CN112215458A - Disaster analysis method and electronic device - Google Patents

Abstract

The invention discloses a disaster analysis method and electronic equipment, wherein after an induced event possibly causing a target disaster occurs, parameter values of the induced event of the target disaster are acquired, a pre-trained Bayesian network model is adopted to analyze the parameter values of the induced event, and the probability of the target disaster and secondary derivative disaster possibly caused by the induced event is determined according to the analysis result of the parameter values of the induced event. Since the pre-trained bayesian network model in the embodiment of the present disclosure is constructed based on the secondary derivative disaster event graph, by analyzing and processing the accident data through the model, the direct disaster (i.e., target disaster) and the secondary derivative disaster related influence possibly caused by the induction time can be obtained, so as to prevent the disaster in advance, thereby improving the capability of preventing the disaster risk in the handling of the urban dangerous disaster.

Description

Disaster analysis method and electronic device

Technical Field

The present disclosure relates to the field of information processing technologies, and in particular, to a disaster analysis method and an electronic device.

Background

In recent years, after dangerous chemical substance disaster accidents occur along with the construction of emergency management departments, the emergency management departments can start emergency plans of corresponding levels according to the types and severity of the accidents and conduct commands in a unified manner.

And the judgment of the accident type and the severity degree in the current situation depends on human experience. Therefore, the related art lacks a method capable of analyzing and reasoning secondary derivative disasters possibly caused by leakage and explosion of hazardous chemicals.

Disclosure of Invention

The present disclosure is directed to a disaster analysis method and an electronic device for solving a problem that a secondary derivative disaster can be inferred through accident-related data analysis after a disaster accident occurs.

In a first aspect, the present disclosure provides an electronic device comprising: a memory and a controller, wherein:

the memory for storing a computer program executable by the controller;

the controller is connected with the memory and is configured to:

acquiring parameter values of an induced event of a target disaster; the parameter value of the evoked event is used to represent the severity of the evoked event;

analyzing the parameter values of the induced events by adopting a pre-trained Bayesian network model, and determining the occurrence probability of the target disaster and the secondary derivative disaster caused by the induced events; wherein the Bayesian network model is constructed based on a secondary derivative disaster event graph, entities in the secondary derivative disaster event graph comprise the evoked event, the target disaster and the secondary derivative disaster caused by the target disaster, and relationships between different entities are used for representing causal relationships between entities;

and outputting the occurrence probability of each of the target disaster and the secondary derivative disaster.

In some possible embodiments, the secondary derivative disaster event graph is constructed based on analysis results of known cases and expert knowledge.

In some possible embodiments, the controller is further configured to, when constructing the secondary derivative disaster event graph, analyze the known cases based on a machine learning method, and extract a triple including a cause, a causal relationship, and an accident caused based on the cause, wherein the entities in the secondary derivative disaster event graph include the cause and the accident.

In some possible embodiments, a knowledge graph is constructed in advance, and entities in the knowledge graph are different ground objects and attributes of the different ground objects, wherein the attributes of the different ground objects comprise danger labels and position information; the target disaster and a designated disaster in the secondary derivative disasters have an incidence relation with a danger label;

after the controller determines respective probabilities of occurrence of the target disaster and secondary derivative disaster based on the induced event, the controller is further configured to:

if the occurrence probability of the designated disaster meets the preset risk point mining condition, searching a target entity meeting the preset condition from the knowledge graph; the preset condition is that the distance between the target entity and the occurrence place of the induced event is determined to be within a specified distance range based on the position information of the target entity, and the target entity has a danger label having an association relationship with the specified disaster;

and if the target entity is searched, outputting the target entity.

In some possible embodiments, if the controller searches for the target entity, the controller is further configured to:

acquiring a plan corresponding to the target entity and used for dealing with disasters; and outputting the predetermined plan.

In some possible embodiments, the controller is further configured to:

training the Bayesian network model according to the following method;

respectively executing the following operations for each known case to obtain a quantization result of the known case;

carrying out quantization processing on the severity of the induced event in the known case according to a preset quantization rule to obtain the parameter value of the induced event;

determining a conditional probability that the occurrence of the evoked event caused the target disaster based on the parameter values for the evoked event for the known case;

respectively executing the following steps for each target condition in a plurality of actual conditions of the predicted place:

adjusting the condition probability by adopting the similarity between the target condition of a forecast place and the case condition of the known case place; wherein, a positive correlation exists between the similarity and the adjusted conditional probability;

determining the occurrence probability of secondary derivative disasters in the known case based on the adjusted conditional probability of the target disasters;

after the quantization results of all the known cases are obtained, the quantization results of all the known cases are adopted to train the Bayesian network model;

when the controller analyzes the parameter values of the evoked events using a pre-trained bayesian network model, the controller is configured to:

and analyzing the parameter values of the induced events by adopting a pre-trained Bayesian network model corresponding to a target condition closest to the current condition of the forecast place.

In some possible embodiments, when the controller represents the condition with at least one random variable, the controller is further configured to:

determining a similarity between the target condition of the forecast location and a case condition of the known case location according to the following method;

respectively acquiring quantized values of the random variables of the forecasting place and the place of the known case;

determining a difference degree between a target condition of the predicted place and a case condition of the place where the known case occurs based on the quantized values of the random variables;

and determining the similarity between the target condition of the predicted place and the case condition of the place of the known case by adopting the difference.

In some possible embodiments, the controller determines a similarity between the target condition of the predicted place and the case condition of the place of the known case using the difference degree, the controller being configured to:

determining a similarity between the target condition of the forecast location and the case condition of the venue of the known case based on the following formula:

wherein, ω is_jkRepresenting the degree of similarity, b_jkRepresents the degree of difference, b_maxRepresenting the degree of difference between the random variables of which the degree of difference is greatest, b_minIndicating the degree of difference between the random variables with the least degree of difference.

In a second aspect, the present disclosure also provides a disaster analysis method, including:

acquiring parameter values of an induced event of a target disaster; the parameter value of the evoked event is used to represent the severity of the evoked event;

analyzing the parameter values of the induced events by adopting a pre-trained Bayesian network model, and determining the occurrence probability of the target disaster and the secondary derivative disaster caused by the induced events; wherein the Bayesian network model is constructed based on a secondary derivative disaster event graph, entities in the secondary derivative disaster event graph comprise the evoked event, the target disaster and the secondary derivative disaster caused by the target disaster, and relationships between different entities are used for representing causal relationships between entities;

and outputting the occurrence probability of each of the target disaster and the secondary derivative disaster.

In some possible embodiments, the secondary derivative disaster event graph is constructed based on analysis results of known cases and expert knowledge.

In some possible embodiments, when the secondary derivative disaster event graph is constructed, the machine learning-based method analyzes the known cases, and extracts the triples including the cause, the causal relationship and the accident caused by the cause, wherein the entities in the secondary derivative disaster event graph include the cause and the accident.

In some possible embodiments, a knowledge graph is constructed in advance, and entities in the knowledge graph are different ground objects and attributes of the different ground objects, wherein the attributes of the different ground objects comprise danger labels and position information; the target disaster and a designated disaster in the secondary derivative disasters have an incidence relation with a danger label;

after the determining is based on the respective probabilities of occurrence of the target disaster and the secondary derivative disaster caused by the induced event, the method further comprises:

if the occurrence probability of the designated disaster meets the preset risk point mining condition, searching a target entity meeting the preset condition from the knowledge graph; the preset condition is that the distance between the target entity and the occurrence place of the induced event is determined to be within a specified distance range based on the position information of the target entity, and the target entity has a danger label having an association relationship with the specified disaster;

and if the target entity is searched, outputting the target entity.

In some possible embodiments, if the target entity is searched, the method further includes:

acquiring a plan corresponding to the target entity and used for dealing with disasters; and outputting the predetermined plan.

In some possible embodiments, the method further comprises:

training the Bayesian network model according to the following method;

respectively executing the following operations for each known case to obtain a quantization result of the known case;

carrying out quantization processing on the severity of the induced event in the known case according to a preset quantization rule to obtain the parameter value of the induced event;

determining a conditional probability that the occurrence of the evoked event caused the target disaster based on the parameter values for the evoked event for the known case;

respectively executing the following steps for each target condition in a plurality of actual conditions of the predicted place:

adjusting the condition probability by adopting the similarity between the target condition of a forecast place and the case condition of the known case place; wherein, a positive correlation exists between the similarity and the adjusted conditional probability;

determining the occurrence probability of secondary derivative disasters in the known case based on the adjusted conditional probability of the target disasters;

after the quantization results of all the known cases are obtained, the quantization results of all the known cases are adopted to train the Bayesian network model;

the analyzing the parameter values of the induced events by adopting a pre-trained Bayesian network model comprises the following steps:

and analyzing the parameter values of the induced events by adopting a pre-trained Bayesian network model corresponding to a target condition closest to the current condition of the forecast place.

In some possible embodiments, the condition is represented using at least one random variable, the method further comprising:

determining a similarity between the target condition of the forecast location and a case condition of the known case location according to the following method;

respectively acquiring quantized values of the random variables of the forecasting place and the place of the known case;

determining a difference degree between a target condition of the predicted place and a case condition of the place where the known case occurs based on the quantized values of the random variables;

and determining the similarity between the target condition of the predicted place and the case condition of the place of the known case by adopting the difference.

In some possible embodiments, the determining the similarity between the target condition of the predicted place and the case condition of the place of the known case by using the difference degree includes:

determining a similarity between the target condition of the forecast location and the case condition of the venue of the known case based on the following formula:

wherein, ω is_jkRepresenting the degree of similarity, b_jkRepresents the degree of difference, b_maxRepresenting the degree of difference between the random variables of which the degree of difference is greatest, b_minIndicating the degree of difference between the random variables with the least degree of difference.

According to the embodiment of the disclosure, after an induced event possibly causing a target disaster occurs, parameter values of the induced event of the target disaster are acquired, the parameter values of the induced event are analyzed by adopting a pre-trained Bayesian network model, and the probability of occurrence of the target disaster and a secondary derivative disaster possibly caused by the induced event is determined according to the analysis result of the parameter values of the induced event. Since the pre-trained bayesian network model in the embodiment of the present disclosure is constructed based on the secondary derivative disaster event graph, the analysis and processing of the accident data by the model can obtain the direct disaster (i.e. target disaster) and the secondary derivative disaster related influence possibly caused by the induction time, so as to prevent the disaster in advance, thereby improving the capability of preventing the disaster risk in the handling of the urban dangerous disaster.

Additional features and advantages of the disclosure will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by the practice of the disclosure. The objectives and other advantages of the disclosure may be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings needed to be used in the embodiments of the present disclosure will be briefly described below, and it is apparent that the drawings described below are only some embodiments of the present disclosure, and it is obvious for a person skilled in the art to obtain other drawings based on the drawings without creative efforts.

Fig. 1 is a diagram of an application scenario provided by some embodiments of the present disclosure;

FIG. 2 is a block diagram of a hardware configuration provided by some embodiments of the present disclosure;

fig. 3 is a schematic diagram illustrating hazardous chemical substance disaster accident derivation provided in some embodiments of the present disclosure;

FIG. 4 is a schematic diagram of an example graph of secondary derivative disasters provided by some embodiments of the present disclosure;

fig. 5a is an overall flowchart of a disaster analysis method according to some embodiments of the present disclosure;

FIG. 5b is a schematic illustration of discretization of data for a known case venue provided by some embodiments of the present disclosure;

fig. 6a is another schematic flow chart of a disaster analysis method according to some embodiments of the present disclosure;

fig. 6b is a flow chart of hazardous chemical substance disaster reasoning provided by some embodiments of the present disclosure;

FIG. 7 is a schematic diagram of a case graph and knowledge graph fusion provided by some embodiments of the present disclosure;

fig. 8 is a schematic flow chart of a disaster analysis method according to some embodiments of the present disclosure.

Detailed Description

The technical solutions in the embodiments of the present disclosure will be described in detail with reference to the accompanying drawings. In the description of the embodiments of the present disclosure, "/" means or, unless otherwise specified, for example, a/B may mean a or B; "and/or" in the text is only an association relationship describing an associated object, and means that three relationships may exist, for example, a and/or B may mean: three cases of a alone, a and B together, and B alone exist, and in addition, "a plurality" means two or more than two in the description of the embodiments of the present disclosure.

In the description of the embodiments of the present disclosure, unless otherwise specified, the term "plurality" means two or more, and other terms are similar. It is to be understood that the embodiments described herein are merely for purposes of illustrating and explaining the present disclosure and are not intended to limit the present disclosure, and features of the embodiments and examples of the present disclosure may be combined with each other without conflict.

To further illustrate the technical solutions provided by the embodiments of the present disclosure, the following detailed description is made with reference to the accompanying drawings and the detailed description. Although the disclosed embodiments provide method steps as shown in the following embodiments or figures, more or fewer steps may be included in the method based on conventional or non-inventive efforts. In the case of steps that are logically not necessarily causal, the order of execution of the steps is not limited to that provided by the disclosed embodiments. The method can be executed in sequence or in parallel according to the method shown in the embodiment or the figures when the method is executed in an actual processing procedure or a control device.

In view of the lack of a method capable of analyzing and reasoning the induced time of a target disaster and possibly bringing secondary derivative disasters in the related art, the disaster analysis method provided by the embodiment of the disclosure has the following inventive concept: and constructing a secondary derivative disaster event map based on the known cases for recording the historical disasters, and then constructing a Bayesian network model structure based on the secondary derivative disaster event map. The known cases are adopted to train the leaf network model, so that the target disaster (namely the target disaster) caused by the induction time and a series of secondary derivative disasters caused by the target disaster can be deduced based on some related information of the induction event of the target disaster, and the occurrence probability of each disaster can be given. And based on the obtained reasoning results of the target disaster and the secondary derivative disaster, the reasoning results can be used as reference information for relevant parts to correspondingly process the disaster. Therefore, the secondary derivative disaster event-based graph and the Bayesian network model can be fused to realize reasonable reasoning on the target disaster and the secondary derivative disaster and give the occurrence probability, reasonable and reliable information can be provided for disaster processing so that the secondary derivative disaster can be prevented in advance, and the disaster analysis method provided by the embodiment of the disclosure also has generalization capability, is not only suitable for dangerous chemical substance disaster analysis, but also can be used for any disaster analysis which can cause the secondary derivative disaster, such as earthquake, typhoon, tsunami and the like.

Furthermore, the method and the device can integrate the secondary derivative disaster case diagram and the Bayesian network model to improve the generalization capability of disaster analysis, and in order to improve the accuracy of the inference result output by the Bayesian network model, the difference between the condition of the predicted place and the condition of the known case is used as an important parameter for training when the Bayesian network model is trained, so that the prediction result of the Bayesian network model is optimized.

In addition, in the embodiment of the disclosure, the system can be linked with the knowledge graph, potential risk points are excavated from the knowledge graph of each ground feature in the forecast area and early-warning is performed, and the system can be further linked with plans of different risk points to provide plans of the potential risk points, so that valuable reference information can be conveniently output for disaster early-warning and disposal of related personnel.

The disaster analysis method in the embodiment of the present disclosure is described in detail below with reference to the drawings.

As shown in fig. 1, an application environment of a disaster analysis method provided by the present disclosure is schematically illustrated, and the application environment may include anetwork 10, aserver 20, and at least oneterminal device 30, where: theterminal device 30 may include some sensors, such as a temperature and humidity sensor, a fire sensor, and a toxic gas sensor, to acquire some information of the monitored area, and report the information to the server for processing. The Internet of things can be established by different sensors, so that the server can conveniently perform centralized management on the different sensors. In addition, theterminal device 30 may further include an intelligent terminal device such as a smart phone, a desktop computer, a smart watch, a smart bracelet, a tablet computer, a notebook computer, etc. so as to communicate with the sensor and theserver 20 to obtain corresponding information as needed.

Theterminal device 30 may collect the severity of an induced event that may induce the target disaster, and report the severity of the induced event as a parameter value to theserver 20. Theserver 20 stores a bayesian network model trained in advance, analyzes the parameter values of the input induced event, infers the occurrence probability of the target disaster and the secondary derivative disaster caused by the induced event, and outputs the result to theterminal device 30 capable of displaying.

Meanwhile, theserver 20 may combine the knowledge map of the emergency records to dig out potential risk points and provide related plans for disaster disposal.

In the embodiment of the present disclosure, the server may perform disaster analysis, the terminal may perform disaster analysis, or a part of the steps of the disaster analysis may be performed by the server and a part of the steps may be performed by the terminal device, which is not limited in the present disclosure.

To facilitate understanding of the disaster analysis method provided by the present disclosure, a hardware configuration block diagram of the electronic device 130 for disaster analysis is exemplarily illustrated in fig. 2. The electronic device may be a server as described above, or may be a terminal device with disaster handling capability.

The electronic device 130 according to this embodiment of the present disclosure is described below with reference to fig. 2, and the electronic device 130 shown in fig. 2 is only an example and should not bring any limitation to the functions and the use range of the embodiment of the present disclosure.

As shown in fig. 2, the electronic device 130 is represented in the form of a general electronic device. The components of the electronic device 130 may include, but are not limited to: at least oneprocessor 131, at least one memory 132, and abus 133 that connects the various system components, including the memory 132 and theprocessor 131.

Bus 133 represents one or more of any of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a processor, or a local bus using any of a variety of bus architectures.

The memory 132 may include readable media in the form of volatile memory, such as Random Access Memory (RAM)1321 and/orcache memory 1322, and may further include Read Only Memory (ROM) 1323.

Memory 132 may also include a program/utility 1325 having a set (at least one) ofprogram modules 1324,such program modules 1324 including, but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may comprise an implementation of a network environment.

The electronic device 130 may also communicate with one or more external devices 134 (e.g., keyboard, pointing device, monitor, sensor, etc.), with one or more devices that enable a user to interact with the electronic device 130, and/or with any devices (e.g., router, modem, etc.) that enable the electronic device 130 to communicate with one or more other electronic devices. Such communication may occur via input/output (I/O) interfaces 135. Also, electronic device 130 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network such as the Internet) vianetwork adapter 136. As shown in FIG. 2,network adapter 136 communicates with other modules for electronic device 130 viabus 133. It should be understood that although not shown in the figures, other hardware and/or software modules may be used in conjunction with electronic device 130, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, among others.

In some possible embodiments, the disaster analysis method provided by the present disclosure may also be implemented in the form of a program product including program code for causing an electronic device to perform the steps of a disaster analysis method according to various exemplary embodiments of the present disclosure described above, when the program product is run on a computer device.

The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

The disaster analysis program product of the present disclosure may employ a portable compact disc read only memory (CD-ROM) and include program code, and may be run on an electronic device. However, the program product of the present disclosure is not so limited, and in the present disclosure, the readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.

Readable signal media may include a propagated data signal with readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A readable signal medium may also be any readable medium that is not a readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.

Program code embodied on a readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

Program code for carrying out operations for the present disclosure may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the consumer electronic device, partly on the consumer electronic device, as a stand-alone software package, partly on the consumer electronic device and partly on a remote electronic device, or entirely on the remote electronic device or server. In the case of remote electronic devices, the remote electronic devices may be connected to the consumer electronic device through any kind of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or may be connected to an external electronic device (e.g., through the internet using an internet service provider).

In the disaster analysis method provided by the present disclosure, the severity of an induced event of a target disaster is analyzed through a pre-trained bayesian network model, and the target disaster and secondary derivative disasters caused by the induced event are further inferred according to the analysis result. In the following, how to train to obtain the bayesian network model, how to infer the occurrence probability of the target disaster and the secondary derivative disaster through the bayesian network model, and how to mine the risk points of the accident potential are explained in detail.

It should be noted that, the disaster analysis method provided in the present disclosure is described below by taking hazardous chemical substance disasters as an example, and it should be understood that other disaster analysis methods are similar to the hazardous chemical substance disaster analysis method and will not be described in detail below.

Part 1: construction of Bayesian network model

Because the bayesian network model in the present disclosure is constructed based on the case map, the case map needs to be constructed before the model is constructed, and the construction process of the case map is as follows:

firstly, accident reasons (namely induced events), induced secondary derivative disasters and disaster consequences are screened from a large number of dangerous chemical accident cases, and an evolution relation among accidents is established by combining experience and conclusion given by experts.

In order to learn more evolutionary relationships among accidents, in some possible embodiments, a machine-based learning method may be used to extract logical triple relationships among hazardous chemical substance disaster cases as a supplement to the evolutionary relationships among the accidents. Wherein the triplet comprises: the cause and cause of the dangerous chemical substance accident and the bad results based on the cause.

And then, constructing a hazardous chemical substance disaster mode on the basis of the evolution process of the hazardous chemical substance disaster. In order to express the hazardous chemical substance disaster mode more intuitively, the present disclosure takes the damage of the hazardous chemical substance device as an example of an accident cause. As shown in fig. 3, in the schematic diagram of the hazardous chemical substance disaster derivation mode, the main cause of the hazardous chemical substance leakage accident is that the container device of the hazardous chemical substance is damaged, the target disasters generated by the accident are the leakage of poisons and flammable and explosive substances and fires, the secondary derived disasters derived from the target disasters are poisoning, fires and explosions, and the accident finally causes serious consequences of environmental pollution, casualties and traffic interruption.

And finally, based on the hazardous chemical substance disaster mode, establishing accident entities and establishing the relationship among the entities to obtain a secondary derivative disaster event map. Taking a dangerous chemical substance leakage accident as an example, the secondary derivative disaster event graph is shown in fig. 4, arrows in fig. 4 represent causal relationships, and if the dangerous chemical substance leakage causes: pipe damage, tank rupture, production equipment damage. And the leakage accidents of target disasters and hazardous chemicals are caused, and further the leakage of flammable and explosive gases and toxic gases are caused in derived disasters. The secondary derived disasters include: explosion, fire, land pollution, atmospheric pollution, water pollution, building damage, casualties, traffic jams, crop death, and human poisoning.

And after the secondary derivative disaster event graph is obtained, constructing a Bayesian network model. And the nodes of the Bayesian network of the entity machine in the secondary derivative disaster event graph are used for constructing a Bayesian network model structure.

In some embodiments, the bayesian network model may calculate the probability of occurrence of the disaster through a probability density function of gaussian distribution for continuous values, i.e. under the condition of satisfying the normal distribution. In the embodiment of the disclosure, in order to obtain a more accurate result and make the calculation process simple and effective, the related data in the secondary derivative disaster event graph is discretized, and the occurrence probability of each disaster in the secondary derivative disaster event graph is calculated and output based on the discrete value through the bayesian network structure. As shown in table 1, a discretization example of different nodes in the bayesian network model. In table 1, taking the index a as an example, the status value is used to describe the severity, and the severity corresponds to a value of 1, the severity corresponds to a value of 2, and the general degree corresponds to a value of 3. Some nodes are described by severity and some nodes can be described by occurrence or non-occurrence, for example crop death, where a "yes" status value indicates a probability of occurrence of crop death and a "no" status value indicates a probability of occurrence of non-occurrence of crop death. It should be noted that the setting of the specific state value and the corresponding value thereof may be set according to actual requirements, and table 1 is only an example and is not used to limit the embodiment of the present disclosure.

TABLE 1 discretization example of different nodes in a Bayesian network model

Serial number	Input variable/index	Status value	Numerical value corresponding to node
				A	Destruction of production equipment	Severe/major/general	1/2/3
B	Tank rupture	Severe/major/general	1/2/3
				C	Destruction of pipelines	Severe/major/general	1/2/3
D	Dangerous chemical leakage accident	Severe/major/general	1/2/3
				E	Inflammable and explosive gas leakage	Yes/no	4/5
F	Toxic gas leakage	Yes/no	4/5
				G	Fire hazard	Is serious-General/non-accident	1/2/3
H	Traffic block	Yes/no	4/5
				I	Explosion of the vessel	Yes/no	4/5
J	Casualty	Yes/no	4/5
				K	Destruction of buildings	Yes/no	4/5
L	Poisoning by	Yes/no	4/5
				M	Pollution of water	Severe/general/noaccident	1/2/3
N	Pollution of the atmosphere	Severe/general/noaccident	1/2/3
				O	Pollution of soil	Severe/general/noaccident	1/2/3
P	Poisoning of people	Yes/no	4/5
				Q	Death of crops	Yes/no	4/5

To complete the training of the bayesian network model, it is implemented with known cases as training samples, and then as shown in fig. 5a, in step 501: respectively executing the following operations for each known case to obtain the quantization result of the known case, wherein the operations comprise the following substeps;

step 5011: carrying out quantization processing on the severity of the induced event in the known case according to a preset quantization rule to obtain a parameter value of the induced event;

the three items A, B, C in Table 1 are used as the evoked events, and the parameter values of the evoked events in each known case are obtained.

Step 5012: based on the parameter values of the induced events of the known cases, the conditional probability that the induced event occurs to cause the target disaster is determined.

The conditional probability obtained in step 5012 is a discretized value, i.e., a value as described in table 1.

In the embodiment of the present disclosure, the known cases may be from different regions and induced disaster accidents under different conditions. When learning is performed on the known case, the learning can be performed indiscriminately. The bayesian network model thus obtained reflects an average case. However, disaster analysis is related to the conditions of the region, and hazardous chemical substance disasters in different time or conditions may cause different results (for example, flammable and explosive substances leak under rainfall conditions, and leaks under dry weather conditions may cause completely different results). The conditions of the terrain may include, for example, as shown in table 2, terrain conditions (e.g., hills, plains, etc.), weather conditions, and rainfall concerns. Of course, in specific implementation, the conditions of the regions may be set according to actual requirements, and table 2 is only an example and is not used to limit the embodiment of the present disclosure.

TABLE 2

Region(s)	Precipitation amount	Climatic conditions	……	Topographic conditions
					City of Enshi Hubei
Guizhou Qinglong county
					Jiangsu Xiangshui county
Prediction ground

Since the situation of the area may affect the disaster situation, the indiscriminate learning may cause a certain difference between the prediction result of the prediction area and the actual situation. In view of this, in the embodiment of the present disclosure, when training the bayesian network model, the difference learning may be performed according to the actual situation of the predicted place and the case situations of each known case. The principle can be summarized as increasing the learning strength of known cases with similar conditions. For example, the conditional probability of occurrence of the target disaster may be adjusted by using the similarity between the actual situation of the prediction site and the case situation of the known case as a weight. The Bayesian network model is then trained based on the adjusted conditional probabilities.

In some embodiments, in order to further optimize the bayesian network model to obtain a more accurate inference result, the embodiments of the present disclosure may also establish bayesian network models under different conditions for the same prediction. It can be implemented that instep 502, the known cases are learned to obtain the corresponding bayesian network model in each target situation. May include the following substeps: respectively executing the following steps for each target condition in a plurality of actual conditions of the predicted place:

step 5021, adjusting the condition probability by adopting the similarity between the target condition of the forecast place and the case condition of the known case place; wherein, the similarity and the adjusted conditional probability have positive correlation.

In some embodiments, as shown in table 2, the conditions of different regions may be represented by a plurality of dimensions, each dimension (e.g., terrain condition, climate condition) is respectively used as a random variable, and each random variable has its corresponding value after being quantized. In implementation, the similarity between the target condition of the forecast place and the case condition of the known case place can be realized as comprising the following steps;

step A1: respectively acquiring quantized values of random variables of a prediction place and a place where a known case occurs;

step A2: determining a difference degree between a target condition of a predicted place and a case condition of a place of known case based on the quantized values of the random variables;

step A3: and determining the similarity between the target condition of the predicted place and the case condition of the place of the known case by adopting the difference degree.

When the method is implemented, firstly, survey needs to be implemented on the random variable of the accident prediction place to obtain the data of the random variable of the target condition of the prediction place. Then, according to the principle of variance, the difference degree b between the accident prediction place and the accident area can be obtained by calculating the sum of the cumulative squared differences of the random variables of the known case occurrence place and the accident prediction place_jkAnd then according to the obtained difference degree b_jkObtaining the similarity omega by the proportional relation of the total distance of the difference degree between the place where the known case occurs and the place where the accident is predicted_jk. Wherein the differenceDegree b_jkThe calculation method of (2) is shown in formula (1):

note that the method of calculating the degree of difference for any random variable is not limited to that shown in formula (1), and for example, the difference between the ratio of the two and 1 may be used as the degree of difference.

Similarity omega_jkThe calculation method of (2) is shown in formula (2):

wherein, y_iIs the ith random variable, x, of the place where the known pattern occurs_iIs the ith random variable of the accident prediction place, b_maxRepresenting the degree of difference between the random variables of which the degree of difference is greatest, b_minIndicating the degree of difference between the random variables with the least degree of difference.

It can be understood that the higher the similarity between the place where the known case occurs and the accident prediction place, the more valuable the similar accidents occurring in the area are.

Finally, the obtained similarity ω is used_jkAnd substituting a Bayesian formula to obtain a corrected probability after correcting the conditional probability. For more intuitive expression, the conditional probability is adjusted by using the similarity as a weight, and the adjusted conditional probability is obtained. For example, taking the probability of occurrence of a serious hazardous chemical leakage accident under the condition that the damage condition of the production equipment is serious as an example, the method for optimizing the condition probability is shown as formula (3):

V_correction＝V_Initial*ω_jkP (dangerouschemical leakage accident 1| -production equipment damage 1) × ω_jkFormula (3)

Wherein, p (dangerouschemical leakage accident 1| damage of production equipment 1) indicates that when the production equipment is damaged and the damage degree is serious, the probability of serious dangerous chemical leakage occurs, and "═ 1" is that when the accident state is serious, the discretization value of the corresponding accident is 1.

Step 5022, determining the occurrence probability of secondary derivative disasters in the known case based on the adjusted conditional probability of the target disasters;

step 5023, after the quantization result of each known case is obtained, the quantization result of each known case is adopted to train the Bayesian network model.

As shown in fig. 5b, wherein the numbers 1-16 represent 16 known cases, the discretization values of the known cases in each action, wherein a-U represent the discretization values of the evoked events, the target disasters and various derived disasters, respectively.

For different target conditions of the forecast place, the similarity between the forecast place and the case condition of each known case under each target condition can be respectively determined, and then the conditional probability of the target disaster (such as dangerous chemical leakage) sending is modified. Assuming that term D in FIG. 5b represents a hazardous chemical leakage and three events A-C are the three triggering events that lead to the hazardous chemical leakage, the conditional probability is modified, e.g., V_{Initial value}P (dangerouschemical leakage accident 1|,production equipment damage 1,storage tank rupture 2, pipeline rupture 2). Namely, the probability of serious dangerous chemical leakage accidents after serious production equipment damage and serious storage tank and pipeline breakage is known.

Part 2: inferring probability of occurrence of target and secondary derivative disasters

Based on the foregoing, the bayesian network model is constructed based on a secondary derivative disaster event graph, wherein the entities in the secondary derivative disaster event graph include an induced event, a target disaster and a secondary derivative disaster, and the relationships between the entities are used for representing causal relationships between the entities. After the bayesian network model is constructed and trained, the parameter values of the induced events can be analyzed and processed by the model, and the disasters can be inferred and predicted. As shown in fig. 6a, a schematic flow chart of a disaster analysis method provided in the embodiment of the present disclosure includes:

601, acquiring parameter values of an induced event of a target disaster; the parameter values of the evoked events are used to indicate the severity of the evoked events.

In the implementation process, the severity of each induced event is acquired by taking the induced event of each target disaster in the case graph of the secondary derivative disaster as a node, wherein the severity of each induced event comprises the following steps: the severity state of the induced event and the preset parameter value corresponding to each state. The quantization data shown in table 1 are not described herein.

Step 602, analyzing the parameter values of the induced events by using a pre-trained bayesian network model, and determining respective occurrence probabilities of the target disaster and the secondary derivative disaster caused by the induced events.

In another embodiment, there are multiple Bayesian network models that analyze the same disaster in the same forecast. That is, as previously described, one model is trained for each of the different target conditions. Therefore, in implementation, after obtaining the parameter value of the induced event of the target disaster, the bayesian network model corresponding to the target condition closest to the current condition of the prediction site can be found according to the current condition of the prediction site. And then analyzing the parameter values of the induced events by adopting the Bayesian network model to obtain an inference result.

In implementation, for example, simplicity and reduced calculation amount, the occurrence probability of different disasters in the embodiment of the disclosure is a discretization result. Taking the hazardous chemical substance disaster in fig. 6B as an example, the induced events are shown as a production equipment damage, a storage tank rupture and a pipeline damage, and the parameter values of the induced events are 3, 2 and 1 respectively as shown in fig. 6B. The target disaster is the D-risk chemical leakage accident shown in fig. 6 b; the probability of occurrence of each disaster including E flammable and explosive gas leakage and F toxic gas leakage D, E, F shown in fig. 6b can be inferred through a bayesian network model. In fig. 6b, the occurrence probability of the hazardous chemical substance leakage accident includes 70% of the occurrence probability when the result is "serious", 20% of the occurrence probability when the result is "significant", and 10% of the occurrence probability when the result is "normal". And the probability of occurrence of toxic gas leak includes 70% when the result is "yes" and 30% when the result is "no".

Step 603, outputting the occurrence probability of the target disaster and the secondary derivative disaster.

For example, when a server performs disaster analysis, the occurrence probability of each disaster may be output in the form shown in fig. 6b so that a user can know the inducing condition of the disaster and the occurrence probability of each disaster. In the output display, as shown in fig. 6b, a disaster having a high occurrence probability may be highlighted so that the user can pay attention to the dangerous disaster.

In another embodiment, it can be inferred that the occurrence probability of the disaster is only a part of the disaster processing, and the disclosed embodiment helps the user to perform disaster disposal in order to provide more comprehensive information. A knowledge map of the predicted site can be constructed. The knowledge graph comprises all ground features and attributes of the same prediction place. The attributes of the method can comprise geographic positions and information of responsible persons, and danger labels can be additionally added so as to be conveniently linked with the inference result of the Bayesian network model. How to mine the potential risk points in conjunction with the knowledge graph is described below.

Part 3: mining potential risk points of accident

In implementation, danger labels are added to all entities in the knowledge graph, and the danger labels are correlated with target disasters in the secondary derivative disaster event graph and designated disasters in the secondary derivative disasters. And when the secondary derivative disaster affair map is linked with the knowledge map, potential risk points are mined through the association relation.

For example, after the occurrence probability of each of the target disaster and the secondary derivative disaster caused by the induced event is estimated by the bayesian network model, if the occurrence probability of the designated disaster satisfies the preset risk point mining condition, the target entity satisfying the preset condition in the knowledge graph can be searched. The preset condition is that the distance between the target entity and the occurrence place of the induced event is within a specified distance, and the target entity and the specified disaster have an association relation.

An exemplary embodiment is shown in fig. 7, assuming that the release of flammable and explosive substances in the secondary derivative disaster event graph leads to chemical fire explosion, the entity in the event graph has an association relationship with the flammable and explosive danger label marked in the knowledge graph. Therefore, when the probability of chemical fire explosion is deduced to be high (namely, when the condition of mining the risk points of chemical fire explosion is met) according to the Bayesian network model, entities with flammable and combustible attributes are searched on the knowledge graph within 2km of the region where the chemical fire explosion occurs. As shown in fig. 7, if thegas station 1 is searched and has flammable and explosive properties, the danger label of thegas station 1 is activated. The activated danger label is combined with the targetentity gas station 1 to find out a corresponding plan and output the plan to the user as reference information for disaster disposal. For example, an example of potential risk points that are not located by mining is shown in table 3.

TABLE 3

	Risk point 1	Risk point 2	Risk points 3
				Entityname	Gasoline station	1
Danger label	Inflammable and explosive
				Location information	Warp and weft value
Person in charge	Zhang three
				Inferring information	Distance is less than or equal to 2Km, attention is paid to protection

After the disaster analysis method provided by the present disclosure is introduced as a whole, how to prevent a secondary derivative disaster that may occur by using the analysis method provided by the present disclosure after a hazardous chemical substance disaster accident occurs is illustrated in fig. 8 for convenience of understanding. The method is used for explaining the overall process of the analysis method of the hazardous chemical derivative disasters, and comprises the following steps:

step 801: analyzing dangerous chemical substance disaster cases, and combining expert conclusions and suggestions after obtaining accident reasons, induced secondary disasters and disaster results screened from dangerous chemical substance accident cases, in step 802: after the second derivative disaster event graph is obtained by constructing the second derivative of the hazardous chemical substance, in step 803, the quantitative operation can be performed on the relevant parameters of each induced event in the event graph, and instep 804, a Bayesian model structure is constructed based on the second derivative disaster event graph; then, with reference to the method for optimizing the bayesian network model provided in this disclosure, instep 805, the bayesian network model is trained based on the known cases so that the occurrence probability of the target disaster and the secondary derivative disaster of the hazardous chemical substance disaster accident can be calculated.

In order to further extract potential risk points, instep 806, the secondary derivative disaster event graph and the knowledge graph are fused to establish an association relationship between the designated disaster and the danger label in the knowledge graph.

After receiving and reporting the accident, quantifying the received and reported accident as an induced event to obtain a parameter value of the induced event, and then performing hazardous chemical substance grade derivative disaster analysis based on the trained bayesian network model instep 807. After the analysis result is obtained, instep 808, when the designated disaster of the analysis result meets the risk point mining condition, potential risk points are mined based on the secondary derivative disaster event graph and knowledge graph fused instep 806, and corresponding plans are searched for output and displayed. For example, when a dangerous chemical substance disaster accident occurs, if the specified disaster in the secondary derivative disaster event graph meets the risk point mining condition, the potential risk point is mined in the knowledge graph according to the risk label of the specified disaster with the association relationship.

In addition, in order to train a bayesian network model with more pertinence and prediction for a specific area, when the analysis of the secondary derivative disasters of the dangerous chemicals is performed instep 807, the relevant parameters of the known case occurrence place and the accident prediction place need to be calculated by the optimized bayesian formula provided by the present disclosure, so as to obtain the potential risk points of the secondary derivative disasters possibly caused by the dangerous chemicals disaster accidents.

Claims (10)

1. An electronic device, characterized in that the electronic device comprises: a memory and a controller, wherein:

the memory for storing a computer program executable by the controller;

the controller is connected with the memory and is configured to:

acquiring parameter values of an induced event of a target disaster; the parameter value of the evoked event is used to represent the severity of the evoked event;

analyzing the parameter values of the induced events by adopting a pre-trained Bayesian network model, and determining the occurrence probability of the target disaster and the secondary derivative disaster caused by the induced events; wherein the Bayesian network model is constructed based on a secondary derivative disaster event graph, entities in the secondary derivative disaster event graph comprise the evoked event, the target disaster and the secondary derivative disaster caused by the target disaster, and relationships between different entities are used for representing causal relationships between entities;

and outputting the occurrence probability of each of the target disaster and the secondary derivative disaster.

2. The electronic device of claim 1, wherein the secondary derivative disaster event graph is constructed based on analysis results of known cases and expert knowledge.

3. The electronic device of claim 2, wherein the controller is further configured to analyze known cases based on a machine learning method to extract triples including causes, causal relationships and incidents based on the causes when constructing the secondary derivative disaster event graph, wherein entities in the secondary derivative disaster event graph include the causes and the incidents.

4. The electronic equipment according to claim 1, wherein a knowledge graph is constructed in advance, and the entities in the knowledge graph are different ground objects and attributes of the different ground objects, wherein the attributes of the different ground objects comprise danger labels and position information; the target disaster and a designated disaster in the secondary derivative disasters have an incidence relation with a danger label;

after the determining is based on respective probabilities of occurrence of the target disaster and a secondary derivative disaster caused by the induced event, the controller is further configured to:

if the occurrence probability of the designated disaster meets the preset risk point mining condition, searching a target entity meeting the preset condition from the knowledge graph; the preset condition is that the distance between the target entity and the occurrence place of the induced event is determined to be within a specified distance range based on the position information of the target entity, and the target entity has a danger label having an association relationship with the specified disaster;

and if the target entity is searched, outputting the target entity.

5. The electronic device of claim 4, wherein if the controller searches for the target entity, the controller is further configured to:

acquiring a plan corresponding to the target entity and used for dealing with disasters;

and outputting the plan.

6. The electronic device of any of claims 1-4, wherein the controller is further configured to:

training the Bayesian network model according to the following method;

respectively executing the following operations for each known case to obtain a quantization result of the known case;

carrying out quantization processing on the severity of the induced event in the known case according to a preset quantization rule to obtain the parameter value of the induced event;

determining a conditional probability that the occurrence of the evoked event caused the target disaster based on the parameter values for the evoked event for the known case;

respectively executing the following steps for each target condition in a plurality of actual conditions of the predicted place:

adjusting the conditional probability by adopting the similarity between the target condition of a forecast place and the case condition of the known case place; wherein, a positive correlation exists between the similarity and the adjusted conditional probability;

determining the occurrence probability of secondary derivative disasters in the known case based on the adjusted conditional probability of the target disasters;

after the quantization results of the known cases are obtained, the Bayesian network model is trained by adopting the quantization results of the known cases;

the controller, in performing the analysis of the parameter values of the evoked events using a pre-trained bayesian network model, is configured to:

and analyzing the parameter values of the induced events by adopting a pre-trained Bayesian network model corresponding to a target condition closest to the current condition of the forecast place.

7. The electronic device of claim 6, wherein when the controller represents the condition with at least one random variable, the controller is further configured to:

determining a similarity between the target condition of the forecast location and a case condition of the known case location according to the following method;

respectively acquiring quantized values of the random variables of the forecasting place and the place of the known case;

determining a degree of difference between a target condition of the predicted place and a case condition of the place of occurrence of the known case based on the quantized values of the respective random variables;

and determining the similarity between the target condition of the predicted place and the case condition of the place of the known case by adopting the difference.

8. The electronic device of claim 7, wherein the controller, when determining the similarity between the target condition of the forecast location and the case condition of the venue of the known case using the degree of difference, is configured to:

determining a similarity between the target condition of the forecast location and the case condition of the venue of the known case based on the following formula:

wherein, ω is_jkRepresenting the degree of similarity, b_jkRepresents the degree of difference, b_maxRepresenting the degree of difference between the random variables of which the degree of difference is greatest, b_minIndicating the degree of difference between the random variables with the least degree of difference.

9. A method for disaster analysis, the method comprising:

acquiring parameter values of an induced event of a target disaster; the parameter value of the evoked event is used to represent the severity of the evoked event;

analyzing the parameter values of the induced events by adopting a pre-trained Bayesian network model, and determining the occurrence probability of the target disaster and the secondary derivative disaster caused by the induced events; wherein the Bayesian network model is constructed based on a secondary derivative disaster event graph, entities in the secondary derivative disaster event graph comprise the evoked event, the target disaster and the secondary derivative disaster caused by the target disaster, and relationships between different entities are used for representing causal relationships between entities;

and outputting the occurrence probability of each of the target disaster and the secondary derivative disaster.

10. The method according to claim 9, wherein a knowledge graph is constructed in advance, and the entities in the knowledge graph are different ground objects and attributes of the different ground objects, wherein the attributes of the different ground objects comprise danger labels and position information; the target disaster and a designated disaster in the secondary derivative disasters have an incidence relation with a danger label;

after the determining is based on the respective probabilities of occurrence of the target disaster and the secondary derivative disaster caused by the induced event, the method further comprises:

if the occurrence probability of the designated disaster meets the preset risk point mining condition, searching a target entity meeting the preset condition from the knowledge graph; the preset condition is that the distance between the target entity and the occurrence place of the induced event is determined to be within a specified distance range based on the position information of the target entity, and the target entity has a danger label having an association relationship with the specified disaster;

and if the target entity is searched, outputting the target entity.

Images