A hierarchical prediction method for emergency control of power systems based on post-control power flow states.

Abstract

This invention relates to the field of power system operation and control technology, specifically to a hierarchical prediction method for emergency control in power systems based on post-control power flow states. The method includes: calculating the post-control power flow states of centralized samples in a simulation dataset, prioritizing them, and generating a priority sequence of control measures; generating several sample sets based on the feature dimensions corresponding to the types of emergency control measures; and using electromechanical transient data during a fault to output predicted emergency control measures level by level. This invention, based on the post-control power flow states, can accurately reflect the impact of different control measures on system stability, improving the scientific rigor and real-time performance of emergency control decisions. By extracting corresponding feature dimensions based on the types of emergency control measures and constructing a multi-model system, it achieves collaborative modeling of multi-dimensional features and control types, enhancing the relevance and accuracy of predictions. Through a hierarchical prediction mechanism, optimal control schemes can be output level by level, improving response efficiency under complex operating conditions.

Description

Technical Field

[0001] This invention relates to the field of power system operation and control technology, specifically to a hierarchical prediction method for emergency control of power systems based on the power flow state after control. Background Technology

[0002] With the continuous expansion of the power system and the large-scale integration of new energy sources, the operating environment of the power grid is becoming increasingly complex. When encountering faults or disturbances, the power system is more prone to transient instability phenomena, such as generator loss of synchronism, system voltage collapse, and large-scale power angle oscillations. If these unsafe conditions are not effectively controlled in a timely manner, they will lead to serious consequences such as system disconnection and large-scale power outages.

[0003] Traditional emergency control measures in power systems mainly include disconnecting generators and disconnecting loads. These methods typically determine emergency control schemes based on fixed rules or offline simulation results, lacking dynamic adaptability to real-time operating conditions. Especially when facing multiple fault scenarios and complex operating conditions, traditional control strategies often struggle to balance control effectiveness and economy, resulting in problems such as unreasonable measure selection, over-control, or under-control.

[0004] In recent years, with the development of artificial intelligence and data-driven technologies, researchers have begun to try to apply machine learning methods to predict emergency control of power system faults. However, existing research mainly focuses on the prediction of single-measure types and has not yet formed a complete hierarchical prediction and prioritization framework, which is difficult to support the actual needs of power systems for rapid and accurate control. Summary of the Invention

[0005] (a) Purpose of the invention

[0006] The purpose of this invention is to provide a hierarchical prediction method for emergency control of power systems based on the post-control power flow state, which overcomes the problems of existing power system emergency control measure prediction methods, such as limited measure selection, lack of hierarchical mechanism, and insufficient adaptability to complex operating conditions, and ensures that the power system can recover to a safe and stable operating state under sudden disturbances.

[0007] (II) Technical Solution

[0008] To address the above problems, this invention provides a hierarchical prediction method for emergency control of power systems based on post-control power flow states, comprising:

[0009] The power flow state of the central samples in the simulation dataset after emergency control measures are applied is calculated.

[0010] The power flow states are prioritized to generate a priority sequence of control measures corresponding to the samples in the simulation dataset;

[0011] Based on the feature dimensions corresponding to the types of emergency control measures, features of samples in the simulation dataset are extracted to generate several sample sets.

[0012] Based on the sample set of emergency control measures and corresponding feature dimensions, train and generate the optimal prediction model corresponding to the emergency control measures type;

[0013] Using electromechanical transient data during a fault, and based on the priority of the corresponding emergency control measure type and the optimal prediction model corresponding to the emergency control measure type, the predicted emergency control measures are output step by step.

[0014] The emergency control measures include a preset set of emergency control measures, which includes several types of emergency control measures, and each type of emergency control measure includes several emergency control measures.

[0015] In another aspect of the present invention, the method for obtaining the simulation dataset includes:

[0016] Based on different fault scenarios, electromechanical transient simulations of the power system are performed to obtain the initial simulation dataset;

[0017] Based on a preset transient safety criterion, the initial simulation data is analyzed to obtain a simulation dataset, which consists of transiently unsafe samples.

[0018] The transient simulation refers to performing electromechanical transient simulations under anticipated faults to verify the transient stability of the power system.

[0019] In another aspect of the present invention, the preset transient safety criterion includes: power angle stability criterion and voltage safety criterion;

[0020] The power angle stability criterion is used to determine the relative power angle change between generators after a fault disturbance. When the maximum power angle difference between any two generators within the set time window before the end of the simulation exceeds the preset power angle threshold, it is determined that the power angle is unstable.

[0021] The voltage safety criterion is used to evaluate the transient recovery characteristics of the bus voltage. When the amplitude of any bus voltage is lower than a preset proportion of the rated voltage and continues to exceed a set time threshold, it is determined that the voltage is unsafe.

[0022] When the power angle is unstable or the voltage is unsafe, it is a sample that is transiently unsafe.

[0023] In another aspect of the present invention, the power angle stability criterion and voltage safety criterion are expressed using the following formulas:

[0024] ;

[0025] ;

[0026] in, and Indicates the first Taiwan and the The rotor power angle of the generator, Indicates the total number of generators; Indicates the total number of busbars. This represents the voltage amplitude of the b-th bus at time t. Indicates duration.

[0027] In another aspect of the present invention, the current flow state is calculated using the following formula:

[0028] ;

[0029] in, Indicates the total number of busbars. Indicates the number of samples in the simulation dataset at the 1st... The voltage value of the b-th bus under the emergency control measures Indicates the rated voltage. This indicates the total number of emergency control measures. Represented as the first The current state of the emergency control measures, i.e., the first The average deviation of the bus voltage for each emergency control measure is considered, with smaller voltage deviations indicating higher priority for the emergency control measure.

[0030] In another aspect of the present invention, the feature dimensions include time dimension data and spatial dimension data. The time dimension includes data before the fault, during the fault, and after the fault is cleared. The spatial dimension includes the power angle, speed, electromagnetic power of all generators, and voltage amplitude of all bus lines.

[0031] In another aspect of the present invention, the prediction model includes at least one of: support vector machine, logistic regression, decision tree, random forest, long short-term memory network, K-nearest neighbor algorithm, Gaussian Naive Bayes, multilayer perceptron and linear discriminant analysis.

[0032] In another aspect of the present invention, the priority of the corresponding emergency control measure type is obtained through the following steps:

[0033] The first power flow state after calculating the electromechanical transient data during the fault and the preset emergency control measures is controlled.

[0034] Prioritize the emergency control measures based on the first power flow state to obtain the priority of the emergency control measures type corresponding to the electromechanical transient data at the time of the fault.

[0035] In another aspect of the present invention, the step of using electromechanical transient data during a fault to output predicted emergency control measures in stages, based on the priority of the corresponding emergency control measure type and the optimal prediction model corresponding to the emergency control measure type, includes:

[0036] Based on the priority of the aforementioned emergency control measures types, multiple optimal prediction models are ranked;

[0037] The electromechanical transient data during the fault is input into the highest priority optimal prediction model for prediction, and the output result is obtained.

[0038] If the output result is empty, the electromechanical transient data at the time of the fault is input into the optimal prediction model of the next priority until the output result is not empty;

[0039] Perform simulation verification on the non-empty output results to obtain the simulation verification results;

[0040] Based on the simulation verification results, the feasibility of the output results is obtained;

[0041] Control measures are implemented based on the feasibility results of the output.

[0042] In another aspect of the present invention, control is performed based on the feasibility result of the output result, including:

[0043] Based on the simulation verification results and the preset power grid security criteria, the feasibility results of the output results are obtained;

[0044] If the feasibility result indicates stable control, then control shall be implemented based on the output result.

[0045] If the feasibility result is unstable, the electromechanical transient data at the time of failure is input into the next priority optimal prediction model and simulated for verification until the feasibility result is stable. Control is then implemented based on the output result.

[0046] (III) Beneficial Effects

[0047] The above-described technical solution of the present invention has the following beneficial technical effects:

[0048] This invention, based on the post-control power flow state, accurately reflects the impact of different control measures on system stability, improving the scientific rigor and real-time performance of emergency control decisions. By extracting corresponding feature dimensions based on the type of emergency control measure and constructing a multi-model system, it achieves collaborative modeling of multi-dimensional features and control types, enhancing the relevance and accuracy of predictions. Through a hierarchical prediction mechanism, it can output optimal control schemes at each level, improving response efficiency under complex operating conditions. It can quickly identify effective control measures in the early stages of a fault, reducing the risk of system instability and significantly improving the operational safety and disturbance resistance of the power system, demonstrating good engineering applicability and promotional value. Attached Figure Description

[0049] Figure 1 This is an overall flowchart of one embodiment of the present invention;

[0050] Figure 2 This is a topology diagram of an IEEE-39 node computational example according to an embodiment of the present invention;

[0051] Figure 3 This is a flowchart of a power system transient security criterion according to an embodiment of the present invention;

[0052] Figure 4 These are sampling points of input features in the time dimension of an embodiment of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be understood that these descriptions are merely exemplary and not intended to limit the scope of the invention. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0054] Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0055] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0056] The invention will now be described in more detail with reference to the accompanying drawings. In the various drawings, the same elements are indicated by similar reference numerals. For clarity, the various parts in the drawings are not drawn to scale.

[0057] Example 1

[0058] A hierarchical prediction method for emergency control of power systems based on post-control power flow state. Figure 1 An overall flowchart of one embodiment of the present invention is shown; Figure 2 The diagram illustrates the topology of an IEEE-39 node computational example according to an embodiment of the present invention, as follows: Figure 1 and Figure 2 As shown, this embodiment uses the IEEE-39-bus system to study the transient stability of the power system, and its topology is as follows. Figure 2 As shown. The system consists of 10 generators, G0 to G9 representing the generators, 46 transmission lines, and 19 loads. Generator G1 is set as the balancing node, and the excitation system adopts the IEEE DC Exciter Type 1 model.

[0059] Hierarchical prediction methods include:

[0060] Based on different fault scenarios, electromechanical transient simulations of the power system are performed to obtain an initial simulation dataset. Since the main factors affecting the transient stability of the power system include fault type, fault branch, fault location, fault clearing time, and load level, this embodiment performs batch settings for these factors during the simulation process, generating a set of initial simulation data. A number of transient simulation samples were used to form the initial simulation dataset. The specific simulation parameters are configured as shown in Table 1.

[0061] Table 1. Specific Simulation Parameter Configuration

[0062]

[0063] Based on the preset transient safety criteria, the initial simulation data is analyzed to obtain a simulation dataset, which is a sample of transient insecurity; the transient simulation refers to performing electromechanical transient simulation under the expected fault to verify the transient stability of the power system.

[0064] Figure 3 A flowchart of a power system transient security criterion according to an embodiment of the present invention is shown, as follows: Figure 3 As shown, in this embodiment, the preset transient safety criteria include: power angle stability criteria and voltage safety criteria;

[0065] The power angle stability criterion is used to determine the relative power angle change between generators after a fault disturbance. When the maximum power angle difference between any two generators within the set time window before the end of the simulation exceeds the preset power angle threshold, it is determined that the power angle is unstable.

[0066] The voltage safety criterion is used to evaluate the transient recovery characteristics of the bus voltage. When the amplitude of any bus voltage is lower than a preset proportion of the rated voltage and continues to exceed a set time threshold, it is determined to be voltage unsafe. When the power angle is unstable or the voltage is unsafe, it is considered a transiently unsafe sample. Specifically, within 2 seconds before the simulation ends, if the maximum power angle difference between any two generators is greater than 180 degrees, it is determined to be power angle unstable; if the amplitude of any bus voltage remains less than or equal to 0.80 pu for more than or equal to 400 ms, it is determined to be voltage unsafe. The power angle stability criterion and the voltage safety criterion are expressed using the following formulas:

[0067] ;

[0068] ;

[0069] in, and Indicates the first Taiwan and the The rotor power angle of the generator, Indicates the total number of generators; Indicates the total number of busbars. This represents the voltage amplitude of the b-th bus at time t. Indicates duration.

[0070] The simulation data was analyzed and judged based on transient safety criteria. Out of 14,850 simulation samples, 10,761 samples were judged as transiently safe, while the remaining 4,089 samples were judged as transiently unsafe. For the samples judged as transiently unsafe, emergency control measures were implemented to restore the power system to normal operation.

[0071] The power flow state of the samples in the simulation dataset after emergency control measures are applied is calculated. The power flow state is then prioritized to generate a priority sequence of control measures corresponding to the samples in the simulation dataset. The preset set of emergency control measures includes several emergency control measure types, and each emergency control measure type includes several emergency control measures. Emergency control measure types include: cutting off one generator, cutting off two generators, cutting off one generator and one load node, cutting off two generators and one load node, and combinations like cutting off one load node (cutting off generators, cutting off loads, and combinations of both). Emergency control measures under each emergency control measure type, such as cutting off generators, include: cutting off one generator (cutting off generator 1, generator 2, ..., or generator N); cutting off two generators: a combination of two generators from a set of generators prioritized by control measures; cutting off k generators: a combination of k generators from a set of generators prioritized by control measures, etc.

[0072] In this embodiment, the power flow state is represented by the average deviation of the bus voltage, and the power flow state is calculated using the following formula:

[0073] ;

[0074] in, Indicates the total number of busbars. Indicates the number of samples in the simulation dataset at the 1st... The voltage value of the b-th bus under the emergency control measures Indicates the rated voltage. This indicates the total number of emergency control measures. Represented as the first The current state of the emergency control measures, i.e., the first The average deviation of the bus voltage for each emergency control measure is considered, with smaller voltage deviations indicating higher priority for the emergency control measure.

[0075] Based on the feature dimensions corresponding to the emergency control measure types, features of the samples in the simulation dataset are extracted to generate several sample sets. The feature dimensions include time dimension data and spatial dimension data. The time dimension includes sample data before, during, and after the fault is cleared. The spatial dimension includes the power angle, speed, electromagnetic power of all generators, and voltage amplitude of all bus lines. Figure 4 This illustrates sampling points of input features in the time dimension according to an embodiment of the present invention, such as... Figure 4 As shown, in the time dimension, the transient security of a power system mainly involves three stages: pre-fault, during-fault, and post-fault clearance. The pre-fault stage reflects the system's safety margin, the during-fault stage reflects the severity of the fault, and the post-fault clearance stage reflects the role of the system's first line of defense. To quantitatively analyze the different contributions of these three stage predictions to emergency control measures, the above three stages are combined into four different input feature sequences, and each input feature sequence is modeled separately.

[0076] In the spatial dimension, for the IEEE-39 node power system, the core of transient security assessment lies in determining the changes in the maximum power angle difference and bus voltage amplitude during transient processes. Therefore, the power angle and bus voltage amplitude constitute key characteristics. The change in power angle is influenced by factors such as rotational speed, electromagnetic power, and mechanical power, and is expressed as:

[0077] ;

[0078] ;

[0079] in, Indicates the angle of attack; Indicates rotational speed; Indicates mechanical power; Indicates electromagnetic power; Indicates the damping coefficient; This represents the time constant.

[0080] In time-domain simulation, electromagnetic power It can be represented as:

[0081] ;

[0082] in, and They are shaft voltage and Shaft voltage; and They are shaft current and Axis current.

[0083] Because the transient process develops very rapidly, we assume that the mechanical power remains constant during this process and neglect the damping coefficient. When a system fault occurs, the electromagnetic power changes with variations in voltage and current. The imbalance between electromagnetic power and mechanical power will cause speed fluctuations, thereby inducing power angle oscillations. Therefore, the transient safety of the system is closely related to both the speed and the electromagnetic power.

[0084] Table 2 shows the simulation dataset information. As shown in Table 2, the features of the time dimension and the spatial dimension are combined. The time series of four input features and the four different spatial dimension input features are combined as input features, and emergency control measures are used as labels.

[0085] Table 2 Simulation Dataset

[0086]

[0087] Based on a sample set of emergency control measure types and corresponding feature dimensions, an optimal prediction model corresponding to each emergency control measure type is trained and generated. In this embodiment, multi-dimensional feature data from the sample set is input into multiple prediction models, where each optimal prediction model corresponds to an input feature and a corresponding emergency control measure type. Each input feature consists of a sample set composed of multiple feature dimensions to achieve prediction for different measure types. The prediction models include at least one of the following: Support Vector Machine, Logistic Regression, Decision Tree, Random Forest, Long Short-Term Memory Network, K-Nearest Neighbors Algorithm, Gaussian Naive Bayes, Multilayer Perceptron, and Linear Discriminant Analysis.

[0088] Table 3 shows the classification accuracy of commonly used machine learning algorithms. As shown in Table 3, the classification accuracy of nine commonly used machine learning algorithms in transient safety prediction using a single electrical quantity feature is compared. The results show that Random Forest (RF), based on a tree model, performs best on all four feature classes, with an accuracy of around 85%, demonstrating strong feature adaptability and robustness. Decision Tree (DT) and K-Nearest Neighbors (KNN) algorithms are next, with accuracies of around 80%, showing above-average predictive ability. Multilayer Perceptron (MLP) achieves high accuracy on electromagnetic power and positive-sequence voltage features, while Support Vector Machine (SVM) and Logistic Regression (LR) perform moderately on different features. Long Short-Term Memory (LSTM) has relatively low accuracy in single feature prediction, especially on electromagnetic power and positive-sequence voltage, showing its limited effectiveness when sequence information is insufficient. Gaussian Naive Bayes (GNB) performs the worst overall, showing particularly low sensitivity to velocity deviation and electromagnetic power. Linear Discriminant Analysis (LDA) has low to medium accuracy on all four feature classes.

[0089] Table 3 Classification accuracy of commonly used machine learning algorithms

[0090]

[0091] To further consider the impact of time series information on prediction performance, a random forest model was used for evaluation under different time dimension feature combinations. Table 4 shows the classification accuracy of feature combinations under different time dimensions. As shown in Table 4, the prediction accuracy using only post-fault data is low; combining pre-fault and post-fault data slightly improves the accuracy, with the accuracy of electromagnetic power increasing from 85.13% to 86.93%; using a combination of during and after the fault does not significantly change the accuracy; however, combining data from the pre-fault, during, and post-fault stages further improves the accuracy of features such as power angle and electromagnetic power to approximately 86–87%. This demonstrates that fully utilizing dynamic information from before, during, and after a fault can effectively enhance the reliability of transient safety prediction, especially for electromagnetic power features, where the improvement is most significant.

[0092] Table 4. Classification accuracy of feature combinations at different time dimensions

[0093]

[0094] Using electromechanical transient data during a fault, and based on the priority of the corresponding emergency control measure type and the optimal prediction model corresponding to the emergency control measure type, the predicted emergency control measures are output step by step.

[0095] The priority of the corresponding emergency control measure type is obtained through the following steps:

[0096] The first power flow state after calculating the electromechanical transient data during the fault and the preset emergency control measures is controlled.

[0097] Prioritize the emergency control measures based on the first power flow state to obtain the priority of the emergency control measure type corresponding to the electromechanical transient data at the time of the fault. The first power flow state is calculated using the power flow state formula described above.

[0098] Furthermore, in this embodiment, the step of utilizing electromechanical transient data during a fault, and outputting predicted emergency control measures step by step according to the priority of the corresponding emergency control measure type and the optimal prediction model corresponding to the emergency control measure type, includes:

[0099] Based on the priority of the emergency control measure types, multiple optimal prediction models are ranked, forming a model invocation sequence;

[0100] The electromechanical transient data during the fault is input into the highest priority optimal prediction model for prediction, and the output result is obtained.

[0101] If the output result is empty, the electromechanical transient data at the time of the fault is input into the optimal prediction model of the next priority until the output result is not empty;

[0102] The non-empty output results are simulated and verified to obtain the simulation verification results. The simulation verification step aims to verify the feasibility and stability of the prediction results under specific power grid topology, operating conditions, and dynamic constraints. By substituting the predicted emergency control measures into the system electromechanical transient simulation module, the power angle stability, voltage stability, and frequency response characteristics of the system after control are calculated to obtain the simulation verification results.

[0103] Based on the simulation verification results, the feasibility of the output results is obtained;

[0104] Control measures are implemented based on the feasibility results of the output.

[0105] Furthermore, in this embodiment, control is performed based on the feasibility results of the output, including:

[0106] Based on the simulation verification results and the preset power grid security criteria, the feasibility of the output result is obtained. The preset power grid security criteria may include the power angle difference limit, the lower limit of voltage amplitude, and the frequency offset threshold, etc., to determine the feasibility of the predicted output and obtain the feasibility of the output result.

[0107] If the feasibility result indicates stable control, then control is implemented based on the output result. When the feasibility result indicates stable control, the output result can be used as the final control command and sent to the scheduling execution layer to complete the implementation of emergency control measures.

[0108] If the feasibility result is unstable, the electromechanical transient data at the time of the fault is input into the next priority optimal prediction model and simulated for verification until the feasibility result is control stable. Based on the output result, control is implemented. The system will automatically call the next priority optimal prediction model and re-perform prediction and simulation verification until a prediction output of "control stable" is obtained. Finally, based on the feasible output result, the corresponding emergency control operation is executed to achieve rapid restoration of power grid transient safety. The dynamic decision-making mechanism based on priority-level prediction and simulation verification not only ensures the physical feasibility and control stability of the prediction results, but also significantly improves the response efficiency and reliability of emergency control under multiple measures and multiple models.

[0109] This invention, based on the post-control power flow state, accurately reflects the impact of different control measures on system stability, improving the scientific rigor and real-time performance of emergency control decisions. By extracting corresponding feature dimensions based on the type of emergency control measure and constructing a multi-model system, it achieves collaborative modeling of multi-dimensional features and control types, enhancing the relevance and accuracy of predictions. Through a hierarchical prediction mechanism, it can output optimal control schemes at each level, improving response efficiency under complex operating conditions. It can quickly identify effective control measures in the early stages of a fault, reducing the risk of system instability and significantly improving the operational safety and disturbance resistance of the power system, demonstrating good engineering applicability and promotional value.

[0110] It should be understood that the specific embodiments described above are merely illustrative or explanatory of the principles of the invention and do not constitute a limitation thereof. Therefore, any modifications, equivalent substitutions, improvements, etc., made without departing from the spirit and scope of the invention should be included within the protection scope of the invention. Furthermore, the appended claims are intended to cover all variations and modifications falling within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries.

[0111] The present invention has been described above with reference to embodiments thereof. However, these embodiments are merely illustrative and not intended to limit the scope of the invention. The scope of the invention is defined by the appended claims and their equivalents. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of the invention, and all such substitutions and modifications should fall within the scope of the invention.

[0112] Although embodiments of the present invention have been described in detail, it should be understood that various changes, substitutions, and modifications can be made to the embodiments of the present invention without departing from the spirit and scope of the invention.

[0113] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A hierarchical prediction method for emergency control of a power system based on post-control power flow state, characterized in that, The method comprises the following steps: calculating the power flow state of the simulation data set sample after control by the emergency control measure; prioritizing the power flow state to generate a control measure priority sequence corresponding to the simulation data set sample; extracting the features of the simulation data set sample according to the feature dimensions corresponding to the type of emergency control measure to generate a plurality of sample sets; training an optimal prediction model corresponding to the type of emergency control measure based on the sample set of the type of emergency control measure and the corresponding feature dimension; outputting the predicted emergency control measure step by step according to the priority of the corresponding type of emergency control measure and the optimal prediction model corresponding to the type of emergency control measure by using the electromechanical transient data at the time of fault; wherein the emergency control measure comprises a preset emergency control measure set, and the preset emergency control measure set comprises a plurality of types of emergency control measures, and each type of emergency control measure comprises a plurality of emergency control measures; the power flow state is calculated by the following formula: ； wherein, represents the total number of buses, represents the voltage value of the bth bus under the th emergency control measure in the sample of the simulation data set, represents the rated voltage, represents the total number of emergency control measures, represents the power flow state of the th emergency control measure, i.e. the average deviation of the bus voltage of the th emergency control measure, and the smaller the voltage deviation, the higher the priority of the emergency control measure; the priority of the corresponding type of emergency control measure is obtained by the following steps: calculating the first power flow state after the electromechanical transient data at the time of fault is controlled by the preset emergency control measure set; prioritizing the first power flow state to obtain the priority of the type of emergency control measure corresponding to the electromechanical transient data at the time of fault; the output of the predicted emergency control measure step by step according to the priority of the corresponding type of emergency control measure and the optimal prediction model corresponding to the type of emergency control measure by using the electromechanical transient data at the time of fault comprises: sorting a plurality of optimal prediction models based on the priority of the type of emergency control measure; inputting the electromechanical transient data at the time of fault into the optimal prediction model with the highest priority for prediction to obtain an output result; if the output result is empty, inputting the electromechanical transient data at the time of fault into the optimal prediction model with the next priority until the output result is not empty; performing simulation checking on the output result that is not empty to obtain a simulation checking result; obtaining the feasibility result of the output result according to the simulation checking result; controlling according to the feasibility result of the output result.

2. The power system emergency control hierarchical prediction method based on post-control power flow state according to claim 1, characterized in that, The method for obtaining the simulation data set comprises: performing electromechanical transient simulation on the power system based on different fault conditions to obtain an initial simulation data set; analyzing the initial simulation data based on a preset transient security criterion to obtain a simulation data set, which is a sample of transient instability; the transient simulation refers to electromechanical transient simulation under a contingency to check the transient stability of the power system.

3. The power system emergency control hierarchical prediction method based on post-control power flow state according to claim 2, characterized in that, the preset transient security criterion comprises a power angle stability criterion and a voltage security criterion; the power angle stability criterion is used to determine the relative power angle change between generators after fault disturbance. When the maximum power angle difference between any two generators within a set time window before the end of simulation exceeds a preset power angle threshold, it is determined that the power angle is unstable; the voltage security criterion is used to evaluate the transient recovery characteristics of bus voltage. When the voltage amplitude of any bus is lower than the preset proportion of the rated voltage and lasts for more than a set time threshold, it is determined that the voltage is unsafe; When the power angle is unstable or the voltage is unsafe, the sample is transiently unsafe.

4. The power system emergency control hierarchical prediction method based on post-control power flow state according to claim 3, characterized in that, The power angle stability criterion and the voltage safety criterion are expressed by the following formulas: ； ； wherein, and denotes the rotor angle of the station and the rotor angle of the station generator, denotes the total number of generators; denotes the total number of buses, denotes the voltage amplitude of the b-th bus at time t, denotes the duration.

5. The power system emergency control hierarchical prediction method based on post-control power flow state according to claim 1, characterized in that, The characteristic dimensions include time dimension and spatial dimension data, the time dimension includes data before the fault, during the fault and after the fault is removed, and the spatial dimension includes power angles, rotating speeds, electromagnetic powers of all generators and voltage amplitudes of all buses.

6. The power system emergency control hierarchical prediction method based on post-control power flow state according to claim 1, characterized in that, The prediction model includes at least one of support vector machine, logistic regression, decision tree, random forest, long short-term memory network, K-nearest neighbor algorithm, Gaussian naive Bayes, multi-layer perceptron and linear discriminant analysis.

7. The power system emergency control hierarchical prediction method based on post-control power flow state according to claim 1, characterized in that, According to the feasibility result of the output result, control is performed, including: According to the simulation checking result and a preset power grid safety criterion, a feasibility result of the output result is obtained; If the feasibility result is control stable, control is performed according to the output result; If the feasibility result is unstable, electromechanical transient data at the time of the fault is input to an optimal prediction model of next priority, and simulation checking is performed until the feasibility result is control stable, and control is performed according to the output result.

Images