Fault risk sensing and early warning method for high-low voltage switchgear assembly

Abstract

The invention relates to the technical field of electric power system safety early warning, and discloses a fault risk sensing early warning method for a high-low voltage switchgear assembly, and the method carries out the data preprocessing and feature extraction through collecting the state monitoring data, environmental parameters and basic attribute information of a device in historical operation. And based on the feature data sets, constructing a fault risk prediction model to predict the fault risk probability of the equipment in a period of time in the future, obtaining the operation state data and environmental parameters of the equipment in real time, and inputting the prediction model to obtain the fault risk probability. Meanwhile, the equipment is monitored in real time, original monitoring data is cleaned, abnormal values are removed, real-time monitoring data is compared with prediction data, whether the difference exceeds a threshold value or not is judged according to a preset risk threshold value, and risk points are marked. And evaluating the equipment fault risk according to the risk point and the evaluation strategy, and sending a result to related personnel to take prevention measures.

Description

Technical Field

[0001] The present invention relates to the technical field of power system safety early warning, and in particular to a fault risk perception and early warning method for high and low voltage complete switchgear. Background Art

[0002] In the power system, high and low voltage switchgear, as key components, bears important responsibilities for power distribution, control and protection. The stable operation of these devices is directly related to the safety and reliability of the power system. As the equipment operates in a complex and changeable environment for a long time, it is affected by voltage fluctuations, current shocks, ambient temperature changes, humidity, and the aging of the equipment itself. Its performance will gradually decline, and may even cause failures, posing a threat to the normal operation of the power system. Traditional high and low voltage switchgear management mainly relies on regular inspections and post-maintenance, which has obvious limitations. Regular inspections are difficult to detect potential equipment failure risks in a timely manner, and can only be handled passively after the failure occurs, affecting the stability and reliability of the power system; post-maintenance is often accompanied by high maintenance costs and long power outages, which brings inconvenience and economic losses to power users and power companies.

[0003] With the rapid development of the Internet of Things, big data and artificial intelligence technologies, intelligent management of high and low voltage switchgear has become possible. By collecting the equipment's status monitoring data, environmental parameters and basic attribute information in historical operation, the equipment's operating status can be comprehensively and accurately evaluated. However, most of the existing fault warning methods are based on a single data source or simple threshold judgment, which makes it difficult to accurately predict the probability of equipment failure risk in the future.

[0004] Therefore, there is an urgent need for a high and low voltage switchgear failure risk perception and early warning method that can comprehensively consider equipment status monitoring data, environmental parameters and basic equipment attribute information. This method needs to be able to build an accurate failure risk prediction model, monitor the operating status of the equipment in real time, and accurately assess the failure risk of the equipment, so that preventive measures can be taken in time before the failure occurs to ensure the stable operation of the power system. Summary of the invention

[0005] The purpose of the present invention is to provide a fault risk perception and early warning method for high and low voltage complete switchgear to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a fault risk perception and early warning method for high and low voltage complete switchgear, the method comprising: Step 1: Collect the status monitoring data, environmental parameters and basic attribute information of high and low voltage complete switchgear in historical operation; the status monitoring data includes voltage, current, power factor, temperature, and partial discharge level; the environmental parameters include ambient temperature, humidity, and altitude; the basic attribute information of the equipment includes equipment model, manufacturer, commissioning date, and design life; Step 2: Preprocess and extract features of the data collected in step 1 to construct a feature data set; Step 3: Build a fault risk prediction model for high and low voltage switchgear. This model is based on the feature data set, combined with environmental parameters and basic equipment properties, to predict the probability of equipment failure risk in the future. Step 4: Obtain the current operating status data and environmental parameters of the high and low voltage switchgear; Step 5: Input the data obtained in step 4 into the fault risk prediction model, and receive the predicted fault risk probability output by the model; Step 6: Monitor the high and low voltage switchgear in real time, obtain the actual status monitoring data, and clean the original monitoring data to remove abnormal values; Step 7: Compare the real-time monitoring data obtained in step 6 with the predicted data output by the model in step 5, and align the timestamps; Step 8: Determine whether the difference between the real-time monitoring data and the predicted data exceeds the threshold according to the preset risk threshold. If so, mark it as a risk point; Step 9: Based on the discovered risk points and the preset assessment strategy, assess the failure risk of the high and low voltage switchgear and send the assessment results to relevant personnel.

[0007] Preferably, the feature extraction method comprises: Step 201: Calculate the statistical characteristics of the condition monitoring data, including mean, variance, maximum value, minimum value and rate of change; Step 202: Combine the selected environmental parameters with the basic attribute information of the equipment to extract cross features; Step 203: Standardize all features obtained in step 201 and step 202 so that each feature has zero mean and unit variance, construct a covariance matrix of feature data, calculate eigenvalues ​​and eigenvectors of the covariance matrix, sort the eigenvalues ​​from large to small, and select the eigenvectors corresponding to the first k eigenvalues ​​with cumulative contribution rates exceeding 90% as principal components; Step 204: Project the original feature data onto the selected principal component to obtain a feature data set after dimensionality reduction.

[0008] Preferably, in step 3, a deep neural network (DNN) algorithm is used to construct a fault risk prediction model.

[0009] Preferably, the step of building a fault risk prediction model includes: Step 301: Use the feature data set as input and divide it into a training set and a validation set. The training set is used for model training, and the validation set is used for model selection. Step 302: construct a deep neural network DNN structure, set an input layer, multiple hidden layers and an output layer, and select a suitable activation function; Step 303: Train the DNN model using the training set, optimize the model parameters by stochastic gradient descent method, and minimize the cross entropy loss between the predicted fault risk probability and the actual fault situation; Step 304: Use the validation set to select the trained DNN model and evaluate the model performance based on accuracy, recall, and F1 score; Step 305: According to the evaluation results, adjust the hyperparameters of the DNN model, including the number of hidden layers, the number of neurons in each layer, and the learning rate, until the model performance is satisfactory; Step 306: Save the trained fault risk prediction model.

[0010] Preferably, the network structure of the fault risk prediction model is: Input layer: receives feature data sets, including statistical features of condition monitoring data, cross-features of environmental parameters and basic equipment attributes, and principal component analysis results; Hidden layer: contains three fully connected hidden layers, each with 128, 64, and 32 neurons, using the ReLU activation function; Output layer: receives the output of the last hidden layer and outputs the failure risk probability through a Sigmoid activation function.

[0011] Preferably, the DNN model is iteratively trained using the training set data. In each iteration, a batch of feature data sets are fed into the DNN model, the output of each layer is calculated by forward propagation, and the model parameters are updated by back propagation. The cross entropy loss function is used to evaluate the difference between the model prediction probability and the actual label. The cross entropy loss function is defined as:

[0012] in, is the sample size, is the true label of the sample, fault is 1, no fault is 0, is the failure risk probability predicted by the model.

[0013] Preferably, in step 8, the steps of implementing risk point marking include: Step 801: aligning the timestamps of real-time monitoring data and prediction data; Step 802: Setting a fault risk threshold, which is preset based on historical fault data and equipment type; Step 803: For each time point, the difference between the real-time monitoring data and the predicted data is calculated, which represents the deviation of the corresponding relationship between the actual state monitoring index and the predicted risk probability; Step 804: Compare the calculated difference with the preset threshold value, and if it exceeds the threshold value, mark it as a risk point, and record the timestamp, difference value and corresponding monitoring indicator.

[0014] Preferably, in step 9, the step of fault risk assessment includes: Step 901: Count the risk points marked in step 8, and record the timestamp, monitoring index, and difference value; Step 902: Analyze risk points according to the preset assessment strategy, taking into account the number of risk points, the size of the difference value, and the duration; Step 903: Generate a failure risk assessment report, including an overview of equipment failure risks, detailed information on risk points, and recommended maintenance measures; Step 904: Send the fault risk assessment report to relevant personnel.

[0015] Preferably, the preset evaluation strategy includes: When the number of risk points exceeds 5 within 24 consecutive hours, and the difference values ​​of at least 3 risk points are greater than 2 times the preset threshold, the device is judged to be in a high failure risk state; When the risk point lasts for more than 48 hours and the difference value continues to increase, it is determined that the equipment may have serious hidden dangers of failure and needs to be inspected immediately.

[0016] Preferably, in step 2, the data preprocessing method includes: For missing values, the average method of the previous and next values ​​was used to fill them; For outliers, the 3σ principle is used, and data points beyond the range of three times the standard deviation of the mean are regarded as outliers and replaced with the median; Using the Min-Max normalization method, each eigenvalue is linearly transformed to the [0,1] interval. The normalization formula is:

[0017] in, is the original data, is the minimum value of the data, is the maximum value of the data, is the standardized data.

[0018] Compared with the prior art, the present invention has the following beneficial effects: The present invention constructs a comprehensive feature data set by comprehensively considering the historical status monitoring data, environmental parameters and basic attribute information of high and low voltage complete switchgear. Based on this data set, the fault risk prediction model can more accurately predict the probability of equipment failure risk in the future, significantly improving the accuracy and reliability of fault prediction. The method of the present invention includes real-time monitoring of the operating status of the equipment and obtaining actual status monitoring data. By comparing these data with the output of the fault risk prediction model, abnormal changes in the operation of the equipment can be discovered in time, thereby giving an early warning before the failure occurs, effectively avoiding power system interruptions and losses caused by the failure.

[0019] Through accurate fault risk prediction and real-time monitoring, the present invention can detect potential fault risks of equipment in advance and take corresponding preventive measures. This greatly reduces the need for subsequent maintenance, reduces maintenance costs, shortens power outages caused by faults, and improves the availability and stability of the power system. The application of the method of the present invention helps to timely discover and deal with the fault risks of high and low voltage complete switchgear, thereby reducing the impact of equipment failures on the power system. This helps to improve the safety and reliability of the entire power system and ensure the stability and continuity of power supply.

[0020] The present invention combines the Internet of Things, big data and artificial intelligence technologies to achieve intelligent management of high and low voltage switchgear. This management method not only improves the efficiency of equipment management, but also provides strong support for the intelligent development of power systems and promotes technological progress and upgrading in the power industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 A processing step diagram of a fault risk perception and early warning method for high and low voltage complete switchgear according to the present invention; Figure 2 The whole feature extraction flow chart from the original data to the feature data set after dimensionality reduction; Figure 3 Flowchart for building a failure risk prediction model. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] See also Figure 1-3The present invention provides a technical solution: a fault risk perception and early warning method for high and low voltage complete switchgear, the method comprising: Step 1: Collect the status monitoring data, environmental parameters and basic attribute information of high and low voltage switchgear in historical operation.

[0024] Condition monitoring data includes but is not limited to voltage, current, power factor, temperature, partial discharge level, etc. These data are collected in real time through sensors installed on the equipment.

[0025] Environmental parameters include ambient temperature, humidity, altitude, etc. These parameters reflect the external environmental conditions in which the equipment operates.

[0026] The basic attribute information of the equipment includes the equipment model, manufacturer, commissioning date, design life, etc. This information provides the basic identity and life characteristics of the equipment.

[0027] Step 2: Preprocess and extract features from the data collected in step 1 to build a feature dataset.

[0028] Preprocessing includes data cleaning, missing value filling, data normalization, etc. to ensure the accuracy and consistency of the data.

[0029] Feature extraction is to extract features useful for fault risk prediction from raw data, such as voltage fluctuation rate, current peak value, temperature change trend, etc.

[0030] Step 3: Construct a fault risk prediction model for high and low voltage switchgear.

[0031] The model is based on feature data sets, combined with environmental parameters and basic equipment properties, and uses machine learning or deep learning algorithms such as random forests, support vector machines, neural networks, etc. to predict the probability of equipment failure risk in the future.

[0032] During the model training process, fault cases in historical data are used as labels, and model parameters are adjusted through an iterative optimization algorithm to improve prediction accuracy.

[0033] Step 4: Obtain the current operating status data and environmental parameters of the high and low voltage switchgear.

[0034] This data is obtained through the real-time monitoring system and used for subsequent failure risk assessment.

[0035] Step 5: Input the data obtained in step 4 into the fault risk prediction model, and receive the predicted fault risk probability output by the model.

[0036] Based on the input data and historical learning experience, the model outputs the failure risk probability of the equipment in the future.

[0037] Step 6: Monitor the high and low voltage switchgear in real time, obtain the actual status monitoring data, and clean the original monitoring data to remove abnormal values.

[0038] Real-time monitoring data is collected in real time through sensors, and the data cleaning process includes removing noise, correcting erroneous data, and eliminating outliers to ensure the accuracy of the data.

[0039] Step 7: Compare the real-time monitoring data obtained in step 6 with the predicted data output by the model in step 5 and align the timestamps.

[0040] By aligning timestamps, the temporal consistency of real-time monitoring data and forecast data is ensured, facilitating subsequent comparison and analysis.

[0041] Step 8: Determine whether the difference between the real-time monitoring data and the predicted data exceeds the threshold based on the preset risk threshold.

[0042] Set a reasonable risk threshold. When the difference between real-time monitoring data and predicted data exceeds the threshold, it is marked as a risk point, indicating that the equipment may be at risk of failure.

[0043] Step 9: Based on the discovered risk points and the preset assessment strategy, assess the failure risk of the high and low voltage switchgear and send the assessment results to relevant personnel.

[0044] The assessment strategy includes risk level classification, risk assessment indicator setting, etc., and conducts a comprehensive assessment of the equipment failure risk based on the number and severity of risk points.

[0045] The evaluation results are sent to equipment managers or maintenance personnel via email, text messages, system notifications, etc., so that they can take timely measures to prevent or handle failures.

[0046] The present invention will be further described below in conjunction with Examples 1 to 3:

[0047] Embodiment 1: This embodiment provides a feature extraction method in a method for sensing and warning of fault risks of high and low voltage complete switchgear. The method specifically comprises the following steps: Step 201: Calculate statistical characteristics of condition monitoring data.

[0048] For the collected condition monitoring data, such as voltage, current, power factor, temperature, partial discharge level, etc., their statistical characteristics are calculated first. Specifically, the mean of each condition monitoring data is calculated to reflect its average level; the variance is calculated to measure the degree of dispersion of the data; the maximum and minimum values ​​are found to understand the fluctuation range of the data; and the rate of change is calculated to capture the trend of data changes over time.

[0049] For example, for voltage data, its mean, variance, maximum value, minimum value and rate of change over a period of time can be calculated. These statistical features can fully describe the stability and fluctuation of the voltage.

[0050] Step 202: Combine the selected environmental parameters with the basic attribute information of the equipment to extract cross-features.

[0051] Environmental parameters such as ambient temperature, humidity, altitude, etc., as well as basic equipment attribute information such as equipment model, manufacturer, commissioning date, design life, etc., are all important factors affecting the risk of equipment failure.

[0052] Combine these parameters and information to extract cross-features between them. For example, the combined features of ambient temperature and equipment model can be considered to reflect the performance of different models of equipment at different temperatures; or the combined features of commissioning date and design life can be considered to evaluate the remaining service life of the equipment.

[0053] Step 203: Standardize all features obtained in step 201 and step 202, construct a covariance matrix of feature data, and perform principal component analysis.

[0054] First, all features are standardized so that each feature has zero mean and unit variance. This is to ensure that different features are comparable in subsequent analysis. Then, the covariance matrix of the feature data is constructed to reflect the correlation between the features. Finally, the eigenvalues ​​and eigenvectors of the covariance matrix are calculated. The eigenvalues ​​reflect the variance explained by each principal component, and the eigenvectors define the direction of each principal component.

[0055] According to the order of eigenvalues ​​from large to small, the eigenvectors corresponding to the first k eigenvalues ​​with cumulative contribution rates exceeding 90% are selected as principal components. This is done to retain the main information of the data while reducing the dimension of the data.

[0056] Step 204: Project the original feature data onto the selected principal component to obtain a feature data set after dimensionality reduction.

[0057] The original feature data (including all features extracted in step 201 and step 202) are projected onto the selected principal components. This process actually converts the original high-dimensional data into a low-dimensional space composed of principal components.

[0058] The projected data is the feature data set after dimensionality reduction. It retains the main information of the original data, but has a lower dimension, which is more convenient for subsequent analysis and processing.

[0059] Example 2: This embodiment describes in detail the steps of constructing a high and low voltage switchgear fault risk prediction model using a deep neural network (DNN) algorithm and the specific structure of the model.

[0060] Step 301: Data division and preparation.

[0061] Use a feature dataset as input, which includes the statistical characteristics of condition monitoring data, cross-features of environmental parameters and basic equipment properties, and principal component analysis results. Divide the feature dataset into a training set and a validation set. The training set is used for model training, and the validation set is used for model selection to ensure that the model performs well on unseen data.

[0062] For example, 70% of the data can be divided into a training set and 30% of the data can be divided into a validation set.

[0063] Step 302: Construct a DNN structure.

[0064] Construct a deep neural network DNN structure, set up an input layer, multiple hidden layers, and an output layer. The input layer receives the feature data set as the input of the model. The hidden layer contains three fully connected hidden layers, each with 128, 64, and 32 neurons. Select ReLU (Rectified Linear Unit) as the activation function to increase the nonlinear expression ability of the model. The output layer receives the output of the last hidden layer and outputs the fault risk probability through a Sigmoid activation function. The Sigmoid function can map the output value to between 0 and 1, indicating the probability of a fault.

[0065] Step 303: Model training.

[0066] The DNN model is trained using the training set. During the training process, the model parameters are optimized using stochastic gradient descent (SGD) or its variants (such as Adam). The output of each layer is calculated through forward propagation, and the model parameters are updated through back propagation.

[0067] The cross entropy loss function is used to evaluate the difference between the model's predicted probability and the actual label. The cross entropy loss function is defined as:

[0068] in, is the sample size, is the true label of the sample, fault is 1, no fault is 0, is the failure risk probability predicted by the model.

[0069] In each iteration, a batch of feature datasets is fed into the DNN model, the loss is calculated and the parameters are updated until a preset number of iterations is reached or the loss converges.

[0070] Step 304: Model selection.

[0071] Use the validation set to select the trained DNN model. Evaluate the performance of the model on the validation set based on evaluation indicators such as accuracy, recall, and F1 score. Accuracy indicates the proportion of correct predictions made by the model; recall indicates the proportion of faulty samples that the model can correctly identify; and the F1 score is the harmonic average of accuracy and recall, which is used to comprehensively evaluate model performance.

[0072] Step 305: Model tuning.

[0073] According to the evaluation results, adjust the hyperparameters of the DNN model, including the number of hidden layers, the number of neurons in each layer, the learning rate, etc. Find the optimal hyperparameter combination through grid search, random search, or Bayesian optimization. Repeat steps 303 and 304 until the model performance reaches a satisfactory level.

[0074] Step 306: Save the model.

[0075] Save the trained fault risk prediction model so that it can be loaded and used in actual applications later. You can save it in common model saving formats (such as TensorFlow's SavedModel, PyTorch's .pt file, etc.).

[0076] Example 3: This example describes in detail the implementation steps of risk point marking and fault risk assessment, aiming to timely identify and assess the fault risk of high and low voltage switchgear through comparative analysis of real-time monitoring data and prediction data.

[0077] Steps to implement risk point marking: Step 801: Align the timestamps of real-time monitoring data and prediction data.

[0078] Real-time monitoring data comes from the device's sensors, recording the device's current operating status; predicted data is generated by the fault risk prediction model, reflecting the probability of equipment failure risk in the future. In order to accurately compare these two types of data, their timestamps need to be aligned first to ensure that the real-time monitoring data and predicted data at each time point correspond.

[0079] Step 802: Set a failure risk threshold.

[0080] Based on historical fault data and equipment type, a fault risk threshold is preset to determine whether the difference between real-time monitoring data and predicted data has reached a level that requires attention.

[0081] For example, for a certain type of switchgear, historical data shows that when the difference value of a certain monitoring indicator exceeds a certain value, the probability of equipment failure increases significantly, so this value is set as the failure risk threshold.

[0082] Step 803: Calculate the difference between the real-time monitoring data and the predicted data.

[0083] For each time point, the difference between the real-time monitoring data and the predicted data is calculated. This difference represents the deviation of the corresponding relationship between the actual state monitoring indicator and the predicted risk probability. The difference can be calculated in absolute value, relative rate of change, etc., depending on the nature of the monitoring indicator and the output of the prediction model.

[0084] Step 804: Mark risk points.

[0085] Compare the calculated difference with the preset threshold. If the difference exceeds the threshold, mark the time point as a risk point. Record the timestamp, difference value and corresponding monitoring indicator of the risk point for subsequent analysis.

[0086] Failure risk assessment steps: Step 901: Count risk points.

[0087] Count the risk points marked in step 8 and record the timestamp, monitoring index and difference value of each risk point. This information will serve as the basis for fault risk assessment.

[0088] Step 902: Analyze risk points.

[0089] Analyze risk points according to the preset assessment strategy. Consider factors such as the number of risk points, the size of the difference value, and the duration. For example, when the number of risk points exceeds 5 within 24 consecutive hours, and the difference value of at least 3 risk points is greater than 2 times the preset threshold, the device is judged to be in a high failure risk state. When the risk point lasts for more than 48 hours and the difference value continues to increase, it is judged that the device may have a serious failure risk and needs to be inspected immediately.

[0090] Step 903: Generate a fault risk assessment report.

[0091] Generate a failure risk assessment report based on the risk point analysis results. The report includes an overview of equipment failure risks, detailed information on risk points (such as timestamps, monitoring indicators, and difference values), and recommended maintenance measures. The report should be clear and accurate so that relevant personnel can quickly understand the equipment's failure risk status.

[0092] Step 904: Send a fault risk assessment report.

[0093] The generated fault risk assessment report is sent to relevant personnel, such as equipment maintenance personnel, management personnel, etc. The report can be sent by email, SMS, system notification, etc. to ensure that relevant personnel can receive it in time and take corresponding measures.

[0094] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0095] 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 fault risk perception and early warning method for high and low voltage complete switchgear, characterized in that: The method comprises: Step 1: Collect the status monitoring data, environmental parameters and basic attribute information of high and low voltage complete switchgear in historical operation; the status monitoring data includes voltage, current, power factor, temperature, and partial discharge level; the environmental parameters include ambient temperature, humidity, and altitude; the basic attribute information of the equipment includes equipment model, manufacturer, commissioning date, and design life; Step 2: Preprocess and extract features of the data collected in step 1 to construct a feature data set; Step 3: Build a fault risk prediction model for high and low voltage switchgear. This model is based on the feature data set, combined with environmental parameters and basic equipment properties, to predict the probability of equipment failure risk in the future. Step 4: Obtain the current operating status data and environmental parameters of the high and low voltage switchgear; Step 5: Input the data obtained in step 4 into the fault risk prediction model, and receive the predicted fault risk probability output by the model; Step 6: Monitor the high and low voltage switchgear in real time, obtain the actual status monitoring data, and clean the original monitoring data to remove abnormal values; Step 7: Compare the real-time monitoring data obtained in step 6 with the predicted data output by the model in step 5, and align the timestamps; Step 8: Determine whether the difference between the real-time monitoring data and the predicted data exceeds the threshold according to the preset risk threshold. If so, mark it as a risk point; Step 9: Based on the discovered risk points and the preset assessment strategy, assess the failure risk of the high and low voltage switchgear and send the assessment results to relevant personnel.

2. A fault risk perception and early warning method for high and low voltage complete switchgear according to claim 1, characterized in that: The feature extraction method comprises: Step 201: Calculate the statistical characteristics of the condition monitoring data, including mean, variance, maximum value, minimum value and rate of change; Step 202: Combine the selected environmental parameters with the basic attribute information of the equipment to extract cross features; Step 203: Standardize all features obtained in step 201 and step 202 so that each feature has zero mean and unit variance, construct a covariance matrix of feature data, calculate eigenvalues ​​and eigenvectors of the covariance matrix, sort the eigenvalues ​​from large to small, and select the eigenvectors corresponding to the first k eigenvalues ​​with cumulative contribution rates exceeding 90% as principal components; Step 204: Project the original feature data onto the selected principal component to obtain a feature data set after dimensionality reduction.

3. A fault risk perception and early warning method for high and low voltage complete switchgear according to claim 1, characterized in that: In step 3, a deep neural network (DNN) algorithm is used to build a fault risk prediction model.

4. A fault risk perception and early warning method for high and low voltage complete switchgear according to claim 3, characterized in that: The steps to build a failure risk prediction model include: Step 301: Use the feature data set as input and divide it into a training set and a validation set. The training set is used for model training, and the validation set is used for model selection. Step 302: construct a deep neural network DNN structure, set an input layer, multiple hidden layers and an output layer, and select a suitable activation function; Step 303: Train the DNN model using the training set, optimize the model parameters by stochastic gradient descent method, and minimize the cross entropy loss between the predicted fault risk probability and the actual fault situation; Step 304: Use the validation set to select the trained DNN model and evaluate the model performance based on accuracy, recall, and F1 score; Step 305: According to the evaluation results, adjust the hyperparameters of the DNN model, including the number of hidden layers, the number of neurons in each layer, and the learning rate, until the model performance is satisfactory; Step 306: Save the trained fault risk prediction model.

5. A fault risk perception and early warning method for high and low voltage complete switchgear according to claim 4, characterized in that: The network structure of the fault risk prediction model is: Input layer: receives feature data sets, including statistical features of condition monitoring data, cross-features of environmental parameters and basic equipment attributes, and principal component analysis results; Hidden layer: contains three fully connected hidden layers, each with 128, 64, and 32 neurons, using the ReLU activation function; Output layer: receives the output of the last hidden layer and outputs the failure risk probability through a Sigmoid activation function.

6. A fault risk perception and early warning method for high and low voltage complete switchgear according to claim 5, characterized in that: The DNN model is iteratively trained using the training set data. In each iteration, a batch of feature data sets are fed into the DNN model, the output of each layer is calculated by forward propagation, and the model parameters are updated by back propagation. The cross entropy loss function is used to evaluate the difference between the model's predicted probability and the actual label. The cross entropy loss function is defined as: ； in, is the sample size, is the true label of the sample, fault is 1, no fault is 0, is the failure risk probability predicted by the model.

7. A method for sensing and warning fault risks of high and low voltage complete switchgear according to claim 1, characterized in that: In step 8, the steps of implementing risk point marking include: Step 801: aligning the timestamps of real-time monitoring data and prediction data; Step 802: Setting a fault risk threshold, which is preset based on historical fault data and equipment type; Step 803: For each time point, the difference between the real-time monitoring data and the predicted data is calculated, which represents the deviation of the corresponding relationship between the actual state monitoring index and the predicted risk probability; Step 804: Compare the calculated difference with the preset threshold value, and if it exceeds the threshold value, mark it as a risk point, and record the timestamp, difference value and corresponding monitoring indicator.

8. A method for sensing and warning fault risks of high and low voltage complete switchgear according to claim 7, characterized in that: In step 9, the steps of failure risk assessment include: Step 901: Count the risk points marked in step 8, and record the timestamp, monitoring index, and difference value; Step 902: Analyze risk points according to the preset assessment strategy, taking into account the number of risk points, the size of the difference value, and the duration; Step 903: Generate a failure risk assessment report, including an overview of equipment failure risks, detailed information on risk points, and recommended maintenance measures; Step 904: Send the fault risk assessment report to relevant personnel.

9. A method for sensing and warning fault risks of high and low voltage complete switchgear according to claim 8, characterized in that: The preset evaluation strategy includes: When the number of risk points exceeds 5 within 24 consecutive hours, and the difference values ​​of at least 3 risk points are greater than 2 times the preset threshold, the device is judged to be in a high failure risk state; When the risk point lasts for more than 48 hours and the difference value continues to increase, it is determined that the equipment may have serious hidden dangers of failure and needs to be inspected immediately.

10. A method for sensing and warning fault risks of high and low voltage complete switchgear according to claim 1, characterized in that: In step 2, the data preprocessing method includes: For missing values, the average of the previous and next values ​​was used to fill them; For outliers, the 3σ principle is used, and data points beyond the range of three times the standard deviation of the mean are regarded as outliers and replaced with the median; Using the Min-Max normalization method, each eigenvalue is linearly transformed to the [0,1] interval. The normalization formula is: ； in, is the original data, is the minimum value of the data, is the maximum value of the data, is the standardized data.

Images