CN116011722A

CN116011722A - Large power grid-oriented man-machine cooperation regulation and control method, module and device

Info

Publication number: CN116011722A
Application number: CN202210515238.2A
Authority: CN
Inventors: 范士雄; 郭剑波; 马士聪; 李立新; 卜广全; 王铁柱; 刘幸蔚; 吕晨; 王国政; 周子涵; 徐浩田; 罗魁; 荆逸然; 侯玮琳
Original assignee: State Grid Corp of China SGCC; State Grid Zhejiang Electric Power Co Ltd; China Electric Power Research Institute Co Ltd CEPRI
Current assignee: State Grid Corp of China SGCC; State Grid Zhejiang Electric Power Co Ltd; China Electric Power Research Institute Co Ltd CEPRI
Priority date: 2022-05-11
Filing date: 2022-05-11
Publication date: 2023-04-25

Abstract

The invention discloses a man-machine cooperation regulation and control method, a module and a device for a large power grid, wherein the method comprises the following steps: determining a target task of power grid operation regulation, and decomposing the target task into a plurality of sub-tasks with association relations based on an execution step of the target task; distributing an execution mode for each of a plurality of subtasks; executing the subtasks based on the distributed execution mode, inputting the decision result of the subtasks after execution to the associated next subtask, and executing the next subtask until all the subtasks of the target task are executed, and acquiring the decision result of all the subtasks; and regulating and controlling the operation of the power grid based on the decision results of all the subtasks. According to the invention, through the man-machine cooperative control enhancement mode, the situation awareness and operation decision-making capability of the power grid are improved, and the working intensity of regulatory personnel is reduced.

Description

Large power grid-oriented man-machine cooperation regulation and control method, module and device

Technical Field

The invention relates to the technical field of power grid dispatching operation and control, in particular to a large power grid-oriented man-machine cooperation regulation and control method, a large power grid-oriented man-machine cooperation regulation and control module and a large power grid-oriented man-machine cooperation regulation and control device.

Background

The power system has the characteristics of openness, nonlinearity, complexity and the like as a maximum manual system. The power grid regulation and control system is used as a control center of the power system, and plays a very important role in ensuring safe and stable operation and reliable power supply of the power system. At present, the power grid regulation and control system is a highly informationized and automated information system integrating data acquisition, analysis decision and control, and is a typical man-machine interaction automated system. The power grid regulation and control personnel obtain power grid situation information, auxiliary decision making and other information in a man-machine interaction mode, and analyze and decide operation risk problems in power grid regulation and control according to scheduling regulation rules and plans by more relying on individual regulation and control experience, so that the problem of influencing safe and stable operation of the power grid is effectively solved.

With the continuous and rapid development of new energy sources in China and the rapid construction of large-scale long-distance transmission lines, the scale of an extra-high voltage interconnected power grid is continuously enlarged, and the reform of an electric power market is continuously advanced, so that the operation mode of the power grid is complex and changeable, and the current power grid regulation and control system and scheduling personnel face totally new challenges in the aspects of power grid operation characteristics and analysis control. On one hand, the learning and self-adaptation capability of the power grid regulation and control system is relatively weak, and the comprehensive processing capability of analysis and decision making is poor in the face of unknown complex working condition scenes, and still depends on the experience analysis of regulation and control personnel. On the other hand, the deep reconstruction of the power system network leads to more complex running characteristics of the power grid, the system is subjected to multivariable interleaving, the perception, decision and control behaviors of regulatory personnel are easily influenced by factors such as decision dimension, psychology and physiology of the personnel, and the like, and the misoperation accidents are easily caused under the extremely weather and other complex working conditions. The artificial intelligence technology based on data driving has stronger perception prediction and decision making capability, can be effectively combined with a power grid regulation and control system, and provides an effective means for improving the intelligent level of power grid regulation and control and reducing the work light of regulation and control personnel.

In the prior art, the intelligent regulation and control of a certain service by a power grid based on an artificial intelligence technology is realized by selecting proper input characteristic data and a model based on the understanding of people on the certain service, and belongs to the end-to-end intelligence. The introduction of artificial intelligence technology improves the intelligent degree of the power grid regulation and control system to a certain extent, but the regulation and control system has larger limitation in dynamic and complex man-machine cooperative tasks. The grid regulation and control man-machine cooperative control is a typical man-machine cooperative mixed enhancement intelligent system in a loop, and the key problem faced by the system is how to integrate the intelligence of regulation and control personnel and the machine intelligence, and the problems of uncertainty, vulnerability, openness and the like in the actual application scene of the grid regulation and control are solved in a man-machine cooperative mode.

In summary, in order to realize the efficient application of the artificial intelligence technology in the power grid regulation, a man-machine mixing enhanced intelligent regulation technology oriented to the large power grid regulation is needed.

Disclosure of Invention

The technical scheme of the invention provides a large power grid-oriented man-machine cooperation regulation and control method, a large power grid-oriented man-machine cooperation regulation and control module and a large power grid-oriented man-machine cooperation regulation and control device, so that the problem of how to intelligently regulate and control a large power grid based on man-machine cooperation is solved.

In order to solve the problems, the invention provides a man-machine cooperation regulation and control method for a large power grid, which comprises the following steps:

determining a target task of power grid operation regulation, and decomposing the target task into a plurality of sub-tasks with association relations based on an execution step of the target task;

distributing an execution mode for each of a plurality of subtasks;

executing the subtasks based on the distributed execution mode, inputting the decision result of the subtasks after execution to the associated next subtask, and executing the next subtask until all the subtasks of the target task are executed, and acquiring the decision result of all the subtasks;

and regulating and controlling the operation of the power grid based on the decision results of all the subtasks.

Preferably, the implementation manner includes: machine execution and man-machine cooperation execution;

the main body of the machine execution is an intelligent machine;

the man-machine cooperation execution main body is an intelligent machine and a regulating and controlling person.

Preferably, it comprises: when a machine execution mode is selected to execute the subtasks, acquiring decision results of the subtasks after execution;

evaluating the decision result, and inputting the decision result of the subtask after execution to the associated next subtask when the decision result is higher than a preset threshold value;

When the decision result is not higher than a preset threshold value and the state of the subtask is non-urgent, training a machine learning model through training data, and re-executing the subtask through the trained machine learning model; or when the state of the subtask is urgent, executing the subtask by a regulating and controlling person.

Preferably, the allocating an execution mode for each of the plurality of subtasks includes:

the set of target tasks is t= { Task ₁ ，Task ₂ ，…Task _n The decision variables for subtask allocation are }

The subtask allocation execution mode is defined as:

the capability vector of a person to perform a subtask is defined as h= { H ₁ ，H ₂ ，…H _n }, wherein element H _i Representing the ability of a person to execute subtask i, executing a machine to execute subtaskThe capacity vector of a transaction is defined as m= { M ₁ ，M ₂ ，…M _n Element Mi, where element Mi represents the machine's ability to perform subtask i.

Preferably, the allocating an execution manner for each of the plurality of subtasks includes:

the mathematical model of human-computer collaborative task allocation can be expressed as:

wherein: f (x) is an objective function; g _j (x) And g' _j (x) Sub-objective functions related to the capabilities of the machine and man-machine, respectively; t is a set of regulation tasks; alpha _j The weight value of the sub-target;

λand/>

δAn upper threshold and a lower threshold of capability values of the human and the machine for executing subtasks respectively; m is the number of sub-targets; n is the number of subtasks.

Preferably, the method further comprises:

when the execution mode is executed by man-machine cooperation, the decision right during the execution of man-machine cooperation is distributed based on the judgment condition, and the method comprises the following steps: switching of man-machine decision rights, man-machine auxiliary decision control and man-machine joint decision control.

Preferably, the method comprises, among other things,

the man-machine decision right switching comprises: when a preset event of switching the man-machine decision right occurs, the decision right is distributed to a regulating person or an intelligent machine;

the man-machine aided decision control comprises: the intelligent machine learns the behaviors of the regulating personnel, acquires knowledge based on the behaviors of the regulating personnel, and provides auxiliary decision-making results for executing subtasks;

the man-machine joint decision control comprises: the regulating personnel and the intelligent machine jointly execute subtasks.

Preferably, the executing the subtasks based on the selected execution mode includes:

when the execution mode of the subtasks is machine execution, the subtasks are sent to a machine learning model or a digital twin system for calculation through an intelligent machine, and a decision result of the subtasks is obtained;

When the execution mode of the subtasks is the man-machine cooperation, the subtasks are sent to a digital twin system for calculation through a regulating person and an intelligent machine, and a decision result of the subtasks is obtained.

Preferably, the method further comprises: training sample data is generated through a digital twin system, and a machine learning model is trained and knowledge is acquired through the training sample data.

Preferably, the digital twin system comprises a plurality of operation modes, and the tasks of different scenes in the power grid operation regulation are processed through different operation modes;

the operation mode includes: real-time, research, planning, training, testing.

Preferably, the method further comprises: acquiring training data, the training data comprising: historical data related to grid operation and digital twin system generation data; the method comprises the steps of carrying out a first treatment on the surface of the

Training the intelligent machine through the training data to generate a machine learning model;

and extracting knowledge elements in the training data based on the machine learning model to generate a knowledge base.

Based on another aspect of the invention, the invention provides a man-machine cooperation regulation and control module facing to a large power grid, wherein the module comprises:

The system comprises an initial unit, a control unit and a control unit, wherein the initial unit is used for determining a target task for power grid operation regulation and control, and decomposing the target task into a plurality of subtasks with association relations based on an execution step of the target task;

the processing unit is used for distributing an execution mode for each sub-task in the plurality of sub-tasks; executing the subtasks based on the distributed execution mode, inputting the decision result of the subtasks after execution to the associated next subtask, and executing the next subtask until all the subtasks of the target task are executed, and acquiring the decision result of all the subtasks;

and the result unit is used for regulating and controlling the operation of the power grid based on the decision results of all the subtasks.

the main body of the machine execution is an intelligent machine;

Preferably, the processing unit is further configured to: when a machine execution mode is selected to execute the subtasks, acquiring decision results of the subtasks after execution;

Preferably, the processing unit is configured to allocate an execution manner to each of a plurality of subtasks, and the allocation rule includes:

The subtask allocation execution mode is defined as:

to take the personThe capability vector of a row subtask is defined as h= { H ₁ ，H ₂ ，…H _n }, wherein element H _i Expressed as the ability of a person to perform subtask i, the ability vector of a machine to perform subtask is defined as m= { M ₁ ，M ₂ ，…M _n Element Mi, where element Mi represents the machine's ability to perform subtask i.

Preferably, the processing unit allocates an execution mode for each of a plurality of subtasks, including:

λand/>

Preferably, the processing unit is further configured to:

Preferably, the processing unit is configured to execute the subtasks based on the selected execution mode, and includes:

Preferably, the processing unit is further configured to: training sample data is generated through a digital twin system, and the intelligent learning model is trained and knowledge is acquired through the training sample data.

the operation mode includes: real-time, research, planning, training, testing.

Preferably, the processing unit is further configured to: acquiring training data, the training data comprising: historical data related to grid operation and digital twin system generation data;

Based on another aspect of the invention, the invention provides a man-machine hybrid power grid regulation intelligent device, which comprises:

The system comprises a power grid regulation and control automatic system, an AI intelligent system, a man-machine interaction system and a power grid digital twin system;

the system comprises a power grid regulation automatic system, a power grid digital twin system and a man-machine interaction system, wherein the power grid regulation automatic system part comprises original power grid safety protection, automatic control part functions and data acquisition and instruction issuing functions, and is used for acquiring power grid operation data and sending the power grid operation data to the power grid digital twin system and the man-machine interaction system; the instruction issuing function is used for executing a decision instruction according to a decision result to complete a power grid regulation task;

the AI intelligent system and the man-machine interaction system are respectively used for executing subtasks after decomposing the power grid regulation task;

the power grid digital twin system is used for executing calculation, analysis and checking tasks of the AI system or the man-machine interaction system in the process of executing the subtasks, and feeding back the obtained task decision result to the AI intelligent system or the man-machine interaction system of the task initiator. Preferably, the AI intelligent system takes the historical data of power grid operation and the data generated by the digital twin system as training data, trains a machine learning model and generates a machine learning model and knowledge elements; and executing the target task through the machine learning model.

Preferably, the AI intelligent system further comprises a knowledge base, wherein the knowledge base utilizes the machine learning model to extract knowledge elements in sample data, and the knowledge elements are represented and stored in different forms.

Preferably, the AI intelligent system and the man-machine interaction system decompose and execute the target task, including:

determining an execution step of the target task, and decomposing the target task into a plurality of subtasks with association relations;

and allocating an execution mode for each of the plurality of subtasks.

Preferably, the AI intelligent system and the man-machine interaction system execute the subtasks based on the distributed execution mode, input the decision result of the subtasks after execution to the associated next subtask, execute the next subtask until all the subtasks of the target task are executed, and acquire the decision result of all the subtasks.

The technical scheme of the invention provides a man-machine cooperation regulation and control method, a module and a device for a large power grid, wherein the method comprises the following steps: determining a target task of power grid operation regulation, and decomposing the target task into a plurality of subtasks with association relations based on an execution step of the target task; distributing an execution mode for each of a plurality of subtasks; executing the subtasks based on the distributed execution mode, inputting the decision result of the executed subtasks to the associated next subtask, executing the next subtask until all the subtasks of the target task are executed, and acquiring the decision result of all the subtasks; and regulating and controlling the operation of the power grid based on the decision results of all the subtasks. The man-machine cooperation regulation and control technology for large power grid regulation and control provided by the technical scheme of the invention aims at realizing the hybrid enhancement of man-machine intelligence by virtue of the respective advantages of man intelligence and machine intelligence and through a man-machine cooperation control mode, forming bidirectional information communication and control, constructing a power grid regulation and control man-machine hybrid enhancement intelligent system of 1+1>2, improving the power grid situation sensing and operation decision-making capability, and further reducing the working strength of regulation and control personnel.

Drawings

Exemplary embodiments of the present invention may be more completely understood in consideration of the following drawings:

FIG. 1 is a flow chart of a man-machine cooperation regulation method for a large power grid according to a preferred embodiment of the invention;

FIG. 2 is a flow chart of a man-machine cooperation regulation method for a large power grid according to a preferred embodiment of the invention;

FIG. 3 is a schematic diagram of intelligent decision making mode of grid regulation human-computer collaboration according to a preferred embodiment of the invention;

FIG. 4 is a schematic diagram of a grid regulation physical architecture of a hybrid man-machine enhanced smart according to a preferred embodiment of the present invention;

fig. 5 is a schematic diagram of a grid regulation and control function framework of man-machine hybrid enhanced intelligent according to a preferred embodiment of the present invention;

FIG. 6 is a schematic diagram of a man-machine cooperation regulation module for a large power grid according to a preferred embodiment of the present invention;

fig. 7 is a schematic structural diagram of a man-machine hybrid power grid regulation intelligent device according to a preferred embodiment of the invention.

Detailed Description

The exemplary embodiments of the present invention will now be described with reference to the accompanying drawings, however, the present invention may be embodied in many different forms and is not limited to the examples described herein, which are provided to fully and completely disclose the present invention and fully convey the scope of the invention to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the accompanying drawings is not intended to be limiting of the invention. In the drawings, like elements/components are referred to by like reference numerals.

Unless otherwise indicated, terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art. In addition, it will be understood that terms defined in commonly used dictionaries should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense.

Fig. 1 is a flowchart of a man-machine cooperation regulation method for a large power grid according to a preferred embodiment of the present invention. The invention provides a man-machine hybrid enhanced intelligent system and a regulation method for large power grid regulation and control by combining the requirements of the current power grid regulation and control system and regulation and control service, and aims to realize the hybrid enhancement of man-machine intelligence by virtue of the respective advantages of man-machine cooperative control mode, form bidirectional information communication and control, construct a 1+1>2 power grid regulation and control man-machine hybrid enhanced intelligent system, improve power grid situation sensing and operation decision-making capability, and further reduce the working strength of regulation and control personnel.

As shown in fig. 1, the invention provides a man-machine cooperation regulation and control method for a large power grid, which comprises the following steps:

step 101: determining a target task of power grid operation regulation, and decomposing the target task into a plurality of subtasks with association relations based on an execution step of the target task;

Step 102: distributing an execution mode for each of a plurality of subtasks;

step 103: executing the subtasks based on the distributed execution mode, inputting the decision result of the executed subtasks to the associated next subtask, executing the next subtask until all the subtasks of the target task are executed, and acquiring the decision result of all the subtasks;

step 104: and regulating and controlling the operation of the power grid based on the decision results of all the subtasks.

the main body of the machine execution is an intelligent machine;

evaluating the decision result, and inputting the decision result of the sub-task after execution to the associated next sub-task when the decision result is higher than a preset threshold value;

when the decision result is not higher than a preset threshold value and the state of the subtask is non-urgent, training the machine learning model through training data, and re-executing the subtask through the trained machine learning model; or when the state of the subtask is urgent, executing the subtask by a regulating and controlling person.

Preferably, the method further comprises:

when the execution mode is the man-machine cooperation execution, the decision right during the man-machine cooperation execution is distributed based on the judgment condition, and the method comprises the following steps: switching of man-machine decision rights, man-machine auxiliary decision control and man-machine joint decision control.

Preferably, the method comprises, among other things,

the man-machine decision right switching comprises the following steps: when a preset event of switching the man-machine decision right occurs, the decision right is distributed to a regulating person or an intelligent machine;

the man-machine joint decision control comprises the following steps: the regulating personnel and the intelligent machine jointly execute subtasks.

The functional technical framework based on the large power grid regulation and control man-machine hybrid enhanced intelligent system provided by the invention constructs a regulation and control method based on man-machine cooperation for the power grid real-time regulation and control service, reasonably distributes the real-time regulation and control tasks to regulation and control personnel and machine execution, and realizes the perception and decision control of the power grid through man-machine hybrid decision. Fig. 2 is a schematic flow chart of a power grid regulation and control method based on man-machine cooperation, which mainly comprises a power grid regulation and control task planning and decision execution process, and mainly comprises the following four steps:

Step 1: and decomposing the current power grid regulation task into a plurality of mutually connected sub-tasks according to the steps in the task execution process. According to the type and the complexity of the subtasks, the method combines the intelligent level of the power grid artificial intelligent agent, and reasonably distributes the decomposed subtasks to people and machines for common execution. The subtask type can be divided into subtasks executed by a machine and subtasks executed by man-machine cooperation according to the man-machine cooperation mode.

Step 2: the subtasks that the machine is responsible for performing use a network model generated by training an algorithm such as machine learning to generate results and input the results into the next subtask. The result of the machine executing task is required to be used for evaluating the practical effect of the training model by establishing corresponding model evaluation indexes, so that the model precision under different power grid operation scenes is ensured. When the result of the evaluation index is higher than a certain prescribed threshold value, the result output by the model is input into the next subtask. When the evaluation index result is lower than the threshold, the retraining of the model or the decision making by a person is decided by judging the urgency of the subtask.

Step 3: the subtask of man-machine cooperation execution is executed by adopting a proper man-machine cooperation mode, and the man-machine mixed decision is input to the next subtask.

Step 4: judging whether the current task is completed or not, if not, executing the next subtask, and executing the flow according to the type of the subtask and the step 2 or the step 3. And if the task is completed, ending the execution of the task processing flow based on the man-machine cooperation.

The machine execution subtasks realize subtasks such as power grid fault identification, transient stability assessment, power grid running track and load prediction, auxiliary decision making and the like by utilizing a network model generated by training machine learning algorithms such as a Convolutional Neural Network (CNN), a long and short term memory neural network (LSTM), reinforcement learning (Q learning) and the like, and input the result into the next subtask. The result of the machine executing task is required to be used for evaluating the practical effect of the training model in terms of model accuracy, generalization capability and the like by establishing corresponding model evaluation indexes, so that model precision in different power grid operation scenes is ensured. And judging the subsequent operation by comparing the evaluation result with a set threshold value. When the result of the evaluation index is higher than a certain prescribed threshold value, the result output by the model is input into the next subtask. When the evaluation index result is below the threshold, the retraining of the model or the decision by a person should be decided according to the urgency of the subtask.

The key of the power grid regulation man-machine decision control link is that the distribution and coordination of decision weights under man-machine cooperation are realized by considering the states of regulation personnel, power grid and environment, so that the conflict of people and machines in the combined decision process is avoided, and the overall performance of a man-machine cooperation system is improved. Based on the safety requirement of the power grid regulation and the degree of the cooperative coupling of the man and the machine, the mode of intelligent decision of the power grid regulation and the man-machine cooperation can be divided into 3 modes of man-machine decision right switching, man-machine auxiliary decision control and man-machine combined decision control, as shown in fig. 3. The three decision control modes are adopted according to the current state of the power grid and the man-machine intelligence and capacity level. Decision control in fig. 3 (a) is actively or passively switched between regulatory personnel and the machine. When a small probability event or an emergency control scene occurs, the regulation and control personnel take over the system actively, and the decision right is classified to the regulation and control personnel. When the internal functions of the regulation and control system are wrong or tasks are too complex to exceed the functional range of the system, the regulation and control system can temporarily send out a request to complete the passive switching of regulation and control personnel. In addition, when the machine can independently process a task or regulate the fatigue state of personnel, the machine can take over decision-making rights actively or passively. Fig. 3 (b) is a diagram of human-machine-aided decision control, in which the machine refines the experience of the regulatory personnel into knowledge through the mining analysis of the regulatory behaviors of the regulatory personnel, further optimizes the decision goal, automatically provides auxiliary decisions for grid operation and abnormal treatment, and guides and helps the regulatory personnel to actively, rapidly, comprehensively and accurately master the current grid state and development trend. Fig. 3 (c) is a human-machine joint decision control, and a regulatory person and a machine jointly cooperate to control the power grid. Aiming at a power grid regulation task, a machine obtains an expected control strategy by using optimization algorithms such as machine learning, a regulation personnel control strategy prediction model is established, and a mixed decision under man-machine cooperation is obtained according to the expected control strategy and the prediction model, so that the total control quantity of the machine and the regulation personnel tends to the control quantity of the expected control strategy, thereby realizing intervention correction and control on the operation of the regulation personnel, improving the cooperation efficiency and reducing the adverse effect caused by unknown random behaviors of the personnel.

Based on the different man-machine cooperative intelligent decision control modes, the power grid regulation man-machine cooperative intelligent decision and control are realized by using an evaluation algorithm, an optimization algorithm, a machine learning algorithm and other algorithms, and the safety control capability of the complex power grid regulation is further improved.

Preferably, an execution mode is allocated to each of the plurality of subtasks, and the allocation principle includes:

The subtask allocation execution mode is defined as:

the capability vector of a person to perform a subtask is defined as h= { H ₁ ，H ₂ ，…H _n }, wherein element H _i Expressed as the ability of a person to perform subtask i, the ability vector of a machine to perform subtask is defined as m= { M ₁ ，M ₂ ，…M _n Element Mi, where element Mi represents the machine's ability to perform subtask i.

Preferably, allocating an execution mode to each of the plurality of subtasks includes:

λand/>

The main task of the power system scheduling of the invention is to monitor the real-time operation of the power grid, determine the arrangement of the power grid operation mode and the safety control measures according to the current power grid state and by combining the experience and the cognitive level of the regulating personnel, and ensure the safe and stable operation and the reliable supply of the power of the large power grid. During the operation of the power grid, regulatory personnel are faced with different types of regulatory tasks, which can be classified into simple tasks and complex tasks according to task complexity. Simple tasks can be allocated to machines for independent completion, complex tasks are usually decomposed into mutually connected or parallel sub-tasks according to basic functional units and boundary conditions of human-machine division, and then the decomposed sub-tasks are allocated to regulatory personnel and machines for execution.

In order to fully exert the initiative and the intelligence of the machine and further lighten the working intensity of the regulating personnel, a mathematical model for distributing the power grid regulation man-machine cooperation tasks needs to be established.

In consideration of the specificity of the power grid regulation and control field, the operation safety of the power grid and the intelligent degree of a man-machine system are improved in the power grid regulation and control task distribution process, and man-machine task optimization distribution is realized through the construction, quantitative evaluation and analysis of a man-machine intelligent capability model and a regulation and control personnel state behavior model. Assume that the set of regulation tasks is t= { Task ₁ ，Task ₂ ，…Task _n Decision variables for task allocation are }

The definition is shown in formula (1):

defining the capacity vector of a regulating person to execute a certain task as H= { H ₁ ，H ₂ ，…H _n }, wherein element H _i Expressed as the ability of the regulatory personnel to perform task i, such as risk factors, tolerance, etc. Defining the capability vector of the intelligent machine to perform a certain task as m= { M ₁ ，M ₂ ，…M _n Element Mi, where element Mi is expressed as the ability of the machine to perform task i, such as level of intelligence, processing task capabilities, etc. The mathematical model of human-computer collaborative task allocation can be expressed as:

λand/>

The objective function of the model can be set according to the important consideration factors of human-machine task allocation, such as minimum participation degree of regulation personnel, minimum cost of a human-machine system or minimum risk coefficient. And according to the objective function and the constraint condition of the task allocation model, reasonably dividing the power grid regulation task into human and machine by solving through an intelligent optimization algorithm.

Preferably, executing the subtasks based on the selected execution mode includes:

when the execution mode of the subtasks is the man-machine cooperation, the subtasks are sent to the digital twin system for calculation through the regulating and controlling personnel and the intelligent machine, and the decision result of the subtasks is obtained.

the operation modes include: real-time, research, planning, training, testing.

Preferably, the method further comprises: acquiring training data, the training data comprising: historical data related to grid operation and digital twin system generation data;

training the intelligent machine through training data to generate a machine learning model;

knowledge elements in the training data are extracted based on the machine learning model, and a knowledge base is generated.

The invention provides a man-machine mixing enhancement intelligent regulation method for large power grid regulation, which aims at the power grid regulation business requirement and the artificial intelligent technical characteristics, constructs a technical framework of the large power grid regulation man-machine mixing enhancement intelligent system physical and function and a regulation method based on man-machine cooperation, fully plays the advantage of intelligent fusion of regulation personnel and machines, and realizes analysis and control of the power grid regulation man-machine mixing enhancement intelligent. The invention has higher engineering applicability, can provide technical reference guidance for realizing the power grid regulation and control man-machine hybrid enhanced intelligent system based on the artificial intelligent technology, can reduce the operation risk caused by strategy mismatch due to the limitation of machine intelligence, and further improves the intelligent level of the power grid regulation and control service.

Fig. 6 is a schematic structural diagram of a man-machine cooperation regulation module for a large power grid according to a preferred embodiment of the present invention. As shown in fig. 6, the present invention provides a man-machine cooperation regulation module for a large power grid, where the module includes:

the initial unit 601 is configured to determine a target task for power grid operation regulation, and decompose the target task into a plurality of subtasks having an association relationship based on an execution step of the target task;

A processing unit 602, configured to allocate an execution manner to each of a plurality of subtasks; executing the subtasks based on the distributed execution mode, inputting the decision result of the executed subtasks to the associated next subtask, executing the next subtask until all the subtasks of the target task are executed, and acquiring the decision result of all the subtasks;

and the result unit 603 is used for regulating and controlling the operation of the power grid based on the decision results of all the subtasks.

the main body of the machine execution is an intelligent machine;

Preferably, the processing unit 602 is further configured to: when a machine execution mode is selected to execute the subtasks, acquiring decision results of the subtasks after execution;

Preferably, the processing unit 602 is further configured to:

Preferably, the processing unit 602 is configured to execute the subtasks based on the selected execution mode, including:

Preferably, the processing unit 602 is further configured to:

Preferably, the processing unit 602 is configured to allocate an execution manner to each of the plurality of subtasks, and the allocation rule includes:

The subtask allocation execution mode is defined as:

Preferably, the processing unit 602 assigns an execution mode to each of the plurality of subtasks, including:

λand/>

Preferably, the processing unit 602 is further configured to: training sample data is generated through a digital twin system, and a machine learning model is trained and knowledge is acquired through the training sample data.

the operation modes include: real-time, research, planning, training, testing.

Preferably, the processing unit 602 is further configured to: acquiring training data, the training data comprising: historical data related to grid operation and digital twin system generation data; the method comprises the steps of carrying out a first treatment on the surface of the

Preferably, the processing unit 602 is further configured to: and distributing an execution mode for each sub-task in the plurality of sub-tasks according to the complexity degree of the sub-tasks. The invention provides a man-machine hybrid enhanced intelligent system and a man-machine hybrid enhanced intelligent system control method for large power grid control, which are characterized by mainly comprising a man-machine hybrid enhanced intelligent system overall architecture design based on a D5000 control system, an AI intelligent system, a power grid digital twin system and control personnel, interaction of unit modules of all subsystems and a power grid dispatching operation control method based on man-machine cooperation.

The man-machine cooperation regulation module 600 facing the large power grid in the preferred embodiment of the present invention corresponds to the man-machine cooperation regulation method 100 facing the large power grid in the preferred embodiment of the present invention, and will not be described herein.

Fig. 7 is a schematic structural diagram of a man-machine hybrid power grid regulation intelligent device according to a preferred embodiment of the invention. The invention provides a man-machine mixing enhancement intelligent device for large power grid regulation and control by combining the current power grid regulation and control system and the requirements of regulation and control service. The device needs to rely on the data acquisition at the front end and the system control at the rear end of the existing regulation and control system to serve as key support links for intelligent enhancement of man-machine mixing, and is an upgrade superposition on the existing D5000 regulation and control system. The invention provides a man-machine mixing enhancement intelligent device for large power grid regulation, which comprises a part of power grid regulation system D5000 scheduling control system, a power grid digital twin system, an AI intelligent system and a man-machine interaction system.

As shown in fig. 7, the present invention provides a man-machine hybrid power grid regulation intelligent device, which includes: a power grid regulation and control automatic system 701, an AI intelligent system 702, a man-machine interaction system 703 and a power grid digital twin system 704;

the power grid regulation and control automatic system 701, wherein the power grid regulation and control automatic system 701 part comprises the original functions of power grid safety protection and automatic control part and the functions of data acquisition and instruction issuing; the power grid regulation and control automatic system 701 is used for acquiring power grid operation data and transmitting the power grid operation data to the power grid digital twin system 704 and the man-machine interaction system 703; the instruction issuing function is used for executing a decision instruction according to a decision result to complete a power grid regulation task;

The power grid regulation and control automatic system 701 is exemplified by a D5000 system, the D5000 system collects data such as power grid operation steady state, dynamic and transient state information, secondary equipment state information and auxiliary detection information in real time by using a data collection and monitoring control (Supervisory control and data acquisition, SCADA) system, and provides power grid real-time/historical data and power grid model information to a power grid digital twin system for offline and online analysis and calculation of a power grid. The data acquisition information and the stability control, AGC and AVC related advanced application information in the D5000 system interact with regulatory personnel through a man-machine display interface and provide power grid state information, so that the capability of the regulatory personnel for system perception is improved.

The AI intelligent system 702 and the man-machine interaction system 703 are respectively used for executing subtasks after decomposing the power grid regulation task;

the grid digital twin system 704 is used for executing the calculation, analysis and check tasks of the AI system 702 or the man-machine interaction system 703 in the process of executing the subtasks, and feeding back the obtained task decision result to the task initiator AI intelligent system or the man-machine interaction system.

Preferably, the AI intelligent system 702 trains the machine learning model with the grid operation history data and the digital twin system generation data as training data to generate the machine learning model and the knowledge elements; the target task is performed by a machine learning model.

Preferably, the AI intelligence system 702 further includes a knowledge base that utilizes a machine learning model to extract knowledge elements in the sample data, which are represented and stored in different forms.

Preferably, the AI intelligent system 702 and the man-machine interaction system 703 decompose and execute the target task, including:

determining an execution step of a target task, and decomposing the target task into a plurality of subtasks with association relations;

and allocating an execution mode for each of the plurality of subtasks.

The AI intelligence system 702 of the present invention is an important part of a man-machine hybrid enhancement device that interacts with other systems for data and knowledge. Sources of data and knowledge include primarily data and knowledge related to the system internals and external data and knowledge related to the application requirements. The system internal data is structured and unstructured data collected and recorded by the D5000 system on one hand, and unstructured text data aiming at the existing power grid, such as stable operation regulations, regulation operation regulations, fault treatment plans and the like on the other hand. External data is mainly non-power system information such as weather information from other systems and the internet, social information, and the like. Because the operation interval of the power system is mainly concentrated in the typical operation mode range, the collected data cannot cover all operation modes and scenes of the system, the data-driven AI intelligent system needs to solve the problem of faults or small samples of atypical scenes by giving simulation data generated by the grid twin system tasks to the AI intelligent system through people and the AI intelligent system, and unbiased grid training data samples are realized. The AI intelligent system performs data screening and key feature extraction aiming at key factors affecting the problem solving, selects a reasonable and efficient machine learning model, performs model training through numerical value type, category type, text, image and other multi-type data, and builds the machine learning model with high accuracy and strong generalization capability. The knowledge base utilizes a machine learning model to extract knowledge elements from training data samples, and utilizes different knowledge representation forms to realize symbolization, formalization or modeling of knowledge and experience in the power grid regulation field. The model library and the knowledge base of the AI intelligent system can judge whether the model and the knowledge need to be updated and expanded in a mode of model evaluation model (such as recall rate, accuracy, precision and the like) or man-machine interaction, and can upgrade and enhance the model and the knowledge in the AI intelligent system by combining the data obtained by a man-machine given task and an external database and the knowledge base so as to ensure the accuracy of the model and the knowledge in the AI intelligent system in the application process. The output of the AI intelligent system is mainly the prediction result of the machine learning model and the related knowledge established by the knowledge base, and the output is provided for regulatory personnel and a D5000 regulatory system execution link in a man-machine interaction mode, so that the AI intelligent system is used for power grid perception and decision making process under man-machine cooperation.

The man-machine interaction system 703 of the present invention is used as an intelligent machine system service object and final "value judgment", which is an arbitrator and a daemon for the safe operation of the power grid, and the interaction cooperation with the machine intelligence is through the power grid state sensing and decision making process. The control personnel acquire data and information related to the operation of the power grid through a visual human-computer interaction interface of each system of the power grid control, and convert the acquired data and information into knowledge through a human brain induction summary and deduction mode for perception and decision of the power grid. The physiological information and the ability evaluation information of the regulatory personnel are transmitted to the AI intelligent system 702 to be used as the basis of the man-machine hybrid intelligent decision. Besides the traditional keyboard, mouse and other input modes, the interaction mode of the regulation personnel and the system can also perform man-machine interaction through voice, gestures, eye, and other modes, and the advanced multichannel man-machine interaction technology can better realize state perception and intention understanding of the machine on the regulation personnel, so that man-machine cooperation capacity in a power grid regulation task is improved, and further the degree of man-machine fusion is promoted.

The power grid digital twin system 704 realizes accurate mapping of the power grid through big data acquisition and digital model construction, is a virtual digital representation mode of a real power grid from data to a model, and can be used for describing and describing different modes of a physical entity of the power grid. The power grid digital twin system 704 has power grid topology and physical model, power grid history and real-time operation data and other information, and can simulate and predict various business processes under different time scales of the power grid based on the information, and realize quick check. In the intelligent system for the man-machine hybrid enhancement of power grid regulation, the system mainly bears the work of offline and online analysis and calculation and check, and the functions related to the analysis and calculation, auxiliary decision and the like in the original power grid regulation can be integrated in the digital twin system 704 of the power grid.

The grid twin system can adopt a robot flow automation technology (Robotic process automation, RPA) to realize the automatic execution of the tasks in the process of executing the specific tasks, and the process is to convert the grid data to the level of the grid knowledge. And the obtained result after the task is executed is also provided for a regulating person and an AI intelligent system in the form of data samples and knowledge, so that the man-machine knowledge collaboration and the optimal evolution are realized.

The AI intelligent system 702 and the man-machine interaction system 703 execute the subtasks based on the distributed execution mode, input the decision result of the executed subtask to the associated next subtask, execute the next subtask until all the subtasks completing the target task are executed, and obtain the decision result of all the subtasks. The protection and automatic control loop in the power grid regulation and control device utilizes the real-time operation information of the power grid to realize the automatic execution of the protection and control of the power grid, and forms a rapid closed-loop control device with the power grid. In the calculation and execution process, the participation degree of regulatory personnel is limited, and the automation and the intellectualization of the system are more relied on. The man-machine joint decision loop is a link in which people (regulatory personnel) and machines (power grid dispatching control system) participate together, relates to interaction cooperation between people and machines, and is a key link for realizing man-machine mixing enhancement intelligence by applying an artificial intelligence technology to the loop. The power grid regulation and control man-machine hybrid intelligent device is composed of an AI intelligent system, a power grid regulation and control system, a power grid digital twin system and a man-machine interaction system.

Preferably, the grid digital twin system 704 and the AI intelligent system 702 are deployed in a management information area of the device;

the machine learning model and knowledge base of the AI intelligence system 702 is deployed at a production control large area of the device.

In order to meet the requirements of safety protection of the power system, the device is deployed in different power system information safety partitions according to the service and functional characteristics, the physical architecture design of the device is shown in figure 4,

(1) At present, multi-source data such as power grid model data, real-time operation data and management data are collected to a regulation and control cloud platform, and together with infrastructure (CPU resources, storage resources and network resources) of the regulation and control cloud, the system can provide data, calculation power, algorithms and services for an artificial intelligent platform applied to the field of power grid regulation and control. Therefore, the AI intelligent system based on data driving and the regulation cloud platform are deployed in the same management information area (namely a safety III area), so that the interaction of data information between the systems and the learning and training of a model are facilitated. The trained model library and the constructed knowledge library in the AI intelligent system can be directly used for sensing and deciding the state of the power grid, so that mirror images of the AI intelligent system are required to be deployed in a production control large area (namely a safe I area and a safe II area), and the AI intelligent system updates the model library and the knowledge library from a management information large area to the production control large area through a reverse isolation device, so that the full life cycle management of the AI model is realized.

(2) The twin system is used as an interaction environment for analysis and calculation of regulatory personnel and the AI intelligent system and is not directly used for power grid control, so that the twin system and the AI intelligent system are deployed in the same management information area, interaction with the AI intelligent system and the regulatory personnel is facilitated, and autonomous cooperation and optimal evolution of man-machine hybrid intelligence are realized.

(3) And regulatory personnel interact and cooperate with the AI intelligent system and the grid twin system deployed under different network safety partitions through man-machine interfaces, so that the capability of dispatching and controlling the complex grid is improved.

Fig. 5 is a framework diagram of a power grid regulation and control man-machine hybrid enhanced intelligent device, and the functions of each module and mutual information of a man-machine intelligent system are clearly formed, so that the mutual data information bidirectional intercommunication among a power grid regulation and control system, a power grid twin system, an AI system and regulation and control personnel is realized. The device is constructed by adding an AI intelligent system and a power grid twin system on the aspects of the existing D5000 system and regulation personnel, and the specific functions are as follows:

(1) The AI intelligent system is a core subsystem of a power grid regulation and control system based on man-machine hybrid enhanced intelligent, and mainly comprises an AI platform and AI application facing power grid regulation and control perception and decision service, wherein the bottom layer of the AI platform is composed of hardware facilities and data, the hardware facilities comprise computing resources (CPU, GPU, TPU and the like), storage resources and network equipment, the data comprise structured and semi/unstructured regulation and control data from the structural point of view, the data stored in a relational database are structured data, files such as XML and JSON belong to the semi-structured data, the regulation and control text data (stable operation regulation, fault treatment plan and the like) and pictures/videos/audios are unstructured data. The AI platform provides the functions of data preprocessing, model and algorithm, model training, evaluation and the like based on the hardware facilities and the data, on the basis, a model library and a knowledge base which are suitable for the power grid regulation and control service are constructed, the performance of the model is analyzed and evaluated by establishing evaluation indexes, the update and upgrading of the model library and the knowledge base are realized, and the accuracy and the generalization capability of the model are further improved. The AI platform provides basic support in data and algorithm for AI application of upper power grid regulation.

The model library is a collection of artificial intelligence models constructed for regulatory business applications. The AI platform performs model training by using hardware calculation power and regulation service data training samples and algorithms such as machine learning, deep learning, reinforcement learning and the like, so as to obtain AI models with different model structures and weight parameters. The model is incorporated into unified management of a model library, and provides model support for model sharing and regulation business intelligent application of cross-information security partition.

The knowledge base is a set of various knowledge constructed for regulation and control service application, such as power grid characteristic knowledge, causal knowledge, associated knowledge and the like, and can organize a knowledge subset through various knowledge representation modes such as first-order logic predicates, production rules, RDF triples and the like, so that an expert system and a knowledge map system are constructed and applied to a power grid regulation and control service scene. The expert system pays attention to logical reasoning and is suitable for regulation and control business with smaller knowledge scale. The knowledge graph is more related knowledge representation, a large-scale knowledge base is formed by carrying out knowledge extraction, knowledge fusion and knowledge processing through a machine, the retrieval of the fact knowledge is emphasized, a certain reasoning can be completed, and the knowledge graph is suitable for intelligent searching, intelligent recommending, intelligent question answering and intelligent decision making in the power grid regulation and control business. The knowledge base can be constructed together in a manual mode and an automatic mode such as machine learning, natural language processing and the like, so that knowledge subsystems with different data scales are formed and applied to specific tasks of power grid regulation.

(2) The power grid digital twin system can support analysis and calculation under the power grid polymorphism mode, wherein the analysis and calculation comprises real-time state, research state, planning state, training state and testing state. The power grid analysis and calculation under different modes is used for solving different scene tasks in power grid regulation. The real-time state is that the twin system utilizes data such as real-time acquisition data and offline mode to analyze and calculate the current state of the power grid operation mode, gives out corresponding scheduling auxiliary decisions, and can realize the application functions such as power grid steady state analysis and online safety and stability analysis. The research state is that the twin system performs case analysis work such as power grid characteristics, power grid operation mode calculation, fault inversion and the like according to tasks to be processed by regulatory personnel and operation mode personnel, and can realize application functions such as power grid mode calculation, safety control strategy generation and the like. The planning state is the prediction of the twin system for trend information such as future power grid new energy power generation, load and the like, the prediction information is used for power grid planning research analysis under the future power grid structure change, auxiliary decision calculation is carried out aiming at potential safety hazards, and the application functions such as prediction, scheduling planning, safety check and the like can be realized. The training state is a regulation and control environment which is completely the same as an actual control center and can simulate the static and dynamic response of the power system, so that a dispatcher is familiar with and grasps various functions of the system in the simulated dispatching environment, the operation experience of the regulator under different working conditions of a power grid is improved by carrying out targeted autonomous training and anti-accident exercise on the knowledge skill level evaluation of the regulator, and the function is that the existing dispatcher training simulation system is tested and debugged in the aspects of network model, advanced application function and the like.

The invention has been described with reference to a few embodiments. However, as is well known to those skilled in the art, other embodiments than the above disclosed invention are equally possible within the scope of the invention, as defined by the appended patent claims.

Generally, all terms used in the claims are to be interpreted according to their ordinary meaning in the technical field, unless explicitly defined otherwise therein. All references to "a// the [ means, component, etc ]" are to be interpreted openly as referring to at least one instance of means, component, etc., unless explicitly stated otherwise. The steps of any method disclosed herein do not have to be performed in the exact order disclosed, unless explicitly stated.

Claims

1. A method for man-machine cooperation regulation and control for a large power grid, the method comprising:

distributing an execution mode for each of a plurality of subtasks;

2. The method of claim 1, the performing comprising: machine execution and man-machine cooperation execution;

the main body of the machine execution is an intelligent machine;

3. The method according to claim 2, comprising: when a machine execution mode is selected to execute the subtasks, acquiring decision results of the subtasks after execution;

4. The method of claim 1, wherein the allocating an execution manner to each of the plurality of subtasks includes:

The subtask allocation execution mode is defined as:

the capability vector of a person to perform a subtask is defined as h= { H ₁ ，H ₂ ，…H _n }, wherein element H _i The ability to perform subtask i for a personForce, the capability vector of a machine to execute a subtask is defined as m= { M ₁ ，M ₂ ，…M _n Element Mi, where element Mi represents the machine's ability to perform subtask i.

5. The method of claim 4, wherein allocating an execution mode to each of the plurality of subtasks comprises:

the mathematical model of the human-computer cooperation task allocation is expressed as:

L

an upper threshold and a lower threshold of capability values of the human and the machine for executing subtasks respectively; m is the number of sub-targets; n is the number of subtasks.

6. The method of claim 2, further comprising:

7. The method of claim 6, wherein,

8. The method of claim 2, the performing the subtask based on the selected execution mode comprising:

9. The method of claim 2, further comprising: training sample data is generated through a digital twin system, and the intelligent learning model is trained and knowledge is acquired through the training sample data.

10. The method of claim 8, the digital twin system comprising a plurality of modes of operation, the tasks of different scenarios in grid operation regulation being handled by different modes of operation;

the operation mode includes: real-time, research, planning, training, testing.

11. The method of claim 2, further comprising: acquiring training data, the training data comprising: historical data related to grid operation and digital twin system generation data;

12. A large grid-oriented man-machine collaboration regulation module, the module comprising:

13. The module of claim 12, the execution means comprising: machine execution and man-machine cooperation execution;

the main body of the machine execution is an intelligent machine;

14. The module of claim 13, the processing unit further to: when a machine execution mode is selected to execute the subtasks, acquiring decision results of the subtasks after execution;

15. The module of claim 12, wherein the processing unit is configured to allocate an execution manner for each of a plurality of subtasks, and the allocation rule includes:

The subtask allocation execution mode is defined as:

16. The module of claim 15, the processing unit to allocate execution for each of a plurality of subtasks, comprising:

L

17. The module of claim 13, the processing unit further to:

18. The module of claim 17, the processing unit further to:

19. The module of claim 13, the processing unit to perform the subtasks based on a selected execution mode, comprising:

20. The module of claim 13, the processing unit further to: training sample data is generated through a digital twin system, and the intelligent learning model is trained and knowledge is acquired through the training sample data.

21. The module of claim 19, the digital twinning system comprising a plurality of modes of operation, the tasks of different scenarios in grid operation regulation being handled by different modes of operation;

the operation mode includes: real-time, research, planning, training, testing.

22. The module of claim 13, the processing unit further to: acquiring training data, the training data comprising: historical data related to grid operation and digital twin system generation data; training the intelligent machine through the training data to generate a machine learning model;

23. An intelligent man-machine hybrid power grid regulation device, the device comprising: the system comprises a power grid regulation and control automatic system, an AI intelligent system, a man-machine interaction system and a power grid digital twin system;

the power grid digital twin system is used for executing calculation, analysis and checking tasks of the AI system or the man-machine interaction system in the process of executing the subtasks, and feeding back the obtained task decision result to the AI intelligent system or the man-machine interaction system of the task initiator.

24. The apparatus of claim 23, the AI intelligence system to train a machine learning model with grid operation history data and digital twin system generation data as training data to generate a machine learning model and knowledge elements; and executing the target task through the machine learning model.

25. The apparatus of claim 24, the AI intelligence system further comprising a knowledge base that utilizes the machine learning model to extract knowledge elements in sample data, the knowledge elements being represented and stored in different forms.

26. The apparatus of claim 23, the AI intelligence system and human-machine interaction system to decompose and execute the target task, comprising:

And allocating an execution mode for each of the plurality of subtasks.

27. The apparatus of claim 26, wherein the AI intelligent system and the human-computer interaction system execute the subtasks based on the assigned execution mode, input decision results of the subtasks after execution to an associated next subtask, and execute the next subtask until all the subtasks that complete the target task are executed, and obtain decision results of all the subtasks.

28. The apparatus of claim 23, the grid digital twin system and the AI intelligent system deployed in a management information community of the apparatus;

the machine learning model and the knowledge base of the AI intelligent system are deployed in a production control large area of the device.