Power plant intelligent safety perception and automatic response system and method based on MCP

The MCP-based intelligent security perception and automated response system enables standardized collection and processing of multi-source data. Combined with a large-scale language model, it facilitates in-depth threat analysis and human-machine collaborative response, thereby resolving cybersecurity threats to the power plant's industrial control system and improving security monitoring efficiency and automated response capabilities.

CN121940156APending Publication Date: 2026-04-28HUANENG POWER INT INC +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202511833710.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-12-08
Publication Date
2026-04-28

AI Technical Summary

Technical Problem

Existing technologies struggle to effectively identify advanced attacks launched using legitimate industrial control protocols. Security system alarms are numerous and scattered, lacking effective correlation analysis, placing a heavy burden on security operations personnel. Traditional security devices struggle to integrate contextual information from industrial control systems and information systems, resulting in low efficiency in automated response.

Method used

The system employs an intelligent security perception and automated response system based on MCP (Multi-Centered Security Processing). Through multi-level detection modules, intelligent analysis modules, and decision execution modules, it achieves standardized collection and processing of multi-source data. Combined with a large-scale language model, it conducts in-depth threat assessment and human-machine collaborative response, and establishes a continuous evolution mechanism.

Benefits of technology

It improved the safety monitoring efficiency and threat response capabilities of the power plant's industrial control system, solved the data silo problem, significantly increased the detection coverage of abnormal behavior, ensured the safety and reliability of critical operations, and formed a complete intelligent security protection closed loop.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121940156A_ABST
    Figure CN121940156A_ABST
Patent Text Reader

Abstract

The invention provides an MCP-based power plant intelligent safety perception and automatic response system and method, and the system comprises a data access module which is used for achieving the standardized collection and processing of multi-source heterogeneous data of a power plant through a multi-class MCP server; the multi-stage detection module is used for realizing comprehensive threat perception by adopting a three-stage mechanism of grammar detection, behavior analysis and correlation analysis; the intelligent analysis module is used for performing deep threat research and judgment and influence evaluation based on a large language model; the decision execution module is used for starting a differentiated man-machine collaborative response process according to the threat level; and the knowledge management module is used for realizing continuous evolution of the system capability through case library learning and a feedback mechanism. According to the invention, intelligent perception, accurate research and judgment and cooperative response of power plant industrial control network security threats are realized, and the automation level and the operation and maintenance efficiency of power plant network security protection are effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of cybersecurity technology for industrial control systems, specifically to an intelligent safety perception and automated response system and method for power plants based on MCP. Background Technology

[0002] Industrial control systems are the core of the safe and stable operation of power plants. Currently, power plants generally deploy multi-layered defense systems, including firewalls and intrusion detection systems (IDS). These systems play a crucial protective role at the boundary between the production control area and the management information area. With the advancement of "smart power plant" construction, the network structure of power plants is shifting from closed and isolated to open and interconnected. While this trend improves operational efficiency, it also exposes power generation control networks to unprecedented cybersecurity threats. Attacks such as advanced persistent threats (APTs) and ransomware can penetrate traditional defense systems and directly cause physical damage to generator units and power grid stability.

[0003] However, existing technical solutions still face several challenges in practice: First, traditional security devices, based on rule-based and feature-based detection methods, struggle to effectively identify advanced attack behaviors that exploit legitimate industrial control protocols and possess normal syntax but anomalous semantics. Second, the large volume and scattered nature of alarm information generated by various security systems, coupled with a lack of effective correlation analysis tools, places significant pressure on security operations personnel. Third, while existing security operations processes include some automation capabilities, the intervention of security experts is still necessary for analysis and decision-making in complex attack scenarios, and response efficiency needs improvement. Furthermore, due to differences in protocols and environments between industrial control systems and information systems, it is often difficult to effectively integrate contextual information such as device operating status during security analysis.

[0004] To address the aforementioned challenges, this invention proposes a power plant intelligent safety perception and automated response system and method based on the Model Context Protocol (MCP). By employing the MCP protocol to construct a standardized data access framework and integrating the reasoning capabilities of a large-scale language model, it enables correlation analysis and intelligent decision support for multi-source safety data, aiming to improve the safety monitoring efficiency and threat response capabilities of power plant industrial control systems. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art, and provides a power plant intelligent safety perception and automated response system and method based on MCP.

[0006] In a first aspect, embodiments of the present invention provide a power plant intelligent safety perception and automated response system based on MCP, the system comprising: The data access module is used to achieve standardized collection and processing of multi-source heterogeneous data through three types of MCP servers; A multi-level detection module is used to achieve comprehensive threat perception through a three-level mechanism of syntax detection, behavior analysis, and correlation analysis. The intelligent analysis module is used for in-depth threat assessment and impact evaluation based on LLM and power generation knowledge base; The decision execution module is used to initiate differentiated human-machine collaborative response processes based on the threat level. The knowledge management module is used to enable the continuous evolution of system capabilities through a case study learning and feedback mechanism.

[0007] Secondly, embodiments of the present invention provide a method for intelligent safety perception and automated response in power plants based on MCP, the method comprising: Multi-data standardization access based on MCP protocol: By deploying three types of MCP servers—industrial control protocol, equipment status, and security log—unified collection and standardized processing of multi-source heterogeneous data from power plants can be achieved, solving the data silo problem; Multi-level collaborative security threat perception: Employing a three-level detection mechanism of syntax detection, behavior analysis, and correlation analysis, it achieves comprehensive perception of known attacks and abnormal behaviors, thereby improving detection coverage; LLM-based deep threat assessment: Utilizing the reasoning capabilities of large-scale language models and combining them with a power generation scenario knowledge base to conduct multi-dimensional correlation analysis, identify attack intentions, and assess the degree of impact; Human-machine collaborative intelligent decision-making and response: Initiating differentiated response processes based on threat levels, and ensuring the security and reliability of critical operations through a combination of manual approval and automated execution; Continuously evolving system optimization mechanism: Based on the case handling library, the detection model and response strategy are continuously optimized, and the system protection capability is adaptively improved through a feedback learning mechanism.

[0008] In conjunction with the second aspect, the standardized access to multiple data types based on the MCP protocol includes: Three types of dedicated MCP servers are deployed. Among them, the industrial control protocol MCP server adopts deep packet inspection technology to parse the syntax and semantic information of industrial control protocols such as IEC 104 and Modbus TCP in real time. The equipment status MCP server collects operating parameters of key equipment such as generator sets and transformers through the OPC UA interface and establishes equipment status data models; The security log MCP server adopts a dual access mode of Syslog and API, aggregating alarm information from security devices such as firewalls and IDS; Each MCP server outputs data in a unified JSON format and uses an asynchronous communication mechanism to ensure the real-time performance and integrity of the data collection.

[0009] In conjunction with the second aspect, the multi-level collaborative security threat perception includes: First, protocol syntax level checks are performed. An anomaly detection rule base is established based on industrial control protocol specifications to detect basic anomalies such as function code anomalies and address out-of-bounds errors. Secondly, behavioral analysis is implemented, using a machine learning-based traffic baseline model to detect abnormal behavior patterns such as communication frequency and data packet size in real time through a sliding time window mechanism. Finally, correlation analysis is performed to identify coordinated attack events based on pre-defined cross-system correlation rules. The three-level detection results are then comprehensively evaluated using a weighted fusion algorithm to output security events with confidence scores.

[0010] In conjunction with the second aspect, the LLM-based deep threat assessment includes: Construct prompt word templates optimized for power generation scenarios and use a chain-of-thought reasoning model to guide the LLM analysis process; LLM comprehensively analyzes multi-source contextual information, including network traffic characteristics, device status change timing, and security alarm correlation. Referencing the MITRE ATT&CK for ICS framework, attack techniques are identified and categorized. By combining equipment operating status and power generation process, the potential physical impact and economic losses of security threats are assessed, and a threat assessment report including confidence levels is generated.

[0011] In conjunction with the second aspect, the human-machine collaborative intelligent decision-making response includes: Establish a tiered response strategy library and configure differentiated response processes based on threat levels; For low- to medium-risk events, the system automatically executes handling operations through the MCP action server; For high-risk operations, a manual approval process is initiated to provide operations and maintenance personnel with detailed analysis reports and handling suggestions; Digital signatures and timestamps are used to ensure the traceability of operations; All operation processes and results are recorded in the audit log.

[0012] In conjunction with the second aspect, the continuously evolving system optimization mechanism includes: Establish a case knowledge base and use a graph database to store data on security incidents and their entire handling process; Regularly use new case data to incrementally train the detection model and optimize the model parameters; Based on feedback on the treatment effect, a reinforcement learning algorithm is used to adjust the weights of the response strategy; Establish a system performance evaluation index system, generate optimization evaluation reports regularly, and guide the continuous improvement of the system.

[0013] Thirdly, embodiments of the present invention provide an electronic device, comprising: One or more processors; A storage unit is used to store one or more programs, which, when executed by one or more processors, enable the one or more processors to implement any of the above-mentioned MCP-based intelligent safety perception and automated response methods for power plants.

[0014] Fourthly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, can implement any of the above-mentioned methods for intelligent safety perception and automated response of power plants based on MCP.

[0015] Compared with existing technologies, this invention provides an intelligent safety perception and automated response system and method for power plants based on MCP (Multi-Center Security Processing). Through a data access module, it achieves standardized collection and processing of multi-source heterogeneous data, effectively solving the data silo problem in power plants. Through a multi-level detection module, it constructs a multi-layered security threat perception system, significantly improving the detection coverage of abnormal behaviors. Through an intelligent analysis module, it introduces the reasoning capabilities of a large-scale language model, enabling in-depth analysis and impact assessment of security threats. Through a decision execution module, it establishes a human-machine collaborative response mechanism, ensuring the safety and reliability of critical operations. Finally, through a knowledge management module, it achieves the system's continuous evolution capability, forming a complete intelligent security protection closed loop. This method effectively improves the network security protection level and security operation and maintenance efficiency of power plant industrial control systems, providing a guarantee for the stable operation and emergency response of power plants. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 A schematic diagram of the overall architecture of the MCP-based intelligent safety perception and automated response system for power plants provided in an embodiment of the present invention; Figure 2 A flowchart illustrating the MCP-based intelligent safety perception and automated response method for power plants provided in this embodiment of the invention; Figure 3 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] Unless otherwise specifically stated, the technical or scientific terms used in the embodiments of this invention should be understood in their ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains. The terms "comprising" or "including," as used in the embodiments of this invention, do not limit the shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof mentioned, nor do they exclude the appearance or addition of one or more other different shapes, numbers, steps, actions, operations, components, elements, and / or groups thereof, or the inclusion of these.

[0020] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale, and techniques, methods, and apparatus known to those skilled in the art may not be discussed in detail; however, where appropriate, the illustrated techniques, methods, and apparatus should be considered part of the specification. In all the examples shown and discussed herein, any other specific example may have different values. It should be noted that similar symbols and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0021] In the description of the embodiments of the present invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In the embodiments of the present invention, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in the embodiments of the present invention, as well as the features of different embodiments or examples.

[0022] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0023] like Figure 1 The diagram shown is a structural schematic of an intelligent safety perception and automated response system for power plants based on MCP, according to an embodiment of the present invention. The system includes: Data access module 110 is used to achieve standardized collection and processing of multi-source heterogeneous data through three types of MCP servers; The multi-level detection module 120 is used to achieve comprehensive threat perception through a three-level mechanism of syntax detection, behavior analysis and correlation analysis. The intelligent analysis module 130 is used for in-depth threat analysis and impact assessment based on LLM and power generation knowledge base; The decision execution module 140 is used to initiate differentiated human-machine collaborative response processes based on the threat level; The knowledge management module 150 is used to enable the continuous evolution of system capabilities through a case study learning and feedback mechanism.

[0024] In the specific implementation of the data access module 110, three types of dedicated MCP servers are deployed to achieve multi-source data acquisition.

[0025] The MCP server for industrial control protocols is developed based on the DPDK (Data Plane Development Kit) high-performance packet processing framework. It employs zero-copy technology to process network packets and supports deep parsing of protocols such as IEC 104, Modbus TCP, and OPC UA. The server automatically identifies protocol types using protocol fingerprinting technology and extracts key fields such as function codes, data addresses, and data values ​​using specific protocol parsing plugins. The parsing engine uses a multi-threaded architecture, with each thread processing a data stream independently, ensuring parsing latency is less than 10 milliseconds. The server maintains a protocol state table, recording the context information of each session, and supports cross-packet content correlation analysis.

[0026] The equipment status MCP server adopts an OPC UA Client architecture and can be implemented based on the open62541 open-source library. Configured with a 500ms sampling period, it collects over 200 operating parameters of the generator set in real time, including temperature, pressure, and vibration, through an OPC UA subscription mechanism. The server has built-in data preprocessing functions, including data validity checks, unit conversion, and data smoothing. An equipment status data model is established, organizing data in a hierarchical structure, including three levels: equipment level, subsystem level, and component level. The data model supports real-time calculation of equipment health status, classifying status (normal, warning, abnormal) based on set threshold ranges.

[0027] The security log MCP server supports three access methods: Syslog, Kafka, and RESTful API. The Syslog receiver supports the RFC5424 standard and can handle structured data; the Kafka consumer supports parallel consumption of multiple topics; and the RESTful API interface uses OAuth 2.0 authentication. The log parser includes more than 10 parsing plugins, supporting the parsing of log formats from mainstream security devices, including firewalls, IDS, and WAF. The parsing process includes log standardization, field extraction, and event classification, outputting data in a unified JSON format.

[0028] All MCP servers use a unified JSON Schema to define the data format, including required and optional fields. Data is transmitted asynchronously via message queue middleware such as RabbitMQ, and message routing is implemented using topic exchanges. Producers are configured with message acknowledgment mechanisms to ensure no data loss; consumers use a load-balanced model to support horizontal scaling. The message queue cluster is deployed in a mirrored mode to ensure high availability.

[0029] The multi-level detection module 120 adopts a three-level detection mechanism to achieve comprehensive threat perception.

[0030] The syntax detection layer is based on the Snort rule engine, employing an improved rule syntax to support specific detection conditions for industrial control protocols. It is configured with over 200 industrial control protocol anomaly detection rules, covering types such as function code anomalies, address out-of-bounds errors, and data range anomalies. The rule management system supports online rule updates and hot reloading, ensuring that ongoing detection tasks are not affected during updates. The detection engine uses a multi-pattern matching algorithm, supporting high-speed rule matching.

[0031] The behavior analysis layer employs an LSTM-based traffic baseline model, using a sliding time window (configurable from 1 to 5 minutes) to analyze traffic characteristics in real time. Model inputs include features such as packets per second, packet length distribution, and protocol type distribution. During training, historical normal traffic data is used to establish the traffic baseline using unsupervised learning. In the detection phase, the deviation between real-time traffic and the baseline is calculated, and a dynamic threshold is used to identify anomalies. The model supports online updates and is periodically retrained using new normal traffic data.

[0032] The correlation analysis layer employs the Drools rule engine, supporting the definition of complex cross-system correlation rules. Rules are written in DRL (Drools Rule Language) and support complex conditions such as time windows, event sequences, and logical operations. The rule engine operates in inference mode, supporting the dynamic insertion and revocation of facts. Correlation analysis includes multiple levels, such as event correlation within the same protocol, event correlation between different protocols, and cross-system event correlation.

[0033] The Level 3 detection results are comprehensively judged using a confidence-based weighted fusion algorithm. Each detection layer outputs a detection result and a corresponding confidence score, and the fusion algorithm calculates the final result based on the confidence weights. The weight parameters can be adjusted using training data, supporting optimized configurations for different scenarios. Standardized security event descriptions are output, including event type, severity level, occurrence time, and related devices.

[0034] The Intelligent Analysis Module 130 uses LLM to perform in-depth threat assessment.

[0035] A finely tuned LLaMA2-13B model is used as the analysis engine, and further pre-training is performed using text data from the power generation field to improve domain adaptability. The model is deployed on a GPU cluster, employing model parallelism to address GPU memory limitations. The inference service provides a gRPC interface, supporting batch request processing and streaming responses.

[0036] A professional knowledge base containing information on power generation equipment parameters, technological processes, and safety standards has been constructed, exceeding 100,000 entities. The knowledge base is stored using a graph database, with entity types including equipment, parameters, processes, and standards, and relationship types including membership, influence, and dependency. The knowledge base supports multi-dimensional queries, including attribute queries, relationship queries, and path queries.

[0037] The design incorporates a thought-chain-based prompt template, guiding a 20-step analysis process. The prompts are structured, including context descriptions, analysis tasks, and output requirements. Each analysis step corresponds to a specific sub-task, such as attack identification, impact analysis, and remediation recommendations. The prompt template supports parameterized configuration, dynamically generating prompt content based on specific events.

[0038] It integrates the MITRE ATT&CK for ICS attack knowledge base, supporting automatic classification of over 100 attack techniques. An attack technique signature database has been established, containing information such as characteristic indicators, implementation steps, and detection methods for each attack technique. The classification algorithm employs a multi-classification model, combining rule matching and machine learning classification to improve classification accuracy.

[0039] An equipment impact assessment model was established, employing the analytic hierarchy process (AHP) to quantify the potential impact of safety incidents on equipment lifespan, power generation, and other aspects. The assessment model considers multiple dimensions of influencing factors, including equipment importance, scope of impact, and recovery difficulty. Each influencing factor is assigned a weight parameter, determined through a combination of expert scoring and statistical data analysis. The assessment results include a confidence score, calculated based on the sufficiency of evidence and data quality.

[0040] The decision execution module 140 implements a human-machine collaborative response mechanism.

[0041] A tiered strategy library with six response levels has been established, and over 50 preset response scripts have been configured. Each response level corresponds to a different response speed and handling intensity, ranging from Level 1 information notification to Level 6 emergency shutdown. Response scripts are defined in YAML format and include elements such as trigger conditions, execution actions, and expected effects. The strategy management system supports script version management and canary releases.

[0042] Develop an MCP action server that encapsulates operation interfaces for over 10 types of security devices. The interface adaptation layer supports multiple communication protocols, including SSH, NETCONF, and RESTful. Each operation interface provides atomic operation commands, supporting command combination and parameterized calls. The execution engine provides transaction support to ensure the atomicity and consistency of operations. Response latency is less than 100 milliseconds, and concurrent operation processing is supported.

[0043] The manual approval workflow is designed using the BPMN (Business Process Model and Notation) standard to define the approval process. The workflow supports advanced features such as conditional branching, parallel approval, and timeout handling. The approval interface provides detailed analysis reports and handling suggestions, including event details, impact analysis, and handling plans. It supports multi-person review and digital signature approval, with approval opinions recorded in a structured manner.

[0044] Digital signatures are generated using the SM2 national cryptographic algorithm, and the signing process includes hash calculation, signature generation, and signature verification. The signature key is protected by a Hardware Security Module (HSM), supporting key rotation and access control. All operation records are stored in a Fabric-based blockchain evidence storage system, employing a Byzantine fault-tolerant consensus algorithm to ensure data consistency. The stored data includes operation content, execution results, and signature information, supporting post-event auditing and traceability.

[0045] The knowledge management module 150 enables the continuous evolution of system capabilities.

[0046] The Neo4j graph database is used to store security incident cases, constructing a knowledge graph containing entity relationships such as events, devices, and operations. The graph schema includes over 20 entity types and over 30 relationship types, supporting attribute graph queries. The graph computing engine supports complex network analysis, including algorithms for path discovery, community detection, and influence propagation.

[0047] An incremental training mechanism was established, using new data weekly to fine-tune and optimize the detection model and the LLM model. Training data underwent rigorous quality control, including data cleaning, annotation validation, and sample balancing. Incremental training employed online learning algorithms, supporting hot updates of model parameters to avoid service interruptions. Training process monitoring included metrics such as loss function, accuracy, and convergence speed.

[0048] We design a reinforcement learning algorithm based on Q-learning, which automatically adjusts the weights of response policies based on feedback from the treatment effect. The state space includes dimensions such as threat level, system state, and environmental factors, while the action space includes various response measures. The reward function is designed based on the treatment effect and treatment cost. The learning process employs an ε-greedy strategy to balance exploration and exploitation, with the learning rate dynamically adjusted.

[0049] A performance evaluation system has been established, encompassing over 10 indicators, including detection accuracy and response time. Each indicator has a clearly defined calculation formula and data collection method, and the evaluation cycle is configurable. The evaluation report employs multi-dimensional analysis, including time trend analysis, scenario comparison analysis, and root cause analysis. The report generation process is automated and supports customized reports and visualizations.

[0050] like Figure 2 The diagram shown is a flowchart illustrating an MCP-based intelligent safety perception and automated response method for power plants according to an embodiment of the present invention, including: S201: Multi-source data standardization access based on MCP protocol: By deploying three types of MCP servers—industrial control protocol, equipment status, and security log—unified collection and standardized processing of multi-source heterogeneous data from power plants can be achieved, solving the data silo problem; S202: Multi-level collaborative security threat perception: It adopts a three-level detection mechanism of syntax detection, behavior analysis and correlation analysis to achieve comprehensive perception of known attacks and abnormal behaviors and improve detection coverage; S203: LLM-based in-depth threat assessment: Utilizing the reasoning capabilities of large-scale language models, combined with a power generation scenario knowledge base, to conduct multi-dimensional correlation analysis, identify attack intentions, and assess the degree of impact; S204: Intelligent decision-making and response through human-machine collaboration: Initiate differentiated response processes based on threat levels, and ensure the security and reliability of critical operations through a combination of manual approval and automated execution; S205: Continuously evolving system optimization mechanism: Based on the case handling library, the detection model and response strategy are continuously optimized, and the system protection capability is adaptively improved through the feedback learning mechanism.

[0051] Furthermore, the standardized access to multiple data types based on the MCP protocol includes: Specifically, three types of dedicated MCP server clusters are deployed. The industrial control protocol MCP server uses deep packet inspection technology to perform real-time parsing of mainstream industrial control protocols such as IEC 104 and Modbus TCP, extracting protocol fields and payload data, and maintaining session state tracking. The equipment status MCP server, based on the OPC UA interface specification, establishes a stable connection with the monitoring system, collects operating parameters of key equipment such as generator sets and transformers according to the configured cycle, and constructs equipment status data models. The security log MCP server adopts a multi-protocol access method, aggregating alarm and log information generated by various security devices through a Syslog receiver and a RESTful API interface.

[0052] Specifically, data standardization is implemented. Various MCP servers convert the collected raw data into a unified JSON format, using a predefined data schema to ensure format consistency. The data schema includes required and optional fields. Required fields include metadata such as data source, timestamp, and data type, while optional fields are dynamically expanded based on the specific data type.

[0053] Specifically, a reliable data transmission mechanism is established. Asynchronous data transmission is achieved using message queues, a producer-configured acknowledgment mechanism ensures no data loss, and a load-balancing approach is employed by consumers to improve processing capacity. Simultaneously, data quality monitoring indicators are established to continuously monitor the completeness, timeliness, and accuracy of data collection.

[0054] Furthermore, the multi-level collaborative security threat perception includes: Specifically, syntax checking is implemented, and an anomaly detection rule base is established based on the industrial control protocol specification. The rule base includes detection rules for function code anomalies, address out-of-bounds errors, data range anomalies, etc., and supports online updates and hot reloading.

[0055] Specifically, behavioral analysis is implemented, and machine learning methods are used to establish a traffic baseline model. A sliding time window mechanism is used to analyze traffic characteristics in real time and detect abnormal communication patterns. Historical normal traffic data is used during the model training phase, and the deviation between real-time traffic and the baseline is calculated during the detection phase.

[0056] Specifically, correlation analysis is implemented, using a rule engine to perform cross-system correlation analysis. By defining complex correlation rules, coordinated attack events are identified. The three-level detection results are comprehensively judged using a weighted fusion algorithm, and a security event report with a confidence score is output.

[0057] Furthermore, the LLM-based deep threat assessment includes: Specifically, a domain knowledge base is constructed by systematically collecting and organizing professional knowledge in the power generation field, including structured data such as equipment technical parameters, typical process flows, and safety production standards. Knowledge graph technology is used for storage and management to establish semantic relationships between entities and support multi-hop queries and semantic reasoning.

[0058] Specifically, a thought chain prompt template is designed, and a phased, structured prompt engineering method is adopted to guide the large language model to analyze according to the reasoning process of "attack identification - evidence association - impact assessment". Each stage includes clear input and output specifications and contextual constraints to ensure the logic and interpretability of the analysis process.

[0059] Specifically, an attack pattern recognition mechanism is established. Referring to international standard frameworks such as MITRE ATT&CK for ICS, an attack tactics knowledge base is constructed. A deep learning-based multi-classification model combined with rule matching is used to accurately identify and classify attack behaviors, thereby improving the accuracy of threat attribution.

[0060] Specifically, an impact assessment model is developed, and a multi-dimensional assessment index system is constructed based on the analytic hierarchy process. This model comprehensively considers multiple assessment dimensions such as equipment safety operation, production process flow, and economic loss. The degree of threat impact is calculated using a combination of quantitative and qualitative methods, and a comprehensive assessment report with confidence scores is output.

[0061] Furthermore, the human-machine collaborative intelligent decision-making response includes: Specifically, a tiered response strategy library is established, adopting the design principle of matching threat levels with response measures. Response strategies are formulated at multiple levels from low to high, including differentiated handling measures such as information notification, early warning notification, equipment isolation, and system network disconnection. The triggering conditions and execution standards for each level are also clearly defined.

[0062] Specifically, the automated execution process is designed by defining standardized device operation interface specifications, encapsulating complex security operations into reusable atomic operations, and supporting operation combination and process orchestration through a workflow engine to ensure the reliability and consistency of automated execution.

[0063] Specifically, a manual approval mechanism is constructed, with multi-level review processes designed for high-risk operations. Digital certificate-based identity authentication and digital signature technologies are used to ensure the non-repudiation and integrity of the approval process, while providing complete decision-making basis and background information to support manual judgment.

[0064] Specifically, operational audit tracking is implemented, a full lifecycle operational log recording mechanism is established, key information such as operational instructions, execution environment, and operational results are recorded in detail, and anti-tampering technologies such as blockchain are used to ensure the integrity and verifiability of audit data, supporting post-event traceability and responsibility determination.

[0065] Furthermore, the continuously evolving system optimization mechanism includes: Specifically, a case knowledge base is constructed, which uses graph database technology to store the full lifecycle data of security incident cases, including multi-dimensional information such as event characteristics, handling process, and handling results. Semantic relationships between cases are established through knowledge graph modeling methods, supporting case retrieval and reuse based on similarity.

[0066] Specifically, an incremental training mechanism for the model is established, a regular model update plan is formulated, new case data is transformed into training samples through online learning technology, and the performance of the detection model and analysis model is continuously optimized by incremental training. At the same time, a model version management and rollback mechanism is established to ensure system stability.

[0067] Specifically, a reinforcement learning algorithm is designed to model the safety response process as a Markov decision process. A reward function with multiple dimensions, including the effect of the response, the cost of the response, and the impact on the system, is defined. The response strategy parameters are automatically optimized through a policy iteration algorithm to achieve the optimal balance between the effect of the response and the cost.

[0068] Specifically, improve the performance evaluation system, establish evaluation indicators that include multiple dimensions such as detection accuracy, response time, and false alarm rate, use automated evaluation tools to regularly generate system performance analysis reports, and use data-driven methods to identify optimization directions to guide the continuous improvement and iterative upgrade of the system.

[0069] This invention provides an intelligent safety perception and automated response method for power plants based on MCP (Multi-Center Security Processing). Through a data access module, it achieves standardized collection and processing of multi-source heterogeneous data, effectively solving the data silo problem in power plants. A multi-level detection module constructs a multi-layered security threat perception system, significantly improving the detection coverage of abnormal behaviors. An intelligent analysis module introduces the reasoning capabilities of a large-scale language model, enabling in-depth analysis and impact assessment of security threats. A decision execution module establishes a human-machine collaborative response mechanism, ensuring the safety and reliability of critical operations. Finally, a knowledge management module enables continuous system evolution, forming a complete intelligent security protection closed loop. This method effectively improves the network security protection level and operational efficiency of power plant industrial control systems, providing a guarantee for stable operation and emergency response in power plants.

[0070] Electronic device 300 can be a desktop computer, laptop, handheld computer, cloud server, or other electronic device. Electronic device 300 may include, but is not limited to, processor 301 and memory 302. Those skilled in the art will understand that... Figure 3 This is merely an example of electronic device 300 and does not constitute a limitation on electronic device 300. It may include more or fewer components than shown, or combine certain components, or different components. For example, electronic device may also include input / output devices, network access devices, buses, etc.

[0071] Processor 301 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0072] The memory 302 can be an internal storage unit of the electronic device 300, such as a hard disk or RAM of the electronic device 300. The memory 302 can also be an external storage device of the electronic device 300, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the electronic device 300. Furthermore, the memory 302 can include both internal and external storage units of the electronic device 300. The memory 302 is used to store the computer program 303 and other programs and data required by the electronic device. The memory 302 can also be used to temporarily store data that has been output or will be output.

[0073] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0074] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0075] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments of the invention herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of the invention.

[0076] In the embodiments provided by this invention, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. Multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0077] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0078] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0079] If integrated modules / units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program may include computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. Computer-readable media may include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in a computer-readable medium may be appropriately added to or subtracted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0080] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.

Claims

1. A power plant intelligent safety perception and automated response system based on MCP, characterized in that, The system includes: The data access module is used to achieve standardized collection and processing of multi-source heterogeneous data through three types of MCP servers; A multi-level detection module is used to achieve comprehensive threat perception through a three-level mechanism of syntax detection, behavior analysis, and correlation analysis. The intelligent analysis module is used for in-depth threat assessment and impact evaluation based on LLM and power generation knowledge base; The decision execution module is used to initiate differentiated human-machine collaborative response processes based on the threat level; The knowledge management module is used to enable the continuous evolution of system capabilities through a case study learning and feedback mechanism.

2. The intelligent safety perception and automated response system for power plants based on MCP as described in claim 1, characterized in that, In the data access module: The industrial control protocol MCP server uses deep packet inspection technology to parse the syntax and semantic information of IEC 104 and Modbus TCP industrial control protocols in real time, and extract key fields such as function codes, data addresses and data values. The equipment status MCP server collects temperature, pressure, and vibration operating parameters of key equipment such as generator sets and transformers through the OPC UA interface, and establishes an equipment status data model. The security log MCP server adopts a dual access mode of Syslog and RESTful API, aggregating alarm information from security devices such as firewalls and IDS, and supports parsing of multiple log formats.

3. The intelligent safety perception and automated response system for power plants based on MCP as described in claim 1, characterized in that, The multi-level detection module includes: The syntax detection layer establishes an anomaly detection rule base based on industrial control protocol specifications to detect basic anomalies such as function code anomalies and address out-of-bounds errors. The behavior analysis layer adopts an LSTM-based traffic baseline model and uses a sliding time window mechanism to detect abnormal behavior patterns in communication frequency and data packet size in real time. The correlation analysis layer, based on the Drools rule engine, enables cross-system correlation analysis to detect collaborative attack events. The results of the Level 3 detection are comprehensively judged using a confidence-based weighted fusion algorithm.

4. The intelligent safety perception and automated response system for power plants based on MCP according to any one of claims 1 to 3, characterized in that, The intelligent analysis module includes: The domain knowledge base integrates professional knowledge of power generation equipment parameters, process flow and safety standards, and uses knowledge graph technology for storage and management. The mind chain prompt template adopts a phased structured design to guide large language models to analyze according to the process of attack identification - evidence association - impact assessment; The attack pattern identification mechanism, referencing the MITRE ATT&CK for ICS framework, employs a multi-classification model combined with rule matching methods for classification and attribution. The impact assessment model, based on the analytic hierarchy process, quantifies the degree of threat impact from the dimensions of equipment safety, production operation, and economic benefits.

5. The intelligent safety perception and automated response system for power plants based on MCP according to claim 1, characterized in that, The decision execution module includes: A tiered response strategy library allows for the configuration of differentiated response processes based on threat levels; The automated execution process encapsulates atomic operation instructions for various security devices through standardized interfaces. The manual approval mechanism employs digital signature technology to provide a multi-level review process for high-risk operations. Operation audit tracking uses blockchain technology to record operation content, execution results, and information about the operators throughout the entire process.

6. The intelligent safety perception and automated response system for power plants based on MCP according to any one of claims 1 to 3, characterized in that, The knowledge management module includes: The case knowledge base uses the Neo4j graph database to store data on security incidents and their handling processes, and constructs a knowledge graph that includes relationships between events, devices, and operational entities. The model incremental training mechanism uses new case data to incrementally train and optimize the detection model and LLM model every week; The reinforcement learning algorithm, based on Q-learning, automatically adjusts the weights of the response strategy according to the feedback of the treatment effect; A performance evaluation system is established, which includes indicators such as detection accuracy, response time, and false alarm rate, and regularly generates optimization evaluation reports.

7. A method for intelligent safety perception and automated response in power plants based on MCP, characterized in that, The method includes: Multi-data standardization access based on MCP protocol: By deploying three types of MCP servers—industrial control protocol, equipment status, and security log—unified collection and standardized processing of multi-source heterogeneous data from power plants can be achieved, solving the data silo problem; Multi-level collaborative security threat perception: Employing a three-level detection mechanism of syntax detection, behavior analysis, and correlation analysis, it achieves comprehensive perception of known attacks and abnormal behaviors, thereby improving detection coverage; LLM-based deep threat assessment: Utilizing the reasoning capabilities of large-scale language models and combining them with a power generation scenario knowledge base to conduct multi-dimensional correlation analysis, identify attack intentions, and assess the degree of impact; Human-machine collaborative intelligent decision-making and response: Initiating differentiated response processes based on threat levels, and ensuring the security and reliability of critical operations through a combination of manual approval and automated execution; Continuously evolving system optimization mechanism: Based on the case handling library, the detection model and response strategy are continuously optimized, and the system protection capability is adaptively improved through a feedback learning mechanism.

8. An electronic device, characterized in that, include: One or more processors; A storage unit is used to store one or more programs, which, when executed by one or more processors, enable the one or more processors to implement the MCP-based intelligent safety perception and automated response method for power plants as described in claim 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it can realize the intelligent safety perception and automated response method for power plants based on MCP as described in claim 7.