Urban underground power supply pipe gallery abnormal event prediction method and electronic equipment

The method and system for predicting abnormal events in urban underground power supply corridors use sensor nodes and federated learning to enhance real-time detection and prediction, addressing the inefficiencies of existing systems and improving response times and accuracy.

CN119067447APending Publication Date: 2024-12-03GUANGZHOU ZHONGKE RONGDA TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411177506.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-26
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Existing power corridor monitoring systems lack efficient real-time detection and prediction mechanisms for abnormal events, leading to delayed responses and increased accident risks in urban underground power supply corridors.

Method used

A method and system utilizing sensor nodes with linear regression training, federated learning, and multiple predictive models to detect and predict abnormal events in real-time, ensuring data privacy and improving prediction accuracy through model aggregation and feature extraction.

Benefits of technology

Enables rapid detection and early warning of potential hazards, reducing accident risks by enhancing response speed and accuracy, while protecting user privacy and improving data processing efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067447A_ABST
    Figure CN119067447A_ABST
Patent Text Reader

Abstract

The invention provides an urban underground power supply pipe gallery abnormal event prediction method and electronic equipment, and belongs to the technical field of artificial intelligence. Comprising the following steps: acquiring original data by various sensing nodes deployed in a pipe gallery; processing the original data through a linear regression data processing module to obtain a first model parameter; the first model parameters are uploaded to a central server to be aggregated, and second model parameters are generated; issuing the second model parameters to various sensing nodes, and updating the second model parameters to obtain third model parameters; the central server extracts basic features in the third model parameters and calculates derivative features, the basic features and the derivative features are combined to obtain a first data set, and the central server carries out training prediction on the first data set to obtain a first prediction result; and performing regression prediction training on the first prediction result to obtain a second prediction result. According to the method, the accuracy of predicting the abnormal events in the pipe gallery is high, the information interaction safety is high, and the risk of accidents is reduced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a method and electronic equipment for predicting abnormal events in an urban underground power supply corridor. Background Art

[0002] Power supply corridors, as the lifeline of underground power transmission in modern cities, bear the important responsibility of ensuring the continuity and stability of urban power supply. They are an indispensable part of urban infrastructure and are directly related to the convenience of residents' daily lives and the normal operation of the social economy. In recent years, with the rapid development of technologies such as the Internet of Things, big data, and artificial intelligence, the construction of smart cities has become an important direction for improving the efficiency of urban management.

[0003] As an important part of smart cities, smart power corridors are designed to achieve comprehensive perception, real-time monitoring and intelligent early warning of the internal environment of power corridors by integrating advanced sensing technology, network communication technology and data analysis technology. Existing corridor monitoring systems often lack efficient real-time detection and prediction mechanisms for abnormal events, making it difficult to respond to emergencies in a timely manner, increasing the risk of accidents. Summary of the invention

[0004] The purpose of this application is to provide a method and electronic equipment for predicting abnormal events in an urban underground power supply corridor, so as to realize real-time detection and early warning of abnormal events in the corridor through intelligent means.

[0005] In order to achieve the above purpose, the technical solution adopted in the embodiment of the present application is as follows:

[0006] In a first aspect, an embodiment of the present application provides a method for predicting abnormal events in an urban underground power supply corridor, which is applied to a system for predicting abnormal events in an urban underground power supply corridor. The system includes a central server, and the method includes the following steps:

[0007] Obtain the original data of target attributes through various sensor nodes deployed in the tunnel;

[0008] The raw data is trained through a linear regression data processing module set by each type of sensor node to obtain a first model parameter;

[0009] Uploading the first model parameters to the central server through the wireless communication modules provided inside the various sensor nodes, and the central server aggregates the first model parameters to generate second model parameters;

[0010] The central server sends the second model parameters to the various sensor nodes, and the various sensor nodes obtain third model parameters after updating;

[0011] The third model parameters are uploaded to the central server via the wireless communication module, and the central server extracts basic features from the third model parameters;

[0012] The central server calculates the derived features of the basic features;

[0013] The basic features are combined with the derived features to obtain a first data set, and the central server performs training prediction on the first data set to obtain a first prediction result;

[0014] The central server performs regression prediction training on the first prediction result to obtain a second prediction result, and the second prediction result is the final prediction result.

[0015] Furthermore, the linear regression data processing module only processes the raw data collected by the corresponding sensor nodes.

[0016] Furthermore, the second model parameter aggregation formula is:

[0017]

[0018] Among them, W is the second model parameter, k is the type of sensor node, N is the number of sensor nodes participating in the training, n k is the number of samples of the kth sensor node, w k is the model parameter of the kth sensor node.

[0019] Furthermore, the central server aggregates the first model parameters to generate second model parameters, including:

[0020] The central server aggregates the first model parameters using an adaptive learning rate adjustment mechanism to generate the second model parameters.

[0021] Furthermore, the central server calculates the derived features of the basic features, including:

[0022] The central server calculates the sliding average value of the past period of time using a sliding average calculation formula; wherein the sliding average calculation formula is:

[0023]

[0024] In the formula, SMA t is the sliding average at time t, x i is the observed value at time i, and n is the number of sensor nodes.

[0025] Furthermore,

[0026] The central server calculates the derived features of the basic features, including:

[0027] The central server calculates the exponential smoothing value of the past period of time using an exponential average calculation formula; wherein the exponential average calculation formula is:

[0028] S t =ax t +(1-a)S t-1

[0029] In the formula, S t is the exponential smoothing value at time t, x t is the observed value at time t, S t-1 is the exponential smoothing value at time t-1, a is the smoothing constant (0 <a<1)。

[0030] Furthermore, the central server calculates the derived features of the basic features, including:

[0031] The central server calculates the standard deviation of each basic feature in the past period of time using a standard deviation calculation formula; wherein the standard deviation calculation formula is:

[0032]

[0033] Where σ is the standard deviation, N is the number of sensor nodes, and x i is the observed value at time i, and u is the mean of the observed values ​​at multiple time points.

[0034] Furthermore, the central server includes a basic prediction model module having one or more combinations of random forest (RF), gradient boosted decision tree (GBDT) and support vector machine (SVM), and the basic model calculation module is used to perform prediction training on the first data set to obtain one or more first prediction results.

[0035] Furthermore, the central server includes a logistic regression (LR) model module, and one or more first prediction results are input into the logistic regression (LR) model module as new features to obtain one or more second prediction results.

[0036] In a second aspect, an embodiment of the present application further provides an electronic device, including:

[0037] a memory for storing one or more programs; a processor;

[0038] When the one or more programs are executed by the processor, any of the above-mentioned methods for predicting abnormal events in urban underground power supply corridors is implemented.

[0039] Compared with the prior art, this application has the following beneficial effects:

[0040] Through the real-time collection and fusion of various sensor data, the system can realize the instant detection and early warning of abnormal events in the pipe gallery, effectively improving the response speed and reducing the risk of accidents. The system can timely discover and deal with potential safety hazards, such as overheating of cable equipment and abnormal humidity, effectively avoiding safety accidents caused by cable equipment failure or environmental factors. At the same time, the system's rapid response mechanism can quickly locate faults and take measures to ensure the continuity and stability of power supply. Using the federated learning algorithm, each sensor node trains the model locally and only uploads the model parameters to the central server for aggregation, which not only protects user privacy but also improves data processing efficiency. Random forests, gradient boosting decision trees and support vector machines are used as base models, and the prediction results of these base models are combined with the logistic regression model to optimize the prediction results and improve the accuracy and reliability of early warnings.

[0041] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, preferred embodiments are specifically cited below and described in detail with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 A schematic diagram of the structure of an abnormal event prediction system for an urban underground power supply corridor provided in an embodiment of the present application.

[0044] Figure 2 A schematic diagram of a module of an electronic device provided in an embodiment of the present application.

[0045] Figure 3 A schematic diagram of a method for predicting abnormal events in an urban underground power supply corridor provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, 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 part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations.

[0047] Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application for which protection is sought, but merely represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0048] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0049] It should be noted that, in this document, relational terms such as first and second, etc. are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0050] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0051] As mentioned in the background technology, currently, comprehensive perception, real-time monitoring and intelligent early warning of the internal environment of the power corridor are achieved by integrating advanced sensing technology, network communication technology and data analysis technology. Existing power corridor monitoring systems often lack efficient abnormal event detection and early warning mechanisms, making it difficult to respond to emergencies in a timely manner, increasing the risk of accidents.

[0052] In view of this, in order to solve the above problems, an embodiment of the present application provides a method and electronic equipment for predicting abnormal events in an urban underground power supply corridor, so as to perform real-time detection and early warning of abnormal events in the corridor through intelligent means.

[0053] The following is an exemplary description of the method for predicting abnormal events in urban underground power supply corridors provided in this application.

[0054] It should be noted that the method for predicting abnormal events in urban underground power supply corridors provided in this application can be applied to electronic devices. Figure 2 A schematic structural block diagram of an electronic device 100 provided in an embodiment of the present application is shown, and the electronic device 100 includes a memory 102, a processor 101, and a communication bus 103. The memory 102, the processor 101, and the communication bus 103 are electrically connected to each other directly or indirectly to achieve data transmission or interaction. For example, these elements can be electrically connected to each other through one or more communication buses or signal lines.

[0055] The memory 102 can be used to store software programs and modules, such as program instructions or modules corresponding to the method for predicting abnormal events in an underground power supply corridor in a city provided in an embodiment of the present application. The processor 101 executes various functional applications and data processing by executing the software programs and modules stored in the memory 102, thereby executing the method for predicting abnormal events in an underground power supply corridor in a city provided in an embodiment of the present application. The communication bus 103 can be used to communicate signaling or data with other node devices.

[0056] Among them, the memory 102 can be, but is not limited to, random access memory (Random Access Memory, RAM), read only memory (Read Only Memory, ROM), programmable read-only memory (Programmable Read-Only Memory, PROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, EPROM), electrically erasable programmable read-only memory (Electric Erasable Programmable Read-Only Memory, EEPROM), etc.

[0057] The processor 101 may be an integrated circuit chip with signal processing capability. The processor 101 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0058] Understandably, Figure 2 The structure shown is for illustration only. The electronic device 100 may also include Figure 2 More or fewer components as shown, or with Figure 2 Different configurations are shown. Figure 2 Each component shown in the figure can be implemented by hardware, software or a combination thereof.

[0059] The following uses the electronic device 100 as an illustrative execution entity to exemplarily illustrate the Internet of Things-based power supply corridor management system provided in the embodiment of the present application.

[0060] Each electronic device 100 contains a central server, which is mainly used for calculation, training and information transmission of various data and model parameters.

[0061] Furthermore, the sensor node mainly includes a sensor, a linear regression operation module, a power supply module, and a communication module; the sensor, the linear regression operation module, the power supply module, and the communication module are electrically connected.

[0062] Furthermore, various types of sensor nodes are deployed at different locations in the tunnel. The sensors in the sensor nodes include temperature sensors, humidity sensors, displacement sensors, smoke sensors, current transformers, and light sensors. These sensors can monitor various environmental parameters and physical conditions inside the power tunnel in real time. For some key areas or important parameters (such as temperature, humidity, smoke concentration, etc.), a redundant sensor design is adopted, that is, multiple sensor nodes of the same type are deployed at the same location or nearby areas. This can improve the reliability and fault tolerance of the data. Even if a sensor fails or the data is abnormal, accurate raw data can be obtained through other sensor nodes.

[0063] Furthermore, the sensor nodes collect data on target attributes in real time through wireless communication modules (such as 4G / 5G, GPRS, etc.) and transmit the data to the central server.

[0064] See also Figure 1 and Figure 3 The method for predicting abnormal events in urban underground power supply corridors provided in this application includes:

[0065] Step 1: Deploy various types of sensor nodes at different locations in the city’s underground power supply corridors to obtain real-time raw data of target attributes, such as temperature, humidity, gas concentration, displacement change, smoke concentration, and light intensity, and transmit the data to the central server.

[0066] Step 2: Each sensor node uses its own linear regression operation module to perform linear regression training on the collected raw data to obtain the first model parameters.

[0067] Furthermore, during the training process, each sensor node only processes local data and does not involve data exchange, ensuring user privacy and data security. After the training is completed, the first model parameters of each sensor node are uploaded to the central server.

[0068] It can be understood that the first model parameters are the model parameters after linear regression of the data collected by each sensor node itself. In the process of generating the first model parameters, each sensor node only uses local data for model training, and does not involve data exchange, thereby ensuring user privacy and data security. By training the first model parameters, the generalization ability is improved for subsequent aggregation. In addition, since the linear regression training is completed on each sensor node, the computing burden of the central server is effectively reduced, and the overall learning efficiency and speed are improved.

[0069] Step 3: On the central server, the FedAvg federated learning module aggregates the model parameters of each node to generate global model parameters, i.e., the second model parameters. By weighted averaging, the local knowledge of each node is integrated into global knowledge, which improves the generalization ability and prediction accuracy of the model.

[0070] It can be understood that the second model parameters are formed after the first model parameters are uploaded to the central server and aggregated by the FedAvg federated learning module. The second model parameter aggregation formula is:

[0071]

[0072] Among them, W is the second model parameter, k is the various sensor nodes, N is the number of sensor nodes participating in the training, n k is the number of samples of the kth sensor node, w k is the model parameter of the kth sensor node.

[0073] It can be understood that when the FedAvg federated learning module aggregates the first model parameters, its essential core is to perform weighted averaging on the first model parameters to obtain a global model parameter, which is the second model parameter.

[0074] Furthermore, the FedAvg federated learning module also includes an adaptive learning rate adjustment mechanism. For example, optimization algorithms such as Adam and RMSprop can be used to dynamically adjust the learning rate based on historical gradient information, thereby improving the efficiency and stability of FedAvg federated learning model training.

[0075] In step 4, the aggregated second model parameters are then distributed to each sensor node, and each node uses the global model parameters to update its own model to obtain the third model parameters. Distributing them to each sensor node not only ensures data privacy (because the data itself does not leave the local node), but also realizes cross-node sharing of model knowledge.

[0076] It is understandable that by aggregating the local models of multiple nodes, the FedAvg algorithm can integrate data features from different regions and environments to generate a more comprehensive and more generalized global model. This helps to achieve more accurate abnormal event detection in various areas of the tunnel. The distribution of global model parameters not only updates the local models of each node, but also provides a basis for continuous optimization and iteration of the model. Through continuous federated learning iterations, the performance of the global model can be continuously improved, so as to better adapt to the dynamic changes of the tunnel environment.

[0077] Step 5: The third model parameters are uploaded to the central server via the wireless communication module, and the central server extracts basic features from the third model parameters.

[0078] It can be understood that the third model parameters include basic features such as temperature, humidity, smoke concentration, etc. at a certain moment. The central server uses common feature extraction tools to extract the basic features in the third model parameters. The extracted information includes the temperature value, humidity value, smoke concentration value, etc. at a certain moment.

[0079] Step 6: The central server calculates the derived features of the basic features.

[0080] It is understandable that the following time interval sequence of temperature data (unit: °C) is obtained from a sensor node: T = {25, 26, 25, 27, 28}. These temperature values ​​are the basic characteristics, which directly reflect the temperature changes in the power pipeline corridor. In order to calculate the rate of change of temperature over time, it is necessary to calculate the temperature change rate, and the calculation formula is:

[0081]

[0082] Assuming the time interval is 1 hour, the temperature change rate is:

[0083]

[0084]

[0085] Since there is no temperature data before the first moment, the temperature change rate 1 cannot be calculated and can be directly represented by 0. Then the temperature change rate for this hour is 0, 1, -1, 2, 1. This well reflects the change in the temperature inside the power pipeline corridor and lays the foundation for subsequent predictions.

[0086] In order to further improve the stability and reliability of the data, it is necessary to smooth the basic features and reduce the impact of noise.

[0087] Furthermore, the central server calculates the derived features of the basic features by calculating the sliding average over the past period of time, and the formula is:

[0088]

[0089] Among them, SMA t is the sliding average at time t, x i is the observed value at time i, and n is the number of sensor nodes.

[0090] Furthermore, the central server calculates the derived features of the basic features by calculating the exponential smoothing value over a period of time in the past, and the formula is:

[0091] S t =ax t +(1-a)S t-1

[0092] Among them, S t is the exponential smoothing value at time t, x t is the observed value at time t, S t-1 is the exponential smoothing value at time t-1, a is the smoothing constant (0 <a<1)。

[0093] Furthermore, the central server calculates the standard deviation of each of the basic features over the past period of time, and the standard deviation calculation formula is:

[0094]

[0095] Where σ is the standard deviation, N is the number of sensor nodes, and x i is the observed value at time i, and u is the mean of the observed values ​​at multiple time points.

[0096] It is understandable that by performing derivative calculations on the basic features, the speed and magnitude of the change in the internal environment of the power pipeline corridor can be obtained. The basic features and derived features together constitute the basis for model prediction, improving the richness of data and the accuracy of prediction. Through calculations such as sliding average and exponential average, the data is effectively smoothed and the impact of noise on the prediction results is reduced. This helps the model to more accurately capture the true trend of the data and improve the stability and reliability of the prediction.

[0097] Step 7, combining the basic features with the derived features to obtain a first data set, and after the first data set is preprocessed, the central server performs training prediction on the first data set to obtain a first prediction result.

[0098] Furthermore, the central server includes a basic prediction model module having one or more combinations of random forest (RF), gradient boosted decision tree (GBDT) and support vector machine (SVM), and the basic model calculation module is used to perform prediction training on the first data set to obtain one or more first prediction results.

[0099] It is understandable that random forest (RF) can improve the stability and predictive ability of the model by constructing multiple decision trees and randomly selecting features and data sets. Gradient boosted decision tree (GBDT) can continuously improve the predictive ability of the model by gradually building decision trees and learning based on the residuals of the previous round. Support vector machine (SVM) can find the optimal classification hyperplane in high-dimensional space, which is suitable for handling classification problems and can also be used for regression. The three basic models of random forest, gradient boosted decision tree and support vector machine are trained using the prepared first data set to produce three independent prediction results. The prediction result generated by the basic model is called the first prediction result.

[0100] It can be understood that the central server selects one or more combinations of appropriate basic models depending on the actual situation inside the power pipeline corridor, which will then generate one or more first prediction results.

[0101] Furthermore, it is best to have multiple basic models in the central server. When facing the complex and changeable underground power supply corridor environment, a single base model may lead to unstable prediction results due to the limitations of the model and the uncertainty of the data. By combining multiple base models, the uncertainty of the prediction results of a single model can be effectively reduced and the stability of the prediction can be improved. In addition, a single base model is overfitted due to excessive complexity or limited training data, resulting in poor generalization ability on unknown data. By introducing multiple basic models, the risk of overfitting can be effectively reduced and the generalization ability and robustness of the model can be improved.

[0102] Step 8: The central server performs regression prediction training on the first prediction result to obtain a second prediction result, which is the final prediction result.

[0103] It is understandable that the central server includes a logistic regression (LR) model module, and one or more first prediction results are respectively input as new features into the logistic regression (LR) model module to obtain one or more second prediction results. The second prediction result is the final prediction result.

[0104] Understandably, in the early warning of abnormal events in the pipeline corridor, it is usually necessary to convert the prediction results into a binary classification problem of abnormality or normality. The logistic regression (LR) model is particularly effective for binary classification problems. Compared with other models, the training and prediction speed of the logistic regression (LR) model is usually faster. At the same time, it is not easily affected by data noise. Therefore, the logistic regression (LR) model can generate a more accurate and stable final prediction result.

[0105] Step 9: Use cross-validation to evaluate the overall performance and perform necessary parameter tuning. By continuously adjusting model parameters and optimizing model structure, the prediction accuracy and generalization ability of the model can be further improved.

[0106] Step 10: Real-time detection and early warning of abnormal events in the pipeline corridor. When an abnormal event is predicted, the early warning mechanism is immediately triggered to send alarm information to relevant personnel so that timely measures can be taken to deal with it.

[0107] In summary, the embodiment of the present application also provides a system for predicting abnormal events in an urban underground power supply corridor, including: a perception layer; the perception layer includes various types of sensor nodes, each sensor node is provided with a linear regression data processing module and a wireless communication module, and each sensor node does not interact with data. A central server, a wireless communication module is provided inside the central server, and the wireless communication module is responsible for information communication between each sensor node and the central server. The central server includes: a FedAvg federated learning module, an adaptive learning rate adjustment mechanism, an extraction module, a derivative calculation module, a basic prediction model module, and a logistic regression (LR) model module. These modules jointly use the above-mentioned prediction method to predict abnormal events. The abnormal event management layer, when an abnormal event is predicted, immediately triggers the early warning mechanism and sends an alarm message to relevant personnel so that timely measures can be taken to deal with it.

[0108] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods may also be implemented in other ways. The device embodiments described above are merely schematic, for example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions and operations of the devices, methods and computer program products according to the embodiments of the present application. In this regard, each box in the flowchart or block diagram may represent a module, a program segment or a part of a code, and the module, program segment or a part of a code includes one or more executable instructions for implementing a specified logical function.

[0109] It should also be noted that in some alternative implementations, the functions marked in the blocks may also occur in a different order than that marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the opposite order, depending on the functions involved.

[0110] It should also be noted that each box in the block diagram and / or flowchart, and combinations of boxes in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0111] In addition, the functional modules in the embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0112] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product in essence or in other words, the part that contributes to the prior art or the part of the technical solution. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in the embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk. Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, so the scope of protection of the present invention should be based on the scope defined by the claims.

Claims

1. A method for predicting abnormal events in urban underground power supply corridors. It is characterized in that Applied to the abnormal event prediction system of the urban underground power supply corridor, the system includes a central server; the method includes the following steps: The original data of target attributes are obtained in real time through various sensor nodes deployed in the tunnel; The raw data is trained through a linear regression data processing module set by each type of sensor node to obtain a first model parameter; , uploading the first model parameters to the central server through the wireless communication modules provided inside the various sensor nodes, and the central server aggregates the first model parameters to generate second model parameters; The central server sends the second model parameters to the various sensor nodes, and the various sensor nodes obtain third model parameters after updating; The third model parameters are uploaded to the central server via the wireless communication module, and the central server extracts basic features from the third model parameters; The central server calculates the derived features of the basic features; Combining the basic features with the derived features to obtain a first data set, and the central server performing training prediction on the first data set to obtain a first prediction result; and The central server performs regression prediction training on the first prediction result to obtain a second prediction result, and the second prediction result is the final prediction result.

2. The prediction method according to claim 1, It is characterized in that The linear regression data processing module only processes the raw data collected by the corresponding sensor nodes.

3. The prediction method according to any one of claims 1 or 2, It is characterized in that The second model parameter aggregation formula is: Among them, W is the second model parameter, k is the type of sensor node, N is the number of sensor nodes participating in the training, n k is the number of samples of the kth sensor node, w k is the model parameter of the kth sensor node.

4. The prediction method according to claim 3, It is characterized in that The central server aggregates the first model parameters to generate second model parameters, including: The central server aggregates the first model parameters using an adaptive learning rate adjustment mechanism to generate the second model parameters.

5. The prediction method according to claim 3, It is characterized in that The central server calculates the derived features of the basic features, including: The central server calculates the sliding average value of the past period of time using a sliding average calculation formula; wherein the sliding average calculation formula is: In the formula, SMA t is the sliding average at time t, x i is the observed value at time i, and n is the number of sensor nodes.

6. The prediction method according to claim 5, It is characterized in that The central server calculates the derived features of the basic features, including: The central server calculates the exponential smoothing value of the past period of time using an exponential average calculation formula; wherein the exponential average calculation formula is: S t =ax t +(1-a)S t-1 In the formula, S t is the exponential smoothing value at time t, x t is the observed value at time t, S t-1 is the exponential smoothing value at time t-1, a is the smoothing constant (0 <a<1)。 7. The prediction method according to claim 3, It is characterized in that The central server Calculate the derived features of the basic features, including: The central server calculates the standard deviation of each basic feature by using the standard deviation calculation formula. The standard deviation over a period of time; wherein the standard deviation calculation formula is: Where σ is the standard deviation, N is the number of sensor nodes, and x i is the observed value at time i, and u is the mean of the observed values ​​at multiple time points.

8. The prediction method according to claim 7, It is characterized in that The central server package A basic prediction model module including one or more combinations of random forest (RF), gradient boosted decision tree (GBDT) and support vector machine (SVM) is used to perform prediction training on the first data set to obtain one or more first prediction results.

9. The prediction method according to claim 8, It is characterized in that The central server package The method comprises a logistic regression (LR) model module, and inputting one or more first prediction results as new features into the logistic regression (LR) model module to obtain one or more second prediction results.

10. An electronic device, It is characterized in that include: A memory for storing one or more programs; processor; When the one or more programs are executed by the processor, the implementation as claimed in claims 1-9 is achieved. The method for predicting abnormal events in an urban underground power supply corridor as described in any one of the above.