Automatic control method and system for transformer substation overhaul, operation and maintenance supervision based on data analysis

Abstract

The invention discloses an automatic control method and system for transformer substation maintenance, operation and maintenance supervision based on data analysis, and belongs to the technical field of transformer substation maintenance, and the method comprises the steps: collecting multi-source monitoring data of a transformer substation, including voltage, current, transformer oil temperature, partial discharge signals, GIS equipment vibration frequency and insulating oil chromatographic data; performing time sequence processing on the multi-source monitoring data to generate time sequence monitoring data, performing filtering processing on power frequency noise in the time sequence monitoring data by adopting a Daubechies wavelet basis function, and performing feature extraction to obtain time sequence processing data; inputting the time sequence processing data into a pre-constructed long short-term memory network model for fault prediction to obtain an equipment fault probability prediction value; when the equipment fault probability prediction value exceeds a preset threshold value, triggering a multi-stage early warning signal and executing an automatic control operation; and according to the maintenance record data, calculating a fault recovery rate and an operation stability index, and generating a maintenance evaluation report.

Description

Technical Field

[0001] The present invention relates to the technical field of substation maintenance and operation and maintenance, and specifically relates to an automatic control method and system for substation maintenance and operation and maintenance supervision based on data analysis. Background Art

[0002] With the continuous expansion of the scale of the power system and the continuous improvement of the intelligent level, the number and complexity of substation equipment have increased significantly. The traditional maintenance and operation and maintenance mode relying on manual inspection and experience judgment has been difficult to meet the high-reliability power supply requirements. Especially in the context of high proportion of new energy access and increasing load fluctuations, the concealment of potential equipment failure risks has increased, and the problems of low efficiency and poor real-time performance of manual inspection have become increasingly prominent. It is urgent to achieve accurate fault prediction and optimized scheduling of maintenance resources through data-driven technologies.

[0003] The prior art document CN119515340A discloses a substation maintenance method and system based on deep learning in the field of substation maintenance. It extracts and reduces the dimensionality of characteristic data such as current and voltage through the Pearson correlation coefficient method and the principal component analysis method, and constructs a fault diagnosis model using a recurrent neural network (RNN). However, its feature extraction is not optimized for the specific power frequency noise and electromagnetic interference in substations, resulting in the effective signal being easily polluted by noise. Moreover, the RNN model has a gradient attenuation problem when dealing with long-term time series dependencies, and the prediction accuracy is limited. At the same time, it lacks automatic control in case of faults and cannot achieve rapid fault isolation and closed-loop control. Summary of the Invention

[0004] The purpose of the present invention is to provide an automatic control method and system for substation maintenance and operation and maintenance supervision based on data analysis to solve the deficiencies in the background art.

[0005] The purpose of the present invention can be achieved through the following technical solutions: In the first aspect, the present invention provides an automatic control method for substation maintenance and operation and maintenance supervision based on data analysis, including the following steps: Collect multi-source monitoring data of the substation; Perform time series processing on the multi-source monitoring data to generate time series monitoring data, filter the power frequency noise in the time series monitoring data using the Daubechies wavelet basis function, and then perform feature extraction to obtain time series processed data; Input the time series processed data into a pre-constructed long short-term memory network model for fault prediction to obtain a predicted value of the equipment fault probability; When the predicted value is within the first preset range, automatically implement fault emergency blocking; when the predicted value is within the second preset range, automatically assign maintenance tasks to maintenance personnel.

[0006] Optionally, the processing of the multi-source monitoring data to obtain time series processed data includes: The multi-source monitoring data includes voltage, current, transformer oil temperature, partial discharge signal, vibration frequency of GIS equipment, and insulating oil chromatogram data; Perform time-series processing on the multi-source monitoring data according to the recorded timestamps to generate time-series monitoring data; Perform wavelet decomposition on the time-series monitoring data to extract the detail coefficients corresponding to power frequency noise; Set the detail coefficients to zero by the threshold zeroing method, and perform inverse wavelet transform to reconstruct the signal to obtain the denoised time-series monitoring data; Extract features from the denoised time-series monitoring data, including time-domain features and frequency-domain features, to generate time-series processed data.

[0007] Optionally, the time-domain features include the change rate of transformer oil temperature and the voltage peak volatility, and the frequency-domain features include the fundamental frequency amplitude of the vibration frequency and the proportion of the third harmonic.

[0008] Optionally, the construction of the long short-term memory network model for fault prediction includes: Set the model time step to 24 hours to match the daily load cycle fluctuation characteristics of the substation; The number of neurons in the hidden layer is consistent with the dimension of the equipment status parameters, and the equipment status parameters include voltage, current, oil temperature, and fundamental frequency amplitude of the vibration frequency; Initialize the forget gate weight matrix based on historical fault records to preferentially suppress the memory of non-critical time-series features; The output layer uses the Sigmoid activation function to generate the predicted value of the equipment fault probability.

[0009] Optionally, the taking of corresponding measures according to the predicted value within the preset range includes: When the equipment fault predicted value ≥ 0.8, automatically implement fault emergency blocking, including triggering the switching of the standby line and the shutdown protection; When 0.5 ≤ the equipment fault predicted value < 0.8, automatically allocate maintenance tasks, generate maintenance work orders and dynamically allocate them.

[0010] Optionally, the triggering of the standby line switching and the shutdown protection includes: Disconnect the current faulty line, close the circuit breaker of the standby power supply line to ensure seamless switching of the power supply; Remotely disconnect the high-voltage side circuit breaker of the transformer and start the forced cooling system to prevent overheating damage of the equipment.

[0011] Optionally, the generating of maintenance work orders and dynamic allocation includes: Calculate the priority weight of the work order. The priority weight is equal to the predicted value weight multiplied by the equipment fault probability predicted value, plus the equipment risk level weight multiplied by the equipment risk level; Generate maintenance work orders that include fault prediction values, equipment locations, and historical maintenance records, and mark the priorities. Determine whether to assign the maintenance work orders to high-risk teams or regular teams according to the work order priorities.

[0012] Optionally, the automatic assignment of maintenance tasks to operation and maintenance personnel includes that the operation and maintenance personnel execute the maintenance tasks and record the maintenance data. Generate a maintenance evaluation report based on the maintenance record data, including: Periodically traverse the maintenance logs to obtain the number of fault repairs, count the number of successfully repaired faults, and calculate the fault repair rate. Statistically calculate the total time of fault maintenance and calculate the operation stability of the substation. Calculate the maintenance evaluation value, compare it with the evaluation threshold, and generate a maintenance evaluation report.

[0013] In a second aspect, the present invention provides an automatic control system for substation maintenance and operation and maintenance supervision based on data analysis. Based on the automatic control method for substation maintenance and operation and maintenance supervision described in the first aspect of the present invention, the system includes: A data acquisition module, configured to collect voltage, current, transformer oil temperature, partial discharge signal, vibration frequency of GIS equipment, and insulating oil chromatogram data in real time through a substation online monitoring device and a SCADA system. A data processing module, including: a timing processing unit, configured to sort the monitoring data according to timestamps to generate timing monitoring data; a wavelet denoising unit, using Daubechies wavelet basis functions to filter power frequency noise; a feature extraction unit, configured to extract the oil temperature change rate, voltage peak volatility, vibration frequency fundamental amplitude, and third harmonic ratio. A fault prediction module, including: an LSTM model unit, configured to obtain a predicted value of equipment fault probability. An automatic control module, including a multi-level early warning unit, triggering the switching of standby lines, remote shutdown protection, or the assignment of maintenance work orders according to the predicted value; a SCADA interface unit, configured to send control instructions to the SCADA system. An evaluation and feedback module, including: an index calculation unit, configured to calculate the fault repair rate and operation stability; a report generation unit, configured to generate an evaluation report and mark the non-compliant equipment.

[0014] Optionally, the automatic control module includes an intelligent circuit breaker, configured to receive instructions from the SCADA system and perform the opening and closing operations of the line; the SCADA system sends control instructions to the circuit breaker through an API interface and monitors the line status in real time.

[0015] Compared with the existing solutions, the beneficial effects achieved by the present invention include: 1. The present invention solves the problems of large noise interference and difficult extraction of effective features in the prior art by integrating substation scenario-based data acquisition and wavelet adaptive denoising technology, improving data quality and the reliability of fault prediction.

[0016] 2. The present invention solves the problems of low utilization rate of time series features and slow convergence speed of general models by constructing an LSTM model adapted to the load cycle and a fault-driven weight initialization strategy, improving the accuracy and efficiency of fault prediction.

[0017] 3. The present invention solves the problems of rigid response of traditional single thresholds and delay in manual operation by linking multi-level warning thresholds with the automatic control of the SCADA system, improving the speed of fault isolation and power supply reliability.

[0018] 4. The present invention solves the problems of strong subjectivity and lack of quantitative basis in the existing evaluation methods by introducing a comprehensive evaluation system bound by industry standards, improving the scientific nature of maintenance management and decision-making efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The present invention will be further described below with reference to the accompanying drawings.

[0020] Figure 1 is a flowchart of an automatic control method for substation maintenance and operation supervision based on data analysis proposed by the present invention; Figure 2 is a module structure diagram of an automatic control system for substation maintenance and operation supervision based on data analysis proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0022] In Embodiment 1 of the present invention, an automatic control method for substation maintenance and operation supervision based on data analysis is provided, as Figure 1 shown, including the following steps: Step 1, collect multi-source monitoring data of the substation, including voltage, current, transformer oil temperature, partial discharge signal, vibration frequency of GIS equipment, and insulating oil chromatogram data.

[0023] Preferably, Step 1 includes: Collect the monitoring data of the substation in real time through the substation on-line monitoring device and the SCADA system.

[0024] Step 2: Perform time series processing on the multi-source monitoring data collected in Step 1 to generate time series monitoring data. Then, use the Daubechies wavelet basis function to filter the power frequency noise in the time series monitoring data, and perform feature extraction to obtain time series processed data.

[0025] Preferably, Step 2 includes: Step 2.1: Perform time series processing on the multi-source monitoring data according to the recorded timestamps to generate time series monitoring data; Step 2.2: Perform wavelet decomposition on the time series monitoring data to extract the detail coefficients corresponding to the power frequency noise; Exemplarily, the wavelet decomposition decomposes the signal into approximation coefficients (low frequency) and detail coefficients (high frequency) through multi-scale analysis. The specific formula is:

[0026] where, is the original time series monitoring signal; is the Daubechies wavelet basis function; is the decomposition level; is the translation parameter; is the detail coefficient of the

[0027] layer. According to the energy distribution characteristics of the substation power frequency noise (50 Hz), selecting 5-layer decomposition can effectively separate the noise from the effective signal. The power frequency noise corresponds to the detail coefficients of the 3rd layer, and the threshold setting to zero range is 10% of the absolute value of the detail coefficients of this layer. Specifically, the threshold setting to zero method processes the detail coefficients through the following formula:

[0028] where, is the processed detail coefficient; is the threshold, which is set to 10% of the absolute value of the detail coefficient exemplarily.

[0029] Step 2.4: Perform feature extraction on the denoised time series monitoring data. The time domain features include the transformer oil temperature change rate and the voltage peak volatility, and the frequency domain features include the fundamental frequency amplitude of the vibration frequency and the proportion of the third harmonic, to generate time series processed data.

[0030] Further preferably, the time domain features include: Oil temperature change rate:

[0031] where, is the oil temperature at the current time point; is the sampling interval. Exemplarily, it can be set to 5 minutes; Voltage peak volatility:

[0032] Among them, is the voltage sequence within the time window; The frequency domain features include: Amplitude of the fundamental vibration frequency: calculated by Fast Fourier Transform (FFT):

[0033] Among them, is the rated vibration frequency of the device; Ratio of the third harmonic:

[0034] Among them, is the amplitude of the frequency domain signal; is the harmonic order. In the substation environment, the 3rd harmonic with high correlation to mechanical looseness can be focused on.

[0035] It should be noted that the noise processing mechanism based on Fourier transform in the current technical system is mainly applicable to the analysis of stationary signals and is difficult to effectively cope with non-stationary characteristics such as transient electromagnetic interference in the complex time-varying noise environment of substations. This technical limitation leads to the inevitable weakening of effective high-frequency components during the noise suppression process by traditional frequency domain filtering - especially the partial discharge pulse characteristics reflecting the insulation state of the equipment. And the present invention can better accurately identify and specifically eliminate the power frequency interference components through the introduction of wavelet multi-scale analysis and adaptive threshold zeroing, a decomposition strategy that dynamically matches the time-frequency characteristics of the signal, while retaining the key high-frequency feature components representing the equipment state, and constructs a better balance mechanism of noise suppression and feature retention in the time-frequency joint domain.

[0036] Step 3, input the time-series processed data into a pre-constructed long short-term memory network model for fault prediction to obtain the predicted value of the equipment fault probability.

[0037] Specifically, among them, the LSTM model is a neural network model specifically used for processing time-series data, capturing the long-term dependencies and dynamic change trends in the data. During the training process, by adjusting the parameters and structure of the model, the model can accurately learn the internal mapping relationship between the equipment state and the input features, so as to achieve accurate prediction of the future state of substation equipment; Preferably, in the said Step 3, constructing a long short-term memory network model for fault prediction includes: Step 3.1: Set the model time step to 24 hours to match the daily load cycle fluctuation characteristics of the substation; Further preferably, the time step of the LSTM model is defined as the window length of the input sequence and can be set to 24 hours to match the daily periodicity of the substation load (such as peak load during the day and low load at night).

[0038] Step 3.2: The number of neurons in the hidden layer is the same as the dimension of the equipment status parameters, and the equipment status parameters include voltage, current, oil temperature, and fundamental frequency amplitude of vibration frequency; Further preferably, the number of neurons in the hidden layer is the same as the dimension of the input features to ensure that the model capacity matches the data complexity and avoid overfitting or underfitting; the dimension of the hidden state vector is to capture temporal dependence relationships.

[0039] Step 3.3: Initialize the forget gate weight matrix based on historical fault records to preferentially suppress the memory of non-critical temporal features; Further preferably, the initialization strategy of the forget gate weight matrix is as follows:

[0040] where is the basic weight matrix initialized with a standard normal distribution; a mask matrix that attenuates the weights of non-critical features according to historical fault records. Exemplarily, it can be set to attenuate by 50%:

[0041] The mask matrix suppresses the memory weights of irrelevant features (such as ambient temperature) through historical fault analysis (e.g., transformer overload events are strongly correlated with oil temperature).

[0042] Step 3.4: The output layer uses a Sigmoid activation function to generate the predicted value of the equipment failure probability.

[0043] Further preferably, Step 3.4 includes that the output layer calculates the predicted value of the failure probability:

[0044] where is the Sigmoid activation function that maps the output to the interval; is the weight matrix of the output layer with a dimension of ; is the bias term of the output layer.

[0045] The LSTM model of the present invention improves the accuracy and efficiency of substation fault prediction by adapting the time step in cycles, matching the dimensions of the hidden layer, initializing the weights driven by faults, and probabilistically outputting, and solves the problems of low utilization rate of time series features, slow convergence speed, and high false alarm rate in traditional methods.

[0046] Exemplarily, an input layer is constructed to receive input data, and the input data is sequentially input into the LSTM layer according to time steps for processing; An LSTM layer is constructed to perform in-depth feature learning and processing on the input data to obtain the hidden state at the last time step; Among them, the LSTM layer includes a forget gate, an input gate, cell state update, and an output gate; Through the formula Perform the forget gate operation on the input data; where Is the output of the forget gate at the current time step t; t is the current time step, t = 1, 2,..., T, and T is the last time step; Is the weight matrix of the forget gate, used to perform a linear transformation on the input data; Is the hidden state at the previous time step; Is the input data at the current time step t; Is the bias vector of the forget gate, used to perform a translation adjustment on the result of the linear transformation to better fit the data; Is the sigmoid activation function; Through the formula And Perform the input gate operation on the input data; where Is the output of the input gate at the current time step t; Is the weight matrix of the input gate; Is the bias vector of the input gate; The candidate cell state at the current time step t; tanh is the hyperbolic tangent function; Is the weight matrix used to calculate the candidate cell state; Is the bias vector used to calculate the candidate cell state; Through the formula Combine the forget gate output and the input gate output values to complete the cell state update operation; where Is the cell state at the current time step t; Is the cell state at the previous time step t - 1; information is screened and updated through the forget gate and the input gate; Through the formula And Complete the output gate operation; where The output value of the output gate at the current time step t; is the weight matrix of the output gate; is the bias vector of the output gate; is the hidden state at the current time step t; Construct an output layer to obtain the hidden state of the last time step after processing by the LSTM layer, and output the predicted output value corresponding to the input data; The involved calculation formula is: ; In the formula, is the predicted output value of the model; is the weight matrix of the output layer; is the bias vector of the output layer; is the hidden state of the last time step; Set the prediction threshold to 0.5. If the calculated predicted output value is greater than or equal to the prediction threshold, the prediction result corresponding to the input data is the fault state; if the calculated predicted output value is less than the prediction threshold, the prediction result corresponding to the input data is the normal state; among them, the fault state or the normal state constitutes the prediction result; Exemplarily, the training of the LSTM model further includes: Traverse the historical monitoring data of the substation to obtain historical time-series processed data. According to the fault records, obtain the true labels corresponding to the historical time-series processed data, and divide the historical time-series processed data into a training set and a validation set according to a preset ratio; among them, the true labels include normal labels or fault labels; the preset ratio is usually set to 7:3; Further preferably, the training of the LSTM model with the training set as the input data includes: Use binary cross-entropy as the loss function to measure the difference between the model prediction result and the true label, and optimize the model through the Adam optimizer to minimize the loss function value; among them, the Adam optimizer is an existing technology, and the optimization process will not be elaborated here; Specifically, based on the prediction result obtained by inputting the training set into the LSTM model, compare the prediction result with the true label and calculate the loss function; The calculation formula is as follows: ; In the formula, L is the loss function value; represents the true label corresponding to the mth sample; among them, is the fault label, is the normal label; m = 1, 2,..., M, and M is the total number of samples in the training set; is the predicted output value corresponding to the mth sample; The evaluation of the performance of the LSTM model using the validation set includes: Obtain the first predicted output value of the validation set output and make a judgment based on the prediction threshold to obtain the corresponding first prediction result; compare the first prediction result with its true label and calculate the prediction accuracy; The calculation formula is as follows: ; Where Q is the prediction accuracy; YZ is the number of verification samples whose first prediction result is consistent with its true label; R is the total number of verification samples; If the calculated prediction accuracy is greater than or equal to the preset standard value, the performance of the LSTM model meets the standard and a trained LSTM model is obtained; if the calculated prediction accuracy is less than the preset standard value, the LSTM model needs to be optimized and adjusted; the preset standard value is determined according to the specific requirements of the model in actual applications; It should be noted that the optimization and adjustment of the LSTM model specifically includes: adjusting the number of LSTM layers and the number of neurons in each layer to capture more complex time dependencies; adding a Dropout layer or regularization technology to prevent model overfitting; adjusting model parameters, such as changing the batch size, to find the best batch size suitable for the current task; and continuously optimizing the LSTM model until its prediction accuracy is improved to a preset standard value.

[0047] Step 4: When the predicted value of the equipment failure probability exceeds the preset threshold, a multi-level warning signal is triggered and an automatic control operation is performed.

[0048] Preferably, step 4 comprises: When the equipment failure prediction value is ≥0.8, the backup line switching and shutdown protection are triggered, and the SCADA system is linked to perform protection operations; When 0.5≤equipment failure prediction value<0.8, a maintenance work order is generated and dynamically allocated.

[0049] Exemplarily, the triggering of standby line switching and shutdown protection includes: Disconnect the current faulty line and close the circuit breaker of the backup power supply line to ensure seamless switching of power supply; Remotely disconnect the transformer high-voltage side circuit breaker and start the forced cooling system to prevent equipment from overheating and damage.

[0050] Further preferably, the generating and dynamically allocating maintenance work orders includes: Calculate the ticket priority weight:

[0051] in, is the predicted value weight, is the equipment risk level weight, is the risk level of the equipment; Exemplarily, the predicted value weight can be set to 0.7, and the equipment risk level weight can be set to 0.3. The weight values are determined through actual operation and maintenance data analysis to balance the prediction probability and the inherent risk of the equipment. Based on the comprehensive scoring of the equipment's historical failure rate, operation years, and maintenance records, the equipment risk level can be divided into levels 1-5, with level 5 being the highest risk. Generate a maintenance work order that includes the fault prediction value, equipment location, and historical maintenance records, and mark the priority.

[0052] Determine whether to assign the maintenance work order to a high-risk team or a regular team according to the calculated value of the work order priority weight. Exemplarily, if the priority weight ≥ 0.6, it is assigned to the high-risk team; otherwise, it is assigned to the regular team.

[0053] It should be noted that, in view of the problems of single-threshold triggering and manual control delay in the prior art, the present invention solves the problems of insufficient distinction of fault severity and response lag through multi-level warning thresholds and automatic control linkage, and improves the fault handling efficiency and automation level. In view of the problem of low resource utilization rate caused by static priority allocation, the present invention quantifies the dynamic allocation strategy of priority through a weight formula, solves the problem of mismatch between work order allocation and real-time risk, and improves the response speed of high-risk equipment and the fault repair rate. In view of the lack of monitoring of overload of standby lines, the present invention solves the risk of secondary faults caused by passive operation and maintenance through real-time monitoring of the load rate and a secondary warning system, and improves the operation stability of the power system.

[0054] Step 5: Calculate the fault repair rate and operation stability index based on the maintenance record data, and generate a maintenance evaluation report.

[0055] Preferably, the said step 5 includes: Step 5.1: Periodically traverse the maintenance log to obtain the number of fault repairs, count the number of successfully repaired faults, and calculate the fault repair rate:

[0056] Wherein, is the number of successfully repaired faults, is the number of fault repairs during the period; Step 5.2: Statistically calculate the total fault repair time and calculate the operation stability of the substation:

[0057] Wherein, is the total fault repair time, is the cycle time; Step 5.3: Calculate the maintenance evaluation value, compare it with the evaluation threshold, and generate a maintenance evaluation report. The formula for the maintenance evaluation value is:

[0058] Among them, and are the weight coefficients corresponding to the fault repair rate and operation stability respectively; Exemplarily, and can be set to 0.6 and 0.4 respectively; The comparison with the evaluation threshold includes that if the calculated maintenance evaluation value is greater than or equal to the evaluation threshold, it indicates that the comprehensive performance of the current maintenance work in terms of fault repair and maintaining the operation stability of the substation is qualified; otherwise, the overall quality of the maintenance work needs to be improved; Among them, the evaluation threshold is set by industry insiders according to the power industry standards and specifications; In the embodiments of the present invention, by calculating the maintenance evaluation value and comparing it with the evaluation threshold numerically, a comprehensive and quantitative evaluation of the maintenance work is achieved, ensuring the normal operation of the substation and power supply reliability.

[0059] Through comprehensive and in-depth data collection and integration, the present invention improves the utilization value of data; By constructing an LSTM model and training it with a large amount of historical time-series processing data, the system can deeply mine the time-series characteristics and potential laws in the substation monitoring data, so as to accurately predict equipment failures; Compared with the traditional early warning methods based on fixed thresholds or simple rules, this system can detect the subtle changes in the equipment operation state in advance and send out early warning signals in time before the failure occurs; Based on the accurate fault prediction results, the present invention reasonably arranges maintenance personnel for maintenance, avoiding blind patrols and unnecessary maintenance operations. At the same time, by calculating indicators such as the fault repair rate and operation stability, it provides data support for the optimization of maintenance resources, improves maintenance efficiency, and reduces operation and maintenance costs.

[0060] In Embodiment 2 of the present invention, an automatic control system for substation maintenance and operation supervision based on data analysis is provided, as Figure 2 shown, based on an automatic control method for substation maintenance and operation supervision based on data analysis described in Embodiment 1 of the present invention, this system includes: A data collection module, used to collect voltage, current, transformer oil temperature, partial discharge signal, GIS equipment vibration frequency, and insulating oil chromatogram data in real time through substation on-line monitoring devices and the SCADA system; A data processing module, including: A time-series processing unit, used to sort the monitoring data according to the time stamp to generate time-series monitoring data; A wavelet denoising unit, using the Daubechies wavelet basis function to filter the power frequency noise; A feature extraction unit, used to extract the oil temperature change rate, voltage peak volatility, vibration frequency fundamental amplitude, and third harmonic ratio; The fault prediction module includes: an LSTM model unit with a time step set to 24 hours, the dimension of the hidden layer matching the device parameters, and the weight of the forget gate initialized based on historical faults; a prediction output unit that uses the Sigmoid function to generate a predicted fault probability value. The automatic control module includes a multi-level early warning unit that triggers the switching of backup lines, remote shutdown protection, or the assignment of maintenance work orders according to the predicted value; an SCADA interface unit for sending control instructions to the SCADA system. The evaluation and feedback module includes: a metric calculation unit for calculating the fault repair rate and operation stability; a report generation unit that generates an evaluation report according to relevant industry test regulations and marks the equipment that fails to meet the standards.

[0061] Further, the automatic control module includes intelligent circuit breakers, such as vacuum circuit breakers and SF6 circuit breakers, which can receive instructions from the SCADA system and perform the opening and closing operations of the line; the SCADA system sends control instructions to the circuit breaker through the API interface and monitors the line status in real time.

[0062] Further, the automatic control module also includes a remote terminal unit (RTU) and a programmable logic controller (PLC) for receiving shutdown instructions, controlling the opening operation of the transformer circuit breaker, and starting the cooling system.

[0063] Further, the evaluation and feedback module includes a work order management system that supports real-time data synchronization with mobile terminals.

[0064] It should be noted that the system of the present invention realizes the full-process automation from early warning to execution by integrating intelligent circuit breakers, SCADA systems, RTU / PLC, and work order management systems, solves the problems of slow response, unreasonable resource allocation, and isolated devices in traditional technologies, and significantly improves the efficiency and safety of substation fault handling.

[0065] In the embodiments provided by the present invention, it should be understood that the disclosed system can be implemented in other ways. For example, the above-described invention embodiments are merely illustrative. For example, the division of modules is only a logical function division, and there can be other division methods in actual implementation.

[0066] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they can be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0067] In addition, in each embodiment of the present invention, each functional module can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module. The above integrated module can be implemented in the form of hardware, or in the form of a combination of hardware and software functional modules.

[0068] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the basic characteristics of the present invention, the present invention can be implemented in other specific forms.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An automatic control method for substation maintenance and operation supervision based on data analysis, characterized in that, It includes the following steps: Collect multi-source monitoring data of the substation; Perform time-series processing on the multi-source monitoring data to generate time-series monitoring data, filter the power frequency noise in the time-series monitoring data using Daubechies wavelet basis function, and then perform feature extraction to obtain time-series processed data; Input the time-series processed data into a pre-constructed long short-term memory network model for fault prediction to obtain the predicted value of the equipment fault probability; When the predicted value is within the first preset range, automatically implement fault emergency blocking; When the predicted value is within the second preset range, automatically assign maintenance tasks to the operation and maintenance personnel.

2. The automatic control method for substation maintenance and operation supervision based on data analysis according to claim 1, characterized in that: The processing of the multi-source monitoring data to obtain time-series processed data includes: The multi-source monitoring data includes voltage, current, transformer oil temperature, partial discharge signal, GIS equipment vibration frequency and insulating oil chromatogram data; Perform time-series processing on the multi-source monitoring data according to the recorded timestamp to generate time-series monitoring data; Perform wavelet decomposition on the time-series monitoring data to extract the detail coefficients corresponding to the power frequency noise; Set the detail coefficients to zero by the threshold zeroing method, and perform wavelet inverse transform to reconstruct the signal to obtain the denoised time-series monitoring data; Perform feature extraction on the denoised time-series monitoring data, including time-domain features and frequency-domain features, to generate time-series processed data.

3. The automatic control method for substation maintenance and operation supervision based on data analysis according to claim 2, characterized in that: The time-domain features include the change rate of transformer oil temperature and the voltage peak volatility, and the frequency-domain features include the fundamental frequency amplitude of the vibration frequency and the proportion of the third harmonic.

4. The automatic control method for substation maintenance and operation supervision based on data analysis according to claim 3, characterized in that: The construction of the long short-term memory network model for fault prediction includes: Set the model time step to 24 hours to match the daily load cycle fluctuation characteristics of the substation; The number of neurons in the hidden layer is the same as the dimension of the equipment state parameters, and the equipment state parameters include voltage, current, oil temperature, and fundamental frequency amplitude of the vibration frequency; Initialize the forget gate weight matrix based on historical fault records to preferentially suppress the memory of non-critical time-series features; The output layer uses the Sigmoid activation function to generate the predicted value of the equipment fault probability.

5. The automatic control method for substation maintenance and operation supervision based on data analysis according to claim 4, characterized in that: The corresponding measures taken according to the preset range where the predicted value is located include: When the predicted value of the equipment fault ≥ 0.8, automatically implement fault emergency blocking, including triggering the switching of the standby line and the shutdown protection; When 0.5 ≤ the predicted value of the equipment fault < 0.8, automatically assign maintenance tasks, generate maintenance work orders and dynamically allocate them.

6. The automatic control method for substation maintenance and operation supervision based on data analysis according to claim 5, characterized in that: The triggering of the standby line switching and the shutdown protection includes: Disconnect the current faulty line, close the circuit breaker of the standby power supply line to ensure seamless switching of the power supply; Remotely disconnect the high-voltage side circuit breaker of the transformer and start the forced cooling system to prevent equipment overheating and damage.

7. The automatic control method for substation maintenance and operation supervision based on data analysis according to claim 6, characterized in that: The generating and dynamically allocating maintenance work orders includes: Calculating the work order priority weight, where the priority weight is equal to the predicted value weight multiplied by the predicted value of the equipment failure probability, plus the equipment risk level weight multiplied by the equipment risk level; Generating a maintenance work order including the predicted value of the failure, the equipment location, and the historical maintenance records, and marking the priority; Determining to allocate the maintenance work order to a high-risk team or a regular team according to the work order priority.

8. The automatic control method for substation maintenance and operation supervision based on data analysis according to claim 7, characterized in that: The automatically allocating maintenance tasks to maintenance personnel includes that the maintenance personnel execute the maintenance tasks and record the maintenance data; Generating a maintenance evaluation report according to the maintenance record data, including: Periodically traversing the maintenance log to obtain the number of fault repairs, counting the number of successfully repaired faults, and calculating the fault repair rate; Statistical total time of fault repair, and calculating the operation stability of the substation; Calculating the maintenance evaluation value, comparing it with the evaluation threshold, and generating a maintenance evaluation report.

9. The automatic control system for substation maintenance and operation supervision based on data analysis, based on the automatic control method for substation maintenance and operation supervision based on data analysis according to any one of claims 1-8, characterized in that, The system includes: A data acquisition module for real-time collecting voltage, current, transformer oil temperature, partial discharge signal, GIS equipment vibration frequency, and insulating oil chromatogram data through substation on-line monitoring devices and the SCADA system; A data processing module, including: a time series processing unit for sorting the monitoring data according to the time stamp to generate time series monitoring data; a wavelet denoising unit for filtering power frequency noise using the Daubechies wavelet basis function; a feature extraction unit for extracting the oil temperature change rate, voltage peak volatility, vibration frequency fundamental amplitude, and third harmonic ratio; A fault prediction module, including: an LSTM model unit for obtaining the predicted value of the equipment failure probability; An automatic control module, including a multi-level warning unit for triggering standby line switching, remote shutdown protection, or maintenance work order allocation according to the predicted value; an SCADA interface unit for sending control instructions to the SCADA system; An evaluation feedback module, including: an index calculation unit for calculating the fault repair rate and operation stability; a report generation unit for generating an evaluation report and marking the equipment that fails to meet the standard.

10. The automatic control system for substation maintenance and operation supervision based on data analysis according to claim 9, characterized in that: The automatic control module includes an intelligent circuit breaker for receiving instructions from the SCADA system and performing the opening and closing operations of the line; the SCADA system sends control instructions to the circuit breaker through the API interface and monitors the line status in real time.

Images