Method and system for extracting transient stability key response characteristics of sending-end system
By constructing an equivalent negative impedance model for new energy generating units and using the Relief algorithm to screen features, the problem of incomplete information extraction after new energy grid connection in large power grids is solved, improving the accuracy and efficiency of transient stability assessment and adapting to the complex characteristics of the power grid.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-24
- Publication Date
- 2026-03-13
AI Technical Summary
Existing technologies struggle to effectively handle the complex characteristics of new energy sources after grid integration in large power grids, resulting in incomplete information extraction and affecting the accuracy and efficiency of transient stability assessment.
By constructing an equivalent negative impedance model of new energy units and combining it with the Relief algorithm for feature dimensionality reduction, key transient stability response features of the sending-end system are extracted, highly correlated key features are selected, and invalid or weakly discriminative features are eliminated.
It improves the computational accuracy and prediction efficiency of transient stability assessment models, clarifies the influencing factors of new energy grid connection on system power angle stability, and adapts to the complex characteristics of large power grids.
Smart Images
Figure CN121663465A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of power system operation and control technology, specifically to a method and system for extracting key transient stability response features of a sending-end system. Background Technology
[0002] The widespread integration of renewable energy and the commissioning of multiple ultra-high-voltage direct current (UHVDC) transmission projects have significantly altered the operating characteristics of power systems. Increased uncertainty between power sources and loads, coupled with the high proportion of power electronic devices, profoundly impacts the dynamic characteristics of the power grid. Under the characteristics of "large-scale DC transmission, large-scale renewable energy, and weak AC transmission," AC / DC system faults have a tremendous impact on the power grid, shrinking its operational safety domain, highlighting safety and stability issues, and increasing the risk of cascading failures. Therefore, there is an urgent need for an effective strategy to rapidly implement transient stability assessment (TSA).
[0003] With the development of big data theory and artificial intelligence technology, and the rapid popularization of measurement equipment, data-driven artificial intelligence methods have made it possible to achieve rapid transient stability assessment. However, due to the large scale and numerous components of actual power systems, high-dimensional electrical characteristics can lead to redundancy among input features, severely affecting the performance and prediction efficiency of assessment models. In recent years, in order to fully explore the key features under system disturbances, dynamic time series data have been applied to transient stability analysis, which increases the computational load and the risk of overfitting. Therefore, exploring suitable feature selection schemes to reduce the dimensionality of original features is a key issue in the application of artificial intelligence methods in the transient voltage stability assessment of power systems.
[0004] Traditional methods typically involve manually selecting key features, relying on expert experience to choose variable features for stability analysis. However, due to the unclear fault characteristics of new equipment and the unclear transient instability mechanisms, relying on manual expertise to select variable features is no longer suitable for the complex characteristics of large power grids and is prone to information omissions. Summary of the Invention
[0005] In order to overcome the shortcomings of the existing technology, the purpose of this invention is to provide a method and system for extracting key transient stability response features of the sending-end system, so as to solve the technical problem of how to achieve complete information extraction from the complex characteristics of large power grids.
[0006] This invention is achieved through the following technical solution: In a first aspect, the present invention provides a method for extracting key transient stability response features of a sending-end system, comprising: A mathematical model for grid-connected power transmission of new energy units is constructed by simplifying the negative impedance of the new energy units. Based on the mathematical model, the impact of new energy access on the transient characteristics of the system is analyzed from the perspective of transient energy, and the key factors affecting the stability level of the system's power angle are extracted. Based on the aforementioned key factors, an original feature set is constructed, and the Relief algorithm is used to reduce the dimensionality of the features to obtain the optimal feature subset.
[0007] Preferably, the new energy unit adopts the equivalent negative impedance model, including the equivalent negative impedance model of wind power. Z wr and photovoltaic negative impedance model Z phr Based on the wind power negative impedance model Z wr and photovoltaic negative impedance model Z phr Negative impedance of equivalent renewable energy power station Z r The negative impedance Z r The formula is as follows:
[0008]
[0009] in, P r and Q r These are the active power and reactive power output by the new energy power station, respectively. U This refers to the bus voltage at the new energy grid connection point.
[0010] Preferably, the mathematical model for grid-connected power transmission from new energy sources is the electromagnetic power-power angle equation of the regional traditional generating units, expressed as follows:
[0011] Among them, Z 11 Z 12 These are the self-impedance of the traditional unit and the mutual impedance with the receiving-end system, respectively. , These are the self-impedance angle and the mutual impedance angle, respectively. This refers to the internal potential phasor of the synchronous generator; E1 is the conjugate phasor of the output current phasor of the synchronous generator; E2 is the internal potential phasor of the synchronous generator. The amplitude of E1; E2 is the amplitude of the equivalent potential of the receiving-end system; For the angle of attack.
[0012] Furthermore, the self-impedance Z 11 and mutual impedance Z 12 The expression is:
[0013] Where x1 and x2 are the system reactances, U rFor the voltage at the grid connection point of new energy sources, Z r It is a negative impedance; P r and Q r These represent the active power and reactive power output of the new energy power station, respectively; j is the imaginary unit.
[0014] Preferably, the key factors affecting the stability of the system power angle include the ratio of new energy and conventional units connected to the grid at the sending end, the location of the new energy grid connection point, and the active power recovery rate of the new energy units during a fault.
[0015] Furthermore, by analyzing the acceleration and deceleration power of the system before and after the integration of new energy sources, the key factors are obtained. The active power of the new energy unit during a fault satisfies the following expression:
[0016] in, t f Indicates the time when the fault occurred; t c Indicates the fault clearing time; parameter k Used to indicate the recovery rate of the unit's active power output.
[0017] Preferably, the feature weight calculation formula for the Relief algorithm is as follows:
[0018]
[0019] in, Let j be the j-th dimension feature of the sample; It is a difference function; Let i be the value of the i-th sample in the j-th dimension feature; The closest sample among samples of the same type. The closest sample among samples of different classes; and The values of sample k and sample s in the j-th dimension feature represent the values of sample k and sample s.
[0020] Preferably, the specific process of feature dimensionality reduction using the Relief algorithm is as follows: Calculate the weights of each input feature of the sample, average the weights of all sample features, and, given a threshold, remove features with small weights or weak correlations to obtain the optimal feature subset, thus achieving feature dimensionality reduction.
[0021] Preferably, the original feature set is constructed based on the ratio of new energy and conventional units connected to the sending end, the location of the new energy grid connection point, and the active power recovery rate of the new energy units during the fault.
[0022] Secondly, the present invention also provides a system for extracting key transient stability response features of a sending-end system, used to implement the above-described method for extracting key transient stability response features of a sending-end system, comprising: The model building module is used to simplify the negative impedance of new energy units and build a mathematical model for the grid-connected transmission of new energy. The factor extraction module is used to analyze the impact of new energy access on the transient characteristics of the system from the perspective of transient energy, based on the mathematical model, and to extract the key factors affecting the stability level of the system's power angle. The feature extraction module is used to construct an original feature set based on the key factors, and to perform feature dimensionality reduction using the Relief algorithm to obtain the optimal feature subset.
[0023] Compared with the prior art, the present invention has the following beneficial technical effects: This invention provides a method for extracting key transient stability response features of a power transmission system. Based on the response characteristics of renewable energy units, it constructs an equivalent negative impedance model and a grid-connected transmission mathematical model, overcoming the limitations of the ambiguity in the renewable energy grid connection mechanism in traditional analysis. By comparing the changes in system power angle and acceleration / deceleration power characteristics before and after renewable energy access from a transient energy perspective, it accurately extracts key influencing factors such as the ratio of renewable energy to conventional units, the location of the grid connection point, and the active power recovery rate during faults. This provides a theoretical basis for the synchronous stability assessment under large-scale renewable energy grid connection and clarifies the core direction of feature extraction from the source, avoiding the blindness of traditional manual feature selection. This invention constructs an original feature set based on the above key influencing factors and uses the Relief algorithm to achieve feature dimensionality reduction. This algorithm assigns weights to the distinguishing ability of similar / dissimilar samples by quantifying features, which can accurately screen highly correlated key features and effectively eliminate invalid / weakly discriminative features, solving the curse of dimensionality problem introduced by dynamic time series data and significantly improving the computational accuracy and prediction efficiency of the stability assessment model. Attached Figure Description
[0024] Figure 1 This is a flowchart of the method for extracting key transient stability response features of the sending-end system in an embodiment of the present invention; Figure 2 This is a schematic diagram of a power transmission system including wind power and photovoltaic power in an embodiment of the present invention; Figure 3 This is a schematic diagram of the equivalent network of the regional new energy unit power transmission system in an embodiment of the present invention; Figure 4 This is a flowchart of the feature selection process based on the Relief algorithm in an embodiment of the present invention; Figure 5 This is a schematic diagram of the 750kV network structure in northern Shaanxi Province in an embodiment of the present invention. Figure 6This is a schematic diagram of the power angle response curve of the sending-end system in an embodiment of the present invention; Figure 7 This is a normalized feature weight distribution diagram of the original feature set in an embodiment of the present invention; Figure 8 This is a histogram showing the frequency distribution of feature weights in an embodiment of the present invention. Figure 9 This is a schematic diagram of the system for extracting key transient stability response features of the sending-end system in an embodiment of the present invention; In the diagram: 1. Model building module; 2. Factor extraction module; 3. Feature extraction module. Detailed Implementation
[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0027] The purpose of this invention is to provide a method and system for extracting key transient stability response features of a sending-end system, so as to solve the technical problem of how to achieve complete information extraction from the complex characteristics of a large power grid.
[0028] The present invention will now be described in further detail with reference to the accompanying drawings: See Figure 1 This invention provides a method for extracting key transient stability response features of a sending-end system, comprising: Step 1: Simplify the negative impedance of the new energy unit and construct a mathematical model for the grid connection and transmission of new energy. Specifically, considering the response characteristics of new energy units under normal and transient conditions, a simplified equivalent of new energy grid connection was performed, and a mathematical model for new energy grid connection and power transmission was constructed.
[0029] by Figure 2 Taking the sending-end system containing wind power and photovoltaics as an example, the synchronous generator adopts the second-order classical model, and the wind turbine model adopts the mainstream doubly-fed wind turbine.
[0030] In the sending-end system, the bus voltage at the grid connection point of the renewable energy source is U. Considering the active power generated by the renewable energy source under normal operating conditions using maximum power point tracking (MPPT), it can be equivalently represented as a negative resistance model. During a fault, according to low voltage ride-through control, the active power generated by the renewable energy source decreases during the transient process, and a certain amount of reactive power is generated. This can be equivalently represented as a negative resistance and negative reactance model. Therefore, the renewable energy source can be represented as a negative impedance model. Simultaneously, considering the negative impedance model of the wind turbine... Z wr The photovoltaic negative impedance model is Z phr Then its parallel connection points to the grid are: For equivalent new energy power stations, the following are: (1) (2) In the formula, P r and Q r These represent the active and reactive power outputs of the renewable energy power plants, respectively. Ignoring the resistance in transmission lines and transformers, the regional renewable energy power transmission system can be simplified as follows: Figure 3 As shown. The electromagnetic power-power angle equation for traditional generating units in the region before the integration of new energy sources can be expressed as: (3) according to Figure 3 For a medium-value circuit, writing the loop current equation, the electromagnetic power-power angle equation for a conventional regional generator unit is: (4) In the formula, Z 11 Z 12 These are the self-impedance of the traditional unit and the mutual impedance with the receiving-end system, respectively. , These are the self-impedance angle and the mutual impedance angle, respectively. The self-impedance and mutual impedance are expressed as follows: (5) As can be seen from (4), the increase in regional renewable energy power will reduce the static stability limit of traditional units.
[0031] Step 2: Based on the mathematical model, analyze the impact of new energy access on the transient characteristics of the system from the perspective of transient energy, and extract the key factors affecting the stability level of the system's power angle. Specifically, based on the equivalent model of the sending end for new energy access, this study investigates the impact mechanism of new energy unit grid connection on the transient characteristics of the system, analyzes the changes in the operating state and transient characteristics of the system before and after new energy access from the perspective of transient energy, and extracts the key factors affecting the stability level of the system's power angle.
[0032] Before the new energy source is connected to the system, it is assumed that the output of the traditional unit is P G According to the DC power flow model, (6) After the new energy source is connected, (6) is rewritten as: (7) In the formula, k p This represents the ratio of the equivalent impedance at the new energy grid connection point to the equivalent impedance from the internal potential node of the synchronous generator to the receiving end bus.
[0033] Therefore, the change in power angle before and after the connection of the new energy unit satisfies the following formula: (8) Therefore, the large-scale integration of new energy sources reduces the steady-state power angle and the unit swing amplitude, which is conducive to improving transient stability.
[0034] During a fault, according to the transient ride-through control requirements for renewable energy sources, the active power output of renewable energy sources decreases during the transient process and recovers at a certain rate. Therefore, during a fault, renewable energy units meet the following characteristics: (9) In the formula, t f Indicates the time when the fault occurred; t c Indicates the fault clearing time; parameter k This is used to indicate the recovery rate of the unit's active power output. Considering the most severe three-phase short-circuit fault, during which the synchronous motor's electromagnetic power approaches zero.
[0035] For the operating scenario before the integration of new energy sources, the acceleration power of the synchronous machine when a fault occurs and the deceleration power after the fault is cleared are respectively... (10) (11) For the operation scenario after the connection of new energy units, the acceleration power of the synchronous machine when a fault occurs and the deceleration power after the fault is cleared are respectively (12) (13) Comparing the system acceleration and deceleration power before and after the addition of new energy sources, it can be seen that, from the perspective of transient energy, the output of centralized new energy sources, the location of the new energy grid connection point, and the active power recovery rate of new energy sources during faults all affect the transient stability of the system. Based on the above analysis, the main influencing factors reflecting transient power angle stability are: the ratio of new energy to conventional units connected at the sending end, the location of the new energy grid connection point, and the active power recovery rate of new energy units during faults.
[0036] Step 3: Construct the original feature set based on the key factors, and use the Relief algorithm to reduce the dimensionality of the features to obtain the optimal feature subset.
[0037] Specifically, based on the key influencing factors of the power angle stability level mentioned above, an original feature set of the power system is constructed, and a key response feature selection method based on the Relief algorithm is proposed to screen key features that are highly correlated with the power angle stability level, which helps to improve the calculation accuracy and speed of transient stability assessment.
[0038] In the application of data-driven methods, the construction of the sample set is crucial, and its core lies in selecting appropriate input features. These features should be able to comprehensively characterize the system's operating state and be available online. Simultaneously, the feature dimensionality should not be too high; otherwise, changes in the system topology can easily trigger the curse of dimensionality, significantly extending the online update time of the model. Therefore, it is necessary to determine the optimal model input feature set through offline analysis. Considering the convenience of data acquisition, the reasonableness of data dimensionality, and the ability of data to reflect the system state, this invention proposes a feature selection method based on the Relief algorithm.
[0039] The Relief algorithm is a filtering feature selection method and also a feature weighting algorithm. This algorithm can be used for binary classification problems, assigning weights based on the correlation between features and classes, describing the ability of features to distinguish nearby samples. The principle of Relief's feature selection is as follows: For the sample set Choose one sample from them. x i Select the closest sample among its similar samples. x i,h And select the closest sample from the samples of its different classes. x i,m The distance is calculated using Euclidean distance; for each feature of the sample, if x i and x i,hThe distance on this feature is less than x i and x i,m If the distance between the features is greater than or equal to the distance between the features and the target samples, then the feature is more likely to help identify neighboring samples of different categories and is therefore assigned a higher weight; conversely, if the distance is less than or equal to the distance between the features and the target samples, then the feature is assigned a lower weight. The weight of each feature can be obtained by averaging the feature weights of all samples; the higher the weight, the stronger the expressive power of the feature. Sort the features by weight and then remove invalid or weakly discriminative features with low weights to obtain the optimal feature subset, thus achieving feature dimensionality reduction. The formulas for calculating the weights of each feature are as follows: (14) In the formula: superscript j Represents the first corresponding sample j Dimensional features; w j Let be the weight value of the j-th dimension sample. Since the power system response data is a continuous variable, we have: (15) After obtaining the feature weight vector, the weights of each feature are sorted, and given an appropriate threshold, the features with higher effectiveness are selected for the classification task. The specific process of feature dimensionality reduction based on the Relief algorithm is as follows: Figure 4 As shown.
[0040] Example 1 This embodiment provides a method for extracting key response features of transient stability in a sending-end system. Taking the extraction of key response features reflecting the power angle stability level of the sending end as an example, the test system is built in PSASP based on the 2025 Shaanxi operating mode. Northern Shaanxi, as a concentrated area for the transmission of new energy from Northwest China, transmits power through three ±800kV DC lines: Shaanxi-Wuhan DC, Shaanxi-Henan DC, and Shaanxi-Anhui DC, each with a rated DC power of 8000MW. AC lines connect to the Guanzhong Plain load area and then to the main grid in Gansu and other regions, achieving multi-regional interconnection. This invention only focuses on the 750kV grid structure of the sending-end system; therefore, the power grid in northern Shaanxi is simplified as follows: Figure 5 As shown, the baseline capacity of the simulation system is 100MW.
[0041] As mentioned above, the main factors affecting the severity of power angle stability at the sending end in large-scale centralized renewable energy access scenarios include: renewable energy penetration rate, renewable energy power station location, and the dynamic characteristics of renewable energy before and after a fault. Therefore, when transient instability occurs, in addition to the generator rotor variation characteristics and terminal characteristics directly reflecting the system's power angle stability level, based on the key influencing factors of power angle stability and data availability requirements, relevant response information that characterizes the dynamic characteristics of equipment before and after a grid fault is selected from the electrical measurement data of synchronous generator sets, renewable energy generator sets, and line transmission power. The original input feature set is shown in Table 1, with a dimension of 166.
[0042] Table 1 Initial Input Feature Construction
[0043] To fully cover the changing characteristics of the actual power grid response information under different scenarios, and considering different system operating conditions, combined with the actual situation of the Shaanxi power grid, the output of new energy sources was adjusted in this simulation system, and operating modes with new energy penetration rates of 50%, 60%, and 70% were set. Simultaneously, the load factors for different regions were adjusted to ensure that the system load level increased in 5% increments, from 90% to 110%. Under the corresponding operating modes, actual power grid fault conditions were simulated, setting different fault types, including eight typical fault scenarios such as three-phase ground faults, DC blocking, and complex AC / DC faults. Among these, the AC fault location was set at 10% of the AC line, increasing in 10% increments to 90%, for a total of nine fault points. The fault duration ranged from 0.1s to 0.25s, increasing in 0.03s increments, with a total simulation time of 10s. A total of 5040 basic scenarios were generated during the simulation. The changes in each variable during the sample generation phase are shown in Table 2.
[0044] Table 2. Changes in variables during the sample generation phase.
[0045] First, a power angle stability margin evaluation index is constructed, and the rotor power angle transient stability coefficient (TSI), which is commonly used in engineering, is used to measure the degree of instability of the sample: (16) In the formula, The maximum power angle difference between any two generators within the simulation period is represented by TSI. When TSI > 0, the system is transiently stable; otherwise, the system is unstable. Under operating conditions of 110% load and 50% renewable energy penetration, the power transmitted by the sending-end system exceeds the static stability limit. An N-2 fault occurs at t = 1s. The power angle response curve of the sending-end system is shown below. Figure 6 As shown, the units are arranged in two groups, and the power angle is unstable. The calculated results of the above stability assessment indicators are less than 0.
[0046] Based on the aforementioned practical engineering criteria, the generated samples were used to determine their power angle stability, ultimately yielding 4274 stable samples and 766 unstable samples. The correlation between the features obtained using the proposed method and the stability labels was calculated. The correlation weights between all features and the stability labels in each sample were calculated, and then the average of each feature weight across all samples was taken to obtain the final weight of each feature. A positive feature weight indicates that the feature has a positive effect on the classification task, while a negative feature weight indicates that the feature has a negative effect on the classification task. A larger feature weight indicates a stronger correlation, which is more beneficial for accurately assessing the system's power angle stability. Based on this algorithm, the correlation of the 166-dimensional original features was calculated, and the normalized feature weight distribution is as follows: Figure 7 As shown in the figure. Simultaneously, the frequency of occurrence of feature weight magnitudes was statistically analyzed, and its distribution histogram is shown in the figure. Figure 8 As shown. By Figure 8 The data shows that most feature weights are between 0.4 and 0.6, indicating significant redundancy among features that require selective removal. A small number of features have weights below 0.4, contributing little to the effective classification of system stability. A small subset of features have weights above 0.6, even approaching 1; these features are closely related to power angle stability and are crucial for stability determination.
[0047] To verify the effectiveness of the proposed feature selection method and the rationality of the selected key input features, a classic CNN neural network was used to classify and predict the stability level of the power angle.
[0048] Using the selected features as input, the positive and negative values of the power angle stability margin corresponding to the above engineering practical criteria are output labels (0, 1 labels), of which 80% of the samples are used as training datasets and the remaining 20% are used as test datasets.
[0049] The confusion matrix in statistics is used to evaluate the classification prediction performance of the model, as shown in Table 3.
[0050] Table 3 Confusion Matrix
[0051] (17) (18) (19) In the formula, accuracy A Recall is the proportion of samples that correctly predict the true value of a "stable" sample. R To determine the proportion of correct results out of the total observations; misdiagnosis rate. FPR The proportion of samples that are actually stable and are incorrectly predicted as unstable.
[0052] Table 4 compares the prediction performance of the CNN neural network under different feature extraction states. As shown in Table 4, the model's prediction performance is poor with the initial input feature set. However, as the feature extraction process progresses, the model shows improvement across various metrics, demonstrating the effectiveness of the proposed method. It is worth noting that when using 0.6 as the feature weight selection threshold, underfitting may occur due to the limited number of retained feature dimensions, negatively impacting model training. Therefore, considering the data distribution, using a threshold of 0.4 to select features with weight values greater than 0.4 as the input feature set is more appropriate. Simultaneously, selective removal of redundant features of the same type is necessary. In this invention, after feature selection, the final feature dimension is determined to be 74. Features with higher weights include the active power injected into the new energy access point bus, the terminal current of some new energy access points, the rotor angular velocity of some conventional generator units, and the power angle of some conventional generator units, consistent with theoretical analysis results.
[0053] Table 4. Prediction performance under different feature extraction states
[0054] In summary, this embodiment provides a method for extracting key response features of transient stability in a sending-end system. Based on the response characteristics of new energy units, a new energy grid-connected power transmission model is constructed. From the perspective of transient energy, the operating status and transient characteristic changes of the system before and after the new energy connection are analyzed. Key factors affecting the power angle stability level of the system are extracted, providing theoretical support for clarifying the synchronous stability assessment under large-scale grid connection of new energy and providing a basis for the input feature selection of the feature extraction method.
[0055] This embodiment proposes a feature extraction method based on the Relief algorithm, which combines the main influencing factors reflecting the power angle stability level of the sending end. It extracts the key response features reflecting the transient stability level of the system, which helps to optimize the feature subset of high-dimensional electrical characteristics, improve the performance and prediction efficiency of the stability assessment model, and is more adaptable to the complex characteristics of large power grids.
[0056] Example 2 according to Figure 9 As shown, this embodiment also provides a system for extracting key transient stability response features of a sending-end system, used to implement the above-described method for extracting key transient stability response features of a sending-end system, including: Model building module 1 is used to simplify the negative impedance of new energy units and build a mathematical model for new energy grid connection and transmission. Factor extraction module 2 is used to analyze the impact of new energy access on the transient characteristics of the system from the perspective of transient energy, based on the mathematical model, and to extract the key factors affecting the stability level of the system's power angle.
[0057] Feature extraction module 3 is used to construct an original feature set based on the key factors, and to perform feature dimensionality reduction using the Relief algorithm to obtain the optimal feature subset.
[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A method for extracting key transient stability response features of a sending-end system, characterized in that, include: A mathematical model for grid-connected power transmission of new energy units is constructed by simplifying the negative impedance of the new energy units. Based on the mathematical model, the impact of new energy access on the transient characteristics of the system is analyzed from the perspective of transient energy, and the key factors affecting the stability level of the system's power angle are extracted. Based on the aforementioned key factors, an original feature set is constructed, and the Relief algorithm is used to reduce the dimensionality of the features to obtain the optimal feature subset.
2. The method for extracting key transient stability response features of a sending-end system according to claim 1, wherein the new energy unit adopts an equivalent negative impedance model, including a wind power negative impedance model. Z wr and photovoltaic negative impedance model Z phr Based on the wind power negative impedance model Z wr and photovoltaic negative impedance model Z phr Negative impedance of equivalent renewable energy power station Z r The negative impedance Z r The formula is as follows: in, P r and Q r These are the active power and reactive power output by the new energy power station, respectively. U This refers to the bus voltage at the new energy grid connection point.
3. The method for extracting key transient stability response features of a power-generating system according to claim 1, wherein the mathematical model for the grid-connected transmission of new energy is the electromagnetic power-power angle equation of a regional traditional generating unit, expressed as follows: in, Z 11 Z 12 These are the self-impedance of the traditional unit and the mutual impedance with the receiving-end system, respectively. , These are the self-impedance angle and the mutual impedance angle, respectively. This refers to the internal potential phasor of the synchronous generator; E1 is the conjugate phasor of the output current phasor of the synchronous generator; E2 is the internal potential phasor of the synchronous generator. The amplitude of E1; E2 is the amplitude of the equivalent potential of the receiving-end system; For the angle of attack.
4. The method for extracting key transient stability response features of a sending-end system according to claim 3, wherein the self-impedance Z 11 and mutual impedance Z 12 The expression is: in, x1 and x2 are the system reactances, U r For the voltage at the grid connection point of new energy sources, Z r It is a negative impedance; P r and Q r These represent the active power and reactive power output of the new energy power station, respectively; j is the imaginary unit.
5. The method for extracting key response features of transient stability of a sending-end system according to claim 1, wherein the key factors affecting the system's power angle stability level include the ratio of new energy and conventional units connected to the sending-end system, the location of the new energy grid connection point, and the active power recovery rate of the new energy units during a fault.
6. The method for extracting key transient stability response features of a sending-end system according to claim 5, characterized in that, By analyzing the acceleration and deceleration power of the system before and after the integration of new energy sources, the key factors are obtained. Among them, the active power of the new energy unit during the fault period satisfies the following expression: in, t f Indicates the time when the fault occurred; t c Indicates the fault clearing time; parameter k Used to indicate the recovery rate of the unit's active power output.
7. The method for extracting key transient stability response features of a sending-end system according to claim 1, characterized in that, The formula for calculating the feature weights in the Relief algorithm is as follows: in, Let j be the j-th dimension feature of the sample; It is a difference function; Let i be the value of the i-th sample in the j-th dimension feature; The closest sample among samples of the same type. The closest sample among samples of different classes; and The values of sample k and sample s in the j-th dimension feature represent the values of sample k and sample s.
8. The method for extracting key transient stability response features of a sending-end system according to claim 1, characterized in that, The specific process of feature dimensionality reduction using the Relief algorithm is as follows: Calculate the weights of each input feature of the sample, average the weights of all sample features, and, given a threshold, remove features with small weights or weak correlations to obtain the optimal feature subset, thus achieving feature dimensionality reduction.
9. The method for extracting key transient stability response features of a sending-end system according to claim 1, characterized in that, The original feature set was constructed based on the ratio of new energy and conventional units connected to the grid at the sending end, the location of the new energy grid connection point, and the active power recovery rate of the new energy units during the fault.
10. A system for extracting key transient stability response features of a sending-end system, characterized in that, A method for extracting key transient stability response features of a sending-end system as described in any one of claims 1-9, comprising: The model building module is used to simplify the negative impedance of new energy units and build a mathematical model for the grid-connected transmission of new energy. The factor extraction module is used to analyze the impact of new energy access on the transient characteristics of the system from the perspective of transient energy, based on the mathematical model, and to extract the key factors affecting the stability level of the system's power angle. The feature extraction module is used to construct an original feature set based on the key factors, and to perform feature dimensionality reduction using the Relief algorithm to obtain the optimal feature subset.