Power grid monitoring bionic robot intelligent decision-making method supporting rule self-learning

Abstract

The invention relates to the technical field of intelligent power grid monitoring, and discloses a power grid monitoring bionic robot intelligent decision-making method supporting rule self-learning, and the method comprises the following steps: S1, sensing and fusing multi-source heterogeneous data; s2, dynamic rule self-evolution and knowledge modeling are carried out; and S3, multi-mode collaborative decision and closed-loop control are carried out. Through a visual acquisition unit, a voice interaction unit and a plurality of industrial communication protocols, such as OPCUA, Modbus TCP and the like, multi-source heterogeneous data such as meter reading, indicator lamp state, mechanical part displacement, monitor voice instructions, equipment operation noise and the like of power grid equipment are comprehensively acquired, cross-system association is established by taking equipment codes as main keys through processing such as wavelet denoising, semantic coding conversion and the like, and multi-source heterogeneous data of the power grid equipment is acquired. A unified data integration view is constructed, and the problem that data acquisition is limited in a traditional monitoring mode is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of smart grid monitoring, and in particular to an intelligent decision-making method for a bionic robot for grid monitoring that supports rule self-learning. Background Art

[0002] With the rapid development and intelligent transformation of power systems, the scale of power grids continues to expand and their structure becomes increasingly complex. Their safe and stable operation is becoming increasingly important for socioeconomic development and people's daily lives. Traditional power grid monitoring relies primarily on fixed sensor networks and manual inspections. However, these methods are gradually exposing their limitations when dealing with the complexity and dynamism of modern power grids.

[0003] After searching, the Chinese patent number CN118081746A discloses a self-learning method and self-learning device for robot friction parameters. The self-learning method includes: obtaining the real-time joint torque, position, velocity, and acceleration of the robot; inputting the real-time joint torque, position, velocity, and acceleration into the robot dynamics model, and calculating the real-time friction force exerted on the robot through the robot dynamics model; inputting the real-time friction force and velocity into the Coulomb viscous friction model, and updating the friction parameters corresponding to the Coulomb viscous friction model in real time through the least squares method with a forgetting factor. The method disclosed by the present invention can solve the problem in the prior art that a set of dynamic parameters cannot meet the requirements of all robots of the same model, nor can it meet the requirements of the same robot at different times.

[0004] However, the above technical solution does not address the rule self-learning requirements unique to the field of power grid monitoring in the power grid monitoring scenario, such as the dynamic optimization of power grid accident anomaly analysis rules and multi-source data fusion strategies, making it difficult to adapt to the intelligent decision-making requirements under complex power grid operation scenarios; in addition, the technical solution lacks the ability to integrate heterogeneous data from multiple systems such as scheduling, centralized control, and operation and maintenance, and cannot build a unified knowledge rule base and real-time information pool. Power grid monitoring requires real-time response to massive data changes, which leads to lags in the dynamic adjustment of rules and self-evolution mechanisms in existing methods, and cannot meet the real-time requirements of power grid accident anomaly root cause reasoning, disposal strategy optimization, etc. Based on this, the present invention designs a power grid monitoring bionic robot intelligent decision-making method that supports rule self-learning to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide an intelligent decision-making method for a bionic robot for power grid monitoring that supports rule self-learning, thereby solving the problem in the background technology that the robot cannot adapt to the operation under complex power grid operation scenarios.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] A bionic robot intelligent decision-making method for power grid monitoring supporting rule self-learning includes the following steps:

[0008] Step S1, multi-source heterogeneous data perception and fusion, includes the following specific processes: Step S1.1, using the target detection algorithm through the visual acquisition unit to identify equipment meter readings, indicator light status, and mechanical component displacement; Step S1.2, using the circular microphone array of the voice interaction unit to collect the monitor's voice commands and equipment operating noise, and using WaveNet voiceprint separation technology to extract power terminology command features; Step S1.3, using the OPC UA protocol to connect to the dispatching D5000 system to obtain real-time telemetry data, using the Modbus TCP protocol to connect to the centralized control E6000 system to obtain alarm signals, and using a direct database connection to obtain equipment ledger data from the OMS / PMS system; Step S1.4, performing wavelet denoising on the telemetry data, performing semantic encoding conversion on the alarm text, and establishing a cross-system association with the equipment code as the primary key;

[0009] Step S2: Dynamic rule self-evolution and knowledge modeling: Build a dynamic weight rule base based on the historical case library, and use a neural network architecture to achieve rule evolution; build a knowledge graph of equipment health status and form a feature model of the equipment's entire life cycle;

[0010] Step S3: Multimodal collaborative decision-making and closed-loop control. When an accident anomaly is triggered, voice, vision, and motion collaborative decision-making are executed; the rule base is dynamically modified according to on-site detection data, a trusted disposal chain is generated, and the equipment status prediction is updated.

[0011] Preferably, the industrial communication protocol in step S1.3 includes: step S1.3.1, using the OPC UA protocol to connect to the scheduling D5000 system and collect telemetry data at a frequency of 10Hz; step S1.3.2, using the Modbus TCP protocol to connect to the centralized control E6000 system to achieve millisecond-level response to alarm signals; step S1.3.3, connecting to the OMS system through JDBC to regularly synchronize equipment maintenance ledger data.

[0012] Preferably, the standardization processing in step S1.4 includes: performing threshold denoising with 5-layer decomposition using the sym5 wavelet basis; generating a 768-dimensional semantic vector for the warning text using the BERT model, and calculating the cosine similarity with historical cases.

[0013] Preferably, the rule evolution implementation method in step S2 includes: step S2.1, constructing a dual neural network architecture, including a strategy generation network, processing time series data features based on LSTM, and outputting an initial disposal strategy; also including a rule evaluation network, evaluating the strategy effect based on DQN, and generating a weight adjustment amount; step S2.2, setting a rule validity attenuation function: Wt=W0×e^(-λt), triggering the elimination mechanism when Wt<0.3; step S2.3, establishing a device association influence node in the knowledge graph, including electrical topology connections and homologous fault propagation paths.

[0014] Preferably, the rule evaluation network adopts a dual reward mechanism, including: immediate reward function R_imm = 0.6×(processing time reduction rate) + 0.4×(equipment damage reduction rate); long-term reward function R_long = 0.7×equipment health score + 0.3×change in rule base entropy value.

[0015] Preferably, the collaborative decision-making of step S3 includes the following steps: step S3.1, the voice layer generates a TTS alarm containing the risk level and receives a voice confirmation instruction; step S3.2, the visual layer displays the three-dimensional anatomical model of the device and the thermal map of the abnormal part on the AR interface; step S3.3, the motion layer controls the robotic arm to perform infrared temperature measurement and partial discharge detection.

[0016] Preferably, the robotic arm detection includes: setting the end contact force threshold F_contact≤0.2×F_device_max; automatically starting redundant detection when the detection data variation coefficient CV>15%; the condition for feedback data triggering rule backtracking training is |detection value-prediction value|>3σ.

[0017] Preferably, the step S3 further includes: generating an accident report chain with a digital signature, including: a hash value of the original data, a video summary of the handling process, and a rule version identifier; updating the equipment remaining life prediction, using an improved Weibull model.

[0018] Preferably, the method also includes initialization configuration: calibrating the vision-robotic arm coordinate conversion matrix, with the error controlled within ±0.1mm; setting the communication heartbeat period: D5000 system ≤50ms, E6000 system ≤100ms; and also includes loading emergency response whitelist rules, including priority response strategies for main transformer overload and busbar pressure loss.

[0019] Preferably, the method further includes a performance optimization and iterative upgrade module, the specific implementation of which includes:

[0020] Performance monitoring and evaluation: During the operation of the method, key performance indicators such as the execution efficiency, data processing speed, and decision accuracy of each step are monitored in real time, and performance evaluation reports are generated regularly; adaptive parameter adjustment: based on the performance evaluation results, the parameters involved in the method are automatically adaptively adjusted; model compression and optimization: the neural network architecture in step S2 is compressed, and pruning, quantization and other technologies are used to reduce the model size without significantly reducing performance, thereby improving the model's operation speed and resource utilization; at the same time, the model's training and inference algorithms are optimized in a targeted manner.

[0021] Compared with the prior art, the present invention has the following beneficial effects:

[0022] 1. In the present invention, a visual acquisition unit, a voice interaction unit, and various industrial communication protocols, such as OPC UA and Modbus TCP, are used to comprehensively collect multi-source heterogeneous data such as meter readings, indicator light status, mechanical component displacement, monitor voice commands, and equipment operating noise of power grid equipment. After processing such as wavelet denoising and semantic coding conversion, cross-system associations are established with the equipment code as the primary key, and a unified data integration view is constructed to address the data acquisition limitations of traditional monitoring methods.

[0023] 2. In the present invention, a dynamic weight rule base is constructed based on the historical case library, and a dual neural network architecture of a strategy generation network and a rule evaluation network is adopted to realize the dynamic evolution of rules. At the same time, a rule validity decay function is set. When the weight decays to a threshold, an elimination mechanism is triggered to ensure continuous optimization of the rule base.

[0024] 3. In the present invention, when an accident abnormality is triggered, the voice layer generates a TTS alarm and receives a voice confirmation instruction. The visual layer displays the three-dimensional anatomical model of the equipment and the thermal map of the abnormal part on the AR interface. The motion layer controls the robotic arm to perform infrared temperature measurement and partial discharge detection. At the same time, the rule base is dynamically corrected according to the on-site detection data to generate a trusted disposal chain and update the equipment status prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a diagram of the overall system architecture of the present invention;

[0026] Figure 2 This is a multi-source data fusion flow chart of the present invention;

[0027] Figure 3 Schematic diagram of the rule self-evolution mechanism of the present invention;

[0028] Figure 4 This is a multimodal collaborative decision-making flow chart of the present invention;

[0029] Figure 5 This is a module diagram for performance optimization and iterative upgrading of the present invention. DETAILED DESCRIPTION

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0031] Example 1;

[0032] See also Figure 1-Figure 5 , a bionic robot intelligent decision-making method for power grid monitoring that supports rule self-learning, including the following steps: Step S1, multi-source heterogeneous data perception and fusion, including the following specific processes: Step S1.1, using the target detection algorithm through the visual acquisition unit to identify the equipment meter readings, indicator light status and mechanical component displacement; Step S1.2, using the circular microphone array of the voice interaction unit to collect the monitor's voice commands and equipment operation noise, and using WaveNet voiceprint separation technology to extract power terminology command features; Step S1.3, using the OPC UA protocol to connect to the scheduling D5000 system to obtain real-time telemetry data, using Modbus The TCP protocol is used to connect to the centralized control system E6000 to obtain alarm signals, and the equipment inventory data of the OMS / PMS system is obtained by direct database connection. In step S1.4, wavelet denoising is performed on the telemetry data, the semantic encoding conversion of the alarm text is performed, and a cross-system association with the equipment code as the primary key is established. In step S2, dynamic rule self-evolution and knowledge modeling are carried out. A dynamic weight rule base is constructed based on the historical case library, and a neural network architecture is used to realize rule evolution. A knowledge graph of equipment health status is constructed to form a feature model of the equipment's entire life cycle. In step S3, multimodal collaborative decision-making and closed-loop control are carried out. When an accident anomaly is triggered, voice, vision, and motion collaborative decision-making are executed. The rule base is dynamically revised according to on-site detection data, a trusted disposal chain is generated, and equipment status predictions are updated.

[0033] The industrial communication protocol in step S1.3 includes: step S1.3.1, using the OPC UA protocol to connect to the scheduling D5000 system and collect telemetry data at a frequency of 10Hz; step S1.3.2, using the Modbus TCP protocol to connect to the centralized control E6000 system to achieve millisecond-level response to alarm signals; step S1.3.3, connecting to the OMS system through JDBC to regularly synchronize equipment maintenance ledger data.

[0034] The rule evolution implementation in step S2 includes: Step S2.1, constructing a dual neural network architecture, including a policy generation network that processes time series data features based on LSTM and outputs an initial response strategy; and a rule evaluation network that evaluates policy effectiveness based on DQN and generates weight adjustments; Step S2.2, setting a rule effectiveness decay function: Wt = W0 × e^(-λt), triggering an elimination mechanism when Wt < 0.3; Step S2.3, establishing device-related impact nodes in the knowledge graph, including electrical topology connections and common fault propagation paths. The rule evaluation network uses a dual reward mechanism, including an immediate reward function R_imm = 0.6 × (response time reduction rate) + 0.4 × (equipment damage reduction rate); and a long-term reward function R_long = 0.7 × device health score + 0.3 × rule base entropy change.

[0035] The working principle of this embodiment of the present invention is as follows: In a power grid monitoring scenario, the method first comprehensively collects various key information through multi-source heterogeneous data perception and fusion. The visual acquisition unit uses target detection algorithms to accurately identify meter readings, indicator light status, and mechanical component displacement of power grid equipment, providing visual evidence for decision-making. The voice interaction unit's circular microphone array captures the monitor's voice commands and equipment operating noise, and uses WaveNet voiceprint separation technology to extract power terminology command features. For data integration, the OPC UA protocol collects real-time telemetry data from the D5000 system at a 10Hz frequency, and the Modbus TCP protocol achieves millisecond-level response to alarm signals from the E6000 system. Simultaneously, a direct database connection to the OM obtains equipment inventory data from the S / PMS system. The collected telemetry data undergoes wavelet denoising, and the alarm text undergoes semantic encoding conversion. Finally, cross-system associations are established using the device code as the primary key, forming a unified data integration view.

[0036] In the dynamic rule evolution and knowledge modeling phase, a dynamic weighted rule base is constructed based on a historical case library to provide rule support for power grid monitoring decisions. A dual neural network architecture is used to drive rule evolution. The strategy generation network uses LSTM to process time series data features and outputs an initial response strategy. The rule evaluation network uses a DQN to evaluate the effectiveness of the strategy and generate and adjust the weights. Furthermore, a rule effectiveness decay function (Wt = W0 × e^(-λt)) is set. When Wt < 0.3, a elimination mechanism is triggered to ensure continuous optimization of the rule base.

[0037] When an abnormality occurs in the power grid, the system enters a multimodal collaborative decision-making and closed-loop control phase. The voice layer generates a text-to-speech (TTS) alert with a risk level and receives voice confirmation commands. The visual layer displays anomalies and thermal maps of the device's 3D anatomical model on an augmented reality interface. The motion layer controls the robotic arm to perform infrared temperature measurement and partial discharge detection, with the end contact force threshold controlled within the range F_contact ≤ 0.2 × F_device_max. Redundancy detection is automatically initiated when the coefficient of variation (CV) of the detection data exceeds 15%. The condition for triggering rule backtracking for feedback data is |detection value - prediction value| > 3σ. A digitally signed incident report chain is generated, including a hash of the original data, a video summary of the handling process, and a rule version identifier. The improved Weibull model is then used to update the remaining life prediction of the equipment.

[0038] Example 2;

[0039] See also Figure 1-Figure 5 In this embodiment of the present invention, the collaborative decision-making process in step S3 includes the following steps: Step S3.1: The voice layer generates a text-to-speech (TTS) alert containing the risk level and receives a voice confirmation command; Step S3.2: The visual layer displays a three-dimensional anatomical model of the device and a thermal map of the abnormal area on an augmented reality (AR) interface; Step S3.3: The motion layer controls the robotic arm to perform infrared temperature measurement and partial discharge detection. Robotic arm detection includes: setting the end contact force threshold F_contact ≤ 0.2 × F_device_max; automatically initiating redundant detection when the coefficient of variation (CV) of the detection data exceeds 15%; and triggering rule backtracking training based on feedback data when |detection value - prediction value| > 3σ. Step S3 also includes: generating a digitally signed incident report chain containing: a hash value of the original data, a video summary of the handling process, and a rule version identifier; and updating the equipment's remaining life prediction using an improved Weibull model.

[0040] The method also includes initialization configuration: calibrating the vision-robot coordinate conversion matrix with an error control of ±0.1mm; setting the communication heartbeat period: ≤50ms for the D5000 system and ≤100ms for the E6000 system; and loading emergency response whitelist rules, including priority response strategies for main transformer overload and busbar voltage loss. The method also includes a performance optimization and iterative upgrade module, whose specific implementation includes: performance monitoring and evaluation. During the operation of the method, key performance indicators such as the execution efficiency, data processing speed, and decision accuracy of each step are monitored in real time, and performance evaluation reports are generated regularly; adaptive parameter adjustment. Based on the performance evaluation results, the parameters involved in the method are automatically and adaptively adjusted; model compression and optimization. The neural network architecture in step S2 is compressed, and pruning, quantization, and other technologies are used to reduce the model size without significantly reducing performance, thereby improving the model's operation speed and resource utilization; and targeted optimization of the model's training and inference algorithms.

[0041] The working principle of this embodiment of the present invention is as follows: During the collaborative decision-making phase, the voice layer not only generates text-to-speech (TTS) alerts containing risk levels but also receives voice confirmation commands, ensuring two-way interaction and confirmation of information transmission. The visual layer intuitively displays a 3D anatomical model of the device and a thermal map of abnormal areas on an augmented reality (AR) interface, helping monitoring personnel quickly locate faults. The motion layer controls the robotic arm to perform infrared temperature measurement and partial discharge detection, strictly controlling the end contact force threshold to avoid damage to the device.

[0042] In addition, this embodiment also includes initialization configuration, such as calibrating the vision-to-robotic arm coordinate conversion matrix with an error control of ±0.1mm, and setting the communication heartbeat period to ensure communication stability with the D5000 and E6000 systems. It also loads emergency response whitelist rules to prioritize handling situations such as main transformer overload and busbar voltage loss. The performance optimization and iterative upgrade module monitors the execution efficiency, data processing speed, and decision accuracy of each step in real time, regularly generates performance evaluation reports, and automatically and adaptively adjusts the parameters in the method based on the evaluation results. It also compresses the neural network architecture, using pruning, quantization, and other techniques to reduce the model size and improve operation speed and resource utilization.

[0043] Example 3;

[0044] See also Figure 1-Figure 5 In the embodiment of the present invention, a 220kV substation monitoring scenario is taken as an example to apply the intelligent decision-making method. The specific parameters and process are as follows:

[0045] During the multi-source heterogeneous data perception and fusion phase, the high-resolution camera in the visual acquisition unit utilizes the YOLOv5 object detection algorithm to identify equipment meter readings 15 times per second with 95% accuracy. This allows for precise identification of subtle changes in meter readings within the range of 0-100 MPa. The system also monitors indicator light status and mechanical component displacement with millimeter-level accuracy. The voice interaction unit's circular microphone array, comprised of eight microphones, captures operator voice commands and equipment operating noise at a 44.1kHz sampling rate. After processing using WaveNet voiceprint separation technology, feature extraction of power terminology commands achieves over 90% accuracy. For data connectivity, the system utilizes the OPC UA protocol to connect to the D5000 dispatching system, collecting real-time telemetry data, including voltage and current, at a 10Hz frequency, with data transmission latency under 50ms. Modbus TCP is used to connect to the E6000 centralized control system to obtain alarm signals, achieving millisecond-level response times of less than 200ms. The system also connects to the OMS via JDBC to synchronize equipment maintenance log data every 15 minutes. The collected telemetry data undergoes wavelet denoising, effectively removing over 90% of high-frequency noise. Alarm text undergoes semantic encoding conversion, establishing a cross-system association with the device code as the primary key. This successfully integrates data from over 200 devices within the substation, creating a unified data integration view.

[0046] During the dynamic rule evolution and knowledge modeling phase, a dynamic weighted rule base was constructed based on over 2,000 historical cases accumulated over the past five years at the substation. Initially, it contained 500 action rules. In the dual neural network architecture employed, the policy generation network is based on an LSTM network with 128 hidden layer neurons and a sequence length of 20. After processing time series data features, it outputs an initial action policy, and policy generation takes approximately 0.5 seconds. The rule evaluation network is based on a DQN network and utilizes Experience Replay technology. Its memory capacity is 10,000, and each training batch consists of 64 samples. Weight adjustments are generated after evaluating policy effectiveness, and the evaluation cycle is repeated every ten policy generations. The rule effectiveness decay function parameter λ is set to 0.1, and a rule elimination mechanism is triggered when the rule weight Wt falls below 0.3. Over the past six months of operation, over 30 ineffective rules have been eliminated. Equipment-related impact nodes are established within the knowledge graph, encompassing the complete electrical topology of the substation's electrical equipment and common common fault propagation paths.

[0047] During the multimodal collaborative decision-making and closed-loop control phase, when an abnormality triggers a power grid incident, such as an abnormally high oil temperature in a substation's main transformer, triggering an alarm, the voice layer immediately generates a text-to-speech (TTS) alert stating "Abnormal main transformer oil temperature, high risk" and receives voice confirmation from the monitor at a 16kHz sampling rate. The visual layer displays a real-time 3D anatomical model of the main transformer on an augmented reality interface, clearly marking the areas with abnormal oil temperature and displaying the temperature distribution as a heat map. The heat map updates at a rate of 10 frames per second. The motion layer controls the robotic arm to perform infrared temperature measurement and partial discharge detection. The contact force at the end of the robotic arm is controlled to within 20% of the maximum allowable contact force of the equipment. The infrared temperature measurement accuracy reaches ±0.5°C, and the partial discharge detection sensitivity is 0.1pC. During the detection process, if the coefficient of variation (CV) of the test data exceeds 15%, or if the ratio of the standard deviation to the mean of five consecutive test data exceeds 15%, redundant detection is automatically initiated, increasing the detection frequency and performing data comparison and verification.

[0048] During the past year's operation, this method has successfully handled more than 30 power grid accidents and anomalies, shortening the average handling time by 40% compared with traditional methods and reducing equipment damage by 35%, thereby improving the intelligence level of power grid monitoring and accident handling efficiency.

[0049] Working Principle: In power grid monitoring scenarios, multi-source heterogeneous data perception and fusion are first performed. Vision acquisition units, voice interaction units, and industrial communication protocols are used to comprehensively collect key information, including equipment meter readings, indicator light status, mechanical component displacement, monitor voice commands, equipment operating noise, telemetry data, alarm signals, and equipment inventory data. After processing, a unified data integration view is formed. Next, a dynamic weighted rule base is constructed based on a historical case library. A dual neural network architecture is used to drive rule evolution, and a rule validity decay function is set to ensure continuous optimization of the rule base. When an abnormality occurs in the power grid, the multimodal collaborative decision-making and closed-loop control phase begins. The voice, vision, and motion layers work together to generate a digitally signed accident report chain and update the equipment's remaining life prediction using an improved Weibull model. Furthermore, the system includes initialization configuration, performance optimization, and iterative upgrade modules to enhance the performance and efficiency of the overall approach.

[0050] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A bionic robot intelligent decision-making method for power grid monitoring supporting rule self-learning, characterized in that: The following steps are involved: Step S1, multi-source heterogeneous data perception and fusion, includes the following specific processes: Step S1.1, using a target detection algorithm through a visual acquisition unit to identify equipment meter readings, indicator light status, and mechanical component displacement; Step S1.2: The voice interaction unit's circular microphone array collects the monitor's voice commands and equipment operating noise, and uses WaveNet voiceprint separation technology to extract power terminology command features. Step S1.3: Connect to the D5000 dispatch system via the OPC UA protocol to obtain real-time telemetry data, connect to the E6000 centralized control system via the Modbus TCP protocol to obtain alarm signals, and obtain equipment inventory data from the OMS / PMS system using a direct database connection. Step S1.4: Perform wavelet denoising on the telemetry data, convert the alarm text into a semantic code, and establish a cross-system association with the device code as the primary key; Step S2: Dynamic rule self-evolution and knowledge modeling: Build a dynamic weight rule base based on the historical case library, and use a neural network architecture to achieve rule evolution; build a knowledge graph of equipment health status and form a feature model of the equipment's entire life cycle; Step S3: Multimodal collaborative decision-making and closed-loop control. When an accident anomaly is triggered, voice, vision, and motion collaborative decision-making are executed; the rule base is dynamically modified according to on-site detection data, a trusted disposal chain is generated, and the equipment status prediction is updated.

2. The method for intelligent decision-making of a bionic robot for power grid monitoring supporting rule self-learning according to claim 1, characterized in that: The industrial communication protocol in step S1.3 includes: In step S1.3.1, the OPC UA protocol is used to connect to the D5000 system and collect telemetry data at a frequency of 10 Hz. Step S1.3.2: Use the Modbus TCP protocol to connect to the centralized control system E6000 to achieve millisecond-level response to alarm signals; Step S1.3.3: Connect to the OMS system through JDBC and synchronize equipment maintenance ledger data regularly.

3. The intelligent decision-making method of a bionic robot for power grid monitoring supporting rule self-learning according to claim 1 is characterized in that: The standardization processing in step S1.4 includes: using the sym5 wavelet basis to perform threshold denoising with 5-layer decomposition; using the BERT model to generate a 768-dimensional semantic vector for the warning text, and calculating the cosine similarity with historical cases.

4. The intelligent decision-making method of a bionic robot for power grid monitoring supporting rule self-learning according to claim 1 is characterized in that: The rule evolution implementation method in step S2 includes: Step S2.1: Build a dual neural network architecture, including a policy generation network that processes time series data features based on LSTM and outputs an initial disposal policy; and a rule evaluation network that evaluates the effectiveness of the policy based on DQN and generates weight adjustments; Step S2.2: Set the rule validity decay function: Wt = W0 × e^(-λt). When Wt < 0.3, the elimination mechanism is triggered. Step S2.3: Establish equipment-related impact nodes in the knowledge graph, including electrical topology connections and homologous fault propagation paths.

5. The intelligent decision-making method of a bionic robot for power grid monitoring supporting rule self-learning according to claim 4 is characterized in that: The rule evaluation network adopts a dual reward mechanism, including: immediate reward function R_imm = 0.6×(processing time reduction rate) + 0.4×(equipment damage reduction rate); long-term reward function R_long = 0.7×equipment health score + 0.3×change in rule base entropy.

6. The method for intelligent decision-making of a bionic robot for power grid monitoring supporting rule self-learning according to claim 1, characterized in that: The collaborative decision-making in step S3 includes the following steps: Step S3.1: The voice layer generates a TTS alert containing the risk level and receives a voice confirmation instruction; Step S3.2: The visual layer displays the 3D anatomical model of the device and the thermal map of the abnormal part on the AR interface; Step S3.3: The motion layer controls the robotic arm to perform infrared temperature measurement and partial discharge detection.

7. The method for intelligent decision-making of a bionic robot for power grid monitoring supporting rule self-learning according to claim 6, characterized in that: The robotic arm detection includes: setting the end contact force threshold F_contact≤0.2×F_device_max; automatically starting redundant detection when the coefficient of variation CV of the detection data is greater than 15%; and the condition for feedback data to trigger rule backtracking training is |detection value-prediction value|>3σ.

8. The bionic robot intelligent decision-making method for power grid monitoring supporting rule self-learning according to claim 1 is characterized in that: The step S3 also includes: generating an accident report chain with a digital signature, including: a hash value of the original data, a video summary of the handling process, and a rule version identifier; updating the equipment remaining life prediction using an improved Weibull model.

9. The bionic robot intelligent decision-making method for power grid monitoring supporting rule self-learning according to claim 1 is characterized in that: The method also includes initialization configuration: calibrating the vision-robotic arm coordinate conversion matrix with an error controlled within ±0.1mm; setting the communication heartbeat period: D5000 system ≤50ms, E6000 system ≤100ms; and loading emergency response whitelist rules, including priority response strategies for main transformer overload and busbar pressure loss.

10. The bionic robot intelligent decision-making method for power grid monitoring supporting rule self-learning according to claim 1, characterized in that: The method also includes a performance optimization and iterative upgrade module, the specific implementation of which includes: Performance monitoring and evaluation: During the method operation, key performance indicators such as the execution efficiency of each step, data processing speed, and decision accuracy are monitored in real time, and performance evaluation reports are generated regularly; Adaptive parameter adjustment: based on the performance evaluation results, the parameters involved in the method are automatically adjusted adaptively; Model compression and optimization: compress the neural network architecture in step S2, and use pruning, quantization and other technologies to reduce the model size without significantly reducing performance, thereby improving the model's running speed and resource utilization; at the same time, specifically optimize the model's training and inference algorithms.

Images