Abnormity detection method and anomaly detection device based on transformer substation equipment portrait

By constructing a deep learning model and combining the correlation coefficient, abnormal detection of primary and secondary equipment in the substation is solved, and the problem of ignoring correlation in the existing technology causes low detection accuracy, achieving more accurate equipment status monitoring.

CN119004330APending Publication Date: 2024-11-22GUANGDONG POWER GRID CO LTD +1
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411123672.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The correlation between primary equipment and secondary equipment in substations is ignored in the prior art, resulting in low accuracy of equipment status detection and cannot meet the standards of intelligent detection.

Method used

Using a deep learning-based method, by constructing the first deep learning model and the second deep learning model, the primary and secondary devices are detected abnormally, and the recurrent neural network and long-term memory network algorithm are used to combine the correlation coefficient between devices. Carry out training to improve detection accuracy.

Benefits of technology

It improves the accuracy of equipment status detection, can accurately determine abnormal conditions of primary and secondary equipment, and enhances the safe and stable operation of the substation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004330A_ABST
    Figure CN119004330A_ABST
Patent Text Reader

Abstract

The invention provides an anomaly detection method and an anomaly detection device based on a transformer substation equipment portrait. The method comprises the following steps: acquiring first equipment data and second equipment data in real time; inputting the first equipment data into a first deep learning model to obtain a first equipment state probability value; inputting the second equipment data into a second deep learning model to obtain a second equipment state probability value; verifying the primary equipment and the secondary equipment according to the first equipment state probability value and the second equipment state probability value so as to judge whether the primary equipment and the secondary equipment are abnormal or not; and under the condition that the primary equipment or the secondary equipment is abnormal, carrying out abnormity repair processing on the primary equipment or the secondary equipment. The problem that in the prior art, relevance of primary equipment and secondary equipment is not considered at the same time, relevance factors between the equipment are ignored, and the equipment state is directly detected, so that the detection precision is low is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of substation anomaly detection, and in particular to an anomaly detection method based on substation equipment profiling, an anomaly detection device, a computer-readable storage medium and a computer program product. Background Art

[0002] In the power system, substations are key nodes for power conversion and transmission, and their safe and stable operation is of vital importance. The primary and secondary equipment in the substation together constitute the core components of the power system. The primary equipment is mainly responsible for the transmission and conversion of power, while the secondary equipment is responsible for monitoring, protecting and controlling the operation of the primary equipment. Therefore, accurate detection of the status of the primary and secondary equipment in the substation is of great significance to ensure the safe and stable operation of the power system. There is a close connection between the equipment portrait and the equipment historical status. The equipment portrait is a comprehensive description and characterization of the equipment characteristics, performance, behavior, etc. It contains the static attributes and dynamic behavior information of the equipment, while the equipment historical status refers to the record of the equipment's operating status, performance, fault occurrence, etc. in the past period of time; At present, the model-based equipment status detection method has become a hot topic of research, which can automatically learn the normal operating status of the equipment and predict the future normal operating status of the equipment. Compared with the simple threshold setting method, it does not require manual adjustment of the threshold, saving manpower.

[0003] However, in the prior art, such as the patent with publication number CN117033913A, there is no targeted detection of the complexity and correlation of primary and secondary devices, which brings challenges to the detection of device status. There are differences in structure and function between primary and secondary devices. At the same time, there is a close correlation between primary and secondary devices, which also affects the detection of device status. If the correlation factor between devices is ignored and the device status is directly detected, the detection accuracy is reduced, the detection error is large, and the standard of intelligent abnormality detection cannot be met. Summary of the invention

[0004] The main purpose of the present application is to provide an anomaly detection method, an anomaly detection device, a computer-readable storage medium and a computer program product based on substation equipment profiling, so as to at least solve the problem in the prior art that the correlation between primary equipment and secondary equipment is not considered at the same time, the correlation factors between the equipment are ignored, and the equipment status is directly detected, resulting in low detection accuracy.

[0005] In order to achieve the above-mentioned purpose, according to one aspect of the present application, there is provided an anomaly detection method based on substation equipment profiling, including: real-time acquisition of first equipment data and second equipment data, the first equipment data being the real-time data of a primary equipment in the substation at the current moment, and the second equipment data being the operating status data of a secondary equipment in the substation at the current moment and the related information between the secondary equipment and the corresponding primary equipment; inputting the first equipment data into a first deep learning model, so as to use the first deep learning model to perform anomaly detection of the primary equipment on the first equipment data, and obtain a first equipment state probability value, the first deep learning model being obtained by training with a recurrent neural network algorithm using multiple groups of first training data, each group of the first training data including first sample equipment data of the primary equipment and a corresponding second sample state probability value of the primary equipment, the first sample equipment data including at least equipment operating parameters, environmental information, equipment normal state information, equipment abnormal state information, interval duration of equipment abnormality and equipment failure record of the primary equipment collected within the same interval, the first equipment state probability value being the probability value of the equipment state of the primary equipment; and the second The device data is input into a second deep learning model so as to use the second deep learning model to perform abnormal detection of the secondary device on the second device data to obtain a second device state probability value. The second deep learning model is obtained by training multiple groups of second training data through a long short-term memory network algorithm. Each group of the second training data includes second sample device data of the secondary device and a corresponding first sample state probability value. The second sample device data at least includes a sample target label, a sample target correlation coefficient and sample operation data of the second device. The target label is the normal state information and abnormal state information of the device obtained by the primary device. The sample target correlation coefficient is used for the degree of association between the primary device and the corresponding secondary device. The second device state probability value is the probability value of the device state of the secondary device. The primary device and the secondary device are verified according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device have an abnormal situation. When the abnormal situation occurs in the primary device or the secondary device, the primary device or the secondary device is repaired abnormally.

[0006] Optionally, before acquiring the first device data and the second device data in real time, the method further includes: acquiring historical data of the substation, the historical data including at least device information data, device operation data, event record data and maintenance management data, the device information data including at least the device name, device type, device rated voltage, current, frequency, power and control parameters of each device in the substation, the device operation data including at least various parameter data of each device in the power transmission process and the discharge process and the device status information of the device at each time point in the historical time period, the event record data including at least the accident type, accident time and various parameter data when the accident occurs when the equipment fails or is abnormal, the maintenance management data including record data of regular equipment inspection, maintenance and calibration, historical record data of equipment failure or abnormality, cause analysis and processing results and parts information of equipment repair and replacement; filtering out status-related data from the historical data, the status-related data being data related to the status of the primary equipment and the status of the secondary equipment; performing preprocessing operations on the status-related data to obtain a training set, the training set including the first training data and the second training data, the preprocessing operations including at least data cleaning.

[0007] Optionally, before inputting the second device data into the second deep learning model to perform abnormal detection of the secondary device on the second device data using the second deep learning model to obtain the second device state probability value of the secondary device, the method further includes: calculating the correlation coefficient of each parameter pair in the primary device and the secondary device according to a first formula, wherein the first formula is r i is the correlation coefficient of the i-th parameter pair, n is the total number of parameter pairs, i=(1,2,3,...,n), x i is the observed value of the primary device in the ith parameter pair, y i is the observed value of the secondary device in the ith parameter pair, the observed value is a data value randomly selected from the historical data, x is the average value of the observed values ​​of the primary device in all the parameter pairs, is the average value of the observed values ​​of the secondary device in all the parameter pairs, the parameter pairs representing the combination of the observed values ​​corresponding to the primary device and the secondary device; according to the correlation coefficient of each parameter pair, each parameter pair is divided into corresponding correlation parameters, and the correlation parameters are divided into strong correlation parameters, weak correlation parameters and uncorrelated parameters; corresponding weights are set for the strong correlation parameters, the weak correlation parameters and the uncorrelated parameters; the sample target correlation coefficient between the primary device and the secondary device is calculated according to the second formula, and the second formula is R is the sample target correlation coefficient, ω i is the weight of the associated parameter to which the i-th parameter pair belongs.

[0008] Optionally, each parameter pair is divided into corresponding associated parameters according to the association coefficient of each parameter pair, including: forming an association matrix according to all the association coefficients of all the parameter pairs in the primary device and the secondary device, the rows of the association matrix represent the observed values ​​of the primary device, the columns of the association matrix represent the observed values ​​of the secondary device, and the elements of the association matrix represent the association coefficients; drawing a heat map according to the association matrix, each cell in the heat map represents the association coefficient of a parameter pair, and the color depth in the heat map represents the strength and direction of the association; the parameter pair whose absolute value of the association coefficient is greater than or equal to a first threshold is determined as the strongly associated parameter; the parameter pair whose absolute value of the association coefficient is greater than or equal to a second threshold and less than the first threshold is determined as the weakly associated parameter; the parameter pair whose absolute value of the association coefficient is less than the second threshold is determined as the unassociated parameter.

[0009] Optionally, the primary device and the secondary device are verified according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device have abnormal conditions, including: when the first device state probability value is greater than or equal to a first state boundary value, determining that the primary device is in a normal state; when the first device state probability value is less than the first state boundary value and greater than or equal to a second state boundary value, issuing a first prompt message, the first prompt message being used to prompt that preventive maintenance measures be taken for the primary device; when the first device state probability value is less than the second state boundary value, determining that the first device is in an abnormal state and the abnormal condition has occurred. When the second device state probability value is greater than or equal to the third state limiting value, the secondary device is determined to be in a normal state; when the second device state probability value is less than the third state limiting value and greater than or equal to the fourth state limiting value, a third prompt message is issued, and the third prompt message is used to prompt to take preventive maintenance measures for the secondary device; when the second device state probability value is less than the fourth state limiting value, the second device is determined to be in an abnormal state and the abnormal situation occurs, and a fourth prompt message is issued, and the fourth prompt message is used to prompt to perform abnormal processing on the secondary device.

[0010] Optionally, before verifying the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device have abnormal conditions, the method further includes: adjusting the first state boundary value and the second state boundary value in real time according to historical data and real-time data corresponding to the primary device; and adjusting the third state boundary value and the fourth state boundary value in real time according to historical data and real-time data corresponding to the secondary device.

[0011] Optionally, when the abnormal situation occurs in the primary device or the secondary device, abnormal repair processing is performed on the primary device or the secondary device, including: when the abnormal situation occurs in the primary device, determining a first abnormal repair strategy according to the abnormal situation, and performing abnormal repair processing on the primary device according to the first abnormal repair strategy; when the abnormal situation occurs in the secondary device, determining a second abnormal repair strategy according to the abnormal situation, and performing abnormal repair processing on the secondary device according to the second abnormal repair strategy.

[0012] According to another aspect of the present application, there is provided an abnormality detection device based on substation equipment portrait, the device comprising: a first acquisition unit, used to acquire first equipment data and second equipment data in real time, the first equipment data being real-time data of a primary equipment in the substation at the current moment, the second equipment data being operating status data of a secondary equipment in the substation at the current moment and related information with the corresponding primary equipment; a first input-output unit, used to input the first equipment data into a first deep learning model, so as to use the first deep learning model to perform abnormality detection of the primary equipment on the first equipment data to obtain a first equipment state probability value, the first deep learning model being obtained by training with a recurrent neural network algorithm using multiple groups of first training data, each group of the first training data comprising first sample equipment data of the primary equipment and a corresponding second sample state probability value of the primary equipment, the first sample equipment data comprising at least equipment operating parameters, environmental information, equipment normal state information, equipment abnormal state information, interval duration of equipment abnormality and equipment failure record of the primary equipment collected within the same interval, the first equipment state probability value being the probability value of the equipment state of the primary equipment; a second input-output unit, used to The second device data is input into a second deep learning model, so as to use the second deep learning model to perform abnormal detection of the secondary device on the second device data to obtain a second device state probability value, wherein the second deep learning model is obtained by training multiple groups of second training data through a long short-term memory network algorithm, each group of the second training data includes second sample device data of the secondary device and a corresponding first sample state probability value, the second sample device data at least includes a sample target label, a sample target correlation coefficient and sample operation data of the second device, the target label is the normal state information and abnormal state information of the device acquired by the primary device, the sample target correlation coefficient is used for the degree of association between the primary device and the corresponding secondary device, and the second device state probability value is the probability value of the device state of the secondary device; a verification unit is used to verify the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device have an abnormal situation; an abnormality processing unit is used to perform abnormal repair processing on the primary device or the secondary device when the abnormal situation occurs in the primary device or the secondary device.

[0013] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute any one of the methods described.

[0014] According to another aspect of the present application, a computer program product is provided, comprising computer instructions, wherein when the computer instructions are executed by a processor, any one of the methods described above is implemented.

[0015] Applying the technical solution of the present application, in an abnormality detection method based on substation equipment profiling, first, the first equipment data and the second equipment data are acquired in real time, the first equipment data being the real-time data of the primary equipment in the substation at the current moment, and the second equipment data being the operating status data of the secondary equipment in the substation at the current moment and the related information with the corresponding primary equipment; the first equipment data is input into the first deep learning model, so as to use the first deep learning model to perform abnormality detection of the primary equipment on the first equipment data, and obtain a first equipment status probability value, the first deep learning model is obtained by training with a recurrent neural network algorithm using multiple groups of first training data, each group of the first training data includes the first sample equipment data of the primary equipment and the corresponding second sample status probability value of the primary equipment, the first sample equipment data at least includes the equipment operating parameters, environmental information, equipment normal status information, equipment abnormal status information, equipment abnormality interval and equipment failure record of the primary equipment collected within the same interval, the first equipment status probability value is the probability value of the equipment status of the primary equipment; the second equipment data is input into In the second deep learning model, the second device data is used to perform abnormal detection of the secondary device by using the second deep learning model to obtain a second device state probability value. The second deep learning model is obtained by training with multiple groups of second training data through a long short-term memory network algorithm. Each group of the second training data includes second sample device data of the secondary device and a corresponding first sample state probability value. The second sample device data at least includes a sample target label, a sample target correlation coefficient and sample operation data of the second device. The target label is the normal state information and abnormal state information of the device obtained by the primary device. The sample target correlation coefficient is used for the degree of association between the primary device and the corresponding secondary device. The second device state probability value is the probability value of the device state of the secondary device. The primary device and the secondary device are verified according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device have abnormal conditions. In the event that the primary device or the secondary device has the abnormal condition, the primary device or the secondary device is repaired abnormally.This application constructs a first deep learning model and a second deep learning model. The first deep learning model predicts the primary device and outputs the probability value of the device state of the primary device. The second deep learning model predicts the device state of the secondary device and outputs the probability value of the device state of the secondary device. The device state is judged according to the probability value. In the training process of the second deep learning model, the parameters related to the secondary device and the device state and the correlation coefficient of the secondary device and the primary device in the device operation state are comprehensively considered as the second training data, so that the second deep learning model trained by using the final correlation coefficient will be more accurate in device state detection and improve the detection accuracy. This application solves the problem that the correlation between the primary device and the secondary device is not considered at the same time in the prior art, and the correlation factors between the devices are ignored to directly detect the device state, resulting in low detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A hardware structure block diagram of a mobile terminal for executing an abnormality detection method based on a substation equipment portrait provided in an embodiment of the present application is shown;

[0017] Figure 2 A schematic flow chart of an abnormality detection method based on substation equipment profiling provided according to an embodiment of the present application is shown;

[0018] Figure 3 A schematic flow chart of an abnormality detection method based on substation equipment portrait provided according to an embodiment of the present application is shown;

[0019] Figure 4 A structural block diagram of an abnormality detection device based on substation equipment profiling provided according to an embodiment of the present application is shown.

[0020] The above drawings include the following reference numerals:

[0021] 102, processor; 104, memory; 106, transmission device; 108, input and output devices. DETAILED DESCRIPTION

[0022] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0023] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0024] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] As introduced in the background technology, the prior art does not perform targeted detection on the complexity and correlation of primary and secondary equipment, which brings challenges to equipment status detection. In order to solve the problem that the correlation between primary and secondary equipment is not considered at the same time in the prior art, the correlation factors between equipment are ignored and the equipment status is directly detected, resulting in low detection accuracy, the embodiments of the present application provide an anomaly detection method, anomaly detection device, computer-readable storage medium and computer program product based on substation equipment profiling.

[0026] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0027] The method embodiments provided in the embodiments of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Taking running on a mobile terminal as an example, Figure 1 1 is a hardware structure block diagram of a mobile terminal of an abnormality detection method based on substation equipment profiling according to an embodiment of the present invention. Figure 1 As shown, the mobile terminal may include one or more ( Figure 1 Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data, wherein the mobile terminal may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1The structure shown is only for illustration and does not limit the structure of the mobile terminal. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations shown.

[0028] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the display method of device information in the embodiment of the present invention. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, the above method is implemented. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories can be connected to the mobile terminal via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof. The transmission device 106 is used to receive or send data via a network. The above-mentioned specific examples of the network may include a wireless network provided by a communication provider of the mobile terminal. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0029] In this embodiment, a method for detecting anomalies based on substation equipment profiling that runs on a mobile terminal, a computer terminal or a similar computing device is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0030] Figure 2 : is a flow chart of an abnormality detection method based on substation equipment profiling according to an embodiment of the present application. Figure 2 As shown, the method comprises the following steps:

[0031] Step S201, real-time acquisition of first device data and second device data, wherein the first device data is real-time data of a primary device in the substation at the current moment, and the second device data is operating status data of a secondary device in the substation at the current moment and related information with the corresponding primary device.

[0032] Specifically, before accurately detecting the status of primary and secondary equipment in a substation, it is necessary to obtain in real time the equipment data generated by the primary and secondary equipment in real time and their specific values ​​(ie, the first equipment data and the second equipment data).

[0033] Step S202, input the above-mentioned first device data into the first deep learning model, so as to use the above-mentioned first deep learning model to perform abnormality detection on the above-mentioned primary device on the above-mentioned first device data, and obtain a first device state probability value. The above-mentioned first deep learning model is obtained by training with a recurrent neural network algorithm using multiple groups of first training data, and each group of the above-mentioned first training data includes the first sample device data of the above-mentioned primary device and the corresponding second sample state probability value of the above-mentioned primary device. The above-mentioned first sample device data at least includes the device operating parameters, environmental information, normal state information of the device, abnormal state information of the device, interval duration of device abnormality and device failure record of the above-mentioned primary device collected within the same interval time, and the above-mentioned first device state probability value is the probability value of the device state of the above-mentioned primary device.

[0034] Specifically, the equipment data generated in real time by the primary equipment in the substation and its specific values ​​are input into the trained first deep learning model in the form of a data set for prediction, and the probability value of the primary equipment status of the substation, that is, the above-mentioned first equipment status probability value, is output.

[0035] It should also be noted that the substation includes primary equipment and secondary equipment. Different types of equipment exhibit different failure / abnormal modes due to their different structures, designs, working principles and usage environments. Primary equipment includes transformers, circuit breakers, disconnectors, and mutual inductors, which are mainly used for power transmission. On the one hand, transformers, circuit breakers, disconnectors, and mutual inductors are exposed to the outside air and are easily affected by wear and corrosion due to the presence of external factors, resulting in failures or abnormalities. On the other hand, when the equipment in the substation exceeds the rated value of the equipment during power consumption, it will cause failures or abnormalities due to overload, short circuit, overvoltage and other factors.

[0036] The status of the primary equipment of the substation changes over time. The first training set data uses the data generated by the time series as the data set. For this reason, the first deep learning model selects the recurrent neural network model (RNN) to predict the equipment status of the primary equipment. The recurrent neural network model (RNN) is particularly effective for processing long-term relationships in time series data. The equipment operating parameters in the first training data include equipment current, equipment voltage, equipment active power, equipment reactive power, insulation level, equipment temperature, equipment switch status, equipment factory date, equipment model, equipment material, and equipment physical parameters. Since transformers, circuit breakers, disconnectors, and mutual inductors can flexibly adjust equipment operating parameters during the power transmission process, they are devices for the stable operation of the substation power grid. For this reason, the equipment current, equipment voltage, equipment active power, equipment reactive power, insulation level, equipment temperature, and equipment switch status of transformers, circuit breakers, disconnectors, and mutual inductors are collected. The first training data constitutes a time series data set; Among them, the reasons for considering equipment operating parameters are as follows: equipment current and equipment voltage can evaluate the equipment load and power quality, equipment reactive power and equipment active power can reflect the equipment's heat conversion efficiency and power quality, insulation level is directly related to the safety and stable operation of the equipment, equipment temperature can evaluate the equipment's thermal performance and heat dissipation efficiency indicators, equipment overheating is also the main factor of equipment failure / abnormality, equipment switch status is used to ensure the on / off status of the equipment switch and then judge the equipment's operating status; the equipment's factory date has a certain correlation with the predicted equipment status (equipment performance, failure rate, maintenance cycle), and different models of equipment have different Design, performance parameters and applicable scenarios help the model distinguish the differences between different models, so as to more accurately predict or analyze the operating status of the equipment. Different materials of the equipment have different properties such as high temperature resistance, wear resistance, and corrosion resistance, which have an important impact on the operation and maintenance of the equipment. Physical parameters are closely related to the operation mode, heat dissipation performance, and installation requirements of the equipment, so they are also important factors to be considered in model analysis; Environmental information includes: rainfall, dust concentration, corrosive gas concentration, temperature, humidity, wind speed, light intensity, and ice cover. As rainfall continues to increase, rainwater penetrates into the equipment and causes a short circuit. Excessive dust concentration and corrosive gas concentration High humidity can cause poor contact due to corrosion, and can also cause the insulation layer of the equipment to age or even break down. Temperature is one of the key factors affecting equipment performance. High temperature may cause the equipment to overheat, accelerate the aging of components, and even cause failures. Low temperature can cause equipment to start up or run unstably. Therefore, understanding temperature information can help the model predict the operating status and possible failure modes of the equipment at different temperatures. Humidity has an important impact on the electrical and insulation properties of the equipment. High humidity can cause condensation inside the equipment, resulting in circuit short circuits or corrosion. Low humidity can cause static electricity accumulation, increasing the risk of equipment failure. Therefore, humidity information is crucial for predicting the electrical performance and life of the equipment.Wind speed affects the heat dissipation and dust prevention performance of the equipment. High wind speed can accelerate the heat dissipation of the equipment, but it will also bring more dust and pollutants. Low wind speed leads to poor heat dissipation of the equipment and increases the risk of failure. Therefore, understanding wind speed information can help the model predict the heat dissipation and dust prevention performance of the equipment under different wind speed conditions. For equipment, light intensity may affect its operating status. For example, the photoelectric conversion efficiency of solar panels is closely related to light intensity. Some sensors or display devices may experience performance degradation or damage under strong light. Therefore, light intensity information is crucial for predicting the performance of these devices. Icing mainly affects the normal operation of outdoor equipment. Icing increases the weight of the equipment and causes structural damage. At the same time, icing may also affect the heat dissipation and electrical performance of the equipment. Therefore, understanding the icing situation can help the model predict the operating risks of the equipment in cold and humid environments. The normal status information of the equipment includes the normal status time of the equipment and the normal status mark of the equipment. The abnormal status information of the equipment includes the abnormal status time of the equipment and the abnormal status mark of the equipment. The normal interval time can reflect the time interval between normal operation and abnormal operation of the equipment. By analyzing this feature, the health status of the equipment and the probability of abnormality can be understood. On the other hand, if the equipment abnormality interval time is short, it may mean that there is a defect or problem in the equipment, which may cause frequent equipment failures. By analyzing the equipment abnormality interval time in the training data, these failure modes can be extracted and useful information can be provided for the prediction model. The equipment failure record includes the time of failure, the type of equipment failure, and the operating parameters before the equipment failure. These three pieces of information can capture important information about equipment failure / abnormality. Since the first training data uses a time series data set, time series analysis can capture the temporal dependence of equipment abnormality, that is, the abnormality may be affected by equipment operating parameters, environmental information, equipment normal status information, equipment abnormal status information, equipment abnormality interval time, and equipment failure records. This analysis method can reveal the trend of equipment status changes over time and better predict and prevent future failures. ;

[0037] Step S203, input the above-mentioned second device data into the second deep learning model, so as to use the above-mentioned second deep learning model to perform abnormal detection of the above-mentioned secondary device on the above-mentioned second device data, and obtain the second device state probability value. The above-mentioned second deep learning model is obtained by training with a plurality of groups of second training data through a long short-term memory network algorithm, and each group of the above-mentioned second training data includes the second sample device data of the above-mentioned secondary device and the corresponding first sample state probability value. The above-mentioned second sample device data at least includes a sample target label, a sample target correlation coefficient and the sample operation data of the above-mentioned second device. The above-mentioned target label is the normal state information and the abnormal state information of the above-mentioned primary device obtained by the above-mentioned primary device. The above-mentioned sample target correlation coefficient is used for the degree of association between the above-mentioned primary device and the corresponding above-mentioned secondary device. The above-mentioned second device state probability value is the probability value of the above-mentioned device state of the above-mentioned secondary device.

[0038] Specifically, the equipment data generated in real time by the secondary equipment in the substation and its specific values ​​are input into the trained second deep learning model in the form of a data set for prediction, and the probability value of the secondary equipment status of the substation is output, that is, the above-mentioned second equipment status probability value.

[0039] It should also be noted that since the equipment status of the secondary equipment is affected by historical data, the second training data is collected from historical data, and the second deep learning model uses a long short-term memory network (LSTM) to predict the equipment status of the secondary equipment. The secondary equipment of the substation includes protection devices, measurement and control devices, automation systems, and communication equipment, which are used to detect and control the primary equipment in real time. Secondary equipment also plays a vital role in the substation. When the secondary equipment is abnormal, it may cause the primary equipment to malfunction, refuse to operate, or expand the fault, thereby affecting the safety and stable operation of the substation. For this reason, the second training data of the second deep learning model for detecting the equipment status of the secondary equipment includes the target label of the primary equipment, the sample target correlation coefficient between the primary equipment and the secondary equipment, the equipment current, the equipment voltage, the equipment temperature, the communication delay time, the packet loss rate, and the deviation between the target value and the feedback value. Protection devices, measurement and control devices, automation systems, and communication equipment are responsible for monitoring, controlling, protecting, and communicating the primary equipment in the substation to ensure safe, stable, and efficient operation of the power system. Power supply is the basis for the stable operation of secondary equipment. To this end, the equipment current, equipment voltage, equipment temperature, communication delay time, packet loss rate, and deviation between the target value and the feedback value of the protection devices, measurement and control devices, automation systems, and communication equipment are collected. The current and voltage of the power supply can reflect the working status of the power supply. For example, abnormal power supply parameters may lead to equipment failure. In addition, the current and voltage can also reflect the load condition of the equipment. For example, the current and voltage data in the protection device can explain the operating status of the power grid and whether there is an abnormal state of overload or short circuit. Secondary equipment will generate heat during operation. The temperature of the equipment can reflect its working status and heat dissipation performance. If the equipment temperature is too high, it means that the equipment is overloaded or short-circuited. Poor heat dissipation. The operating status of the communication equipment has an important impact on the overall performance of the secondary equipment. Parameters such as communication delay time and packet loss rate can reflect the performance and stability of the communication equipment. The measurement and control device has a target value and a feedback value. The target value of the measurement and control device refers to the output value or state expected by the operator or the automatic control system. The target value is based on the operating requirements of the substation and the operating parameters of the equipment. The feedback value is the output value or state actually detected by the measurement and control device, which reflects the current operating status of the equipment. The operating status of the equipment is reflected by calculating the deviation between the target value and the feedback value. A large deviation means that the equipment does not respond to the control signal correctly, or there is a fault or interference, resulting in a decrease or instability in equipment performance. The deviation between the target value and the feedback value can be determined according to the deviation data range set by the equipment and parameters. If the deviation is not within the deviation data range, it means that the deviation is too large.

[0040] It is worth noting that since the secondary device is used to detect the status of the primary device, there is a certain correlation between the primary device and the secondary device. When detecting the device status of the secondary device, it is crucial to consider the correlation between the device status of the primary device and the secondary device. For this reason, the target label of the primary device and the correlation coefficient between the primary device and the secondary device are considered in the second training data to help the model predict the status of the secondary device more accurately, especially when the status of the primary device changes. The target label of the primary device refers to the normal status information and abnormal status information of the device obtained by the primary device. The status and function of the secondary device usually depend on the status of the primary device. If the primary device fails, the secondary device will record abnormal data and trigger an alarm. Therefore, there is a certain correlation between the primary device and the secondary device. Therefore, the correlation coefficient between the primary device and the secondary device is used to achieve the association between the primary device and the secondary device.

[0041] Step S204: verifying the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device are abnormal.

[0042] Specifically, the first device state probability value and the second device state probability value are compared with the state definition range. When the first device state probability value or the second device state probability value is not within the state definition range, it is considered that the current device state is in an abnormal state and requires increased vigilance or maintenance. Alternatively, when the first device state probability value or the second device state probability value exceeds the state definition value, it is considered that the device state is normal; when it is lower than the state definition value, it is considered that the device state is abnormal, that is, an abnormal situation has occurred.

[0043] Step S205: When the above-mentioned abnormal situation occurs in the above-mentioned primary device or the above-mentioned secondary device, abnormal repair processing is performed on the above-mentioned primary device or the above-mentioned secondary device.

[0044] Specifically, when the above-mentioned abnormal situation occurs in the above-mentioned primary equipment or the above-mentioned secondary equipment, preventive maintenance measures are adopted to reduce the probability of equipment failure, and the operator immediately takes measures to repair and handle it to prevent the failure from expanding or affecting production.

[0045] In this embodiment, by constructing a first deep learning model and a second deep learning model, the first deep learning model predicts the primary device and outputs the probability value of the device state of the primary device, and the second deep learning model predicts the device state of the secondary device and outputs the probability value of the device state of the secondary device. The device state is judged according to the probability value, and in the training process of the second deep learning model, the parameters related to the secondary device and the device state and the correlation coefficient of the secondary device and the primary device in the device operation state are comprehensively considered as the second training data, so that the second deep learning model trained by using the final correlation coefficient for model training will be more accurate in device state detection and improve the detection accuracy. This application solves the problem that the correlation between the primary device and the secondary device is not considered at the same time in the prior art, and the correlation factor between the devices is ignored to directly detect the device state, resulting in low detection accuracy.

[0046] In order to enable those skilled in the art to more clearly understand the technical solution of the present application, the implementation process of the anomaly detection method based on substation equipment profiling of the present application will be described in detail below in combination with specific embodiments.

[0047] In order to consider the correlation between the primary device and the secondary device to reduce the detection error, in an optional implementation manner, before the above step S201, the method further includes:

[0048] Step S301, obtaining historical data of the above-mentioned substation, the above-mentioned historical data at least includes equipment information data, equipment operation data, event record data and maintenance management data, the above-mentioned equipment information data at least includes the equipment name, equipment type, equipment rated voltage, current, frequency, power and control parameters of each equipment in the substation, the above-mentioned equipment operation data at least includes various parameter data of each of the above-mentioned equipment in the power transmission process and the discharge process and the equipment status information of the above-mentioned equipment at each time point in the historical time period, the above-mentioned event record data at least includes the accident type, accident time and various parameter data when the equipment fails or is abnormal, and the above-mentioned maintenance management data includes the record data of regular equipment inspection, maintenance and calibration, the historical record data of equipment failure or abnormality, cause analysis and processing results, and equipment repair and replacement parts information;

[0049] Step S302, filtering out state-related data from the historical data, wherein the state-related data is data related to the state of the primary device and the state of the secondary device;

[0050] Step S303, performing a preprocessing operation on the state-related data to obtain a training set, wherein the training set includes the first training data and the second training data, and the preprocessing operation at least includes data cleaning.

[0051] In the above embodiment, the sample data of the substation is the historical data generated by the substation during use. The historical data includes equipment information data, equipment operation data, event record data, maintenance management data, etc. The historical data is obtained from the substation automation system. The equipment information data mainly includes basic equipment information and technical parameters of the equipment, such as equipment name, equipment type, equipment rated voltage, current, frequency, power and control parameters, etc. The equipment operation data includes equipment operation parameter data and equipment operation status data, mainly including various parameter data of the equipment transmission and discharge process and equipment status information at each time point in the equipment historical time period. The event record data mainly includes the accident type, accident time and various parameter data when the accident occurs when the equipment fails or abnormalities. The maintenance management data includes record data of regular equipment inspection, maintenance and calibration, historical record data of equipment failure or abnormality, cause analysis and processing results, and equipment repair and replacement parts information, etc. Since the amount of sample data obtained is relatively large, considering detecting the status of the primary and secondary equipment of the substation, the sample data needs to be preprocessed. First, the data related to the primary equipment status and the secondary equipment status are filtered out from the sample data, and the data irrelevant to the equipment type, equipment operating parameters, environmental information, equipment normal status information, equipment abnormal status information, equipment abnormal interval duration, and equipment fault records are deleted. After the screening, the missing value data and the duplicate value data are deleted to obtain the above training set.

[0052] In order to more accurately evaluate the degree of association between the primary device and the secondary device, in an optional implementation manner, before the above step S203, the method further includes:

[0053] Step S401, calculating the correlation coefficient of each parameter pair in the primary device and the secondary device according to the first formula, the first formula is: r i is the correlation coefficient of the i-th parameter pair, n is the total number of parameter pairs, i = (1, 2, 3, ..., n), x i is the observed value of the primary device in the i-th parameter pair, y i is the observed value of the secondary device in the i-th parameter pair, the observed value is a data value randomly selected from the historical data, x is the average value of the observed values ​​of the primary device in all the parameter pairs, is the average value of the above-mentioned observed values ​​of the above-mentioned secondary equipment in all the above-mentioned parameter pairs, and the above-mentioned parameter pair represents the combination formed by the above-mentioned observed values ​​corresponding to the above-mentioned primary equipment and the above-mentioned secondary equipment;

[0054] Step S402, dividing each of the parameter pairs into corresponding correlation parameters according to the correlation coefficient of each of the parameter pairs, wherein the correlation parameters are divided into strong correlation parameters, weak correlation parameters and no correlation parameters;

[0055] Step S403, setting corresponding weights for the above-mentioned strongly associated parameters, the above-mentioned weakly associated parameters and the above-mentioned unassociated parameters;

[0056] Step S404, calculating the sample target correlation coefficient between the primary device and the secondary device according to a second formula, the second formula is: R is the above sample target correlation coefficient, ω i is the weight of the above-mentioned associated parameter to which the i-th parameter pair belongs.

[0057] In the above embodiment, the correlation coefficient of each parameter pair in the above primary device and the above secondary device is calculated according to the first formula, and different parameter pairs are divided into strong correlation parameters, weak correlation parameters or uncorrelated parameters according to the obtained value of the correlation coefficient between the primary device and the secondary device. Corresponding weights are assigned to the strongly correlated parameters, weakly correlated parameters and uncorrelated parameters of each parameter pair respectively. Strong correlation plays an important role in the correlation evaluation process between the primary device and the secondary device, which means that the primary device and the secondary device are closely and significantly correlated. In order to more accurately evaluate the degree of correlation between the primary device and the secondary device, the weight of the strongly correlated parameter can be assigned to 2, the weight of the weakly correlated parameter can be assigned to 0.5, and the weight of the uncorrelated parameter can be assigned to 0.1; the final correlation coefficient between the primary device and the secondary device is calculated according to the correlation coefficient and weight of the primary device and the secondary device of each parameter pair, that is, the above sample target correlation coefficient.

[0058] In order to detect the status of the primary device and the secondary device in a targeted manner, in an optional implementation manner, the above step S402 includes:

[0059] Step S4021, forming a correlation matrix according to all the correlation coefficients of all the parameter pairs in the primary device and the secondary device, wherein the rows of the correlation matrix represent the observed values ​​of the primary device, the columns of the correlation matrix represent the observed values ​​of the secondary device, and the elements of the correlation matrix represent the correlation coefficients;

[0060] Step S4022, drawing a heat map according to the association matrix, wherein each cell in the heat map represents the association coefficient of one of the parameter pairs, and the color depth in the heat map represents the strength and direction of the association;

[0061] Step S4023, determining the parameter pairs whose absolute values ​​of the correlation coefficients are greater than or equal to the first threshold as the strongly correlated parameters;

[0062] Step S4024, determining the parameter pairs whose absolute values ​​of the correlation coefficients are greater than or equal to the second threshold and less than the first threshold as the weakly correlated parameters;

[0063] Step S4025: Determine the parameter pairs whose absolute values ​​of the correlation coefficients are smaller than the second threshold as the uncorrelated parameters.

[0064] In the above embodiment, firstly, according to the correlation coefficient between the primary device and the secondary device, an association matrix is ​​formed, including the correlation coefficient between all parameter pairs, and a heat map is drawn according to the association matrix. Each cell in the heat map represents the correlation coefficient of a parameter pair, and the color depth indicates the strength and direction of the correlation, red indicates strong correlation, yellow indicates weak correlation, and white indicates no correlation. Then, the strong correlation parameter and strong correlation coefficient at the red position, the weak correlation parameter and weak correlation coefficient at the yellow position, and the no correlation parameter and irrelevant connection number at the white position are obtained respectively. By comprehensively analyzing the strong correlation coefficient, the weak correlation coefficient and the irrelevant connection number, the correlation coefficient of the strong correlation parameter is defined as an absolute value greater than 0.7, the correlation coefficient of the weak correlation parameter is defined as an absolute value between 0.3 and 0.7, and the correlation coefficient of the no correlation parameter is defined as an absolute value less than 0.3. According to the range of the relationship coefficient corresponding to the parameter pair, the parameter pair is determined as a specific correlation parameter. It is possible to detect the device status of the primary device and the secondary device in a targeted manner, so as to better detect the device status of the primary device and the secondary device according to the strength of the relationship between the primary device and the secondary device.

[0065] In order to handle the exception in time, in an optional implementation, the above step S204 includes:

[0066] Step S2041, when the first device state probability value is greater than or equal to the first state boundary value, determining that the primary device is in a normal state;

[0067] Step S2042, when the probability value of the first equipment state is less than the first state limit value and greater than or equal to the second state limit value, issuing a first prompt message, wherein the first prompt message is used to prompt to take preventive maintenance measures for the primary equipment;

[0068] Step S2043, when the state probability value of the first device is less than the second state limit value, it is determined that the first device is in an abnormal state and the abnormal situation occurs, and a second prompt message is issued, wherein the second prompt message is used to prompt the primary device to handle the abnormal situation;

[0069] Step S2044, when the second device state probability value is greater than or equal to the third state boundary value, determining that the secondary device is in a normal state;

[0070] Step S2045, when the second device state probability value is less than the third state limit value and greater than or equal to the fourth state limit value, issuing a third prompt message, wherein the third prompt message is used to prompt to take preventive maintenance measures for the secondary device;

[0071] Step S2046, when the second device state probability value is less than the fourth state limit value, it is determined that the second device is in an abnormal state and the abnormal situation occurs, and a fourth prompt message is issued, and the fourth prompt message is used to prompt the secondary device to perform abnormal processing.

[0072] In the above embodiment, a sigmoid function is set at the input layer of the deep learning model, which is used as the activation function of the deep learning model, maps any input to the probability distribution, and converts the output of the deep learning model into a probability value. For example, the probability value output by the deep learning model ranges from 0.1 to 1. The closer the probability value output by the model is to 1, the higher the degree to which the device state is in a normal state. The closer the probability value output by the model is to 0.1, the higher the degree to which the device state is in an abnormal state, and measures should be taken immediately to rescue it. A state limit value or state limit range can be set for the probability value output by the model. The state limit value can be fixed, one or more. When the prediction result output by the model is not within the state limit range, it is considered that the current device state is in an abnormal state, and vigilance or maintenance is required. Alternatively, if the prediction result exceeds the state limit value, the device state is considered to be in a normal state, and if it is lower than the state limit value, the device state is considered to be in an abnormal state. The probability values ​​of the first deep learning model and the second deep learning model do not necessarily have to have the same value range, that is, the first state defining value and the third state defining value do not need to be the same, and the second state defining value and the fourth state defining value do not need to be the same, but the second state defining value and the fourth state defining value are usually both 1. The first state defining value is smaller than the second state defining value, and the second state defining value is smaller than the fourth state defining value.

[0073] In order to improve the accuracy of anomaly detection, in an optional implementation, before the above step S204, the method further includes:

[0074] Step S501, adjusting the first state boundary value and the second state boundary value in real time according to the historical data and real-time data corresponding to the primary device;

[0075] Step S502: adjusting the third state boundary value and the fourth state boundary value in real time according to the historical data and real-time data corresponding to the secondary device.

[0076] In the above embodiments, the state boundary value or state boundary range is not a fixed value, but is adjusted in real time according to the historical data and real-time data of the equipment operation. Multiple tests and adjustments are required to find the most suitable state boundary value or state boundary range setting, which can improve the accuracy of abnormality detection.

[0077] In order to prevent the fault from expanding or affecting production, in an optional implementation manner, the above step S205 includes:

[0078] Step S2051, when the above-mentioned abnormal situation occurs in the above-mentioned primary device, determine a first abnormality repair strategy according to the above-mentioned abnormal situation, and perform abnormality repair processing on the above-mentioned primary device according to the above-mentioned first abnormality repair strategy;

[0079] Step S2052: when the above-mentioned abnormal situation occurs in the above-mentioned secondary device, a second abnormality repair strategy is determined according to the above-mentioned abnormal situation, and abnormality repair processing is performed on the above-mentioned secondary device according to the above-mentioned second abnormality repair strategy.

[0080] In the above embodiment, the fused data generated in real time in the substation and its specific values ​​are input into the trained deep learning model in the form of a data set for prediction, and the probability value of the equipment status of the substation is output. The equipment status includes the normal state of the equipment and the abnormal state of the equipment. When the abnormal state of the equipment is predicted, it means that the substation equipment may enter an abnormal state at a certain point in the future. By capturing certain abnormal conditions, the operator should pay attention to the current status of the substation equipment and verify whether there are any abnormal signs. Preventive maintenance measures can also be adopted to reduce the probability of equipment failure. If an abnormal condition occurs in the primary or secondary equipment, the operator should immediately determine the repair strategy according to the abnormal condition, and take measures to repair and process it according to the repair strategy to prevent the fault from expanding or affecting production.

[0081] This embodiment relates to a specific abnormality detection method based on substation equipment portrait, such as Figure 3 As shown, the following steps are included:

[0082] Step S1: Obtain sample data of the substation. The sample data of the substation is historical data generated during the use of the substation. The historical data includes equipment information data, equipment operation data, event record data, maintenance management data, etc. The historical data is obtained from the substation automation system;

[0083] Step S2: pre-process the sample data, filter out data related to the primary equipment status and the secondary equipment status from the sample data, delete data that is not related to the equipment type, equipment operating parameters, environmental information, equipment normal status information, equipment abnormal status information, equipment abnormality interval, and equipment fault records, and delete missing value data and duplicate value data from the filtered sample data.

[0084] Step S3: Establish a deep learning model, the deep learning model includes a first deep learning model and a second deep learning model, the first deep learning model is used to detect the equipment status of the primary equipment of the substation, and the second deep learning model is used to detect the equipment status of the substation.

[0085] Step S4: Obtain a training set of a deep learning model, the training set comprising a first training set and a second training set, the first training set being used to train the first deep learning model, and the second training set being used to train the second deep learning model;

[0086] Step S5: input the training set into the deep learning model for training, and obtain a trained deep learning model;

[0087] Step S6: Input the data to be predicted of the substation into the trained deep learning model, and output the prediction results, namely the first device state probability value and the second device state probability value.

[0088] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0089] The embodiment of the present application also provides an abnormality detection device based on the substation equipment portrait. It should be noted that the abnormality detection device based on the substation equipment portrait of the embodiment of the present application can be used to execute the abnormality detection method based on the substation equipment portrait provided by the embodiment of the present application. The device is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions that have been made will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceived.

[0090] The following introduces the anomaly detection device based on substation equipment profiling provided in an embodiment of the present application.

[0091] Figure 4is a structural block diagram of an abnormality detection device based on substation equipment profiling according to an embodiment of the present application. Figure 4 As shown, the device comprises:

[0092] The first acquisition unit 10 is used to acquire first device data and second device data in real time, wherein the first device data is real-time data of a primary device in the substation at the current moment, and the second device data is operating status data of a secondary device in the substation at the current moment and related information with the corresponding primary device.

[0093] Specifically, before accurately detecting the status of primary and secondary equipment in a substation, it is necessary to obtain in real time the equipment data generated by the primary and secondary equipment in real time and their specific values ​​(ie, the first equipment data and the second equipment data).

[0094] The first input-output unit 20 is used to input the above-mentioned first device data into the first deep learning model, so as to use the above-mentioned first deep learning model to perform abnormal detection of the above-mentioned primary device on the above-mentioned first device data to obtain a first device state probability value. The above-mentioned first deep learning model is obtained by training with a recurrent neural network algorithm using multiple groups of first training data. Each group of the above-mentioned first training data includes the first sample device data of the above-mentioned primary device and the corresponding second sample state probability value of the above-mentioned primary device. The above-mentioned first sample device data at least includes the device operating parameters, environmental information, normal state information of the device, abnormal state information of the device, interval duration of device abnormality and device failure record of the above-mentioned primary device collected within the same interval time. The above-mentioned first device state probability value is the probability value of the device state of the above-mentioned primary device.

[0095] Specifically, the equipment data generated in real time by the primary equipment in the substation and its specific values ​​are input into the trained first deep learning model in the form of a data set for prediction, and the probability value of the primary equipment status of the substation, that is, the above-mentioned first equipment status probability value, is output.

[0096] It should also be noted that the substation includes primary equipment and secondary equipment. Different types of equipment exhibit different failure / abnormal modes due to their different structures, designs, working principles and usage environments. Primary equipment includes transformers, circuit breakers, disconnectors, and mutual inductors, which are mainly used for power transmission. On the one hand, transformers, circuit breakers, disconnectors, and mutual inductors are exposed to the outside air and are easily affected by wear and corrosion due to the presence of external factors, resulting in failures or abnormalities. On the other hand, when the equipment in the substation exceeds the rated value of the equipment during power consumption, it will cause failures or abnormalities due to overload, short circuit, overvoltage and other factors.

[0097] The status of the primary equipment of the substation changes over time. The first training set data uses the data generated by the time series as the data set. For this reason, the first deep learning model selects the recurrent neural network model (RNN) to predict the equipment status of the primary equipment. The recurrent neural network model (RNN) is particularly effective for processing long-term relationships in time series data. The equipment operating parameters in the first training data include equipment current, equipment voltage, equipment active power, equipment reactive power, insulation level, equipment temperature, equipment switch status, equipment factory date, equipment model, equipment material, and equipment physical parameters. Since transformers, circuit breakers, disconnectors, and mutual inductors can flexibly adjust equipment operating parameters during the power transmission process, they are devices for the stable operation of the substation power grid. For this reason, the equipment current, equipment voltage, equipment active power, equipment reactive power, insulation level, equipment temperature, and equipment switch status of transformers, circuit breakers, disconnectors, and mutual inductors are collected. The first training data constitutes a time series data set; Among them, the reasons for considering equipment operating parameters are as follows: equipment current and equipment voltage can evaluate the equipment load and power quality, equipment reactive power and equipment active power can reflect the equipment's heat conversion efficiency and power quality, insulation level is directly related to the safety and stable operation of the equipment, equipment temperature can evaluate the equipment's thermal performance and heat dissipation efficiency indicators, equipment overheating is also the main factor of equipment failure / abnormality, equipment switch status is used to ensure the on / off status of the equipment switch and then judge the equipment's operating status; the equipment's factory date has a certain correlation with the predicted equipment status (equipment performance, failure rate, maintenance cycle), and different models of equipment have different Design, performance parameters and applicable scenarios help the model distinguish the differences between different models, so as to more accurately predict or analyze the operating status of the equipment. Different materials of the equipment have different properties such as high temperature resistance, wear resistance, and corrosion resistance, which have an important impact on the operation and maintenance of the equipment. Physical parameters are closely related to the operation mode, heat dissipation performance, and installation requirements of the equipment, so they are also important factors to be considered in model analysis; Environmental information includes: rainfall, dust concentration, corrosive gas concentration, temperature, humidity, wind speed, light intensity, and ice cover. As rainfall continues to increase, rainwater penetrates into the equipment and causes a short circuit. Excessive dust concentration and corrosive gas concentration High humidity can cause poor contact due to corrosion, and can also cause the insulation layer of the equipment to age or even break down. Temperature is one of the key factors affecting equipment performance. High temperature may cause the equipment to overheat, accelerate the aging of components, and even cause failures. Low temperature can cause equipment to start up or run unstably. Therefore, understanding temperature information can help the model predict the operating status and possible failure modes of the equipment at different temperatures. Humidity has an important impact on the electrical and insulation properties of the equipment. High humidity can cause condensation inside the equipment, resulting in circuit short circuits or corrosion. Low humidity can cause static electricity accumulation, increasing the risk of equipment failure. Therefore, humidity information is crucial for predicting the electrical performance and life of the equipment.Wind speed affects the heat dissipation and dust prevention performance of the equipment. High wind speed can accelerate the heat dissipation of the equipment, but it will also bring more dust and pollutants. Low wind speed leads to poor heat dissipation of the equipment and increases the risk of failure. Therefore, understanding wind speed information can help the model predict the heat dissipation and dust prevention performance of the equipment under different wind speed conditions. For equipment, light intensity may affect its operating status. For example, the photoelectric conversion efficiency of solar panels is closely related to light intensity. Some sensors or display devices may experience performance degradation or damage under strong light. Therefore, light intensity information is crucial for predicting the performance of these devices. Icing mainly affects the normal operation of outdoor equipment. Icing increases the weight of the equipment and causes structural damage. At the same time, icing may also affect the heat dissipation and electrical performance of the equipment. Therefore, understanding the icing situation can help the model predict the operating risks of the equipment in cold and humid environments. The normal status information of the equipment includes the normal status time of the equipment and the normal status mark of the equipment. The abnormal status information of the equipment includes the abnormal status time of the equipment and the abnormal status mark of the equipment. The normal interval time can reflect the time interval between normal operation and abnormal operation of the equipment. By analyzing this feature, the health status of the equipment and the probability of abnormality can be understood. On the other hand, if the equipment abnormality interval time is short, it may mean that there is a defect or problem in the equipment, which may cause frequent equipment failures. By analyzing the equipment abnormality interval time in the training data, these failure modes can be extracted and useful information can be provided for the prediction model. The equipment failure record includes the time of failure, the type of equipment failure, and the operating parameters before the equipment failure. These three pieces of information can capture important information about equipment failure / abnormality. Since the first training data uses a time series data set, time series analysis can capture the temporal dependence of equipment abnormality, that is, the abnormality may be affected by equipment operating parameters, environmental information, equipment normal status information, equipment abnormal status information, equipment abnormality interval time, and equipment failure records. This analysis method can reveal the trend of equipment status changes over time and better predict and prevent future failures. ;

[0098] The second input-output unit 30 is used to input the above-mentioned second device data into the second deep learning model, so as to use the above-mentioned second deep learning model to perform abnormal detection of the above-mentioned secondary device on the above-mentioned second device data, and obtain the second device state probability value. The above-mentioned second deep learning model is obtained by training with a plurality of groups of second training data through a long short-term memory network algorithm. Each group of the above-mentioned second training data includes the second sample device data of the above-mentioned secondary device and the corresponding first sample state probability value. The above-mentioned second sample device data at least includes a sample target label, a sample target correlation coefficient and the sample operation data of the above-mentioned second device. The above-mentioned target label is the normal state information and the abnormal state information of the device obtained by the above-mentioned primary device. The above-mentioned sample target correlation coefficient is used for the degree of association between the above-mentioned primary device and the corresponding above-mentioned secondary device. The above-mentioned second device state probability value is the probability value of the above-mentioned device state of the above-mentioned secondary device.

[0099] Specifically, the equipment data generated in real time by the secondary equipment in the substation and its specific values ​​are input into the trained second deep learning model in the form of a data set for prediction, and the probability value of the secondary equipment status of the substation is output, that is, the above-mentioned second equipment status probability value.

[0100] It should also be noted that since the equipment status of the secondary equipment is affected by historical data, the second training data is collected from historical data, and the second deep learning model uses a long short-term memory network (LSTM) to predict the equipment status of the secondary equipment. The secondary equipment of the substation includes protection devices, measurement and control devices, automation systems, and communication equipment, which are used to detect and control the primary equipment in real time. Secondary equipment also plays a vital role in the substation. When the secondary equipment is abnormal, it may cause the primary equipment to malfunction, refuse to operate, or expand the fault, thereby affecting the safety and stable operation of the substation. For this reason, the second training data of the second deep learning model for detecting the equipment status of the secondary equipment includes the target label of the primary equipment, the sample target correlation coefficient between the primary equipment and the secondary equipment, the equipment current, the equipment voltage, the equipment temperature, the communication delay time, the packet loss rate, and the deviation between the target value and the feedback value. Protection devices, measurement and control devices, automation systems, and communication equipment are responsible for monitoring, controlling, protecting, and communicating the primary equipment in the substation to ensure safe, stable, and efficient operation of the power system. Power supply is the basis for the stable operation of secondary equipment. To this end, the equipment current, equipment voltage, equipment temperature, communication delay time, packet loss rate, and deviation between the target value and the feedback value of the protection devices, measurement and control devices, automation systems, and communication equipment are collected. The current and voltage of the power supply can reflect the working status of the power supply. For example, abnormal power supply parameters may lead to equipment failure. In addition, the current and voltage can also reflect the load condition of the equipment. For example, the current and voltage data in the protection device can explain the operating status of the power grid and whether there is an abnormal state of overload or short circuit. Secondary equipment will generate heat during operation. The temperature of the equipment can reflect its working status and heat dissipation performance. If the equipment temperature is too high, it means that the equipment is overloaded or short-circuited. Poor heat dissipation. The operating status of the communication equipment has an important impact on the overall performance of the secondary equipment. Parameters such as communication delay time and packet loss rate can reflect the performance and stability of the communication equipment. The measurement and control device has a target value and a feedback value. The target value of the measurement and control device refers to the output value or state expected by the operator or the automatic control system. The target value is based on the operating requirements of the substation and the operating parameters of the equipment. The feedback value is the output value or state actually detected by the measurement and control device, which reflects the current operating status of the equipment. The operating status of the equipment is reflected by calculating the deviation between the target value and the feedback value. A large deviation means that the equipment does not respond to the control signal correctly, or there is a fault or interference, resulting in a decrease or instability in equipment performance. The deviation between the target value and the feedback value can be determined according to the deviation data range set by the equipment and parameters. If the deviation is not within the deviation data range, it means that the deviation is too large.

[0101] It is worth noting that since the secondary device is used to detect the status of the primary device, there is a certain correlation between the primary device and the secondary device. When detecting the device status of the secondary device, it is crucial to consider the correlation between the device status of the primary device and the secondary device. For this reason, the target label of the primary device and the correlation coefficient between the primary device and the secondary device are considered in the second training data to help the model predict the status of the secondary device more accurately, especially when the status of the primary device changes. The target label of the primary device refers to the normal status information and abnormal status information of the device obtained by the primary device. The status and function of the secondary device usually depend on the status of the primary device. If the primary device fails, the secondary device will record abnormal data and trigger an alarm. Therefore, there is a certain correlation between the primary device and the secondary device. Therefore, the correlation coefficient between the primary device and the secondary device is used to achieve the association between the primary device and the secondary device.

[0102] The verification unit 40 is used to verify the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device have abnormal conditions.

[0103] Specifically, the first device state probability value and the second device state probability value are compared with the state definition range. When the first device state probability value or the second device state probability value is not within the state definition range, it is considered that the current device state is in an abnormal state and requires increased vigilance or maintenance. Alternatively, when the first device state probability value or the second device state probability value exceeds the state definition value, it is considered that the device state is normal; when it is lower than the state definition value, it is considered that the device state is abnormal, that is, an abnormal situation has occurred.

[0104] The abnormality processing unit 50 is used to perform abnormality repair processing on the primary device or the secondary device when the above-mentioned abnormal situation occurs in the above-mentioned primary device or the above-mentioned secondary device.

[0105] Specifically, when the above-mentioned abnormal situation occurs in the above-mentioned primary equipment or the above-mentioned secondary equipment, preventive maintenance measures are adopted to reduce the probability of equipment failure, and the operator immediately takes measures to repair and handle it to prevent the failure from expanding or affecting production.

[0106] In this embodiment, by constructing a first deep learning model and a second deep learning model, the first deep learning model predicts the primary device and outputs the probability value of the device state of the primary device, and the second deep learning model predicts the device state of the secondary device and outputs the probability value of the device state of the secondary device. The device state is judged according to the probability value, and in the training process of the second deep learning model, the parameters related to the secondary device and the device state and the correlation coefficient of the secondary device and the primary device in the device operation state are comprehensively considered as the second training data, so that the second deep learning model trained by using the final correlation coefficient for model training will be more accurate in device state detection and improve the detection accuracy. This application solves the problem that the correlation between the primary device and the secondary device is not considered at the same time in the prior art, and the correlation factor between the devices is ignored to directly detect the device state, resulting in low detection accuracy.

[0107] In order to consider the correlation between the primary device and the secondary device to reduce the detection error, in an optional implementation manner, the apparatus further includes:

[0108] A second acquisition unit is used to acquire historical data of the substation before acquiring the first device data and the second device data in real time, wherein the historical data at least includes device information data, device operation data, event record data and maintenance management data, wherein the device information data at least includes the device name, device type, device rated voltage, current, frequency, power and control parameters of each device in the substation, the device operation data at least includes various parameter data of each device in the power transmission process and the discharge process and the device status information of the device at each time point in the historical time period, the event record data at least includes the accident type, accident time and various parameter data when the equipment fails or is abnormal, and the maintenance management data includes record data of regular equipment inspection, maintenance and calibration, historical record data of equipment failure or abnormality, cause analysis and processing results and equipment repair and replacement parts information;

[0109] A screening unit, used to screen out state-related data from the historical data, wherein the state-related data is data related to the state of the primary device and the state of the secondary device;

[0110] The preprocessing unit is used to perform a preprocessing operation on the above-mentioned state-related data to obtain a training set, wherein the above-mentioned training set includes the first training data and the above-mentioned second training data, and the above-mentioned preprocessing operation at least includes data cleaning.

[0111] In the above embodiment, the sample data of the substation is the historical data generated by the substation during use. The historical data includes equipment information data, equipment operation data, event record data, maintenance management data, etc. The historical data is obtained from the substation automation system. The equipment information data mainly includes basic equipment information and technical parameters of the equipment, such as equipment name, equipment type, equipment rated voltage, current, frequency, power and control parameters, etc. The equipment operation data includes equipment operation parameter data and equipment operation status data, mainly including various parameter data of the equipment transmission and discharge process and equipment status information at each time point in the equipment historical time period. The event record data mainly includes the accident type, accident time and various parameter data when the accident occurs when the equipment fails or abnormalities. The maintenance management data includes record data of regular equipment inspection, maintenance and calibration, historical record data of equipment failure or abnormality, cause analysis and processing results, and equipment repair and replacement parts information, etc. Since the amount of sample data obtained is relatively large, considering detecting the status of the primary and secondary equipment of the substation, the sample data needs to be preprocessed. First, the data related to the primary equipment status and the secondary equipment status are filtered out from the sample data, and the data irrelevant to the equipment type, equipment operating parameters, environmental information, equipment normal status information, equipment abnormal status information, equipment abnormal interval duration, and equipment fault records are deleted. After the screening, the missing value data and the duplicate value data are deleted to obtain the above training set.

[0112] In order to more accurately evaluate the degree of association between the primary device and the secondary device, in an optional implementation manner, the apparatus further includes:

[0113] The first calculation unit is used to calculate the correlation coefficient of each parameter pair between the primary device and the secondary device according to the first formula before inputting the second device data into the second deep learning model to perform abnormal detection of the secondary device on the second device data using the second deep learning model to obtain the second device state probability value of the secondary device, wherein the first formula is: r i is the correlation coefficient of the i-th parameter pair, n is the total number of parameter pairs, i = (1, 2, 3, ..., n), x i is the observed value of the primary device in the i-th parameter pair, y i is the observed value of the secondary device in the i-th parameter pair, the observed value is a data value randomly selected from the historical data, x is the average value of the observed values ​​of the primary device in all the parameter pairs, is the average value of the above-mentioned observed values ​​of the above-mentioned secondary equipment in all the above-mentioned parameter pairs, and the above-mentioned parameter pair represents the combination formed by the above-mentioned observed values ​​corresponding to the above-mentioned primary equipment and the above-mentioned secondary equipment;

[0114] A division unit, used for dividing each of the parameter pairs into corresponding correlation parameters according to the correlation coefficient of each of the parameter pairs, wherein the correlation parameters are divided into strongly correlated parameters, weakly correlated parameters and uncorrelated parameters;

[0115] A setting unit, used to set corresponding weights for the above-mentioned strong correlation parameter, the above-mentioned weak correlation parameter and the above-mentioned uncorrelated parameter;

[0116] The second calculation unit is used to calculate the sample target correlation coefficient between the primary device and the secondary device according to a second formula, and the second formula is: R is the above sample target correlation coefficient, ω i is the weight of the above-mentioned associated parameter to which the i-th parameter pair belongs.

[0117] In the above embodiment, the correlation coefficient of each parameter pair in the above primary device and the above secondary device is calculated according to the first formula, and different parameter pairs are divided into strong correlation parameters, weak correlation parameters or uncorrelated parameters according to the obtained value of the correlation coefficient between the primary device and the secondary device. Corresponding weights are assigned to the strongly correlated parameters, weakly correlated parameters and uncorrelated parameters of each parameter pair respectively. Strong correlation plays an important role in the correlation evaluation process between the primary device and the secondary device, which means that the primary device and the secondary device are closely and significantly correlated. In order to more accurately evaluate the degree of correlation between the primary device and the secondary device, the weight of the strongly correlated parameter can be assigned to 2, the weight of the weakly correlated parameter can be assigned to 0.5, and the weight of the uncorrelated parameter can be assigned to 0.1; the final correlation coefficient between the primary device and the secondary device is calculated according to the correlation coefficient and weight of the primary device and the secondary device of each parameter pair, that is, the above sample target correlation coefficient.

[0118] In order to detect the status of the primary device and the secondary device in a targeted manner, in an optional implementation manner, the division unit includes:

[0119] A matrix forming module is used to form a correlation matrix according to all the correlation coefficients of all the parameter pairs in the primary device and the secondary device, wherein the rows of the correlation matrix represent the observed values ​​of the primary device, the columns of the correlation matrix represent the observed values ​​of the secondary device, and the elements of the correlation matrix represent the correlation coefficients;

[0120] A drawing module draws a heat map according to the association matrix, wherein each cell in the heat map represents the association coefficient of one of the parameter pairs, and the color depth in the heat map indicates the strength and direction of the association;

[0121] A first determination module determines the parameter pairs whose absolute values ​​of the correlation coefficients are greater than or equal to a first threshold as the strongly correlated parameters;

[0122] A second determination module determines the parameter pair whose absolute value of the correlation coefficient is greater than or equal to the second threshold and less than the first threshold as the weak correlation parameter;

[0123] The third determination module determines the parameter pair whose absolute value of the correlation coefficient is smaller than the second threshold as the uncorrelated parameter.

[0124] In the above embodiment, firstly, according to the correlation coefficient between the primary device and the secondary device, an association matrix is ​​formed, including the correlation coefficients between all parameter pairs, and a heat map is drawn according to the association matrix. Each cell in the heat map represents the correlation coefficient of a parameter pair, and the color depth indicates the strength and direction of the correlation, red indicates strong correlation, yellow indicates weak correlation, and white indicates no correlation. Then, the strong correlation parameter and strong correlation coefficient at the red position, the weak correlation parameter and weak correlation coefficient at the yellow position, and the unrelated parameter and irrelevant connection number at the white position are obtained respectively. By comprehensively analyzing the strong correlation coefficient, the weak correlation coefficient and the irrelevant connection number, the correlation coefficient of the strong correlation parameter is defined as an absolute value greater than 0.7, the correlation coefficient of the weak correlation parameter is defined as an absolute value between 0.3 and 0.7, and the correlation coefficient of the unrelated parameter is defined as an absolute value less than 0.3. According to the range of the relationship coefficient corresponding to the parameter pair, the parameter pair is determined as a specific correlation parameter. It is possible to detect the device status of the primary device and the secondary device in a targeted manner, so as to better detect the device status of the primary device and the secondary device according to the strength of the relationship between the primary device and the secondary device.

[0125] In order to perform exception handling in a timely manner, in an optional implementation manner, the verification unit includes:

[0126] A first determination module determines that the primary device is in a normal state when the first device state probability value is greater than or equal to a first state limit value;

[0127] A second determination module, when the probability value of the first equipment state is less than the first state limit value and greater than or equal to the second state limit value, issues a first prompt message, wherein the first prompt message is used to prompt to take preventive maintenance measures for the primary equipment;

[0128] A third determination module determines that the first device is in an abnormal state and the abnormal situation occurs when the state probability value of the first device is less than the second state limit value, and issues a second prompt message, wherein the second prompt message is used to prompt the primary device to handle the abnormal situation;

[0129] A fourth determination module determines that the secondary device is in a normal state when the second device state probability value is greater than or equal to the third state boundary value;

[0130] A fifth determination module, when the second device state probability value is less than the third state limit value and greater than or equal to the fourth state limit value, issues a third prompt message, wherein the third prompt message is used to prompt to take preventive maintenance measures for the secondary device;

[0131] The sixth determination module determines that the second device is in an abnormal state and the abnormal situation occurs when the second device state probability value is less than the fourth state limit value, and issues a fourth prompt message, wherein the fourth prompt message is used to prompt the secondary device to perform abnormal processing.

[0132] In the above embodiment, a sigmoid function is set at the input layer of the deep learning model, which is used as the activation function of the deep learning model, maps any input to the probability distribution, and converts the output of the deep learning model into a probability value. For example, the probability value output by the deep learning model ranges from 0.1 to 1. The closer the probability value output by the model is to 1, the higher the degree to which the device state is in a normal state. The closer the probability value output by the model is to 0.1, the higher the degree to which the device state is in an abnormal state, and measures should be taken immediately to rescue it. A state limit value or state limit range can be set for the probability value output by the model. The state limit value can be fixed, one or more. When the prediction result output by the model is not within the state limit range, it is considered that the current device state is in an abnormal state, and vigilance or maintenance is required. Alternatively, if the prediction result exceeds the state limit value, the device state is considered to be in a normal state, and if it is lower than the state limit value, the device state is considered to be in an abnormal state. The probability values ​​of the first deep learning model and the second deep learning model do not necessarily have to have the same value range, that is, the first state defining value and the third state defining value do not need to be the same, and the second state defining value and the fourth state defining value do not need to be the same, but the second state defining value and the fourth state defining value are usually both 1. The first state defining value is smaller than the second state defining value, and the second state defining value is smaller than the fourth state defining value.

[0133] In order to improve the accuracy of anomaly detection, in an optional implementation manner, the device further includes:

[0134] A first adjustment unit is used for adjusting the first state limit value and the second state limit value in real time according to historical data and real-time data corresponding to the primary device before verifying the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device have abnormal conditions;

[0135] The second adjustment unit is used to adjust the third state boundary value and the fourth state boundary value in real time according to the historical data and real-time data corresponding to the secondary device.

[0136] In the above embodiments, the state boundary value or state boundary range is not a fixed value, but is adjusted in real time according to the historical data and real-time data of the equipment operation. Multiple tests and adjustments are required to find the most suitable state boundary value or state boundary range setting, which can improve the accuracy of abnormality detection.

[0137] In order to prevent the failure from expanding or affecting production, in an optional implementation, the exception handling unit includes:

[0138] a first processing module, which, when the above-mentioned abnormal situation occurs in the above-mentioned primary device, determines a first abnormality repair strategy according to the above-mentioned abnormal situation, and performs abnormality repair processing on the above-mentioned primary device according to the above-mentioned first abnormality repair strategy;

[0139] The second processing module determines a second abnormality repair strategy according to the abnormal situation when the abnormal situation occurs in the secondary device, and performs abnormality repair processing on the secondary device according to the second abnormality repair strategy.

[0140] In the above embodiment, the fused data generated in real time in the substation and its specific values ​​are input into the trained deep learning model in the form of a data set for prediction, and the probability value of the equipment status of the substation is output. The equipment status includes the normal state of the equipment and the abnormal state of the equipment. When the abnormal state of the equipment is predicted, it means that the substation equipment may enter an abnormal state at a certain point in the future. By capturing certain abnormal conditions, the operator should pay attention to the current status of the substation equipment and verify whether there are any abnormal signs. Preventive maintenance measures can also be adopted to reduce the probability of equipment failure. If an abnormal condition occurs in the primary or secondary equipment, the operator should immediately determine the repair strategy according to the abnormal condition, and take measures to repair and process it according to the repair strategy to prevent the fault from expanding or affecting production.

[0141] The above-mentioned abnormality detection device based on substation equipment portrait includes a processor and a memory. The above-mentioned units are stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize the corresponding functions. The above-mentioned modules are all located in the same processor; or, the above-mentioned modules are located in different processors in any combination.

[0142] The processor includes a kernel, which calls the corresponding program unit from the memory. One or more kernels can be set, and the kernel parameters can be adjusted to solve the problem that the correlation between the primary device and the secondary device is not considered at the same time in the prior art, and the correlation factor between the devices is ignored to directly detect the device status, resulting in low detection accuracy.

[0143] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0144] An embodiment of the present invention provides a computer-readable storage medium, which includes a stored program, wherein when the program is running, the device where the computer-readable storage medium is located is controlled to execute the above-mentioned anomaly detection method based on substation equipment profiling.

[0145] Specifically, the anomaly detection method based on substation equipment portrait includes:

[0146] Step S201, acquiring first device data and second device data in real time, wherein the first device data is real-time data of a primary device in a substation at the current moment, and the second device data is operating status data of a secondary device in the substation at the current moment and related information with the corresponding primary device;

[0147] Step S202, inputting the first device data into the first deep learning model, so as to use the first deep learning model to perform abnormality detection on the first device data of the primary device to obtain a first device state probability value, wherein the first deep learning model is obtained by training with a plurality of groups of first training data through a recurrent neural network algorithm, and each group of the first training data includes the first sample device data of the primary device and the corresponding second sample state probability value of the primary device, and the first sample device data at least includes the device operation parameters, environmental information, normal state information of the device, abnormal state information of the device, interval duration of abnormality of the device and device failure record of the primary device collected within the same interval, and the first device state probability value is the probability value of the device state of the primary device;

[0148] Step S203, inputting the second device data into the second deep learning model, so as to use the second deep learning model to perform abnormal detection of the secondary device on the second device data, and obtain a second device state probability value, wherein the second deep learning model is obtained by training with a plurality of groups of second training data through a long short-term memory network algorithm, and each group of the second training data includes the second sample device data of the secondary device and the corresponding first sample state probability value, and the second sample device data at least includes a sample target label, a sample target correlation coefficient and the sample operation data of the second device, and the target label is the normal state information and abnormal state information of the device obtained by the primary device, and the sample target correlation coefficient is used for the degree of association between the primary device and the corresponding secondary device, and the second device state probability value is the probability value of the device state of the secondary device;

[0149] Step S204, verifying the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device are abnormal;

[0150] Step S205: When the above-mentioned abnormal situation occurs in the above-mentioned primary device or the above-mentioned secondary device, abnormal repair processing is performed on the above-mentioned primary device or the above-mentioned secondary device.

[0151] An embodiment of the present invention provides a processor, which is used to run a program, wherein the above-mentioned anomaly detection method based on substation equipment profiling is executed when the above-mentioned program is run.

[0152] Specifically, the anomaly detection method based on substation equipment portrait includes:

[0153] Step S201, acquiring first device data and second device data in real time, wherein the first device data is real-time data of a primary device in a substation at the current moment, and the second device data is operating status data of a secondary device in the substation at the current moment and related information with the corresponding primary device;

[0154] Step S202, inputting the first device data into the first deep learning model, so as to use the first deep learning model to perform abnormality detection on the first device data of the primary device to obtain a first device state probability value, wherein the first deep learning model is obtained by training with a plurality of groups of first training data through a recurrent neural network algorithm, and each group of the first training data includes the first sample device data of the primary device and the corresponding second sample state probability value of the primary device, and the first sample device data at least includes the device operation parameters, environmental information, normal state information of the device, abnormal state information of the device, interval duration of abnormality of the device and device failure record of the primary device collected within the same interval, and the first device state probability value is the probability value of the device state of the primary device;

[0155] Step S203, inputting the second device data into the second deep learning model, so as to use the second deep learning model to perform abnormal detection of the secondary device on the second device data, and obtain a second device state probability value, wherein the second deep learning model is obtained by training with a plurality of groups of second training data through a long short-term memory network algorithm, and each group of the second training data includes the second sample device data of the secondary device and the corresponding first sample state probability value, and the second sample device data at least includes a sample target label, a sample target correlation coefficient and the sample operation data of the second device, and the target label is the normal state information and abnormal state information of the device obtained by the primary device, and the sample target correlation coefficient is used for the degree of association between the primary device and the corresponding secondary device, and the second device state probability value is the probability value of the device state of the secondary device;

[0156] Step S204, verifying the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device are abnormal;

[0157] Step S205: When the above-mentioned abnormal situation occurs in the above-mentioned primary device or the above-mentioned secondary device, abnormal repair processing is performed on the above-mentioned primary device or the above-mentioned secondary device.

[0158] An embodiment of the present invention provides a substation equipment image anomaly detection system, the substation equipment image anomaly detection system includes a processor, a memory, and a program stored in the memory and executable on the processor, and when the processor executes the program, at least the following steps are implemented:

[0159] Step S201, acquiring first device data and second device data in real time, wherein the first device data is real-time data of a primary device in a substation at the current moment, and the second device data is operating status data of a secondary device in the substation at the current moment and related information with the corresponding primary device;

[0160] Step S202, inputting the first device data into the first deep learning model, so as to use the first deep learning model to perform abnormality detection on the first device data of the primary device to obtain a first device state probability value, wherein the first deep learning model is obtained by training with a plurality of groups of first training data through a recurrent neural network algorithm, and each group of the first training data includes the first sample device data of the primary device and the corresponding second sample state probability value of the primary device, and the first sample device data at least includes the device operation parameters, environmental information, normal state information of the device, abnormal state information of the device, interval duration of abnormality of the device and device failure record of the primary device collected within the same interval, and the first device state probability value is the probability value of the device state of the primary device;

[0161] Step S203, inputting the second device data into the second deep learning model, so as to use the second deep learning model to perform abnormal detection of the secondary device on the second device data, and obtain a second device state probability value, wherein the second deep learning model is obtained by training with a plurality of groups of second training data through a long short-term memory network algorithm, and each group of the second training data includes the second sample device data of the secondary device and the corresponding first sample state probability value, and the second sample device data at least includes a sample target label, a sample target correlation coefficient and the sample operation data of the second device, and the target label is the normal state information and abnormal state information of the device obtained by the primary device, and the sample target correlation coefficient is used for the degree of association between the primary device and the corresponding secondary device, and the second device state probability value is the probability value of the device state of the secondary device;

[0162] Step S204, verifying the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device are abnormal;

[0163] Step S205: When the above-mentioned abnormal situation occurs in the above-mentioned primary device or the above-mentioned secondary device, abnormal repair processing is performed on the above-mentioned primary device or the above-mentioned secondary device.

[0164] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program for initializing at least the following method steps:

[0165] Step S201, acquiring first device data and second device data in real time, wherein the first device data is real-time data of a primary device in a substation at the current moment, and the second device data is operating status data of a secondary device in the substation at the current moment and related information with the corresponding primary device;

[0166] Step S202, inputting the first device data into the first deep learning model, so as to use the first deep learning model to perform abnormality detection on the first device data of the primary device to obtain a first device state probability value, wherein the first deep learning model is obtained by training with a plurality of groups of first training data through a recurrent neural network algorithm, and each group of the first training data includes the first sample device data of the primary device and the corresponding second sample state probability value of the primary device, and the first sample device data at least includes the device operation parameters, environmental information, normal state information of the device, abnormal state information of the device, interval duration of abnormality of the device and device failure record of the primary device collected within the same interval, and the first device state probability value is the probability value of the device state of the primary device;

[0167] Step S203, inputting the second device data into the second deep learning model, so as to use the second deep learning model to perform abnormal detection of the secondary device on the second device data, and obtain a second device state probability value, wherein the second deep learning model is obtained by training with a plurality of groups of second training data through a long short-term memory network algorithm, and each group of the second training data includes the second sample device data of the secondary device and the corresponding first sample state probability value, and the second sample device data at least includes a sample target label, a sample target correlation coefficient and the sample operation data of the second device, and the target label is the normal state information and abnormal state information of the device obtained by the primary device, and the sample target correlation coefficient is used for the degree of association between the primary device and the corresponding secondary device, and the second device state probability value is the probability value of the device state of the secondary device;

[0168] Step S204, verifying the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device are abnormal;

[0169] Step S205: When the above-mentioned abnormal situation occurs in the above-mentioned primary device or the above-mentioned secondary device, abnormal repair processing is performed on the above-mentioned primary device or the above-mentioned secondary device.

[0170] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0171] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0172] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0173] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0174] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0175] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0176] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0177] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0178] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity 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, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0179] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0180] 1) The anomaly detection method based on substation equipment portrait of the present application constructs a first deep learning model and a second deep learning model. The first deep learning model predicts the primary equipment and outputs the probability value of the equipment state of the primary equipment. The second deep learning model predicts the equipment state of the secondary equipment and outputs the probability value of the equipment state of the secondary equipment. The equipment state is judged according to the probability value. In the training process of the second deep learning model, the parameters related to the secondary equipment and the equipment state and the correlation coefficient of the secondary equipment and the primary equipment in the equipment operation state are comprehensively considered as the second training data, so that the model training is performed using the final correlation coefficient to obtain the trained second deep learning model, which will be more accurate in equipment state detection and improve the detection accuracy. The present application solves the problem that the correlation between the primary equipment and the secondary equipment is not considered at the same time in the prior art, and the correlation factors between the equipment are ignored to directly detect the equipment state, resulting in low detection accuracy.

[0181] 2) The abnormality detection device based on substation equipment portrait of the present application constructs a first deep learning model and a second deep learning model. The first deep learning model predicts the primary equipment and outputs the probability value of the equipment state of the primary equipment. The second deep learning model predicts the equipment state of the secondary equipment and outputs the probability value of the equipment state of the secondary equipment. The equipment state is judged according to the probability value. In the training process of the second deep learning model, the parameters related to the secondary equipment and the equipment state and the correlation coefficient with correlation between the secondary equipment and the primary equipment in the equipment operation state are comprehensively considered as the second training data, so that the model training is performed using the final correlation coefficient to obtain the trained second deep learning model, which will be more accurate in equipment state detection and improve the detection accuracy. The present application solves the problem that the correlation between the primary equipment and the secondary equipment is not considered at the same time in the prior art, and the correlation factors between the equipment are ignored to directly detect the equipment state, resulting in low detection accuracy.

[0182] The above description is only the preferred embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An abnormality detection method based on substation equipment profiling, It is characterized in that include: Acquire first device data and second device data in real time, wherein the first device data is real-time data of a primary device in the substation at the current moment, and the second device data is operation status data of a secondary device in the substation at the current moment and related information with the corresponding primary device; The first device data is input into a first deep learning model, so as to use the first deep learning model to perform abnormality detection of the primary device on the first device data to obtain a first device state probability value, wherein the first deep learning model is obtained by training with a plurality of groups of first training data through a recurrent neural network algorithm, wherein each group of the first training data includes first sample device data of the primary device and a corresponding second sample state probability value of the primary device, wherein the first sample device data includes at least device operation parameters, environmental information, device normal state information, device abnormal state information, device abnormality interval duration and device failure record of the primary device collected within the same interval, and the first device state probability value is a probability value of the device state of the primary device; The second device data is input into a second deep learning model, so as to use the second deep learning model to perform abnormality detection on the second device data for the secondary device, and obtain a second device state probability value, wherein the second deep learning model is obtained by training with a plurality of groups of second training data through a long short-term memory network algorithm, each group of the second training data includes second sample device data of the secondary device and a corresponding first sample state probability value, the second sample device data includes at least a sample target label, a sample target correlation coefficient and sample operation data of the second device, the target label is the normal state information and abnormal state information of the device acquired by the primary device, the sample target correlation coefficient is used for the degree of association between the primary device and the corresponding secondary device, and the second device state probability value is the probability value of the device state of the secondary device; Verifying the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device have abnormal conditions; When the abnormal situation occurs in the primary device or the secondary device, abnormal repair processing is performed on the primary device or the secondary device.

2. The method according to claim 1, It is characterized in that Before acquiring the first device data and the second device data in real time, the method further includes: Obtaining historical data of the substation, the historical data at least including equipment information data, equipment operation data, event record data and maintenance management data, the equipment information data at least including the equipment name, equipment type, equipment rated voltage, current, frequency, power and control parameters of each equipment in the substation, the equipment operation data at least including various parameter data of each equipment in the power transmission process and the discharge process and the equipment status information of the equipment at each time point in the historical time period, the event record data at least including the accident type, accident time and various parameter data when the equipment fails or is abnormal, the maintenance management data includes the record data of regular equipment inspection, maintenance and calibration, the historical record data of equipment failure or abnormality, cause analysis and treatment results and equipment repair and replacement parts information; Filtering out state-related data from the historical data, wherein the state-related data is data related to the state of the primary device and the state of the secondary device; A preprocessing operation is performed on the state-related data to obtain a training set, where the training set includes first training data and the second training data, and the preprocessing operation at least includes data cleaning.

3. The method according to claim 2, It is characterized in that Before inputting the second device data into a second deep learning model to perform abnormality detection of the secondary device on the second device data using the second deep learning model to obtain a second device state probability value of the secondary device, the method further includes: The correlation coefficient of each parameter pair in the primary device and the secondary device is calculated according to a first formula, wherein the first formula is: r i is the correlation coefficient of the i-th parameter pair, n is the total number of parameter pairs, i=(1, 2, 3, ..., n), x i is the observed value of the primary device in the ith parameter pair, y i is the observed value of the secondary device in the i-th parameter pair, wherein the observed value is a data value randomly selected from the historical data, is the average value of the observed values ​​of the primary equipment in all the parameter pairs, is the average value of the observed values ​​of the secondary device in all the parameter pairs, wherein the parameter pair represents a combination of the observed values ​​corresponding to the primary device and the secondary device; According to the correlation coefficient of each parameter pair, each parameter pair is divided into corresponding correlation parameters, and the correlation parameters are divided into strong correlation parameters, weak correlation parameters and no correlation parameters; Setting corresponding weights for the strongly associated parameter, the weakly associated parameter and the unassociated parameter; The sample target correlation coefficient between the primary device and the secondary device is calculated according to a second formula, where the second formula is: R is the sample target correlation coefficient, ω i is the weight of the associated parameter to which the i-th parameter pair belongs.

4. The method according to claim 3, It is characterized in that Dividing each parameter pair into corresponding associated parameters according to the associated coefficient of each parameter pair includes: forming an association matrix according to all the association coefficients of all the parameter pairs in the primary device and the secondary device, wherein the rows of the association matrix represent the observed values ​​of the primary device, the columns of the association matrix represent the observed values ​​of the secondary device, and the elements of the association matrix represent the association coefficients; Draw a heat map according to the association matrix, wherein each cell in the heat map represents the association coefficient of a parameter pair, and the color depth in the heat map indicates the strength and direction of the association; Determine the parameter pair whose absolute value of the correlation coefficient is greater than or equal to a first threshold as the strongly correlated parameter; Determine the parameter pair whose absolute value of the correlation coefficient is greater than or equal to the second threshold and less than the first threshold as the weak correlation parameter; The parameter pairs whose absolute values ​​of the correlation coefficients are smaller than the second threshold are determined as the uncorrelated parameters.

5. The method according to claim 1, It is characterized in that Verifying the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device have abnormal conditions, including: When the first device state probability value is greater than or equal to the first state limit value, determining that the primary device is in a normal state; When the first equipment state probability value is less than the first state limit value and greater than or equal to the second state limit value, issuing a first prompt message, wherein the first prompt message is used to prompt to take preventive maintenance measures for the primary equipment; When the first device state probability value is less than the second state limit value, it is determined that the first device is in an abnormal state and the abnormal situation occurs, and a second prompt message is issued, wherein the second prompt message is used to prompt the primary device to handle the abnormal situation; When the second device state probability value is greater than or equal to the third state limit value, determining that the secondary device is in a normal state; When the second equipment state probability value is less than the third state limit value and greater than or equal to the fourth state limit value, a third prompt message is issued, wherein the third prompt message is used to prompt to take preventive maintenance measures for the secondary equipment; When the second device state probability value is less than the fourth state limit value, it is determined that the second device is in an abnormal state and the abnormal situation occurs, and a fourth prompt message is issued, wherein the fourth prompt message is used to prompt the secondary device to perform abnormal processing.

6. The method according to claim 5, It is characterized in that Before verifying the primary device and the secondary device according to the first device state probability value and the second device state probability value to determine whether the primary device and the secondary device have abnormal conditions, the method further includes: Adjusting the first state limit value and the second state limit value in real time according to the historical data and real-time data corresponding to the primary device; The third state boundary value and the fourth state boundary value are adjusted in real time according to the historical data and real-time data corresponding to the secondary device.

7. The method according to claim 1, It is characterized in that When the abnormal situation occurs in the primary device or the secondary device, performing abnormal repair processing on the primary device or the secondary device includes: In the case where the abnormal situation occurs in the primary device, determining a first abnormality repair strategy according to the abnormal situation, and performing abnormality repair processing on the primary device according to the first abnormality repair strategy; When the abnormal situation occurs in the secondary device, a second abnormality repair strategy is determined according to the abnormal situation, and abnormality repair processing is performed on the secondary device according to the second abnormality repair strategy.

8. An abnormality detection device based on substation equipment portrait, It is characterized in that The device comprises: A first acquisition unit is used to acquire first device data and second device data in real time, wherein the first device data is real-time data of a primary device in the substation at the current moment, and the second device data is operation status data of a secondary device in the substation at the current moment and related information with the corresponding primary device; A first input-output unit is used to input the first device data into a first deep learning model, so as to use the first deep learning model to perform abnormality detection on the first device data of the primary device to obtain a first device state probability value, wherein the first deep learning model is obtained by training with a recurrent neural network algorithm using multiple groups of first training data, each group of the first training data includes first sample device data of the primary device and a corresponding second sample state probability value of the primary device, the first sample device data at least includes device operating parameters, environmental information, device normal state information, device abnormal state information, device abnormality interval duration and device failure record of the primary device collected within the same interval, and the first device state probability value is a probability value of the device state of the primary device; a second input-output unit, configured to input the second device data into a second deep learning model, so as to use the second deep learning model to perform abnormality detection on the secondary device on the second device data, and obtain a second device state probability value; the second deep learning model is obtained by training with a plurality of groups of second training data through a long short-term memory network algorithm; each group of the second training data includes second sample device data of the secondary device and a corresponding first sample state probability value; the second sample device data includes at least a sample target label, a sample target correlation coefficient, and sample operation data of the second device; the target label is the normal state information and abnormal state information of the device acquired by the primary device; the sample target correlation coefficient is used for the degree of association between the primary device and the corresponding secondary device; and the second device state probability value is the probability value of the device state of the secondary device; a verification unit, configured to verify the primary device and the secondary device according to the first device state probability value and the second device state probability value, so as to determine whether an abnormality occurs in the primary device and the secondary device; The abnormality processing unit is used to perform abnormality repair processing on the primary device or the secondary device when the abnormal situation occurs in the primary device or the secondary device.

9. A computer-readable storage medium, It is characterized in that The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

10. A computer program product comprising computer instructions, It is characterized in that When the computer instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Citation Information

Patent Citations

  • Anomaly detection method and device based on electrical equipment portrait, and storage medium

    CN117033913A

  • Transformer area data exception discrimination method and device based on cross validation

    CN110807014A

  • Big data analysis method and system applied to substation inspection

    CN118467975A

  • Electrical grid anomaly detection, classification, and prediction

    WO2024118550A1