample set, and randomly divide it into a training set and a test set according to a certain proportion.

[0055] Data normalization is a data preprocessing technique that sets a new data range for the existing data to make it fall into a small specific interval, eliminates the influence of data dimension and dimension unit, improves the accuracy of the results, and extracts valuable information from the data. The specific steps are as follows:

[0056] Step 2-1: For each feature in the sample set and σ are standardized using the Z-Score formula.

[0057] The Z-Score normalization is defined as:

[0058]

[0059] In the formula: is the average value; σ is the standard deviation; n is the number of the initial sample set.

[0060] Step 2-2: Use the principal component analysis method to map the high-dimensional data of the initial sample set to a low-dimensional space to reduce the dimension of the standardized data, and then obtain an efficient sample set;

[0061] Step 2-3: Randomly divide the efficient sample set into a training set and a test set, with 80% of the data as the training set and 20% of the data as the test set.

[0062] The role of the training set is to train the parameters of the network model so that the model can fit the characteristics and laws of the data; the role of the test set is to judge the performance of the model on an untrained data set and to check whether the model can be fully fitted.

[0063] In step 3, a DBN model is constructed based on the TVSM index to represent the mapping relationship between the input variables and the output variables, and an improved PSO algorithm is used for optimization. Based on the preprocessed sample set obtained in step 2, the DBN model is trained to find the optimal weights and thresholds. The specific steps are as follows:

[0064] Step 3-1: The transient voltage stability margin index is a quantitative index used to evaluate whether the power system can maintain voltage stability after a fault. The purpose is to determine the critical values and safety values of the voltage amplitude, active power, and reactive power in each load bus. The numerical index starts from 0 and goes up to 1. A numerical index of 0 indicates that the power system is in a good and stable state, and an index of 1 indicates that the power system is in a bad and unstable state. This index can be applied to multiple power grid distribution systems for future planning, control systems, and optimization tasks. The specific steps are as follows:

[0065] Step 3-1-1: Use the π model in the power system to determine the key buses that have a greater impact on the system voltage stability;

[0066] Step 3-1-2: Build a dynamic model and conduct time-domain simulation, and record the voltage variation curve of the key buses over time;

[0067] Step 3-1-3: Calculate the TVSM index of the key buses, and then judge the transient voltage stability of the system through the calculated TVSM index.

[0068] The transient voltage stability margin index is as follows:

[0069]

[0070] In the formula: X sr is the line reactance from the sending end to the receiving end; Q r is the reactive power at the receiving end; U s is the sending-end voltage; θ is the line impedance angle; δ is the voltage angle difference between the sending end and the receiving end.

[0071] To ensure the stable operation of the power system, the TVSM value of the power system needs to be kept within a range far less than 1. When the TVSM value approaches 1, it indicates that the line is close to instability, and in severe cases, it may lead to voltage collapse in the power system.

[0072] Step 3-2: Based on the TVSM index obtained in Step 3-1, build a DBN model with an output layer of 2. The input layer of this model receives the operating variables of the power system, and the output layer determines the transient voltage stability of the system according to the calculated TVSM value. The closer the TVSM value is to 0, the more stable the system is.

[0073] The basic element of the deep belief network is the restricted Boltzmann machine (RBM). The model structure of the deep belief network model is as Figure 2 shown, and the structure of the RBM is as Figure 3 shown. The neurons between different network layers of the multi-layer deep belief network are fully connected, while the neurons in the same layer are not connected. The deep belief network consists of multiple RBMs, and each RBM can perform the function of feature extraction. Different RBMs are combined with each other to form a feature extractor with better performance, which can perform higher-level feature extraction.

[0074] The neurons in the deep belief network are divided into two categories: visible neurons and hidden neurons. During the data processing process, the hidden neurons play the role of feature extraction. Let the state of the visible neurons be Y, and the state of the hidden neurons be L. The energy state can be expressed as:

[0075]

[0076] In the formula: Y n is the number of visible neurons; H m is the number of hidden neurons; I a is the bias of the a-th visible neuron; J b is the bias of the b-th hidden neuron; ζ ab is the weight between the a-th visible neuron and the b-th hidden neuron; Y a and S b are the existence forms of visible and hidden neurons.

[0077] By using the energy function to add all the energies together, the energy between the RBM models can be obtained, and then the joint probability of the corresponding visible layer and hidden layer can be calculated. Its expression is as follows:

[0078]

[0079] The partition function can be expressed as:

[0080]

[0081] In the formula: ψ is the sum of all possible values that visible neurons and hidden neurons can take.

[0082] After dividing the visible layer and the hidden layer, the distributed probability for the visible layer is obtained, and its expression is:

[0083]

[0084] Step 3-3: Based on the DBN model obtained in Step 3-2, use the improved PSO algorithm to optimize it, complete the training process, and find the optimal weights and thresholds.

[0085] In Step 3-3, use the improved PSO to optimize the DBN model, complete the training process, and find the optimal weights and thresholds. The specific steps are as follows:

[0086] Step 3-3-1: The PSO algorithm is considered one of the most important swarm intelligence methods. The inspiration for this algorithm comes from social behaviors in animal groups, such as bird flocks or fish schools. That is, starting from a particle, the optimal value is iteratively searched by changing the direction, speed, and position of the particle.

[0087] To update the position and speed of the particle, the formulas are as follows:

[0088]

[0089] In the formula: v i (t) and y i (t) are the speed and position of particle i in the t-th generation; W best,i and G best,i are the individual optimal position of particle i and the global optimal position of the population; x1 and x2 are the individual learning factor and swarm learning factor that determine W best,i and G best,i ; n1 and n2 are random numbers between [0, 1].

[0090] The improved PSO algorithm mainly uses the tangent function to transform the constant inertia weight into a non-linear inertia weight function:

[0091]

[0092] In the formula: ω s is the initial inertia weight; ω e is the termination inertia weight; t is the current iteration number; t max is the maximum iteration number; k is the control factor.

[0093] Step 3-3-2: Represent the weights and thresholds of the DBN model as the position vector of the particle. Using the training set obtained above, use the improved PSO algorithm to optimize the DBN model;

[0094] Step 3-3-3: Solve the fitness value of each particle, continuously update the position of the particle, find the global optimal solution, and then continuously iterate until the maximum iteration number is reached. The final global optimal position is the optimal weights and thresholds of the DBN model.

[0095] In step 4: Based on the optimal weights and thresholds obtained in step 3, set the parameters of the DBN model, and then use the preprocessed test set to test the DBN to judge the transient voltage stability of the power system. The specific steps are as follows:

[0096] Step 4-1: Assign the optimal weights and threshold parameters of the DBN model represented by the position vector of the global optimal particle obtained in step 3 to the DBN model;

[0097] Step 4-2: Use the above-obtained test set containing the operating variables of the power system and the corresponding TVSM values, input its features into the DBN model, and obtain the classification of the transient voltage stability of the system;

[0098] Step 4-3: Use four indicators, namely accuracy (A), recall (R), precision (P), and F1-score (F1), to measure the performance of the DBN model.

[0099]

[0100] The performance of the model is measured by four indicators: accuracy (A), recall (R), precision (P), and F1-score (F1). A is the primary indicator to measure the performance of the model, which represents the accuracy of model evaluation; R represents the proportion of correctly classified samples among all unstable samples, and the larger its value, the less likely the model is to miss unstable alarms; P is used to reflect the probability of false unstable alarms. The F1-score is the harmonic mean of precision and recall, which can more objectively evaluate the performance of the model.

[0101] After training the DBN transient voltage stability evaluation model using the training set, use the test set to test the model, evaluate the stability of the transient voltage, and verify the effectiveness of the model.

[0102] Example:

[0103] To verify the effectiveness of the present invention, tests are carried out on the IEEE 30-bus system. The schematic diagram of the IEEE 30-bus system is as Figure 4 shown. In the IEEE 30-bus system, the synchronous generators at nodes 1, 5, 8, and 13 are replaced with wind farms (WFs) based on doubly-fed induction generators. The efficiency and accuracy of the deep belief network optimized by the improved particle swarm algorithm in transient voltage stability evaluation are verified through simulation.

[0104] After the DBN model is trained with the training set, the DBN model is tested with the test set. The experimental results are verified by the Continuation Power Flow (CPF) method based on the TVSM index. The TVSM values of each bus of the IEEE 30-node system measured by the two methods are obtained, and the maximum value bus and the minimum value bus are obtained by comparing the TVSM values of each bus. Table 1 shows the TVSM minimum value bus and the maximum value bus measured by the two methods.

[0105] Table 1

[0106]

[0107] As can be seen from the above table, comparing CPF as a benchmark with the method proposed in the present invention, the trained DBN gives accurate TVSM values. The TVSM value of the starting node 9 and the ending node 10 is the maximum value, which is closer to 1 than other buses, and it is closer to transient voltage instability. The TVSM value of the starting node 25 and the ending node 27 is lower, indicating that the bus has stronger transient voltage stability. Comparing the method proposed in the present invention with the CPF method, using the deep belief network to evaluate the voltage stability of the power system is not only fast but also more accurate.

[0108] Table 2

[0109]

[0110] Comparing the performance of the trained DBN model with the ANN, RF and CNN models, it can be seen from Table 2 that the DBN model has a higher accuracy rate. The F1 score can be regarded as the weighted average of the model accuracy rate and the recall rate. The DBN model has the highest F1 score and the best comprehensive performance of the model.

[0111] A transient voltage stability evaluation method based on the deep belief network proposed in the present invention is of great significance for ensuring the safe and stable operation of the power system.

Claims

1. A transient voltage stability assessment method based on a deep belief network, characterized in that, It includes the following steps: Step 1: Based on the historical operation data of the power system and the simulation of faults, obtain the system operation data and form an initial sample set; Step 2: Based on the initial sample set obtained in Step 1, use the Z-Score normalization method to normalize it, reduce the dimension of the data, generate an efficient sample set, and randomly divide it into a training set and a test set according to a certain proportion; Step 3: Construct a deep belief network DBN model based on the transient voltage stability margin index to represent the mapping relationship between input variables and output variables, use an improved particle swarm optimization algorithm for optimization, and based on the preprocessed sample set obtained in Step 2, train the DBN model to find the optimal weights and thresholds; Step 4: Set the parameters of the DBN model based on the optimal weights and thresholds obtained in Step 3, and then use the preprocessed test set to test the DBN to judge the transient voltage stability of the power system; The above 4 steps can be used to well perform the transient voltage stability assessment based on the deep belief network.

2. The method according to claim 1, wherein: In Step 2, based on the initial sample set obtained in Step 1, use the Z-Score normalization method to normalize it, reduce the dimension of the data, generate an efficient sample set, and randomly divide it into a training set and a test set according to a certain proportion.

3. The method according to claim 2, wherein: In Step 2, it specifically includes the following steps: Step 2-1: Standardize and σ for each feature in the sample set using the Z-Score formula; where the Z-Score normalization is defined as: In the formula: is the average value; σ is the standard deviation; n is the number of initial sample sets; Step 2-2: Use the principal component analysis method to map the high-dimensional data of the initial sample set into a low-dimensional space to reduce the dimension of the standardized data, and then obtain an efficient sample set; Step 2-3: Randomly divide the efficient sample set into a training set and a test set.

4. The method according to claim 1, characterized in that: In Step 3, construct a deep belief network DBN model based on the transient voltage stability margin TVSM index to represent the mapping relationship between input variables and output variables, use an improved particle swarm optimization PSO algorithm for optimization, and based on the preprocessed sample set obtained in Step 2, train the DBN model to find the optimal weights and thresholds.

5. The method according to claim 4, wherein: In Step 3, it specifically includes the following steps: Step 3-1: Use TVSM to detect the key buses in the power system to obtain the transient voltage stability margin of the power system; Step 3-2: Based on the TVSM index obtained in Step 3-1, construct a DBN model to represent the mapping relationship between input variables and output variables; Step 3-3: Based on the DBN model obtained in Step 3-2, use an improved PSO algorithm to optimize it, complete the training process and find the optimal weights and thresholds.

6. The method according to claim 5, wherein: In Step 3-1, the transient voltage stability index can be used to study the transient voltage stability of the power system. Use TVSM to detect the key buses in the power system to obtain the transient voltage stability margin of the power system.

7. The method according to claim 6, wherein: Specifically, it includes the following steps: Step 3-1-1: Use the π model in the power system to determine the key buses that have a greater impact on the system voltage stability; Step 3-1-2: Build a dynamic model and perform time-domain simulation, and record the voltage change curve of the key buses over time; Step 3-1-3: Calculate the TVSM index of the key bus, and then judge the transient voltage stability of the system based on the calculated TVSM index; The transient voltage stability margin index is as follows: Where: X sr is the line reactance from the sending end to the receiving end; Q r is the reactive power at the receiving end; U s is the sending-end voltage; θ is the line impedance angle; δ is the voltage angle difference between the sending end and the receiving end.

8. The method according to claim 5, characterized in that: In Step 3-2, based on the TVSM index obtained above, a DBN model with an output layer of 2 is built. The input layer of this model receives the operating variables of the power system, and the output layer judges the transient voltage stability of the system according to the calculated TVSM value. The closer the TVSM value is to 0, the more stable the system is.

9. The method according to claim 5, wherein: In Step 3-3, the improved PSO is used to optimize the DBN model, complete the training process and find the optimal weights and thresholds; it specifically includes the following steps: Step 3-3-1: The improved PSO algorithm mainly uses the tangent function to transform the constant inertia weight into a non-linear inertia weight function: Where: ω s is the initial inertia weight; ω e is the termination inertia weight; t is the current iteration number; t max is the maximum iteration number; k is the control factor; Step 3-3-2: The weights and thresholds of the DBN model are represented as the position vector of the particle. Through the training set obtained above, the improved PSO algorithm is used to optimize the DBN model; Step 3-3-3: Solve the fitness value of each particle, continuously update the position of the particle, find the global optimal solution, and then continuously iterate until the maximum number of iterations is reached. The final global optimal position is the optimal weights and thresholds of the DBN model.

10. The method according to claim 1, wherein: In Step 4, based on the optimal weights and thresholds obtained in Step 3, the parameters of the DBN model are set, and then the preprocessed test set is used to test the DBN to judge the transient voltage stability of the power system; It specifically includes the following steps: Step 4-1: Assign the optimal weights and threshold parameters of the DBN model represented by the position vector of the global optimal particle obtained in Step 3 to the DBN model; Step 4-2: Use the test set obtained above that contains the operating variables of the power system and the corresponding TVSM values, input its features into the DBN model, and obtain the classification of the transient voltage stability of the system; Step 4-3: Use four indicators, namely accuracy A, recall R, precision P, and F1-score F1, to measure the performance of the DBN model.

Images