Limited space safety early warning method and device, electronic equipment and storage medium

Abstract

The invention discloses a safety early warning method and device for a limited space, electronic equipment and a storage medium, and relates to the technical field of electric power. The method comprises the steps that an actual safety score in a limited space is determined according to target data in the limited space, and the target data at least comprises environment data, equipment data, personnel state data of operators and operation data in the limited space; inputting the actual safety score into a safety state neural network to obtain a target score which is used for representing the overall safety risk level in the limited space; and early warning information is generated according to an early warning threshold value, the actual safety score and the target score, and the early warning threshold value is periodically adjusted according to the actual safety score and the target score. The technical problem of poor early warning reliability caused by the fact that safety early warning is often carried out on a limited space based on a single data source in the prior art is solved.

Description

Technical Field

[0001] The present application relates to the field of electric power technology, and in particular to a safety warning method, device, electronic device and storage medium for a confined space. Background Art

[0002] In power construction, confined space operations often involve closed or poorly ventilated areas, such as shafts of power equipment, tunnels, and inside substations. There are many challenges when working in these environments. Therefore, it is crucial to ensure the safety of workers working in confined spaces.

[0003] However, current early warning systems mostly rely on a single data source, such as monitoring of gas concentration or oxygen levels. This method has certain limitations. A single data source cannot fully reflect all potential risks in a confined space. At the same time, since the environment in a confined space changes rapidly and is highly dynamic, this dynamic change cannot be captured in real time by a single data source, which in turn affects the accuracy of the early warning. In addition, the physiological state of the operator is also an important factor affecting operational safety, but the existing early warning system usually ignores this point, and the current early warning system cannot adjust the early warning method in real time according to the operator's status and changes in the environment in the confined space, resulting in unreliable early warning.

[0004] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0005] The present application provides a method, device, electronic device and storage medium for safety warning of a confined space, so as to at least solve the technical problem in the prior art that safety warning of a confined space is often performed based on a single data source, resulting in poor warning reliability.

[0006] In order to achieve the above-mentioned purpose, according to one aspect of the present application, a safety warning method for a confined space is provided, comprising: determining an actual safety score within a confined space based on target data within the confined space, wherein the target data at least includes environmental data, equipment data, personnel status data of operators, and operation data within the confined space; inputting the actual safety score into a safety status neural network to obtain a target score, wherein the target score is used to characterize the overall safety risk level within the confined space; generating warning information based on a warning threshold, an actual safety score, and a target score, wherein the warning threshold is periodically adjusted based on the actual safety score and the target score.

[0007] Optionally, the safety status neural network is trained through the following steps, including: obtaining historical data within a limited space, wherein the historical data at least includes historical environmental data within the limited space and personnel status data of historical operators; extracting key features from the historical data, and fusing the key features according to the weight of each key feature to obtain feature vectors at different times, wherein the key features at least include historical environmental parameters within the limited space and physiological characteristics of historical operators, wherein the historical environmental parameters at least include temperature, humidity, gas concentration, and oxygen content, and the physiological characteristics of historical operators at least include heart rate, blood pressure, and blood oxygen saturation; obtaining a multimodal data set based on the feature vector and the risk score value of each feature vector; and obtaining the safety status neural network based on the multimodal data set.

[0008] Optionally, a safe state neural network is obtained based on training of a multimodal data set, including: obtaining an initial learning rate, a number of input layer neurons, and a number of output layer neurons of the initial neural network, wherein the initial learning rate is used to characterize the update frequency of the initial settings of weights and bias vectors in the initial neural network; determining a target learning rate of the initial neural network based on the initial learning rate and a decay factor, wherein the target learning rate is used to improve the convergence speed of the initial neural network, and the decay factor is used to characterize the decay rate of the initial learning rate; determining a target number of hidden layers of the initial neural network and the number of neurons in each hidden layer based on the number of input layer neurons and the number of output layer neurons, wherein the hidden layer is located between the input layer and the output layer of the initial neural network, and the target number and the number of neurons are used to improve the performance of the initial neural network; and performing multiple iterative training on the initial neural network based on the multimodal data set to obtain a safe state neural network through training.

[0009] Optionally, after the safe state neural network is obtained based on the training of the multimodal data set, the safety warning method for a limited space also includes: dividing the multimodal data set into N subsets, where N is an integer greater than 1; performing K rounds of testing operations on the safe state neural network to obtain K prediction results, wherein each round of testing operation is used to select any one of the N subsets as a target test set, and use the subset other than the target test set as a target training set, use the target training set to perform a round of testing on the safe state neural network, and use the tested safe state neural network to predict the target test set to obtain a prediction result; determine K accuracy rates of the safe state neural network based on the K prediction results, wherein the Jth accuracy rate among the K accuracy rates is used to characterize the prediction accuracy of the safe state neural network under the Jth test operation in the K rounds of testing operations; and calculate the average of the K accuracy rates as the prediction accuracy of the safe state neural network.

[0010] Optionally, warning information is generated based on the warning threshold, the actual safety score and the target score, including: calculating the absolute difference between the actual safety score and the target score; updating the warning threshold based on the absolute difference and the adjustment coefficient to obtain the target warning threshold; generating warning information based on the target warning threshold and the actual safety score.

[0011] Optionally, warning information is generated based on the target warning threshold and the actual safety score, including: determining a first numerical value based on environmental data in the confined space and the weight of the environmental data, wherein the first numerical value is used to characterize the risk intensity caused by environmental factors in the confined space; taking the absolute difference between the first numerical value and the second numerical value as the target difference; determining the risk type in the confined space based on the target warning threshold, the first numerical value, the second numerical value and the target difference, wherein the risk type is used to distinguish the source category of the risk in the confined space; determining the risk level in the confined space based on the actual safety score and at least one preset risk threshold, wherein the risk level is used to characterize the degree of risk in the confined space; and generating warning information based on the risk type and the risk level.

[0012] Optionally, the risk type in the confined space is determined according to the target warning threshold, the first numerical value, the second numerical value and the target difference, including: when the first numerical value is greater than the second numerical value and the target difference is greater than the target warning threshold, it is determined that a first type of risk occurs in the confined space, wherein the first type of risk is used to characterize the risk caused by environmental factors; when the first numerical value is less than the second numerical value and the target difference is greater than the target warning threshold, it is determined that a second type of risk occurs in the confined space, wherein the second type of risk is used to characterize the risk caused by physical factors of the operator; when the target difference is less than or equal to the target warning threshold, it is determined that a third type of risk occurs in the confined space, wherein the third type of risk is used to characterize the risk caused by both environmental factors and physical factors of the operator.

[0013] According to another aspect of the present application, a safety warning device for a confined space is also provided, including: a determination unit, which determines an actual safety score within the confined space based on target data within the confined space, wherein the target data at least includes environmental data, equipment data, personnel status data of operators, and operation data within the confined space; an input unit, which inputs the actual safety score into a safety status neural network to obtain a target score, wherein the target score is used to characterize the overall safety risk level within the confined space; a generation unit, which generates warning information based on a warning threshold, an actual safety score, and a target score, wherein the warning threshold is periodically adjusted based on the actual safety score and the target score.

[0014] According to another aspect of the present application, an electronic device is also provided, wherein the electronic device includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by one or more processors, the one or more processors execute the above-mentioned limited space security warning method.

[0015] According to another aspect of the present application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer-readable storage medium, wherein when the computer program is executed, the device where the computer-readable storage medium is located executes the above-mentioned limited space security warning method.

[0016] From the above content, it can be seen that in this application, the actual safety score in the confined space is first determined based on multiple data sources in the confined space to ensure the comprehensiveness of the data source, and then the actual safety score is input into the security status neural network to obtain the target score, wherein the target score is used to characterize the overall safety risk level in the confined space, and then the warning information is generated based on the warning threshold, the actual safety score and the target score, wherein the warning threshold is periodically adjusted according to the actual safety score and the target score to ensure the reliability of the warning.

[0017] Compared with the prior art that provides safety warnings to workers when working in confined spaces based on a single data source, the present application determines the actual safety score of the confined space based on real-time environmental data, equipment data, worker status data, operation data and other relevant data in the confined space, and inputs the actual safety score value determined based on multiple data sources into the safety status neural network to obtain a target score. Then, the warning threshold in the confined space is periodically adjusted based on the actual safety score obtained from the real-time data and the target score obtained from the safety status neural network, thereby achieving the purpose of real-time adjustment of the warning threshold based on the real-time data in the confined space, and generating warning information for reminders, thereby achieving the technical effect of improving the reliability of warnings, and further solving the technical problem that the prior art often provides safety warnings for confined spaces based on a single data source, resulting in poor warning reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0019] Figure 1 is a flow chart of an optional safety warning method for a confined space according to an embodiment of the present application;

[0020] Figure 2It is a schematic diagram of an optional safety warning device for a confined space according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] It should also be noted that the information and data collected by this application are authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of relevant data comply with the relevant laws, regulations and standards of the relevant regions, take necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation portals for users to choose to authorize or refuse. For example, an interface is set up between this system and relevant users or institutions. Before obtaining relevant information, it is necessary to send an acquisition request to the aforementioned user or institution through the interface, and obtain relevant information after receiving the consent information fed back by the aforementioned user or institution.

[0024] According to an embodiment of the present application, an embodiment of a safety warning method for a confined space is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0025] It should be noted that an intelligent early warning system can be used as the execution subject of the safety early warning method for a confined space in the embodiment of the present application. It is understandable that the safety early warning method for a confined space provided in the embodiment of the present application can also be executed by other systems or devices, and the embodiment of the present application does not specifically limit this.

[0026] Figure 1 is a flow chart of an optional safety warning method for a confined space according to an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0027] Step S101, determining an actual safety score within the limited space according to target data within the limited space.

[0028] In step S101 , the target data includes at least environmental data, equipment data, operator status data, and operation data within a limited space.

[0029] Optionally, the environmental data in the confined space includes but is not limited to the temperature, humidity, gas concentration, and oxygen content in the confined space.

[0030] Optionally, the personnel status data of the workers in the confined space include but are not limited to the heart rate, blood pressure, and blood oxygen saturation of the workers.

[0031] Step S102, input the actual safety score into the safety status neural network to obtain the target score.

[0032] In step S102, the target score is used to characterize the overall safety risk level within the confined space.

[0033] Optionally, the intelligent early warning system inputs the actual safety score into a trained safety status neural network to obtain a predicted score value within a limited space.

[0034] Step S103, generating warning information according to the warning threshold, the actual safety score and the target score.

[0035] In step S103, the warning threshold is periodically adjusted according to the actual safety score and the target score.

[0036] Optionally, the warning threshold is adjusted in real time according to the actual safety score and the target score.

[0037] It can be seen from the contents of steps S101 to S103 that in the present application, the actual safety score in the confined space is first determined based on the target data in the confined space, wherein the target data includes at least environmental data, equipment data, personnel status data of the operators, and operation data in the confined space. The actual safety score is then input into the safety status neural network to obtain a target score, wherein the target score is used to characterize the overall safety risk level in the confined space. Finally, warning information is generated based on the warning threshold, the actual safety score, and the target score, wherein the warning threshold is periodically adjusted based on the actual safety score and the target score.

[0038] From the above content, it can be seen that in this application, the actual safety score in the confined space is first determined based on multiple data sources in the confined space to ensure the comprehensiveness of the data source, and then the actual safety score is input into the security status neural network to obtain the target score, wherein the target score is used to characterize the overall safety risk level in the confined space, and then the warning information is generated based on the warning threshold, the actual safety score and the target score, wherein the warning threshold is periodically adjusted according to the actual safety score and the target score to ensure the reliability of the warning.

[0039] Compared with the prior art that provides safety warnings to workers when working in confined spaces based on a single data source, the present application determines the actual safety score of the confined space based on real-time environmental data, equipment data, worker status data, operation data and other relevant data in the confined space, and inputs the actual safety score value determined based on multiple data sources into the safety status neural network to obtain a target score. Then, the warning threshold in the confined space is periodically adjusted based on the actual safety score obtained from the real-time data and the target score obtained from the safety status neural network, thereby achieving the purpose of real-time adjustment of the warning threshold based on the real-time data in the confined space, and generating warning information for reminders, thereby achieving the technical effect of improving the reliability of warnings, and further solving the technical problem that the prior art often provides safety warnings for confined spaces based on a single data source, resulting in poor warning reliability.

[0040] In an optional embodiment, the intelligent early warning system first obtains historical data within a limited space, wherein the historical data at least includes historical environmental data and historical operating personnel status data within the limited space, then extracts key features from the historical data, and fuses the key features according to the weight of each key feature to obtain feature vectors at different times, wherein the key features at least include historical environmental parameters and historical operating personnel physiological characteristics within the limited space, wherein the historical environmental parameters at least include temperature, humidity, gas concentration, and oxygen content, and the physiological characteristics of historical operating personnel at least include heart rate, blood pressure, and blood oxygen saturation, then obtains a multimodal data set based on the feature vector and the risk score value of each feature vector, and finally obtains a safe status neural network based on training the multimodal data set.

[0041] Optionally, the intelligent early warning system first obtains historical data within a limited space and extracts key features of the historical data, wherein the key features in the historical environmental data within the limited space include at least temperature, humidity, gas concentration and oxygen content, and the key features in the historical personnel status data of the operators include at least the heart rate, blood pressure and blood oxygen saturation of the operators. After extracting these key features, these key features are fused to obtain a fused feature vector, and different feature vectors are obtained according to different samples. Then, a multimodal data set is obtained according to each feature vector obtained and the risk score value corresponding to each feature vector. Finally, a safe status neural network is trained based on the multimodal data set.

[0042] Optionally, the intelligent early warning system processes the historical environmental data and calculates the mean value T of all temperatures collected in the historical time period. kz and the mean of all humidity H kz , and calculate the rate of change of gas concentration G within a given time interval, as shown in formula (1):

[0043]

[0044] Among them, ΔG(t) represents the rate of change of the gas concentration G, G(t) represents the gas concentration G at the current time point t, G(t-Δt) represents the gas concentration G corresponding to the time point t-Δt, and Δt represents the time interval (seconds).

[0045] Optionally, the intelligent warning system calculates the rate of change of the oxygen content OC within a given time interval, as shown in formula (2):

[0046]

[0047] Wherein, ΔOC(t) represents the rate of change of the oxygen content OC, OC(t) represents the oxygen content OC at the current time point t, and OC(t-Δt) represents the oxygen content OC corresponding to the time point t-Δt.

[0048] Optionally, the intelligent early warning system processes the personnel status data of historical operators and calculates the mean blood pressure BP of each operator collected during the historical time period. kz , and the heart rate variation coefficient of each operator is calculated to evaluate the amplitude of the heart rate variation of each operator, as shown in formula (3):

[0049]

[0050] Among them, HRV represents the heart rate variation coefficient of each operator, HR i Represents the heart rate measurement value of each operator for the i-th time, HR kz represents the average value of all heart rate measurements of each operator, and n represents the number of heart rate measurements of each operator.

[0051] Optionally, the intelligent early warning system calculates the rate of change of the blood oxygen saturation of each staff member within a given time interval, as shown in formula (4):

[0052]

[0053] Among them, ΔSP(t) represents the rate of change of the blood oxygen saturation SP of each operator, SP(t) represents the blood oxygen saturation SP of each operator at the current time point t, and SP(t-Δt) represents the blood oxygen saturation SP of each operator corresponding to the time point t-Δt.

[0054] Optionally, the intelligent early warning system fuses the extracted features, as shown in formula (5):

[0055] F tud =[w1·ΔG(t)+w2·ΔOC(t)+w3·HRV+w4·ΔSP(t)+w5·T kz +w6·H kz +

[0056] w7 BP kz ](5)

[0057] Among them, F tud is the fused feature vector, w1, w2, w3, w4, w5, w6 and w7 represent ΔG(t), ΔOC(t), HRV, ΔSP(t), T kz , H kz and BP kz The weight of .

[0058] Optionally, the intelligent early warning system obtains a multimodal data set based on the fused features and the risk score value of each feature vector, as shown in formula (6):

[0059] D={(F1, S1), (F2, S2), (F3, S3), ..., (F u , S u )} (6)

[0060] Among them, D is a multimodal dataset, F u is the feature vector after the fusion of the u-th sample, S u Represents the risk score value after the fusion of the u-th sample, where u is the number of samples.

[0061] Optionally, the safe state neural network refers to a BP (Backpropagation Neural Network Model) neural network model.

[0062] Optionally, the BP neural network model includes an input layer, a hidden layer and an output layer, wherein the input layer contains seven nodes corresponding to the fused D, the hidden layer contains 10 nodes, and the output layer contains 1 node. The input is the actual safety score S acl , the output is the comprehensive safety score S zs (target score), where the actual safety score S acl It is pre-set by the management personnel based on the environmental data, equipment data, personnel status data of the operators, and operation data within the limited space.

[0063] Optionally, the intelligent early warning system initializes the weight matrix W and the bias vector b of the BP neural network model, as shown in formula (6):

[0064]

[0065] Among them, n inc is the number of input nodes, 0 represents the mean, which is used to indicate that the expected value of the weight matrix W is 0. Represents the standard deviation, which is used to indicate that the standard deviation of the weight matrix W is n inc Decide, It means the mean is 0 and the variance is Sampling from a normal distribution.

[0066] Optionally, the intelligent early warning system uses forward propagation to calculate the score prediction value S of the output layer pv , as shown in formulas (7)-(9):

[0067] Z (1) =W (1) D+b (1) (7)

[0068] A (1) =f(Z (1) ) (8)

[0069] S pv =W (2) A (1) +b (2) (9)

[0070] Among them, Z (1) represents the linear combination of the hidden layers, A (1) represents the activation output of the hidden layer, W (1) and b (1) are the weight matrix and bias vector input to the hidden layer, W (2) and b (2) are the weight matrix and bias vector from the hidden layer to the output layer, respectively, f(Z (1) ) represents the activation function of the hidden layer.

[0071] Optionally, the intelligent early warning system calculates the score prediction value S by the mean square error pv and comprehensive safety score S zs The difference between them is shown in formula (10):

[0072]

[0073] Among them, L represents the score prediction value S pv and comprehensive safety score S zs The loss function between .

[0074] Optionally, the intelligent early warning system calculates the gradient of the output layer by back-propagation and updates the weight matrix and bias vector, as shown in formula (11):

[0075]

[0076] Among them, δ (2) is the error term of the output layer, is the loss function L with respect to the score prediction value S pv gradient.

[0077] Optionally, the intelligent early warning system calculates the gradient of the hidden layer, as shown in formula (12):

[0078] δ (1) =δ (2) W (2) f d (Z (1) ) (12)

[0079] Among them, δ (1) is the error term of the hidden layer, f d (Z (1) ) is the derivative of the activation function f of the hidden layer.

[0080] Optionally, the intelligent early warning system calculates the gradient of the weight rectangle and the bias, as shown in formulas (13)-(16):

[0081]

[0082] in, is the loss function L with respect to the weight matrix W from the hidden layer to the output layer (2) The gradient of is the loss function L with respect to the bias vector b from the hidden layer to the output layer (2) The gradient of is the loss function L with respect to the weight matrix W from the input layer to the hidden layer (1) The gradient of is the loss function L with respect to the bias vector b from the input layer to the hidden layer (1) gradient.

[0083] Optionally, the intelligent early warning system updates the weight matrix and the bias vector, as shown in formula (17) and formula (18):

[0084]

[0085] Among them, W (i) is the weight matrix of the i-th layer, b (i) is the bias vector of the i-th layer, and η is the learning rate, which is used to control the step size of parameter update.

[0086] From the above content, it can be seen that the intelligent early warning system integrates the historical environmental data in the confined space and the physiological status data of the operators, providing a comprehensive information basis. By extracting key features, the system can identify key factors affecting safety, such as environmental parameters (temperature, humidity, gas concentration, oxygen content) and physiological characteristics (heart rate, blood pressure, blood oxygen saturation). At the same time, the constructed multimodal data set provides a rich source of information for the system to simulate and predict, which helps to more comprehensively understand the safety status of confined space operations. Based on the trained neural network, the system can predict and warn of potential safety risks in confined spaces, so as to take measures in advance and reduce the possibility of accidents.

[0087] In an optional embodiment, the intelligent early warning system obtains an initial learning rate, the number of input layer neurons, and the number of output layer neurons of the initial neural network, wherein the initial learning rate is used to characterize the update frequency of the initial settings of the weights and bias vectors in the initial neural network, and then determines a target learning rate of the initial neural network based on the initial learning rate and the attenuation factor, wherein the target learning rate is used to improve the convergence speed of the initial neural network, and the attenuation factor is used to characterize the attenuation rate of the initial learning rate, and then determines a target number of hidden layers of the initial neural network and the number of neurons in each hidden layer based on the number of input layer neurons and the number of output layer neurons, wherein the hidden layer is located between the input layer and the output layer of the initial neural network, and the target number and the number of neurons are used to improve the performance of the initial neural network, and finally performs multiple iterative training on the initial neural network based on the multimodal data set to obtain a safe state neural network through training.

[0088] Optionally, the intelligent early warning system optimizes the initial parameters of the BP neural network and makes the learning rate η adaptive by attenuation, as shown in formula (19):

[0089]

[0090] Among them, η0 is the initial learning rate, ω is the current iteration number, τ is the time constant of learning rate decay, and e is the base of the exponential function.

[0091] Optionally, the intelligent early warning system determines the optimal number of hidden layers through cross-validation, as shown in formula (20):

[0092]

[0093] Where Q is the number of hidden layers, is the number of neurons in the input layer, is the number of neurons in the output layer.

[0094] Optionally, the intelligent early warning system determines the configuration of the neuron, as shown in formula (21):

[0095]

[0096] Here, Β represents the number of neurons in each hidden layer.

[0097] Optionally, the intelligent early warning system uses an improved PSO algorithm (Improved Particle Swarm Optimization) to further optimize the BP neural network, mainly to improve the weight and bias parameters of the network to achieve better training effect and prediction performance. First, a particle swarm is randomly generated, and each particle represents a set of parameters of the BP neural network. The parameters are the weight matrix Wzl and the bias vector b zl , where the particle position P zl As shown in formula (22):

[0098] P zl =[W zl1 , W zl2 , …, W zlg , b zl1 , b zl2 , …, b zlj ] (twenty two)

[0099] Where g is the number of weight rectangles and j is the number of bias vectors.

[0100] Optionally, the intelligent early warning system uses the parameters of each particle zl to train the BP neural network and calculates the fitness function, as shown in formula (23):

[0101]

[0102] Among them, MSE zl represents the fitness value of the zlth particle, usually the mean square error, N is the number of training samples, t k is the target output of the kth sample, o zlk is the predicted output of the kth sample using the parameters of particle zl.

[0103] Optionally, if the fitness of particle zl is better than the previously recorded individual optimal fitness Then update its individual optimal position, that is, if but in, represents the best fitness value found by the zlth particle so far, represents the best position found by the zlth particle so far. At the same time, if MSE zl is the smallest among all particles, then update the global optimal position, that is, if Then P best =P zl ,in, represents the best fitness value found so far by all particles, P best represents the global optimal position, that is, the parameter set corresponding to the minimum MSE among all particles.

[0104] Optionally, the intelligent early warning system updates the velocity and position of the particle, as shown in formula (24):

[0105]

[0106] Where φ is the inertia weight used to balance exploration and exploitation, c1 and c2 represent cognitive and social acceleration constants, and r 1zl and r 2zl is the random number of particle zl, generated from a uniform distribution [0, 1], v zl (t) represents the velocity vector of the zlth particle at time t, v zl (t+1) represents the new velocity vector of the zlth particle at time t+1, i.e., the velocity of the next iteration, P zl (t+1) represents the new position vector of the zlth particle at time t+1, according to its new velocity v zl (t+1) Update, P zl (t) represents the position of the zlth particle in the search space at time t.

[0107] Optionally, the intelligent early warning system sets a maximum number of iterations J CHG Repeat the update of the particle fitness value, the individual optimal position, the global optimal position, and the particle position and velocity until the maximum number of iterations J is reached. CHG , stop updating.

[0108] From the above content, it can be seen that the intelligent early warning system can automatically adjust the learning rate through the initial learning rate and the attenuation factor, which helps to learn quickly in the early stage of training, and gradually reduce the learning rate in the later stage of training to refine the adjustment of weights, thereby improving the convergence speed of the network. At the same time, the number of hidden layers and the number of neurons in each hidden layer are determined according to the number of neurons in the input layer and the output layer. This structural design helps the network to better capture the complex features in the data, and further optimizes the neural network with the improved PSO algorithm, improves the prediction accuracy and response speed of the model, and makes the warning more timely. Finally, by using multimodal data sets for training, the system can process and learn data of different types and forms, which helps to improve the adaptability and accuracy of the model to different data sources. The model is optimized through multiple iterative training, and each iteration is adjusted according to the results of the previous one. In this way, the performance of the model can be gradually improved until a satisfactory level of safety status prediction is reached.

[0109] In an optional embodiment, the intelligent early warning system divides the multimodal data set into N subsets, where N is an integer greater than 1, performs K rounds of test operations on the safe state neural network, and obtains K prediction results, wherein each round of test operation is used to select any one of the N subsets as a target test set, and uses the subset other than the target test set as a target training set, uses the target training set to perform a round of testing on the safe state neural network, uses the tested safe state neural network to predict the target test set to obtain a prediction result, and determines K accuracy rates of the safe state neural network based on the K prediction results, wherein the Jth accuracy rate among the K accuracy rates is used to characterize the prediction accuracy of the safe state neural network under the Jth test operation in the K rounds of test operations, and then calculates the average of the K accuracy rates as the prediction accuracy of the safe state neural network.

[0110] Optionally, the intelligent early warning system uses a cross-validation method to evaluate the performance of the BP neural network model, dividing the multimodal data set D into N subsets, each of which is roughly equal in size, denoted as D1, D2, ..., D k Perform K rounds of testing on the safe state neural network. In each cycle, select one of the subsets as the test set D tese , all other subsets are used as training sets for model training. In the first cycle, D1 is used to perform one round of testing on the safe state neural network. In the second cycle, D2 is used to perform two rounds of testing on the safe state neural network. And so on, K prediction results are obtained. Based on the K prediction results and the true label of each test set, the K accuracy rates of the safe state neural network are determined, as shown in formula (25):

[0111]

[0112] Among them, Acc represents the accuracy, TP i is the true number of cases in the i-th cycle, TN i is the number of true negative examples in the i-th cycle, FP i is the number of false positives in the i-th cycle, FN i is the number of false negatives in the i-th cycle.

[0113] Optionally, the intelligent early warning system calculates the average accuracy of all k rounds of cycles to obtain the prediction accuracy of the model, as shown in formula (26):

[0114]

[0115] Among them, AAy is the average value of accuracy Acc.

[0116] Optionally, after each cycle, use the test set D teseEvaluate the performance of the model and obtain the overall performance of the model, as shown in formula (27):

[0117] Pfae=f evalue (M ME , D tese ) (27)

[0118] Among them, M ME represents the BP neural network model, Pfae is the performance index after evaluation, and f evalue is the evaluation function.

[0119] From the above content, it can be seen that the intelligent early warning system divides the data set into multiple subsets, so that the model can be trained and tested under different data configurations, which helps to evaluate the performance of the model in different situations. By adopting the cross-validation method, the model performance can be evaluated under different conditions to ensure its stability and effectiveness in a variety of actual operating environments. By calculating the average accuracy of multiple rounds of tests, the system can derive a comprehensive prediction accuracy and provide a quantitative indicator to evaluate the performance of the model. By analyzing the accuracy under different test rounds and the overall performance of the model, the system can identify which aspects of the model have better prediction effects and which aspects need improvement, thereby providing guidance for further optimization of the model.

[0120] In an optional embodiment, the intelligent early warning system calculates the absolute difference between the actual safety score and the target score, and then updates the early warning threshold according to the absolute difference and the adjustment coefficient to obtain the target early warning threshold, and finally generates early warning information according to the target early warning threshold and the actual safety score.

[0121] Optionally, the intelligent early warning system analyzes the output of the BP neural network and automatically adjusts the early warning threshold to adapt to the changing working environment, operating personnel status, etc. in the limited space.

[0122] Optionally, the intelligent early warning system calculation model outputs a safety score S zs and actual safety score S acl The deviation is obtained by calculating the deviation value, as shown in formula (28):

[0123] e sc =|S zs -S acl | (28)

[0124] Among them, e sc Represents the deviation value.

[0125] Optionally, the intelligent warning system adjusts the warning threshold G according to the deviation value and the cumulative error TC , as shown in formula (29):

[0126] G new =G TC +v·∑e sc (29)

[0127] Among them, G new is the adjusted warning threshold, v is the adjustment coefficient, which determines the speed of threshold adjustment, v is a small positive number, ∑e sc Represents all deviation values ​​e sc The accumulated value.

[0128] From the above content, it can be seen that the intelligent early warning system first calculates the absolute difference between the actual safety score and the target score, quantifies the gap between the two, and then updates the early warning threshold based on the calculated absolute difference and the preset adjustment coefficient. By establishing a dynamic threshold adjustment mechanism, the early warning system can adapt to the constant changes in the working environment and staff status, improve the adaptability and accuracy of the early warning, more accurately reflect the current safety situation, reduce false alarms and missed alarms, and at the same time improve the adaptability and flexibility of the system.

[0129] In an optional embodiment, the intelligent early warning system determines a first value based on environmental data in a confined space and the weight of the environmental data, wherein the first value is used to characterize the risk intensity caused by environmental factors in the confined space, and then determines a second value based on the personnel status data and contribution factors of the operators in the confined space, wherein the contribution factor is used to characterize the probability of risk caused by physical factors of the operators, wherein the second value is used to characterize the risk intensity caused by the physical state of the operators, and then takes the absolute difference between the first value and the second value as the target difference, and then determines the risk type in the confined space based on the target early warning threshold, the first value, the second value and the target difference, wherein the risk type is used to distinguish the source category of the risk in the confined space, and then determines the risk level in the confined space based on the actual safety score and at least one preset risk threshold, wherein the risk level is used to characterize the degree of risk in the confined space, and finally generates early warning information based on the risk type and risk level.

[0130] Optionally, the intelligent early warning system establishes a contribution mechanism to determine the main risk sources by comparing the contribution values. First, the total contribution of environmental factors (the first value mentioned above) is calculated, as shown in formula (30):

[0131] E xd =γ 1 T+γ 2 H+γ 3 G+γ 4 ·OC (30)

[0132] Among them, γ 1 , γ 2 , γ3 and γ 4 are the weights corresponding to temperature T, humidity H, gas concentration G and oxygen content OC respectively.

[0133] Optionally, the intelligent early warning system calculates the total contribution of the operator's personal factors (the above-mentioned second value), as shown in formula (31):

[0134]

[0135] Among them, P heth Indicates health status, which is the average value of heart rate HR, blood pressure BP and blood oxygen saturation SP. P heth The weight of (the contribution factor mentioned above).

[0136] Optionally, the intelligent early warning system calculates the absolute difference between the total contribution of environmental factors and the total contribution of personal factors, and then determines the risk type within the confined space based on the adjusted early warning threshold, the absolute difference, the total contribution of environmental factors, and the total contribution of personal factors, determines the risk level within the confined space based on the actual safety score and the preset risk threshold, and finally generates early warning information based on the risk type and risk level.

[0137] From the above content, it can be seen that the intelligent early warning system can evaluate the risk intensity caused by environmental factors and physical factors of personnel by analyzing the environmental data and the status data of the operators in a confined space. By determining the first value and the second value, the system quantifies the risk intensity, making the risk assessment more accurate and operational. According to the risk type and risk level, the system can generate corresponding early warning information to provide decision support for safety management.

[0138] In an optional embodiment, the intelligent warning system determines that a first type of risk occurs in the confined space when the first value is greater than the second value and the target difference is greater than the target warning threshold, wherein the first type of risk is used to characterize the risks caused by environmental factors; determines that a second type of risk occurs in the confined space when the first value is less than the second value and the target difference is greater than the target warning threshold, wherein the second type of risk is used to characterize the risks caused by physical factors of the operators; and determines that a third type of risk occurs in the confined space when the target difference is less than or equal to the target warning threshold, wherein the third type of risk is used to characterize the risks caused by both environmental factors and physical factors of the operators.

[0139] Optionally, the first type of risk represents environmental factor risks, such as excessive temperature, excessive humidity, excessive gas concentration and low oxygen content; the second type of risk represents personal factor risks, such as abnormal heart rate, abnormal blood pressure and abnormal blood oxygen saturation; the third type of risk is comprehensive, which is a combination of environmental factor and personal factor risks.

[0140] Optionally, if E xd > R xd , and E xd - R xd > G new , it is defined as a type I risk; if R xd > E xd , and R xd - E xd > G new , it is defined as a type II risk; if |E xd - R xd | ≤ G new , it is defined as a type III risk.

[0141] Optionally, the intelligent early warning system locates the risk levels as four risk levels: minor, medium, severe, and extreme, and sets the thresholds for the four risk levels of minor, medium, severe, and extreme as FB1, FB2, FB3, and FB4. Then, through threshold comparison, a risk instruction is obtained. If S acl < FB1, it is determined as a minor risk and a minor risk instruction is issued; if FB1 ≤ S acl < FB2, it is determined as a medium risk and a medium risk instruction is issued; if FB2 ≤ S acl < FB3, it is determined as a severe risk and a severe risk instruction is issued; if FB3 ≤ S acl < FB4, it is determined as an extreme risk and an extreme risk instruction is issued.

[0142] Optionally, in the case of a type I risk: at the minor level, the warning light emits purple light to remind the staff to pay attention to environmental changes; at the medium level, the ventilation system is automatically adjusted and the alarm sounds; at the severe level, the emergency lighting is activated, the ventilation system is automatically adjusted to the maximum, and the alarm issues an evacuation message; at the extreme level, all non-essential power supplies are automatically shut off, the emergency evacuation and rescue procedures are initiated.

[0143] Optionally, in the case of a type II risk: at the minor level, the warning light emits blue light to remind the staff to pay attention to their personal health status; at the medium level, the alarm issues a rest message for the staff to take a rest; at the severe level, an emergency medical examination is arranged and the alarm issues an evacuation message; at the extreme level, first aid is immediately provided and the alarm issues an emergency evacuation message.

[0144] Optionally, in the case of a type III risk: at the minor level, the warning light emits red light to remind the staff to pay attention to the environment and their personal health status; at the medium level, the ventilation system is automatically adjusted and the alarm issues a rest message for the staff to take a rest; at the severe level, the emergency lighting is activated and the alarm issues an evacuation message. At the extreme level, all non-essential power supplies are automatically shut off, first aid is immediately provided, and the alarm issues an emergency evacuation message.

[0145] Optionally, if multiple risks are detected at the same time, the intelligent early warning system performs early warning processing in order of priority, with the order of risk types from high to low being Category III risk, Category II risk, Category I risk, and the order of risk levels from high to low being extreme risk level, severe risk level, medium risk level, and slight risk level.

[0146] From the above content, it can be seen that when the intelligent early warning system detects potential risks, it not only issues an early warning, but also automatically triggers corresponding emergency response measures according to the risk level and type, achieving seamless connection from early warning to emergency response, and effectively reducing the risk of accidents.

[0147] According to another aspect of the embodiments of the present application, an embodiment of a safety warning device for a confined space is also provided. Figure 2 is a schematic diagram of an optional safety warning device for a limited space according to an embodiment of the present application, such as Figure 2 As shown, the safety warning device for a confined space includes: a determination unit 201 , an input unit 202 and a generation unit 203 .

[0148] Optionally, the input unit 202 includes: a first acquisition subunit, a second processing subunit, a first determination subunit, and a second determination subunit. The first acquisition subunit is used to acquire historical data in a limited space, wherein the historical data at least includes historical environmental data in the limited space and personnel status data of historical operators; the second processing subunit is used to extract key features from the historical data, and fuse the key features according to the weight of each key feature to obtain feature vectors at different times, wherein the key features at least include historical environmental parameters in the limited space and physiological characteristics of historical operators, wherein the historical environmental parameters at least include temperature, humidity, gas concentration, and oxygen content, and the physiological characteristics of historical operators at least include heart rate, blood pressure, and blood oxygen saturation; the first determination subunit is used to obtain a multimodal data set according to the feature vector and the risk score value of each feature vector; the second determination subunit is used to train a safety status neural network based on the multimodal data set.

[0149] Optionally, the second determination subunit includes: a first acquisition module, a first determination module, a second determination module, and a first training module. The first acquisition module is used to acquire the initial learning rate, the number of input layer neurons, and the number of output layer neurons of the initial neural network, wherein the initial learning rate is used to characterize the update frequency of the initial settings of the weights and bias vectors in the initial neural network; the first determination module is used to determine the target learning rate of the initial neural network based on the initial learning rate and the attenuation factor, wherein the target learning rate is used to improve the convergence speed of the initial neural network, and the attenuation factor is used to characterize the attenuation rate of the initial learning rate; the second determination module is used to determine the target number of hidden layers of the initial neural network and the number of neurons in each hidden layer based on the number of neurons in the input layer and the number of neurons in the output layer, wherein the hidden layer is located between the input layer and the output layer of the initial neural network, and the target number and the number of neurons are used to improve the performance of the initial neural network; the first training module is used to perform multiple iterative training on the initial neural network based on the multimodal data set to obtain a safe state neural network through training.

[0150] Optionally, the second determination subunit includes: a first processing module, a second processing module, a third determination module and a first calculation module. The first processing module is used to divide the multimodal data set into N subsets, where N is an integer greater than 1; the second processing module is used to perform K rounds of test operations on the safe state neural network to obtain K prediction results, wherein each round of test operation is used to select any one of the N subsets as the target test set, and use the subset other than the target test set as the target training set, use the target training set to perform a round of testing on the safe state neural network, and use the tested safe state neural network to predict the target test set to obtain a prediction result; the third determination module is used to determine K accuracy rates of the safe state neural network based on the K prediction results, wherein the Jth accuracy rate among the K accuracy rates is used to characterize the prediction accuracy of the safe state neural network under the Jth test operation in the K rounds of test operations; the first calculation module is used to calculate the average of the K accuracy rates as the prediction accuracy of the safe state neural network.

[0151] Optionally, the generating unit 203 includes: a first calculating subunit, a third processing subunit, and a first generating subunit. The first calculating subunit is used to calculate the absolute difference between the actual safety score and the target score; the third processing subunit is used to update the warning threshold according to the absolute difference and the adjustment coefficient to obtain the target warning threshold; the first generating subunit is used to generate warning information according to the target warning threshold and the actual safety score.

[0152] Optionally, the first generation subunit includes: a fourth determination module, a fifth determination module, a sixth determination module, a seventh determination module, an eighth determination module and a first generation module. The fourth determination module is used to determine a first value according to environmental data in a limited space and a weight of the environmental data, wherein the first value is used to characterize the risk intensity caused by environmental factors in the limited space; the fifth determination module is used to determine a second value according to the personnel status data and contribution factors of the operators in the limited space, wherein the contribution factor is used to characterize the probability of risk caused by the physical factors of the operators, wherein the second value is used to characterize the risk intensity caused by the physical state of the operators; the sixth determination module is used to use the absolute difference between the first value and the second value as the target difference; the seventh determination module is used to determine the risk type in the limited space according to the target warning threshold, the first value, the second value and the target difference, wherein the risk type is used to distinguish the source category of the risk in the limited space; the eighth determination module is used to determine the risk level in the limited space according to the actual safety score and at least one preset risk threshold, wherein the risk level is used to characterize the risk degree in the limited space; the first generation module is used to generate warning information according to the risk type and risk level.

[0153] Optionally, the seventh determination module includes: a first determination submodule, a second determination submodule, and a third determination submodule. The first determination submodule is used to determine that a first type of risk occurs in the confined space when the first value is greater than the second value and the target difference is greater than the target warning threshold, wherein the first type of risk is used to characterize the risk caused by environmental factors; the second determination submodule is used to determine that a second type of risk occurs in the confined space when the first value is less than the second value and the target difference is greater than the target warning threshold, wherein the second type of risk is used to characterize the risk caused by the physical factors of the operator; the third determination submodule is used to determine that a third type of risk occurs in the confined space when the target difference is less than or equal to the target warning threshold, wherein the third type of risk is used to characterize the risk caused by both the environmental factors and the physical factors of the operator.

[0154] According to another aspect of the present application, an electronic device is also provided, wherein the electronic device includes one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the above-mentioned limited space safety warning method.

[0155] According to another aspect of the present application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer-readable storage medium, wherein when the computer program is executed, the device where the computer-readable storage medium is located executes the above-mentioned limited space security warning method.

[0156] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0157] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0158] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units can be a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0159] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0160] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0161] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or optical disk and other media that can store program codes.

[0162] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A safety early warning method for a confined space, characterized in that: include: Determining an actual safety score in the limited space according to target data in the limited space, wherein the target data at least includes environmental data, equipment data, personnel status data of operators, and operation data in the limited space; Inputting the actual safety score into a safety state neural network to obtain a target score, wherein the target score is used to characterize the overall safety risk level in the confined space; The warning information is generated according to a warning threshold, the actual safety score and the target score, wherein the warning threshold is periodically adjusted according to the actual safety score and the target score.

2. The safety early warning method for a confined space according to claim 1, characterized in that: The safe state neural network is trained by the following steps, including: Acquiring historical data within the limited space, wherein the historical data at least includes historical environmental data and historical status data of operators within the limited space; Extract key features from the historical data, and fuse the key features according to the weight of each key feature to obtain feature vectors at different times, wherein the key features at least include historical environmental parameters in the limited space and physiological characteristics of the historical operators, wherein the historical environmental parameters at least include temperature, humidity, gas concentration, and oxygen content, and the physiological characteristics of the historical operators at least include heart rate, blood pressure, and blood oxygen saturation; Obtaining a multimodal data set according to the feature vectors and the risk score value of each feature vector; The safe state neural network is obtained by training based on the multimodal data set.

3. The safety early warning method for a confined space according to claim 2, characterized in that: The safe state neural network is obtained by training based on the multimodal data set, including: Obtaining an initial learning rate, a number of input layer neurons, and a number of output layer neurons of the initial neural network, wherein the initial learning rate is used to characterize an update frequency of initial settings of weights and bias vectors in the initial neural network; Determining a target learning rate of the initial neural network according to the initial learning rate and the decay factor, wherein the target learning rate is used to improve the convergence speed of the initial neural network, and the decay factor is used to characterize the decay rate of the initial learning rate; Determining a target number of hidden layers of the initial neural network and a number of neurons in each hidden layer according to the number of neurons in the input layer and the number of neurons in the output layer, wherein the hidden layer is located between the input layer and the output layer of the initial neural network, and the target number and the number of neurons are used to improve the performance of the initial neural network; The initial neural network is iteratively trained multiple times according to the multimodal data set to obtain the safe state neural network.

4. The safety early warning method for a confined space according to claim 2, characterized in that: After the safety state neural network is obtained through training based on the multimodal data set, the safety early warning method for a limited space further includes: Dividing the multimodal dataset into N subsets, where N is an integer greater than 1; Performing K rounds of testing operations on the safe state neural network to obtain K prediction results, wherein each round of testing operation is used to select any one of the N subsets as a target test set, and use a subset other than the target test set as a target training set, perform a round of testing on the safe state neural network using the target training set, and use the tested safe state neural network to predict the target test set to obtain a prediction result; Determining K accuracy rates of the safety state neural network according to the K prediction results, wherein the Jth accuracy rate among the K accuracy rates is used to characterize the prediction accuracy of the safety state neural network under the Jth round of test operation in the K rounds of test operations; The average value of the K accuracy rates is calculated as the prediction accuracy rate of the safety state neural network.

5. The safety early warning method for a confined space according to claim 1, characterized in that: Generating warning information according to the warning threshold, the actual safety score and the target score includes: Calculating an absolute difference between the actual safety score and the target score; The warning threshold is updated according to the absolute difference and the adjustment coefficient to obtain a target warning threshold; Warning information is generated according to the target warning threshold and the actual safety score.

6. The safety early warning method for a confined space according to claim 5, characterized in that: Generating warning information according to the target warning threshold and the actual safety score includes: Determining a first value according to environmental data in the confined space and a weight of the environmental data, wherein the first value is used to characterize the risk intensity caused by environmental factors in the confined space; Determining a second value according to the personnel status data and a contribution factor of the operator in the limited space, wherein the contribution factor is used to characterize the probability of risk caused by the physical factors of the operator, and wherein the second value is used to characterize the intensity of risk caused by the physical status of the operator; taking the absolute difference between the first value and the second value as a target difference; Determine the risk type in the confined space according to the target warning threshold, the first value, the second value, and the target difference, wherein the risk type is used to distinguish the source category of the confined space risk; Determining a risk level within the confined space according to the actual safety score and at least one preset risk threshold, wherein the risk level is used to characterize the degree of risk within the confined space; Generate warning information according to the risk type and the risk level.

7. The safety early warning method for a confined space according to claim 6, characterized in that: Determining the risk type in the limited space according to the target warning threshold, the first value, the second value, and the target difference includes: When the first value is greater than the second value and the target difference is greater than the target warning threshold, it is determined that a first type of risk occurs in the limited space, wherein the first type of risk is used to characterize the risk caused by environmental factors; When the first value is less than the second value and the target difference is greater than the target warning threshold, it is determined that a second type of risk occurs in the confined space, wherein the second type of risk is used to characterize the risk caused by the physical factors of the operator; When the target difference is less than or equal to the target warning threshold, it is determined that a third type of risk occurs in the confined space, wherein the third type of risk is used to characterize the risk caused by the environmental factors and the physical factors of the operator.

8. A safety warning device for a limited space, characterized in that: include: A determination unit, which determines an actual safety score in the limited space according to target data in the limited space, wherein the target data at least includes environmental data, equipment data, personnel status data of operators, and operation data in the limited space; An input unit, inputting the actual safety score into a safety state neural network to obtain a target score, wherein the target score is used to characterize the overall safety risk level in the confined space; A generating unit generates warning information according to a warning threshold, the actual safety score, and the target score, wherein the warning threshold is periodically adjusted according to the actual safety score and the target score.

9. An electronic device, characterized in that: It includes one or more processors and a memory, wherein the memory is used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors execute the confined space safety warning method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located executes the confined space safety warning method according to any one of claims 1 to 7.

Images