An integrated relay protection and management system based on intelligent algorithm

Abstract

The application discloses a kind of comprehensive relay protection and management system based on intelligent algorithm, to realize the real-time monitoring of power system, fault detection and prediction, and relay protection control by highly integrated technical scheme.The system includes data exchange unit, programmable logic controller (PLC), intelligent data processing unit and touch screen device.Data exchange unit is responsible for the data bidirectional exchange with the sensors and devices of power system, ensures the real-time collection of system data and the effective transmission of relay protection control signal.PLC receives and processes these data, generates relay protection control signal according to the instructions of intelligent data processing unit.Intelligent data processing unit pre-processes, analyzes and predicts the collected data, provides PLC with logic control instructions based on deep learning, to optimize relay protection action.Finally, the touch screen device displays the real-time running state and fault analysis result of power system, and provides user interaction interface.

Description

Technical Field

[0001] This invention relates to the field of relay protection technology, and in particular to a comprehensive relay protection and management system based on intelligent algorithms. Background Technology

[0002] In the operation and management of power systems, relay protection systems play a crucial role. They are responsible for providing necessary protection when power system faults occur, preventing equipment damage and the spread of system failures, and ensuring stable system operation. Existing relay protection systems typically rely on a series of preset, fixed parameters to trigger protection actions. These systems may include various sensors and electrical devices, connected to a central processor or programmable logic controller (PLC) via a data exchange unit to achieve basic monitoring and control functions. However, due to the complexity of power systems and the variability of operating environments, traditional relay protection strategies sometimes struggle to respond promptly and accurately to the actual protection needs of the system.

[0003] The main problem with existing technologies is the lack of sufficient flexibility and intelligence, making the systems slow or inaccurate in responding to unforeseen changes or complex fault scenarios. Traditional methods often rely on fixed protection logic and parameter settings, making it difficult to adapt to real-time changes in the operating state of the power system. Furthermore, the limited fault analysis capabilities of these systems mean that relay protection decisions may be based on outdated or inaccurate information, thereby increasing the risk of system operation.

[0004] Therefore, there is an urgent need to develop a highly adaptive and intelligent relay protection and management system to improve the safety and reliability of the power system. Summary of the Invention

[0005] This application provides an integrated relay protection and management system based on intelligent algorithms to improve the safety and reliability of power systems.

[0006] This application provides a comprehensive relay protection and management system based on intelligent algorithms, including:

[0007] The data exchange unit is equipped with serial communication and Ethernet communication capabilities and is configured to perform bidirectional data exchange with sensors and electrical equipment in the power system. It is used to collect system data in the power system and transmit relay protection control signals generated by the programmable logic controller to the electrical equipment.

[0008] A programmable logic controller (PLC) is connected to the data exchange module and configured to receive system data from the data exchange module; process the received system data according to the logic control instructions provided by the intelligent data processing unit, generate a relay protection control signal, and transmit the generated relay protection control signal to the data exchange unit.

[0009] An intelligent data processing unit is configured to receive system data from a programmable logic controller (PLC); preprocess the received system data to obtain preprocessed data, wherein the preprocessing includes data standardization and cleaning to ensure the data quality for subsequent analysis; analyze the preprocessed data to obtain the real-time operating status of the power system and transmit the real-time operating status of the power system to a touch screen device; perform fault analysis on the power system based on the preprocessed data, obtain fault analysis results, and transmit the fault analysis results to the touch screen device; generate logic control instructions for relay protection based on the real-time operating status and the fault analysis results; and transmit the generated logic control instructions to the PLC.

[0010] The touchscreen device, connected to the programmable logic controller and intelligent data processing unit, is equipped with a graphical user interface to display the real-time operating status and fault analysis results of the power system to the user; it also allows the user to directly input operation commands and configure the system.

[0011] Furthermore, the intelligent data processing unit uses a trained neural network model to obtain the real-time operating status and fault analysis results of the power system based on the preprocessed data, and generates logic control instructions for relay protection. The neural network model includes a feature extraction part, a state analysis part, and a decision generation part. The feature extraction part uses a multi-scale convolutional neural network to process the preprocessed data and obtain a high-dimensional feature representation of the time series data, wherein the preprocessed data includes time series data and equipment status index data. The state analysis part uses a bidirectional long short-term memory network to process the high-dimensional feature representation of the time series data provided by the feature extraction part, and obtains the real-time operating status of the power system, fault analysis results, and feature representations for decision generation. The decision generation part uses a graph attention network to process the real-time operating status of the power system, fault analysis results, and feature representations for decision generation provided by the state analysis part, and generates logic control instructions for relay protection.

[0012] Furthermore, the feature extraction part includes an input layer, a multi-scale convolutional layer, a pooling layer, a dynamic feature fusion module, and an output layer;

[0013] The input layer is used to receive preprocessed time series data and equipment status index data from the power system, and directly use the time series data and status index as the network input. The equipment status index data is processed by an embedding transformation before input to adapt to the network input requirements.

[0014] The multi-scale convolutional layer includes three different scales of convolutional kernels: 3×3, 5×5, and 7×7. Each scale of convolutional kernel is used to process the data passed from the input layer to extract feature information at different scales. The output of the multi-scale convolutional layer is a combination of feature maps extracted at the three scales, which is completed by concatenating the feature maps along the depth direction.

[0015] The pooling layer uses the max pooling method to downsample the combination of feature maps provided by the multi-scale convolutional layer, reducing the feature dimension while retaining important feature information, thereby reducing the complexity and computational cost of the model.

[0016] The dynamic feature fusion module is implemented through a fully connected layer, which fuses the pooled feature maps output by the pooling layer to obtain a feature vector with dynamically adjusted weights.

[0017] The output layer is used to convert the output of the dynamic feature fusion module into a high-dimensional feature representation of the time series data.

[0018] Furthermore, the state analysis part of the neural network model employs an improved bidirectional long short-term memory network, which enhances its ability to understand and predict the time dependence of power system data by introducing an adaptive gating mechanism; wherein, the improved bidirectional long short-term memory network is implemented in the following manner:

[0019] A standard bidirectional long short-term memory network structure is adopted to process the forward and backward data sequences simultaneously at each time step t, in order to capture the forward and backward time dependencies in the time series;

[0020] At each time step t, based on the current input x t And the hidden state h of the previous time step t-1 t-1 Dynamically calculate the gating signal g t Wherein, the gating signal g t It is to change the current input x t and hidden state h t-1 The concatenated result is input into a fully connected layer and activated by the sigmoid function, which is used to adaptively adjust the weights for updating the hidden state.

[0021] The update formula for hidden states is expressed as follows:

[0022]

[0023] Among them, h t This represents the hidden state at time step t; ⊙ represents the candidate hidden state calculated at time step t in a standard bidirectional long short-term memory network without the application of a gating signal; ⊙ represents element-wise multiplication.

[0024] Furthermore, the decision generation part of the neural network model is specifically used for:

[0025] The real-time operating status of the power system, fault analysis results, and decision generation feature representations provided by the state analysis section are used as inputs. These inputs reflect the current comprehensive state of the power system and potential fault information. The inputs are integrated into a unified representation to facilitate processing by the graph attention network.

[0026] A graph attention network is used to process the integrated input data. When dealing with complex relationships between nodes, the graph attention network adaptively allocates attention weights, thereby effectively capturing the interdependencies and influences between the components of the power system; the nodes include various components of the power system.

[0027] Based on the results of graph attention network processing, logical control instructions are generated; the logical control instructions provide relay protection operation suggestions for the identified power system status or potential faults, wherein the relay protection operation suggestions include adjusting the load, activating the backup system, or triggering an alarm.

[0028] Furthermore, the data exchange unit is configured to use an encrypted communication protocol to ensure the security and integrity of data during transmission.

[0029] Furthermore, the touchscreen device includes an alarm system for alerting operators when an abnormal operating state or potential fault of the power system is detected.

[0030] Furthermore, the intelligent data processing unit is configured to automatically adjust the operating parameters of the power system based on the real-time operating status and fault analysis results of the power system, so as to optimize the performance of the power system and prevent potential faults. The operating parameters include power generation, voltage level, or load distribution of the power grid.

[0031] Furthermore, the programmable logic controller is equipped with a self-diagnostic function, which can automatically detect and report its operating status and potential hardware or software faults for timely maintenance and repair.

[0032] The beneficial effects of this application mainly include: (1) By performing real-time analysis and fault prediction on power system data through an intelligent data processing unit, the system can quickly identify potential faults and abnormal states, and promptly generate and issue relay protection control signals. This is faster and more accurate than traditional protection systems based on fixed parameters in detecting and responding to faults. (2) Utilizing advanced data preprocessing technology and intelligent algorithms, this system can dynamically adjust protection strategies according to the real-time operating status of the power system and environmental changes. This adaptive capability enables the system to maintain efficient and stable protection performance even when facing complex and ever-changing operating conditions. (3) Through comprehensive monitoring and intelligent fault analysis of the power system, the system can anticipate and prevent the spread of potential faults, reducing the risk of equipment damage and power outages. This not only improves the overall reliability of the power system but also ensures power supply safety and economic benefits. (4) The touch screen device equipped with a graphical user interface provides users with an intuitive and easy-to-operate platform, allowing users to monitor the operating status of the power system in real time and easily configure the system and handle faults. This human-computer interaction design greatly improves the system's usability and user experience. (5) The system's intelligent analysis and prediction functions reduce the reliance on professional personnel, making the daily maintenance and management of the power system more efficient and economical. Meanwhile, the fault analysis results and operational status information provided by the system help guide maintenance decisions and optimize resource allocation. Attached Figure Description

[0033] Figure 1 This is a schematic diagram of an integrated relay protection and management system based on intelligent algorithms provided in the first embodiment of this application. Detailed Implementation

[0034] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0035] The first embodiment of this application provides a comprehensive relay protection and management system based on intelligent algorithms. Please refer to... Figure 1 This figure is a schematic diagram of the first embodiment of this application. The following is in conjunction with... Figure 1 The first embodiment of this application provides a detailed description of an integrated relay protection and management system based on intelligent algorithms.

[0036] The integrated relay protection and management system includes a data exchange unit 101, a programmable logic controller 102, an intelligent data processing unit 103, and a touch screen device 104.

[0037] The data exchange unit 101 has serial communication and Ethernet communication capabilities and is configured to perform bidirectional data exchange with sensors and electrical equipment in the power system. It is used to collect system data in the power system and transmit relay protection control signals generated by the programmable logic controller to the electrical equipment.

[0038] In the integrated relay protection and management system based on intelligent algorithms, the data exchange unit 101 is a key component of the system. Its design and function ensure the efficient and accurate operation of the entire system. The main responsibility of the data exchange unit 101 is to realize bidirectional data exchange with various sensors and electrical equipment in the power system. This includes, but is not limited to, collecting data generated by these devices and sensors and transmitting control signals from the system back to these devices.

[0039] In power systems, there are numerous types of sensors and electrical devices, which together constitute the sensing and execution layers of the system, playing a crucial role in system monitoring, control, and protection. Below are some examples of common sensors and electrical devices:

[0040] The sensors include:

[0041] 1. Current sensor: Used to measure the magnitude of current in circuits of a power system. Current sensors provide real-time current data, which is crucial for detecting abnormal conditions such as overload and short circuit.

[0042] 2. Voltage sensor: Used to measure voltage levels in a circuit. Voltage fluctuations reflect load changes and grid stability in a power system, and are crucial for voltage control and load management.

[0043] 3. Temperature Sensors: In electrical equipment, temperature sensors are used to monitor the operating temperature of the equipment, such as the temperature of transformer oil. Excessively high temperatures may indicate equipment overload or a cooling system malfunction.

[0044] 4. Vibration sensors: Used to monitor the vibration status of electrical equipment (such as generators and transformers). Abnormal vibration patterns may indicate equipment failure.

[0045] Electrical equipment includes:

[0046] 1. Circuit breaker: Used to automatically disconnect the circuit when a fault is detected, to prevent the fault from spreading and to protect the system safety.

[0047] 2. Relay: As an automatic control device in a power system, it responds to sensor signals and performs operations such as closing and opening to achieve control and protection of the power system.

[0048] 3. Transformers: Used to increase or decrease voltage to meet different transmission or distribution needs. The operating status of transformers needs to be monitored in real time to ensure a stable power supply to the power system.

[0049] 4. Distribution cabinet / switch cabinet: Distribution cabinets contain various control and protection devices used for the distribution and control of the power system. They receive data from sensors and execute corresponding control logic based on this data.

[0050] Through bidirectional data exchange between the data exchange unit 101 and these sensors and electrical devices, the power system can achieve real-time monitoring and rapid response. For example, when a current sensor detects an overload current, the data exchange unit 101 can quickly transmit this information to the programmable logic controller, which then generates a control signal to instruct the circuit breaker to operate and disconnect the faulty circuit, thereby protecting the power system from further damage.

[0051] The data exchange unit 101 is capable of both serial communication and Ethernet communication, meaning it can communicate with devices in the power system in two different ways. Serial communication is a traditional communication method suitable for short-distance communication, while Ethernet communication supports data transmission over longer distances and can handle larger amounts of data, which is particularly important for large power systems.

[0052] In power systems, serial communication and Ethernet communication each play important roles to ensure accurate and real-time transmission of information.

[0053] Serial communication (such as RS-232, RS-485, etc.) is a data transmission method based on serial communication protocols, commonly used for point-to-point communication between devices in power systems. A typical application scenario is communication between protective relays and smart meters. In this scenario, the protective relay needs to monitor the voltage and current data of the meter in real time to determine whether the system is operating normally or whether there are faults such as overload or short circuit.

[0054] For example, an RS-485 serial communication link can connect a protective relay and several smart meters. RS-485 supports multi-point communication, allowing multiple devices to be connected on a single communication line, making it ideal for densely populated power distribution room environments. When a protective relay receives an abnormal current signal from a meter via the RS-485 line, it can react quickly, performing protective actions such as disconnecting the circuit, thereby preventing further damage to the power system.

[0055] Ethernet communication is a network layer-based communication method that supports longer communication distances and higher data transmission rates, making it suitable for wide area network (WAN) communication in power systems. One application of Ethernet communication in power systems is remote monitoring and control. For example, in a distributed generation system, the central control room may need to acquire real-time operating data from wind turbines and photovoltaic power plants located in different geographical locations.

[0056] By deploying Ethernet communication technology, the central control room's monitoring system can remotely access controllers and sensors at various power plants via the internet or a dedicated network to collect key operational data such as power output and equipment status. This data can then be used to optimize power generation scheduling and achieve optimal energy management. Simultaneously, the central control room can also remotely issue control commands via Ethernet, such as adjusting the blade angle of wind turbines to cope with changing wind speeds and load demands.

[0057] The data exchange unit 101, equipped with serial and Ethernet communication capabilities, provides a flexible and reliable communication solution for devices in the power system. By precisely selecting the appropriate communication method, the power system can achieve full coverage from local device control to wide-area system management, ensuring the efficient and stable operation of the power system.

[0058] The configuration of this unit is designed to ensure efficient and reliable data exchange. To achieve this, the data exchange unit 101 is designed with a highly compatible interface, allowing easy connection to a variety of sensors and electrical devices, regardless of their manufacturer or model. Furthermore, the data exchange unit 101 possesses advanced data processing capabilities, enabling real-time processing and forwarding of large amounts of data, ensuring that the system's real-time performance is not affected by data processing delays.

[0059] The data exchange unit 101 is not merely a simple data transmission device; it is also responsible for accurately transmitting the relay protection control signals generated by the programmable logic controller 102 to the corresponding electrical equipment. The accuracy of this process is crucial for ensuring the safe operation of the power system, as any delay or error in the control signals can lead to system failure or equipment damage.

[0060] In integrated relay protection and management systems based on intelligent algorithms, system data is a broad concept encompassing various types of information generated during power system operation. This data is crucial for monitoring system status, enabling fault detection, fault prediction, and relay protection. The following is a detailed explanation of system data, along with specific data examples that may be included in a power system:

[0061] The types of system data include:

[0062] 1. Electrical parameters: These are the most basic and important data types in a power system, including current, voltage, power (active and reactive power), power factor, and frequency. Examples include the input and output voltages of transformers and the current values ​​of transmission lines. These parameters are crucial for assessing the operating status, energy efficiency, and load conditions of the power system.

[0063] 2. Status indication data: This includes the open / closed status of switches and the operational status of protective devices. For example, whether a circuit breaker has been triggered or whether a disconnecting switch is in the closed state. This information is extremely important for understanding the system's operating logic and ensuring its safe operation.

[0064] 3. Environmental Data: This type of data involves various parameters of the operating environment of electrical equipment, such as temperature and humidity. For some equipment, environmental conditions have a direct impact on its performance; therefore, monitoring this data helps prevent failures caused by environmental factors.

[0065] 4. Fault and Abnormal Data: This includes data collected from various sensors and monitoring devices regarding system faults and abnormal states, such as specific indications of short circuits, overloads, and equipment malfunctions. This data is crucial for achieving rapid fault location and isolation, and for protecting system safety.

[0066] For example, consider a typical transmission line; its system data might include:

[0067] - The voltage and current values ​​at both ends of the transmission line, as well as the voltage and current at each node in the line.

[0068] Temperature monitoring data along the line is used to assess the risk of overheating, which may affect the line's conductivity and equipment lifespan.

[0069] - Real-time load data of the line is used to analyze load distribution and predict load change trends.

[0070] - The status of line protection devices, including the operation of overcurrent protection, distance protection, and other devices, is used to analyze the effectiveness of protection strategies and respond to system faults in a timely manner.

[0071] By collecting and analyzing this data, the data exchange unit 101 can provide the programmable logic controller with comprehensive system information, thereby generating accurate relay protection control signals to ensure the stable and safe operation of the power system. The intelligent data processing unit further analyzes this data in depth, not only identifying the current system status and potential fault points, but also predicting the future operating trends of the system, providing support for system management and maintenance.

[0072] Furthermore, the data exchange unit is configured to use an encrypted communication protocol to ensure the security and integrity of data during transmission.

[0073] The data exchange unit, acting as a bridge between the power system and the intelligent algorithm processing module, needs to process various types of information, including real-time monitoring data, equipment status indicators, and fault diagnosis results. This information is crucial for the stable operation of the power system; any data leakage or tampering could lead to inaccurate status judgments or erroneous control commands, thereby jeopardizing system security.

[0074] To protect the security and integrity of data transmission, data exchange units need to employ encrypted communication protocols. These protocols can include, but are not limited to, Secure Sockets Layer (SSL) and Transport Layer Security (TLS). These protocols provide end-to-end data encryption services, ensuring that data is not intercepted or tampered with by third parties during transmission.

[0075] The programmable logic controller 102 is connected to the data exchange module and is configured to receive system data from the data exchange module; process the received system data according to the logic control instructions provided by the intelligent data processing unit, generate a relay protection control signal, and transmit the generated relay protection control signal to the data exchange unit.

[0076] In the integrated relay protection and management system based on intelligent algorithms, the programmable logic controller (PLC) 102 plays a central role, connecting the data exchange unit 101 and the intelligent data processing unit 103. It is the brain and execution arm of the system. This PLC is specifically designed to receive and process data from power system sensors and equipment, as well as execute control commands from the intelligent data processing unit, ensuring the stable operation and safe protection of the power system.

[0077] The programmable logic controller 102 is capable of receiving system data from the data exchange unit 101. This data covers comprehensive information about the power system, including but not limited to measurement data of physical quantities such as current, voltage, and temperature, as well as status information of power equipment. The PLC connects to the data exchange unit through a built-in communication interface, such as a serial port or Ethernet port, to ensure real-time data transmission.

[0078] Upon receiving system data, the PLC processes and analyzes the data according to pre-programmed logic or logic control instructions provided by the intelligent data processing unit 103. During processing, the PLC may need to perform operations such as data filtering, logical judgment, and calculation to determine whether protective measures or adjustments to equipment operating parameters are necessary. For example, when a current exceeding a safety threshold is detected, the PLC generates a corresponding relay protection control signal, triggering a circuit breaker to isolate the faulty circuit and prevent the accident from spreading.

[0079] In addition, the PLC is also responsible for feeding back information such as the real-time operating status of the power system, fault analysis results, and future operating status predictions obtained by the intelligent data processing unit 103 to other parts of the system, such as the touch screen device 104, through the data exchange unit 101, so that operators can monitor the status of the power system in real time and make corresponding operational decisions.

[0080] The following example illustrates this:

[0081] A substation is equipped with the integrated relay protection and management system based on intelligent algorithms provided in this embodiment. One of the core components of this system is the programmable logic controller (PLC) 102, which communicates with various sensors and electrical devices in the power system through the data exchange unit 101.

[0082] Imagine a hot summer day where the heavy use of air conditioners and other appliances causes a sharp increase in the load on a substation, leading to a rapid rise in current in the transmission lines. In this situation, overload could damage equipment or even trigger a more serious power system failure.

[0083] 1. Data Collection: Current sensors in the power system monitor the current of transmission lines in real time. When the current exceeds a safety threshold, the sensor sends this information to PLC 102 via data exchange unit 101.

[0084] 2. Data Processing and Logic Control Instruction Generation: Upon receiving data, PLC 102 first performs preliminary verification and analysis. Then, it sends the data to the intelligent data processing unit 103 for further analysis. The intelligent data processing unit uses advanced algorithms to assess the current load situation and predict whether the load will continue to increase in the short term. Based on these analyses, the intelligent data processing unit generates logic control instructions, such as adjusting the load, activating backup lines, or triggering circuit breakers to isolate overloaded components.

[0085] 3. Execute control commands: These logic control commands are then sent back to PLC102. Based on the received commands, the PLC generates corresponding relay protection control signals, such as issuing a circuit breaker command.

[0086] 4. Control Signal Transmission and Execution: The PLC102 transmits these relay protection control signals to the corresponding electrical equipment, such as circuit breakers, through the data exchange unit 101. Upon receiving the control signal, the circuit breaker immediately activates, disconnecting the overloaded portion, thereby avoiding the risk of equipment damage and power outage.

[0087] 5. Feedback and Monitoring: Throughout the process, the touch screen device 104 displays the system status and operational feedback in real time, allowing the operator to monitor the process and intervene when necessary.

[0088] This example demonstrates the crucial role of the PLC in receiving data, processing instructions, and executing control commands. It not only ensures the stable operation of the power system but also responds rapidly to potential faults. Through intelligent algorithm-optimized control commands, it achieves efficient and precise fault handling, significantly improving the safety and reliability of the power system.

[0089] Furthermore, the programmable logic controller is equipped with a self-diagnostic function, which can automatically detect and report its operating status and potential hardware or software faults for timely maintenance and repair.

[0090] The Programmable Logic Controller (PLC) plays a central role, responsible for executing logic control instructions from the intelligent data processing unit and managing and monitoring the operation of the power system. To ensure high reliability and continuous stable operation of the system, the PLC is equipped with self-diagnostic capabilities. This function enables the controller to automatically detect and report its own operating status and potential hardware or software faults, which is crucial for early problem detection, timely maintenance, and repair.

[0091] The self-diagnostic function mainly relies on built-in monitoring and analysis tools, which can periodically or continuously detect multiple aspects of the controller, including but not limited to processor status, memory usage, integrity of input and output signals, communication interface status, and the correctness of running programs and logic.

[0092] 1. Processor status and memory usage: By detecting processor load and memory usage, the programmable logic controller can identify problems such as overload or memory leaks, which may cause the controller to run slowly or restart abnormally.

[0093] 2. Input / Output Signal Detection: The controller will verify the integrity and correctness of all input / output (I / O) signals, ensuring that the signals are not damaged or interfered with, and that all sensors and actuators are working properly.

[0094] 3. Communication Interface Status: The controller will detect the status of its communication interface, including network connection, serial communication, etc., to ensure smooth and accurate data exchange with other system components.

[0095] 4. Program and logic verification: Regularly check the running programs and implemented logic for errors, such as logical conflicts or program errors, which may affect the normal operation of the system.

[0096] When the self-diagnostic function detects a potential problem, the programmable logic controller (PLC) automatically generates a fault report and sends it to the management interface or directly notifies maintenance personnel via a predefined communication protocol. These fault reports typically contain specific information about the fault, such as the fault type, fault location, possible causes, and recommended maintenance or repair measures.

[0097] By implementing self-diagnostic functions in programmable logic controllers, integrated relay protection and management systems based on intelligent algorithms can achieve higher autonomy and reliability, detect and resolve potential problems in advance, thereby ensuring the stable operation and safety of the power system.

[0098] The intelligent data processing unit 103 is used to receive system data from the programmable logic controller (PLC); preprocess the received system data to obtain preprocessed data, wherein the preprocessing includes data standardization and cleaning to ensure the data quality for subsequent analysis; analyze the preprocessed data to obtain the real-time operating status of the power system and transmit the real-time operating status of the power system to the touch screen device; perform fault analysis on the power system based on the preprocessed data, obtain fault analysis results, and transmit the fault analysis results to the touch screen device; generate logic control instructions for relay protection based on the real-time operating status and the fault analysis results; and transmit the generated logic control instructions to the PLC.

[0099] The intelligent data processing unit 103 plays a core role in the integrated relay protection and management system based on intelligent algorithms described in this patent. Its main task is to perform in-depth analysis of the power system data received from the programmable logic controller 102 in order to monitor the operating status of the power system in real time, predict future operating trends, diagnose possible faults, and generate control commands for relay protection based on this information.

[0100] The intelligent data processing unit 103 first preprocesses the received system data, a step that includes data standardization and cleaning. Data standardization refers to converting data from different sources into a uniform format to facilitate analysis; data cleaning identifies and corrects errors or inconsistencies in the data to ensure the accuracy of the analysis. These two processes are crucial for subsequent data analysis because they directly affect the reliability and accuracy of the analysis results.

[0101] After preprocessing, the intelligent data processing unit 103 can analyze the data using a series of advanced algorithms. These algorithms may include modern intelligent technologies such as machine learning, deep learning, and pattern recognition. They can learn the operating patterns of the power system from historical data, thereby accurately judging the current state of the system and predicting potential future problems. For example, by analyzing grid load data and historical fault records, the intelligent data processing unit can predict potential overloads on a certain line or the risk of a transformer failure.

[0102] After analyzing the real-time operating status and fault analysis results of the power system, the intelligent data processing unit 103 transmits this information to the touch screen device 104, enabling operators to understand the system status in real time and take corresponding measures. Furthermore, based on the real-time operating status and fault analysis results, logic control instructions for relay protection are generated and transmitted to the programmable logic controller (PLC). These instructions are then transmitted to the PLC 102, which is responsible for execution, such as adjusting the settings of protection devices or automatically isolating a portion of the power grid when a potential fault is detected.

[0103] Furthermore, the intelligent data processing unit uses a trained neural network model to obtain the real-time operating status and fault analysis results of the power system based on the preprocessed data, and generates logic control instructions for relay protection. The neural network model includes a feature extraction part, a state analysis part, and a decision generation part. The feature extraction part uses a multi-scale convolutional neural network to process the preprocessed data and obtain a high-dimensional feature representation of the time series data, wherein the preprocessed data includes time series data and equipment status index data. The state analysis part uses a bidirectional long short-term memory network to process the high-dimensional feature representation of the time series data provided by the feature extraction part, and obtains the real-time operating status of the power system, fault analysis results, and feature representations for decision generation. The decision generation part uses a graph attention network to process the real-time operating status of the power system, fault analysis results, and feature representations for decision generation provided by the state analysis part, and generates logic control instructions for relay protection.

[0104] In integrated relay protection and management systems based on intelligent algorithms, time-series data and equipment status index data are two key types of input data. They provide the system with rich information for monitoring and analyzing the operating status and potential faults of the power system. The following are specific examples of these two types of data:

[0105] Time series data refers to a series of data points recorded in chronological order, which typically includes the following in power systems:

[0106] Voltage and current data: These data change over time and record the voltage and current levels at various nodes or circuits in the power system. For example, the inlet and outlet voltages of a substation can be recorded every minute.

[0107] Power data: This includes records of active and reactive power, which helps analyze system load and actual energy consumption. For example, the changes in active and reactive power over time on a transmission line.

[0108] Frequency data: Frequency data of the power system reflects the balance of the power grid. For example, recording the changes in system frequency every second can help monitor for deviations from the normal range.

[0109] Equipment status index data provides information on the current operating status and performance of various devices in the power system, including:

[0110] Equipment temperature: such as the oil temperature of transformers and the conductor temperature of transmission lines, these data are crucial for assessing the operational safety of equipment and preventing overheating.

[0111] Equipment operating status: such as switch status (on or off), activation status of protection devices, etc. For example, whether a circuit breaker is in the open state.

[0112] Equipment failure indicators: These include abnormal alarm records and maintenance records of the equipment. This data is crucial for fault prediction and maintenance planning. For example, a record of an overheat alarm issued by a transformer.

[0113] Maintenance logs and fault history: These records provide information on past maintenance and types of faults experienced by the equipment, helping the model learn fault modes and preventative measures. For example, maintenance records and fault occurrences of a generator over the past year.

[0114] By integrating these time-series data and equipment status index data and inputting them into a comprehensive relay protection and management system based on intelligent algorithms, the neural network model can extract high-dimensional feature representations from these data. This allows for the analysis of the real-time operating status of the power system, identification of potential faults, and the generation of corresponding logical control commands, thereby achieving intelligent monitoring and protection of the power system. This data-driven approach provides strong support for the stable operation and fault handling of power systems.

[0115] Furthermore, the feature extraction part includes an input layer, a multi-scale convolutional layer, a pooling layer, a dynamic feature fusion module, and an output layer;

[0116] The input layer is used to receive preprocessed time series data and equipment status index data from the power system, and directly use the time series data and status index as the network input. The equipment status index data is processed by an embedding transformation before input to adapt to the network input requirements.

[0117] The multi-scale convolutional layer includes three different scales of convolutional kernels: 3×3, 5×5, and 7×7. Each scale of convolutional kernel is used to process the data passed from the input layer to extract feature information at different scales. The output of the multi-scale convolutional layer is a combination of feature maps extracted at the three scales, which is completed by concatenating the feature maps along the depth direction.

[0118] The pooling layer uses the max pooling method to downsample the combination of feature maps provided by the multi-scale convolutional layer, reducing the feature dimension while retaining important feature information, thereby reducing the complexity and computational cost of the model.

[0119] The dynamic feature fusion module is implemented through a fully connected layer, which fuses the pooled feature maps output by the pooling layer to obtain a feature vector with dynamically adjusted weights.

[0120] The output layer is used to convert the output of the dynamic feature fusion module into a high-dimensional feature representation of the time series data.

[0121] In the feature extraction section, this embodiment provides a novel multi-scale convolutional neural network (MSCNN) structure, designed to deeply mine the complex features of time-series data and equipment status index data in power systems. This network structure can not only capture multi-level features of power system data, but also adapt to pattern changes at different scales, thereby providing richer and more accurate information for subsequent status analysis and decision generation.

[0122] The core design of Multi-Scale Convolutional Neural Networks (MSCNN) lies in using convolutional kernels of different sizes to process input data in parallel within the same layer. This allows the network to capture patterns from fine-grained to coarse-grained in time-series data. Each scale of convolutional kernel focuses on extracting features of different sizes, enabling the network to comprehensively consider both short-term fluctuations and long-term trends.

[0123] Multi-scale convolutional neural networks include:

[0124] 1. Input Layer: This layer receives preprocessed time-series data and equipment status index data from the power system and uses the time-series data and status indexes directly as network input. The equipment status index data undergoes an embedding transformation process before input to adapt to the network's input requirements.

[0125] In multi-scale convolutional neural networks (MSCNNs), embedding transformation of device status metrics is a common approach to convert these typically discrete or scalar status metrics into high-dimensional vector forms that the network can efficiently process. This transformation not only improves the model's ability to handle discrete data but also enhances its ability to learn complex relationships between status metrics. The following is a concrete implementation example demonstrating how to implement this process in PyTorch.

[0126] Suppose there is a power system with equipment whose status indicators include: equipment operating mode (4 types), temperature level (3 types), and load level (5 types). These status indicators are discrete, and the number of possible values ​​for each indicator is different.

[0127] The steps of the embedding transformation process include:

[0128] 1. Define an embedding layer for each state indicator: For each state indicator, define an embedding layer based on the number of its possible values. The size of the embedding layer (i.e., the dimension of the output vector) can be adjusted according to the actual situation. For example, for the state indicators of 4 operating modes, an embedding layer might be defined, whose output is an 8-dimensional vector.

[0129] 2. Convert state indicators into embedding vectors: Before processing, each state indicator is first converted into an index (usually an integer), and then this index is used as the input of the embedding layer, which looks up the corresponding embedding vector based on the index.

[0130] 3. Concatenate the embedding vectors of all state indicators: Concatenate the embedding vectors obtained from the transformation of all state indicators along a specific dimension to form a unified representation for further processing by subsequent network layers.

[0131] The implementation code based on PyTorch is as follows:

[0132]

[0133]

[0134] This example illustrates how to use PyTorch's `nn.Embedding` module to perform embedding transformations on device state indicators in a power system. By converting discrete state indicators into continuous vector representations, the network can more effectively learn the complex relationships between state indicators, thereby improving the overall model performance.

[0135] 2. Multi-scale convolutional layers: Three convolutional kernels are set: 3×3 (small scale), 5×5 (medium scale), and 7×7 (large scale). Each convolutional kernel extracts patterns of different sizes.

[0136] 3. Pooling layer: Max pooling is used to reduce feature dimensionality while retaining important feature information, which helps to reduce model complexity and computational cost.

[0137] 4. Output layer: After processing by several convolutional and pooling layers, a fully connected layer is used to convert multi-scale feature information into high-dimensional feature representation for further processing by the state analysis part.

[0138] A dynamic feature fusion module is introduced after the multi-scale convolutional layer. This module adjusts the weights of features at different scales to further improve the information content and accuracy of feature representation.

[0139] The following is a reference implementation code for this section, which demonstrates a complete neural network structure including an input layer, multi-scale convolutional layers, pooling layers, a dynamic feature fusion module, and an output layer.

[0140]

[0141]

[0142] print("Output dimensions:",output.shape) # It should be [32, 10]

[0143] This code first implements a multi-scale convolutional layer using `MultiScaleConv`, which includes convolutional operations at three different scales to capture spatial features of varying sizes. Then, it uses the `DynamicFeatureFusion` module to dynamically fuse these features, adjusting the weights of features at different scales through a fully connected layer (`fc`) to optimize feature representation. Finally, it reduces the feature dimensionality through pooling layers and then maps the result to the final output feature space through an output layer.

[0144] The state analysis section employs a bidirectional long short-term memory (LSTM) network to process the high-dimensional feature representations of the time-series data provided by the feature extraction section, thereby obtaining the real-time operating status of the power system, fault analysis results, and feature representations for decision generation. Here, the state analysis section can be implemented using a standard bidirectional LSM network from existing technologies, which will not be elaborated further. The following will introduce an improved LSM network provided in this embodiment.

[0145] Furthermore, the state analysis part of the neural network model employs an improved bidirectional long short-term memory network, which enhances its ability to understand and predict the time dependence of power system data by introducing an adaptive gating mechanism; wherein, the improved bidirectional long short-term memory network is implemented in the following manner:

[0146] A standard bidirectional long short-term memory network structure is adopted to process the forward and backward data sequences simultaneously at each time step t, in order to capture the forward and backward time dependencies in the time series;

[0147] At each time step t, based on the current input x t And the hidden state h of the previous time step t-1 t-1 Dynamically calculate the gating signal g t Wherein, the gating signal g t It is to change the current input x t and hidden state h t-1 The concatenated result is input into a fully connected layer and activated by the sigmoid function, which is used to adaptively adjust the weights for updating the hidden state.

[0148] The update formula for hidden states is expressed as follows:

[0149]

[0150] Among them, h t This represents the hidden state at time step t; ⊙ represents the candidate hidden state calculated at time step t in a standard bidirectional long short-term memory network without the application of a gating signal; ⊙ represents element-wise multiplication.

[0151] In the field of real-time monitoring and fault prediction of power systems, accurate analysis of time-series data is crucial. Therefore, this embodiment proposes an improved bidirectional long short-term memory (Bi-LSTM) network, which significantly improves the understanding and prediction capabilities of the time dependencies of power system data by introducing an adaptive gating mechanism. The following is a detailed description of this improved Bi-LSTM network.

[0152] This network is based on a standard bidirectional long short-term memory (BLSTM) network structure. Bi-LSTM was designed to better capture the temporal dependencies in time-series data. In power system monitoring data, these temporal dependencies are crucial for accurately predicting system status and potential faults.

[0153] Standard Bi-LSTM processes time series data simultaneously in both the forward (past to future) and reverse (future to past) directions by setting up two processing paths. This design allows the network to consider information from both the past and future at each time step t, thus providing a comprehensive view of the data.

[0154] The key innovation lies in the introduction of an adaptive gating mechanism. This mechanism dynamically adjusts the model's emphasis on information at each time step t, thereby achieving more refined state updates. Specifically, this mechanism is implemented through the following steps:

[0155] - Calculation of the gating signal: At each time step t, based on the current input data x t and the hidden state h from the previous moment t-1 Calculate a gate signal g t This calculation is performed by using the current input x. t With hidden state h t-1 The concatenation is performed, and the concatenated result is then passed to a fully connected layer, where the gating signal g is obtained through a sigmoid activation function. t The gating signal is limited to a range of 0 and 1 to dynamically adjust the degree of hidden state updates.

[0156] - Update hidden state: using gating signal g t Combined with the candidate hidden states calculated by standard Bi-LSTM at the current time step And the hidden state h from the previous moment t-1 According to the following formula:

[0157]

[0158] Update the hidden state h of the current time step t Here, ⊙ represents element-wise multiplication, meaning the gate signal g... t The candidate hidden state will be adjusted element by element. Compared to the previous hidden state h t-1 The proportion of its contribution.

[0159] To implement an improved Bi-LSTM model with an adaptive gating mechanism, and to output real-time running status, fault analysis results, and feature representations for decision generation as required, the following is a specific PyTorch code example.

[0160]

[0161]

[0162]

[0163] This code first defines an `AdaptiveGateBiLSTM` class, which dynamically adjusts the model output by adding an adaptive gating signal layer to the traditional Bi-LSTM model. The gating signal is calculated based on the model's current input and the Bi-LSTM output and is used to adjust the hidden layer state updates. Finally, three fully connected layers are established to output the model's real-time operating status of the power system, fault analysis results, and feature representations for subsequent decision generation, respectively, to meet different application requirements.

[0164] In the reference code above, the real-time operating status and fault analysis results output by the model are usually not presented directly in text form, but rather in numerical or vector form. These numerical values ​​or vectors represent specific states or potential fault information of the power system. To convert these numerical values ​​or vectors into easily understandable textual descriptions, the following methods can be used:

[0165] 1. Threshold Determination: For real-time operating status, a series of thresholds can be predefined. Based on the comparison between the model output value and these thresholds, the value is converted into the corresponding state description. For example, if the model output value represents the voltage level, it can be compared with the threshold of normal operating voltage to determine whether the power system is in a normal state, an overload state, or a low voltage state.

[0166] 2. Classification Mapping: For fault analysis results, if the model output is the result of a multi-class classification problem, each class can be directly mapped to a textual description of a specific fault type. For example, if the model output represents multiple possible fault types, each type can correspond to an index. Through the mapping between the index and the predefined fault type description, the numerical results are converted into textual descriptions.

[0167] Suppose the real-time operating status output by the model is a set of values ​​[0.2, 0.8, 0.1], representing the probabilities of three possible states: normal, overload, and low voltage. By selecting the state with the highest probability, the output can be converted into the textual description "overload".

[0168] Similarly, if the fault analysis result is [0.05, 0.95], it represents the probability of two possible fault types: line fault and equipment fault. Select the fault type with the highest probability and convert the output to "equipment fault".

[0169] The above conversion method transforms the numerical values ​​or vectors output by the improved Bi-LSTM model into concrete and intuitive textual descriptions, enabling non-experts to understand the model's analysis results and make corresponding decisions or take appropriate measures. This conversion mechanism is a key step in achieving effective interaction between artificial intelligence systems and human users. After conversion, the fault analysis results and real-time operating status can be transmitted to the touchscreen device.

[0170] Furthermore, the decision generation part of the neural network model is specifically used for:

[0171] The real-time operating status of the power system, fault analysis results, and decision generation feature representations provided by the state analysis section are used as inputs. These inputs reflect the current comprehensive state of the power system and potential fault information. The inputs are integrated into a unified representation to facilitate processing by the graph attention network.

[0172] A graph attention network is used to process the integrated input data. When dealing with complex relationships between nodes, the graph attention network adaptively allocates attention weights, thereby effectively capturing the interdependencies and influences between the components of the power system; the nodes include various components of the power system.

[0173] Based on the results of graph attention network processing, logical control instructions are generated; the logical control instructions provide relay protection operation suggestions for the identified power system status or potential faults, wherein the relay protection operation suggestions include adjusting the load, activating the backup system, or triggering an alarm.

[0174] In integrated relay protection and management systems based on intelligent algorithms, the decision generation component plays a crucial role, directly impacting the power system's response measures and relay protection efficiency. The core of this component lies in its ability to comprehensively utilize the power system's real-time operating status, fault analysis results, and decision generation feature representations. Through sophisticated algorithmic processing, it ultimately generates specific logical control commands for implementing relay protection.

[0175] First, this section collects and integrates key information from the state analysis section, including the real-time operating status of the power system, identified potential faults, and feature representations to support decision generation. This information comprehensively reflects the current state of the power system and the problems it faces, forming the basis for generating accurate control commands.

[0176] Next, by introducing advanced Graph Attention Network (GAT) technology, the system is able to perform in-depth analysis of the integrated input data. The introduction of GAT is particularly valuable for its ability to handle the complex interactions between the various components within a power system. In a power system, complex dependencies exist between different components, and these relationships are crucial for the stable operation of the system. By adaptively allocating attention weights to the relationships between these nodes, GAT ensures that these dependencies are fully considered when analyzing and processing this information, thereby effectively capturing the dynamic changes and potential risks within the power system.

[0177] Finally, based on the results processed by the graph attention network, this decision generation part can generate a series of specific logical control instructions. These instructions will propose corresponding relay protection operation suggestions based on the identified system state and potential faults. These suggestions may include adjusting the system load, activating the backup system to ensure power supply, or triggering an alarm in a timely manner when a potential fault is detected to prevent the fault from spreading and causing greater losses.

[0178] Here's an example. Consider a simplified power system consisting of power plants, substations, transmission lines, and consumption areas. These components form a complex network where each element is not only interconnected but also influences the others, affecting the overall performance and stability of the system. In this system, Graph Attention Networks (GATs) can be used to effectively analyze and understand the complex interactions between these components.

[0179] In a graph attention network, each component of the power system can be represented as a node. For example:

[0180] - Power plant node: Represents the power generation facility in the power system. Its characteristics may include power generation capacity, current power generation, and power generation type (such as hydropower, thermal power, wind power, etc.).

[0181] - Substation node: Represents a substation in the power system. Its characteristics may include transformer capacity, operating status, maintenance information, etc.

[0182] - Transmission line node: Represents the transmission line in the power system. Its characteristics may include line length, voltage level, line load, loss rate, etc.

[0183] - Consumption area node: Represents the final electricity consumption area of ​​the power system. Its characteristics may include current electricity consumption, predicted electricity consumption, historical electricity consumption data, etc.

[0184] The edges in the diagram represent the connections between the components within a power system. These connections reflect not only physical connections but also dependencies between components. For example, transmission lines directly connect power plants and substations, and substations are then connected to consumer areas via other transmission lines.

[0185] Graph Attention Networks (GATs) adaptively emphasize relationships between components that have a greater impact on system stability and performance by assigning attention weights to edges between nodes. For example, during analysis:

[0186] If a transmission line's load is close to its maximum capacity, GAT (Gas Adaptive Tactics Authority) may assign higher attention weights to the edges between this line and its connected power plants and substations. This reflects the importance of this part of the system, prompting system operators or automated control systems to prioritize how to adjust operating strategies to avoid overloading.

[0187] - When predicting potential system failures, if the operating status data of a substation is abnormal, GAT can highlight the potential impact of this abnormal state on the entire system by increasing the attention weight of the edges between the substation node and other nodes (such as power plants and consumption areas).

[0188] The following is sample code for the decision generation part.

[0189]

[0190]

[0191] This code demonstrates how to use a Graph Attention Network (GAT) to process critical information about a power system in the decision generation section and generate logical control instructions. `GATConv` is the graph attention layer, which adaptively allocates attention weights based on the complex relationships between nodes, effectively capturing the interdependencies and influences between the various components of the power system. Here, `edge_index` represents the connection relationships of edges in the graph; it is a key component of the graph structure data, indicating which nodes have direct connections.

[0192] By adding such a graph attention layer to the model, corresponding relay protection logic control commands can be accurately generated based on the real-time operating status of the power system, fault analysis results, and decision generation feature representations to address various situations and potential faults in the system. This approach provides a powerful tool for enhancing the intelligent monitoring and automated management capabilities of power systems.

[0193] The logical control instructions generated in the code are usually represented in numerical form. They represent specific operations or states and require subsequent processing or interpretation modules to convert these numbers into actual control actions or understandable instructions. This numerical form of logical control instructions is easy for computers to process and can be efficiently mapped to a series of predefined operations.

[0194] Assuming the output feature (logic control instructions) has a dimension of 4, this means each logical control instruction is a vector consisting of 4 numbers. For example, the model might generate one of the following logical control instruction vectors:

[0195] [0.1,0.9,0.05,0.05]

[0196] This vector can represent the probability distribution of different action suggestions or state evaluations. In practical applications, these four dimensions can be mapped to predefined actions or states, for example:

[0197] 1. The first dimension represents the probability that the system is operating normally.

[0198] 2. The second dimension represents the probability that the load needs to be adjusted immediately.

[0199] 3. The third dimension represents the probability that the backup system needs to be activated.

[0200] 4. The fourth dimension represents the probability of triggering an alarm.

[0201] Based on the logic control command vector [0.1, 0.9, 0.05, 0.05] in the example above, it can be interpreted that the model suggests the highest probability (0.9) of immediately adjusting the load, while the probability of other operations is lower. Therefore, based on this command vector, the most likely decision is to adjust the load on the power system to prevent potential overload or optimize system performance.

[0202] To translate these numerical instructions into actual control actions, a parsing module or rule engine is typically needed. This module determines which action has a probability exceeding the execution threshold based on preset rules or thresholds, and then outputs corresponding operation suggestions or directly triggers the corresponding control command.

[0203] For example, if the execution threshold is set to 0.8, then an operation will only be considered for execution if its probability exceeds 0.8. Based on the previous example, since the probability of load adjustment is 0.9, which exceeds the execution threshold, the load adjustment operation will be selected for execution.

[0204] In this way, the improved Bi-LSTM model can not only provide real-time status monitoring and fault prediction for power systems, but also generate specific logic control instructions based on the current state of the system and potential fault conditions, thereby realizing intelligent control and management of the power system.

[0205] In integrated relay protection and management systems based on intelligent algorithms, training a pre-built neural network model is a crucial step. This ensures the model can accurately analyze the real-time operating status and fault analysis results of the power system, thereby generating effective logic control commands. The following is a detailed description of the neural network model training steps:

[0206] First, a training dataset needs to be prepared. This dataset should include historical operating data of the power system, such as time-series data and equipment status indicators, as well as the corresponding outputs, namely, the real-time operating status, fault analysis results, and executed logic control commands from the historical data. This data needs to be preprocessed, such as through normalization, to ensure that the data input into the model meets the model's processing requirements.

[0207] Next, the prepared dataset is divided into training, validation, and test sets. This step ensures the model performs well on unseen data, avoiding overfitting. Typically, the data is divided into approximately 70% training set for model training; 15% validation set for parameter tuning and performance evaluation; and 15% test set for final model performance testing.

[0208] When starting model training, appropriate hyperparameters need to be set, including the learning rate, batch size, and number of training epochs. The learning rate determines the step size for updating model parameters, the batch size affects the efficiency and effectiveness of model training, and the number of training epochs determines the total training time. These hyperparameters need to be adjusted according to specific circumstances to obtain the best training results.

[0209] During training, the model gradually adjusts its internal parameters using backpropagation to minimize the difference between the predicted and actual results. This process is repeated across multiple training epochs, each using data from the training set. To monitor the training process and prevent overfitting, the model's performance is periodically evaluated using a validation set. If the model's performance on the validation set stops improving or begins to decline, the training process can be terminated early.

[0210] After training, the model is evaluated using a test set to ensure it has good generalization ability. If the model performs as expected on the test set, it can be applied to practical power system relay protection and management to generate logic control commands based on real-time data.

[0211] Finally, based on the test results and practical application requirements, further adjustments and optimizations to the model may be necessary. This might include adjusting the model's structure, reselecting hyperparameters, or using more or less training data. This is an iterative process aimed at continuously improving the model's performance and accuracy in real-world applications.

[0212] Through this series of training steps, it can be ensured that the neural network model fully learns and understands the operating mechanism and fault modes of the power system, thereby providing strong intelligent support for the stability and security of the power system in practical applications.

[0213] Furthermore, the intelligent data processing unit is configured to automatically adjust the operating parameters of the power system based on the real-time operating status and fault analysis results of the power system, so as to optimize the performance of the power system and prevent potential faults. The operating parameters include power generation, voltage level, or load distribution of the power grid.

[0214] In integrated relay protection and management systems based on intelligent algorithms, the intelligent data processing unit is a core component. One of its main functions is to automatically adjust key operating parameters of the power system using real-time operating status and fault analysis results. These parameters include power generation, voltage levels, and grid load distribution, aiming to optimize the overall performance of the power system and prevent potential faults. The following details how to implement this intelligent data processing unit to ensure that those skilled in the art can implement it according to these guidelines.

[0215] The intelligent data processing unit first needs to receive and process data from various sensors and monitoring devices in the power system. This includes time-series data such as current, voltage, and frequency, as well as the operating status and performance indicators of the equipment. By employing machine learning or deep learning models, the unit can analyze this data in real time to identify the current operating status of the power system, such as whether it is in optimal operating condition, whether there are overloads, underloads, or other abnormalities, and whether there is a risk of a fault occurring or about to occur.

[0216] Based on the above analysis results, the intelligent data processing unit will then take action to automatically adjust key operating parameters of the power system, including but not limited to:

[0217] Power generation adjustment: Automatically adjusts the power generation of the power plant based on current and predicted load demand to ensure a balance between power supply and demand and avoid waste or supply shortages.

[0218] Voltage level regulation: By changing the tap position of the transformer or using voltage regulation equipment, the voltage level in the power grid is regulated to ensure that the voltage at each load point is stable within the specified range, thereby improving power quality and the safe operation of equipment.

[0219] Grid load distribution: Optimizing load distribution in the power grid by adjusting the switching status of transmission lines or reconfiguring the grid topology can reduce transmission losses and improve the operating efficiency and reliability of the power grid.

[0220] The touch screen device 104 is connected to the programmable logic controller and the intelligent data processing unit, and is equipped with a graphical user interface to display the real-time operating status and fault analysis results of the power system to the user; it also allows the user to directly input operation commands and configure the system.

[0221] In the integrated relay protection and management system based on intelligent algorithms provided in this embodiment, the touch screen device 104 plays a crucial role as an information interaction interface. It not only serves as a window for human-machine interaction, allowing operators to intuitively monitor the real-time operating status of the power system, but also displays in real-time fault analysis results and system prediction information derived by the intelligent data processing unit 103. Simultaneously, it provides an operating interface, allowing users to make immediate control decisions and configuration adjustments based on system feedback.

[0222] The touchscreen device 104 is designed to connect to the programmable logic controller 102 and the intelligent data processing unit 103, enabling it to receive and display data and instructions from both units. This connection ensures data flow and interactivity, allowing operators to access all necessary information on an integrated interface for real-time maintenance and troubleshooting decisions.

[0223] To achieve these functions, the touchscreen device 104 is equipped with a high-definition graphical user interface (GUI) that displays information in a user-friendly manner through intuitive visual elements such as charts, indicator lights, alarm messages, and system logs. Furthermore, the GUI design includes multiple operating interfaces, such as system overview, real-time data monitoring, historical data analysis, fault logging, system configuration, and control command input interfaces, which can be easily switched via menus or icons on the touchscreen.

[0224] Operators can input operating commands via a virtual keyboard or buttons on the touchscreen, such as adjusting protection settings, resetting alarms, or switching operating modes. These commands are transmitted via the touchscreen device to the programmable logic controller 102 and the intelligent data processing unit 103 to implement corresponding system adjustments or control actions.

[0225] Furthermore, the touchscreen device includes an alarm system for alerting operators when an abnormal operating state or potential fault of the power system is detected.

[0226] In the integrated relay protection and management system based on intelligent algorithms, the touch screen device not only serves as a human-machine interface, displaying the real-time operating status and fault analysis results of the power system, but also includes a crucial function—an alarm system. The purpose of this alarm system is to promptly alert operators when abnormal operating conditions occur in the power system or when a potential fault is detected, thereby allowing for timely response and handling to avoid potential accidents or system damage. The following is a detailed description of the implementation of this alarm system.

[0227] The alarm system is a component of the touchscreen device, relying on the power system status and fault diagnosis results analyzed by the intelligent data processing unit. The main functions of the alarm system include:

[0228] Anomaly and Fault Detection: Utilizing the results analyzed by the intelligent data processing unit, the alarm system can identify abnormal operating states and potential faults in the power system. This may include abnormal fluctuations in voltage or current, degraded equipment performance, and predicted loads exceeding normal ranges.

[0229] Alarm activation: Once an anomaly or malfunction is detected, the alarm system will immediately alert the operator. This alarm can be a visual signal (such as a flashing alarm icon or color change on a touchscreen), an audio signal (such as an alarm sound or voice prompt), or a combination of both.

[0230] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A comprehensive relay protection and management system based on intelligent algorithms, characterized in that, include: The data exchange unit is equipped with serial communication and Ethernet communication capabilities and is configured to perform bidirectional data exchange with sensors and electrical equipment in the power system. It is used to collect system data in the power system and transmit relay protection control signals generated by the programmable logic controller to the electrical equipment. A programmable logic controller (PLC) is connected to the data exchange unit and configured to receive system data from the data exchange unit; process the received system data according to the logic control instructions provided by the intelligent data processing unit; generate a relay protection control signal; and transmit the generated relay protection control signal to the data exchange unit. Intelligent data processing unit, used to receive system data from programmable logic controller; The received system data is preprocessed to obtain preprocessed data, which includes data standardization and cleaning to ensure the data quality for subsequent analysis. The preprocessed data is then analyzed to obtain the real-time operating status of the power system, and this status is transmitted to the touchscreen device. Based on the preprocessed data, fault analysis is performed on the power system to obtain fault analysis results, which are also transmitted to the touchscreen device. Based on the real-time operating status and the fault analysis results, logic control instructions for relay protection are generated and transmitted to the programmable logic controller (PLC). The touchscreen device, connected to the programmable logic controller and intelligent data processing unit, is equipped with a graphical user interface to display the real-time operating status and fault analysis results of the power system to the user; it also allows the user to directly input operation commands and configure the system. The intelligent data processing unit uses a trained neural network model to obtain the real-time operating status and fault analysis results of the power system based on the preprocessed data, and generates logic control instructions for relay protection. The neural network model includes a feature extraction part, a state analysis part, and a decision generation part. The feature extraction part uses a multi-scale convolutional neural network to process the preprocessed data and obtain a high-dimensional feature representation of the time series data, wherein the preprocessed data includes time series data and equipment status index data. The state analysis part uses a bidirectional long short-term memory network to process the high-dimensional feature representation of the time series data provided by the feature extraction part to obtain the real-time operating status of the power system, fault analysis results, and feature representations for decision generation. The decision generation part uses a graph attention network to process the real-time operating status of the power system, fault analysis results, and feature representations for decision generation provided by the state analysis part to generate logic control instructions for relay protection.

2. The integrated relay protection and management system based on intelligent algorithms according to claim 1, characterized in that, The feature extraction part includes an input layer, a multi-scale convolutional layer, a pooling layer, a dynamic feature fusion module, and an output layer; The input layer is used to receive preprocessed time series data and equipment status index data from the power system, and directly use the time series data and status index as the network input. The equipment status index data is processed by an embedding transformation before input to adapt to the network input requirements. The multi-scale convolutional layer includes three different scales of convolutional kernels: 3×3, 5×5, and 7×7. Each scale of convolutional kernel is used to process the data passed from the input layer to extract feature information at different scales. The output of the multi-scale convolutional layer is a combination of feature maps extracted at the three scales, which is completed by concatenating the feature maps along the depth direction. The pooling layer uses the max pooling method to downsample the combination of feature maps provided by the multi-scale convolutional layer, reducing the feature dimension while retaining important feature information, thereby reducing the complexity and computational cost of the model. The dynamic feature fusion module is implemented through a fully connected layer, which fuses the pooled feature maps output by the pooling layer to obtain a feature vector with dynamically adjusted weights. The output layer is used to convert the output of the dynamic feature fusion module into a high-dimensional feature representation of the time series data.

3. The integrated relay protection and management system based on intelligent algorithms according to claim 1, characterized in that, The state analysis part of the neural network model employs an improved bidirectional long short-term memory network, which enhances its ability to understand and predict the time dependence of power system data by introducing an adaptive gating mechanism; wherein, the improved bidirectional long short-term memory network is implemented in the following manner: A standard bidirectional long short-term memory network structure is adopted to process the forward and backward data sequences simultaneously at each time step t, in order to capture the forward and backward time dependencies in the time series; At each time step t, based on the current input x t And the hidden state h of the previous time step t-1 t-1 Dynamically calculate the gating signal g t Wherein, the gating signal g t It is to change the current input x t and hidden state h t-1 The concatenated result is input into a fully connected layer and activated by the sigmoid function, which is used to adaptively adjust the weights for updating the hidden state. The update formula for hidden states is expressed as follows: ； Among them, h t This represents the hidden state at time step t; ⊙ represents the candidate hidden state calculated at time step t in a standard bidirectional long short-term memory network without the application of a gating signal; ⊙ represents element-wise multiplication.

4. The integrated relay protection and management system based on intelligent algorithms according to claim 1, characterized in that, The decision generation part of the neural network model is specifically used for: The real-time operating status of the power system, fault analysis results, and decision generation feature representations provided by the state analysis section are used as inputs. These inputs reflect the current comprehensive state of the power system and potential fault information. The inputs are integrated into a unified representation to facilitate processing by the graph attention network. A graph attention network is used to process the integrated input data. When dealing with complex relationships between nodes, the graph attention network adaptively allocates attention weights, thereby effectively capturing the interdependencies and influences between the components of the power system; the nodes include various components of the power system. Based on the results of graph attention network processing, logical control instructions are generated; the logical control instructions provide relay protection operation suggestions for the identified power system status or potential faults, wherein the relay protection operation suggestions include adjusting the load, activating the backup system, or triggering an alarm.

5. The integrated relay protection and management system based on intelligent algorithms according to claim 1, characterized in that, The data exchange unit is configured to use an encrypted communication protocol to ensure the security and integrity of data during transmission.

6. The integrated relay protection and management system based on intelligent algorithms according to claim 1, characterized in that, The touchscreen device includes an alarm system for alerting operators when an abnormal operating state or potential fault of the power system is detected.

7. The integrated relay protection and management system based on intelligent algorithms according to claim 1, characterized in that, The intelligent data processing unit is configured to automatically adjust the operating parameters of the power system based on the real-time operating status and fault analysis results of the power system, so as to optimize the performance of the power system and prevent potential faults. The operating parameters include power generation, voltage level, or load distribution of the power grid.

8. The integrated relay protection and management system based on intelligent algorithms according to claim 1, characterized in that, The programmable logic controller is equipped with a self-diagnostic function, which can automatically detect and report its operating status and potential hardware or software faults, so as to enable timely maintenance and repair.

Images