A safety training method for thermal power generation enterprises based on virtual reality technology

Abstract

The present invention discloses a safety training method for thermal power generation enterprises based on virtual reality technology, and the content includes data collection, constructing a training basic framework, constructing a training role function model, safety hazard monitoring, and comprehensive training. The present invention relates to the technical field of training safety management, and specifically refers to a safety training method for thermal power generation enterprises based on virtual reality technology. This solution improves the realism and practicability of the scenario by generating a training scenario and constructing a virtual training role framework; by designing an incentive function and performing multi-dimensional quantum fusion, it improves the generalization ability and feature extraction ability of the model, improves the efficiency and accuracy of information fusion, and improves the accuracy and reliability of the prediction result; by designing a matching degree calculation function, designing a data cluster center point update function, and setting an abnormal inspection threshold to perform real-time hazard monitoring, it improves the real-time performance and accuracy of monitoring and improves the identification efficiency of safety hazards.

Description

Technical Field

[0001] The present invention relates to the technical field of training safety management, and specifically refers to a safety training method for thermal power generation enterprises based on virtual reality technology. Background Art

[0002] A safety training method for thermal power generation enterprises based on virtual reality technology combines virtual reality technology and artificial intelligence technology to construct a real and comprehensive scene model, providing a simulation environment with a high sense of immersion and realism, and making intelligent feedback based on the operations and responses of trainees. However, the traditional safety training methods for thermal power generation enterprises have problems such as insufficiently realistic scene construction, insufficient complexity assessment, and a single role framework when constructing the training basic framework; problems such as insufficient model generalization ability, insufficient feature extraction, and lack of an efficient information fusion mechanism when constructing the training role function model; problems such as insufficient monitoring accuracy, poor real-time performance, and difficulty in effectively identifying potential safety hazards in safety hazard monitoring. Summary of the Invention

[0003] In view of the above situation, to overcome the defects of the prior art, the present invention provides a safety training method for thermal power generation enterprises based on virtual reality technology. Aiming at the problems of insufficiently realistic scene construction, insufficient complexity assessment, and a single role framework existing in the traditional safety training methods for thermal power generation enterprises when constructing the training basic framework, this solution improves the realism and practicality of the scene and makes the virtual training role more three-dimensional and rich by defining element complexity, calculating scene complexity, generating training scenes, and constructing a virtual training role framework; aiming at the problems of insufficient model generalization ability, insufficient feature extraction, and lack of an efficient information fusion mechanism existing in the traditional safety training methods for thermal power generation enterprises when constructing the training role function model, this solution improves the model generalization ability and feature extraction ability, improves the efficiency and accuracy of information fusion, and improves the accuracy and reliability of prediction results by designing an incentive function and performing multi-dimensional quantum fusion; aiming at the problems of insufficient monitoring accuracy, poor real-time performance, and difficulty in effectively identifying potential safety hazards in the traditional safety training methods for thermal power generation enterprises in safety hazard monitoring, this solution improves the real-time performance and accuracy of monitoring and improves the identification efficiency and accuracy of safety hazards by defining the enhanced correlation distance, designing a matching degree calculation function, designing a data cluster center point update function, setting an abnormal inspection threshold, and performing real-time hazard monitoring.

[0004] The technical solution adopted by the present invention is as follows: A safety training method for thermal power generation enterprises based on virtual reality technology, the method comprising the following steps:

[0005] Step S1: Data collection;

[0006] Step S2: Construct the training basic framework;

[0007] Step S3: Construct the training role function model;

[0008] Step S4: Monitor potential safety hazards;

[0009] Step S5: Comprehensive training.

[0010] Furthermore, in Step S1, the data collection is to collect equipment operation data, personnel operation data, historical fault data, virtual scenario data, and training role data; the equipment operation data includes the operation parameters of generators, boilers, cooling systems, and flue gas treatment systems; the personnel operation data includes the operation content and operation results of operators, and the operation results include normal and abnormal; the historical fault data is the historical equipment fault records and the equipment parameters at the time of the fault; the virtual scenario data includes the equipment models, building models, and topographic models of the training site; the training role data is the information of training instructors with training experience in reality.

[0011] Furthermore, in Step S2, the construction of the training basic framework specifically includes the following steps:

[0012] Step S21: Construct the training scenario. A virtual training scenario is constructed through the collected equipment models, building models, and terrain models of the training site, which specifically includes the following steps:

[0013] Step S211: Define the element complexity. Each equipment, building, and terrain is set as an element, and the element complexity is defined as follows:

[0014] ;

[0015] where q represents the index of the element, represents the complexity of the q-th element, , and represent the complexity weights, represents the volume of the q-th element, represents the number of components of the q-th element, represents the number of interactive functions of the q-th element;

[0016] Step S212: Calculate the scenario complexity. The equipment models, building models, and terrain models of the training site are combined to generate an initial training scenario, which is expressed as follows:

[0017] ;

[0018] where U represents the index of the initial training scenario, Represents the scenario complexity of the U-th training scenario, Represents the total number of elements within the training scenario, Represents the balance weight of the q-th element;

[0019] Step S213: Generate a training scenario, calculate the scenario complexity of the generated initial training scenario, and select the scenario with the lowest complexity as the final training scenario;

[0020] Step S22: Construct a virtual training role framework, create attribute items for the virtual training tutor. The attributes of the virtual training tutor include: name, gender, appearance characteristics, personality, professional knowledge, behavior style, and background story. Input the collected training role data into the created attribute items to construct the framework of the virtual training role.

[0021] Furthermore, in step S3, the construction of the training role function model specifically includes the following steps:

[0022] Step S31: Set the model label data, and set the operation result data as the label data of the model;

[0023] Step S32: Design the input layer, expressed as follows:

[0024] ;

[0025] Among them, Represents the output value of the input layer, Represents the initial weight, d represents the total number of input features, , And Respectively represent the 1st, 2nd, and d-th input features;

[0026] Step S33: Design the directional excitation function, expressed as follows:

[0027] ;

[0028] Among them, Represents the input value of the directional excitation function, And Represent the scale parameters, Represents the directional excitation function;

[0029] Step S34: Design the hidden layer, expressed as follows:

[0030] ;

[0031] Among them, Represents the output value of the l-th hidden layer, Represents the weight of the l-th hidden layer, represents the input value of the l-th hidden layer, represents the bias of the l-th hidden layer;

[0032] Step S35: Multi-dimensional quantum fusion, specifically including the following steps:

[0033] Step S351: Quantum state initialization, expressed as follows:

[0034] ;

[0035] where r represents the index of the qubit, R represents the maximum number of qubits, represents the initial state of the r-th qubit, and are complex numbers that satisfy and and are the basis states of the qubit;

[0036] Step S352: Quantum feature encoding, expressed as follows:

[0037] ;

[0038] where Qe represents a quantum feature vector composed of R quantum states, represents the initial state of the first qubit, represents the initial state of the second qubit, represents the initial state of the R-th qubit;

[0039] Step S353: Quantum transformation, expressed as follows:

[0040] ;

[0041] where, represents the transformed state of the r-th qubit after quantum transformation, represents a 2x2 complex matrix, deo represents a random integer between 1 and R, represents the initial state of the deo-th qubit, represents the tensor product operator of the qubit;

[0042] Step S354: Quantum feature fusion, expressed as follows:

[0043] ;

[0044] where, represents the feature vector after quantum feature fusion, represents the quantum feature fusion weight;

[0045] Step S36: Design the output layer, expressed as follows:

[0046] ;

[0047] Among them, represents the operation prediction result output by the output layer, represents taking the result with the highest probability in the prediction result, represents the normalized exponential activation function, , and respectively represent the weights, input values, and biases of the output layer.

[0048] Furthermore, in step S4, the safety hazard monitoring specifically includes the following steps:

[0049] Step S41: Define the enhanced association distance, expressed as follows:

[0050] ;

[0051] Among them, i and j represent the indices of data points, and respectively represent the positions of the i-th data point and the j-th data point, represents the enhanced association distance function, represents the enhanced association distance between the i-th data point and the j-th data point, represents taking the maximum value, and respectively represent the abscissa and ordinate of the i-th data point, and respectively represent the abscissa and ordinate of the j-th data point, represents taking the absolute value, represents the enhanced weight, represents taking the square root;

[0052] Step S42: Design the matching degree calculation function, expressed as follows:

[0053] ;

[0054] Among them, represents the matching degree between the i-th data point and the j-th data point, represents the dimension index of the data, represents the maximum dimension of the data, represents taking the Euclidean distance, represents the -th data dimension value of the i-th data point, represents the -th data dimension value of the j-th data point;

[0055] Step S43: Design the data cluster center point update function, which is expressed as follows:

[0056] ;

[0057] Among them, k represents the index of the data cluster, t represents the number of data cluster updates, represents the position of the k-th data cluster center at the (t + 1)-th data cluster update, represents taking the data point with the closest distance, represents the total number of data points in the k-th data cluster, ki represents the data point index of the k-th data cluster, represents the position of the ki-th data point, represents the position of the k-th data cluster center at the t-th data cluster update;

[0058] Step S44: Cluster construction. Randomly select B data points from all data points as the initial cluster center points, calculate the matching degree between each data point and each cluster center point, assign the data points to the cluster to which the cluster center with the highest matching degree belongs, update the position of the cluster center, increment the number of updates by one, and iterate the update process until the position of the cluster center no longer changes, forming the final data cluster;

[0059] Step S45: Set the anomaly detection threshold, which is expressed as follows:

[0060] ;

[0061] Among them, represents the anomaly detection threshold, represents the Euclidean distance between the two farthest cluster center points in the final data cluster;

[0062] Step S46: Real-time hidden danger monitoring. Real-time collect the user's training data, calculate the Euclidean distance between the user's training data point and each cluster center point. If there is a Euclidean distance less than the anomaly detection threshold between any cluster center point, then the training data point is normal at this time; otherwise, there is a potential safety hazard, and a warning message is issued.

[0063] Furthermore, in step S5, the comprehensive training is to collect the data in the user's training in real time, input it into the training role function model, the model processes and analyzes the user's training operations, outputs the predicted operation results, conducts real-time training guidance, and at the same time monitors the potential safety hazards of the user's training data.

[0064] The beneficial effects achieved by the present invention using the above solution are as follows:

[0065] (1) Aiming at the problems of insufficiently realistic scenario construction, insufficient complexity evaluation, and single role framework in the construction of the training basic framework of traditional thermal power generation enterprise safety training methods, this solution improves the realism and practicality of the scenario and makes the virtual training roles more three-dimensional and rich by defining element complexity, calculating scenario complexity, generating training scenarios, and constructing a virtual training role framework.

[0066] (2) Aiming at the problems of insufficient model generalization ability, insufficient feature extraction, and lack of an efficient information fusion mechanism in the construction of the training role function model of traditional thermal power generation enterprise safety training methods, this solution improves the model generalization ability and feature extraction ability, improves the efficiency and accuracy of information fusion, and improves the accuracy and reliability of prediction results by designing an activation function and performing multi-dimensional quantum fusion.

[0067] (3) Aiming at the problems of insufficient monitoring accuracy, poor real-time performance, and difficulty in effectively identifying potential safety hazards in the safety hazard monitoring of traditional thermal power generation enterprise safety training methods, this solution improves the real-time performance and accuracy of monitoring and improves the identification efficiency and accuracy of safety hazards by defining the enhanced correlation distance, designing a matching degree calculation function, designing a data cluster center point update function, setting an abnormal inspection threshold, and performing real-time hazard monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 It is a schematic flow chart of a safety training method for thermal power generation enterprises based on virtual reality technology provided by the present invention;

[0069] Figure 2 It is a schematic flow chart of step S2;

[0070] Figure 3 It is a schematic flow chart of step S3;

[0071] Figure 4 It is a schematic flow chart of step S4.

[0072] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0073] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments; based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0074] In the description of the present invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc. indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on the present invention.

[0075] Example 1, refer to Figure 1 , a safety training method for thermal power generation enterprises based on virtual reality technology provided by the present invention, the method comprising the following steps:

[0076] Step S1: Data collection, collecting equipment operation data, personnel operation data, historical fault data, virtual scene data, and training role data;

[0077] Step S2: Constructing a training basic framework, constructing a training basic framework by constructing a training scene and a virtual training role framework;

[0078] Step S3: Constructing a training role function model, constructing a training role function model by designing an input layer, designing an activation function, designing a hidden layer, multi-dimensional quantum fusion, and designing an output layer;

[0079] Step S4: Safety hazard monitoring, performing safety hazard monitoring by defining a reinforcement correlation distance, designing a matching degree calculation function, designing a data cluster center point update function, and setting an abnormal inspection threshold;

[0080] Step S5: Comprehensive training, collecting data in real-time during user training, inputting it into the training role function model, the model providing real-time training guidance, and at the same time performing safety hazard monitoring on the user's training data.

[0081] Example 2, refer to Figure 1 , this embodiment is based on the above embodiment. In step S1, the data collection is to collect equipment operation data, personnel operation data, historical fault data, virtual scene data, and training role data; the equipment operation data includes the operation parameters of generators, boilers, cooling systems, and flue gas treatment systems; the personnel operation data includes the operation content and operation results of operators, and the operation results include normal and abnormal; the historical fault data is the historical equipment fault records and the equipment parameters at the time of the fault; the virtual scene data includes the equipment models, building models, and terrain and landform models of the training site; the training role data is the information of training tutors with training experience in reality.

[0082] Example 3, refer to Figure 1 and Figure 2, this embodiment is based on the above embodiment. In step S2, the construction of the training basic framework specifically includes the following steps:

[0083] Step S21: Construct a training scenario. A virtual training scenario is constructed by collecting the equipment model, building model, and terrain model of the training site, which specifically includes the following steps:

[0084] Step S211: Define the element complexity. Each piece of equipment, building, and terrain is set as an element, and the element complexity is defined as follows:

[0085] ;

[0086] where q represents the index of the element, represents the complexity of the q-th element, , and represent the complexity weights, represents the volume of the q-th element, represents the number of components of the q-th element, represents the number of interactive functions of the q-th element;

[0087] Step S212: Calculate the scene complexity. The equipment model, building model, and terrain model of the training site are combined to generate an initial training scene, which is expressed as follows:

[0088] ;

[0089] where U represents the index of the initial training scene, represents the scene complexity of the U-th training scene, represents the total number of elements in the training scene, represents the balance weight of the q-th element;

[0090] Step S213: Generate the training scene. Calculate the scene complexity of the generated initial training scene, and select the scene with the lowest complexity as the final training scene;

[0091] Step S22: Construct a virtual training role framework. Create the attribute items of the virtual training tutor. The attributes of the virtual training tutor include: name, gender, appearance characteristics, personality, professional knowledge, behavior style, and background story. Input the collected training role data into the created attribute items to construct the framework of the virtual training role.

[0092] By performing the above operations, for the problems of insufficiently realistic scenario construction, insufficient complexity evaluation, and single role framework in the construction of the training basic framework of traditional thermal power generation enterprises, this solution improves the realism and practicality of the scenario and makes the virtual training roles more three-dimensional and rich by defining element complexity, calculating scenario complexity, generating training scenarios, and constructing a virtual training role framework.

[0093] Example 4, refer to Figure 1 and Figure 3 , based on the above example, the specific steps of constructing the training role function model are as follows:

[0094] Step S31: Set the model label data, and set the operation result data as the label data of the model;

[0095] Step S32: Design the input layer, which is expressed as follows:

[0096] ;

[0097] Among them, represents the output value of the input layer, represents the initial weight, d represents the total number of input features, , and respectively represent the 1st, 2nd, and dth input features;

[0098] Step S33: Design the directional excitation function, which is expressed as follows:

[0099] ;

[0100] Among them, represents the input value of the directional excitation function, and represent the scale parameters, represents the directional excitation function;

[0101] Step S34: Design the hidden layer, which is expressed as follows:

[0102] ;

[0103] Among them, represents the output value of the lth hidden layer, represents the weight of the lth hidden layer, represents the input value of the lth hidden layer, represents the bias of the lth hidden layer;

[0104] Step S35: Multidimensional quantum fusion, specifically including the following steps:

[0105] Step S351: Quantum state initialization, expressed as follows:

[0106] ;

[0107] where r represents the index of the qubit, R represents the maximum number of qubits, represents the initial state of the r-th qubit, and are complex numbers satisfying , and are the basis states of the qubit;

[0108] Step S352: Quantum feature encoding, expressed as follows:

[0109] ;

[0110] where Qe represents a quantum feature vector composed of R quantum states, represents the initial state of the first qubit, represents the initial state of the second qubit, represents the initial state of the R-th qubit;

[0111] Step S353: Quantum transformation, expressed as follows:

[0112] ;

[0113] where represents the transformed state of the r-th qubit after quantum transformation, represents a 2x2 complex matrix, deo represents a random integer between 1 and R, represents the initial state of the deo-th qubit, represents the tensor product operator of the qubit;

[0114] Step S354: Quantum feature fusion, expressed as follows:

[0115] ;

[0116] where represents the feature vector after quantum feature fusion, represents the quantum feature fusion weight;

[0117] Step S36: Design the output layer, expressed as follows:

[0118] ;

[0119] where represents the operation prediction result output by the output layer, It represents taking the result with the highest probability in the prediction results. It represents the sigmoid activation function. 、 and respectively represent the weights, input values, and biases of the output layer.

[0120] By performing the above operations, for the problems of insufficient model generalization ability, insufficient feature extraction, and lack of an efficient information fusion mechanism in the traditional safety training method for thermal power generation enterprises when constructing a training role function model, this solution improves the model's generalization ability and feature extraction ability, as well as the efficiency and accuracy of information fusion, by designing an activation function and performing multi-dimensional quantum fusion, thus improving the accuracy and reliability of the prediction results.

[0121] Example 5, referring to Figure 1 and Figure 4 , based on the above example, in step S4, the safety hazard monitoring specifically includes the following steps:

[0122] Step S41: Define the enhanced correlation distance, expressed as follows:

[0123] ;

[0124] where i and j represent the indices of data points, and respectively represent the positions of the i-th data point and the j-th data point, represents the enhanced correlation distance function, represents the enhanced correlation distance between the i-th data point and the j-th data point, represents taking the maximum value, and respectively represent the abscissa and ordinate of the i-th data point, and respectively represent the abscissa and ordinate of the j-th data point, represents taking the absolute value, represents the enhanced weight, represents taking the square root;

[0125] Step S42: Design a matching degree calculation function, expressed as follows:

[0126] ;

[0127] where, represents the matching degree between the i-th data point and the j-th data point, represents the dimension index of the data, represents the maximum dimension of the data, denotes taking the Euclidean distance, denotes the th data dimension value of the i-th data point, denotes the th data dimension value of the j-th data point;

[0128] Step S43: Design a function for updating the center point of the data cluster, which is expressed as follows:

[0129] ;

[0130] where k represents the index of the data cluster, t represents the number of times the data cluster is updated, denotes the position of the k-th data cluster center at the (t + 1)-th data cluster update, denotes taking the data point with the closest distance, denotes the total number of data points in the k-th data cluster, ki represents the data point index in the k-th data cluster, denotes the position of the ki-th data point, denotes the position of the k-th data cluster center at the t-th data cluster update;

[0131] Step S44: Cluster construction. Randomly select B data points from all data points as the initial cluster center points, calculate the matching degree between each data point and each cluster center point, assign the data points to the cluster to which the cluster center point with the highest matching degree belongs, update the position of the cluster center, increment the number of update times, and iterate the update process until the position of the cluster center no longer changes, forming the final data cluster;

[0132] Step S45: Set the anomaly detection threshold, which is expressed as follows:

[0133] ;

[0134] where, denotes the anomaly detection threshold, denotes the Euclidean distance between the two cluster center points with the farthest distance in the final data cluster;

[0135] Step S46: Real-time hidden danger monitoring. Real-time collect the training data of users, calculate the Euclidean distance between the training data points of users and each cluster center point. If there exists a Euclidean distance less than the anomaly detection threshold between any cluster center point, then the training data point is normal at this time; otherwise, there is a potential safety hazard, and a warning message is issued.

[0136] By performing the above operations, in view of the problems of insufficient monitoring accuracy, poor real-time performance, and difficulty in effectively identifying potential safety hazards in the safety training methods of traditional thermal power generation enterprises, this solution improves the real-time performance and accuracy of monitoring, as well as the identification efficiency and accuracy of safety hazards, by defining the enhanced correlation distance, designing the matching degree calculation function, designing the data cluster center point update function, setting the abnormal inspection threshold, and performing real-time hazard monitoring.

[0137] Embodiment 6. Refer to Figure 1 , based on the above embodiment, in step S5, the comprehensive training is to collect the data in the user training in real time, input it into the training role function model, the model processes and analyzes the user's training operations, outputs the predicted operation results, conducts real-time training guidance, and at the same time monitors the safety hazards of the user's training data.

[0138] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprises", "comprising" or any other variation thereof is intended to cover a non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements, but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0139] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention.

[0140] The above describes the present invention and its implementation manners. Such a description is not restrictive. What is shown in the drawings is only one of the implementation manners of the present invention, and the actual structure is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and design similar structural manners and embodiments without creative efforts without departing from the purpose of the present invention, they shall fall within the protection scope of the present invention.

Claims

1. A safety training method for thermal power generation enterprises based on virtual reality technology, characterized in that, The method includes the following steps: Step S1: Data collection, collecting equipment operation data, personnel operation data, historical fault data, virtual scenario data, and training role data; Step S2: Construct a training basic framework, and construct the training basic framework by constructing a training scenario and a virtual training role framework; Step S3: Construct a training role function model, and construct the training role function model by designing an input layer, designing an activation function, designing a hidden layer, multi-dimensional quantum fusion, and designing an output layer; In step S3, the construction of the training role function model specifically includes the following steps: Step S31: Set model label data, and set the operation result data as the label data of the model; Step S32: Design the input layer, which is expressed as follows: ； Among them, represents the output value of the input layer, represents the initial weight, d represents the total number of input features, , and represent the 1st, 2nd, and dth input features respectively; Step S33: Design a directional activation function, which is expressed as follows: ； Among them, represents the input value of the directional excitation function, e represents the natural constant, and represent the scale parameter, represents the directional excitation function; Step S34: Design the hidden layer, which is expressed as follows: ； Among them, represents the output value of the l-th hidden layer, represents the weight of the l-th hidden layer, represents the input value of the l-th hidden layer, represents the bias of the l-th hidden layer; Step S35: Multi-dimensional quantum fusion, specifically including the following steps: Step S351: Quantum state initialization, which is expressed as follows: ； where r represents the index of the qubit, and R represents the maximum number of qubits, represents the initial state of the r-th qubit, and are complex numbers satisfying and and are the basis states of the qubits; Step S352: Quantum feature encoding, which is expressed as follows: ； Among them, Qe represents a quantum feature vector composed of R quantum states, represents the initial state of the first qubit, represents the initial state of the second qubit, represents the initial state of the R-th qubit; Step S353: Quantum transformation, which is expressed as follows: ； Among them, represents the transformed state of the r-th qubit after quantum transformation, represents a 2×2 complex matrix, and deo represents a random integer between 1 and R, represents the initial state of the deo-th qubit, represents the tensor product operator of qubits; Step S354: Quantum feature fusion, which is expressed as follows: ； Among them, represents the eigenvector after quantum feature fusion, represents the quantum feature fusion weight; Step S36: Design the output layer, which is expressed as follows: ； Among them, represents the operation prediction result output by the output layer, represents taking the result with the highest probability in the prediction result, represents the sigmoid activation function, 、 and respectively represent the weight, input value, and bias of the output layer; Step S4: Safety hazard monitoring, and conduct safety hazard monitoring by defining the enhanced correlation distance, designing a matching degree calculation function, designing a data cluster center point update function, and setting an anomaly detection threshold; Step S5: Comprehensive training, collect the data in the user's training in real time, input it into the training role function model, the model conducts training guidance in real time, and at the same time conduct safety hazard monitoring on the user's training data.

2. The safety training method for thermal power generation enterprises based on virtual reality technology according to claim 1, characterized in that: In step S4, the safety hazard monitoring specifically includes the following steps: Step S41: Define the enhanced correlation distance, which is expressed as follows: ； where i and j represent the indices of data points, and represent the positions of the i-th data point and the j-th data point respectively, represents the enhanced correlation distance function, represents the enhanced correlation distance between the i-th data point and the j-th data point, represents taking the maximum value, and represent the abscissa and ordinate of the i-th data point respectively, and represent the abscissa and ordinate of the j-th data point respectively, represents taking the absolute value, represents the enhanced weight, represents taking the square root; Step S42: Design a matching degree calculation function, which is expressed as follows: ； Among them, represents the matching degree between the i-th data point and the j-th data point, represents the dimension index of the data, represents the maximum dimension of the data, represents taking the Euclidean distance, represents the -th data dimension value of the i-th data point, represents the -th data dimension value of the j-th data point; Step S43: Design a data cluster center point update function, which is expressed as follows: ； Among them, k represents the index of the data cluster, and t represents the number of data cluster updates. represents the position of the k-th data cluster center at the (t + 1)-th data cluster update. represents taking the data point with the closest distance. represents the total number of data points in the k-th data cluster, and ki represents the index of the data point in the k-th data cluster. represents the position of the ki-th data point. represents the position of the k-th data cluster center at the t-th data cluster update. Step S44: Cluster construction, randomly select B data points from all data points as the initial cluster center points, calculate the matching degree between each data point and each cluster center point, assign the data points to the cluster to which the cluster center with the highest matching degree belongs, update the position of the cluster center, increment the number of updates by one, and iterate the update process until the position of the cluster center no longer changes, forming the final data cluster; Step S45: Set the anomaly detection threshold, which is expressed as follows: ； Among them, represents the abnormal detection threshold, represents the Euclidean distance between the two cluster centers that are farthest apart in the final data clusters; Step S46: Real-time hazard monitoring, collect the user's training data in real time, calculate the Euclidean distance between the user's training data points and each cluster center point, if the Euclidean distance between any cluster center point exists and is less than the anomaly detection threshold, then the training data point is normal at this time; otherwise, there is a safety hazard and a warning message is issued.

3. The safety training method for thermal power generation enterprises based on virtual reality technology according to claim 1 is characterized in that: In step S2, the construction of the training basic framework specifically includes the following steps: Step S21: Construct a training scenario, and construct a virtual training scenario through the equipment model, building model, and terrain model of the training site collected, specifically including the following steps: Step S211: Define the element complexity. Set each device, building, and terrain as an element, and define the element complexity as follows: ； where q represents the index of the element, represents the complexity of the q-th element, , and represent the complexity weights, represents the volume of the q-th element, represents the number of components of the q-th element, represents the number of interactive functions of the q-th element; Step S212: Calculate the scene complexity. Combine the device model, building model, and terrain model of the training ground to generate an initial training scene, which is expressed as follows: ； where U represents the index of the initial training scenario, represents the scenario complexity of the U-th training scenario, represents the total number of elements within the training scenario, represents the balance weight of the q-th element; Step S213: Generate the training scene. Calculate the scene complexity of the generated initial training scene, and select the scene with the lowest complexity as the final training scene; Step S22: Construct the virtual training role framework. Create the attribute items of the virtual training tutor. The attributes of the virtual training tutor include: name, gender, appearance characteristics, personality, professional knowledge, behavior style, and background story. Input the collected training role data into the created attribute items to construct the framework of the virtual training role.

4. A safety training method for thermal power generation enterprises based on virtual reality technology according to claim 1, characterized in that: In Step S1, the data collection is to collect device operation data, personnel operation data, historical fault data, virtual scene data, and training role data; the device operation data includes the operation parameters of generators, boilers, cooling systems, and flue gas treatment systems; the personnel operation data includes the operation content and operation results of operators, and the operation results include normal and abnormal; the historical fault data is the historical device fault records and the device parameters at the time of the fault; the virtual scene data includes the device model, building model, and topographic and geomorphic model of the training ground; the training role data is the information of the training tutor with training experience in reality; In Step S5, the comprehensive training is to collect the data in the user's training in real time, input it into the training role function model, and the model processes and analyzes the user's training operations, outputs the predicted operation results, conducts real-time training guidance, and monitors the safety hazards of the user's training data at the same time.

Images