Short-circuit parameter measurement algorithm development method and system

Abstract

The invention relates to the technical field of power system short circuit detection, and discloses a short circuit parameter measurement algorithm development method and system, and the method comprises the steps: carrying out the short circuit parameter measurement background analysis, and collecting a non-fault disturbance signal; non-fault disturbance signals are preprocessed, and key features are extracted; and establishing a short-circuit parameter prediction algorithm model, and performing classified early warning according to a prediction result. According to the invention, the reliability of short-circuit parameter prediction is improved. The response speed and accuracy to current and voltage changes are improved, and the problems of delay and false alarm in a traditional method can be effectively avoided. The model can retain key features when processing multi-source signals, and the accuracy and robustness of short-circuit parameter prediction are improved. According to the invention, accurate measurement of short-circuit current and voltage is realized, rapid identification and response can be ensured when a fault occurs in a power system, the fault recovery time of the system is reduced, and the safety and reliability of power grid operation are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of short-circuit detection of electric power systems, and in particular to a method and system for developing a short-circuit parameter measurement algorithm. Background Art

[0002] With the continuous expansion of the scale of power systems, the frequency of short-circuit faults has gradually increased, which not only threatens the stability and reliability of the power grid, but may also cause serious damage to equipment and large-scale power outages. Therefore, accurate and timely detection and measurement of short-circuit parameters are crucial for the operation and maintenance of power systems. Traditional short-circuit detection methods usually rely on the action of relay protection equipment and circuit breakers to determine faults, and identify short-circuit types through traditional current and voltage acquisition and simple analysis. However, these methods have certain limitations in complex power grid environments, especially when faced with multiple interference signals and complex fault modes, it is difficult to accurately measure and locate short-circuit parameters.

[0003] In recent years, with the development of digital and intelligent technology, methods based on signal processing and data analysis have been gradually applied to short-circuit detection. Modern power system fault detection methods have introduced more sophisticated algorithms and intelligent analysis tools, such as signal processing technologies based on time domain, frequency domain and wavelet transform, to achieve short-circuit fault detection by analyzing current and voltage waveforms. Although these methods have improved the accuracy of detection to a certain extent, due to the complexity of the power system, the diversity of disturbance signals and real-time requirements, the existing technology still faces problems such as insufficient accuracy, response delay and high false alarm rate. Especially in the face of complex backgrounds such as non-fault disturbances and equipment switching, traditional algorithms often cannot effectively distinguish between disturbances and real short-circuit events, resulting in inaccurate measurement of short-circuit parameters, further affecting the performance of the short-circuit detection system. At this time, how to improve the accurate measurement accuracy of short-circuit current and voltage, and how to eliminate errors in the complex signal background have become technical problems that need to be solved in the current power system fault detection field. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the existing technology still faces problems such as insufficient accuracy, delayed response and high false alarm rate. Especially in the face of complex backgrounds such as non-fault disturbances and equipment switching, traditional algorithms are often unable to effectively distinguish disturbances from real short-circuit events, resulting in inaccurate measurement of short-circuit parameters, further affecting the performance of the short-circuit detection system. At this time, how to improve the accurate measurement accuracy of short-circuit current and voltage, and how to eliminate errors under complex signal backgrounds have become technical problems that need to be solved in the current field of power system fault detection.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: a method for developing a short-circuit parameter measurement algorithm, comprising: performing short-circuit parameter measurement background analysis and collecting non-fault disturbance signals.

[0007] Preprocess the non-fault disturbance signal and extract key features.

[0008] A short-circuit parameter prediction algorithm model is established, and classified warnings are carried out based on the prediction results.

[0009] As a preferred solution of the short-circuit parameter measurement algorithm development method described in the present invention, the short-circuit parameter measurement background analysis includes obtaining historical data of power grid operation from the dispatching center of the power system, including current, voltage, frequency, and load changes, and accessing real-time monitoring SCADA data to ensure coverage of a wide range of power grid operation status information.

[0010] Extract historical short-circuit fault event records, including the current, voltage drop value, short-circuit time, and short-circuit location when the short-circuit fault occurred. Perform cluster analysis on these data to divide the parameter range of typical short-circuit faults.

[0011] Analyze the target power grid’s wiring method, transformer capacity, and line impedance, and calculate the system’s short-circuit capacity as an important reference value for subsequent measurements.

[0012] As a preferred solution of the short-circuit parameter measurement algorithm development method described in the present invention, the acquisition of non-fault disturbance signals includes determining the time points of frequently occurring load switching events through scheduling plans and historical records, extracting the time records of start and stop of power generation equipment as potential disturbance sources, and extracting relevant disturbance signals from the switching records of reactive compensation equipment.

[0013] High-precision current transformers and voltage transformers are installed on the bus side and load side of the substation respectively to collect three-phase current and voltage signals in real time.

[0014] The acquisition equipment uses a data acquisition device with a high sampling rate to ensure that the rapidly changing characteristics of the disturbance signal are captured.

[0015] The signal acquisition parameters are configured as follows:

[0016] The sampling frequency is set to 10kHz to capture high-frequency components and transient characteristics.

[0017] The time window is set to 2 seconds to cover the entire process of the disturbance.

[0018] Data storage uses edge computing devices to store the collected signal data in real time and add time stamps.

[0019] The collected signals are classified according to the disturbance sources, the non-disturbance signal samples are labeled, and clear disturbance and non-disturbance data sets are constructed.

[0020] As a preferred solution of the short-circuit parameter measurement algorithm development method described in the present invention, the preprocessing of the non-fault disturbance signal includes signal denoising and preprocessing, denoising uses a bandpass filter, sets the frequency range between 1Hz and 1000Hz, and removes low-frequency noise and high-frequency noise.

[0021] The filtered signal is band-stopped using a notch filter to remove the fundamental frequency of 50Hz / 60Hz.

[0022] Finally, the signal of each channel is normalized to ensure that the signal amplitudes of different channels are consistent to avoid analysis errors caused by amplitude differences.

[0023] As a preferred solution of the short-circuit parameter measurement algorithm development method described in the present invention, the key feature extraction includes segmenting the continuous non-fault disturbance signal according to the sliding window, and the overlap between the windows is set to 50%. After segmentation, the time domain and frequency domain features are extracted for each time period. The time domain features calculate the peak value, root mean square value and signal amplitude change rate, and the frequency domain features extract the power spectrum density and frequency band energy.

[0024] The continuous wavelet transform is used to extract the time-frequency characteristics of the signal, and based on the results of the wavelet transform, the instantaneous frequency of the signal is extracted through the Hilbert transform.

[0025] Principal component analysis is used to reduce the dimensionality of the extracted time domain, frequency domain and time-frequency features.

[0026] As a preferred solution of the short-circuit parameter measurement algorithm development method described in the present invention, the establishment of a short-circuit parameter prediction algorithm model includes: an input layer receives feature data, a convolution layer is used to extract high-frequency features in the signal, a nonlinear activation layer processes data through a high-order nonlinear activation function, an interaction layer is used to interact and fuse multiple features, and an output layer predicts short-circuit parameters.

[0027] In the input layer, the feature of the input signal after normalization is , including time domain, frequency domain and time-frequency domain features.

[0028] In the convolution layer, the convolution layer is used to extract the high-frequency disturbance features in the signal and design an integral operator based on the convolution kernel. , used to extract detailed features of signals at different scales.

[0029] Through the convolution operation, the signal and Perform convolution to obtain the convolution feature map , describing the local frequency characteristics of the signal at different scales.

[0030] In the nonlinear activation layer, in the convolutional feature map Afterwards, a nonlinear activation function is applied To introduce nonlinear relationships and enhance the expressiveness of the model.

[0031] In the interaction layer, the features extracted from each layer are effectively fused through high-order filtering functions and weighted summation methods to obtain deep interactive information in the signal.

[0032] The features output by multiple convolutional layers are fused through composite weighted summation, and nonlinear filtering is introduced. The fused features are expressed as:

[0033]

[0034] in, represents the weighting coefficient, represents a high-order time-dependent filter function, Indicates time.

[0035] In the output layer, the short-circuit parameters are predicted based on the features extracted from the interaction layer, and a composite exponential-power function model is designed, which is expressed as:

[0036]

[0037] in, represents the predicted short-circuit parameters, specifically the short-circuit current and short-circuit voltage, represents the output features of the interaction layer, Represents the adjustable parameters of the model.

[0038] As a preferred solution of the short-circuit parameter measurement algorithm development method described in the present invention, the classification warning according to the prediction results includes classification warning based on the predicted short-circuit parameters, setting multiple thresholds to distinguish different short-circuit event categories. According to the size of the predicted value, it is divided into normal state, warning level fault, serious fault, and critical fault.

[0039] The predicted short-circuit parameters are short-circuit current and short-circuit voltage. The threshold of short-circuit current is as follows:

[0040] T1=5.0kA, a short-circuit current less than T1 indicates a normal state.

[0041] T2=15.0kA, a short-circuit current greater than T1 and less than T2 indicates a warning-level fault.

[0042] T3=30.0kA, a short-circuit current greater than T2 and less than T3 indicates a serious fault.

[0043] T4=50.0kA, a short-circuit current greater than T4 indicates a critical fault.

[0044] The short-circuit voltage thresholds are as follows:

[0045] V1=0.7pu, which is 70% of the rated voltage. A short-circuit voltage less than V1 indicates a normal state.

[0046] V2=0.5pu, a short-circuit voltage greater than V1 and less than V2 indicates a warning-level fault.

[0047] V3=0.3pu, a short-circuit voltage greater than V2 and less than V3 indicates a serious fault.

[0048] V4=0.1pu, a short-circuit voltage less than V4 indicates a critical fault.

[0049] In the case of short-circuit current, the system will update the monitoring status in real time through the cloud in normal state without manual intervention. Every 10 minutes, the key operating parameters will be automatically recorded to the data storage center, and a normal operation report will be generated. The system will provide relevant data for operators to review. If the system detects long-term stable operation, it will automatically adjust the data collection frequency to reduce resource usage.

[0050] When a warning-level fault occurs, the system activates the warning mechanism, sends an audible and visual alarm signal, and notifies the relevant maintenance personnel, providing the current short-circuit current and the equipment that may be affected. At the same time, automatic data backup is activated, and the disturbance signals in the past hour are recorded to a secure cloud storage platform. The system will extract this data and mark it as a warning sample of potential faults.

[0051] In case of serious faults, the system switches to semi-automatic control mode, suspends non-critical tasks, and prioritizes computing and communication resources for data analysis in the fault area. The system displays the current change curve in the current area and monitors and analyzes the data in real time. Before issuing an isolation command, the system requires maintenance personnel to confirm the fault isolation scope and execution timing through a multi-select interactive interface. The change trend and impact range of the short-circuit current will be highlighted on the interactive interface, and recommended power-off operations for related equipment and lines will be provided.

[0052] In case of critical failure, the system immediately starts the emergency processing procedure, automatically cuts off the power supply of the faulty area, and reduces the impact on other areas. The estimated recovery time is calculated in real time, and the estimated time and specific steps of fault recovery are displayed on the interactive interface. At the same time, emergency warning information is sent to the maintenance team and the management center, including the fault location, the scope of impact, and the preliminary repair plan.

[0053] In the case of short-circuit voltage, the voltage data is recorded every 10 minutes in the normal state and uploaded to the cloud platform to generate a report for analysis. The system displays a comparison chart of the current voltage value and the historical voltage fluctuation for reference by maintenance personnel.

[0054] In case of warning-level faults, the system activates the voltage warning mechanism, providing warning information and equipment inspection suggestions. The system pops up equipment inspection suggestions and actionable troubleshooting steps, automatically backs up the voltage fluctuation data of the past hour and uploads it to the cloud storage platform to facilitate subsequent data analysis and identify potential risks of voltage anomalies.

[0055] In case of a serious fault, when the short-circuit voltage approaches the serious fault range, the system switches to semi-automatic mode, automatically stops non-essential operations, and allocates more resources to the voltage recovery algorithm. The operation interface displays the fault voltage data and recovery plan, and requires the operator to select the fault isolation operation through a multiple-selection interface. The voltage recovery progress will be displayed and it will recommend switching to the backup power supply or restarting the equipment.

[0056] In case of critical fault, the system immediately performs emergency isolation operation, automatically cuts off the voltage supply in the faulty area, and switches to the backup power supply. The system starts the power restoration plan, pushes the repair progress and expected restoration time. Emergency warning information is sent to all relevant maintenance personnel and management, including the faulty area, voltage level, and repair plan, and the real-time restoration status is displayed through the HMI interface, and detailed processing steps are provided.

[0057] A short-circuit parameter measurement algorithm development system, characterized in that: it includes:

[0058] The signal acquisition module performs background analysis of short-circuit parameter measurements and collects non-fault disturbance signals.

[0059] The preprocessing module preprocesses the non-fault disturbance signal and extracts key features.

[0060] The prediction module establishes a short-circuit parameter prediction algorithm model and performs classified warnings based on the prediction results.

[0061] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the above method when executing the computer program.

[0062] A computer-readable storage medium stores a computer program, which implements the steps of the method described above when executed by a processor.

[0063] The beneficial effects of the present invention are as follows: the reliability of short-circuit parameter prediction is improved. The response speed and accuracy to current and voltage changes are improved, and the delay and false alarm problems in traditional methods can be effectively avoided. The model can retain key features when processing multi-source signals, and improve the accuracy and robustness of short-circuit parameter prediction. Accurate measurement of short-circuit current and voltage is achieved, ensuring that the power system can quickly identify and respond to faults, reducing the system's fault recovery time, and improving the safety and reliability of power grid operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0065] Figure 1 An overall flow chart of a short-circuit parameter measurement algorithm development method and system provided in the first embodiment of the present invention. DETAILED DESCRIPTION

[0066] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0067] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a method for developing a short-circuit parameter measurement algorithm, comprising:

[0068] S1: Perform short-circuit parameter measurement background analysis and collect non-fault disturbance signals.

[0069] It should be noted that the main purpose of the short-circuit parameter measurement background analysis is to provide effective basic data and context information for short-circuit detection and prediction models. By deeply analyzing the working status of the power system, common short-circuit modes, and the structural characteristics of the power system, a more accurate reference can be provided for subsequent short-circuit detection and early warning. In the background analysis stage, it is first necessary to obtain operating data such as current, voltage, and frequency from the power grid dispatching center, and analyze the system behavior under different operating conditions in combination with the existing power system architecture.

[0070] During the background analysis process, it is also necessary to extract the historical fault records of the power system, especially the detailed information of short-circuit faults, including the time, location, type of fault, and related current and voltage changes. Clustering and statistical analysis of these data can help identify common fault modes in the system and clarify the current and voltage characteristics generated by different types of faults. Through background analysis, it is possible to provide effective parameter ranges and preliminary judgment criteria for subsequent short-circuit parameter measurement and prediction models, thereby improving the accuracy of subsequent model predictions.

[0071] Background analysis can also help design reasonable feature extraction and data preprocessing strategies, determine the appropriate time window and signal acquisition range, avoid interference from irrelevant data, and ensure accurate capture of key features. Through sufficient background analysis, the system can reduce misjudgment of faults during operation, thereby achieving more accurate early warning and recovery.

[0072] Obtain historical data on grid operation from the power system dispatch center, including current, voltage, frequency, and load changes, and access real-time monitoring SCADA data to ensure coverage of a wide range of grid operation status information.

[0073] Extract historical short-circuit fault event records, including the current, voltage drop value, short-circuit time, and short-circuit location when the short-circuit fault occurred. Perform cluster analysis on these data to divide the parameter range of typical short-circuit faults.

[0074] Analyze the target power grid’s wiring method, transformer capacity, and line impedance, and calculate the system’s short-circuit capacity as an important reference value for subsequent measurements.

[0075] The time points of frequently occurring load switching events are determined through scheduling plans and historical records, and the time records of start and stop of power generation equipment are extracted as potential disturbance sources. The relevant disturbance signals are extracted from the switching records of reactive compensation equipment.

[0076] High-precision current transformers and voltage transformers are installed on the bus side and load side of the substation respectively to collect three-phase current and voltage signals in real time.

[0077] The acquisition equipment uses a data acquisition device with a high sampling rate to ensure that the rapidly changing characteristics of the disturbance signal are captured.

[0078] The signal acquisition parameters are configured as follows:

[0079] The sampling frequency is set to 10kHz to capture high-frequency components and transient characteristics.

[0080] The time window is set to 2 seconds to cover the entire process of the disturbance.

[0081] Data storage uses edge computing devices to store the collected signal data in real time and add time stamps.

[0082] The collected signals are classified according to the disturbance sources, the non-disturbance signal samples are labeled, and clear disturbance and non-disturbance data sets are constructed.

[0083] S2: Preprocess the non-fault disturbance signal and extract key features.

[0084] Perform signal denoising and preprocessing. Use a bandpass filter for denoising and set the frequency range between 1Hz and 1000Hz to remove low-frequency noise and high-frequency noise.

[0085] The filtered signal is band-stopped using a notch filter to remove the fundamental frequency of 50Hz / 60Hz.

[0086] Finally, the signal of each channel is normalized to ensure that the signal amplitudes of different channels are consistent to avoid analysis errors caused by amplitude differences.

[0087] The continuous non-fault disturbance signal is segmented by sliding windows, and the overlap between windows is set to 50%. After segmentation, the time domain and frequency domain features are extracted for each time period. The time domain features calculate the peak value, root mean square value and signal amplitude change rate, and the frequency domain features extract the power spectrum density and frequency band energy.

[0088] The continuous wavelet transform is used to extract the time-frequency characteristics of the signal, and based on the results of the wavelet transform, the instantaneous frequency of the signal is extracted through the Hilbert transform.

[0089] Principal component analysis is used to reduce the dimensionality of the extracted time domain, frequency domain and time-frequency features.

[0090] S3: Establish a short-circuit parameter prediction algorithm model and conduct classified warnings based on the prediction results.

[0091] The input layer receives feature data, the convolution layer is used to extract high-frequency features in the signal, the nonlinear activation layer processes the data through a high-order nonlinear activation function, the interaction layer is used to interact and fuse multiple features, and the output layer predicts short-circuit parameters.

[0092] In the input layer, the feature of the input signal x(t) after normalization is , including time domain, frequency domain and time-frequency domain features, expressed as:

[0093]

[0094] in, represents the normalized RMS value, represents the peak value, represents the rate of change of the signal, represents the power spectral density, represents the frequency band energy, Represents the time-frequency characteristics of wavelet transform.

[0095] In the convolution layer, the convolution layer is used to extract the high-frequency disturbance features in the signal. An integral operator based on the convolution kernel is designed to extract the detailed features of the signal at different scales. The convolution kernel is expressed as:

[0096]

[0097] in, Represents the standard deviation of the Gaussian function and controls the smoothness. Represents the center time of the convolution kernel. Represents the frequency of the convolution kernel.

[0098] It should be noted that the integral operator of the convolution kernel extracts local features in the signal through the convolution operation. In the short-circuit parameter measurement, the convolution kernel is used to extract the high-frequency and detail features of the signal so that the subsequent model can make accurate predictions. When designing the convolution kernel, the smoothness, frequency components and time locality of the signal are usually considered.

[0099] Furthermore, the convolution kernel is generally based on a combination of a smooth function (such as a Gaussian function) and a high-frequency carrier (such as a cosine function), which can adapt to different frequency ranges of the signal. The role of the Gaussian function is to smooth the signal, eliminate noise, and maintain the stability of the signal. The cosine function can maintain the periodicity of the signal and adapt to the periodic disturbances that may occur in the power system. By integrating the input signal and the convolution kernel, the detailed features in the signal can be extracted, and the subsequent model can be helped to better understand the local features of the signal.

[0100] The benefits of this design are that, first, it can capture local changes in signals at multiple scales, which not only meets the detection needs of high-frequency disturbances, but also can handle complex instantaneous changes. Secondly, the integral operator based on the convolution kernel can adapt to different signal types through flexible kernel parameter adjustment, making the model more generalizable. Ultimately, this design method can help the system extract highly discriminative features from complex time domain signals and improve the accuracy and robustness of short-circuit parameter prediction.

[0101] Through the convolution operation, the signal and Perform convolution to obtain the convolution feature map It is expressed as:

[0102]

[0103] Convolutional feature map Describes the local frequency characteristics of the signal at different scales.

[0104] In the nonlinear activation layer, in the convolutional feature map After that, a nonlinear activation function is applied to introduce nonlinear relationships and enhance the expressiveness of the model. The high-order exponential activation function is expressed as:

[0105]

[0106] in, represents the adjustment parameter, represents the adjustment parameter, Represents the adjustment parameter, controls the gain and attenuation speed of the activation function, and enhances the nonlinear response of the signal.

[0107] It should be noted that the combination of compound weighted summation and nonlinear filtering can enhance the performance of deep learning models when processing complex signals. Through compound weighted summation, the system can weightedly fuse the features extracted by multiple convolutional layers to obtain a more comprehensive and diverse signal representation. The purpose of weighted summation is to give different features more important weights, optimize the learning process of the model, and effectively combine local features into higher-level abstract features.

[0108] During the design process, the features of each convolution output layer need to be weighted first. The weights can be learned during the training process and adjusted according to the influence of different features. Next, nonlinear filtering is introduced to further increase the nonlinear expression ability of the model by applying nonlinear functions (such as ReLU or Sigmoid) to the weighted summed features. This process can help the model recognize more complex patterns in the signal, especially when the signal has a strong nonlinear relationship (such as the current and voltage changes when a short circuit fault occurs).

[0109] The benefits of this design method are: first, it can effectively combine the features of multiple convolutional layers, enhance the model's ability to fuse multi-scale information, and extract richer signal features. Second, nonlinear filtering can enhance the model's expressiveness, allowing the network to handle complex, nonlinear data relationships and improve the accuracy of short-circuit event predictions. Finally, by combining weighted summation with nonlinear activation, the robustness of the model can be significantly improved, reducing sensitivity to local noise, thereby improving the stability of the system in practical applications.

[0110] In the interaction layer, the features extracted from each layer are effectively fused through high-order filtering functions and weighted summation methods to obtain deep interactive information in the signal.

[0111] The features output by multiple convolutional layers are fused through composite weighted summation, and nonlinear filtering is introduced. The fused features are expressed as:

[0112]

[0113] in, represents the weighting coefficient, represents a high-order time-dependent filter function, expressed as:

[0114]

[0115] in, and Control the periodic variation of time differences, Indicates the attenuation speed of the control filter function.

[0116] In the output layer, the short-circuit parameters are predicted based on the features extracted from the interaction layer, and a composite exponential-power function model is designed, which is expressed as:

[0117]

[0118] in, Represents the predicted short-circuit parameters, specifically the short-circuit current I sc Or short circuit voltage V sc , represents the output features of the interaction layer, Represents the adjustable parameters of the model.

[0119] Based on the predicted short-circuit parameters, the warning is classified and multiple thresholds are set to distinguish different short-circuit event categories. According to the size of the predicted value, it is divided into normal state, warning level fault, serious fault, and critical fault.

[0120] The predicted short-circuit parameters are short-circuit current and short-circuit voltage. The threshold of short-circuit current is as follows:

[0121] T1=5.0kA, a short-circuit current less than T1 indicates a normal state.

[0122] T2=15.0kA, a short-circuit current greater than T1 and less than T2 indicates a warning-level fault.

[0123] T3=30.0kA, a short-circuit current greater than T2 and less than T3 indicates a serious fault.

[0124] T4=50.0kA, a short-circuit current greater than T4 indicates a critical fault.

[0125] The short-circuit voltage thresholds are as follows:

[0126] V1=0.7pu, which is 70% of the rated voltage. A short-circuit voltage less than V1 indicates a normal state.

[0127] V2=0.5pu, a short-circuit voltage greater than V1 and less than V2 indicates a warning-level fault.

[0128] V3=0.3pu, a short-circuit voltage greater than V2 and less than V3 indicates a serious fault.

[0129] V4=0.1pu, a short-circuit voltage less than V4 indicates a critical fault.

[0130] It should be noted that the short-circuit current and voltage thresholds are set to achieve accurate fault classification and early warning. In short-circuit parameter measurement, short-circuit current and voltage are the two most critical parameters. By monitoring and analyzing these two parameters, it is possible to accurately determine whether a short-circuit fault occurs in the power system and the severity of the fault. The threshold setting is based on the following aspects:

[0131] First, the threshold setting needs to refer to the working standards and safety requirements of the power system. For example, for most power systems, the current and voltage will be maintained within a certain range during normal operation, and any fluctuations beyond this range may indicate the occurrence of a fault. Therefore, the threshold is usually set as the dividing line between different fault categories, which can accurately distinguish between normal status, warning level faults, serious faults and critical faults.

[0132] Secondly, the setting of the threshold value needs to be adjusted in combination with the actual operation data and historical fault records of the power system. By analyzing a large amount of fault data and non-fault data, the range of current and voltage changes can be determined according to the actual situation. For example, when setting the threshold value of short-circuit current, the actual response of various loads in the power system and the typical changes in current when a fault occurs can be taken into account. The voltage threshold can be determined by analyzing the characteristics of voltage drops to ensure that voltage anomalies can be accurately identified.

[0133] Finally, considering the dynamic response characteristics of the power system, the setting of thresholds also needs to adapt to changes in different times and regions. By monitoring the operating status of the system in real time and dynamically adjusting the thresholds, it can be ensured that the fault warning mechanism can play the best effect under different loads, different seasons or different weather conditions.

[0134] In the case of short-circuit current, the system will update the monitoring status in real time through the cloud in normal state without manual intervention. Every 10 minutes, the key operating parameters will be automatically recorded to the data storage center, and a normal operation report will be generated. The system will provide relevant data for operators to review. If the system detects long-term stable operation, it will automatically adjust the data collection frequency to reduce resource usage.

[0135] When a warning-level fault occurs, the system activates the warning mechanism, sends an audible and visual alarm signal, and notifies the relevant maintenance personnel, providing the current short-circuit current and the equipment that may be affected. At the same time, automatic data backup is activated, and the disturbance signals in the past hour are recorded to a secure cloud storage platform. The system will extract this data and mark it as a warning sample of potential faults.

[0136] In case of serious faults, the system switches to semi-automatic control mode, suspends non-critical tasks, and prioritizes computing and communication resources for data analysis in the fault area. The system displays the current change curve in the current area and monitors and analyzes the data in real time. Before issuing an isolation command, the system requires maintenance personnel to confirm the fault isolation scope and execution timing through a multi-select interactive interface. The change trend and impact range of the short-circuit current will be highlighted on the interactive interface, and recommended power-off operations for related equipment and lines will be provided.

[0137] In case of critical failure, the system immediately starts the emergency processing procedure, automatically cuts off the power supply of the faulty area, and reduces the impact on other areas. The estimated recovery time is calculated in real time, and the estimated time and specific steps of fault recovery are displayed on the interactive interface. At the same time, emergency warning information is sent to the maintenance team and the management center, including the fault location, the scope of impact, and the preliminary repair plan.

[0138] In the case of short-circuit voltage, the voltage data is recorded every 10 minutes in the normal state and uploaded to the cloud platform to generate a report for analysis. The system displays a comparison chart of the current voltage value and the historical voltage fluctuation for reference by maintenance personnel.

[0139] In case of warning-level faults, the system activates the voltage warning mechanism, providing warning information and equipment inspection suggestions. The system pops up equipment inspection suggestions and actionable troubleshooting steps, automatically backs up the voltage fluctuation data of the past hour and uploads it to the cloud storage platform to facilitate subsequent data analysis and identify potential risks of voltage anomalies.

[0140] In case of a serious fault, when the short-circuit voltage approaches the serious fault range, the system switches to semi-automatic mode, automatically stops non-essential operations, and allocates more resources to the voltage recovery algorithm. The operation interface displays the fault voltage data and recovery plan, and requires the operator to select the fault isolation operation through a multiple-selection interface. The voltage recovery progress will be displayed and it will recommend switching to the backup power supply or restarting the equipment.

[0141] In case of critical fault, the system immediately performs emergency isolation operation, automatically cuts off the voltage supply in the faulty area, and switches to the backup power supply. The system starts the power restoration plan, pushes the repair progress and expected restoration time. Emergency warning information is sent to all relevant maintenance personnel and management, including the faulty area, voltage level, and repair plan, and the real-time restoration status is displayed through the HMI interface, and detailed processing steps are provided.

[0142] The above embodiments also include a short-circuit parameter measurement algorithm development system, specifically:

[0143] The signal acquisition module performs background analysis of short-circuit parameter measurements and collects non-fault disturbance signals.

[0144] The preprocessing module preprocesses the non-fault disturbance signal and extracts key features.

[0145] The prediction module establishes a short-circuit parameter prediction algorithm model and performs classified warnings based on the prediction results.

[0146] The computer device may be a server. The computer device includes a processor, a memory, an input / output interface (I / O for short) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data cluster data of the power monitoring system. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a short-circuit parameter measurement algorithm development method is implemented.

[0147] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0148] Embodiment 2 is an embodiment of the present invention, which provides a method and system for developing a short-circuit parameter measurement algorithm. In order to verify the beneficial effects of the present invention, a scientific demonstration is carried out through simulation experiments.

[0149] In order to verify the innovation and effectiveness of the short-circuit parameter measurement algorithm of the present invention, a comparative experiment was conducted. The experiment aims to compare the performance of the existing traditional short-circuit parameter detection method and the short-circuit parameter measurement algorithm based on non-fault disturbance technology proposed in the present invention under different fault conditions, especially the differences in the accuracy, response speed and false alarm rate of short-circuit current and voltage measurement.

[0150] Prior art steps: Prior art mainly relies on the traditional short-circuit protection method based on threshold triggering. Its basic steps include:

[0151] Signal acquisition: Signal acquisition is performed through conventional current transformers (CT) and voltage transformers (PT), and the sampling frequency is usually 1kHz.

[0152] Signal analysis: Determine whether a short circuit occurs by setting the current and voltage thresholds, and record the short circuit parameters. This method does not consider the impact of disturbance signals and can only respond based on simple threshold judgments, resulting in a high probability of false alarms and missed alarms.

[0153] The steps of the present invention: The short-circuit parameter measurement algorithm of the present invention includes the following steps:

[0154] Signal acquisition: Similar to the existing technology, high-precision current transformers (CT) and voltage transformers (PT) are used for signal acquisition, but the sampling frequency is increased to 10kHz to capture high-frequency disturbance signals.

[0155] Non-fault disturbance signal processing: Remove 50Hz / 60Hz fundamental frequency noise and other irrelevant signals through bandpass filter and notch filter.

[0156] Feature extraction and prediction model establishment: Use convolutional neural networks to extract the time domain, frequency domain, and time-frequency domain features of the signal, and use the prediction model to predict the short-circuit current and voltage.

[0157] Classification warning and response mechanism: According to the prediction results, short circuit events are classified into normal, warning, serious, critical by setting different thresholds, and corresponding warnings and automatic responses are triggered at different fault levels.

[0158] During the experiment, different short-circuit current and voltage parameters were set, and these parameters were measured and predicted using the existing technology and the algorithm of the present invention respectively. In the experiment, different situations such as normal state, warning level fault, serious fault and critical fault were simulated, and the measurement results of the two methods were recorded in each case. All data were collected by the on-site real-time monitoring system and processed for comparative analysis. The experimental results are shown in Table 1.

[0159] Table 1 Experimental results

[0160] Through comparative analysis of experimental data, it can be clearly seen that the short-circuit parameter measurement algorithm proposed in the present invention is superior to the prior art in many aspects. The following is a detailed analysis based on the table data:

[0161] In the prior art, the measurement accuracy of short-circuit current and voltage is relatively low, especially under high-intensity short-circuit faults (such as 50.0 kA and 0.1 pu), the measurement accuracy is only 80%. However, the measurement accuracy of the present invention is significantly improved. Under the same fault conditions, the measurement accuracy can reach more than 95%, and up to 99%. This shows that the present invention effectively improves the prediction ability of short-circuit parameters and reduces the measurement errors caused by equipment limitations by adopting higher sampling frequency signal acquisition and deep learning algorithms.

[0162] The false alarm rate and missed alarm rate of the existing technology are relatively high, especially in more complex fault scenarios, such as severe faults (25.0 kA and 0.3 pu), with a false alarm rate of up to 18% and a missed alarm rate of 12%. These errors mainly come from the inherent defects of the existing technology based on threshold judgment, which cannot effectively deal with the impact of disturbance signals. In the present invention, the false alarm rate and missed alarm rate are significantly reduced, especially in the case of high-intensity faults, the false alarm rate is reduced to less than 6%, and the missed alarm rate is reduced to 5%. By introducing feature extraction and prediction models based on non-fault disturbance technology, the present invention can more accurately distinguish between fault and non-fault signals, significantly reducing the occurrence of false alarms and missed alarms.

[0163] The measurement response time of the existing technology is relatively long, especially under complex short-circuit faults with high current and high voltage, the response time is 36ms on average. In contrast, the present invention reduces the response time to 18ms by adopting a higher sampling frequency and an efficient data processing algorithm, which is of great significance for the real-time protection and rapid response of the power system. The fast response time can ensure that the power grid can start protection measures in time when a short circuit occurs, reducing the risk of power equipment damage and power outages.

[0164] Compared with the prior art, the present invention has significant innovation and advantages. By introducing high sampling frequency (10kHz), non-fault disturbance signal processing and deep learning algorithms, the present invention can more accurately measure short-circuit current and voltage, reduce the risk of false alarms and missed alarms, and has a shorter response time, and can make fault judgments in a short time. In addition, the composite weighted summation and nonlinear filtering technology used in the invention shows stronger robustness when multi-source signals are fused, so that the present invention can still maintain a high accuracy in a complex power grid environment.

[0165] Through these data comparisons, it is clearly demonstrated that the present invention has obvious advantages in improving the measurement accuracy of short-circuit parameters, reducing the false alarm rate and missed alarm rate, shortening the response time, etc. Especially in the case of high-intensity faults, the measurement accuracy and response speed of the present invention far exceed the existing technology, and it has significant innovation and application value.

[0166] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for developing a short-circuit parameter measurement algorithm, characterized in that: include: Conduct short-circuit parameter measurement background analysis and collect non-fault disturbance signals; Preprocess the non-fault disturbance signal and extract key features; Establish a short-circuit parameter prediction algorithm model and conduct classified warnings based on the prediction results; The background analysis includes obtaining historical data of power grid operation from the dispatch center of the power system, including current, voltage, frequency, and load changes, and accessing real-time monitoring SCADA data to ensure coverage of a wide range of power grid operation status information; Extract historical short-circuit fault event records, including the current, voltage drop value, short-circuit time, and short-circuit location when the short-circuit fault occurred; Perform cluster analysis on these data to divide the parameter range of typical short-circuit faults; Analyze the target power grid’s wiring mode, transformer capacity, and line impedance, and calculate the system’s short-circuit capacity as an important reference value for subsequent measurements; The key feature extraction includes segmenting the continuous non-fault disturbance signal according to the sliding window, setting the overlap between the windows to 50%, and extracting the time domain and frequency domain features of each time period after segmentation; The time domain features calculate the peak value, RMS value and rate of change of signal amplitude, and the frequency domain features extract the power spectrum density and frequency band energy; Use continuous wavelet transform to extract the time-frequency characteristics of the signal, and based on the results of wavelet transform, extract the instantaneous frequency of the signal through Hilbert transform; Principal component analysis is used to reduce the dimensionality of the extracted time domain, frequency domain and time-frequency features.

2. The method for developing a short-circuit parameter measurement algorithm according to claim 1, wherein: The collecting of non-fault disturbance signals includes determining the time points of frequently occurring load switching events through scheduling plans and historical records, extracting the time records of the start and stop of power generation equipment as potential disturbance sources, and extracting relevant disturbance signals from the switching records of reactive power compensation equipment; High-precision current transformers and voltage transformers are installed on the bus side and load side of the substation to collect three-phase current and voltage signals in real time; The acquisition equipment uses a data acquisition device with a high sampling rate to ensure that the fast-changing characteristics of the disturbance signal are captured; The signal acquisition parameters are configured as follows: The sampling frequency was set to 10kHz to capture high-frequency components and transient characteristics; The time window is set to 2 seconds, covering the entire process of disturbance occurrence; Data storage uses edge computing devices to store the collected signal data in real time and add time stamps; The collected signals are classified according to the disturbance sources, the non-disturbance signal samples are labeled, and clear disturbance and non-disturbance data sets are constructed.

3. The method for developing a short-circuit parameter measurement algorithm according to claim 2, wherein: The preprocessing of the non-fault disturbance signal includes signal denoising and preprocessing, wherein the denoising uses a bandpass filter, sets the frequency range between 1 Hz and 1000 Hz, and removes low-frequency noise and high-frequency noise; The filtered signal is band-stopped using a notch filter to remove the fundamental frequency of 50Hz / 60Hz; Finally, the signal of each channel is normalized.

4. The method for developing a short-circuit parameter measurement algorithm according to claim 3, wherein: The short-circuit parameter prediction algorithm model is established, including an input layer receiving feature data, a convolution layer for extracting high-frequency features in the signal, a nonlinear activation layer for processing data through a high-order nonlinear activation function, an interaction layer for interacting and fusing multiple features, and an output layer for predicting short-circuit parameters; In the input layer, the feature of the input signal after normalization is , including time domain, frequency domain and time-frequency domain features; In the convolution layer, the convolution layer is used to extract the high-frequency disturbance features in the signal and design an integral operator based on the convolution kernel. , used to extract detailed features of signals at different scales Through the convolution operation, the signal and Perform convolution to obtain the convolution feature map , describing the local frequency characteristics of the signal at different scales; In the nonlinear activation layer, in the convolutional feature map Afterwards, a nonlinear activation function is applied To introduce nonlinear relationships and enhance the expressiveness of the model; In the interaction layer, the features extracted from each layer are effectively fused through high-order filtering functions and weighted summation methods to obtain deep interactive information in the signal; The features output by multiple convolutional layers are fused through composite weighted summation, and nonlinear filtering is introduced; the fused features are expressed as: ,in, represents the weighting coefficient, represents a high-order time-dependent filter function, Indicates time; In the output layer, the short-circuit parameters are predicted based on the features extracted from the interaction layer, and a composite exponential-power function model is designed, which is expressed as: ,in, represents the predicted short-circuit parameters, specifically the short-circuit current and short-circuit voltage, represents the output features of the interaction layer, Represents the adjustable parameters of the model.

5. The method for developing a short-circuit parameter measurement algorithm according to claim 4, wherein: The classified warning according to the prediction results includes classified warning based on the predicted short-circuit parameters, and setting multiple thresholds to distinguish different short-circuit event categories; According to the size of the predicted value, it is divided into normal state, warning level fault, serious fault and critical fault; The predicted short-circuit parameters are short-circuit current and short-circuit voltage. The threshold of short-circuit current is as follows: T1=5.0kA, a short-circuit current less than T1 indicates a normal state; T2=15.0kA, a short-circuit current greater than T1 and less than T2 indicates a warning-level fault; T3=30.0kA, a short-circuit current greater than T2 and less than T3 indicates a serious fault; T4=50.0kA, a short-circuit current greater than T4 indicates a critical fault; The short-circuit voltage thresholds are as follows: V1=0.7pu, which is 70% of the rated voltage. A short-circuit voltage less than V1 indicates a normal state. V2=0.5pu, a short-circuit voltage greater than V1 and less than V2 indicates a warning-level fault; V3=0.3pu, a short-circuit voltage greater than V2 and less than V3 indicates a serious fault; V4=0.1pu, a short-circuit voltage less than V4 indicates a critical fault; In the case of short-circuit current, the system will update the monitoring status in real time through the cloud in normal state without manual intervention; key operating parameters will be automatically recorded to the data storage center every 10 minutes, and a normal operation report will be generated. The system will provide relevant data for operators to review; If the system detects long-term stable operation, it will automatically adjust the data collection frequency to reduce resource usage; In case of a warning-level fault, the system activates the warning mechanism, sends an audible and visual alarm signal, and notifies the relevant maintenance personnel, providing the current short-circuit current and the equipment that may be affected; at the same time, it starts automatic data backup, recording the disturbance signals in the past hour to a secure cloud storage platform. The system will extract this data and mark it as a warning sample of potential faults; In case of serious faults, the system switches to semi-automatic control mode, suspends non-critical tasks, and prioritizes the allocation of computing and communication resources for data analysis in the fault area. The system displays the current change curve of the current area and monitors and analyzes the data in real time. Before issuing an isolation command, the system requires maintenance personnel to confirm the fault isolation scope and execution timing through a multi-select interactive interface. The changing trend and impact range of short-circuit current will be highlighted on the interactive interface, and recommended power-off operations for related equipment and lines will be provided; In case of critical faults, the system immediately initiates the emergency processing procedure, automatically cuts off the power supply in the faulty area, and reduces the impact on other areas; calculates the estimated recovery time in real time, and displays the estimated time and specific steps for fault recovery on the interactive interface; and sends emergency warning information to the maintenance team and management center, including the fault location, impact range, and preliminary repair plan; In the case of short-circuit voltage, the voltage data is recorded every 10 minutes in normal state and uploaded to the cloud platform to generate a report for analysis; the system displays a comparison chart of the current voltage value and historical voltage fluctuations for reference by maintenance personnel; In case of warning level fault, the system activates the voltage warning mechanism, providing warning information and equipment inspection suggestions; The system pops up equipment inspection suggestions and actionable troubleshooting steps, automatically backs up the voltage fluctuation data of the past hour and uploads it to the cloud storage platform to facilitate subsequent data analysis and identify potential risks of voltage anomalies; In case of a serious fault, when the short-circuit voltage approaches the serious fault range, the system switches to semi-automatic mode, automatically stops non-essential operations, and allocates more resources to the voltage recovery algorithm; the fault voltage data and recovery plan are displayed in the operation interface, and the operator is required to select the fault isolation operation through a multiple-selection interface, which will display the voltage recovery progress and recommend operations such as switching to a backup power source or restarting the equipment; In case of critical fault, the system will immediately perform emergency isolation operation, automatically cut off the voltage supply to the faulty area, and switch to the backup power supply; The system initiates the power restoration plan, pushes the repair progress and expected restoration time; sends emergency warning information to all relevant maintenance personnel and management, including the fault area, voltage level, and repair plan, displays the real-time restoration status through the HMI interface, and provides detailed processing steps.

6. A short-circuit parameter measurement algorithm development system using the method according to any one of claims 1 to 5, characterized in that: The signal acquisition module performs background analysis of short-circuit parameter measurements and acquires non-fault disturbance signals; The preprocessing module preprocesses the non-fault disturbance signal and extracts key features; Prediction module, establishes a short-circuit parameter prediction algorithm model and performs classified warnings based on the prediction results; The background analysis includes obtaining historical data of power grid operation from the dispatch center of the power system, including current, voltage, frequency, and load changes, and accessing real-time monitoring SCADA data to ensure coverage of a wide range of power grid operation status information; Extract historical short-circuit fault event records, including the current, voltage drop value, short-circuit time, and short-circuit location when the short-circuit fault occurred; Perform cluster analysis on these data to divide the parameter range of typical short-circuit faults; Analyze the target power grid’s wiring mode, transformer capacity, and line impedance, and calculate the system’s short-circuit capacity as an important reference value for subsequent measurements; The key feature extraction includes segmenting the continuous non-fault disturbance signal according to the sliding window, setting the overlap between the windows to 50%, and extracting the time domain and frequency domain features of each time period after segmentation; The time domain features calculate the peak value, RMS value and rate of change of signal amplitude, and the frequency domain features extract the power spectrum density and frequency band energy; Use continuous wavelet transform to extract the time-frequency characteristics of the signal, and based on the results of wavelet transform, extract the instantaneous frequency of the signal through Hilbert transform; Principal component analysis is used to reduce the dimensionality of the extracted time domain, frequency domain and time-frequency features.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images