CN-121980446-A - Power system fault risk detection method, device, equipment and storage medium

CN121980446ACN 121980446 ACN121980446 ACN 121980446ACN-121980446-A

Abstract

The application discloses a method, a device, equipment and a storage medium for detecting the fault risk of an electric power system, wherein the method comprises the steps of obtaining the operation data of the electric power system; the method comprises the steps of calculating energy change rates of all elements according to three types of elements including a branch, a generator and a load based on operation data, constructing a feature vector based on the energy change rates, inputting the feature vector as a model, carrying out power system risk detection by adopting a preset two-class XGBoost model to obtain detection results, wherein the detection results comprise risk and no risk, and carrying out risk positioning by adopting a preset multi-class XGBoost model based on the feature vector when the detection results are risk to obtain element positions with fault risks. The method and the system improve timeliness and accuracy of fault risk early warning and effectively guarantee reliable operation of the power grid.

Inventors

ZHOU XIN
ZHOU YUMIN
YAO HAICHENG
YANG XINYI
XIAO LIANG
MAO ZHENYU
WANG WEI
YUAN QUAN
DU XU
ZHOU YONGCAN
ZHANG QIANG

Assignees

中国南方电网有限责任公司

Dates

Publication Date: 20260505
Application Date: 20251229

Claims (10)

1. A power system fault risk detection method, comprising: Acquiring operation data of a power system; Based on the operation data, respectively calculating the energy change rate of each element according to three types of elements of a branch, a generator and a load; Constructing a feature vector based on the energy change rate; Inputting the feature vector as a model, and detecting the risk of the power system by adopting a preset two-class XGBoost model to obtain a detection result, wherein the detection result comprises risk and risk-free; and when the detection result is that the risk exists, performing risk positioning by adopting a preset multi-classification XGBoost model based on the feature vector to obtain the element position with fault risk.
2. The power system fault risk detection method of claim 1, wherein the acquiring the operation data of the power system includes: Acquiring historical wind power output, photovoltaic power output, generator data and grid structure parameters in a power system; and carrying out annual time sequence production simulation of the power system considering the tide net rack based on the wind power output, the photovoltaic output, the generator data and the power grid structure parameter and combining constraint conditions to obtain annual operation data of the simulated power system, wherein the constraint conditions comprise, but are not limited to, power balance constraint, unit output constraint, climbing rate constraint, start-stop constraint, system standby constraint and inter-subsystem net rack constraint.
3. The power system fault risk detection method as claimed in claim 1, wherein the calculating the energy change rate of each element according to three kinds of elements of branch, generator and load based on the operation data comprises: based on the operation data, respectively calculating the element accumulated energy of three types of elements of a branch circuit, a generator and a load; and calculating the energy change rate of each element according to the difference value of the accumulated energy of each element at adjacent time.
4. A power system fault risk detection method as claimed in claim 3, wherein the component accumulated energy comprises branch node accumulated energy, generator node accumulated energy, load node accumulated energy; Said calculating, based on said operational data, the element accumulated energy of the three classes of elements, branch, generator, load, respectively, comprises: according to the operation data, carrying out integral solution based on the instantaneous power of the branch, and obtaining the accumulated energy of the node of the branch; According to the generator frequency in the operation data, calculating to obtain the kinetic energy of a generator rotor as the accumulated energy of a generator node; And obtaining apparent power of the load according to the operation data, and carrying out trapezoidal integral solution based on the apparent power of the load to obtain accumulated energy of the load node.
5. The power system fault risk detection method of claim 1, wherein the constructing a feature vector based on the energy change rate comprises: based on the operation data, acquiring basic operation characteristics and time sequence statistical characteristics of each element, wherein the basic operation characteristics comprise active power, reactive power and voltage, and the time sequence statistical characteristics comprise the maximum value, standard deviation and fluctuation times of power in preset time; After the basic operation characteristics, the time sequence statistical characteristics and the energy change rate are subjected to standardized processing, splicing the basic operation characteristics, the time sequence statistical characteristics and the energy change rate into an initial characteristic matrix; And based on correlation analysis and feature importance analysis, feature screening is carried out on the initial feature matrix to obtain a final feature vector.
6. The method for detecting risk of power system failure as set forth in claim 1, wherein said detecting risk of power system using a preset classification XGBoost model to obtain a detection result includes: respectively inputting the feature vectors of the elements corresponding to each subsystem into a trained two-class XGBoost model, and outputting the risk probability corresponding to the subsystems, wherein the power system is divided into a plurality of subsystems according to the region; If the risk probability is not greater than the probability threshold, judging that the detection result of the corresponding subsystem is risk-free; the classification XGBoost model is obtained through training according to the historical feature vector, the label of training data is 1 or 0,1 indicates risk, and 0 indicates no risk.
7. The method for detecting risk of failure in a power system according to claim 1, wherein the risk localization using a preset multi-classification XGBoost model to obtain the location of the element at risk of failure comprises: Inputting the feature vector corresponding to the subsystem with risk as the detection result into a trained multi-classification XGBoost model to obtain risk probability distribution of each element in the subsystem; in the risk probability distribution, selecting a set number of elements as candidate elements according to the order of the probabilities from high to low; and determining the element position with fault risk by verifying whether the energy change rate of the candidate element is larger than a preset threshold value.
8. A power system fault risk detection apparatus, comprising: The data acquisition module is used for acquiring the operation data of the power system; The energy change rate module is used for respectively calculating the energy change rate of each element according to three types of elements of a branch, a generator and a load based on the operation data; The feature vector module is used for constructing a feature vector based on the energy change rate; the risk detection module is used for inputting the feature vector as a model, and carrying out power system risk detection by adopting a preset two-class XGBoost model to obtain detection results, wherein the detection results comprise risks and no risks; and the risk positioning module is used for performing risk positioning by adopting a preset multi-classification XGBoost model based on the feature vector when the detection result is that the risk exists, so as to obtain the element position with fault risk.
9. A computer device comprising a processor and a memory, the memory having a computer program stored therein and the computer program being configured to be executed by the processor, when executing the computer program, implementing the power system fault risk detection method according to any one of claims 1 to 7.
10. A computer readable storage medium, wherein the computer readable storage medium stores a computer program, and wherein the apparatus in which the computer readable storage medium is located implements the power system fault risk detection method according to any one of claims 1 to 7 when the computer program is executed by the apparatus.

Description

Power system fault risk detection method, device, equipment and storage medium Technical Field The present application relates to the field of power systems, and in particular, to a method, an apparatus, a device, and a storage medium for detecting a fault risk of a power system. Background As the scale of power systems is continuously enlarged and the structure is increasingly complex, especially in the context of large amount of new energy access and wide application of power electronics, the operation of power grids becomes more complex and the risks are increasing. Therefore, the risk identification and positioning of the power grid are accurately and timely carried out, and the method is very important for guaranteeing the reliable operation of the power system. The traditional power system risk identification and positioning method generally takes a physical model drive as a core and relies on accurate power grid parameters and off-line simulation analysis. For example, stability analysis based on transient energy functions or off-line time domain simulation methods such as PSASP, BPA and the like are highly dependent on accurate parameters of elements and preset operation scenes, and parameter calibration is time-consuming and easy to generate larger deviation when new energy output fluctuates or loads suddenly change. In addition, the traditional model is often analyzed from the perspective of the whole system, and local coupling relations among elements are ignored, so that risk positioning is fuzzy, and specific fault elements are difficult to lock. Disclosure of Invention In order to solve the technical problems, the application provides the method, the device, the equipment and the storage medium for detecting the fault risk of the power system, which have strong generalization capability and high positioning precision, improve the timeliness and the accuracy of fault risk early warning and effectively ensure the reliable operation of a power grid. The application provides a power system fault risk detection method, which comprises the following steps: Acquiring operation data of a power system; Based on the operation data, respectively calculating the energy change rate of each element according to three types of elements of a branch, a generator and a load; Constructing a feature vector based on the energy change rate; Inputting the feature vector as a model, and detecting the risk of the power system by adopting a preset two-class XGBoost model to obtain a detection result, wherein the detection result comprises risk and risk-free; and when the detection result is that the risk exists, performing risk positioning by adopting a preset multi-classification XGBoost model based on the feature vector to obtain the element position with fault risk. As an improvement of the above-mentioned aspect, the acquiring operation data of the electric power system includes: Acquiring historical wind power output, photovoltaic power output, generator data and grid structure parameters in a power system; and carrying out annual time sequence production simulation of the power system considering the tide net rack based on the wind power output, the photovoltaic output, the generator data and the power grid structure parameter and combining constraint conditions to obtain annual operation data of the simulated power system, wherein the constraint conditions comprise, but are not limited to, power balance constraint, unit output constraint, climbing rate constraint, start-stop constraint, system standby constraint and inter-subsystem net rack constraint. As an improvement of the above-mentioned scheme, the calculating the energy change rate of each element according to the three kinds of elements of the branch, the generator and the load based on the operation data includes: based on the operation data, respectively calculating the element accumulated energy of three types of elements of a branch circuit, a generator and a load; and calculating the energy change rate of each element according to the difference value of the accumulated energy of each element at adjacent time. As an improvement of the above-mentioned scheme, the element accumulated energy includes branch node accumulated energy, generator node accumulated energy, load node accumulated energy; Said calculating, based on said operational data, the element accumulated energy of the three classes of elements, branch, generator, load, respectively, comprises: according to the operation data, carrying out integral solution based on the instantaneous power of the branch, and obtaining the accumulated energy of the node of the branch; According to the generator frequency in the operation data, calculating to obtain the kinetic energy of a generator rotor as the accumulated energy of a generator node; And obtaining apparent power of the load according to the operation data, and carrying out trapezoidal integral solution based on the apparent power of the loa