Disaster relief and early warning method and system based on artificial intelligence

Abstract

The application discloses a disaster reduction and rescue early warning method and system based on artificial intelligence, and the method comprises data acquisition, disaster diffusion prediction, disaster diffusion range quantification, prediction result evaluation, parameter optimization and disaster reduction and rescue early warning. The application belongs to the field of rescue early warning, and particularly relates to a disaster reduction and rescue early warning method and system based on artificial intelligence. In the scheme, boundary continuous processing is adopted to map angle error to unit circular arc length, and a disaster output normalization layer is used to avoid the rescue team advancing towards the disaster source; based on a dynamic error sensitive weighting mechanism, a dynamic bandwidth error distribution estimation is introduced to ensure that the evacuation radius contains a complete risk area; the decision effectiveness is improved based on the risk adaptability of the prediction result evaluation; the stage regulation and control of the disturbance amplitude is introduced to accelerate the optimization convergence speed; the diversity of the current solution is reserved based on the dynamic balance of the core control weight; the parameter combination group diversity is maintained based on the replacement of the inefficient solution; and then the rescue early warning effect is improved.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of rescue early warning, in particular to a disaster reduction and rescue early warning method and system based on artificial intelligence. BACKGROUND

[0002] The disaster reduction and rescue early warning method is a technical means for collecting and analyzing disaster-related data and predicting the disaster diffusion trend by means of a model. The core is to generate early warning information to provide a basis for decision-making such as evacuation planning and rescue deployment to cope with disasters in advance and reduce losses. However, the general disaster reduction and rescue early warning method has the problems of boundary fracture in direction prediction, equalization defects in error processing, static nature of uncertainty quantification, and risk mismatch of evaluation indexes, which leads to poor accuracy of rescue early warning. The general disaster reduction and rescue early warning method has the problems of static iteration process of parameter optimization and degeneration of group diversity, which leads to poor rescue early warning effect. SUMMARY

[0003] In view of the above problems, the application provides a disaster reduction and rescue early warning method and system based on artificial intelligence to overcome the defects of the prior art. The application solves the problems of boundary fracture in direction prediction, equalization defects in error processing, static nature of uncertainty quantification, and risk mismatch of evaluation indexes in the general disaster reduction and rescue early warning method, which leads to poor accuracy of rescue early warning. The application maps the angle error to the unit circular arc length by boundary continuous processing, cooperates with the disaster output normalization layer, and forcibly constrains the results to solve the boundary fracture and avoid the rescue team advancing towards the disaster source. The application introduces error classification weight based on the dynamic error sensitivity weighting mechanism to realize the nonlinear incremental of large error weight. The application introduces the dynamic bandwidth error distribution estimation combined with the cross-border range correction to ensure that the evacuation radius contains the complete risk area. The application improves the decision effectiveness based on the risk adaptability of the prediction result evaluation, and further improves the accuracy of rescue early warning. The application introduces the phased regulation and control of disturbance amplitude to speed up the optimization convergence speed. The application focuses on retaining the diversity of the current solution based on the dynamic balance of the core control weight. The application ensures the maintenance of the diversity of the parameter combination group based on the replacement of inefficient solutions, and further improves the effect of rescue early warning.

[0004] The technical scheme adopted by the application is as follows: The application provides a disaster reduction and rescue early warning method based on artificial intelligence, which comprises the following steps:

[0005] Step S1: data collection;

[0006] Step S2: disaster diffusion prediction;

[0007] Step S3: disaster diffusion range quantification;

[0008] Step S4: prediction result evaluation;

[0009] Step S5: parameter optimization;

[0010] Step S6: disaster relief early warning.

[0011] Further, in step S1, the data collection is to obtain historical disaster spread data; the label is labeled as the disaster spread direction; and preprocessing is performed; and a disaster spread dataset is obtained.

[0012] Further, in step S2, the disaster spread prediction is to establish a disaster spread prediction model based on the disaster spread dataset and the bidirectional LSTM, predict the disaster spread direction, the model structure includes: an input layer, a bidirectional LSTM layer, a full connection layer and a disaster output normalization layer; a disaster point loss function is constructed, a dynamic error sensitive weighting mechanism is introduced, a correction error of a single sample is defined; an error grading weight is designed; a final loss function is obtained; and a disaster output normalization layer is constructed.

[0013] Further, in step S3, the disaster spread range quantification is to perform error distribution estimation based on the point prediction error, fit the error distribution, and introduce a dynamic bandwidth of dynamic time scale; and perform cross-border range processing on the predicted range obtained by combining the predicted disaster spread direction and the prediction error.

[0014] Further, in step S4, the prediction result evaluation is to divide the disaster spread dataset into a test set and a training set, train the disaster spread prediction model based on the training set, and verify the performance of the disaster spread prediction model based on the test set; construct a disaster spread prediction evaluation index; construct a range prediction evaluation index, including a prediction range coverage rate and a prediction range normalized average width, and introduce a coverage rate calibration term; when the disaster spread prediction model converges on the training set loss, the disaster spread prediction model training is completed, a verification threshold is set, the model performance is verified based on the verification threshold, if it meets the standard, the disaster spread prediction model is established, otherwise parameter optimization is performed.

[0015] Further, in step S5, the parameter optimization is to optimize the parameters of the disaster spread prediction model, minimize the loss function of the verification set, and specifically includes the following steps:

[0016] Step S51: initial solution generation; generate an initial solution with a parameter combination mapping; each solution represents a set of hyperparameter combinations;

[0017] Step S52: fitness function design; for each solution, train the bidirectional LSTM and calculate the on the verification set as the fitness value;

[0018] Step S53: perturbation amplitude adjustment; dynamically adjust the perturbation amplitude according to the iteration process; and perform mutation operation;

[0019] Step S54: inefficient replacement; if the fitness value increment of K consecutive iterations is less than the threshold, it is marked as inefficient; and the generated initial solution is used to replace the inefficient solution;

[0020] Step S55: solution update;

[0021] Step S56: solution evaluation; set the maximum number of iterations, if the maximum number of iterations is reached or the fitness value of the optimal solution converges, the performance of the disaster spread prediction model trained based on the optimal solution is evaluated, if the standard is met, the disaster spread prediction model is established; otherwise, return to step S51.

[0022] Further, in step S6, the disaster mitigation and rescue warning is real-time collection of disaster spread data, which is input into the disaster spread prediction model after preprocessing, and the obtained disaster spread direction and error distribution are used to warn the management personnel. The disaster mitigation and rescue warning system based on artificial intelligence provided by the application comprises a data acquisition module, a disaster spread prediction module, a disaster spread range quantification module, a prediction result evaluation module, a parameter optimization module and a disaster mitigation and rescue warning module.

[0024] The data acquisition module acquires historical disaster spread data to obtain a disaster spread data set.

[0025] The disaster spread prediction module is based on the disaster spread data set and bidirectional LSTM, and a loss function comprising a dynamic error sensitive weighting mechanism is constructed to design a disaster spread prediction model.

[0026] The disaster spread range quantification module introduces a dynamic bandwidth fitting error distribution of dynamic time scale to quantify the disaster spread prediction range and perform cross-border range processing.

[0027] The prediction result evaluation module evaluates the prediction performance by the index of boundary correction average error to determine whether the disaster spread prediction model meets the standard.

[0028] The parameter optimization module optimizes the parameters of the disaster spread prediction model by dynamically adjusting the disturbance amplitude and the core control weight.

[0029] The disaster mitigation and rescue warning module is based on the disaster spread prediction model to perform rescue warning on the real-time collected disaster spread data.

[0030] The application has the following beneficial effects by adopting the above scheme:

[0031] (1) In view of the problems of boundary breakage of direction prediction, equalization defects of error processing, static nature of uncertainty quantification, and risk mismatch of evaluation index in general disaster mitigation and rescue early warning methods, which further leads to poor accuracy of rescue early warning, the scheme solves the boundary breakage by boundary continuous processing and maps the angle error to the unit circular arc length, cooperates with the disaster output normalization layer to forcibly constrain the results, avoids the rescue team advancing towards the disaster source direction; based on the dynamic error sensitive weighting mechanism, the error classification weight is introduced to realize the nonlinear incremental of large error weight; the dynamic bandwidth error distribution estimation is introduced, combined with the cross-border range correction to ensure that the evacuation radius contains the complete risk area; the risk adaptability based on the prediction result evaluation improves the decision effectiveness; and further improves the accuracy of rescue early warning.

[0032] (2) In view of the problems of static iteration process of parameter optimization, degeneration of group diversity, which further leads to poor rescue early warning effect, the scheme introduces the stage regulation of disturbance amplitude to speed up the optimization convergence speed; based on the dynamic balance of core control weight, the diversity of current solution is reserved; based on the inefficient solution replacement to ensure the maintenance of parameter combination group diversity; and further improve the rescue early warning effect. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 A flowchart of the disaster mitigation and rescue early warning method based on artificial intelligence provided by the present application is provided.

[0034] Figure 2 A schematic diagram of the disaster mitigation and rescue early warning system based on artificial intelligence provided by the present application.

[0035] The accompanying drawings are used to provide a further understanding of the present application, and constitute a part of the specification, together with the embodiments of the present application, to explain the present application, and do not constitute a limitation of the present application. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0037] In the description of the present application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the systems or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0038] Embodiment one, refer to Figure 1 The present application provides a disaster mitigation and rescue early warning method based on artificial intelligence, which comprises the following steps:

[0039] Step S1: data acquisition; obtain historical disaster diffusion data to obtain a disaster diffusion data set;

[0040] Step S2: disaster diffusion prediction; based on the disaster diffusion data set and the bidirectional LSTM, a loss function containing a dynamic error sensitive weighting mechanism is constructed to design a disaster diffusion prediction model;

[0041] Step S3: disaster diffusion range quantification; the dynamic bandwidth fitting error distribution of the dynamic time scale is introduced to quantify the disaster diffusion prediction range and perform cross-border range processing;

[0042] Step S4: prediction result evaluation; the prediction performance is evaluated by the index of boundary correction average error to determine whether the disaster diffusion prediction model meets the standard;

[0043] Step S5: parameter optimization; the disaster diffusion prediction model is parameter optimized by dynamically adjusting the disturbance amplitude and the core control weight;

[0044] Step S6: disaster mitigation and rescue early warning; based on the disaster diffusion prediction model, the disaster diffusion data collected in real time is rescued and early warned.

[0045] Embodiment two, refer to Figure 1 This embodiment is based on the above embodiment, in step S1, the data acquisition is to obtain historical disaster diffusion data; the historical disaster diffusion data includes wind direction, wind speed, air humidity, air temperature, air pressure, disaster type, disaster intensity, precipitation, and diffusion source position; the label is disaster diffusion direction, 0-360 degrees; and preprocessing is performed, including interpolation, filtering and normalization processing; to obtain a disaster diffusion data set.

[0046] Embodiment three, refer to Figure 1, the embodiment based on the above embodiment, in step S2, the disaster spread prediction is based on the disaster spread dataset and the bidirectional LSTM to establish a disaster spread prediction model, to predict the disaster spread direction in the future 1-3 days, and to provide a time reservation for evacuation route planning and rescue force deployment in disaster relief, and the boundary discontinuity problem needs to be solved to avoid misjudgment of the spread direction; the disaster spread prediction model is constructed, and the time sequence characteristics of the disaster spread data are extracted; the model structure includes: an input layer; a bidirectional LSTM layer, forward LSTM+backward LSTM, capturing past and future dependency; a fully connected layer, gradually compressing the features to 1-dimensional output; a disaster output normalization layer, outputting the predicted disaster spread direction; a disaster point loss function is constructed , a dynamic error sensitivity weighted mechanism is introduced, the loss weight is dynamically adjusted according to the size of the prediction error, the large error that may lead to serious consequences is corrected in priority, and the small error within the acceptable range is secondary concerned, the correction error l of a single sample is defined and represented as: ; the loss is calculated by the unit circular arc length distance to avoid the prediction deviation caused by the boundary discontinuity; the error grading weight is designed and represented as: ; ; E is the prediction error after boundary correction; is the sensitivity coefficient; is the enhancement coefficient, which ensures that the weight increase of large error is faster; through the weight grading, the loss function is matched with the risk sensitivity of rescue decision, and the high-risk scene corresponding to large error is optimized in priority; the final loss function is represented as: ; wherein, is the predicted disaster spread direction; is the actual disaster spread direction; H is the total number of samples; v is the sample index; the disaster output normalization layer is constructed and represented as: ; wherein, is the normalized predicted disaster spread direction; is the modulo operation; the output is forced to be constrained in 0-360°, to ensure that the prediction result conforms to the physical meaning of the actual disaster spread direction;

[0047] High-precision point prediction provides the core basis for rescue decision, and boundary continuity processing avoids fatal misjudgment.

[0048] Embodiment four, refer to Figure 1 , the embodiment based on the above embodiment, in step S3, the disaster spread range quantification is to quantify the prediction uncertainty, to give the possible range of the spread direction, to provide a fault tolerance space for rescue decision, and to avoid rescue failure caused by deviation of single prediction value; the specific operation is: error distribution estimation is carried out based on point prediction error, error distribution is fitted, and a dynamic bandwidth of dynamic time scale is introduced, so that the error distribution estimation adapts to different time scales, and the error distribution is represented as: ; ; where G is the prediction error; ; is the probability density function estimate of the prediction error, describing the likelihood of the error taking a certain value, if the prediction error is 5° with the highest value, then the error is 5°; N is the total number of error samples, n is the error sample index; h is the bandwidth, used to control the smoothness of the error distribution; is the base bandwidth, is the time scale coefficient, d is the prediction day; is the prediction error of the nth error sample; the cross-border range processing of the prediction range obtained by combining the prediction error of the predicted disaster diffusion direction is represented as: ; where Inl is the processed cross-border range; and are the original lower and upper bounds of the range, respectively; is used to describe the start and end relationship of the angle range; the confidence range is determined by the percentile method, if the 5% and 95% quantile values of the error distribution are taken as the upper and lower boundaries at the 90% confidence level, 1 day is selected as the time scale, and the directions corresponding to the 5% and 95% quantile values of the error distribution are and , respectively, then the warning result that the disaster diffusion direction in the future 1 day has a 90% probability of falling within is obtained;

[0049] Quantifying uncertainty through range prediction avoids over-reliance on a single prediction value in decision-making, and in typhoon rescue, the possible fluctuation range of disaster diffusion direction is determined to expand the evacuation radius to leave sufficient safety redundancy; and the range is ensured to conform to the actual geographical direction through cross-border processing.

[0050] Embodiment five, see Figure 1 , this embodiment is based on the above embodiment, in step S4, the prediction result evaluation is to divide the disaster diffusion data set into a test set and a training set, train the disaster diffusion prediction model based on the training set, and verify the performance of the disaster diffusion prediction model based on the test set; a disaster diffusion prediction evaluation index is constructed, represented as: ; ; where M is the average error of boundary correction, measuring the overall deviation degree of the prediction after boundary correction; R is the mean square error of boundary correction; Q is the total number of test set samples, q is the test set sample index; a range prediction evaluation index is constructed, including a prediction range coverage and a prediction range normalized average width, and a coverage calibration term is introduced, respectively represented as: ; ; ; where PP is the prediction range coverage, is the base calibration value,​ is a calibration coefficient, CL is a preset signal level; is a coverage indicator variable of the qth sample, when the real direction falls in the predictor range after the boundary correction, , otherwise ; PW is the average width of the prediction range; is the total width of the prediction range of the qth sample; the smaller the PW, the more accurate the early warning; when the disaster spread prediction model converges on the training set loss, the disaster spread prediction model training is completed, a verification threshold is set, the model performance is verified based on the verification threshold, if it meets the standard, the disaster spread prediction model is established, otherwise parameter optimization is performed; M, R, PP and PW all have corresponding verification thresholds, for each sample in the test set, the point prediction error is calculated, and then M and R are obtained; for the range prediction result of each sample, it is judged whether the real value is covered by the range, PP is calculated, and the width of each range is calculated to obtain PW; if M is lower than the M verification threshold and R is lower than the R verification threshold, the spread prediction meets the standard; if PP is higher than the PP threshold and PW is lower than the PW threshold, the range prediction meets the standard; at the same time, the disaster spread prediction model is established when the spread prediction meets the standard and the range prediction meets the standard.

[0051] By performing the above operation, for the general disaster mitigation and rescue early warning method, the boundary fracture problem of direction prediction, the equalization defect of error processing, the static nature of uncertainty quantification, and the risk mismatch of evaluation index, and then the poor accuracy of rescue early warning, the present scheme solves the boundary fracture by boundary continuous processing, maps the angle error to the unit circular arc length, cooperates with the disaster output normalization layer, and forcibly constrains the result to solve the boundary fracture and avoid the rescue team advancing to the disaster source direction; based on the dynamic error sensitive weighting mechanism, the error classification weight is introduced to realize the nonlinear incremental of large error weight; the dynamic bandwidth error distribution estimation is introduced, combined with the cross-border range correction, to ensure that the evacuation radius contains the complete risk area; the risk adaptability based on the prediction result evaluation improves the decision effectiveness; and then the rescue early warning accuracy is improved.

[0052] Embodiment six, referring to Figure 1 This embodiment is based on the above embodiment, in step S5, the parameter optimization is to optimize the parameters of the disaster spread prediction model to minimize the disaster point loss function on the verification set, which specifically includes the following steps:

[0053] Step S51: initial solution generation; generate initial solution with parameter combination mapping to ensure uniform initial distribution and cover the hyperparameter space; each solution represents a set of hyperparameter combinations, including bidirectional LSTM hidden layer unit number, learning rate, time series window size, regularization coefficient, batch size, basic bandwidth, time scale coefficient, sensitivity coefficient, enhancement coefficient and calibration coefficient; the generated initial solution is represented as: ; ; where h is the position of the initial solution; and are the lower and upper bounds of the search space, respectively; and are the mapping values generated in the u-th and u+1-th time, respectively; is the control parameter;

[0054] Step S52: Fitness function design; for each solution, train the bidirectional LSTM and calculate the on the validation set as the fitness value; the smaller the value, the better the solution performance, i.e., the better the hyperparameter combination;

[0055] Step S53: Perturbation amplitude adjustment; dynamically adjust the perturbation amplitude according to the iteration process, strengthen global exploration in the early iteration, and strengthen local development in the later period; represented as: ; and perform mutation operation, represented as: ; where is a random perturbation vector, each dimension follows a uniform distribution [-1, 1] and is independent of each other; is the perturbation amplitude; and are the minimum and maximum intensities, respectively; t is the current search iteration number, and maxt is the maximum search iteration number; and are the positions of the i-th solution after mutation and before mutation in the t-th search iteration, respectively;

[0056] Step S54: Inefficient replacement; identify and replace inefficient solutions that have not been improved for a long time to avoid population diversity degradation and premature convergence; the inefficient judgment rule is: if the fitness value improvement of the solution in the continuous K iteration is less than the threshold, it is marked as inefficient; replace the inefficient with the generated initial solution;

[0057] Step S55: Solution update; design core control weight, which changes dynamically with iteration to adapt to different search stage requirements, represented as: ; Position update is represented as: ; where is the core control weight; and are the minimum and maximum weights, respectively; is the updated position of the j-th dimension of the solution; is the value of the j-th dimension of the optimal solution; is the value of the j-th dimension of the ; and

[0058] Step S56: de-evaluation; set the maximum number of iterations, if the maximum number of iterations is reached or the fitness value of the optimal solution converges, the performance of the disaster spread prediction model trained based on the optimal solution is evaluated, if the standard is met, the disaster spread prediction model is established; otherwise, return to step S51.

[0059] By performing the above operation, for the general disaster reduction and rescue early warning method, the iteration process of parameter optimization is static, the group diversity degenerates, and then the rescue early warning effect is poor. The scheme introduces the stage regulation and control of the disturbance amplitude, accelerates the optimization convergence speed; the dynamic balance based on the core control weight focuses on retaining the diversity of the current solution; the low-efficiency solution replacement ensures the maintenance of the group diversity of the parameter combination; and then the rescue early warning effect is improved.

[0060] Embodiment seven, refer to Figure 1 This embodiment is based on the above-mentioned embodiment, in step S6, the disaster reduction and rescue early warning is real-time collection of disaster spread data, after pretreatment, input into the disaster spread prediction model, the obtained disaster spread direction and error distribution are used for early warning to the management personnel.

[0061] Embodiment eight, refer to Figure 2 This embodiment is based on the above-mentioned embodiment, the disaster reduction and rescue early warning system based on artificial intelligence provided by the application includes a data acquisition module, a disaster spread prediction module, a disaster spread range quantification module, a prediction result evaluation module, a parameter optimization module and a disaster reduction and rescue early warning module.

[0062] The data acquisition module obtains historical disaster spread data to obtain a disaster spread data set;

[0063] The disaster spread prediction module is based on the disaster spread data set and the bidirectional LSTM, and a loss function containing a dynamic error sensitive weighting mechanism is constructed to design a disaster spread prediction model;

[0064] The disaster spread range quantification module introduces a dynamic bandwidth fitting error distribution of dynamic time scale to quantify the disaster spread prediction range and perform cross-border range processing;

[0065] The prediction result evaluation module evaluates the prediction performance by the index of boundary correction average error to judge whether the disaster spread prediction model meets the standard;

[0066] The parameter optimization module adjusts the disturbance amplitude and the core control weight dynamically to optimize the parameters of the disaster spread prediction model;

[0067] The disaster reduction and rescue early warning module is based on the disaster spread prediction model to perform rescue early warning on the real-time collected disaster spread data.

[0068] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0069] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0070] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A disaster relief and early warning method based on artificial intelligence, characterized in that: The method comprises the following steps: Step S1: data collection; obtain historical disaster diffusion data to obtain a disaster diffusion data set; Step S2: disaster diffusion prediction; based on the disaster diffusion data set and the bidirectional LSTM, a loss function comprising a dynamic error-sensitive weighting mechanism is constructed to design a disaster diffusion prediction model; Step S3: disaster diffusion range quantification; the dynamic bandwidth fitting error distribution of the dynamic time scale is introduced to quantify the disaster diffusion prediction range and perform cross-border range processing; Step S4: prediction result evaluation; the prediction performance is evaluated by the index of boundary correction average error to determine whether the disaster diffusion prediction model meets the standard; Step S5: parameter optimization; the disaster diffusion prediction model is parameter optimized by dynamically adjusting the disturbance amplitude and the core control weight; Step S6: disaster relief and early warning; based on the disaster diffusion prediction model, the real-time collected disaster diffusion data is subjected to rescue early warning; In step S2, the disaster spread prediction is based on a disaster spread dataset and a bidirectional LSTM to establish a disaster spread prediction model, and the model structure includes: an input layer; a bidirectional LSTM layer; a full connection layer; a disaster output normalization layer, which outputs a predicted disaster spread direction; and a disaster point loss function , a dynamic error sensitivity type weighting mechanism is introduced, the loss weight is dynamically adjusted according to the size of the prediction error, the correction error l of a single sample is defined, and is expressed as: ; the loss is calculated by the unit circular arc length distance; the error grading weight is designed , which is expressed as: ; ; E is the prediction error corrected by the boundary; is the sensitivity coefficient; is the enhancement coefficient; the final loss function is expressed as: ; wherein, is the predicted disaster spread direction; is the actual disaster spread direction; H is the total number of samples; v is the sample index; a disaster output normalization layer is constructed, which is expressed as: ; wherein, is the normalized predicted disaster spread direction; is the modulo operation; In step S3, the disaster spread range quantification is based on point prediction error to perform error distribution estimation fitting error distribution, and a dynamic bandwidth of dynamic time scale is introduced, and the error distribution is expressed as: ; ; wherein G is the prediction error; ; is the probability density function estimation value of the prediction error; N is the total amount of error samples, n is the error sample index; h is the bandwidth; is the basic bandwidth, is the time scale coefficient, and d is the prediction day; is the prediction error of the nth error sample; the prediction range obtained by combining the prediction error with the prediction disaster spread direction is processed across the border range, and the cross-border correction is expressed as: ; wherein Inl is the processed cross-border range; and are the original lower and upper limits of the range respectively; is used to describe the start and end relationship of the angle range. 2.The AI-based disaster mitigation and rescue early warning method according to claim 1, characterized in that: In step S4, the prediction result evaluation is to divide the disaster diffusion data set into a test set and a training set, train the disaster diffusion prediction model based on the training set, and verify the performance of the disaster diffusion prediction model based on the test set; a disaster diffusion prediction evaluation index is constructed; a range prediction evaluation index is constructed, including a prediction range coverage rate and a prediction range normalized average width, and a coverage rate calibration term is introduced; when the disaster diffusion prediction model converges to the training set loss, the disaster diffusion prediction model training is completed, a verification threshold is set, the model performance is verified based on the verification threshold, if the standard is met, the disaster diffusion prediction model is established, otherwise parameter optimization is performed. 3.The AI-based disaster mitigation and rescue early warning method according to claim 2, characterized in that: In step S5, the parameter optimization is to optimize the parameters of the disaster diffusion prediction model to minimize the loss function of the verification set, which specifically comprises the following steps: Step S51: initial solution generation; an initial solution is generated by parameter combination mapping; each solution represents a combination of hyperparameters; Step S52: Fitness function design; for each solution, train the bidirectional LSTM and compute the validation set accuracy on the validation set as the fitness value. , as the fitness value; Step S53: disturbance amplitude adjustment; the disturbance amplitude is dynamically adjusted according to the iteration process; and mutation operation is performed; Step S54: inefficient replacement; if the fitness value improvement of the solution is less than the threshold for K consecutive iterations, it is marked as inefficient; the initial solution is replaced with the generated initial solution; Step S55: solution update; Step S56: solution evaluation; a maximum number of iterations is set, if the maximum number of iterations is reached or the fitness value of the optimal solution converges, the performance of the disaster diffusion prediction model trained based on the optimal solution is evaluated, if the standard is met, the disaster diffusion prediction model is established; otherwise, return to step S51. 4.The AI-based disaster mitigation and rescue early warning method according to claim 3, characterized in that: In step S5, the solution update is to design a core control weight and perform position update. 5.The artificial intelligence-based disaster mitigation and rescue early warning method according to claim 4, characterized in that: In step S1, the data collection is to obtain historical disaster diffusion data; the label is marked as the disaster diffusion direction; and pre-processing is performed; a disaster diffusion data set is obtained. 6.The disaster relief and rescue early warning method based on artificial intelligence according to claim 5, characterized in that: In step S6, the disaster relief and early warning is to collect real-time disaster diffusion data, which is input into the disaster diffusion prediction model after pre-processing, and the obtained disaster diffusion direction and error distribution are used to warn the management personnel.

7. The disaster mitigation and rescue early warning system based on artificial intelligence, used to realize the disaster mitigation and rescue early warning method based on artificial intelligence as claimed in any one of claims 1-6, characterized in that: It comprises a data collection module, a disaster diffusion prediction module, a disaster diffusion range quantification module, a prediction result evaluation module, a parameter optimization module and a disaster relief and early warning module; The data acquisition module acquires historical disaster diffusion data to obtain a disaster diffusion data set; The disaster diffusion prediction module is based on the disaster diffusion data set and the bidirectional LSTM, constructs a loss function containing a dynamic error-sensitive weighting mechanism, and designs a disaster diffusion prediction model; The disaster diffusion range quantification module introduces a dynamic bandwidth fitting error distribution of a dynamic time scale, quantifies the disaster diffusion prediction range, and performs cross-border range processing; The prediction result evaluation module evaluates the prediction performance through the index of boundary correction average error, and determines whether the disaster diffusion prediction model meets the standard; The parameter optimization module optimizes the parameters of the disaster diffusion prediction model by dynamically adjusting the disturbance amplitude and the core control weight; The disaster mitigation and rescue early warning module is based on the disaster diffusion prediction model and performs rescue early warning on the real-time collected disaster diffusion data.

Images