RTU abnormity early warning system based on AI algorithm

Abstract

The invention relates to the technical field of RTU anomaly detection, in particular to an RTU anomaly early warning system based on an AI algorithm, and the system comprises an analysis early warning subsystem which is used for carrying out the anomaly analysis of data transmitted by an RTU; the analysis early warning subsystem comprises a data acquisition unit, a data identification unit, a data analysis unit and a feedback early warning unit, the data acquisition unit is used for monitoring the RTU in real time, and the data identification unit is used for identifying, judging and processing data acquired by the data acquisition unit; according to the system, the running state of the RTU can be detected through the analysis early warning subsystem, the reason of the RTU fault can be obtained according to subsequent identification analysis, multi-mode RTU fault reason analysis can be realized through data fusion, the integrity of RTU fault analysis is improved, the abnormal reason can be quickly found when the RTU fault occurs, and the fault diagnosis efficiency is improved. And targeted maintenance by workers is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of RTU anomaly detection, and in particular to an RTU anomaly early warning system based on an AI algorithm. Background Art

[0002] RTU is a remote terminal unit, a special computer measurement and control unit with a modular structure designed for long communication distances and harsh industrial field environments. It connects the terminal detection instruments and actuators with the main computer of the remote control center. It has remote data acquisition, control and communication functions, can receive operating instructions from the main computer, and control the action of the terminal actuators.

[0003] For example, an intelligent RTU monitoring system with disaster recovery function, whose publication (announcement) number is CN205541380U and whose publication (announcement) date is 2016-08-31, is characterized in that the system includes at least one intelligent RTU, at least one sensor module, a sensor unit, a server module and a client module; the intelligent RTU is provided with a first wireless communication unit; the sensor module is provided with a second wireless communication unit; the sensor unit is data connected with the sensor module; the intelligent RTU, the sensor module, the server module and the client module are communicatively connected via the first wireless communication unit; the sensor module is communicatively connected with the first wireless communication unit via the second wireless communication unit; the intelligent RTU is provided with a backup RTU; multi-site networking and network management of data can be realized, and a backup RTU is provided at the same time, so that the system has stronger stability.

[0004] In order to improve the stability of RTU operation, the disaster recovery capability of RTU will be improved through backup, but this method is only a transitional means. In order to ensure the stability of RTU operation, RTU will be inspected and repaired to find out the cause of RTU failure. However, during the troubleshooting process, workers need to spend a lot of time and the troubleshooting is difficult. It is impossible to quickly and accurately get the cause of RTU operation failure. Therefore, it is urgent to design an RTU abnormal warning system based on AI algorithm to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide an RTU abnormality warning system based on AI algorithm to solve the above-mentioned deficiencies in the prior art.

[0006] In order to achieve the above object, the present invention provides the following technical solutions: An RTU abnormality warning system based on an AI algorithm includes an analysis and warning subsystem, wherein the analysis and warning subsystem is used to perform abnormal analysis on data transmitted by the RTU; The analysis and early warning subsystem includes a data acquisition unit, a data identification unit, a data analysis unit, and a feedback early warning unit; The data acquisition unit is used to monitor the RTU in real time. The data acquisition unit adopts a multi-source data acquisition module, and the specific detection targets are temperature, humidity, communication transmission rate, voltage and current, action command execution effect, and energy consumption; The data identification unit is used to identify and judge the data collected by the data collection unit, and the specific process is as follows: Step S1-1. Identify the data collected by the data acquisition unit, extract features from the data, and classify and summarize the data based on the features; Step S1-2. The classified data is judged according to the calculation formula to determine whether the data meets the data standard range during normal RTU operation. The specific calculation formula is as follows: ; Among them, A is the data value of the classified summary data, It is the minimum value of the relevant RTU operation data corresponding to the classified summary data. is the maximum value of the relevant RTU operation data corresponding to the classified summary data. If A is located at , It indicates that the data value of the classified summary data is within the normal range. If A is between , In addition, it indicates that the data values ​​of the classified and summarized data do not conform to the normal range; Step S1-3. After the preliminary judgment is completed, the abnormal data found is marked and transmitted to the feedback warning unit for warning processing, and the data without abnormalities is transmitted to the data analysis unit.

[0007] The data analysis unit is used to analyze and process the non-abnormal data identified by the data identification unit. The data analysis unit selects a fusion and anomaly detection model. The analysis process is as follows: Step S2-1. Standardize the data to eliminate the dimension effect; Step S2-2. Using data fusion technology to fuse the data; Step S2-3. Detect the fused data through the anomaly detection model, and the detection process is divided into the following steps: (1) Deep convolutional neural network input: convert RTU related data into a time series signal spectrum graph; (2) Autoencoder input: structured data of ambient temperature and humidity; (3) Reconstruction error > 3σ is considered abnormal.

[0008] The feedback warning unit is used to feed back the results processed by the data identification unit and the data analysis unit to the base station staff. The feedback warning unit is established based on information feedback technology, and specifically provides feedback through text messages, pop-up push, and voice reminders.

[0009] In another embodiment provided by the present invention, it also includes a computing model building subsystem and a storage and optimization subsystem. The computing model building subsystem builds an AI algorithm model by analyzing the previous abnormal data of the RTU. The storage and optimization subsystem is used to optimize the analysis and early warning system and store the detected RTU operation data. The computing model building subsystem includes a data acquisition unit and a model building unit. The data acquisition unit is used to acquire the data stored in the storage and optimization subsystem. The model building unit builds the recognition model required by the data recognition unit and the fusion and anomaly detection model required by the data analysis unit based on the data acquired by the data acquisition unit. The data acquisition unit includes a data call module and a data category module. The data category module is used to collect the RTU data category collected by the data acquisition unit. The data call module includes a traversal submodule and an extraction submodule. The traversal submodule is used to perform traversal marking processing on the data stored in the storage and optimization subsystem. The extraction submodule extracts the data marked by the traversal submodule according to the data category module. The construction steps of the model building unit are as follows: Step S3-1. Filter, denoise, and normalize the data acquired by the data acquisition unit to reduce noise interference and extract effective features; Step S3-2. Construct recognition model, fusion and anomaly detection model through deep convolutional neural network and autoencoder based on the extracted features; Step S3-3. Train the constructed recognition model, fusion and anomaly detection model, optimize the recognition model, fusion and anomaly detection model according to the training results, and construct a complete recognition model, fusion and anomaly detection model.

[0010] In another embodiment provided by the present invention, the storage and optimization subsystem includes a storage unit and an optimization unit. The storage unit is a solid-state hard disk. The storage unit is used to store the RTU data collected by the data acquisition unit and the data of the normal operation of the RTU. The storage unit also has a cloud upload function. The storage unit can automatically upload the stored data to the cloud database. The optimization unit includes an energy adjustment module and a connection and relay module. The energy adjustment module adjusts the power supply of the RTU according to the analysis results of the analysis and early warning subsystem. The connection and relay module is used to connect multiple RTU devices together to build an RTU operation network. The connection and relay module will also use the operation steps of other RTU devices to control the operation of the terminal actuator when the RTU device fails.

[0011] In the above technical solution, the present invention provides an RTU abnormal warning system based on AI algorithm, which has the following beneficial effects: (1) The present invention can detect the operating status of the RTU through the analysis and early warning subsystem, and can obtain the cause of the RTU failure based on subsequent identification and analysis. In addition, through the fusion of data, multi-modal RTU failure cause analysis can be realized, thereby improving the completeness of the RTU failure analysis, and can quickly find the abnormal cause when the RTU fails, facilitating workers to carry out targeted maintenance; (2) The present invention can connect multiple RTUs together through the storage and optimization subsystem. When one RTU fails, the storage and optimization subsystem can control the operation of the terminal actuator according to the operation of other RTUs, thereby improving the disaster recovery capability of the RTU; (3) The present invention uses the "DCNN+AE" hybrid model to solve the limitations of the traditional single model for complex anomaly detection, and can break through the fixed threshold rules, adapt to the problems of equipment aging and changes in working conditions, and improve the perfect effect of anomaly detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0013] Figure 1 The present invention provides a system structure block diagram of an RTU abnormal warning system embodiment based on an AI algorithm.

[0014] Figure 2 A schematic diagram of a data acquisition unit provided for an embodiment of an RTU abnormality warning system based on an AI algorithm of the present invention.

[0015] Figure 3 A schematic diagram of a data identification unit process provided for an embodiment of an RTU abnormal warning system based on an AI algorithm of the present invention.

[0016] Figure 4 A schematic diagram of a data analysis unit process provided for an embodiment of an RTU abnormal warning system based on an AI algorithm of the present invention.

[0017] Figure 5 A schematic diagram of a model building unit process provided for an embodiment of an RTU abnormal warning system based on an AI algorithm of the present invention.

[0018] Figure 6A schematic diagram of an optimization unit provided for an embodiment of an RTU abnormality warning system based on an AI algorithm of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0020] like Figure 1-Figure 6 As shown, an RTU abnormality warning system based on an AI algorithm provided in an embodiment of the present invention includes an analysis and warning subsystem, and the analysis and warning subsystem is used to perform abnormal analysis on data transmitted by the RTU; The analysis and early warning subsystem includes a data collection unit, a data identification unit, a data analysis unit, and a feedback and early warning unit; The data acquisition unit is used to monitor the RTU in real time. The data acquisition unit adopts a multi-source data acquisition module, and the specific monitoring targets are temperature, humidity, communication transmission rate, voltage and current, action command execution effect, and energy consumption; The data identification unit is used to identify and judge the data collected by the data collection unit, and the specific process is as follows: Step S1-1. Identify the data collected by the data acquisition unit, extract features from the data, and classify and summarize the data based on the features; Step S1-2. The classified data is judged according to the calculation formula to determine whether the data meets the data standard range during normal RTU operation. The calculation formula is as follows: ; Among them, A is the data value of the classified summary data, It is the minimum value of the relevant RTU operation data corresponding to the classified summary data. is the maximum value of the relevant RTU operation data corresponding to the classified summary data. If A is located at , It indicates that the data value of the classified summary data is within the normal range. If A is between , In addition, it indicates that the data values ​​of the classified and summarized data do not conform to the normal range; Step S1-3. After the preliminary judgment is completed, the abnormal data found is marked and transmitted to the feedback warning unit for warning processing, and the data without abnormalities is transmitted to the data analysis unit.

[0021] The data analysis unit is used to analyze and process the normal data identified by the data identification unit. The data analysis unit uses a fusion and anomaly detection model. The analysis process is as follows: Step S2-1. Standardize the data to eliminate the dimension effect; Step S2-2. Using data fusion technology to fuse the data; It should be noted that the data fusion algorithm aims to improve the accuracy and reliability of information processing by integrating multi-source heterogeneous data. The following is a summary of the key points: ‌Common data fusion algorithm classification:‌ ‌Basic Statistics‌ ‌Weighted average method‌: Weighted processing of redundant data from multiple sensors is performed. It has strong real-time performance but requires manual setting of weights.

[0022] ‌Kalman Filter (KF): It estimates the state of dynamic systems through recursive optimization and is suitable for real-time low-level sensor data fusion (such as autonomous vehicle positioning)‌.

[0023] Probabilistic Reasoning ‌Bayesian estimation‌: Processing uncertain information based on probability models, often used for confidence fusion of multi-source data‌.

[0024] ‌D-S evidence reasoning‌: solves decision-making problems under conflicting evidence, suitable for complex scenarios such as target recognition‌.

[0025] ‌Intelligent Computing‌ Fuzzy logic: It can process imprecisely described data and realize semantic fusion through membership function.

[0026] ‌Neural network‌: uses deep learning to extract features and is good at processing unstructured data fusion such as images and texts. ‌ Step S2-3. Detect the fused data through the anomaly detection model, and the detection process is divided into the following steps: (1) Deep convolutional neural network input: convert RTU related data into a time series signal spectrum graph; (2) Autoencoder input: structured data of ambient temperature and humidity; (3) A reconstruction error > 3σ is considered abnormal.

[0027] It should be noted that the anomaly detection process combining ‌DCNN‌ (processing the conversion of RTU-related data into a time series signal spectrum) and ‌AE‌ (processing environmental parameters) is as follows: ‌1. Data preprocessing‌ ‌a. Time series vibration signal processing‌ ‌Spectrum graph generation‌: Perform short-time Fourier transform (STFT) or wavelet transform on the original vibration signal to generate a time-frequency domain spectrum graph as DCNN input‌ 5.

[0028] ‌Normalize‌: Normalize the spectrogram.

[0029] ‌b. Environmental parameter processing‌ ‌Structuring unstructured data‌: Convert temperature and humidity parameters into numerical vectors‌.

[0030] ‌Missing value filling‌: Use mean or interpolation to fill missing values.

[0031] ‌2. Model construction and training‌ ‌a.DCNN model (vibration signal feature extraction)‌ ‌Input layer‌: Receives the spectrogram.

[0032] ‌Convolutional layer‌: Use 3×3 convolution kernels to extract local features (such as edges, periodic patterns)‌.

[0033] ‌Pooling Layer‌: Max pooling reduces dimensionality.

[0034] ‌Output layer‌: The fully connected layer generates the feature vector.

[0035] ‌b.AE model (environmental parameter reconstruction)‌ ‌Encoder‌: compresses the input vector into a low-dimensional latent space.

[0036] ‌Decoder‌: Reconstructs the original input from the latent space, and the loss function is the mean squared error (MSE).

[0037] ‌3. Reconstruction error calculation and anomaly determination‌ ‌a. Normal data training phase‌ Use normal working condition data to train DCNN and AE, and record the reconstruction error (MSE of AE output and input).

[0038] Calculate the mean (μ) and standard deviation (σ) of the reconstruction error, and establish the threshold: ‌μ+3σ‌‌.

[0039] ‌b. Real-time detection stage‌ ‌DCNN Inference‌: Input is a spectrogram and output is a vibration feature vector.

[0040] ‌AE Reasoning‌: Input environment parameters and calculate reconstruction error.

[0041] Abnormality judgment: If the AE reconstruction error > μ+3σ, it is judged as abnormal.

[0042] The feedback warning unit is used to feed back the results processed by the data identification unit and the data analysis unit to the base station staff. The feedback warning unit is established based on information feedback technology, and feedback is provided specifically through text messages, pop-up push notifications, and voice reminders.

[0043] In another embodiment provided by the present invention, a computing model building subsystem and a storage and optimization subsystem are also included. The computing model building subsystem builds an AI algorithm model by analyzing the previous abnormal data of the RTU. The storage and optimization subsystem is used to optimize the analysis and early warning system and store the detected RTU operation data. The computing model building subsystem includes a data acquisition unit and a model building unit. The data acquisition unit is used to acquire the data stored in the storage and optimization subsystem. The model building unit builds the recognition model required by the data recognition unit and the fusion and anomaly detection model required by the data analysis unit based on the data acquired by the data acquisition unit. The data acquisition unit includes a data call module and a data category module. The data category module is used to collect the RTU data category collected by the data acquisition unit. The data call module includes a traversal submodule and an extraction submodule. The traversal submodule is used to perform traversal marking processing on the data stored in the storage and optimization subsystem. The extraction submodule extracts and processes the data marked by the traversal submodule based on the data category module. The model building unit construction steps are as follows: Step S3-1. Filter, denoise, and normalize the data acquired by the data acquisition unit to reduce noise interference and extract effective features; Step S3-2. Construct recognition model, fusion and anomaly detection model through deep convolutional neural network (DCNN) and autoencoder (AE) based on the extracted features; Step S3-3. Train the constructed recognition model, fusion and anomaly detection model, optimize the recognition model, fusion and anomaly detection model according to the training results, and construct a complete recognition model, fusion and anomaly detection model.

[0044] It should be noted that the traversal submodule uses the traversal and marking algorithm. The traversal and marking algorithm is a core method used for data exploration, state tracking and feature identification. The traversal and marking algorithm specifically uses the anomaly marking algorithm.

[0045] In another embodiment provided by the present invention, the storage and optimization subsystem includes a storage unit and an optimization unit. The storage unit is a solid-state hard disk. The storage unit is used to store the RTU data collected by the data acquisition unit and the data of the normal operation of the RTU. The storage unit also has a cloud upload function. The storage unit can automatically upload the stored data to the cloud database. The optimization unit includes an energy adjustment module and a connection and relay module. The energy adjustment module adjusts the power supply of the RTU according to the analysis results of the analysis and early warning subsystem. The connection and relay module is used to connect multiple RTU devices together to build an RTU operation network. The connection and relay module will also use the operation steps of other RTU devices to control the operation of the terminal actuator when the RTU device fails.

[0046] It should be noted that multiple RTU devices are connected through software drivers and permission management, and the details are as follows: Driver compatibility: Ensure that the device driver matches the operating system version. In multi-device scenarios, batch installation or update of drivers is required.

[0047] ‌Permission Check‌: Verify device access permissions to prevent connection interruptions due to insufficient permissions.

[0048] Working principle: When the system is running, the data acquisition unit in the computational model building subsystem will acquire the data stored in the storage and optimization subsystem, and then the model building unit will build the recognition model required by the data recognition unit and the fusion and anomaly detection model required by the data analysis unit based on the data acquired by the data acquisition unit; after the model building is completed, the data acquisition unit in the analysis and early warning subsystem will monitor the RTU in real time, and the monitoring targets are temperature, humidity, communication transmission rate, voltage and current, action command execution effect, and energy power consumption. The subsequent data recognition unit will identify and judge the data collected by the data acquisition unit, mark the abnormal data found, and transmit it to the feedback early warning unit for early warning processing. The data without abnormalities is transmitted to the data analysis unit, and the data analysis unit will analyze and process the data without abnormalities identified by the data recognition unit. The feedback early warning unit will feed back the results of the data recognition unit and the data analysis unit to the base station staff, so that the base station staff can carry out maintenance. At the same time, the storage and optimization subsystem can adjust the power supply of the RTU according to the analysis results of the analysis and early warning subsystem, and will also use the operation steps of other RTU devices to control the operation of the terminal actuator when the RTU device fails.

[0049] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. An RTU abnormal warning system based on AI algorithm, including an analysis and warning subsystem, characterized in that: The analysis and early warning subsystem is used to perform abnormal analysis on the data transmitted by the RTU; The analysis and early warning subsystem includes a data acquisition unit, a data identification unit, a data analysis unit, and a feedback early warning unit. The data acquisition unit is used to monitor the RTU in real time, and the data identification unit is used to identify and judge the data collected by the data acquisition unit. The specific process is as follows: Step S1-1. Identify the data collected by the data acquisition unit, extract features from the data, and classify and summarize the data based on the features; Step S1-2. According to the calculation formula, the classified data is judged to determine whether the data meets the data standard range during normal RTU operation; Step S1-3. After the preliminary judgment is completed, the abnormal data found is marked and transmitted to the feedback warning unit for warning processing, and the data without abnormality is transmitted to the data analysis unit; The data analysis unit is used to analyze and process the non-abnormal data identified by the data identification unit. The data analysis unit selects a fusion and anomaly detection model. The analysis process is as follows: Step S2-1. Standardize the data to eliminate the dimension effect; Step S2-2. Using data fusion algorithm to fuse the data; Step S2-3. Detect the fused data using an anomaly detection model; The feedback warning unit is used to feed back the results processed by the data identification unit and the data analysis unit to the base station staff.

2. According to claim 1, the RTU abnormal warning system based on AI algorithm is characterized in that: The data acquisition unit adopts a multi-source data acquisition module, and the specific monitoring targets are temperature, humidity, communication transmission rate, voltage and current, action command execution effect, and energy power consumption; the feedback warning unit is established based on information feedback technology, and feedback is specifically provided through text messages, pop-up push notifications, and voice reminders.

3. According to claim 1, the RTU abnormal warning system based on AI algorithm is characterized in that: The calculation formula selected by the data identification unit is as follows: ; Where A is the data value of the classified summary data, It is the minimum value of the relevant RTU operation data corresponding to the classified summary data. is the maximum value of the relevant RTU operation data corresponding to the classified summary data. If A is located at , It indicates that the data value of the classified summary data is within the normal range. If A is between , In addition, it indicates that the data values ​​of the classified and summarized data do not conform to the normal range; The data analysis unit anomaly detection model detection process is divided into the following steps: (1) Deep convolutional neural network input: convert RTU related data into a time series signal spectrum graph; (2) Autoencoder input: structured data of ambient temperature and humidity; (3) A reconstruction error > 3σ is considered abnormal.

4. According to claim 1, the RTU abnormal warning system based on AI algorithm further comprises a calculation model construction subsystem and a storage and optimization subsystem, characterized in that: The computing model building subsystem builds an AI algorithm model by analyzing the previous abnormal data of the RTU. The storage and optimization subsystem is used to optimize the analysis and early warning system and store the detected RTU operation data. The computing model building subsystem includes a data acquisition unit and a model building unit.

5. The RTU abnormality warning system based on AI algorithm according to claim 4 is characterized in that: The data acquisition unit is used to acquire data stored in the storage and optimization subsystem. The data acquisition unit includes a data calling module and a data category module. The data category module is used to collect the RTU data categories collected by the data acquisition unit.

6. The RTU abnormality warning system based on AI algorithm according to claim 5 is characterized in that: The model building unit builds the recognition model required by the data recognition unit and the fusion and anomaly detection model required by the data analysis unit based on the data acquired by the data acquisition unit. The data calling module includes a traversal submodule and an extraction submodule. The traversal submodule is used to perform traversal marking processing on the data stored in the storage and optimization subsystem, and the extraction submodule extracts and processes the data marked by the traversal submodule based on the data category module.

7. The RTU abnormality warning system based on AI algorithm according to claim 4 is characterized in that: The steps of constructing the model construction unit are as follows: Step S3-1. Filter, denoise, and normalize the data acquired by the data acquisition unit to reduce noise interference and extract effective features; Step S3-2. Construct recognition model, fusion and anomaly detection model through deep convolutional neural network and autoencoder based on the extracted features; Step S3-3. Train the constructed recognition model, fusion and anomaly detection model, optimize the recognition model, fusion and anomaly detection model according to the training results, and construct a complete recognition model, fusion and anomaly detection model.

8. The RTU abnormality warning system based on AI algorithm according to claim 4 is characterized in that: The storage and optimization subsystem includes a storage unit and an optimization unit. The storage unit is a solid-state hard disk. The storage unit is used to store the RTU data collected by the data acquisition unit and the data of the normal operation of the RTU. The storage unit also has a cloud upload function. The storage unit automatically uploads the stored data to the cloud database.

9. The RTU abnormality warning system based on AI algorithm according to claim 8 is characterized in that: The optimization unit includes an energy adjustment module and a connection and relay module. The energy adjustment module adjusts the power supply of the RTU according to the analysis result of the analysis and early warning subsystem.

10. The RTU abnormality warning system based on AI algorithm according to claim 9 is characterized in that: The connection and relay module is used to connect multiple RTU devices to build an RTU operation network. The connection and relay module will also use the operation steps of other RTU devices to control the operation of the terminal actuator when the RTU device fails.

Images