CN115239153A - Emergency command decision-making method, system and medium for chemical industry park accidents - Google Patents
Emergency command decision-making method, system and medium for chemical industry park accidents Download PDFInfo
- Publication number
- CN115239153A CN115239153A CN202210890450.7A CN202210890450A CN115239153A CN 115239153 A CN115239153 A CN 115239153A CN 202210890450 A CN202210890450 A CN 202210890450A CN 115239153 A CN115239153 A CN 115239153A
- Authority
- CN
- China
- Prior art keywords
- accident
- plan
- node
- label
- chemical industry
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
Abstract
The invention discloses an emergency command decision method for accidents in a chemical industry park, which comprises the steps of collecting accident data in real time, analyzing a sensitive point and an accident model through accident real-time sensing, analyzing an analysis result based on a preset multi-level characteristic label library to obtain a label, and evaluating through a preset decision tree model to obtain an event node corresponding to a current accident; and carrying out accident evolution and comprehensive risk analysis on the event nodes under the assistance of a multi-level feature tag library by utilizing the Bayesian network to generate a plan link and an execution strategy which accord with the current accident scene. And continuously circulating until the accident is solved. Feature extraction is carried out by means of Jieba word segmentation, N-gram characterization and KNN clustering, a multi-level structured plan label library is output, inference prediction of accident situation development and dynamic strategy suggestion generation can be carried out, and plan adaptability and satisfaction evaluation are carried out on future possible unknown and known risks by considering the dynamic evolution characteristics of the accident situation development so as to carry out real-time adjustment suggestion.
Description
Technical Field
The invention belongs to the field of risk processing, and particularly relates to an emergency command decision method, an emergency command decision system and an emergency command decision medium for accidents in a chemical industry park.
Background
Along with the rapid development of economy in recent years, the petrochemical industry has also made great progress, and in addition, the country requires that petrochemical enterprises enter chemical industry gardens, and the quantity and the scale of chemical industry gardens increase with each day, and the chemical industry garden becomes potential accident high-rise area. Therefore, how to construct a rapid and effective structured plan in a petrochemical park and how to scientifically and reasonably dynamically adjust the plan in actual accident emergency command become important research hotspots.
Although the text plan provides an important guiding function in the emergency command process of accidents in the park, accidents of all the times still remind people that the emergency still exceeds the guiding scope of a specific plan object, along with the popularization and the application of digital technologies, the combined utilization of various models and analysis and application can help to improve the accident command efficiency and correct operation, and the accident risk, particularly the possibility of occurrence of secondary derivative risks is reduced. Through analyzing the accident characteristics and rules of chemical and dangerous chemicals, the accident development situation is scientifically researched and judged in the accident commanding process, the effectiveness and the availability of the plan are analyzed in real time, the problems are timely found, effective measures are taken aiming at the problems, and the accident consequence is very necessary to be reduced.
The structured plan is to carry out structured decomposition on the elements such as emergency events, event levels, organizations and responsibilities, emergency response processes, emergency materials and the like on the basis of a text plan. In the prior art, from emergency service logic, an event information stream is used as a driving mechanism to bind a process, a node and a task, activate execution of the task in an emergency process chain, and finally implement the process of an emergency plan. In the prior art, an optimization strategy and a method based on scene construction are also provided, the optimization of a plan system is promoted from a micro level, and the structural problems of the emergency plan system at the micro, meso and macro levels are solved to a certain extent. The prior art also has the characteristics of multi-source isomerism, complex association, dynamic evolution and the like aiming at abnormal situation data, joint mining is carried out on the multi-source isomerism data, a knowledge graph is constructed, and a data base and an auxiliary decision support are provided for the dynamic generation of an intelligent emergency plan. In addition, an accident mechanism model is also based, the accident disaster consequences are calculated in real time according to an emergency plan flow and emergency rescue measures, and a chemical accident emergency plan three-dimensional dynamic deduction simulation system which can simulate scenes, edit plans and intelligently evaluate the scenes is designed and researched. In addition, in the prior art, a theoretical method of an accident tree (FTA) is also used, the importance of the structured instructions in the emergency process is comprehensively analyzed and evaluated, the importance of the emergency instructions is sequenced, and the plan instructions are intelligently pushed according to the emergency importance to assist in on-site command. In addition, a semi-structured emergency plan is uniformly expressed by using an extensible markup language, and new emergency problem description is extracted; preferentially calculating the content of each part with high similarity of the same type of emergency plan, and generating a semi-structured emergency plan which conforms to the new emergency; and evaluating the generated emergency plan by using an analytic hierarchy process according to the emergency plan evaluation index system.
At present, the structured emergency plan still faces many challenges in the practical application process: 1) In the dynamic emergency decision command process, the structured plan cannot be sensed and flexibly adjusted in real time along with the characteristic that the event is rapidly changed, the real-time correction capability is weak, and the application value of the plan is influenced; 2) Research is focused on how a structured plan is generated reasonably according to the rapid and dynamic state of an event, and the data interaction and evolution adaptive capacity under the multi-dimensional scene of characters, events, space, time and the like involved in the response process of the plan is ignored.
Disclosure of Invention
The invention aims to provide an emergency command decision method, an emergency command decision system and an emergency command decision medium for accidents in a chemical industry park, so as to solve the problem that real-time updating and adaptation to accident judgment are difficult to carry out.
In order to solve the problems, the technical scheme of the invention is as follows:
an emergency command and decision method for accidents in chemical industrial parks comprises the following steps
S1: acquiring accident data in real time, analyzing the sensitive points and the accident model through real-time accident sensing to respectively obtain sensitive point information and an accident model analysis result, analyzing the accident data, the sensitive point information and the accident model analysis result based on a preset multi-level feature label library to obtain labels, and evaluating through a preset decision tree model to obtain event nodes corresponding to the current accident;
s2: accident evolution and comprehensive risk analysis are carried out on event nodes by utilizing a Bayesian network based on a multi-level feature tag library, and a pre-arranged link and an execution strategy which are in line with the current accident scene are generated;
s3: and repeating the steps S1 and S2 until the accident is solved.
Further preferably, before step S1 is executed, a text plan library, a structured plan library, an emergency resource archive and a knowledge base need to be acquired to construct a campus foundation support framework for providing data support for a multi-level feature tag library, a decision tree model and a bayesian network.
Further preferably, before step S1 is executed, a multi-level feature tag library is further constructed based on the campus foundation support framework, and the specific step is
A1: vectorizing the single n-gram after word segmentation to obtain an n-gram vector;
a2: performing superposition averaging on the words and the n-gram vectors in the park basic support frame to obtain a document vector;
a3: and selecting part of typical documents and corresponding document vectors thereof, and labeling the typical documents with labels to obtain a multilevel characteristic label library.
Wherein, the specific steps of obtaining the label based on the analysis of the preset multi-level feature label library in the step S1 are
B1: acquiring the calculation difference between an unknown document and a typical document in a multilevel feature tag library;
b2: and obtaining a plurality of typical documents with the minimum difference with the unknown document, and determining the label of the unknown document according to a voting method.
Further preferably, a decision tree model is constructed before step S1 is performed
C1: acquiring historical detection data, manual identification information, historical sensitive point information, historical accident model analysis results and corresponding labels as training data;
c2: determining the optimal division attribute a by carrying out the calculation of the Gini index according to the training data * And optimally dividing the attribute a * As a parent node;
c3: traversing optimal partition attribute a * Each value of
C4: generating a branch node for the father node, and taking values
Forming a subset of samples D v If the labels in the sample subset are uniform or the decision tree model is reachedIf the maximum depth or no other separable attributes exist, taking the branch node as a leaf node, and taking the maximum label of the sample subset as a label of the branch node; otherwise, the sample subset D is discarded v The optimal partition attribute a of * Repeating steps C2 to C4 for the remaining attributes;
c5: and after all the leaf nodes are generated, outputting the decision tree model.
Specifically, in step C2, the formula for the calculation of the kini index is as follows:
wherein the content of the first and second substances,
wherein D is the data set of the father node, y is the label kind, p k Is the proportion of the k-th class attribute in the data set D.
Wherein, the step S2 specifically comprises the following steps
D1: establishing a Bayesian network of an accident development situation according to the association among the event objects;
d2: obtaining a current event node, discretizing a label in the event node to determine a state value of the event node, and determining prior probability through statistics;
d3: performing accident reasoning according to the labels in the event nodes, and calculating the joint probability of each accident development scheme;
d4: determining the accident development with the maximum possibility according to the joint probability and generating a corresponding structured plan;
d5: evaluating the structured plan, and outputting the structured plan if the evaluation is passed; otherwise, deleting the structured plan and repeating the steps D2 to D5.
Specifically, the prior probability is calculated by the formula
Wherein X ni Is the ith node event of the nth layer, pi ni Is X ni A set of parent nodes;
the joint probability is calculated as
An emergency command decision-making system based on chemical industry park accidents comprises a memory and a processor, wherein computer instructions capable of running on the processor are stored in the memory, and the processor executes the steps of any one of the methods when running the computer instructions.
A computer readable storage medium having stored thereon computer instructions which, when executed, perform the steps of any of the methods described above.
Due to the adoption of the technical scheme, compared with the prior art, the invention has the following advantages and positive effects:
the method comprises the steps of extracting the characteristics of an emergency plan by utilizing Jieba word segmentation, N-gram characterization and KNN clustering, outputting a multi-level structured plan label library, and mainly having the following functions: firstly, the method is used as a Bayesian network analysis node, which is beneficial to reasoning and predicting the accident situation development; secondly, generating a dynamic strategy suggestion according to the prediction result of the Bayesian network;
in consideration of the dynamic evolution characteristic of the event situation development, the method can start to make a plan evaluation in the accident response stage, analyze important sensitive data such as temperature, pressure, combustibility, toxicity and harm and the like collected by a system, and qualitatively determine the current accident node by using a decision tree model;
the Bayesian network is used for analyzing the event development, so that the event development can be predicted in real time in the command process, the optimal execution strategy under the current node can be output,
the method can enhance the overall risk insights of emergency command, perform plan adaptability and satisfaction evaluation on possible unknown and known risks in the future, and perform real-time adjustment suggestions according to the defects of plan links, thereby avoiding the situation that a commander completely depends on experience and subjectivity in the command process.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention.
FIG. 1 is a flow chart of an emergency command decision method for chemical industry park accidents according to an embodiment of the present invention;
FIG. 2 is a block diagram of a multi-level feature tag library according to an embodiment of the present invention;
FIG. 3 is a diagram of a Bayesian network architecture according to an embodiment of the present invention;
fig. 4 is a structural framework diagram of the emergency command decision system for chemical industry park accidents of the invention.
Detailed Description
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following description will be made with reference to the accompanying drawings. It is obvious that the drawings in the following description are only some examples of the invention, and that for a person skilled in the art, other drawings and embodiments can be derived from them without inventive effort.
For the sake of simplicity, only the parts relevant to the present invention are schematically shown in the drawings, and they do not represent the actual structure as a product. In addition, in order to make the drawings concise and understandable, components having the same structure or function in some of the drawings are only schematically illustrated or only labeled. In this document, "a" means not only "only one of this but also a case of" more than one ".
The emergency command decision method, system and medium for chemical industry park accidents according to the present invention will be further described in detail with reference to the accompanying drawings and specific embodiments. Advantages and features of the present invention will become apparent from the following description and from the claims.
Example 1
Referring to fig. 1 to 4, the invention provides an emergency command decision method for accidents in chemical industry parks, which utilizes Jieba word segmentation, N-gram characterization, and KNN clustering generation and utilizes a multi-level feature tag library, so that the current accident situation development and prediction can be actively evaluated according to the features in command, and meanwhile, when an accident responds, the current accident progress is positioned through a decision tree, so that the event development can be predicted in real time in the command process, and an optimal execution strategy according with the current node is output.
A pre-construction is required before this embodiment can be implemented, as follows.
Preferably, a park basic support framework is constructed by acquiring data such as a text plan library, a structured plan library, an emergency resource archive and a knowledge base, and is used for providing data support for a multi-level feature tag library, a decision tree model and a Bayesian network. The subsequent label extraction can be carried out only by the text plan library and the structured plan library, and support is provided for actions such as accident analysis and the like; if emergency resource archive data exist, the rescue data can be called and displayed back to actually support rather than theoretical results when the later decision is output; only if there is a knowledge base (chemical knowledge base) that has the results of the model analysis, the model will use some parameters in the chemical.
Preferably, referring to FIG. 2, next, a multi-level feature tag library is constructed based on the campus basis support framework. The specific steps are that the single n-gram after word segmentation is vectorized to obtain the n-gram vector. And then carrying out superposition averaging on the words and the n-gram vectors in the park basic support frame to obtain a document vector. And selecting part of typical documents and corresponding document vectors thereof, and labeling the typical documents to obtain a multi-level feature label library. Wherein, the typical document is a special plan (a special plan for leakage disposal in a crude oil tank region) under a specific scene, and the selection standard can represent the plan of typical accidents of the chemical industry park and is preferably the special plan.
Wherein, the model can learn partial information of the local word sequence by utilizing the n-gram, and the method has the advantages that: 1. generating a better word vector for rare words; 2. in lexical words, word vectors for words can be constructed from character-level n-grams even if the words do not appear in the training corpus; n-grams allow the model to learn part of the information of local word order, so that several adjacent words are related by means of n-grams, which allows the model to maintain word order information during training.
Referring to fig. 2, in the present embodiment, the multi-level feature tag library selects an event type as a root node tag, that is, the multi-level feature tag library is a target for ensuring a plan. In the secondary labels, the labels in the execution stage contain important node information such as accident handling links, treatment tasks and the like, and are the key for ensuring the structured plan flow to be correct; and the information contained in the ' accident grade ' and ' application place ' labels in the secondary labels is used as the specific attribute of the plan, and after the model determines the ' type ' -execution stage ', auxiliary support is added for matching the model to the plan more fitting to the reality.
The label identification for the unknown document, namely the document data in the accident occurrence, needs to be obtained by comparing with the multi-level feature label library. The method comprises the following steps of obtaining the calculation difference between an unknown document and a typical document in a multi-level feature tag library. In particular, vectorizing unknown documents and typical documents, it can be understood that the documents are mapped into a multidimensional space, different documents exist at different points in the space, and the difference size of the unknown documents and the typical documents can be characterized by calculating the separation distance of the different documents in the space through Euclidean distance. The concrete formula is as follows
Wherein x is i And y i Attribute values for the ith dimension of example x and y, respectively. After the document is segmented into a plurality of words (the documents need to be segmented and then compared when being compared), each word represents a dimension, the value of each dimension is 0 or 1, and the document is representedIf this word is present.
And then obtaining K typical documents with the minimum difference from the unknown documents through a K proximity algorithm, and determining the labels of the unknown documents according to a minority majority-compliant voting method. The K-nearest algorithm is a supervised classification algorithm, and the working principle is as follows: and (3) a sample set exists and also becomes a training sample, the sample contains a label, each feature of the new data is compared with the corresponding feature of the data of the sample set, then the most similar classification label of the sample is extracted, k is the selected most similar data point, and the classification with the highest frequency of occurrence in k points is selected and is the classification of the new data.
Referring to fig. 1 and 4, preferably, the decision tree model also needs to be constructed in advance. And (3) performing risk and strategy evaluation at the beginning of an accident response stage, wherein the grading of the accident is usually determined according to the size of a specific characteristic value, so that the CART decision tree model is selected as a classification model to determine the class label of the current accident. The steps of the decision tree model construction are as follows.
Firstly, inputting historical detection data (five kinds of collected data, namely combustible data, temperature data, liquid level data, pressure data, toxic data, harmful data, site information and the like), manual identification information (accident qualitative rating), historical sensitive point information, historical accident model analysis results and corresponding labels as training data. Then, the optimal partition attribute a is determined by carrying out the calculation of the Gini index according to the training data * And the optimal partition attribute a is determined * As a parent node. Specifically, the formula for the calculation of the kini index is as follows:
wherein the content of the first and second substances,
wherein D is a data set of a parent node, D v Data set D as parent node at attribute a * Up value is a v OfThis subset, y is the tag type, p k Is the proportion of the kth class attribute sample in the data set D.
Then traversing the optimal partition attribute a * Each value of (1)
Then, a branch node is generated for the father node, and a sample subset D is constructed v . And if the labels in the sample subset are uniform or the decision tree model reaches the maximum depth or has no other separable attributes, taking the branch node as a leaf node, and taking the maximum label of the sample subset as the label of the branch node. Otherwise, the sample subset D is discarded v The optimal partition attribute a * And repeating the steps for the rest attributes and regenerating branch nodes. And after all the leaf nodes are generated, all the nodes reach the stop condition, and then the decision tree model is output.
Now, to illustrate more intuitively the construction of the decision tree model, we now assume that our data consists of five attributes (a, B, C, D, E), and in the previous steps, we assume we find that classification can be performed according to attribute C, and the overall Gini index can be minimized, so we determine C as the best attribute of the current round, then get the subsets D1, D2, \ 8230;, dm classified via C (the specific number depends on the label category), then the attribute to be considered is (ABDE), find the best distinguishing attribute in each subset according to the above logic, and then loop through this until the stop condition is reached.
Referring to fig. 1 and 4, the present embodiment is implemented after the above steps are completed.
Firstly, in step S1, accident data is collected in real time, and the sensitive points and the accident model are analyzed through accident real-time sensing to respectively obtain sensitive point information and an accident model analysis result. And analyzing the accident data, the sensitive point information and the accident model analysis result based on a preset multi-level characteristic label library to obtain a label, and evaluating through a preset decision tree model to obtain an event node corresponding to the current accident. The event nodes comprise labels such as scenes, types, stages, special items and the like which can be obtained according to accident data analysis; labels such as grades, hazard degrees, purposes and the like can be obtained according to decision tree classification analysis.
Then, in step S2, accident evolution and comprehensive risk analysis are performed on the event node based on the multi-level feature tag library by using a bayesian network, so as to generate a plan link and an execution strategy that conform to the current accident scenario. The planning link is a structured output result, the link is formed by a digital means according to the actual accident handling process (such as accident response, accident handling and the like), the task is a task included in each link (such as the accident handling link including site rescue, environment monitoring and the like), and the task field includes: task name, responsible person, contact, coordination, required equipment, required materials, emergency team, emergency specialist, emergency vehicle, specific action, remarks, etc.
Referring to fig. 3 and 4, in step S2, a bayesian network is used to quantify causal association between event objects, a risk inference network with accident development as a main line is established, a network structure and node positions are determined, and occurrence probability of risk nodes is analyzed, so as to calculate sensitivity of state value change of network nodes to inference results. The specific process is as follows: firstly, in step D1, a bayesian network of an accident development situation is established according to the association between the event objects, as shown in fig. 3, for example, if a link and a task have a progressive relationship, that is, an association relationship, and if the current node is in the field rescue link, the current node is correspondingly associated with an expected result of the link and subsequent operations. Then in step D2, acquiring a current event node from the decision tree, discretizing a label in the event node to determine a state value of the event node, and determining prior probability through statistics; specifically, the prior probability is calculated by the formula
Wherein X ni Is the ith node event of the nth layer, pi ni Is X ni N is an unknown quantity and is an index, for example, only three layers are shown in fig. 3, and the value of n is obtainedThe range is 1,2,3. And D3, performing accident reasoning according to the labels in the event nodes, and calculating the joint probability of each accident development scheme.
Further, in step D4, the accident development with the highest probability is determined according to the joint probability, and a corresponding structured plan, that is, a plan link and an execution strategy are generated. The joint probability is calculated as
In step D5, evaluating the obtained structured plan, and outputting the structured plan if the evaluation is passed; otherwise, deleting the structured plan and repeating the steps D2 to D5.
And finally, in step S3, repeating the steps S1 and S2 until the accident is completely solved, and ending the emergency command of the accident of the chemical industry park.
Example 2
The embodiment of the invention relates to an emergency command decision-making system based on a chemical industry park accident, which comprises a memory and a processor, wherein the memory stores computer instructions capable of running on the processor, and the processor executes the steps of the method in the embodiment 1 when executing the computer instructions.
Example 3
A computer readable storage medium having stored thereon computer instructions which when executed perform the steps of the method of any of embodiment 1.
The embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above embodiments. Even if various changes are made to the present invention, it is still within the scope of the present invention if they fall within the scope of the claims of the present invention and their equivalents.
Claims (10)
1. An emergency command decision-making method for accidents in chemical industry parks is characterized by comprising the following steps
S1: acquiring accident data in real time, analyzing a sensitive point and an accident model through accident real-time sensing to respectively obtain sensitive point information and an accident model analysis result, analyzing the accident data, the sensitive point information and the accident model analysis result based on a preset multi-level feature label library to obtain labels, and evaluating through a preset decision tree model to obtain an event node corresponding to the current accident;
s2: performing accident evolution and comprehensive risk analysis on the event node by using a Bayesian network based on the multi-level feature tag library to generate a plan link and an execution strategy which accord with the current accident scene;
s3: and repeating the steps S1 and S2 until the accident is solved.
2. The chemical industry park accident-based emergency command decision method according to claim 1, wherein a text plan library, a structured plan library, an emergency resource archive and a knowledge base are further required to be obtained before the step S1 is executed, so as to construct a park basic support framework for providing data support for the multilevel feature tag library, the decision tree model and the bayesian network.
3. The chemical industry park accident-based emergency command and decision method according to claim 2, wherein the multilevel characteristic tag library is further required to be constructed based on the park foundation support framework before the step S1 is executed, and the specific steps are as follows
A1: vectorizing the single n-gram after word segmentation to obtain a n-gram vector;
a2: performing superposition averaging on the words in the park basic support frame and the n-gram vector to obtain a document vector;
a3: and selecting part of typical documents and corresponding document vectors thereof, and labeling the typical documents to obtain the multi-level feature label library.
4. The emergency command and decision method based on chemical industry park accident according to claim 3, wherein the specific step of obtaining the label based on the analysis of the preset multi-level feature label library in the step S1 is
B1: obtaining the difference between an unknown document and the typical document in the multilevel feature label library through Euclidean formula calculation;
b2: and obtaining a plurality of typical documents with the smallest difference with the unknown document, and determining the label of the unknown document according to a voting method.
5. The chemical industry park accident-based emergency command decision method according to claim 3, wherein the decision tree model is further constructed before the step S1 is executed
C1: acquiring historical detection data, manual identification information, historical sensitive point information, historical accident model analysis results and corresponding labels as training data;
c2: performing a Gini index calculation according to the training data to determine an optimal partition attribute a * And the optimal division attribute a is set * As a parent node;
c3: traversing the optimal partition attribute a * Each value of (1)
C4: generating a branch node for the father node, and taking a plurality of values
Forming a subset of samples D v If the labels in the sample subset are uniform or the decision tree model reaches the maximum depth or has no other separable attributes, taking the branch node as a leaf node and taking the maximum label of the sample subset as the label of the branch node; otherwise, the subset of samples D is discarded v Said optimal partition attribute of a * Repeating said steps C2 to C4 for the remaining attributes;
c5: and outputting the decision tree model after all the leaf nodes are generated.
6. An emergency command decision method based on chemical industry park accident according to claim 5, wherein in the step C2, the formula of the calculation of the Kini index is as follows:
wherein the content of the first and second substances,
wherein D is the data set of the father node, y is the label type, p k Is the proportion of the kth class attribute in the data set D.
7. The chemical industry park accident-based emergency command and decision method according to claim 3, wherein the step S2 specifically comprises the following steps
D1: establishing a Bayesian network of an accident development situation according to the association among the event objects;
d2: obtaining the current event node, discretizing the label in the event node to determine the state value of the event node, and determining the prior probability through statistics;
d3: performing accident reasoning according to the labels in the event nodes, and calculating the joint probability of each accident development scheme;
d4: determining the accident development with the maximum possibility according to the joint probability, and generating a corresponding structured plan;
d5: evaluating the structured plan, and outputting the structured plan if the evaluation is passed; otherwise, deleting the structured plan and repeating the steps D2 to D5.
8. The chemical industry park accident-based emergency command and decision method according to claim 7, wherein the prior probability is calculated by the formula
Wherein X ni Is the ith node event of the nth layer, pi ni Is X ni A set of parent nodes;
the joint probability is calculated as
9. An emergency command decision system based on chemical park accidents, comprising a memory and a processor, the memory having stored thereon computer instructions executable on the processor, wherein the processor, when executing the computer instructions, performs the steps of the method according to any one of claims 1 to 8.
10. A computer readable storage medium having computer instructions stored thereon, wherein the computer instructions when executed perform the steps of the method of any one of claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210890450.7A CN115239153A (en) | 2022-07-27 | 2022-07-27 | Emergency command decision-making method, system and medium for chemical industry park accidents |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210890450.7A CN115239153A (en) | 2022-07-27 | 2022-07-27 | Emergency command decision-making method, system and medium for chemical industry park accidents |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115239153A true CN115239153A (en) | 2022-10-25 |
Family
ID=83675448
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210890450.7A Pending CN115239153A (en) | 2022-07-27 | 2022-07-27 | Emergency command decision-making method, system and medium for chemical industry park accidents |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115239153A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117520796A (en) * | 2023-11-23 | 2024-02-06 | 交通运输部规划研究院 | Knowledge-based method road map roadbed water damage evaluation method and system |
-
2022
- 2022-07-27 CN CN202210890450.7A patent/CN115239153A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117520796A (en) * | 2023-11-23 | 2024-02-06 | 交通运输部规划研究院 | Knowledge-based method road map roadbed water damage evaluation method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112612902B (en) | Knowledge graph construction method and device for power grid main equipment | |
| CN113723632A (en) | Industrial equipment fault diagnosis method based on knowledge graph | |
| Gakii et al. | A classification model for water quality analysis using decision tree | |
| CN106067094A (en) | A kind of dynamic assessment method and system | |
| Finlay et al. | Data stream mining for predicting software build outcomes using source code metrics | |
| CN111966076A (en) | Fault positioning method based on finite-state machine and graph neural network | |
| CN112507136A (en) | Knowledge-driven business operation map construction method | |
| CN109947898B (en) | Equipment fault testing method based on intellectualization | |
| CN115357726A (en) | Fault disposal plan digital model establishing method based on knowledge graph | |
| CN113268370B (en) | Root cause alarm analysis method, system, equipment and storage medium | |
| CN107103363A (en) | A kind of construction method of the software fault expert system based on LDA | |
| Wu et al. | Human resource allocation based on fuzzy data mining algorithm | |
| CN115794803B (en) | Engineering audit problem monitoring method and system based on big data AI technology | |
| CN105912773A (en) | Novel intelligent stamping process design method based on data mining technology | |
| CN114118524A (en) | Knowledge reasoning-based equipment state comprehensive analysis method | |
| CN115238197A (en) | Expert thinking model-based field business auxiliary analysis method | |
| CN115438199A (en) | Knowledge platform system based on smart city scene data middling platform technology | |
| CN115239153A (en) | Emergency command decision-making method, system and medium for chemical industry park accidents | |
| CN115358481A (en) | Early warning and identification method, system and device for enterprise ex-situ migration | |
| Sreenivasula Reddy et al. | Intuitionistic fuzzy rough sets and fruit fly algorithm for association rule mining | |
| CN114219096A (en) | Training method and device of machine learning algorithm model and storage medium | |
| Chen et al. | Extracting classification rule of software diagnosis using modified MEPA | |
| CN113837554B (en) | Multi-mode key information matching-based food safety risk identification method and system | |
| CN114881154A (en) | Natural gas station fault detection method and system based on PCA and deep forest | |
| Jasiulewicz-Kaczmarek et al. | Modeling a hierarchical structure of factors influencing exploitation policy for water distribution systems using ISM approach |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |