An intelligent monitoring method and system for power equipment

Abstract

The present invention discloses an intelligent monitoring method and system for power equipment, relating to the field of power equipment monitoring. An intelligent monitoring system for power equipment includes: a digital twin module, a data acquisition module, a data processing module, an abnormal monitoring module, a fault detection module, and an intelligent alarm module. The present invention reduces the amount of data to be processed by performing preliminary abnormal monitoring and independent real-time abnormal monitoring on the data of power equipment, predicts faults based on the results, obtains the power equipment that needs to be re-inspected according to the fault prediction results, performs associated abnormal monitoring on the re-inspected power equipment, and finally performs intelligent alarm according to the obtained independent abnormal monitoring results, fault prediction results, and associated abnormal monitoring results, improving the monitoring performance of power equipment.

Description

Technical Field

[0001] The present invention relates to the field of power equipment monitoring, and particularly to an intelligent monitoring method and system for power equipment. Background Art

[0002] With the increasing dependence of modern society on electricity, the stability and reliability of the power system have become particularly important. Power equipment, such as transformers, circuit breakers, transmission lines, etc., is the core component of the power system, and its normal operation is directly related to the safety and power supply quality of the entire power system. However, as the operation time of the equipment extends, power equipment may be affected by various factors and fail, such as insulation aging, mechanical wear, overload, etc. Traditional power equipment monitoring methods mainly rely on regular manual inspections and maintenance, which have problems such as low efficiency, poor timeliness, and insufficient accuracy.

[0003] To solve these problems, in recent years, the power industry has gradually introduced intelligent monitoring technologies. By real-time monitoring the operation status of power equipment, intelligent evaluation of equipment health status and fault prediction are realized. Such an intelligent monitoring system can provide more timely and accurate equipment status information and give early warnings before faults occur, thus effectively reducing power outages and maintenance costs caused by equipment failures. The current intelligent monitoring systems have limited data processing capabilities. With the increase in monitoring equipment and the amount of collected data, traditional data processing technologies are difficult to meet the processing requirements of large-scale and real-time data. The prediction models are not perfect enough. Existing fault prediction models often rely on simple rules or single algorithms and are difficult to comprehensively capture the complexity of equipment failures, resulting in low prediction accuracy. The system integration degree is not high. Many intelligent monitoring systems are designed for specific equipment or specific faults and lack the comprehensive monitoring ability for the entire power system, resulting in the inability to share data between systems and unsatisfactory monitoring effects. Summary of the Invention

[0004] The present invention provides an intelligent monitoring method and system for power equipment with high integration degree, strong data processing ability, and high prediction accuracy to meet the higher requirements of modern power systems for the safe, reliable, and economic operation of equipment.

[0005] An intelligent monitoring method for power equipment includes:

[0006] S1. Obtain the equipment information and operation status information of the power equipment, perform a set simulation on the power equipment through the equipment information to obtain the association information between the power equipment, and use the operation status information of the power equipment for virtual-real mapping to construct a digital twin model of the operation status of the power equipment;

[0007] S2. Collect the real-time operation data of the power equipment through high-precision sensors at a set period;

[0008] S3. Perform preliminary anomaly analysis and processing on the real-time operation data to obtain the data to be analyzed, perform noise reduction processing on the data to be analyzed, and obtain the noise-reduced data;

[0009] S4. Send the noise-reduced data of each power device into the real-time anomaly monitoring model for independent anomaly monitoring to obtain the real-time independent anomaly monitoring results of a single power device;

[0010] S5. Take the power devices with abnormal preliminary anomaly analysis and processing results and real-time independent anomaly monitoring results as the power devices with faults to be predicted. According to the association information of the power devices with faults to be predicted, obtain the power devices associated with the power devices with faults to be predicted and the specific association relationships, and obtain the data to be analyzed of the power devices with faults to be predicted and the power devices associated with the power devices with faults to be predicted in the current cycle and the previous 9 cycles;

[0011] Generate a sample to be predicted. Take the power device with a fault to be predicted as the main object, the data to be analyzed in its corresponding 10 cycles as the main object data, take each power device associated with the main object as an associated object, the data to be analyzed in the corresponding 10 cycles of the associated object as the associated object data, and add relationship weights to the associated object data according to the types of association relationships;

[0012] Send the sample to be predicted into the fault prediction model for fault prediction to obtain the fault prediction results. Take the power devices corresponding to the abnormal fault prediction results as the retest devices, obtain the noise-reduced data of the retest devices and the power devices associated with them, which is called the retest data, and send the retest data into the anomaly monitoring model for associated anomaly monitoring to obtain the associated anomaly monitoring results of the retest devices;

[0013] S6. Record the independent anomaly monitoring results, fault prediction results, and associated anomaly monitoring results of the power devices in the intelligent monitoring log, and perform intelligent alarm according to the recorded situation.

[0014] As a preferred technical solution of the present invention, the step of performing preliminary anomaly analysis and processing on the real-time operation data to obtain the data to be analyzed specifically includes:

[0015] Cache the operation data collected within a period of time, calculate the standard deviation of each item of the real-time operation data and the operation data in the cache, take the data exceeding the set threshold as abnormal data, detect the missing values in the real-time operation data, perform moving smoothing on the abnormal data and the missing values, and record the abnormal data and the missing values in the review log.

[0016] As a preferred technical solution of the present invention, the step of performing noise reduction processing on the data to be analyzed to obtain the noise-reduced data specifically includes:

[0017] The data to be analyzed of the power equipment is represented by X, and it is divided and processed according to the data type to obtain N modal components, and the modal components are represented by Z N ={z 1 ,z 2 ,···,z n}, calculate the energy density Q n and the average period T n , and the calculation formula is: , where M 1 represents the acquisition density of the equipment operation data; V 2 represents the sampling volume; K Y represents the negative amplitude coefficient, t 0 represents the time required for data noise reduction processing;

[0018] Calculate the variational mode component noise threshold G h , G h =Q n *T n , detect whether the variational mode component noise threshold G h has a mutation. If there is a mutation, the previous modal component is noise, and all the detected noises are deleted to obtain the noise reduction processed data X ’ .

[0019] As a preferred technical solution of the present invention, the real-time anomaly monitoring model includes a mode selection layer and an anomaly monitoring layer;

[0020] The mode selection layer is used to select a model according to the input data type. If the input data is the noise reduction processed data of a single power equipment, it enters the independent anomaly monitoring mode. If the input data is the noise reduction processed data of multiple power equipment, it enters the associated anomaly monitoring mode;

[0021] The anomaly monitoring layer is used to extract the fault characteristic parameters in the data information, describe the fault characteristic parameters through a multi-dimensional feature vector, and output the anomaly monitoring results of the power equipment.

[0022] As a preferred technical solution of the present invention, it further includes constructing a real-time anomaly monitoring model, and the specific steps are as follows:

[0023] A1. Obtain the real-time operation data of several power devices. The obtained data includes the real-time operation data of a single power device and the real-time operation data of the power devices associated with it. Perform preprocessing operations on the real-time operation data and mark the abnormal monitoring labels. Compose the real-time operation data of each power device into an independent abnormal monitoring data set, and compose the real-time operation data of the power device and the power devices associated with it into an associated abnormal monitoring data set. Split the independent abnormal monitoring data set and the associated abnormal monitoring data set to obtain an independent abnormal test set, an associated abnormal test set, an independent abnormal training set, and an associated abnormal training set;

[0024] A2. Send the independent abnormal training set and the associated abnormal training set into the initial real-time abnormal monitoring model based on the BP neural network in batches for training, with the abnormal monitoring label as the target, and train to obtain an optimized real-time abnormal monitoring model;

[0025] A3. Send the independent abnormal test set and the associated abnormal test set into the optimized real-time abnormal monitoring model respectively for accuracy evaluation, with the abnormal monitoring label as the target, adjust the hyperparameters of the model, and obtain a real-time abnormal monitoring model with the accuracy evaluation within the set range.

[0026] As a preferred technical solution of the present invention, it further includes a fault prediction model. The fault prediction model is trained based on a recurrent neural network, and the goal of the model is to predict the fault prediction result of the sample, and is trained by the time-series device operation data of several power devices and the power devices associated with them.

[0027] As a preferred technical solution of the present invention, the steps of constructing a digital twin model of the operation state of a power device specifically include:

[0028] B1. Construct an entity model for each power device, construct a three-dimensional space to store the model object, perform a set simulation on the entity model, clarify the specifications and equipment parameter data of the power device, and add them as attribute values to the entity model;

[0029] B2. Obtain the association information between different power devices, establish a data interface and build an entity model association to construct a digital model of the operation of the power device;

[0030] B3. Use the operation state information of the power device for virtual-real mapping, collect the data set uploaded to the cloud platform, including the operation state, working conditions, and data monitored by sensors of the power device, and use a data-driven model to realize the mapping of the digital virtual body and construct a digital twin model of the operation state of the power device.

[0031] As a preferred technical solution of the present invention, the real-time operation data includes the voltage, current, temperature, vibration, and humidity parameters of the power device.

[0032] As a preferred technical solution of the present invention, intelligent alarm is performed according to the recording situation, and the specific steps are as follows:

[0033] C1. Obtain the independent abnormal monitoring results, fault prediction results, and associated abnormal monitoring results in the intelligent monitoring log;

[0034] C2. If there is one or more abnormalities among the three results, but not all are abnormal, a yellow-level alarm is issued, the yellow warning light is turned on, and the yellow alarm result is recorded;

[0035] C3. If all three results are abnormal, a red-level alarm is issued, a sound alarm is given, and the red warning light is turned on;

[0036] C4. Associate the true result of the power equipment being abnormal with the record, calculate the weight according to the weight value of the record situation of the true result of the power equipment being abnormal, and issue a red-level alarm for the record situation whose weight exceeds the set threshold.

[0037] An intelligent monitoring system for power equipment includes:

[0038] A digital twin module, which is used to obtain the device information and operating status information of the power equipment, perform a set simulation of the power equipment through the device information to obtain the associated information between the power equipment, and use the power equipment operating status information for virtual-real mapping to construct a digital twin model of the power equipment operating status;

[0039] A data acquisition module, which is used to collect the real-time operating data of the power equipment at a set period through high-precision sensors;

[0040] A data processing module, which performs preliminary abnormal analysis and processing on the real-time operating data to obtain the data to be analyzed, and performs noise reduction processing on the data to be analyzed to obtain the noise-reduced processed data;

[0041] An abnormal monitoring module, which is used to send the noise-reduced processed data of each power equipment into the real-time abnormal monitoring model for independent abnormal monitoring to obtain the real-time independent abnormal monitoring results of a single power equipment; send the re-inspection data into the abnormal monitoring model for associated abnormal monitoring to obtain the associated abnormal monitoring results of the re-inspection equipment;

[0042] A fault prediction module, which is used to regard the power equipment whose preliminary abnormal analysis and processing results and real-time independent abnormal monitoring results are not normal as the equipment to be predicted for faults, obtain the power equipment associated with the equipment to be predicted for faults and the specific association relationship according to the associated information of the equipment to be predicted for faults, and obtain the data to be analyzed of the equipment to be predicted for faults and the power equipment associated with the equipment to be predicted for faults in the current period and the previous 9 periods;

[0043] Generate a sample to be predicted. Take the faulty device to be predicted as the main object, and the data to be analyzed for 10 cycles corresponding to it as the main object data. Take each power device having an associated relationship with the main object as an associated object, and the data to be analyzed for 10 cycles corresponding to the associated object as the associated object data, and add a relationship weight to the associated object data according to the type of the associated relationship.

[0044] Send the sample to be predicted into the fault prediction model for fault prediction, obtain the fault prediction result, and take the power device corresponding to the fault prediction result that is not normal as the device to be re-inspected. Obtain the data to be analyzed of the device to be re-inspected and the power devices having an associated relationship with it, which is called the re-inspection data.

[0045] The intelligent alarm module is used to record the independent abnormal monitoring result, fault prediction result and associated abnormal monitoring result of the power device in the intelligent monitoring log and perform intelligent alarm according to the recorded situation.

[0046] The present invention has the following advantages:

[0047] 1. By performing preliminary abnormal monitoring and independent real-time abnormal monitoring on the data of power devices, the present invention reduces the amount of data to be processed, performs fault prediction according to the results, obtains the power devices that need to be re-inspected according to the fault prediction results, performs associated abnormal monitoring on the re-inspected power devices, and finally performs intelligent alarm according to the obtained independent abnormal monitoring results, fault prediction results and associated abnormal monitoring results, improving the monitoring performance of power devices.

[0048] 2. By constructing a digital twin model of power devices and performing associated analysis on power devices according to the associated relationship of the model, the present invention realizes the joint analysis of multiple power devices and improves the integration performance of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic structural diagram of an intelligent monitoring system for power devices adopted in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0050] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention.

[0051] Embodiment 1, an intelligent monitoring method for power devices, includes:

[0052] S1. Obtain the device information and operating status information of power devices, perform set simulation on power devices through the device information to obtain the associated information between power devices, and perform virtual-real mapping using the operating status information of power devices to construct a digital twin model of the operating status of power devices.

[0053] The steps of constructing the digital twin model of the operation state of power equipment in step S1 specifically include:

[0054] B1. Construct an entity model for each power equipment, construct a three-dimensional space to store the model object, perform a set simulation on the entity model, clarify the specifications and equipment parameter data of the power equipment, and add them as attribute values to the entity model;

[0055] B2. Obtain the association information between different power equipment, establish a data interface and build an association of entity models to construct a digital model of the operation of power equipment;

[0056] B3. Use the operation state information of power equipment for virtual-real mapping, collect the data set uploaded to the cloud platform, including the operation state, working conditions, and data monitored by sensors of power equipment, use the data-driven model to realize the mapping of the digital virtual body, and construct the digital twin model of the operation state of power equipment.

[0057] S2. Collect the real-time operation data of power equipment at a set period through high-precision sensors;

[0058] The real-time operation data includes the voltage, current, temperature, vibration, and humidity parameters of power equipment. The data is obtained through the sensors of the operating equipment. The data of the sensors is first transmitted to the local computer and then uploaded to the data center for processing through the network. The data center can be a large computer room or a cloud platform.

[0059] S3. Perform preliminary abnormal analysis and processing on the real-time operation data to obtain the data to be analyzed, and perform noise reduction processing on the data to be analyzed to obtain the noise-reduced processed data;

[0060] The steps of performing preliminary abnormal analysis and processing on the real-time operation data to obtain the data to be analyzed specifically include:

[0061] Cache the operation data collected within a period of time, calculate the standard deviation of each item of the real-time operation data and the operation data in the cache, regard the data exceeding the set threshold as abnormal data, detect the missing values in the real-time operation data, perform moving smoothing processing on the abnormal data and missing values, and record the abnormal data and missing values in the review log.

[0062] The moving smoothing processing method adopted is 5-point smoothing processing, that is, use the average value of a total of 4 data, namely 2 data before and after the processing point, to replace the data at the processing point;

[0063] The preliminary abnormal analysis and processing removes the abnormal values and missing values in the data, reducing the abnormal interference received during abnormal processing;

[0064] The steps of performing noise reduction processing on the data to be analyzed to obtain the noise-reduced processed data specifically include:

[0065] Let the data to be analyzed of the power equipment be represented by X, and it is divided according to the data type to obtain N modal components, and the modal components are represented by Z N ={z 1 ,z 2 ,···,z n}}, calculate the energy density Q n and the average period T n , and the calculation formula is: , where M 1 represents the acquisition density of the equipment operation data; V 2 represents the sampling volume; K Y represents the negative amplitude coefficient, t 0 represents the time required for data noise reduction processing;

[0066] Calculate the variational mode component noise threshold G h , G h =Q n *T n , detect whether the variational mode component noise threshold G h has a mutation. If there is a mutation, the previous modal component is noise, and all the detected noise is deleted to obtain the noise reduction processed data X ’ ;

[0067] Further removing abnormal data through noise reduction processing enables the model to perform more targeted analysis on the data change situation;

[0068] S4. Send the noise reduction processed data of each power equipment into the real-time anomaly monitoring model for independent anomaly monitoring to obtain the real-time independent anomaly monitoring results of a single power equipment;

[0069] The real-time anomaly monitoring model includes a mode selection layer and an anomaly monitoring layer;

[0070] The mode selection layer is used to select the model according to the input data type. If the input data is the noise reduction processed data of a single power equipment, it enters the independent anomaly monitoring mode. If the input data is the noise reduction processed data of multiple power equipment, it enters the associated anomaly monitoring mode;

[0071] The anomaly monitoring layer is used to extract the fault characteristic parameters in the data information, describe the fault characteristic parameters through a multi-dimensional feature vector, and output the anomaly monitoring results of the power equipment.

[0072] Finally, output the real-time anomaly monitoring results of the power equipment, and its output calculation formula is:

[0073]

[0074] wherein represents the output abnormal result, is the mapping function representing the i-th item of data, represents the output feature vector;

[0075] Build a real-time anomaly monitoring model, and the specific steps are as follows:

[0076] A1. Obtain the real-time operation data of several power equipment. The obtained data includes the real-time operation data of a single power equipment and the real-time operation data of the power equipment associated with it. Perform preprocessing operations on the real-time operation data and mark the anomaly monitoring labels. Form independent anomaly monitoring data sets for the real-time operation data of each power equipment, and form associated anomaly monitoring data sets for the real-time operation data of the power equipment and the power equipment associated with it. Split the independent anomaly monitoring data sets and the associated anomaly monitoring data sets to obtain independent anomaly test sets, associated anomaly test sets, independent anomaly training sets, and associated anomaly training sets;

[0077] The sample objects in the associated anomaly monitoring data set include the main object and the associated object, which are the current power equipment and the power equipment associated with it respectively. The associated object has an associated weight value, which is obtained through the type of association relationship;

[0078] A2. Send the independent anomaly training set and the associated anomaly training set into the initial real-time anomaly monitoring model based on the BP neural network in batches for training. Using the anomaly monitoring label as the target, train to obtain an optimized real-time anomaly monitoring model;

[0079] The initial real-time anomaly monitoring model includes 1 integrated input layer, 8 hidden layers, and 1 integrated output layer. The integrated input layer is used to receive input data. The first 3 layers of the 8 hidden layers are used to analyze the feature relationship between the real-time operation data of the current power equipment, and the last 5 layers are used to analyze the feature relationship between the real-time operation data of the current power equipment and the power equipment associated with it. When the mode is the independent anomaly monitoring mode, only the first 3 layers of the hidden layer are open. When the mode is the associated anomaly monitoring mode, the data is analyzed through all hidden layers. The integrated output layer is used to output the real-time anomaly monitoring result;

[0080] A3. Send the independent anomaly test set and the associated anomaly test set into the optimized real-time anomaly monitoring model respectively for accuracy evaluation. Using the anomaly monitoring label as the target, adjust the hyperparameters of the model to obtain a real-time anomaly monitoring model with the accuracy evaluation within the set range.

[0081] S5. Take the power equipment whose preliminary abnormal analysis processing results and real-time independent abnormal monitoring results are not normal as the equipment to be predicted for faults. Obtain the power equipment associated with the equipment to be predicted for faults and the specific association relationships according to the association information of the equipment to be predicted for faults. Obtain the data to be analyzed for the equipment to be predicted for faults and the power equipment associated with the equipment to be predicted for faults in the current cycle and the previous 9 cycles.

[0082] Generate a sample to be predicted. Take the equipment to be predicted for faults as the main object, the data to be analyzed for 10 cycles corresponding to it as the main object data. Take each power equipment having an association relationship with the main object as an associated object, the data to be analyzed for 10 cycles corresponding to the associated object as the associated object data, and add a relationship weight to the associated object data according to the type of the association relationship.

[0083] The specific relationship weight also needs to be calculated through the distance between the equipment and the type of the association relationship. The specific formula is:

[0084]

[0085] In the formula, d is the distance between the equipment, D is the standard value of the distance, and α j is the type weight of the association relationship.

[0086] Send the sample to be predicted into the fault prediction model for fault prediction to obtain the fault prediction result. Obtain the power equipment corresponding to the fault prediction result that is not normal as the equipment for re-inspection. Obtain the noise reduction processing data of the equipment for re-inspection and the power equipment having an association relationship with it, which is called the re-inspection data. Send the re-inspection data into the abnormal monitoring model for associated abnormal monitoring to obtain the associated abnormal monitoring result of the equipment for re-inspection.

[0087] The fault prediction model is trained based on the recurrent neural network. The goal of the model is the fault prediction result of the prediction sample, and it is trained through the time-series equipment operation data of several power equipment and the power equipment having an association relationship with it.

[0088] The fault prediction model adopted by this method includes 1 integrated input layer, 4 hidden layers and 1 integrated output layer. The integrated input layer is used to receive the data of the input prediction sample. The 4 hidden layers are used to analyze the connection between the data of the input prediction sample. The integrated output layer is used to output the fault prediction result.

[0089] In the recurrent neural network, the states of different layers are represented by time steps. Through the propagation of the recurrent connection relationship between different layers in the time steps, the dependence relationship in the data is obtained.

[0090] The hidden layer state is jointly determined by the hidden state of the previous time step and the current input. The specific formula is:

[0091]

[0092] Among them, represents the hidden state at the time step, tanh represents the hyperbolic tangent activation function, represents the weight of the input layer, represents the current input, represents the state quantity of the input layer, represents the weight of the hidden layer, represents the hidden state of the previous time step, represents the state quantity of the hidden layer;

[0093] The output layer state is determined by the current hidden state, and the specific formula is:

[0094]

[0095] Among them, represents the current output, represents the weight of the output layer, represents the current hidden state, is the bias term of the output layer;

[0096] S6. Record the independent anomaly monitoring results, fault prediction results, and associated anomaly monitoring results of the power equipment in the intelligent monitoring log, and perform intelligent alarm according to the recording situation.

[0097] The specific steps for performing intelligent alarm according to the recording situation in step S6 are as follows:

[0098] C1. Obtain the independent anomaly monitoring results, fault prediction results, and associated anomaly monitoring results in the intelligent monitoring log;

[0099] C2. If there is one or more anomalies among the three results, but not all are anomalies, then perform a yellow-level alarm, turn on the yellow warning light, and record the yellow alarm result;

[0100] C3. If all three results are anomalies, then perform a red-level alarm, issue a sound alarm and turn on the red warning light;

[0101] C4. Associate the true result of the power equipment being abnormal with the record, calculate the weight according to the weight value of the record situation of the true result of the power equipment being abnormal, and perform a red-level alarm for the record situation whose weight exceeds the set threshold.

[0102] The weight of the record situation consists of an initial value and an adjustment value, and the weights of the initial value and the adjustment value will change over time to achieve the purpose of dynamically adjusting the intelligent alarm; the specific formula is:

[0103]

[0104] Wherein are the weights of the initial value and the adjusted value respectively, has a relatively large value at the initial stage, which can reduce the deviation caused by a small amount of anomalies, decreases over time, ensuring the accuracy in the long term. U is the number of actual anomaly results, and T is the time;

[0105] Example 2. An intelligent monitoring system for power equipment. Refer to Figure 1 as shown, including:

[0106] A digital twin module, which is used to obtain the device information and operation status information of the power equipment, perform a collective simulation of the power equipment through the device information to obtain the association information between the power equipment, and use the operation status information of the power equipment to perform virtual-real mapping to construct a digital twin model of the operation status of the power equipment;

[0107] A data acquisition module, which is used to collect the real-time operation data of the power equipment at a set period through high-precision sensors;

[0108] A data processing module, which performs preliminary anomaly analysis and processing on the real-time operation data to obtain the data to be analyzed, and performs noise reduction processing on the data to be analyzed to obtain the noise-reduced processed data;

[0109] An anomaly monitoring module, which is used to send the noise-reduced processed data of each power equipment into the real-time anomaly monitoring model for independent anomaly monitoring to obtain the real-time independent anomaly monitoring results of a single power equipment; send the re-inspection data into the anomaly monitoring model for associated anomaly monitoring to obtain the associated anomaly monitoring results of the re-inspected equipment;

[0110] A fault prediction module, which uses the power equipment with abnormal preliminary anomaly analysis and processing results and real-time independent anomaly monitoring results as the equipment to be predicted for faults, obtains the power equipment associated with the equipment to be predicted for faults and the specific association relationship according to the association information of the equipment to be predicted for faults, and obtains the data to be analyzed of the equipment to be predicted for faults and the power equipment associated with the equipment to be predicted for faults in the current period and the previous 9 periods;

[0111] Generate a sample to be predicted, use the equipment to be predicted for faults as the main object, the data to be analyzed in its corresponding 10 periods as the main object data, use each power equipment associated with the main object as the associated object, the data to be analyzed in the corresponding 10 periods of the associated object as the associated object data, and add relationship weights to the associated object data according to the type of association relationship;

[0112] Send the sample to be predicted into the fault prediction model for fault prediction, obtain the fault prediction results, and take the power equipment corresponding to the fault prediction results that are not normal as the re-inspection equipment. Obtain the data to be analyzed of the re-inspection equipment and the power equipment associated with it, which is called re-inspection data;

[0113] The intelligent alarm module is used to record the independent abnormal monitoring results, fault prediction results and associated abnormal monitoring results of power equipment in the intelligent monitoring log and conduct intelligent alarms according to the recorded situation.

[0114] It should be understood that those of ordinary skill in the art can make improvements or transformations according to the above description, and all such improvements and transformations should fall within the protection scope of the appended claims of the present invention. The parts not described in detail in this specification belong to the prior art well known to those skilled in the art.

Claims

1. A method for intelligent monitoring of electric power equipment, characterized in that: include: S1. Obtain the equipment information and operation status information of the power equipment, perform collective simulation of the power equipment through the equipment information, obtain the correlation information between the power equipment, use the power equipment operation status information to perform virtual-real mapping, and build a digital twin model of the power equipment operation status; S2. Collect real-time operation data of power equipment at a set period through high-precision sensors; S3. Perform preliminary abnormal analysis on the real-time operation data to obtain the data to be analyzed, and perform noise reduction on the data to be analyzed to obtain noise reduction data; S4. Send the noise reduction data of each power device into the real-time abnormal monitoring model for independent abnormal monitoring to obtain the real-time independent abnormal monitoring results of a single power device; S5. The power equipment for which the preliminary abnormal analysis processing results and the real-time independent abnormal monitoring results are not normal is regarded as the fault equipment to be predicted, and the power equipment associated with the fault equipment to be predicted and the specific association relationship are obtained according to the association information of the fault equipment to be predicted, and the data to be analyzed of the fault equipment to be predicted and the power equipment associated with the fault equipment to be predicted in the current cycle and the previous 9 cycles are obtained; Generate samples to be predicted, take the fault device to be predicted as the main object, and its corresponding 10 cycles of data to be analyzed as the main object data, take each power device that has an association relationship with the main object as an associated object, and the 10 cycles of data to be analyzed corresponding to the associated object as the associated object data, and add a relationship weight to the associated object data according to the type of the association relationship; The specific relationship weight needs to be calculated based on the distance between devices and the type of association relationship. The specific formula is: Where d is the distance between devices, D is the standard value of the distance, and α j is the type weight of the association relationship; The samples to be predicted are sent to the fault prediction model for fault prediction to obtain fault prediction results, the power equipment corresponding to the abnormal fault prediction results is obtained as the re-inspection equipment, the noise reduction processing data of the re-inspection equipment and the power equipment associated with it is obtained, which is called re-inspection data, and the re-inspection data is sent to the real-time abnormality monitoring model for associated abnormality monitoring to obtain the associated abnormality monitoring results of the re-inspection equipment; S6. Record the independent abnormal monitoring results, fault prediction results and associated abnormal monitoring results of the power equipment in the intelligent monitoring log, and make intelligent alarms according to the records; The real-time anomaly monitoring model includes a mode selection layer and an anomaly monitoring layer; The mode selection layer is used to select a model according to the input data type. If the input data is the noise reduction processing data of a single power device, it enters the independent abnormality monitoring mode. If the input data is the noise reduction processing data of multiple power devices, it enters the associated abnormality monitoring mode. The abnormal monitoring layer is used to extract fault characteristic parameters from data information, describe the fault characteristic parameters through multi-dimensional feature vectors, and output abnormal monitoring results of power equipment; It also includes building a real-time anomaly monitoring model. The specific steps are: A1. Acquire real-time operation data of several power equipment, including the real-time operation data of a single power equipment and the real-time operation data of power equipment associated with it, perform preprocessing operations on the real-time operation data and mark abnormal monitoring labels, form the real-time operation data of each power equipment into an independent abnormal monitoring data set, form the real-time operation data of the power equipment and the power equipment associated with it into an associated abnormal monitoring data set, split the independent abnormal monitoring data set and the associated abnormal monitoring data set to obtain an independent abnormal test set, an associated abnormal test set, an independent abnormal training set and an associated abnormal training set; A2. Send the independent anomaly training set and the associated anomaly training set in batches to the initial real-time anomaly monitoring model based on the BP neural network for training, and train the optimized real-time anomaly monitoring model with the anomaly monitoring label as the target; A3. Send the independent anomaly test set and the associated anomaly test set to the optimized real-time anomaly monitoring model for accuracy evaluation. Taking the anomaly monitoring label as the target, adjust the model's hyperparameters to obtain a real-time anomaly monitoring model with an accuracy evaluation within the set range.

2. The method for intelligent monitoring of electric power equipment according to claim 1, characterized in that: The steps of performing preliminary abnormal analysis on real-time operation data to obtain the data to be analyzed specifically include: The operation data collected over a period of time is cached, the standard deviation of each data item between the real-time operation data and the operation data in the cache is calculated, the data exceeding the set threshold is regarded as abnormal data, the missing values ​​in the real-time operation data are detected, the abnormal data and missing values ​​are processed using moving smoothing, and the abnormal data and missing values ​​are recorded in the review log.

3. The method for intelligent monitoring of electric power equipment according to claim 1, characterized in that: The steps of performing noise reduction processing on the data to be analyzed and obtaining the noise reduction processed data specifically include: The data to be analyzed of the power equipment is represented by X, and it is divided according to the data type to obtain N modal components. The modal components are represented by Z N ={z1,z2,···,z n } indicates that the energy density Q of the nth modal component is calculated. n and the average period T n , the calculation formula is: , where M1 represents the collection density of equipment operation data; V2 represents the sampling volume; K Y represents the negative amplitude coefficient, and t0 represents the time required for data noise reduction processing; Calculate the variational modal component noise threshold G h , G h =Q n *T n , detection variational modal component noise threshold G h Whether a mutation occurs. If a mutation occurs, the previous modal component is noise. All the monitored noise is deleted to obtain the noise reduction processing data X ’ .

4. The method for intelligent monitoring of electric power equipment according to claim 1, characterized in that: It also includes a fault prediction model, which is based on recurrent neural network training. The goal of the model is to predict the fault prediction results of the samples, and is obtained through training with the timing equipment operation data of several power equipment and the power equipment associated with them.

5. The method for intelligent monitoring of electric power equipment according to claim 1, characterized in that: The steps to build a digital twin model of the operating status of power equipment include: B1. Build a physical model for each power equipment, construct a three-dimensional space for storing model objects, perform collective simulation on the physical model, clarify the specifications and equipment parameter data of the power equipment, and add them as attribute values ​​to the physical model; B2. Obtain the correlation information between different power equipment, establish data interface and build physical model association, and construct a digital model of power equipment operation; B3. Use the operating status information of power equipment for virtual-reality mapping, collect data sets uploaded to the cloud platform, including the operating status, working conditions, and sensor monitoring data of power equipment, use data-driven models to achieve mapping of digital virtual bodies, and build a digital twin model of the operating status of power equipment.

6. The method for intelligent monitoring of electric power equipment according to claim 1, characterized in that: Real-time operating data includes voltage, current, temperature, vibration and humidity parameters of power equipment.

7. The method for intelligent monitoring of electric power equipment according to claim 1, characterized in that: Intelligent alarm is carried out according to the recorded situation. The specific steps are as follows: C1. Obtain independent abnormal monitoring results, fault prediction results and associated abnormal monitoring results in the intelligent monitoring log; C2. If one or more of the three results are abnormal, but not all are abnormal, a yellow level alarm will be issued, a yellow warning light will be lit, and the yellow alarm result will be recorded; C3. If all three results are abnormal, a red level alarm will be triggered, with an audible alarm and a red warning light; C4. Associate the actual result of the abnormality of the power equipment with the record, perform weight calculation according to the weight of the record of the actual result of the abnormality of the power equipment, and issue a red level alarm for the record whose weight exceeds the set threshold.

8. An intelligent monitoring system for electric power equipment, characterized in that: The system applies an intelligent monitoring method for electric power equipment according to any one of claims 1 to 7, including: The digital twin module is used to obtain the equipment information and operation status information of the power equipment, perform collective simulation of the power equipment through the equipment information, obtain the correlation information between the power equipment, use the power equipment operation status information to perform virtual-real mapping, and build a digital twin model of the power equipment operation status; A data acquisition module is used to collect real-time operation data of power equipment at a set period through high-precision sensors; The data processing module performs preliminary abnormal analysis on the real-time operation data to obtain the data to be analyzed, and performs noise reduction on the data to be analyzed to obtain the noise reduction processed data; The abnormality monitoring module is used to send the noise reduction processing data of each power device into the real-time abnormality monitoring model for independent abnormality monitoring, and obtain the real-time independent abnormality monitoring result of a single power device; send the re-inspection data into the real-time abnormality monitoring model for associated abnormality monitoring, and obtain the associated abnormality monitoring result of the re-inspected device; A fault prediction module is used to take the power equipment for which the preliminary abnormality analysis processing results and the real-time independent abnormality monitoring results are not normal as the fault equipment to be predicted, obtain the power equipment associated with the fault equipment to be predicted and the specific association relationship according to the association information of the fault equipment to be predicted, and obtain the data to be analyzed of the fault equipment to be predicted and the power equipment associated with the fault equipment to be predicted in the current cycle and the previous 9 cycles; Generate samples to be predicted, take the fault device to be predicted as the main object, and its corresponding 10 cycles of data to be analyzed as the main object data, take each power device that has an association relationship with the main object as an associated object, and the 10 cycles of data to be analyzed corresponding to the associated object as the associated object data, and add a relationship weight to the associated object data according to the type of the association relationship; The samples to be predicted are sent to the fault prediction model for fault prediction to obtain fault prediction results, and the power equipment corresponding to the abnormal fault prediction results is obtained as the re-inspection equipment, and the data to be analyzed of the re-inspection equipment and the power equipment associated with it are obtained, which is called re-inspection data; The intelligent alarm module is used to record the independent abnormal monitoring results, fault prediction results and associated abnormal monitoring results of the power equipment in the intelligent monitoring log, and to make intelligent alarms according to the recorded conditions.

Images