Power grid oscillation monitoring system and method based on intelligent sensor

Abstract

The invention relates to the technical field of power system monitoring and control, and discloses a power grid oscillation monitoring system and method based on an intelligent sensor, and the system comprises an intelligent sensor module which is used for collecting the signal data of the operation of a power grid, is internally provided with a self-adaptive sampling mechanism, dynamically adjusts the sampling frequency to achieve the capturing of the signal change, and meanwhile, transmits the signal data to the intelligent sensor module; de-noising and preprocessing are carried out on the collected signal data, the processed signal data are transmitted to an edge calculation module, and high-quality input is provided for oscillation signal feature extraction; and the edge calculation module is used for receiving the preprocessed signal data from the intelligent sensor module. Through the intelligent sensor module, a self-adaptive sampling mechanism is combined with multi-mode signal acquisition, the rapid change characteristic of sampling frequency capture signals is dynamically adjusted, accurate capture of oscillation signals is ensured, the redundant calculation burden and energy consumption of the system can be reduced while the precision is ensured, and technical innovation is provided for real-time monitoring of a complex oscillation scene of a power grid.

Description

Technical Field

[0001] The present invention relates to the technical field of power system monitoring and control, and in particular to a power grid oscillation monitoring system and method based on intelligent sensors. Background Art

[0002] Grid oscillation is a common and complex dynamic phenomenon in power systems. It is caused by faults, control instability and the access of new energy sources. Its characteristics are rapid changes in voltage, current and frequency. If the oscillation cannot be discovered and effectively controlled in time, it will lead to equipment damage, reduced power quality and even grid collapse.

[0003] Traditional power grid monitoring equipment usually adopts a fixed sampling method, which has limited ability to capture rapid changes in signals. If the frequency of power grid oscillation changes rapidly and the amplitude fluctuates significantly, fixed sampling will lead to reduced timeliness of collected data and information loss. At the same time, traditional equipment lacks multi-modal acquisition capabilities and cannot obtain multi-dimensional data, which limits the ability to fully analyze oscillation characteristics. Current oscillation pattern recognition technologies mainly rely on shallow machine learning algorithms and rule-based methods, which perform well under single-source oscillations and simple grid topologies. However, when faced with complex grid structures and multi-source oscillations that propagate across regions, the recognition accuracy and traceability are significantly reduced. Most existing oscillation monitoring systems are in passive response mode, which detects and alarms after oscillation occurs, but cannot provide timely predictions of future oscillation trends, leaving grid operators in a passive state and unable to take measures in advance against potential risks. Most existing systems are deployed independently, and the signal acquisition, analysis and control links are scattered and lack linkage, resulting in a delayed response to oscillation control. In addition, traditional systems cannot directly convert monitoring results into executable control signals, and have a low degree of integration with the power grid protection system, which limits the rapid response capability of oscillation control.

[0004] Therefore, those skilled in the art provide a power grid oscillation monitoring system and method based on smart sensors to solve the above-mentioned problems. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a power grid oscillation monitoring system and method based on intelligent sensors to solve the problems raised in the above background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a power grid oscillation monitoring system based on intelligent sensors, comprising: The intelligent sensor module is used to collect signal data of power grid operation, and has a built-in adaptive sampling mechanism to dynamically adjust the sampling frequency to capture signal changes. At the same time, the collected signal data is pre-processed and transmitted to the edge computing module to provide high-quality input for oscillation signal feature extraction; The edge computing module receives the pre-processed signal data from the intelligent sensor module, extracts the characteristic information of the oscillation signal using the signal analysis algorithm, and determines the oscillation state through the threshold dynamic adjustment method to preliminarily classify the oscillation type. On this basis, it supports collaborative analysis among multiple nodes, transmits the results of feature extraction and oscillation classification to the data fusion module, and synchronously transmits them to the cloud platform module; The data fusion module is used to receive the signal data from the intelligent sensor module and the feature extraction and oscillation classification results from the edge computing module, and uniformly process the signal data through the data integration algorithm. The integrated data has the characteristics of time alignment, providing consistent input for the cloud platform module; The cloud platform module receives the integrated signal data from the data fusion module and the oscillation classification results from the edge computing module, further identifies complex oscillation patterns, establishes a power grid topology model based on a deep learning algorithm, locates the oscillation source, and at the same time, deeply mines the oscillation pattern in combination with historical data, generates the oscillation propagation path, and transmits the analysis results to the trend analysis module; The trend analysis module predicts the future development trend of power grid oscillation based on the analysis results of oscillation mode, oscillation source location and propagation path from the cloud platform module and historical data, and generates short-term and medium-term oscillation trend data. The prediction results will be further transmitted to the communication and control module to support control decisions; The communication and control module realizes data transmission between the intelligent sensor module, edge computing module, data fusion module and cloud platform module, provides time synchronization function for modules within the system, and at the same time, receives oscillation analysis results from the cloud platform module and trend analysis module, converts them into executable protection instructions, and sends corresponding control signals to the power grid control center to achieve effective oscillation control of the power grid.

[0007] Preferably, the adaptive sampling mechanism of the intelligent sensor module dynamically adjusts the sampling frequency based on the signal fluctuation characteristics. , its sampling frequency satisfies the following relationship: , in, is the real-time sampling frequency, is the basic sampling frequency, is the adjustment factor, is the standard deviation of the signal amplitude, The preset threshold for signal fluctuation.

[0008] Preferably, the intelligent sensor module includes a signal filtering unit, which uses Kalman filtering to denoise the signal, and the filtering process includes the following steps: Status prediction: , in, For the predicted state, For the previous moment The posterior estimate of the state, For the previous moment The control input, is the control matrix, is the state transfer matrix; Error covariance prediction: , in, is the forecast error covariance matrix, For the previous moment The posterior error covariance matrix of is the transpose of the state transfer matrix, is the state transfer matrix, is the process noise covariance; Status Update: , in, is the updated state estimate, For the predicted state, is the Kalman gain, is the actual observed signal of the sensor, is the observation matrix; Kalman gain calculation: , in, is the Kalman gain, is the observation noise covariance, is the observation matrix, is the transpose of the observation matrix, is the prediction error covariance matrix.

[0009] Preferably, the edge computing module performs feature extraction on the transmission signal. The feature extraction process is based on fast Fourier transform, which converts the time domain signal x(t) into the frequency domain to satisfy the following relationship: , in, is the frequency domain representation of the signal, x(t) is the input signal, is the frequency variable, is the time variable, The angular frequency component of the signal, Indicates the frequency The complex exponential form of the fundamental wave.

[0010] Preferably, the edge computing module detects the oscillation state through a dynamic threshold algorithm, and its dynamic threshold The following relations are satisfied: , in, Dynamic threshold, is the mean of the signal eigenvalues, The standard deviation of the signal eigenvalues, is the threshold adjustment factor.

[0011] Preferably, the data fusion module performs weighted fusion on the signal data from the smart sensor, and the fused signal data is calculated by the following formula: , in, is the fused signal data, is the original signal data from the i-th sensor, is the weight of the i-th sensor, N is the total number of signals, is the time variable; It is proportional to the inverse of the sensor measurement noise and satisfies the following relationship: , in, is the standard deviation of the noise of the i-th sensor measurement signal, with N being the total number of signals.

[0012] Preferably, the cloud platform module extracts the time-frequency features of the signal data through a convolutional neural network, and the convolution operation satisfies the following relationship: , in, The position in the feature matrix of the convolution output The value of is the position in the input signal feature matrix The value of is the position in the convolution kernel matrix The weight of is the bias value, is the activation function, and The row and column size of the convolution kernel.

[0013] Preferably, the cloud platform module realizes oscillation source positioning through a graph neural network, and the update of node features satisfies the following relationship: , in, For node v in The feature vector of the layer, is the feature vector of node u in layer k, is the weight matrix of the kth layer, is the neighbor set of node v, , is the degree of nodes v and u, is the bias vector of the kth layer, is the activation function.

[0014] Preferably, the trend analysis module predicts the time series characteristics of the oscillation signal based on the long short-term memory network, and the state update of the long short-term memory network unit satisfies the following relationship: Input Gate:

[0015] Forget Gate:

[0016] Output Gate:

[0017] Unit status update:

[0018] Hide status updates:

[0019] in, , , is the activation value of the input gate, forget gate, and output gate, , , , is the weight matrix, , , , is the bias value, is the unit state, is the hidden state, is the input signal at the current moment, is the activation function, is the hyperbolic tangent function.

[0020] A method for monitoring power grid oscillation based on an intelligent sensor, comprising: Step 1: Collect signal data in power grid operation through the intelligent sensor module. The intelligent sensor module has a built-in adaptive sampling mechanism to dynamically adjust the sampling frequency according to signal changes. At the same time, the collected signal data is pre-processed and the pre-processed high-quality signal data is transmitted to step 2; Step 2, receiving the pre-processed signal data from the intelligent sensor module, processing the signal data using a signal analysis algorithm, extracting characteristic information of the oscillation signal, determining the oscillation state based on the extracted characteristic information using a threshold dynamic adjustment method, and preliminarily classifying the oscillation type. At the same time, supporting collaborative analysis among multiple nodes, transmitting the results of oscillation feature extraction and preliminary classification to step 3, and synchronously transmitting them to step 4; Step 3: Receive the signal data from the smart sensor module and the oscillation characteristics and classification results from the edge computing module, process the signal data through a data integration algorithm, and transmit the integrated data to step 4 as input for further analysis; Step 4: Receive the integrated signal data from step 3 and the oscillation classification result from step 2, analyze the signal using a deep learning algorithm, identify complex oscillation patterns, establish a power grid topology model, and locate the oscillation source by calculation. At the same time, further mine the oscillation pattern in combination with historical data, generate the oscillation propagation path, and pass the result to step 5; Step 5: receiving the analysis results of the oscillation mode, oscillation source location and propagation path provided in step 4, combining historical data, predicting the short-term and medium-term development trends of the power grid oscillation, generating oscillation trend data, and transmitting the prediction results to step 6; Step 6: After receiving the oscillation analysis results provided in step 4 and step 5, unified processing is performed to generate protection control instructions. The protection instructions are converted into specific power grid control signals and sent to the power grid control center for implementing oscillation control of the power grid. At the same time, internal data transmission and time synchronization functions are provided in the system.

[0021] The present invention provides a power grid oscillation monitoring system and method based on intelligent sensors, which have the following beneficial effects: 1. The present invention combines the adaptive sampling mechanism with multimodal signal acquisition through an intelligent sensor module, dynamically adjusts the sampling frequency to capture the rapid change characteristics of the signal, ensures the accurate capture of the oscillation signal, reduces the redundant computing burden and energy consumption of the system while ensuring accuracy, and provides technical innovation for real-time monitoring of complex oscillation scenarios in the power grid.

[0022] 2. The present invention identifies complex oscillation patterns through deep learning algorithms and combines them with the grid topology to achieve accurate modeling of oscillation propagation paths and multi-source oscillation tracing, thereby improving the accuracy and practicality of analysis in scenarios with multi-source and multi-path oscillations, and laying the foundation for accurate prediction and control of grid operation risks.

[0023] 3. The present invention introduces a long short-term memory network to achieve short-term and medium- to long-term trend prediction of oscillation signals, providing strong support for real-time control and operation planning of power grids. By using deep learning to mine the timing characteristics of oscillation signals, the present invention can capture the potential change trend of oscillations in advance and effectively avoid the risk of sudden oscillations in power grid operation.

[0024] 4. The present invention proposes a complete monitoring process from signal acquisition, feature extraction, pattern analysis, trend prediction to linkage control to achieve the coordinated work of various functional units. In particular, the communication and control module can convert the analysis results into specific protection instructions, achieve seamless connection with the power grid control system, support the control operations of generator excitation adjustment and load shedding, and solve the problems of data islands and response delays in traditional systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 is a system diagram of the present invention; Figure 2 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0026] In order to make the technical personnel in the technical field understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in combination with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a partial embodiment of the present invention, not a complete embodiment. Based on the embodiment of the present invention, other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.

[0027] The present invention is described in detail below in conjunction with the accompanying drawings: Example: Please see attached Figure 1 and attached Figure 2 The embodiment of the present invention provides a power grid oscillation monitoring system based on an intelligent sensor, comprising: The intelligent sensor module is used to collect signal data of power grid operation, and has a built-in adaptive sampling mechanism to dynamically adjust the sampling frequency to capture signal changes. At the same time, the collected signal data is pre-processed and transmitted to the edge computing module to provide high-quality input for oscillation signal feature extraction; The edge computing module receives the pre-processed signal data from the intelligent sensor module, extracts the characteristic information of the oscillation signal using the signal analysis algorithm, and determines the oscillation state through the threshold dynamic adjustment method to preliminarily classify the oscillation type. On this basis, it supports collaborative analysis among multiple nodes, transmits the results of feature extraction and oscillation classification to the data fusion module, and synchronously transmits them to the cloud platform module; The data fusion module is used to receive the signal data from the intelligent sensor module and the feature extraction and oscillation classification results from the edge computing module, and uniformly process the signal data through the data integration algorithm. The integrated data has the characteristics of time alignment, providing consistent input for the cloud platform module; The cloud platform module receives the integrated signal data from the data fusion module and the oscillation classification results from the edge computing module, further identifies complex oscillation patterns, establishes a power grid topology model based on a deep learning algorithm, locates the oscillation source, and at the same time, deeply mines the oscillation pattern in combination with historical data, generates the oscillation propagation path, and transmits the analysis results to the trend analysis module; The trend analysis module predicts the future development trend of power grid oscillation based on the analysis results of oscillation mode, oscillation source location and propagation path from the cloud platform module and historical data, and generates short-term and medium-term oscillation trend data. The prediction results will be further transmitted to the communication and control module to support control decisions; The communication and control module realizes data transmission between the intelligent sensor module, edge computing module, data fusion module and cloud platform module, provides time synchronization function for modules within the system, and at the same time, receives oscillation analysis results from the cloud platform module and trend analysis module, converts them into executable protection instructions, and sends corresponding control signals to the power grid control center to achieve effective oscillation control of the power grid.

[0028] Benefits of smart sensor modules: Support multi-modal signal acquisition, provide comprehensive grid operation information, improve the ability to capture fast-changing oscillation signals, reduce acquisition power consumption and data redundancy, and improve data quality; Benefits of edge computing module: It receives smart sensor data in real time and extracts key oscillation features, realizes local rapid analysis, reduces the computing burden on the cloud, quickly identifies anomalies at the early stage of oscillation, and adapts to the needs of distributed oscillation analysis under complex power grid topology; Benefits of the data fusion module: The data of the intelligent sensor is integrated through a weighted fusion algorithm to avoid data deviation caused by sensor measurement errors, improve signal consistency, and achieve time alignment processing, so that multi-source data can be analyzed under a unified time reference, ensuring the synchronization and consistency of input signals; Benefits of the cloud platform module: It uses deep learning algorithms to efficiently identify complex oscillation patterns and model oscillation propagation paths, adapt to multi-source and multi-path oscillation environments, provide accurate traceability capabilities, support grid operation optimization and control strategy formulation, and provide support for long-term grid operation stability analysis; The benefits of the trend analysis module are that it can identify future oscillation trends, help take preventive measures in advance, provide operational references for real-time control, provide operating data basis for power grid planning and dispatching, and enhance the system's perception and defense capabilities against uncertainties in power grid operation; The benefits of the communication and control module are to ensure the consistency and accuracy of multi-module data transmission, connect the intelligent sensor module, edge computing module, data fusion module and cloud platform module to form a complete data flow and feedback link, and convert the analysis results into specific protection instructions, including generator excitation adjustment and load shedding operations, which are quickly transmitted to the power grid control center to achieve linkage control.

[0029] The adaptive sampling mechanism of the intelligent sensor module dynamically adjusts the sampling frequency based on the signal fluctuation characteristics , its sampling frequency satisfies the following relationship: , in, is the real-time sampling frequency, is the basic sampling frequency, is the adjustment factor, is the standard deviation of the signal amplitude, The preset threshold for signal fluctuation.

[0030] The adaptive sampling mechanism can dynamically adjust the sampling frequency according to the standard deviation of the signal amplitude. If the signal fluctuates violently, the sampling frequency will be automatically increased to capture the details of the rapidly changing signal. If the signal changes steadily, the sampling frequency will be reduced to reduce redundant acquisition. By dynamically adjusting the sampling frequency, necessary high-frequency sampling data is transmitted, unnecessary redundant data transmission is avoided, the load of the communication network is reduced, and efficient and accurate input data is provided for the edge computing module and the cloud analysis module; The introduction of adjustment factors provides flexible adjustment capabilities for sampling strategies, optimizes the dynamic adjustment range of sampling frequency according to different power grid operation scenarios, and adapts to various complex monitoring needs; In the early stage of the oscillation signal, the sampling frequency can be increased rapidly, and the adaptive sampling mechanism can capture the early oscillation characteristics, providing more timely data support for subsequent edge computing modules and cloud platforms, and improving the system's response speed to oscillation events.

[0031] The intelligent sensor module includes a signal filtering unit, which uses Kalman filtering to denoise the signal. The filtering process includes the following steps: Status prediction: , in, For the predicted state, For the previous moment The posterior estimate of the state, For the previous moment The control input, is the control matrix, is the state transfer matrix; Error covariance prediction: , in, is the forecast error covariance matrix, For the previous moment The posterior error covariance matrix of is the transpose of the state transfer matrix, is the state transfer matrix, is the process noise covariance; Status Update: , in, is the updated state estimate, For the predicted state, is the Kalman gain, is the actual observed signal of the sensor, is the observation matrix; Kalman gain calculation: , in, is the Kalman gain, is the observation noise covariance, is the observation matrix, is the transpose of the observation matrix, is the prediction error covariance matrix.

[0032] Kalman filtering combines state prediction with state update to optimize the estimated value of the real signal in multiple samplings. The filter uses the state information of the previous moment to infer the current state in the prediction stage, and corrects the prediction result in combination with the actual observation value of the sensor in the update stage, which significantly improves the accuracy of the signal. The signals processed by Kalman filtering have high accuracy and consistency, providing high-quality input data for subsequent edge computing modules and cloud platforms. The errors and noise in the signals are significantly reduced, ensuring the accuracy of feature extraction and pattern analysis.

[0033] The edge computing module extracts features from the transmission signal. The feature extraction process is based on fast Fourier transform, which converts the time domain signal x(t) into the frequency domain to satisfy the following relationship: , in, is the frequency domain representation of the signal, x(t) is the input signal, is the frequency variable, is the time variable, The angular frequency component of the signal, Indicates the frequency The complex exponential form of the fundamental wave.

[0034] The edge computing module uses fast Fourier transform to extract features of the transmission signal, converting the time domain signal into the frequency domain, and effectively extracting the frequency, amplitude and phase information of the oscillation signal. The fast Fourier transform can quickly identify the fundamental and harmonic components with a frequency of by analyzing the angular frequency components of the signal, revealing the spectral characteristics of the oscillation signal, and achieving efficient frequency domain analysis with low computational complexity, so that the edge computing module can analyze sensor data in real time, quickly capture the frequency changes and amplitude characteristics of the oscillation signal, and provide accurate and comprehensive feature data support for subsequent oscillation state determination and type classification.

[0035] The edge computing module detects the oscillation state through a dynamic threshold algorithm. The following relations are satisfied: , in, Dynamic threshold, is the mean of the signal eigenvalues, The standard deviation of the signal eigenvalues, is the threshold adjustment factor.

[0036] The edge computing module detects the oscillation state through a dynamic threshold algorithm, and dynamically calculates the threshold using the mean and standard deviation of the signal characteristic values. It can adaptively adjust the threshold according to the actual characteristics of the signal fluctuation, avoiding misjudgment and missed judgment caused by the fixed threshold method when the signal changes drastically and the characteristics fluctuate. It can accurately distinguish between normal signal fluctuations and abnormal oscillations, and improve the accuracy and sensitivity of oscillation state judgment. At the same time, it provides real-time and reliable detection results for subsequent oscillation classification and response.

[0037] The data fusion module performs weighted fusion on the signal data from the intelligent sensor. The fused signal data is calculated using the following formula: , in, is the fused signal data, is the original signal data from the i-th sensor, is the weight of the i-th sensor, N is the total number of signals, is the time variable; It is proportional to the inverse of the sensor measurement noise and satisfies the following relationship: , in, is the standard deviation of the noise of the i-th sensor measurement signal, with N being the total number of signals.

[0038] The data fusion module integrates the signal data from multiple intelligent sensors through a weighted fusion algorithm, and uses the relationship that the weight of each sensor is proportional to the inverse of its measurement noise to effectively suppress the interference of noise sensors on the fusion results. The signal data calculated by the fusion formula has higher accuracy and robustness, and time alignment is achieved at the same time, ensuring that data from different sensors can be analyzed under a unified time reference, significantly improving the consistency and credibility of multi-source data, and providing reliable input support for subsequent complex analysis on the cloud platform. It is especially suitable for oscillation signal processing in large-scale distributed monitoring scenarios.

[0039] The cloud platform module extracts the time-frequency features of the signal data through a convolutional neural network. The convolution operation satisfies the following relationship: , in, The position in the feature matrix of the convolution output The value of is the position in the input signal feature matrix The value of is the position in the convolution kernel matrix The weight of is the bias value, is the activation function, and The row and column size of the convolution kernel.

[0040] The cloud platform module uses convolutional neural networks to extract time-frequency features from signal data. Through convolution operations, it can efficiently capture the local spatial correlation and characteristic patterns of the signal. The weights of the convolution kernel automatically learn the key characteristics of the signal, and the activation function introduces nonlinearity to enhance the expression of complex signals. It can extract deep time-frequency information from the input data and significantly improve the recognition accuracy of complex oscillation patterns, laying a solid foundation for subsequent oscillation tracing and trend prediction, especially in multi-source and multi-path oscillation analysis.

[0041] The cloud platform module locates the oscillation source through graph neural network, and the update of node features satisfies the following relationship: , in, For node v in The feature vector of the layer, is the feature vector of node u in layer k, is the weight matrix of the kth layer, is the neighbor set of node v, , is the degree of nodes v and u, is the bias vector of the kth layer, is the activation function.

[0042] The cloud platform module uses graph neural networks to locate the oscillation source. Through the node feature update formula, it can efficiently model the complex relationship between nodes and neighboring nodes in the power grid topology. The recursive update of the node feature vector enables each node to integrate the information of neighboring nodes layer by layer, and then capture the propagation path of the oscillation signal in the power grid. By learning the structural characteristics of the power grid and the laws of signal propagation, this module can accurately locate the source of oscillation, provide key data support for power grid dispatching and fault control, and is particularly suitable for accurate tracing in multi-source oscillation and complex topology scenarios.

[0043] The trend analysis module predicts the time series characteristics of the oscillation signal based on the long short-term memory network. The state update of the long short-term memory network unit satisfies the following relationship: Input Gate:

[0044] Forget Gate:

[0045] Output Gate:

[0046] Unit status update:

[0047] Hide status updates:

[0048] in, , , is the activation value of the input gate, forget gate, and output gate, , , , is the weight matrix, , , , is the bias value, is the unit state, is the hidden state, is the input signal at the current moment, is the activation function, is the hyperbolic tangent function.

[0049] The trend analysis module predicts the time series characteristics of the oscillation signal through the long short-term memory network, and uses the activation values ​​of the input gate, forget gate and output gate to selectively remember and forget the historical information of the time series. At the same time, through the update of the unit state and hidden state, it can flexibly capture the long-term dependence and short-term dynamic change characteristics of the signal. It can process the complex time series characteristics of the power grid oscillation signal, predict the frequency, amplitude and change trend of the future signal, and provide reliable short-term and medium- and long-term prediction results for power grid regulation.

[0050] A method for monitoring power grid oscillation based on an intelligent sensor, comprising: Step 1: Collect signal data in power grid operation through the intelligent sensor module. The intelligent sensor module has a built-in adaptive sampling mechanism to dynamically adjust the sampling frequency according to signal changes. At the same time, the collected signal data is pre-processed and the pre-processed high-quality signal data is transmitted to step 2; Step 2, receiving the pre-processed signal data from the intelligent sensor module, processing the signal data using a signal analysis algorithm, extracting characteristic information of the oscillation signal, determining the oscillation state based on the extracted characteristic information using a threshold dynamic adjustment method, and preliminarily classifying the oscillation type. At the same time, supporting collaborative analysis among multiple nodes, transmitting the results of oscillation feature extraction and preliminary classification to step 3, and synchronously transmitting them to step 4; Step 3: Receive the signal data from the smart sensor module and the oscillation characteristics and classification results from the edge computing module, process the signal data through a data integration algorithm, and transmit the integrated data to step 4 as input for further analysis; Step 4: Receive the integrated signal data from step 3 and the oscillation classification result from step 2, analyze the signal using a deep learning algorithm, identify complex oscillation patterns, establish a power grid topology model, and locate the oscillation source by calculation. At the same time, further mine the oscillation pattern in combination with historical data, generate the oscillation propagation path, and pass the result to step 5; Step 5: receiving the analysis results of the oscillation mode, oscillation source location and propagation path provided in step 4, combining historical data, predicting the short-term and medium-term development trends of the power grid oscillation, generating oscillation trend data, and transmitting the prediction results to step 6; Step 6: After receiving the oscillation analysis results provided in step 4 and step 5, unified processing is performed to generate protection control instructions. The protection instructions are converted into specific power grid control signals and sent to the power grid control center for implementing oscillation control of the power grid. At the same time, internal data transmission and time synchronization functions are provided in the system.

[0051] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A power grid oscillation monitoring system based on intelligent sensors, characterized in that: include: The intelligent sensor module is used to collect signal data of power grid operation, and has a built-in adaptive sampling mechanism to dynamically adjust the sampling frequency to capture signal changes. At the same time, the collected signal data is pre-processed and transmitted to the edge computing module to provide high-quality input for oscillation signal feature extraction; The edge computing module receives the pre-processed signal data from the intelligent sensor module, extracts the characteristic information of the oscillation signal using the signal analysis algorithm, and determines the oscillation state through the threshold dynamic adjustment method to preliminarily classify the oscillation type. On this basis, it supports collaborative analysis among multiple nodes, transmits the results of feature extraction and oscillation classification to the data fusion module, and synchronously transmits them to the cloud platform module; The data fusion module is used to receive the signal data from the intelligent sensor module and the feature extraction and oscillation classification results from the edge computing module, and uniformly process the signal data through the data integration algorithm. The integrated data has the characteristics of time alignment, providing consistent input for the cloud platform module; The cloud platform module receives the integrated signal data from the data fusion module and the oscillation classification results from the edge computing module, further identifies complex oscillation patterns, establishes a power grid topology model based on a deep learning algorithm, locates the oscillation source, and at the same time, deeply mines the oscillation pattern in combination with historical data, generates the oscillation propagation path, and transmits the analysis results to the trend analysis module; The trend analysis module predicts the future development trend of power grid oscillation based on the analysis results of oscillation mode, oscillation source location and propagation path from the cloud platform module and historical data, and generates short-term and medium-term oscillation trend data. The prediction results will be further transmitted to the communication and control module to support control decisions; The communication and control module realizes data transmission between the intelligent sensor module, edge computing module, data fusion module and cloud platform module, provides time synchronization function for modules within the system, and at the same time, receives oscillation analysis results from the cloud platform module and trend analysis module, converts them into executable protection instructions, and sends corresponding control signals to the power grid control center to achieve effective oscillation control of the power grid.

2. A power grid oscillation monitoring system based on intelligent sensors according to claim 1, characterized in that: The adaptive sampling mechanism of the intelligent sensor module dynamically adjusts the sampling frequency based on the signal fluctuation characteristics , its sampling frequency satisfies the following relationship: ， in, is the real-time sampling frequency, is the basic sampling frequency, is the adjustment factor, is the standard deviation of the signal amplitude, The preset threshold for signal fluctuation.

3. The power grid oscillation monitoring system based on intelligent sensors according to claim 1, characterized in that: The intelligent sensor module includes a signal filtering unit, which uses Kalman filtering to denoise the signal. The filtering process includes the following steps: Status prediction: , in, For the predicted state, For the previous moment The posterior estimate of the state, For the previous moment The control input, is the control matrix, is the state transfer matrix; Error covariance prediction: , in, is the forecast error covariance matrix, For the previous moment The posterior error covariance matrix of is the transpose of the state transfer matrix, is the state transfer matrix, is the process noise covariance; Status Update: , in, is the updated state estimate, For the predicted state, is the Kalman gain, is the actual observed signal of the sensor, is the observation matrix; Kalman gain calculation: , in, is the Kalman gain, is the observation noise covariance, is the observation matrix, is the transpose of the observation matrix, is the prediction error covariance matrix.

4. The power grid oscillation monitoring system based on intelligent sensors according to claim 1, characterized in that: The edge computing module extracts features from the transmission signal. The feature extraction process is based on fast Fourier transform, which converts the time domain signal x(t) into the frequency domain to satisfy the following relationship: ， in, is the frequency domain representation of the signal, x(t) is the input signal, is the frequency variable, is the time variable, is the angular frequency component of the signal, Indicates the frequency The complex exponential form of the fundamental wave.

5. The power grid oscillation monitoring system based on intelligent sensors according to claim 1, characterized in that: The edge computing module detects the oscillation state through a dynamic threshold algorithm, and its dynamic threshold The following relations are satisfied: , in, Dynamic threshold, is the mean of the signal eigenvalues, The standard deviation of the signal eigenvalues, is the threshold adjustment factor.

6. The power grid oscillation monitoring system based on intelligent sensors according to claim 1, characterized in that: The data fusion module performs weighted fusion on the signal data from the intelligent sensor, and the fused signal data is calculated by the following formula: ， in, is the fused signal data, is the original signal data from the i-th sensor, is the weight of the i-th sensor, N is the total number of signals, is the time variable; It is proportional to the inverse of the sensor measurement noise and satisfies the following relationship: ， in, is the standard deviation of the noise of the signal measured by the i-th sensor, and N is the total number of signals.

7. The power grid oscillation monitoring system based on intelligent sensors according to claim 1, characterized in that: The cloud platform module extracts the time-frequency features of the signal data through a convolutional neural network, and the convolution operation satisfies the following relationship: ， in, The position in the feature matrix of the convolution output The value of is the position in the input signal feature matrix The value of is the position in the convolution kernel matrix The weight of is the bias value, is the activation function, and The row and column size of the convolution kernel.

8. The power grid oscillation monitoring system based on smart sensors according to claim 1, characterized in that: The cloud platform module realizes oscillation source positioning through graph neural network, and the update of node features satisfies the following relationship: ， in, For node v in The feature vector of the layer, is the feature vector of node u in layer k, is the weight matrix of the kth layer, is the neighbor set of node v, , is the degree of nodes v and u, is the bias vector of the kth layer, is the activation function.

9. The power grid oscillation monitoring system based on intelligent sensors according to claim 1, characterized in that: The trend analysis module predicts the time series characteristics of the oscillation signal based on the long short-term memory network. The state update of the long short-term memory network unit satisfies the following relationship: Input Gate: Forget Gate: Output Gate: Unit status update: Hide status updates: in, , , is the activation value of the input gate, forget gate, and output gate, , , , is the weight matrix, , , , is the bias value, is the unit state, is the hidden state, is the input signal at the current moment, is the activation function, is the hyperbolic tangent function.

10. A method for monitoring power grid oscillation based on smart sensors, according to a system for monitoring power grid oscillation based on smart sensors according to any one of claims 1 to 9, characterized in that: include: Step 1: Collect signal data in power grid operation through the intelligent sensor module. The intelligent sensor module has a built-in adaptive sampling mechanism to dynamically adjust the sampling frequency according to signal changes. At the same time, the collected signal data is pre-processed and the pre-processed high-quality signal data is transmitted to step 2; Step 2, receiving the pre-processed signal data from the intelligent sensor module, processing the signal data using a signal analysis algorithm, extracting characteristic information of the oscillation signal, determining the oscillation state based on the extracted characteristic information using a threshold dynamic adjustment method, and preliminarily classifying the oscillation type. At the same time, supporting collaborative analysis among multiple nodes, transmitting the results of oscillation feature extraction and preliminary classification to step 3, and synchronously transmitting them to step 4; Step 3: Receive the signal data from the smart sensor module and the oscillation characteristics and classification results from the edge computing module, process the signal data through a data integration algorithm, and transmit the integrated data to step 4 as input for further analysis; Step 4: Receive the integrated signal data from step 3 and the oscillation classification result from step 2, analyze the signal using a deep learning algorithm, identify complex oscillation patterns, establish a power grid topology model, and locate the oscillation source by calculation. At the same time, further mine the oscillation pattern in combination with historical data, generate the oscillation propagation path, and pass the result to step 5; Step 5: receiving the analysis results of the oscillation mode, oscillation source location and propagation path provided in step 4, combining historical data, predicting the short-term and medium-term development trends of the power grid oscillation, generating oscillation trend data, and transmitting the prediction results to step 6; Step 6: After receiving the oscillation analysis results provided in step 4 and step 5, unified processing is performed to generate protection control instructions. The protection instructions are converted into specific power grid control signals and sent to the power grid control center for implementing oscillation control of the power grid. At the same time, internal data transmission and time synchronization functions are provided in the system.

Images