Real-time generation and recommendation method and system for response plan of industrial park, and medium
By generating and recommending response plans in the industrial park system in real time, combining multi-dimensional feature vectors and recommendation models, the problem of insufficient response speed in the emergency response of existing systems is solved, and more efficient emergency management is achieved.
Patent Information
- Application Number
- CN202411689655.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-25
- Publication Date
- 2025-03-14
AI Technical Summary
When an emergency risk event occurs, the existing industrial park system cannot effectively combine the historical data of the risk event, current risk information and response plan characteristics to recommend the best response plan, resulting in insufficient response speed.
A real-time generation and recommendation method for industrial park response plans is proposed. By obtaining the response plan of the current risk type from the management center system, multi-dimensional feature vectors are calculated, and the score of each plan is inputted into the trained recommendation model, and the recommendation list is finally formed and recommended to the on-site processing personnel.
Real-time and accurate recommendation of emergency response plans is achieved, the park’s emergency management efficiency is improved, and emergency response is achieved more quickly.
Smart Images
Figure CN119624109A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of industrial park safety risk management, and in particular to a method, system and medium for real-time generation and recommendation of response plans for industrial parks. Background Art
[0002] Although many industrial enterprise parks have deployed related systems such as CMS, CRM, ERP, WMS, and installed many security, fire protection, and environmental monitoring sensor equipment, when unexpected risks occur, most of these systems can only recommend response plans to on-site processing personnel through manual screening or use simple type matching methods to approximately recommend response plans, which cannot really improve the response speed of risks.
[0003] Through investigation, it was found that the current response plan matching method of the industrial park system or platform is difficult to effectively combine the historical processing data of risk events, current risk event information and the characteristics of the response plan itself for comprehensive analysis, and recommend the best response plan for dealing with current emergencies. Summary of the invention
[0004] The present invention proposes a method, system and medium for real-time generation and recommendation of response plans for industrial parks, which effectively solves the problem that traditional response plan matching methods cannot flexibly respond to emergencies.
[0005] The first aspect of the present invention proposes a method for real-time generation and recommendation of an industrial park response plan, comprising:
[0006] Obtain all response plans corresponding to the current risk type from the industrial park management center system;
[0007] Calculate the multi-dimensional feature vectors of each plan under the current risk event;
[0008] Input the multi-dimensional feature vector into the trained recommendation model and calculate the score of each plan;
[0009] A recommendation list is formed based on the plan scores, and the plans in the recommendation list are recommended to the industrial park management center system, and then recommended to the on-site processing personnel by the industrial park management center system.
[0010] As a preferred solution, the calculation of the multi-dimensional feature vectors of each plan under the current risk event includes:
[0011] Obtain the historical execution characteristics of the plan;
[0012] Obtain the similarity characteristics between the current risk event and the plan;
[0013] Obtain the attribute characteristics of the plan itself;
[0014] A multi-dimensional feature vector is obtained by integrating the historical execution characteristics of the plan, the similarity characteristics between the current risk event and the plan, and the attribute characteristics of the plan itself.
[0015] As a preferred solution, the acquisition of historical execution characteristics of the plan includes:
[0016] Calculate the average recommendation rate of the plan: average recommendation rate = the number of times the plan is included in the recommendation list for this type of event / total number of recommendations;
[0017] Calculate the average execution rate of the plan: Average execution rate = number of times the plan is pushed to on-site processing personnel / total number of pushes.
[0018] As a preferred solution, the method of obtaining the similarity characteristics between the current risk event and the plan includes:
[0019] The cosine similarity between the event feature vector and the response plan feature row vector is calculated to obtain the similarity characteristics between the current risk event and the plan; among them, the event feature vector is a 6-dimensional vector including urgency, duration, difficulty, feasibility, credibility and scope of influence; the response plan feature row vector is a 6-dimensional vector including plan level, plan response time, plan complexity, plan type, source standard and number of executors.
[0020] As a preferred solution, the method of obtaining the attribute characteristics of the plan itself includes:
[0021] Timeliness score: Timeliness score = 1-softmax (plan response time index value) / thredhold;
[0022] Safety score: Safety score = 1-softmax (plan level index value) / thredhold
[0023] Cost score: Cost score = 1-softmax (plan complexity index value) / thredhold;
[0024] Among them, the thredhold value is 0.9. When the softmax() value is greater than thredhold, the score is 0.
[0025] As a preferred solution, the training process of the recommendation model includes:
[0026] Build deep neural network models;
[0027] The training set is formed based on the RAG-based response plan acquisition and annotation technology;
[0028] Extract the multi-dimensional feature vectors in the training set, complete the training of the deep neural network model, and obtain the recommendation model.
[0029] As a preferred solution, in the deep neural network model, the output layer adopts an improved activation function, specifically:
[0030]
[0032] As a preferred solution, the RAG-based response plan acquisition and annotation technology to form a training set specifically includes:
[0033] Merge the existing response plans of various industries into an Excel spreadsheet as the knowledge base storage of the RAG model;
[0034] Select risks and use prompts to query the large model RAG to obtain the labeled data of risks with high credibility and corresponding response plans;
[0035] The obtained response plans are screened to exclude samples with incomprehensible syntax and those that do not conform to the facts. The remaining response plans are combined with the corresponding input risks and converted into matching scores to form the training subset of the DNN model. The training subset with a matching score greater than 60 is used as the positive sample of the training set.
[0036] Then randomly sample some risk and response plan data, exclude the subsets that are repeated with the positive samples, and save the rest as negative samples in the training set.
[0037] A second aspect of the present invention proposes a system including a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and execute the real-time generation of the industrial park response plan and the recommended method described in the first aspect.
[0038] The third aspect of the present invention proposes a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, are used to implement the process of real-time generation of an industrial park response plan corresponding to the recommended method described in the first aspect.
[0039] Compared with the prior art, the beneficial effect of adopting the above technical solution is that the present invention can realize real-time and accurate recommendation of emergency response plans, thereby improving the efficiency of park emergency management. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] Figure 1 The present invention is a flowchart of a method for real-time generation and recommendation of an industrial park response plan according to an embodiment of the present invention.
[0041] Figure 2 The present invention uses the response plan dataset in the large model RAG knowledge base.
[0042] Figure 3 The present invention uses the large model RAG to generate a risk response plan training set flow chart.
[0043] Figure 4 A flow chart is generated for the present invention using a response scenario training set.
[0044] Figure 5 A flowchart of a recommended plan for the industrial park management center system deployed in an embodiment of the present invention. DETAILED DESCRIPTION
[0045] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar modules or modules with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be construed as limitations on the present application. On the contrary, the embodiments of the present application include all changes, modifications and equivalents that fall within the spirit and connotation of the appended claims.
[0046] In view of the problem that the traditional response plan matching method cannot flexibly deal with emergencies, the embodiment of the present invention proposes a real-time generation and recommendation method for industrial park response plans, which can achieve accurate recommendation of response plans by extracting multi-dimensional feature vectors of risk events and plans. Figure 1 , the specific plan is as follows:
[0047] Step 1: Obtain all response plans corresponding to the current risk type from the industrial park management center system.
[0048] In practical applications, the existing response plans of various industries can be directly merged and written into an Excel spreadsheet to be stored as the knowledge base of the RAG model. This can include various published standards and factory specification manuals, as well as the scenarios, locations, conditions of occurrence, etc. that these response plans target.
[0049] Step 2: Calculate the multi-dimensional feature vectors of each plan under the current risk event.
[0050] In order to make the recommended response plan more scientific, in this embodiment, the historical execution characteristics of the plan, the similarity characteristics between the current risk event and the plan, and the attribute characteristics of the plan itself are combined to form a multi-dimensional feature vector to achieve the recommendation of the response plan.
[0051] (1) The historical execution characteristics of the plan include the average recommendation rate a and the average execution rate b of plan i. The calculation formula is as follows: a = the number of times plan i is included in the recommendation list for this type of event / the total number of recommendations, b = the number of times plan i is pushed to the on-site handling personnel / the total number of pushes
[0052] (2) The similarity feature between the current risk event and the plan includes the similarity c between the event and the plan. In this embodiment, the cosine similarity value c of the event feature vector and the response plan feature vector is calculated.
[0053] In one embodiment, both the event feature vector and the response plan feature vector are 6-dimensional vectors. The value range of the event feature vector is shown in Table 1, and the value range of the response plan feature vector is shown in Table 2.
[0054] Table 1 Event feature vector value range
[0055]
[0056] Table 2 Range of response plan feature vectors
[0057] Plan level Low(3), Medium(2), High(1) Plan response time Within 30 minutes (3), 30-60 minutes (2), More than 60 minutes (1) Complexity of the plan Execution step length < 10 (3), 10-20 (2), Execution step length > 20 = 1 Plan Type How fast the system returns to normal after treatment, Environmental and Energy (3), Security, Fire Protection, Operation (2), Other (1) Source Standard None(3), Other(2), National Standard(1) Number of executors <2 people (3), 2-3 people (2), >3 people (1)
[0058] (3) The attribute characteristics of the plan itself are mainly reflected by the scores of timeliness (d), safety (e), and cost (f). In this embodiment, the timeliness score d is related to the response time of the plan. The longer the response time of the plan, the lower the timeliness score; the safety score e is related to the level of the plan (3 / 2 / 1). The higher the level value, the lower the score; the cost score f is related to the complexity of the plan. The longer the steps, the lower the score. In one embodiment, the score of each indicator can be calculated using the following formula: score of each indicator = 1-softmax (plan response time / plan level / plan complexity indicator value) / thredhold, where thredhold = 0.9, and when the softmax value is greater than thredhold, the score is 0.
[0059] After calculating the historical execution characteristics of the plan, the similarity characteristics between the current risk event and the plan, and the plan’s own attribute characteristics, a multi-dimensional feature vector S = [a, b, c, d, e, f] is generated.
[0060] Step 3: Input the multi-dimensional feature vector into the trained recommendation model and calculate the score of each plan.
[0061] In this embodiment, the recommendation model is a pre-trained deep neural network. At the same time, this embodiment also provides a specific training process, which is as follows:
[0062] (1) Build a deep neural network model.
[0063] The deep neural network model will eventually score and recommend the response plan. In the deep neural network model, a fully connected network is used to implement the scoring module. Generally, the ReLU activation function (i.e., y=max(0,x)) helps to alleviate the gradient vanishing problem. In this embodiment, in order to ensure that the output score value is between 0 and 100, the output layer uses an improved activation function sReLU. The calculation formula is as follows:
[0064]
[0065] (2) The training set is formed based on RAG’s response plan acquisition and annotation technology.
[0066] The RAG (Retrieval-Augmented Generation) technology of the large language model combines the two methods of retrieval and generation to enhance the capabilities of the language model. In the RAG model, the model first retrieves relevant information from external data sources, and then combines this information with the knowledge within the model to generate more accurate and richer answers or texts. In view of the problems of industrial scene replication, multiple risk types, and too many complex response plans, it is too difficult to fully manually annotate the response plans.
[0067] In this embodiment, the RAG-based response plan acquisition and annotation technology is used to generate a training set, and the existing response plans of various industries are merged and written into an Excel table as a knowledge base storage of the RAG model. Figure 2 As shown in the figure, select a specific risk to form a prompt to query the large model RAG, and obtain the labeled data of the risk with higher credibility and the corresponding response plan. Figure 3 A prompt example is shown, where context is the content retrieved from the knowledge base by the RAG model, and question is the risk to be inquired. The obtained response plans are screened, and samples with incorrect syntax and inconsistent with the facts are excluded. The remaining response plans are combined with the corresponding input risks and converted into matching scores (0-100 points) to form a training subset for the deep neural network model. This subset has a higher accuracy and will be used as positive samples in the training set, with a score greater than 60 points. Then some risk and response plan data are randomly sampled, and those that are repeated in the positive samples are excluded. The rest are saved as negative samples in the training set, such as Figure 4 shown.
[0068] Finally, the multi-dimensional feature vectors in the training set are extracted, and the deep neural network model is trained. The trained deep neural network model is saved, and the version with the best performance in the test set is selected and used as the recommended model for the intelligent response plan.
[0069] like Figure 5 The figure shows that an industrial park management center system is deployed in the industrial park. The industrial park management center system obtains the historical execution information, current event information and response plan information of each response plan corresponding to the current risk type through the system acquisition layer. The response plan generation and recommendation method proposed in this embodiment is used to recommend a list of response plans ranked from high to low scores to the processing personnel in real time. In terms of deep learning model training, the large model RAG technology is used to generate reliable and accurate risk and response plan matching results, which are converted into scores after manual filtering and annotation for use as a training set for deep learning.
[0070] An embodiment of the present invention further proposes a system, including a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and execute the aforementioned industrial park response plan in real time and corresponding to the recommended method.
[0071] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer readable medium, and the computer program includes a program code for executing the method shown in the flowchart.
[0072] It should be noted that the computer-readable medium shown in the embodiment of the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, - but not limited to - an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by an instruction execution system, device or device or used in combination with it. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.
[0073] The flowchart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present application. Wherein, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0074] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.
[0075] As another aspect, the present application also provides a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method for real-time generation and recommendation of industrial park response plans described in the above embodiment.
[0076] As another aspect, the present application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device implements the method for real-time generation and recommendation of industrial park response plans described in the above embodiment.
[0077] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.
[0078] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software or by combining software with necessary hardware. Therefore, the technical solution according to the implementation methods of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation methods of the present application.
[0079] For those skilled in the art, the specific meanings of the above terms in the present invention can be understood in specific situations; the drawings in the embodiments are used to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.
[0080] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.
Claims
1. A method for real-time generation and recommendation of response plans for industrial parks, characterized in that: include: Obtain all response plans corresponding to the current risk type from the industrial park management center system; Calculate the multi-dimensional feature vectors of each plan under the current risk event; Input the multi-dimensional feature vector into the trained recommendation model and calculate the score of each plan; A recommendation list is formed based on the plan scores, and the plans in the recommendation list are recommended to the industrial park management center system, and then recommended to the on-site processing personnel by the industrial park management center system.
2. The method for real-time generation and recommendation of industrial park response plans according to claim 1 is characterized in that: The calculation of the multi-dimensional feature vectors of each plan under the current risk event includes: Obtain the historical execution characteristics of the plan; Obtain the similarity characteristics between the current risk event and the plan; Obtain the attribute characteristics of the plan itself; A multi-dimensional feature vector is obtained by integrating the historical execution characteristics of the plan, the similarity characteristics between the current risk event and the plan, and the attribute characteristics of the plan itself.
3. The method for real-time generation and recommendation of industrial park response plans according to claim 2 is characterized in that: The characteristics of obtaining the historical execution status of the plan include: Calculate the average recommendation rate of the plan: average recommendation rate = the number of times the plan is included in the recommendation list for this type of event / total number of recommendations; Calculate the average execution rate of the plan: Average execution rate = number of times the plan is pushed to on-site processing personnel / total number of pushes.
4. The method for real-time generation and recommendation of industrial park response plans according to claim 2 or 3, characterized in that: The similarity features of obtaining the current risk event and the plan include: The cosine similarity between the event feature vector and the response plan feature row vector is calculated to obtain the similarity characteristics between the current risk event and the plan; among them, the event feature vector is a 6-dimensional vector including urgency, duration, difficulty, feasibility, credibility and scope of influence; the response plan feature row vector is a 6-dimensional vector including plan level, plan response time, plan complexity, plan type, source standard and number of executors.
5. The method for real-time generation and recommendation of industrial park response plans according to claim 2 or 3, characterized in that: The acquisition of the plan's own attribute characteristics includes: Timeliness score: Timeliness score = 1-softmax (plan response time index value) / thredhold; Safety score: Safety score = 1-softmax (plan level index value) / thredhold Cost score: Cost score = 1-softmax (plan complexity index value) / thredhold; Among them, the thredhold value is 0.
9. When the softmax() value is greater than thredhold, the score is 0.
6. The method for real-time generation and recommendation of industrial park response plans according to claim 1 is characterized in that: The training process of the recommendation model includes: Build deep neural network models; The training set is formed based on the RAG-based response plan acquisition and annotation technology; Extract the multi-dimensional feature vectors in the training set, complete the training of the deep neural network model, and obtain the recommendation model.
7. The method for real-time generation and recommendation of industrial park response plans according to claim 6 is characterized in that: The method for real-time generation and recommendation of industrial park response plans according to claim 1 is characterized in that, in the deep neural network model, the output layer adopts an improved activation function, specifically: 。 8. The method for real-time generation and recommendation of industrial park response plans according to claim 6 or 7, characterized in that: The method for real-time generation and recommendation of response plans for industrial parks according to claim 1 is characterized in that the RAG-based response plan acquisition and annotation technology to form a training set specifically includes: Merge the existing response plans of various industries into an Excel spreadsheet as the knowledge base storage of the RAG model; Select risks and use prompts to query the large model RAG to obtain the labeled data of risks with high credibility and corresponding response plans; The obtained response plans are screened to exclude samples with incomprehensible syntax and those that do not conform to the facts. The remaining response plans are combined with the corresponding input risks and converted into matching scores to form the training subset of the DNN model. The training subset with a matching score greater than 60 is used as the positive sample of the training set. Then randomly sample some risk and response plan data, exclude the subsets that are repeated with the positive samples, and save the rest as negative samples in the training set.
9. A system, characterized in that: It comprises a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executes the real-time generation of an industrial park response plan and the corresponding recommendation method as described in any one of claims 1 to 8.
10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the program instructions are executed by the processor, they are used to implement the process corresponding to the real-time generation of the industrial park response plan described in any one of claims 1 to 8 and the recommended method.