Power grid signal disturbance analysis method and device based on frequency spectrum characteristics

Abstract

The invention provides a power grid signal disturbance analysis method and device based on frequency spectrum characteristics, and relates to the technical field of power grid signal disturbance analysis, and the method comprises the specific steps: collecting historical data containing various power quality disturbances, carrying out the segmentation through employing equal-length time windows, obtaining a plurality of time sequence signal segments corresponding to the time windows, and carrying out the segmentation of the time sequence signal segments; fourier transform is carried out on each time sequence signal segment, and a frequency spectrum feature corresponding to each time sequence signal segment is obtained; forming a data link by the spectrum features, marking the data link, and constructing a power quality disturbance data training set; and constructing a deep learning model, training the model by using the marked power quality disturbance data training set, and obtaining a trained power quality disturbance recognition model. According to the invention, a traditional neural network structure is transformed, the difficulty of signal processing is reduced, and the accuracy of power quality disturbance signal recognition is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power quality disturbance identification, and in particular to a method and device for analyzing power grid signal disturbance based on spectrum characteristics. Background Art

[0002] With the continuous expansion of the scale of power systems and the increasing complexity of power loads, power quality issues have become an important issue that needs to be urgently addressed in the power industry. The definition of power quality includes voltage fluctuations, frequency offsets, harmonic distortion, transient interference and other phenomena. These interferences not only affect the normal operation of power equipment, but may also cause damage to equipment and unstable power supply. Traditional power quality monitoring methods mostly rely on experience-based analysis and simple statistical methods, and usually use fixed thresholds to judge the quality of power. However, these methods often lack sufficient flexibility and accuracy when faced with complex power quality disturbances and cannot adapt to the ever-changing power environment. In addition, traditional methods often ignore the spectral characteristics of signals during data processing and analysis, making it difficult to accurately identify the source of disturbances and subsequent treatment measures, resulting in the inability to effectively prevent and respond to potential power quality issues.

[0003] Another major shortcoming of existing monitoring technology is its lack of real-time and intelligence level. Many monitoring systems need to rely on manual intervention to interpret power quality data, and the processing process is often time-consuming and susceptible to human factors. This means that when sudden power quality events occur, the power system is often unable to take effective measures in time to respond, further aggravating the power quality problem. In addition, existing methods have limited capabilities in data mining and pattern recognition, and cannot fully utilize rich historical data and real-time data for in-depth analysis and prediction. This defect causes the power system to delay its response to changes in power quality, making it difficult to achieve active management of power quality. Therefore, there is an urgent need for a new method that can improve the ability to identify power quality disturbances and enhance the stability and security of the power system through effective data analysis and intelligent processing methods.

[0004] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0005] The purpose of the present invention is to provide a method and device for analyzing power grid signal disturbance based on spectrum characteristics, so as to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions: A method for analyzing power grid signal disturbance based on spectrum characteristics, the specific steps include: Step 1: Collect historical data containing various power quality disturbances, segment them using time windows of equal length, and obtain time series signal segments corresponding to multiple time windows; Step 2: Perform Fourier transform on each time series signal segment to obtain the spectrum characteristics corresponding to each time series signal segment, wherein the spectrum characteristics include time, frequency and signal intensity; Step 3: The spectral features corresponding to the above-mentioned time series signal segments are used to form a data chain, that is, a data chain composed of time, frequency and intensity. In the data chain, the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency; and the data chain is annotated based on the type of the collected power quality disturbance signal to construct a power quality disturbance data training set; Step 4: Build a deep learning model and train the model using the labeled power quality disturbance data training set to obtain a trained power quality disturbance recognition model. Deploy the trained model to the actual power quality monitoring system to identify and classify power quality disturbances in real time. Step 5: After identifying and classifying the power quality disturbance, collect the total harmonic distortion value and the voltage effective value, perform correlation analysis, calculate the comprehensive power quality disturbance index, compare it with the preset threshold, and perform corresponding processing based on the comparison result.

[0007] Furthermore, the power quality disturbance signal includes: Voltage swell / sag signal, voltage interruption signal, voltage fluctuation signal, pulse transient signal, harmonic signal, transient oscillation signal, voltage swell / sag harmonic signal, frequency deviation signal.

[0008] Furthermore, the process of performing Fourier transform on each time series signal segment includes: Short-time Fourier transform STFT expression: ; in, is the source signal, is the frequency, t represents the time, is a complex function; It is a window added to the time domain signal, which is used to divide the signal into smaller time periods. The result obtained after STFT transformation is the frequency domain information of the original signal, and j is an imaginary unit; a. Select the Hamming window function as the window function, where the window length is 1024 sample points and the overlap ratio is set to 50%; In power quality disturbance identification, power disturbance types include harmonics, transient events, and voltage swells and sags. The Hamming window is used as the window function because it provides a balance between time resolution and frequency resolution and is suitable for power signal analysis scenarios. The Hamming window function expression is: ; in, Represents the index of the sample point in the sequence, the range is , is the length of the window, that is, the number of sample points; Indicates that the Hamming window function is at index The value at the sample point of ; is a cycle of The cosine function of b. Apply the window function to the signal slidingly to form a series of signal fragments within the window; The window function is applied to the original time signal by sliding the window method, starting from the starting point of the signal, and the length of the extraction is The signal segment is divided into two parts and the step size is set. Slide forward to form multiple overlapping window signal fragments; for each window signal fragment, apply the Hamming window function to weight it, and then use Fourier transform to convert it into the frequency domain to obtain the spectrum information of the fragment; Step Length The calculation formula is: ; in, is the step length, is the window length, is the overlap ratio; The Fourier transform formula is as follows: ; in, is the source signal, is a complex function, is the frequency, t is the time, and j is the imaginary unit.

[0009] Furthermore, the power quality disturbance identification model is the same as the ResNet network architecture; The process of constructing the power quality disturbance data training set includes: Perform Fourier transform on each time series signal segment to obtain the spectrum characteristics corresponding to each time series signal segment, wherein the spectrum characteristics include time, frequency and signal strength; construct a data link with the spectrum characteristics corresponding to the time series signal segment, i.e., a data link composed of time, frequency and strength, wherein the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and strength is a frequency index, which points to the signal strength at a specific frequency; and annotate the data link based on the type of power quality disturbance signal collected to construct a power quality disturbance data training set; The method for building a deep learning model and training the model using the labeled power quality disturbance dataset is as follows: In the model, the loss function is used to measure the inconsistency between the predicted value and the true value. Therefore, the method of training the model is: calculate the loss function each time you train, and modify the model parameters by reducing the loss function to meet the expected requirements, that is, to make the predicted value consistent with the true value. The classic cross entropy function is used. The cross entropy is the distance between two probability distributions in mathematical sense. Furthermore, the loss function adopts the cross entropy function, and the formula is as follows: ; in, refers to the true probability distribution, Refers to the predicted probability distribution.

[0010] Furthermore, the process of real-time identification and classification of power quality disturbances includes: Real-time monitoring to obtain power quality disturbance signals; Performing Fourier transform on the acquired power quality disturbance signal to obtain corresponding spectrum characteristics, wherein the spectrum characteristics include time, frequency and signal strength; A data link is constructed based on the spectrum characteristics, i.e., a data link composed of time, frequency and intensity. In the data link, the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency; The data chain is input into the trained power quality disturbance identification model, and the probability corresponding to each power quality disturbance signal is calculated using the Softmax function. Then, the calculated probability values ​​are compared, and the power quality disturbance signal corresponding to the probability value closest to 1 is taken as the target signal to obtain the power quality disturbance identification result.

[0011] Furthermore, the Softmax function formula is: ; in, , , is the input of the Softmax layer, is the number of types of power quality disturbance signals, is a natural constant.

[0012] Furthermore, the calculation obtains a comprehensive power quality disturbance index, specifically: Collecting Total Harmonic Distortion Values ​​Using a Power Quality Analyzer and voltage RMS , use the following formula to calculate the comprehensive power quality disturbance index: ; in, is the comprehensive power quality disturbance index, is the total harmonic distortion value, is the difference between the voltage effective value and the preset value, is the preset scaling factor, is the constant correction exponent; According to the power quality standard, multiple thresholds are set and the comprehensive power quality index is compared with the thresholds: like , determine that the current power quality is good, no special treatment is required, maintain normal monitoring, and record data for trend analysis; like , determine that the current power quality is acceptable, maintain regular monitoring, recommend routine maintenance, and observe whether there are further signs of deterioration; like , determine that the current power quality is critical, immediately conduct detailed analysis, and check the equipment and system status; like , determine that the current power quality is poor, take corrective measures immediately, notify relevant operation and maintenance personnel, and prepare for on-site inspection and maintenance.

[0013] A power grid signal disturbance analysis device based on spectrum characteristics, comprising: The data acquisition module is used to collect historical data containing various power quality disturbances, segment them using time windows of equal length, and obtain time series signal segments corresponding to multiple time windows; A processing module, used to perform Fourier transform on each time series signal segment to obtain a spectrum feature corresponding to each time series signal segment, wherein the spectrum feature includes time, frequency and signal intensity; A training set module is constructed, which is used to construct a data chain of spectral features corresponding to the above-mentioned time series signal fragments, that is, a data chain composed of time, frequency and intensity, in which the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency; and the data chain is annotated based on the type of the collected power quality disturbance signal to construct a power quality disturbance data training set; The model building module is used to build a deep learning model, train the model using the labeled power quality disturbance data training set, obtain a trained power quality disturbance recognition model, and deploy the trained model to the actual power quality monitoring system to identify and classify power quality disturbances in real time; The comprehensive application module is used to collect the total harmonic distortion value and the voltage effective value after identifying and classifying the power quality disturbance, perform correlation analysis, calculate the comprehensive power quality disturbance index, compare it with the preset threshold, and perform corresponding processing according to the comparison result.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention solves the shortcomings of existing monitoring methods in terms of recognition accuracy and real-time performance by introducing a grid signal disturbance analysis method based on spectral characteristics. By conducting in-depth spectral analysis of power quality disturbance signals and combining deep learning technology, efficient recognition and classification of power quality disturbances can be achieved. This method not only improves the intelligence level of power quality monitoring, but also makes real-time data analysis and processing possible, thereby providing strong support for the maintenance and optimization of power systems. Ultimately, the solution significantly improves the response speed and accuracy of power quality management, and reduces the economic losses and equipment failure risks caused by power quality problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the overall method flow of the present invention; Figure 2 It is a schematic diagram of the overall device module of the present invention. DETAILED DESCRIPTION

[0016] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments.

[0017] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the present invention should be understood by people with ordinary skills in the field to which the present invention belongs. The words "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0018] Example: See also Figure 1 , the present invention provides a technical solution: A method for analyzing power grid signal disturbance based on spectrum characteristics, the specific steps include: Step 1: Collect historical data containing various power quality disturbances, segment them using time windows of equal length, and obtain time series signal segments corresponding to multiple time windows; In this embodiment, the power quality disturbance signal includes: Voltage swell / sag signal, voltage interruption signal, voltage fluctuation signal, pulse transient signal, harmonic signal, transient oscillation signal, voltage swell / sag harmonic signal, frequency deviation signal; The above-mentioned various types of power quality disturbance signals can be collected by sensors; various types of power quality disturbance time series signals are collected, and the time domain signals are transformed into frequency domain signals using Fourier transform to obtain the spectral characteristics of the signals, and they are labeled according to the classification, and used as training sets to input into the power quality disturbance identification model for training; therefore, collecting power quality disturbance identification signals is of great significance in power quality management and analysis.

[0019] Step 2: Perform Fourier transform on each time series signal segment to obtain the spectrum characteristics corresponding to each time series signal segment, wherein the spectrum characteristics include time, frequency and signal intensity; In this embodiment, the signal is converted into a time-frequency image using a short-time Fourier transform (STFT) to obtain the time-frequency image required for convolutional neural network training. The specific logic is as follows: Short-time Fourier transform STFT expression: ; in, is the source signal, is the frequency, t represents the time, is a complex function; It is a window added to the time domain signal, which is used to divide the signal into smaller time periods. The result obtained after STFT transformation is the information of the original signal in the time domain and frequency domain. j is an imaginary unit. a. Select the Hamming window function as the window function, where the window length is 1024 sample points and the overlap ratio is set to 50%; In power quality disturbance identification, common power disturbance types include harmonics, transient events, voltage swells and sags, etc. It is more appropriate to use the Hamming window as the window function because the Hamming window provides a good balance between time resolution and frequency resolution and is suitable for most power signal analysis scenarios. The Hamming window function expression is: ; in, Represents the index of the sample point in the sequence, the range is , is the length of the window, that is, the number of sample points; Indicates that the Hamming window function is at index The value at the sample point of ; is a cycle of The cosine function of b. Apply the window function to the signal slidingly to form a series of signal fragments within the window; The window function is applied to the original time signal by sliding the window method, starting from the starting point of the signal, and the length of the extraction is The signal segment is divided into two parts and the step size is set. Slide forward to form multiple overlapping window signal fragments; for each window signal fragment, apply the Hamming window function to weight it, and then use Fourier transform to convert it into the frequency domain to obtain the spectrum information of the fragment; Step Length The calculation formula is: ; in, is the step length, is the window length, is the overlap ratio; The Fourier transform formula is as follows: ; in, is the source signal, is a complex function, is the frequency, t is the time, and j is the imaginary unit; The short-time Fourier transform is a time-frequency analysis method used to analyze the frequency components of non-stationary signals. It calculates the Fourier transform by applying a sliding window to the signal, thereby obtaining the spectrum information of the signal at different time points. The short-time Fourier transform can effectively analyze the time-frequency characteristics of non-stationary signals and is widely used in the field of power quality detection. The short-time Fourier transform (STFT) is used to convert the time-domain power signal into a frequency-domain signal, which can provide a clear display of time-frequency characteristics and help identify and distinguish various types of power quality disturbances. Its dynamic analysis capability makes STFT have obvious advantages in power quality monitoring and fault detection, and is particularly suitable for real-time or online monitoring applications. Through the spectrum diagram, it is easier to find and explain anomalies and disturbances in power signals, thereby improving the reliability and stability of the power system. The convolutional neural network is often used for processing and identifying multidimensional signals in the image field. Traditional signal processing methods usually require manual feature extraction, which is not only time-consuming and labor-intensive, but the extracted features may not be comprehensive enough. The convolutional neural network can automatically learn and extract complex features from the input signal, avoiding the trouble of manual feature extraction, and can capture deeper and more representative features in the signal. This automatic feature extraction capability is more suitable for processing complex and diverse power quality disturbance signals.

[0020] Step 3: The spectral features corresponding to the above-mentioned time series signal segments are used to form a data chain, that is, a data chain composed of time, frequency and intensity. In the data chain, the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency; and the data chain is annotated based on the type of the collected power quality disturbance signal to construct a power quality disturbance data training set; Step 4: Build a deep learning model and train the model using the labeled power quality disturbance data training set to obtain a trained power quality disturbance recognition model. Deploy the trained model to the actual power quality monitoring system to identify and classify power quality disturbances in real time. The power quality disturbance identification model has the same architecture as the ResNet network; this is because the power quality disturbance identification model can take advantage of ResNet's powerful deep feature extraction capabilities, effective training mechanisms, and excellent robustness and generalization performance, which are very beneficial for the recognition and classification of complex signal patterns. During the training process, the power quality disturbance identification model uses a cross-entropy loss function and focuses on the optimization of the classification task. The cross-entropy loss function can effectively evaluate the difference between the probability distribution of the model output and the true label, and promote the accurate classification of the model on various types of power quality disturbances. By labeling and training each type of disturbance, the model can continuously adjust its parameters and optimize the learning of signal features. In addition, in order to further enhance the robustness of the model, data enhancement technology is used during the training process to simulate different power quality disturbance scenarios, so that the model is exposed to more diverse data inputs during training, improving its generalization ability in practical applications. In summary, the power quality disturbance identification model can achieve efficient and accurate disturbance identification and classification in complex power quality data by combining the deep learning capabilities of ResNet with the effectiveness of the cross-entropy loss function.

[0021] The specific logic of the process of constructing the power quality disturbance data training set is as follows: Perform Fourier transform on each time series signal segment to obtain the spectrum characteristics corresponding to each time series signal segment, wherein the spectrum characteristics include time, frequency and signal strength; construct a data link with the spectrum characteristics corresponding to the time series signal segment, i.e., a data link composed of time, frequency and strength, wherein the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and strength is a frequency index, which points to the signal strength at a specific frequency; and annotate the data link based on the type of power quality disturbance signal collected to construct a power quality disturbance data training set; The method for building a deep model and training the model using the labeled power quality disturbance dataset is as follows: In the power quality disturbance identification model, the loss function is used to measure the inconsistency between the predicted value and the true value. Therefore, the method of training the model is: calculate the loss function each time training, and modify the parameters of the power quality disturbance identification model by reducing the loss function to meet the expected requirements, that is, to make the predicted value consistent with the true value. The classic cross entropy function is used. The cross entropy is the distance between two probability distributions in mathematical sense. In this embodiment, the classic cross entropy function is used. The cross entropy is the distance between two probability distributions in mathematical sense, and the expression is: ; in, refers to the true probability distribution, Refers to the predicted probability distribution.

[0022] The power quality disturbance recognition model described here is essentially a convolutional neural network designed for the field of signal recognition, which can effectively maximize the recognition rate when computing resources are limited. Specifically, firstly, the number of network parameters is small, which reduces the possibility of overfitting of the network; secondly, the network has a small amount of computation and is suitable for platforms with limited computing resources.

[0023] The process of real-time identification and classification of power quality disturbances has the following specific logic: Real-time monitoring to obtain power quality disturbance signals; Performing Fourier transform on the acquired power quality disturbance signal to obtain corresponding spectrum characteristics, wherein the spectrum characteristics include time, frequency and signal strength; A data link is constructed based on the spectrum characteristics, i.e., a data link composed of time, frequency and intensity. In the data link, the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency; The data chain is input into the trained power quality disturbance identification model, and the probability corresponding to each power quality disturbance signal is calculated using the Softmax function. Then, the calculated probability values ​​are compared, and the power quality disturbance signal corresponding to the probability value closest to 1 is taken as the target signal to obtain the power quality disturbance identification result.

[0024] The Softmax function formula is: ; in, , , is the input of the Softmax layer, is the number of types of power quality disturbance signals, is a natural constant; for example: if you want to classify six types of power quality disturbance signals, then K is 6, which is converted into a probability between 0 and 1 through the Softmax function, corresponding to the probabilities of the six types of power quality disturbance signals. The probability closest to 1 is the classification result of that type of power quality disturbance signal.

[0025] Step 5: After identifying and classifying the power quality disturbance, collect the total harmonic distortion value and the voltage effective value, perform correlation analysis, calculate the comprehensive power quality disturbance index, compare it with the preset threshold, and perform corresponding processing according to the comparison result; In this embodiment, the calculation obtains the comprehensive power quality disturbance index, specifically: Collecting Total Harmonic Distortion Values ​​Using a Power Quality Analyzer and voltage RMS , use the following formula to calculate the comprehensive power quality disturbance index: ; in, is the comprehensive power quality disturbance index, is the total harmonic distortion value, is the difference between the voltage effective value and the preset value, is the preset scaling factor, is the constant correction exponent; According to the power quality standard, multiple thresholds are set and the comprehensive power quality index is compared with the thresholds: like , determine that the current power quality is good, no special treatment is required, maintain normal monitoring, and record data for trend analysis; like , determine that the current power quality is acceptable, maintain regular monitoring, recommend routine maintenance, and observe whether there are further signs of deterioration; like , determine that the current power quality is critical, immediately conduct detailed analysis, and check the equipment and system status; like , determine that the current power quality is poor, take corrective measures immediately, notify relevant operation and maintenance personnel, and prepare for on-site inspection and maintenance.

[0026] The present invention further provides a power grid signal disturbance analysis device based on spectrum characteristics, the power grid signal disturbance analysis device based on spectrum characteristics is used to perform the above-mentioned power grid signal disturbance analysis method based on spectrum characteristics, including: The data acquisition module is used to collect historical data containing various power quality disturbances, segment them using time windows of equal length, and obtain time series signal segments corresponding to multiple time windows; A processing module, used to perform Fourier transform on each time series signal segment to obtain a spectrum feature corresponding to each time series signal segment, wherein the spectrum feature includes time, frequency and signal intensity; A training set module is constructed, which is used to construct a data chain of spectral features corresponding to the above-mentioned time series signal fragments, that is, a data chain composed of time, frequency and intensity, in which the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency; and the data chain is annotated based on the type of the collected power quality disturbance signal to construct a power quality disturbance data training set; The model building module is used to build a deep learning model, train the model using the labeled power quality disturbance data training set, obtain the trained power quality disturbance recognition model, and deploy the trained model to the actual power quality monitoring system to identify and classify power quality disturbances in real time; The comprehensive application module is used to collect the total harmonic distortion value and the voltage effective value after identifying and classifying the power quality disturbance, perform correlation analysis, calculate the comprehensive power quality disturbance index, compare it with the preset threshold, and perform corresponding processing according to the comparison result.

[0027] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0028] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product. Those skilled in the art may appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein may be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0029] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, and may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0030] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A method for analyzing power grid signal disturbance based on spectrum characteristics, characterized in that: The specific steps include: Step 1: Collect historical data containing various power quality disturbances, segment them using time windows of equal length, and obtain time series signal segments corresponding to multiple time windows; Step 2: Perform Fourier transform on each time series signal segment to obtain the spectrum characteristics corresponding to each time series signal segment, wherein the spectrum characteristics include time, frequency and signal intensity; Step 3: The spectral features corresponding to the above-mentioned time series signal segments are used to form a data chain, that is, a data chain composed of time, frequency and intensity. In the data chain, the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency; and the data chain is annotated based on the type of the collected power quality disturbance signal to construct a power quality disturbance data training set; Step 4: Build a deep learning model and train the model using the labeled power quality disturbance data training set to obtain a trained power quality disturbance recognition model. Deploy the trained model to the actual power quality monitoring system to identify and classify power quality disturbances in real time. Step 5: After identifying and classifying the power quality disturbance, collect the total harmonic distortion value and the voltage effective value, perform correlation analysis, calculate the comprehensive power quality disturbance index, compare it with the preset threshold, and perform corresponding processing based on the comparison result.

2. The method for analyzing power grid signal disturbance based on spectrum characteristics according to claim 1, characterized in that: The power quality disturbance signal includes: Voltage swell / sag signal, voltage interruption signal, voltage fluctuation signal, pulse transient signal, harmonic signal, transient oscillation signal, voltage swell / sag harmonic signal, frequency deviation signal.

3. The method for analyzing power grid signal disturbance based on spectrum characteristics according to claim 2, characterized in that: The process of Fourier transforming each time series signal segment includes: Short-time Fourier transform STFT expression: ； in, is the source signal, is the frequency, t represents the time, is a complex function; It is a window added to the time domain signal, which is used to divide the signal into smaller time periods. The result obtained after STFT transformation is the frequency domain information of the original signal, and j is an imaginary unit; a. Select the Hamming window function as the window function, where the window length is 1024 sample points and the overlap ratio is set to 50%; The Hamming window function expression is: ； in, Represents the index of the sample point in the sequence, the range is , is the length of the window, that is, the number of sample points; Indicates that the Hamming window function is at index The value at the sample point of ; is a cycle of The cosine function of b. Apply the window function to the signal slidingly to form a series of signal fragments within the window; The window function is applied to the original time signal by sliding the window method, starting from the starting point of the signal, and the length of the extraction is The signal segment is divided into two parts and the step size is set. Slide forward to form multiple overlapping window signal fragments; for each window signal fragment, apply the Hamming window function to weight it, and then use Fourier transform to convert it into the frequency domain to obtain the spectrum information of the fragment; Step Length The calculation formula is: ； in, is the step length, is the window length, is the overlap ratio; The Fourier transform formula is as follows: ； in, is the source signal, is a complex function, is the frequency, t is the time, and j is the imaginary unit.

4. The method for analyzing power grid signal disturbance based on spectrum characteristics according to claim 1, characterized in that: The power quality disturbance identification model is the same as the ResNet network architecture; The process of constructing the power quality disturbance data training set includes: Perform Fourier transform on each time series signal segment to obtain the spectrum characteristics corresponding to each time series signal segment, wherein the spectrum characteristics include time, frequency and signal strength; construct a data link with the spectrum characteristics corresponding to the time series signal segment, i.e., a data link composed of time, frequency and strength, wherein the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and strength is a frequency index, which points to the signal strength at a specific frequency; and annotate the data link based on the type of power quality disturbance signal collected to construct a power quality disturbance data training set; The method for building a deep learning model and training the model using the labeled power quality disturbance dataset is as follows: The loss function is calculated each time during training, and the model parameters are modified by reducing the loss function until the predicted value is consistent with the true value.

5. The method for analyzing power grid signal disturbance based on spectrum characteristics according to claim 1, characterized in that: The loss function uses the cross entropy function, and the formula is as follows: ； in, refers to the true probability distribution, Refers to the predicted probability distribution.

6. The method for analyzing power grid signal disturbance based on spectrum characteristics according to claim 1, characterized in that: The process of identifying and classifying power quality disturbances in real time includes: Real-time monitoring to obtain power quality disturbance signals; Performing Fourier transform on the acquired power quality disturbance signal to obtain corresponding spectrum characteristics, wherein the spectrum characteristics include time, frequency and signal strength; A data link is constructed based on the spectrum characteristics, i.e., a data link composed of time, frequency and intensity. In the data link, the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency; The data chain is input into the trained power quality disturbance identification model, and the probability corresponding to each power quality disturbance signal is calculated using the Softmax function. Then, the calculated probability values ​​are compared, and the power quality disturbance signal corresponding to the probability value closest to 1 is taken as the target signal to obtain the power quality disturbance identification result.

7. A method for analyzing power grid signal disturbance based on spectrum characteristics according to claim 1 or 5, characterized in that: The Softmax function formula is: ； in, , , is the input of the Softmax layer, is the number of types of power quality disturbance signals, is a natural constant.

8. The method for analyzing power grid signal disturbance based on spectrum characteristics according to claim 1, characterized in that: The calculation obtains the comprehensive power quality disturbance index, which is specifically: Collecting Total Harmonic Distortion Values ​​Using a Power Quality Analyzer and voltage RMS , use the following formula to calculate the comprehensive power quality disturbance index: ； in, is the comprehensive power quality disturbance index, is the total harmonic distortion value, is the difference between the voltage effective value and the preset value, is the preset scaling factor, is the constant correction exponent; According to the power quality standard, multiple thresholds are set and the comprehensive power quality index is compared with the thresholds: like , determine that the current power quality is good, no special treatment is required, maintain normal monitoring, and record data for trend analysis; like , determine that the current power quality is acceptable, maintain regular monitoring, recommend routine maintenance, and observe whether there are further signs of deterioration; like , determine that the current power quality is critical, immediately conduct detailed analysis, and check the equipment and system status; like , determine that the current power quality is poor, take corrective measures immediately, notify relevant operation and maintenance personnel, and prepare for on-site inspection and maintenance.

9. A power grid signal disturbance analysis device based on spectrum characteristics, characterized in that: The power grid signal disturbance analysis device based on spectrum characteristics is used to execute the power grid signal disturbance analysis method based on spectrum characteristics according to any one of claims 1 to 8, comprising: The data acquisition module is used to collect historical data containing various power quality disturbances, segment them using time windows of equal length, and obtain time series signal segments corresponding to multiple time windows; A processing module, used to perform Fourier transform on each time series signal segment to obtain a spectrum feature corresponding to each time series signal segment, wherein the spectrum feature includes time, frequency and signal intensity; A training set module is constructed, which is used to construct a data chain of spectral features corresponding to the above-mentioned time series signal fragments, that is, a data chain composed of time, frequency and intensity, in which the pointer between time and frequency is a timestamp index, which points to the frequency value at a specific moment, and the pointer between frequency and intensity is a frequency index, which points to the signal intensity at a specific frequency; and the data chain is annotated based on the type of the collected power quality disturbance signal to construct a power quality disturbance data training set; The model building module is used to build a deep learning model, train the model using the labeled power quality disturbance data training set, obtain a trained power quality disturbance recognition model, and deploy the trained model to the actual power quality monitoring system to identify and classify power quality disturbances in real time; The comprehensive application module is used to collect the total harmonic distortion value and the voltage effective value after identifying and classifying the power quality disturbance, perform correlation analysis, calculate the comprehensive power quality disturbance index, compare it with the preset threshold, and perform corresponding processing according to the comparison result.

Images