A man-machine fusion scheduling method and system applied to a DCS operator station of a thermal power plant

Abstract

The application relates to a man-machine fusion method and system applied to a DCS operator station of a thermal power plant, and aims to solve the technical problems of missing man-machine trust and low collaborative efficiency caused by the fact that the existing DCS system simply enhances machine intelligence and ignores man-machine deep fusion. The application builds a system architecture integrating operator physiological and behavioral monitoring, realizes real-time sensing of the states such as fatigue, stress and attention of the operator, and establishes a quantitative model to scientifically measure the dynamic trust level between man and machine. On this basis, the application proposes an adaptive man-machine interaction interface and a collaborative regulation mechanism based on game theory, dynamically allocates control authority according to the operator state and man-machine trust degree, realizes flexible switching of man-machine roles and intelligent resolution of decision conflicts. The application adaptively deeply fuses the experience wisdom of man and the intelligence of machine, and significantly improves the safety and decision efficiency of system operation under complex working conditions.

Description

Technical Field

[0001] This application relates to the field of power plant operation and maintenance, and in particular to a human-machine integrated scheduling method, system, computer equipment, and computer-readable storage medium applied to the DCS operator station of a thermal power plant. Background Technology

[0002] Intelligentization provides technical support for the flexible operation of power units. Among them, the automatic control system is the main focus of power unit intelligence and an effective way to achieve flexible and intelligent thermal power units. The automatic control system of thermal power units is mainly built around the distributed control system (DCS). In recent years, it has evolved into the intelligent power generation operation control system (ICS), and research and applications have been carried out around aspects such as reduced manpower, intelligent monitoring, and autonomous operation and navigation. Its main research direction is to improve the intelligence level of DCS, that is, the machine.

[0003] The increase in machine intelligence has fundamentally changed the way power plant control room operators participate in unit operation. While this change has greatly improved work efficiency and power generation efficiency, it has also revealed new human-machine interaction issues, mainly in the areas of human-machine trust and human-machine collaboration.

[0004] Regarding human-machine trust, the specific manifestations are as follows: First, insufficient trust. Due to individual differences in knowledge background, centralized control operators sometimes resist intelligent systems. In the initial stages of intelligent system operation or when initial application results are unsatisfactory, operators may even completely lose trust in the intelligent system, resulting in its capabilities not being fully utilized. Second, excessive trust. Centralized control operators may exhibit excessive trust in the intelligent system. After the intelligent system has been running reliably for a period of time, operators may find themselves in a state of "having nothing to do," and their operational awareness and feel at the physical layer will weaken. Excessive trust in the system may exceed the capabilities of the intelligent system, thus triggering potential safety risks. Simultaneously, the skill enhancement process for centralized control operators is forced to stop, and their unique initiative is suppressed.

[0005] Regarding human-machine collaboration, the specific manifestations are as follows: First, traditional DCS systems rely on human experience for operation, resulting in response delays and operational errors. Second, purely AI-based control systems lack sufficient reliability in decision-making under complex operating conditions. Third, there is the problem of information overload in human-machine interfaces; for example, displaying too many parameters on a single screen may exceed the operator's cognitive capacity. Fourth, rigid emergency response procedures are difficult to adapt to the requirements of new power systems for low-load and rapid load changes.

[0006] Currently, while some systems possess basic AI-assisted functions, they lack dynamic allocation of human-machine permissions; some systems employ finite state machine models, lacking support for multimodal interaction. Some DCS models integrate knowledge graphs, but have not established a quantified decision weighting mechanism, and their human-machine interfaces typically only offer manual or automatic mode switching.

[0007] In summary, the current DCS field of thermal power plants lacks technologies that deeply integrate human intelligence and machine intelligence, and related technologies for achieving adaptive human-machine trust and human-machine collaboration are still in their infancy or blank stage. Summary of the Invention

[0008] This application provides a human-machine integrated scheduling method, system, computer equipment, and computer-readable storage medium for DCS operator stations in thermal power plants, to at least solve the problem of insufficient accuracy in human-machine collaboration in related technologies.

[0009] In a first aspect, embodiments of this application provide a human-machine integrated scheduling method applied to a DCS operator station in a thermal power plant, the method comprising:

[0010] Acquire multi-source status data from centralized control operators and real-time operational data from the DCS system;

[0011] Based on the real-time operational data and forecast information, control decisions are generated for both the centralized control operators and the DCS system.

[0012] Based on the multi-source status data, determine whether the centralized control operator is in a normal state; if so,

[0013] Based on a preset conflict threshold, the difference between human-machine decision-making between the control decisions of the centralized control operator and the control decisions of the DCS system is measured to determine whether there is a human-machine decision-making conflict.

[0014] When it is determined that there is no human-machine decision conflict, the control decision of the DCS system is adopted as the final control decision.

[0015] When a human-machine decision conflict is determined to exist, a Stackelberg game model is constructed with the centralized control operator as the leader and the DCS system as the follower. The human-machine game equilibrium solution is solved through the Stackelberg game model, and the human-machine game equilibrium solution is used as the final control decision of the DCS operator station.

[0016] In some embodiments, obtaining the multi-source status data of the centralized control operator includes:

[0017] Real-time physiological characteristic data of the centralized control operators are collected, including skin conductance, heart rate, body surface temperature, heart rate variability and blood oxygen saturation.

[0018] Collect operational status data and environmental and event data synchronized with the physiological characteristic data;

[0019] The physiological feature data, the operational condition data, and the environment and events are fused across modally using a multi-source data fusion framework based on an attention mechanism to obtain the multi-source state data.

[0020] In some embodiments, cross-modal fusion via an attention-based multi-source data fusion framework includes:

[0021] The real-time physiological feature data is input into the channel attention mechanism and long short-term memory network for processing, and the first temporal feature is output.

[0022] The unit operating condition data synchronized with the physiological characteristic data is input into the long short-term memory network and the gated loop unit for processing, and the second time-series feature is output.

[0023] Environmental and event data are input into a one-dimensional convolutional neural network and a long short-term memory network for processing, and the third temporal feature is output.

[0024] Based on the first time-series feature, the second time-series feature, and the third time-series feature, the feature expression capability is improved by fusing the features output by each network, and the model's ability to process time-series data is enhanced, thus obtaining the multi-source state data.

[0025] In some embodiments, the method further includes:

[0026] Based on the multi-source status data and the real-time operation data, calculate the real-time trust level between humans and machines;

[0027] Based on the real-time trust level, before constructing the Stackelberg game model, the system roles of the central control operator and the DCS system are determined. When the real-time trust level is lower than a first preset threshold, the central control operator plays the dominant role; when the real-time trust level is higher than a second preset threshold, the DCS system plays the dominant role.

[0028] In some embodiments, calculating a person's real-time trust in the machine includes:

[0029] In the operational behavior data of centralized control operators, the number of times the operators followed machine decisions and the total number of decisions are obtained. The ratio of the number of times the operators followed machine decisions to the total number of decisions is calculated to obtain the behavioral measurement trust level.

[0030] The physiological characteristic data is processed using a lightweight AI model to obtain the physiological measurement confidence level;

[0031] The decision-making capability trust level is calculated based on the perception deviation of the centralized control operators regarding DCS control.

[0032] The real-time trust score is obtained by weighted summation of the behavioral measurement trust score, the physiological measurement trust score, and the decision-making ability trust score.

[0033] In some embodiments, the method further includes:

[0034] The DCS human-machine interface is adaptively adjusted based on the physiological state of the centralized control operators and the real-time trust level.

[0035] The physiological states mentioned include fatigue, abnormal stress, and abnormal concentration.

[0036] The specific adjustments include: when the centralized control operator is fatigued, a dual confirmation mechanism for key parameters is activated and the alarm control is enlarged with a red warning box; when the pressure is abnormal, the auxiliary control loop is automatically activated and the interface is adjusted to a soothing color scheme; when the concentration level is abnormal, the refresh frequency of key parameters is increased and key control points are highlighted.

[0037] In addition, when the physiological state of the centralized control operator is abnormal, the secondary centralized control operator takeover protocol is triggered, and the DCS degradation control is triggered;

[0038] When the physiological state of the centralized control operator is normal, differentiated auxiliary decision-making information is pushed according to the real-time trust level. Specifically, when the real-time trust level is lower than the first preset threshold, parameter indicators and operation process optimization data are pushed; when the real-time trust level is higher than the second preset threshold, automatic inspection data and intelligent alarm data are pushed.

[0039] In some embodiments, constructing a Stackelberg game model with the centralized control operator as the leader and the DCS system as the follower includes:

[0040] The DCS system acts as a follower, acquiring the first control decisions of the centralized control operators under conditions of information asymmetry.

[0041] The DCS system, acting as a follower, optimizes its own control model based on the first control decision and the predicted information on grid AGC, coal quality, and unit output, and generates a second control decision that satisfies the preset Stackelberg game equilibrium conditions.

[0042] The second control decision that meets the conditions is taken as the equilibrium solution of the human-machine game and used as the system control decision to eliminate the conflict between human and machine decisions.

[0043] Secondly, embodiments of this application provide a human-machine integrated scheduling system applied to a DCS operator station in a thermal power plant. The system includes: an acquisition module, a judgment module, and a decision module, wherein:

[0044] The acquisition module is used to acquire multi-source status data of the centralized control operators and real-time operation data of the DCS system;

[0045] Furthermore, based on the real-time operating data and predictive information, control decisions are generated for both the centralized control operators and the DCS system.

[0046] The judgment module is used to determine whether the centralized control operator is in a normal state based on the multi-source state data. If so, based on a preset conflict threshold, it measures the difference between the human-machine decision-making between the control decision of the centralized control operator and the control decision of the DCS system, and determines whether there is a human-machine decision-making conflict.

[0047] The decision module is used to adopt the control decision of the DCS system as the final control decision when it is determined that there is no human-machine decision conflict.

[0048] Furthermore, when it is determined that there is a conflict between human and machine decision-making, a Stackelberg game model is constructed with the centralized control operator as the leader and the DCS system as the follower. The Stackelberg game model is used to solve for the equilibrium solution of the human-machine game, and the equilibrium solution of the human-machine game is used as the final control decision of the DCS operator station.

[0049] In some embodiments, the system further includes a hardware layer, a data layer, and an application layer;

[0050] The hardware layer is used to integrate the raw data of the DCS system with the physiological signals of the centralized control operator, including a physiological signal acquisition device for acquiring the physiological signals and a DCS interface for interfacing with the DCS control station.

[0051] The data layer is used to perform signal filtering and feature extraction, and to deploy a lightweight AI model to predict fatigue levels in real time, and to complete the millisecond-level synchronization and centralized storage of the physiological signals and DCS operation logs.

[0052] The application layer is deployed on the DCS extended server and is used to implement functional modules for human-machine trust measurement, human-machine trust correction, intelligent decision generation, and human-machine collaborative control.

[0053] Thirdly, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in the first aspect above.

[0054] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect above.

[0055] Compared to related technologies, this application provides a human-machine integration method and system for DCS operator stations in thermal power plants, aiming to solve the technical problem of existing DCS systems simply enhancing machine intelligence while neglecting deep human-machine integration, leading to a lack of human-machine trust and low collaborative efficiency. This invention constructs a system architecture integrating operator physiological and behavioral monitoring to perceive operator fatigue, stress, and attention levels in real time, and establishes a quantitative model to scientifically measure the dynamic trust level between humans and machines. Based on this, this invention proposes an adaptive human-machine interface and a game theory-based collaborative control mechanism, dynamically allocating control permissions according to operator status and human-machine trust levels, achieving flexible switching of human-machine roles and intelligent resolution of decision-making conflicts. This invention adaptively and deeply integrates human experience and machine intelligence, significantly improving the safety and decision-making efficiency of system operation under complex working conditions. Attached Figure Description

[0056] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0057] Figure 1 This is a schematic diagram of the human-machine integration system architecture of a DCS operator station in a thermal power plant according to an embodiment of this application;

[0058] Figure 2 This is a flowchart of a human-machine integrated scheduling method applied to a DCS operator station in a thermal power plant, according to an embodiment of this application.

[0059] Figure 3 This is a schematic diagram of a multi-source data fusion framework for measuring the physiological and behavioral measurements of DCS centralized control personnel of thermal power units according to an embodiment of this application;

[0060] Figure 4 This is a schematic diagram of the human-machine collaborative control model for thermal power units based on Stackelberg game theory, according to an embodiment of this application.

[0061] Figure 5 This is a flowchart of conflict resolution based on Stackelberg game theory according to an embodiment of this application;

[0062] Figure 6 This is a schematic diagram of the internal structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this application clearer, the application is described and illustrated below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0064] Obviously, the accompanying drawings described below are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar scenarios based on these drawings without any inventive effort. Furthermore, it is understood that although the efforts made in this development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, any changes to design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as insufficient disclosure of the content of this application.

[0065] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment that is mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0066] Unless otherwise defined, the technical or scientific terms used in this application shall have the ordinary meaning understood by one of ordinary skill in the art to which this application pertains. The terms “a,” “an,” “an,” “the,” and similar words used in this application do not indicate quantity limitation and may indicate singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units not listed, or may include other steps or units inherent to these processes, methods, products, or devices. The terms “connected,” “linked,” “coupled,” and similar words used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. The character " / " generally indicates that the preceding and following objects are in an "or" relationship. The terms "first," "second," and "third" used in this application are merely to distinguish similar objects and do not represent a specific ordering of the objects.

[0067] This application provides a human-machine integration system for DCS operator stations in thermal power plants, which is used to implement the human-machine integration method described later. Figure 1 This is a schematic diagram of the human-machine integration system architecture of a DCS operator station in a thermal power plant according to an embodiment of this application. Figure 1 As shown, the system comprises a hardware layer, a data layer, and an application layer. Specifically:

[0068] The hardware layer integrates raw data from the DCS with physiological signals from the operators. It primarily consists of physiological signal acquisition devices and the DCS interface. The physiological signal acquisition devices, through wearable sensors (such as wristband sensors) and infrared eye trackers, integrate optical heart rate sensors, SpO2 monitors, bioimpedance sensors, electrocardiogram (ECG) sensors, skin conductance sensors, and skin temperature sensors to measure real-time physiological states such as skin conductance, heart rate, body surface temperature, heart rate variability, and blood oxygen saturation. To adapt to industrial environments, these devices feature electromagnetic interference immunity and intrinsic safety certification, and support industrial wireless networks for low-power wireless transmission. The DCS interface connects to the DCS control station via industrial protocols (such as OPCUA and Modbus TCP) and features dual-channel communication redundancy to avoid single points of failure.

[0069] The data layer primarily focuses on data acquisition, handling data computation and storage. Data computation is performed by edge computing nodes deployed on the DCS edge gateway. These nodes perform signal filtering (such as wavelet denoising) and feature extraction (such as heart rate variability (HRV) analysis). Simultaneously, the data layer deploys lightweight AI models, such as efficient channel attention mechanisms (ECA), long short-term memory networks (LSTM), and gated recurrent units (GRUs), for real-time fatigue level prediction. The data layer also utilizes a precise network time synchronization protocol to achieve millisecond-level synchronization and centralized storage of physiological signals and DCS operation logs.

[0070] The application layer is deployed on the DCS extension server and integrated with the DCS Human-Machine Interface (HMI) and control logic. The functional modules implemented in the application layer include: human-machine trust measurement, human-machine trust correction, intelligent decision generation, and human-machine collaborative control. The human-machine trust measurement module establishes comprehensive factors to measure human-machine trust. The intelligent decision generation module establishes a knowledge base and has hierarchical early warning and dynamic access control functions. The human-machine collaborative control module establishes a dynamic allocation model for human-machine control rights.

[0071] This application also provides a human-machine integrated scheduling method for DCS operator stations in thermal power plants. Figure 2 This is a flowchart illustrating a human-machine integrated scheduling method applied to a DCS operator station in a thermal power plant, according to an embodiment of this application. Figure 2 As shown, the process includes the following steps:

[0072] Step S201: Obtain multi-source status data from centralized control operators and real-time operating data from the DCS system;

[0073] This step is fundamental to human-machine fusion decision-making. By comprehensively collecting human state information and machine operational information, it provides the necessary data input for subsequent trust measurement, decision analysis, and collaborative control. This step utilizes multimodal data fusion technology to integrate dispersed and heterogeneous human physiological data with system operating condition and environmental data, forming a unified, high-dimensional state description. This improves the accuracy and depth of the system's perception of the current human-machine interaction context, laying a solid foundation for achieving truly adaptive human-machine collaboration.

[0074] In one embodiment, obtaining multi-source status data from centralized control operators includes the following sub-steps:

[0075] S1, collect real-time physiological characteristic data of the centralized control operators, including skin conductance, heart rate, body surface temperature, heart rate variability, and blood oxygen saturation. Simultaneously, collect operational condition data and environmental and event data synchronized with the physiological characteristic data.

[0076] Step S1 uses wearable devices and other hardware to capture the operator's physiological indicators in real time, and records their synchronized operational data, environmental data, and event data, achieving simultaneous monitoring of the operator's internal state and external environment. This dual-dimensional data collection method provides an objective basis for subsequent accurate assessment of the operator's cognitive load, fatigue state, and attention level, enabling the system to transcend the traditional machine perspective and incorporate human factors into its considerations.

[0077] S2 uses an attention-based multi-source data fusion framework to perform cross-modal fusion of physiological feature data, operational condition data, and environmental and event data to obtain multi-source state data.

[0078] Figure 3 This is a schematic diagram of a multi-source data fusion framework for measuring the physiological and behavioral measurements of DCS centralized control personnel of thermal power units according to an embodiment of this application.

[0079] The specific integration process is as follows:

[0080] Real-time physiological feature data is input into the efficient channel attention mechanism ECA and the long short-term memory network LSTM for processing, and the first temporal feature is output.

[0081] The unit's operating condition data, synchronized with physiological characteristic data, is input into an LSTM and a gated recurrent unit (GRU) for processing, outputting the second time-series feature. Environmental and event data are input into a one-dimensional convolutional neural network (1D-CNN) and an LSTM for processing, outputting the third time-series feature.

[0082] Finally, the first, second, and third time-series features are fused to obtain multi-source state data.

[0083] Step S2 utilizes a deep learning model to effectively integrate data from different sources and with different characteristics, addressing the challenge of directly comparing and using heterogeneous data. By assigning weights to data of different modalities, this framework can extract more expressive and discriminative fusion features, thereby providing high-quality input for subsequent human-computer trust measurement and decision analysis, significantly improving the accuracy of the entire method.

[0084] Step S202: Based on real-time operating data and forecast information, generate control decisions for centralized control operators and the DCS system.

[0085] In this embodiment, this step embodies the parallel working mode of two decision-making entities in the human-machine system. The centralized control operator makes subjective decisions based on their experience and knowledge, while the DCS system makes objective decisions based on precise models and data. This step clarifies the source and basis of each decision-making entity, serving as a prerequisite for subsequently identifying differences in human-machine decision-making and determining whether conflicts exist. By generating two decisions in parallel, the system can compare and analyze them in subsequent stages, thereby initiating corresponding collaborative mechanisms to ensure that the final control behavior leverages both the computational advantages of the machine and incorporates the operator's experience and knowledge. The predictive information refers to prior data used to simulate future operating conditions of the unit and changes in external commands, mainly including grid AGC commands, coal quality entering the furnace, and the unit's output capacity, allowing both humans and machines to predict system trends during decision-making.

[0086] Step S203: Based on a preset conflict threshold, measure the difference between human-machine decision-making between the control decisions of the centralized control operator and the control decisions of the DCS system, and determine whether there is a human-machine decision-making conflict.

[0087] Specifically, the system determines whether the centralized control operator is in a normal state based on the multi-source status data; if so, it determines whether there is a human-machine decision conflict.

[0088] The metric for human-machine decision-making conflict can be expressed by formula (3):

[0089] in, It is the difference in decision-making between humans and machines. It is a human decision. The decision is made by the machine, and T is the conflict threshold. The conflict threshold is a hyperparameter, and it varies for each control decision, requiring setting based on actual operating conditions. When the difference between human and machine decisions is less than the conflict threshold, no human-machine conflict occurs. When the difference between human and machine decisions exceeds the conflict threshold, a human-machine conflict occurs.

[0090] This step introduces a quantitative conflict detection mechanism into the system, transforming disagreements between humans and machines from vague qualitative judgments into clear quantitative analyses. By setting reasonable conflict thresholds, the system can automatically identify when intervention or collaboration procedures need to be initiated, avoiding unnecessary switching and interference, while also ensuring timely responses to major decision-making disagreements. This mechanism is a key element in achieving dynamic and flexible allocation of human-machine permissions, improving the stability and decision-making efficiency of the entire system.

[0091] Step S204: When it is determined that there is no conflict between human and machine decision-making, the control decision of the DCS system is adopted as the final control decision.

[0092] This step establishes the default principle of prioritizing machine decision-making when human and machine opinions are aligned or differences are manageable. This fully leverages the advantages of the DCS system in precise calculation and rapid response, ensuring operational efficiency and stability under normal conditions. This strategy simplifies the decision-making process, reduces unnecessary human intervention, and allows central control operators to focus their efforts on abnormal or complex conditions that require more experience and judgment, achieving optimal allocation of human and machine resources.

[0093] Step S205: When it is determined that there is a conflict between human and machine decision-making, a Stackelberg game model is constructed with the centralized control operator as the leader and the DCS system as the follower. The Stackelberg game model is used to solve the equilibrium solution of the human-machine game, and the equilibrium solution of the human-machine game is used as the final control decision of the DCS operator station.

[0094] This step is the core mechanism for human-machine collaboration and resolving decision-making conflicts. By introducing non-cooperative game theory, the human-machine interaction process is modeled as a dynamic master-slave game, providing a rigorous mathematical framework for conflict resolution. By solving for the Stackelberg equilibrium, this method finds an optimal solution that considers the decision-making considerations of both humans and machines, rather than simply choosing between the two. This achieves a shift from "human-machine confrontation" to "human-machine collaboration," improving the scientific rigor and robustness of the final decision in complex conflict scenarios.

[0095] In an optional embodiment, before constructing the Stackelberg game model, the method further includes: calculating the real-time trust level between humans and machines based on multi-source state data and real-time operational data; and determining the system roles of the central control operator and the DCS system based on the real-time trust level. When the real-time trust level is lower than a first preset threshold, the central control operator plays the dominant role; when the real-time trust level is higher than a second preset threshold, the DCS system plays the dominant role.

[0096] Furthermore, the specific process for calculating the real-time trust level between humans and machines is as follows: First, record the historical and real-time operational behavior data of the centralized control operator. This data includes operation frequency, dialog box switching speed, and the number of manual / automatic switching operations. From this data, obtain the number of times the operator followed machine decisions and the total number of decisions, and calculate the ratio between the two to obtain the behavioral measurement trust level. Second, use a lightweight AI model to process physiological characteristic data to obtain the physiological measurement trust level. Third, calculate the decision-making ability trust level based on the operator's perception deviation of DCS control. Finally, perform a weighted sum of the behavioral measurement trust level, physiological measurement trust level, and decision-making ability trust level to obtain the real-time trust level.

[0097] Specifically, the calculation method for human trust in machines is shown in the following formula.

[0098]

[0099] in, This represents the expected value of a skilled centralized control operator's perception of the difference. This represents the expected value of the perceived differences by the new centralized control operators, reflecting the boundary of the centralized control operators' perception ability of the DCS control of thermal power units; The expected value, representing perceptual bias, is used to measure a person's perceptual ability. The more accurate the perception, the higher the level of trust. It indicates the level of trust people have in machines.

[0100] The calculation method for human-machine trust measurement is shown in the following formula.

[0101]

[0102] Where r(k) represents the real-time level of trust between humans and machines in the DCS operator station of the thermal power unit at time k, and w1 represents the trust weight based on behavioral measurement. This represents the number of times the central control operators followed the machine's decision-making results up to time k. denoted by , w2 represents the total number of decisions up to time k; w2 represents the trust weight based on physiological measurements, F(k) represents the trust based on physiological factors; w3 represents the trust weight based on decision-making ability; and e represents the bias in human understanding of machine decision-making ability.

[0103] It should be noted that, firstly, the "decision-making ability" component of the trust level is calculated. Then, this result is used as an input, and together with the trust level components of other dimensions (behavioral and physiological), the final, comprehensive, real-time trust level is calculated through common weighting.

[0104] This step utilizes a multi-dimensional comprehensive evaluation model to quantify the subjective psychological state of human-machine trust. It considers not only the operator's historical behavior but also incorporates real-time physiological states and individual ability assessments, making the trust calculation results more dynamic and accurate. This quantified trust level provides a reliable basis for the subsequent dynamic allocation of human-machine roles and permissions.

[0105] Figure 4 This is a schematic diagram of the human-machine collaborative control model for thermal power units based on Stackelberg game theory, according to an embodiment of this application. Figure 5This is a flowchart illustrating conflict resolution based on a Stackelberg game according to an embodiment of this application. In the game model, the DCS system, as a follower, obtains the first control decision from the centralized control operator, who is the leader. Based on this decision and predictions of grid AGC, coal quality, and unit output, the DCS system optimizes its own control model and generates a second control decision that satisfies a preset Stackelberg game equilibrium condition. This equilibrium condition is expressed by the following formula: The second regulatory decision that meets the conditions is the human-machine game equilibrium solution that serves as the final regulatory decision.

[0106] It should be noted that in the game theory model, the two parties are designated as "Leader" and "Follower." In the default settings of this embodiment or in scenarios with low human-machine trust, the centralized control operator acts as the "Leader," while the DCS automation system acts as the "Follower." This is an incomplete information game process. The DCS system, as the "Follower," can obtain the control operation decisions of the centralized control operator, who is the "Leader," while the centralized control operator cannot know the machine's control information in advance when making decisions. There is an information asymmetry between the two.

[0107] The game process is as follows: First, the "dominant player" (centralized control operator) optimizes and makes its first control decision based on its own experience and knowledge, as well as predictions of the power grid's AGC commands, the quality of coal entering the furnace, and the unit's output capacity. Subsequently, the "dominant player" notifies the "followers" of this decision and executes it.

[0108] Next, the DCS system, acting as the "follower," after receiving the first control decision from the "dominant" system, optimizes its own control model based on this decision and its own predictions using more comprehensive information such as grid AGC commands, primary frequency regulation commands, and calorific value signals of the coal fed into the furnace. Through this optimization process, the DCS system generates a new second control decision aimed at maximizing its own benefit function. This newly generated decision must satisfy a pre-set Stackelberg game equilibrium condition. This equilibrium condition requires that, given the dominant system's optimal decision, the follower's chosen decision must be the solution that optimizes its own benefit function.

[0109] Ultimately, this second regulatory decision, which satisfies the Stackelberg game equilibrium condition, was determined as the equilibrium solution in the human-machine game and served as the system's final regulatory decision output after eliminating the human-machine decision conflict. Through this mechanism, the system can find a collaborative solution that considers both the dominant intentions of humans and fully utilizes the optimization capabilities of machines when decision disagreements exist.

[0110] In another embodiment, the method of this application further includes adaptively adjusting the DCS human-machine interface based on the physiological state and real-time trust level of the centralized control operator.

[0111] This step deeply integrates human factors engineering theory with DCS system design, transforming the human-machine interface from a static information display platform into an intelligent assistant that proactively adapts to the operator's state. By monitoring the operator's physiological state and trust level in real time, it can dynamically adjust the presentation of information and interaction logic, aiming to reduce the operator's cognitive load, prevent misoperations, and provide differentiated decision support at different trust levels. This significantly improves the efficiency, safety, and comfort of human-machine interaction.

[0112] Specifically, Table 1 shows the main status of centralized control operators, their DCS responses, and HMI adjustment strategies:

[0113] As shown in Table 1, when the system detects that the control room operator is in a state of fatigue, high pressure, or low concentration, it will trigger corresponding adjustments. For example, in a state of fatigue, the system activates a dual-confirmation mechanism for key parameters and amplifies alarm controls; in a state of high pressure, the system automatically activates auxiliary control loops and adjusts the interface to a soothing color scheme; in a state of low concentration, the system increases the refresh rate of key parameters and highlights key control points. When the control room operator's physiological state is abnormal, the system triggers a secondary control room operator takeover protocol and DCS degradation control. When the control room operator's physiological state is normal, differentiated auxiliary decision-making information is pushed based on the real-time trust level. When the trust level is low, parameter indicators and operational process optimization data are pushed; when the trust level is high, automatic monitoring data and intelligent alarm data are pushed.

[0114] Through steps S201 to S205 above, compared with related technologies, the human-machine integration method provided in this application for DCS operator stations in thermal power plants aims to solve the technical problem of existing DCS systems simply enhancing machine intelligence while neglecting deep human-machine integration, resulting in a lack of human-machine trust and low collaborative efficiency. This invention constructs a system architecture integrating operator physiological and behavioral monitoring to perceive operator fatigue, stress, and attention levels in real time, and establishes a quantitative model to scientifically measure the dynamic trust level between humans and machines. Based on this, this invention proposes an adaptive human-machine interface and a game theory-based collaborative decision-making control mechanism, dynamically allocating control permissions according to operator status and human-machine trust levels, achieving flexible switching of human-machine roles and intelligent resolution of decision-making conflicts. This invention adaptively and deeply integrates human experience and machine intelligence, significantly improving the safety and decision-making efficiency of system operation under complex working conditions.

[0115] Furthermore, the human-machine integration method and system provided in this application possess high flexibility and relevance in practical applications, and their specific implementation methods cover aspects such as system integration, operating modes, interface implementation, and application scope:

[0116] To adapt to the DCS architecture and implementation conditions of different power plants, this system provides two integrated deployment solutions:

[0117] Deep Integration Solution: This is Method 1, which involves directly deploying the necessary workstations and server hardware resources on the C / D network (usually engineering workstations or extended networks) planned within the DCS system. Based on this hardware platform, comprehensive integration of the hardware, data, and application layers of the human-machine interface system is achieved, realizing tight coupling with the DCS system.

[0118] External Solution: This is the second approach, which involves setting up a workstation and server independent of the DCS core network. This solution uses a dedicated communication server or communication card to securely and reliably exchange data with the DCS system using industry-standard protocols (such as Modbus TCP or OPC / OPC-UA). Each layer of the system is deployed on this independent hardware, offering advantages such as not affecting the existing DCS system and flexible installation.

[0119] This system is designed with two complementary operating modes: offline and online, to ensure both the accuracy of the model and the security of its operation.

[0120] Offline Mode: In this mode, the system primarily performs learning, modeling, and testing. Utilizing historical data and a simulated environment, it completes the training of the AI ​​model, the establishment of the human factors model, and comprehensive testing of system functions, ensuring the reliability of all decision-making logic before deployment.

[0121] Online Mode: In this mode, the system is put into actual operation, performing real-time closed-loop decision support. The system continuously collects real-time DCS data and operator physiological behavior data, analyzes them, and provides intelligent suggestions or collaborative control.

[0122] To achieve dynamic adaptation of the Human-Machine Interface (HMI), this system will expand the functionality of the existing DCS HMI. First, the system will acquire the plug-ins, components, and related configuration information of the DCS screen. Next, a new interface layout and element label system will be established based on this, and uniformly managed by the management module. Finally, this system will be associated with specific operator user information, thereby dynamically adjusting the interface according to the operator's real-time status (such as fatigue level and trust level) to achieve a personalized and intelligent interactive experience.

[0123] Application targets and control levels:

[0124] To ensure that the real-time performance and stability of the basic control system (such as protection and interlocking) are not interfered with, the human-machine collaborative control function of this system is mainly aimed at higher-level optimized control objects. The response time scale of these objects is typically in the range of seconds to minutes, making them suitable for human-machine negotiation and decision optimization. Specific application scenarios include:

[0125] Unit and plant-level AGC control: such as optimization of automatic generation control (AGC) for a single unit, or allocation and coordination of AGC commands across the entire plant.

[0126] Multi-energy complementary coordinated control: such as intelligent regulation of combined electric and heat loads to achieve optimal comprehensive energy utilization efficiency.

[0127] Start-up, shutdown, and operation optimization of major auxiliary equipment: such as optimizing the start-up and shutdown sequence and load distribution of multiple auxiliary equipment of the same type, such as coal mills and slurry circulation pumps, that are operating in parallel.

[0128] In one embodiment, Figure 6 This is a schematic diagram of the internal structure of an electronic device according to an embodiment of this application, such as... Figure 6 As shown, an electronic device is provided, which can be a server, and its internal structure diagram can be as follows. Figure 6 As shown, the electronic device includes a processor, a network interface, internal memory, and non-volatile memory connected via an internal bus. The non-volatile memory stores the operating system, computer programs, and a database. The processor provides computing and control capabilities, the network interface communicates with external terminals via a network, the internal memory provides the environment for the operating system, the computer programs are executed by the processor to implement a human-machine interface method applied to a DCS operator station in a thermal power plant, and the database stores data.

[0129] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0130] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by hardware related to computer program instructions. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0131] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0132] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A human-machine fusion scheduling method applied to a DCS operator station of a thermal power plant, characterized in that, The method comprises: acquiring multi-source state data of a centralized operation personnel and real-time operation data of a DCS system, wherein an edge computing node deployed at a DCS edge gateway performs signal filtering and feature extraction; the multi-source state data is obtained through cross-modal fusion of physiological feature data, operation condition data, and environment and event data based on a multi-source data fusion framework based on an attention mechanism; based on the real-time operation data and prediction information, generating control decisions of the centralized operation personnel and the DCS system, respectively; determining whether the centralized operation personnel is in a normal state according to the multi-source state data, and if so, measuring a man-machine decision difference between the control decisions of the centralized operation personnel and the control decisions of the DCS system based on a preset conflict threshold to determine whether there is a man-machine decision conflict; when it is determined that there is no man-machine decision conflict, adopting the control decisions of the DCS system as the final control decisions; when it is determined that there is a man-machine decision conflict, constructing a Stackelberg game model with the centralized operation personnel as a leader and the DCS system as a follower, solving a man-machine game equilibrium solution through the Stackelberg game model, and taking the man-machine game equilibrium solution as the final control decisions of a DCS operator station; wherein, based on the multi-source state data and the real-time operation data, a real-time trust degree of a human to a machine is calculated; according to the real-time trust degree, the system roles of the centralized operation personnel and the DCS system are determined before the Stackelberg game model is constructed, wherein when the real-time trust degree is lower than a first preset threshold, the centralized operation personnel is the leader; and when the real-time trust degree is higher than a second preset threshold, the DCS system is the leader; constructing the Stackelberg game model with the centralized operation personnel as the leader and the DCS system as the follower comprises: the DCS system as the follower acquires a first control decision of the centralized operation personnel under the condition of information asymmetry; the DCS system as the follower optimizes its control model based on the first control decision and prediction information of a power grid AGC, coal quality into the furnace, and unit available output, to generate a second control decision that meets a preset Stackelberg game equilibrium condition; and the second control decision that meets the condition is taken as the man-machine game equilibrium solution, which is used as a system control decision to eliminate the man-machine decision conflict.

2. The method of claim 1, wherein, acquiring the multi-source state data of the centralized operation personnel comprises: collecting real-time physiological feature data of the centralized operation personnel, wherein the physiological feature data comprises electrodermal activity, heart rate, body surface temperature, heart rate variability, and blood oxygen saturation; collecting operation condition data and environment and event data synchronized with the physiological feature data.

3. The method of claim 2, wherein, cross-modal fusion through the multi-source data fusion framework based on the attention mechanism comprises: inputting the real-time physiological feature data into a channel attention mechanism and a long short-term memory network for processing to output first time sequence features; The unit operation condition data synchronized with the physiological characteristic data is input into a long short-term memory network and a gated recurrent unit for processing, and second time sequence characteristics are output; The environment and event data are input into a one-dimensional convolutional neural network and a long short-term memory network for processing, and third time sequence characteristics are output; Based on the first time sequence characteristics, the second time sequence characteristics and the third time sequence characteristics, the feature expression capability is improved by fusing the features output by each network, and the processing capability of the model for time sequence data is enhanced, and the multi-source state data is obtained.

4. The method of claim 1, wherein, The real-time trust degree of the human to the machine includes: In the operation behavior data of the centralized operation personnel, the number of times of following the machine decision and the total number of decisions of the centralized operation personnel are obtained, and the ratio of the number of times of following the machine decision to the total number of decisions is calculated to obtain a behavior measurement trust degree; The physiological characteristic data is processed by using a lightweight AI model to obtain a physiological measurement trust degree; According to the perception deviation of the centralized operation personnel to the DCS control perception, a decision capability trust degree is calculated; The behavior measurement trust degree, the physiological measurement trust degree and the decision capability trust degree are weighted and summed to obtain the real-time trust degree.

5. The method of claim 1, wherein, The method further includes: According to the physiological state of the centralized operation personnel and the real-time trust degree, the DCS human-machine interaction interface is adaptively adjusted; The physiological state includes a fatigue state, an abnormal stress state and an abnormal concentration state; The adjustment content specifically includes: when the centralized operation personnel is in the fatigue state, a key parameter double confirmation mechanism is enabled and an alarm control is enlarged with a red warning box; when the stress is in the abnormal state, an auxiliary control loop is automatically enabled and the interface is adjusted to a soothing color tone; when the concentration is in the abnormal state, the key parameter refresh frequency is increased and the key control point is highlighted; And when the physiological state of the centralized operation personnel is abnormal, a secondary centralized operation personnel takeover protocol is triggered, and DCS degradation control is triggered; When the physiological state of the centralized operation personnel is normal, according to the high and low of the real-time trust degree, differential auxiliary decision information is pushed, wherein when the real-time trust degree is lower than a first preset threshold, parameter indicators and operation process optimization data are pushed; when the real-time trust degree is higher than a second preset threshold, automatic patrol data and intelligent alarm data are pushed.

6. A human-machine fusion dispatching system applied to a DCS operator station of a thermal power plant, characterized in that, The system includes an acquisition module, a judgment module and a decision module, wherein: The acquisition module is used to acquire multi-source state data of the centralized operation personnel and real-time operation data of the DCS system, wherein signal filtering and feature extraction are performed by an edge computing node deployed at a DCS edge gateway; the multi-source state data is obtained by cross-modal fusion of physiological characteristic data, operation condition data and environment and events based on a multi-source data fusion framework based on an attention mechanism; And based on the real-time operation data and prediction information, control decisions of the centralized operation personnel and the DCS system are generated respectively; The judgment module is configured to determine whether the centralized operation personnel is in a normal state according to the multi-source state data, and if so, measure a man-machine decision difference between a decision of the centralized operation personnel and a decision of the DCS system based on a preset conflict threshold, and determine whether there is a man-machine decision conflict; The decision module is configured to, when it is determined that there is no man-machine decision conflict, adopt the decision of the DCS system as a final control decision; and when it is determined that there is a man-machine decision conflict, construct a Stackelberg game model with the centralized operation personnel as a leader and the DCS system as a follower, solve a man-machine game equilibrium solution through the Stackelberg game model, and adopt the man-machine game equilibrium solution as a final control decision of a DCS operator station. The real-time trust degree of man to machine is calculated based on the multi-source state data and the real-time operation data. The system roles of the centralized operation personnel and the DCS system are determined according to the real-time trust degree before the Stackelberg game model is constructed, wherein the centralized operation personnel is the leader when the real-time trust degree is lower than a first preset threshold, and the DCS system is the leader when the real-time trust degree is higher than a second preset threshold. The Stackelberg game model with the centralized operation personnel as the leader and the DCS system as the follower includes: The DCS system as the follower acquires a first control decision of the centralized operation personnel under the condition of information asymmetry; The DCS system as the follower optimizes its control model based on the first control decision and prediction information of power grid AGC, coal quality into the furnace and unit available output, generates a second control decision meeting a preset Stackelberg game equilibrium condition, and adopts the second control decision meeting the condition as the man-machine game equilibrium solution, which is used as a system control decision for eliminating man-machine decision conflict. The system further includes a hardware layer, a data layer and an application layer.

7. The system of claim 6, wherein, The hardware layer is configured to integrate original data of a DCS system and physiological signals of the centralized operation personnel, and includes a physiological signal acquisition device for acquiring the physiological signals and a DCS interface for interfacing with a DCS control station. The data layer is configured to perform signal filtering and feature extraction, and deploy a lightweight AI model to predict a fatigue level in real time, complete millisecond-level synchronization and centralized storage of the physiological signals and DCS operation logs. The application layer is deployed on a DCS expansion server, and is configured to implement function modules of man-machine trust measurement, man-machine trust correction, intelligent decision generation and man-machine collaborative control. The processor implements the method of any one of claims 1 to 5 when executing the computer program.

8. A computer device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, ​

Images