A multi-stage emergency response scheduling method for heating systems with large models

Abstract

The present invention discloses a multi-stage emergency response scheduling method for a heating system containing a large model, comprising: forming a large model for emergency response scheduling of the heating system; in the emergency prevention and preparation stage, using the large model to construct a multi-type emergency event question and answer database for the heating system, performing intelligent question and answer of emergency knowledge, emergency event cause analysis, risk hidden danger query and recommendation of emergency response scheduling measures; in the emergency response stage, guiding the large model to output the optimal feature combination representing each type of emergency event, and using machine learning to identify each type of emergency event; using the large model to evaluate the heating loss caused by the emergency event, setting a multi-level emergency response scheduling strategy, and obtaining the optimal emergency response scheduling plan; in the emergency response aftermath stage, using the large model to analyze the emergency response scheduling process, establishing an emergency response scheduling plan library, and using an intelligent agent to simulate the plan, identify potential loopholes and strengthen the emergency response scheduling process.

Description

Technical Field

[0001] The present invention belongs to the technical field of heating systems, and in particular relates to a multi-stage emergency response scheduling method for a heating system containing a large model. Background Art

[0002] With the advancement of urbanization, urban heating services have continued to develop, increasing the difficulty of operation and management. Frequent emergency events in heating systems have also caused serious losses to heating companies. Therefore, timely response and dispatch after an emergency event is particularly important. Emergency dispatch for heating systems involves the timely and accurate development of the optimal emergency response plan based on the type of emergency, such as system failures or extreme weather events, to control and prevent the escalation of the situation and minimize losses and impacts.

[0003] In recent years, the emergence of artificial intelligence technologies such as natural language big models and machine learning has enabled sectors like power and industry to gradually acquire the capabilities of knowledge learning, comprehensive perception, and autonomous decision-making. Research on how to effectively empower emergency response and dispatch scenarios in heating systems based on these technologies, making the emergency prevention, preparation, response, and aftermath phases of heating systems more intelligent and efficient, and thus improving the emergency response and dispatch management of heating systems, is an urgent issue.

[0004] Based on the above technical problems, it is necessary to design a new multi-stage emergency response scheduling method for heating systems containing large models. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the shortcomings of the existing technology and provide a multi-stage emergency response and scheduling method for a heating system containing a large model, integrating large models, machine learning and intelligent agent technologies, giving full play to the advantages of different technologies, and performing emergency knowledge understanding and generation through the large language model of the heating system formed by training, emergency event identification through large models and machine learning, heating loss assessment through large models, and outputting the optimal emergency response scheduling plan through multi-agent emergency response scheduling decision-making, etc., so that each stage of the heating system emergency is more intelligent and scientific, and the emergency response scheduling management level of the heating system is comprehensively improved.

[0006] In order to solve the above technical problems, the technical solution of the present invention is:

[0007] The present invention provides a multi-stage emergency response scheduling method for a heating system containing a large model, including: an emergency prevention and preparation stage, an emergency response stage, and an emergency response and aftermath stage;

[0008] The emergency preparedness phase includes:

[0009] A multi-type emergency event question-and-answer database for the heating system is constructed based on a large model. The database extracts key event information from the heating system's emergency response and dispatch data and generates targeted questions and answers. These targeted questions and answers include intelligent emergency knowledge questions and answers. The database provides emergency event cause analysis, risk hazard query, and recommendation of emergency response and dispatch measures.

[0010] The emergency response phase includes:

[0011] Use prompt word engineering to guide the large model to output the optimal feature combination to characterize various types of emergency events, and identify various types of emergency events;

[0012] Evaluate the heat loss caused by the identified emergency events, and set emergency response scheduling decision-making tasks based on the goals of quickly reconstructing the hydraulic balance of the heating network and reducing heat loss;

[0013] Multiple emergency response dispatch agents output emergency response dispatch plans based on emergency response dispatch decision-making tasks;

[0014] Each agent judges the emergency response and dispatch plans output by other agents;

[0015] Obtain the optimal emergency response scheduling plan based on the judgment and analysis of the emergency response collaborative agent and each agent;

[0016] The emergency response and aftermath phase includes:

[0017] The emergency response and scheduling process is analyzed and summarized using a large model, and emergency response and scheduling plans are simulated using intelligent agents to strengthen the emergency response and scheduling process.

[0018] Furthermore, during the emergency response phase, a feature combination workflow is set up to output various types of emergency events. The feature combination workflow includes uploading the emergency response scheduling dataset, describing the emergency event, describing the task requirements for feature extraction using machine learning, selecting the feature combination mechanism, and outputting the optimal feature combination.

[0019] Use prompt word engineering to guide the large model to sequentially implement the workflow of output feature combination, and output the optimal feature combination to represent various types of emergency events;

[0020] The optimal feature combination of each type of emergency event is input into the preset machine learning model for training and learning, and a heating emergency event recognition model is established to identify the types of emergency events occurring in the heating system.

[0021] Furthermore, the construction of a heating emergency event recognition model includes: setting up a two-layer machine learning model to extract features from the emergency event dataset, where the input end of the first-layer machine learning model is the original emergency response scheduling dataset, and the features of the original emergency response scheduling dataset are abstracted through unsupervised learning;

[0022] The second-layer machine learning model uses local feature extraction capabilities to perform deep extraction of various data features;

[0023] The cross-validation method is used to evaluate different feature combinations, and the feature combinations are optimized and iterated based on the evaluation results to output the optimal feature combination for identifying emergency events in the heating system.

[0024] Furthermore, the heat loss assessment step includes: constructing an economic loss calculation model to calculate the economic losses caused by the emergency event based on the data identified in the emergency event.

[0025] Furthermore, the data in emergency events include the number of affected heat users, heating area, missing values of heat load, and emergency repair time.

[0026] Furthermore, the method for constructing an economic loss calculation model includes taking the number of affected heat users, the duration of heating interruption, the unit price of heating fee, etc. as independent variables and the economic loss value as the dependent variable based on the economic loss data of historical emergency events and the currently determined loss factors, training the model to learn the relationship between loss factors and economic losses, and constructing an economic loss calculation model.

[0027] Furthermore, the hydraulic balance conditions of the heating network are quickly reconstructed based on the hydraulic model of the heating network, including building the hydraulic model of the heating network based on the topological structure data of the heating network, fluid mechanics principles and historical operation data, and calculating the impact of emergency events on the hydraulic balance of the heating network.

[0028] Furthermore, a multi-level emergency response scheduling strategy is set up: in the first layer, multiple emergency response scheduling agents are set up to analyze and reason about the emergency event information occurring in the heating system, and output their own emergency response scheduling plans; in the second layer, each emergency response scheduling agent exchanges information, judges, discusses and puts forward opinions on the emergency response scheduling plans output by other emergency response scheduling agents; in the third layer, the set emergency response collaborative agent integrates the current emergency event problem and the inference of each emergency response scheduling agent for judgment and analysis, and obtains the optimal emergency response scheduling plan.

[0029] Furthermore, the emergency response dispatch decision-making task is expressed as: ;

[0030] Where: D To generate emergency response scheduling plan; A large-scale model for emergency response and dispatching of heating systems; To quickly reconstruct the hydraulic balance of the heating network and reduce heat loss; Provides current heating system operation information; Provide emergency response and dispatch knowledge for heating systems.

[0031] Furthermore, the invention also includes extracting key event information from the emergency response and dispatching data of the heating system, generating targeted questions and answers, and automatically constructing a multi-type emergency event question and answer database for the heating system using a large model. The beneficial effects of the invention are:

[0032] (1) The present invention forms a large model of emergency response and scheduling for the heating system, and has the ability to understand the semantics of emergency response and scheduling knowledge, reason about knowledge, and generate content for emergency response and scheduling for the heating system, so that it can deeply understand the business knowledge and emergency scenarios of the heating system, and establish a learning and reasoning foundation for emergency events and emergency response and scheduling knowledge for the subsequent emergency prevention and preparation stage, emergency response stage, and emergency response and aftermath stage. It can improve the learning and application efficiency of emergency knowledge of heating business personnel, make all emergency stages of the heating system more intelligent and scientific, and comprehensively improve the emergency response and scheduling management level of the heating system; in the emergency prevention and preparation stage, the present invention uses a large model to construct a multi-type emergency event question-and-answer database for the heating system, so that heating business personnel can quickly obtain relevant emergency knowledge in advance, conduct risk and hidden danger inquiries, and analyze the causes of emergency events, etc., deepen the understanding of emergency events by heating business personnel, thoroughly investigate potential risk points, and prevent the occurrence of emergency events in advance;

[0033] (2) In the emergency response stage, the present invention uses prompt word engineering to guide the large model to output the optimal feature combination, and combines it with a machine learning model to identify emergency events, which can significantly improve the accuracy and efficiency of emergency event identification, quickly and accurately identify various types of emergency events, and win valuable time for subsequent emergency disposal scheduling; the large model's assessment of heating losses is more comprehensive and accurate, and it can formulate more targeted strategies to reduce losses, minimizing the impact of emergency events on the heating system and users; and, through the emergency disposal scheduling of multi-level intelligent bodies, it can better adapt to the complex emergency event response environment of the heating system, understand and analyze the current emergency event environment and related emergency issues, and propose an explainable emergency disposal scheduling plan for the heating system, which is convenient for heating business personnel to quickly respond to and handle emergency events, and better achieve the goal of quickly reconstructing the hydraulic balance of the heating network and reducing heating losses;

[0034] (3) In the emergency response and aftermath stage, the present invention uses a large model to analyze and summarize the emergency response scheduling process of emergency events in the heating system, which helps to accumulate valuable experience and integrate these experiences into the heating system safety emergency response scheduling plan library, which can continuously improve the emergency plan to make it more in line with the actual situation and improve the practicality and operability of the plan; and, by using intelligent agents to simulate emergency response scheduling plans, potential loopholes can be discovered in advance. Through the analysis of simulation results, the emergency response scheduling process can be strengthened in a targeted manner, and the response speed and processing capabilities of the heating system in dealing with similar events in the future can be improved, so as to achieve continuous optimization of the emergency response process and ensure the safe and stable operation of the heating system. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the specific embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0036] FIG1 is a flow chart of an embodiment of the present invention;

[0037] FIG2 is a schematic diagram showing the implementation principle of a multi-level emergency response scheduling strategy according to an embodiment of the present invention. DETAILED DESCRIPTION

[0038] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0039] like Figure 1 As shown, one embodiment of the present invention provides a multi-stage emergency response scheduling method for a heating system including a large model, which includes:

[0040] Based on the heating system text data, a large language model of the heating system is formed through pre-training of the base large model. The large language model of the heating system is then fine-tuned using the heating system emergency response and scheduling data to form a large model of the heating system emergency response and scheduling.

[0041] During the emergency prevention and preparation phase, by extracting key event information from the heating system's emergency response and dispatch data, generating targeted questions and answers, and using a large model to automatically build a multi-type emergency event question-and-answer database for the heating system, the system supports heating business personnel in intelligent emergency knowledge question-and-answer sessions, emergency event cause analysis, risk hazard query, and recommendation of emergency response and dispatch measures.

[0042] During the emergency response phase, we use prompt word engineering to guide the large model to output the optimal feature combination to characterize various types of emergency events, and use machine learning models to identify various types of emergency events.

[0043] After using a large model to evaluate the heat loss caused by the identified emergency events, an emergency response scheduling decision-making task is set with the goal of quickly reconstructing the hydraulic balance of the heating network and reducing heat loss. After analysis and reasoning by different intelligent agents, each intelligent agent outputs its own emergency response scheduling plan. Each intelligent agent then judges the emergency response scheduling plans of other intelligent agents. Finally, the optimal emergency response scheduling plan is obtained through judgment and analysis by the emergency response collaborative intelligent agent.

[0044] During the emergency response and aftermath stage, a large model is used to analyze and summarize the emergency response and scheduling process of emergency events in the heating system, establish a heating system safety emergency response and scheduling plan library, and use intelligent agents to simulate emergency response and scheduling plans to identify potential loopholes and strengthen the emergency response and scheduling process.

[0045] It should be noted that during the emergency response and aftermath phase, the heating system emergency response and dispatching large model is used to analyze and summarize the emergency response and dispatching process of the heating system emergency event, generating a detailed analysis including the detailed course of the event, the effectiveness evaluation of various response and dispatching measures, whether resource utilization is reasonable, and whether personnel dispatching is efficient. Based on the analysis report of the large model, combined with past successful emergency response cases and industry standards and specifications, the large model is used to generate a general emergency response and dispatching plan template. For example, for a pipeline rupture incident, the plan template generated by the large model may include:

[0046] Emergency response initiated: After receiving a pipeline rupture alarm, confirm the location and general situation of the incident;

[0047] Emergency measures: Immediately close the valves upstream and downstream of the ruptured pipeline to prevent further leakage of hot water; and perform heat network disconnection and network cutting operations to compensate for heat loss;

[0048] Emergency repair preparation: dispatch emergency repair personnel and necessary materials (such as pipes of corresponding diameters, welding equipment, etc.) to the site;

[0049] On-site emergency repair: Repair pipelines according to standard procedures;

[0050] Restore heating: After the repair is completed, conduct pressure testing and trial operation to ensure there are no leaks, and then gradually restore heating.

[0051] Use intelligent agents to simulate emergency response and dispatch plans, identify potential vulnerabilities, and strengthen emergency response and dispatch processes, including:

[0052] An intelligent agent is set up to simulate the operating environment of the heating system, generate an emergency event scenario, and execute the emergency response dispatch plan. For example, if a heating pipe ruptures in a certain area, the agent executes the response according to the emergency response dispatch plan. Simultaneously, various data during the simulation are tracked, such as whether valve connections are successful and whether heating system parameters are gradually returning to normal. During the simulation, potential vulnerabilities are identified by comparing the actual simulation results with the expected results of the plan. For example, according to the plan, after closing the valve, the pressure in the pipeline should drop to a safe range within a preset time. However, the simulation results show that the pressure drops slowly, exceeding the expected time, indicating a possible vulnerability in the plan's valve operation or pressure control. Simultaneously, in-depth analysis of the simulation data is conducted to identify potential systemic issues. For example, by analyzing the resource allocation situation in multiple simulations, insufficient emergency repair resources in certain areas were found, resulting in prolonged emergency response time, which is also a vulnerability that needs to be addressed. Based on the potential vulnerabilities identified by the intelligent agent simulation, improvement suggestions are generated using the large model and fed back into the emergency response and scheduling plan library to update and improve the corresponding plans. At the same time, the intelligent agent simulation is re-run regularly to verify the effectiveness of the improved plans, continuously strengthen the emergency response and scheduling process, and improve the heating system's ability to respond to emergencies.

[0053] In this embodiment, a large language model of the heating system is formed by pre-training a large base model based on the heating system text data. The large language model of the heating system is fine-tuned using the heating system emergency response scheduling data to form a large model of the heating system emergency response scheduling, including:

[0054] Obtain heating system text data, including: heating system design documents, operation manuals, historical operation data, and industry standards and specifications;

[0055] Input the heating system text data into a pre-selected base model for pre-training, learn the language patterns, terminology, and knowledge related to the heating system, and continuously adjust the model parameters to form a heating system language model that can understand heating system professional knowledge, reason about knowledge, and generate content;

[0056] Obtain emergency response and dispatch data for the heating system in the event of equipment failure, extreme weather impact, or heating network failure, and fine-tune the large language model of the heating system. During the fine-tuning process, while maintaining the original knowledge of the heating system, learn specific knowledge related to emergency response and dispatch to form a large model of emergency response and dispatch for the heating system.

[0057] In this embodiment, during the emergency prevention and preparation phase, key event information from the heating system emergency response and dispatch data is extracted, targeted questions and answers are generated, and a large model is used to automatically construct a multi-type emergency event question and answer database for the heating system, including:

[0058] During the emergency prevention and preparation phase, the analytical capabilities of the large-scale model for emergency response and scheduling of heating systems are utilized to automatically extract key event information from emergency response and scheduling data of heating systems, including emergency event named entity recognition and event description extraction.

[0059] Based on the key event information of different emergency events, the emergency response and dispatch model of the heating system is used to generate questions and answers related to the emergency events.

[0060] The generated questions and answers are sorted and stored in a structured manner to build a Q&A database for multiple types of emergency events in the heating system.

[0061] The text parsing capabilities of the big model are used to identify named entities in emergency events, such as event type, names of equipment involved, geographic location, time, etc. The big model is also used to extract key event description information from the emergency event text, such as the process of the event, the scope of impact, and measures taken. Based on the extracted key event information, the big model is again used to generate relevant questions and answers. For example, for pipeline rupture events, questions such as "Where do pipeline ruptures usually occur?" and "How can you quickly locate the rupture point after a pipeline ruptures?" can be generated. In response to the question "How can you quickly locate the rupture point after a pipeline ruptures?", the big model generates the answer "You can use the pressure monitoring system to compare the normal pressure distribution with the current pressure data. Areas with obvious pressure drops may be the location of the rupture point. Thermal imaging technology can also be used to detect areas with abnormal temperature on the pipeline surface to locate the rupture point."

[0062] It should be noted that in the heating system, common emergency events include:

[0063] Pipeline rupture: This is usually accompanied by a sudden drop in pressure, an abnormal increase in flow, and a drop in temperature in the area near the rupture point. For example, at 2 a.m. one winter day, a main heating network in the city center ruptured, affecting the heating supply to several surrounding communities.

[0064] Equipment failure: For example, a circulation pump failure can significantly reduce the circulation flow of the heating system, causing the heating temperature to drop. A heat exchanger failure can reduce the heat exchange efficiency, resulting in high supply water temperature but also high return water temperature. For example, a circulation pump in a heat exchange station suddenly failed during operation, causing the heating temperature in the area under the station's responsibility to drop by 5°C in a short period of time.

[0065] Heat source interruption: This may be caused by a failure in the thermal power plant, fuel supply problems, etc. At this time, the water supply temperature of the entire heating system will drop rapidly, and the pressure will also drop accordingly. For example, due to a maintenance error in the thermal power plant equipment, the heat source was interrupted for 2 hours, and the temperature of the heating system in the entire city generally dropped. In this embodiment, emergency event entity recognition includes: the time, location, equipment name, emergency event type, severity, and corresponding emergency response scheduling knowledge;

[0066] Emergency event types include: heat source failure, heat network failure, pump and valve failure, and extreme weather conditions that cause heat sources to fail to be handled normally according to the scheduling plan;

[0067] Emergency response and dispatch knowledge includes: analyzing the causes of different emergency events, the resulting heat losses, economic losses, and emergency response and dispatch plans; emergency response and dispatch plans include: adjusting heat source load, starting backup equipment, switching heating network lines, valve disconnection, and regulating pumps and valves at heating stations;

[0068] Event description extraction includes: locating key descriptive sentences by analyzing sentence structure and semantic relationships, and extracting descriptions of the emergency event process, impact, and handling measures.

[0069] In this embodiment, during the emergency response phase, prompt word engineering is used to guide the large model to output the optimal feature combination representing each type of emergency event, and a machine learning model is used to identify each type of emergency event, including:

[0070] During the emergency response phase, a feature combination workflow is set up to output various types of emergency events, including uploading emergency response dispatch data sets, emergency event description, task requirement description for feature extraction using machine learning, selection of feature combination mechanism, and output of optimal feature combination.

[0071] Use prompt word engineering to guide the large model to sequentially implement the workflow of output feature combination, and output the optimal feature combination to represent various types of emergency events;

[0072] The optimal feature combination of each type of emergency event is input into the preset machine learning model for training and learning, and a heating emergency event recognition model is established to identify the types of emergency events occurring in the heating system.

[0073] The task requirement for feature extraction using machine learning is to extract key features that can accurately identify different types of emergency events from the large and complex operating data of the heating system. The extracted features must have the following characteristics:

[0074] High discrimination: The model can clearly distinguish different types of emergency events. For example, the pressure drop characteristics of a pipeline rupture and the flow abnormality characteristics of an equipment failure should be significantly different, so that the model can accurately identify the event type.

[0075] Stability: The characteristic is not affected by the normal daily fluctuations of the heating system, ensuring that it can reliably indicate the occurrence of emergency events under different operating conditions. For example, the normal flow changes of the heating system during peak hours in the morning and evening should not interfere with the judgment of flow anomalies caused by equipment failure;

[0076] Computational efficiency: Considering the real-time requirements of emergency response, the feature calculation process should be simple and efficient, and be able to complete the calculation in a short time to provide support for rapid decision-making.

[0077] In this embodiment, the task requirements for feature extraction using machine learning include: setting up a two-layer machine learning model to extract features from the emergency response scheduling dataset. The input end of the first-layer machine learning model is the original emergency response scheduling dataset, which is used to abstract features through unsupervised learning. The second-layer machine learning model uses local feature extraction capabilities to deeply extract features from each data.

[0078] The feature combination selection mechanism includes: exhaustive search mechanism, greedy search mechanism, and heuristic algorithm mechanism;

[0079] Outputting the optimal feature combination includes: using a cross-validation method to evaluate different feature combinations, and optimizing and iterating the feature combination according to the evaluation results, and outputting the optimal feature combination suitable for identifying emergency events in the heating system.

[0080] It should be noted that the first-layer machine learning model uses an autoencoder as an unsupervised learning model for feature abstraction. The autoencoder consists of an encoder and a decoder. The encoder compresses the original high-dimensional emergency response dispatch dataset into a low-dimensional feature representation, while the decoder reconstructs these features back into the original data space. The autoencoder is trained to reconstruct the input data as accurately as possible, thereby learning the underlying patterns and features in the data. The second-layer machine learning model uses a convolutional neural network (CNN) for deep feature extraction. Because CNNs have strong local feature extraction capabilities, they are well-suited to processing feature data abstracted by the first-layer autoencoder.

[0081] You can choose to use the genetic algorithm in the heuristic algorithm mechanism to perform feature combination. The genetic algorithm simulates the biological evolution process and gradually optimizes the feature combination through operations such as selection, crossover, and mutation. The specific steps are as follows:

[0082] Encoding: Encode each feature combination as a chromosome, for example, using binary encoding, where 1 indicates that the feature is selected and 0 indicates that it is not selected;

[0083] Initialize the population: randomly generate a certain number of chromosomes (feature combinations) to form the initial population;

[0084] Fitness evaluation: Use a fitness function based on a machine learning model (such as a decision tree classifier) to evaluate the quality of each chromosome. The fitness value is the classification accuracy of the model under the feature combination on the training data.

[0085] Selection: Roulette wheel selection method is used to select chromosomes according to their fitness. The higher the fitness, the greater the probability of being selected, thus retaining excellent feature combinations.

[0086] Crossover: Perform a single-point crossover operation on the selected chromosome with a certain crossover probability to generate a new chromosome and introduce new feature combination possibilities;

[0087] Mutation: flipping certain gene positions in chromosomes with a lower mutation probability to increase population diversity;

[0088] Iterative optimization: Repeat the above fitness evaluation, selection, crossover, and mutation operations until the termination condition is met (such as reaching the maximum number of iterations or the fitness no longer improves).

[0089] In this embodiment, a large model is used to evaluate the heat loss caused by the identified emergency event, including:

[0090] The large model of emergency response and dispatch of the heating system is used to obtain the heating operation data before and after the emergency event according to the identified emergency events, and calculate the number of affected heat users and heating area, missing heat load value, and economic loss value.

[0091] It should be noted that the assessment of heat loss specifically includes:

[0092] 1) The large model can process and analyze large amounts of heat network topology data and historical operating data, helping to identify patterns, trends, and correlations in the data, providing deeper insights for building a heat network hydraulic model. For example, by analyzing historical operating data, the large model can discover how the hydraulic characteristics of the heat network change over time and seasons. Secondly, using the heat network topology data (pipeline connections, pipe diameters, lengths, etc.), fluid mechanics principles, and historical operating data, a heat network hydraulic model is constructed (this model can simulate the hydraulic characteristics of the heat network under different operating conditions, such as flow distribution and pressure changes). After this, the large model's optimization algorithm is used to optimize the model's parameters to improve the model's accuracy and reliability. By simulating different operating conditions, the large model can adjust parameters such as pipe diameters and lengths in the model to better simulate the hydraulic characteristics of the actual heat network.

[0093] 2) Simulate the impact of emergency events on the network's hydraulic balance within the network's hydraulic model. Determine the extent of the impact, flow rate changes, and pressure fluctuations. For example, if a pipeline ruptures, the model can quickly calculate the sudden drop in flow and pressure near the rupture point and downstream.

[0094] 3) Loss factor analysis: Use the big model to comprehensively analyze the type of emergency event, the number of affected heat users, the heating area, missing values of heat load, and the economic data of the heating system (such as the unit price of heat fee, equipment maintenance cost, user compensation standard, etc.). Through in-depth mining of these data, the big model can more comprehensively determine the components of heat loss. For example, in the case of a pipeline rupture, the economic losses include emergency repair costs, heat fee losses due to heat interruption, and possible compensation costs to users; in addition, based on the analysis of loss factors, the big model can evaluate and warn of the economic loss risks that may be brought about by different types of emergency events. For example, based on the current operating status of the heating network and the development trend of emergency events, the big model can predict possible heat fee losses, equipment maintenance costs, etc. in advance, and provide decision support for emergency response;

[0095] 4) Loss calculation model construction: The large model leverages its powerful learning capabilities to build an economic loss calculation model based on historical emergency economic loss data and currently identified loss factors. It uses the number of affected heat users, duration of heat interruption, and unit price of heat as independent variables, and the economic loss value as the dependent variable. The model is trained to learn the relationship between these factors and economic losses.

[0096] 5) Loss Assessment: When an emergency occurs, the large model can obtain relevant data in real time and use the constructed economic loss calculation model to input relevant data of the current emergency (such as the number of affected heat users, heating area, missing heat load values, emergency repair time, etc.) to quickly calculate the economic loss value caused by this emergency. At the same time, the large model can also promptly feedback the calculation results to relevant personnel, providing real-time support for emergency dispatch plans.

[0097] In this embodiment, with the goal of rapidly reconstructing the hydraulic balance of the heating network and reducing heat loss, an emergency response scheduling decision-making task is set. Different intelligent agents perform analysis and reasoning, then output their own emergency response scheduling plans. Each intelligent agent then judges the emergency response scheduling plans of other intelligent agents. Finally, the optimal emergency response scheduling plan is obtained through judgment and analysis by the emergency response collaborative intelligent agent, including:

[0098] Set up emergency response and scheduling decision-making tasks: The heating system emergency response and scheduling large model aims to quickly reconstruct the hydraulic balance of the heating network and reduce heating losses. Based on the current heating system operation information and the knowledge of heating system emergency response and scheduling, it makes judgments and generates the optimal emergency response and scheduling plan.

[0099] A multi-level emergency response and scheduling strategy is set up: in the first layer, multiple emergency response and scheduling intelligent agents are set up to analyze and reason about the emergency event information occurring in the heating system, and output their own emergency response and scheduling plans; in the second layer, each emergency response and scheduling intelligent agent exchanges information, and judges, discusses and puts forward opinions on the emergency response and scheduling plans output by other emergency response and scheduling intelligent agents; in the third layer, the set emergency response collaborative intelligent agent integrates the current emergency event problem and the inference of each emergency response and scheduling intelligent agent for judgment and analysis, and obtains the optimal emergency response and scheduling plan.

[0100] As shown in FIG2 , in one embodiment of the present invention, the specific steps for implementing the multi-level emergency response scheduling strategy include:

[0101] 1) In the first layer, the ChatGPT, ChatGLM, and Gemini large models can be integrated into the emergency response and scheduling agent. The large model drives the agent to generate multiple emergency response and scheduling agents (Agent I, Agent II, and Agent III). Based on the emergency events occurring in the current heating system, they perceive the operating parameters of the heating system, collect valuable information for analysis and reasoning, and then output their respective emergency response and scheduling plans.

[0102] 2) In the second layer, based on the emergency response and dispatch plans of Agents I, II, and III, each agent reviews the emergency response and dispatch plans of the other two agents through information exchange, makes judgments and discusses them, and puts forward its own opinions;

[0103] 3) In the third layer, the emergency response collaborative agent integrates the current emergency event problem and the inferences of Agents I, II, and III. After judgment and analysis, it obtains the optimal emergency response scheduling plan that meets the current emergency events of the heating system, providing strong decision-making support for heating business personnel.

[0104] Assume that a serious pipe rupture emergency occurs in the heating system during the winter peak period. The following is a specific example of generating an emergency response scheduling plan based on a multi-level emergency response scheduling strategy:

[0105] 1) Large model drives the agent to generate a plan and initialize the agent:

[0106] The ChatGPT, ChatGLM, and Gemini large models are integrated into the emergency response dispatch agent to generate Agent I, Agent II, and Agent III. After the system identifies a pipeline rupture event, the emergency response dispatch agent begins to work.

[0107] Data Collection and Analysis: Agent I, through its network of sensors connected to the heating system, detected operational parameters such as a sudden drop in temperature, a rapid decrease in pressure, and an abnormal increase in flow near the pipe rupture point. It also collected data on the number of heat users in the area, covering a heating area of 300,000 square meters. The primary water temperature in the pipe network was 80°C, the return water temperature was 50°C, and the ruptured pipe had a diameter of 500 mm, making it a critical trunk pipeline. Based on this information, large-scale model analysis and reasoning were used to generate Emergency Response Plan I: First, valves upstream and downstream of the ruptured pipe were immediately closed to prevent further hot water leakage. The valve closure operation was expected to be completed within one hour. A nearby repair team was mobilized to the site, carrying a 500mm diameter spare pipe and related welding equipment. The team was expected to arrive within two hours. Simultaneously, a nearby backup heat source was activated, increasing its output to partially offset the heating needs of the affected area. The backup heat source was expected to reach full capacity within 30 minutes, meeting approximately 30% of the affected area's heating needs.

[0108] Agent II, ingesting data from the equipment monitoring system and the user information management system, determined that the pipeline rupture could affect three surrounding heat exchange stations, impacting 100 commercial and 20 industrial users. After applying a large-scale model analysis, it developed Emergency Response Plan II: First, reduce the operating load of the three surrounding heat exchange stations, reduce overall network flow, and lower pipeline pressure to minimize hot water leakage. Load adjustments are expected to be completed within 15 minutes. Neighboring commercial and industrial users will be notified of the potential heat supply interruption and the estimated time it will be restored, allowing them to prepare accordingly. A professional inspection team will be deployed, using pipeline inspection robots, to inspect the area surrounding the ruptured pipeline to assess for other potential hazards. Inspections are expected to be completed within four hours.

[0109] Agent III, combining its emergency event monitoring system with historical data, discovered that the pipe rupture occurred at a critical node in the heating network, impacting a wide area. Using the large-scale model, it generated Emergency Response Plan III: Heating resources will be urgently redeployed from other areas. By adjusting regulating valves in the heating network, some heat will be delivered to the affected area. Heating resource redistribution is expected to be completed within two hours. Simultaneously, personnel will conduct a comprehensive inspection of the heating network in the affected area, focusing on branch lines connected to the ruptured pipe. This inspection is expected to be completed within six hours. Temporary heating equipment, such as electric heaters, will be provided to affected residents within three hours.

[0110] 2) Information interaction and judgment between agents

[0111] Agent I's judgment on other agents' plans:

[0112] When reviewing Plan II, Agent I concluded that while reducing the operating load of the heat exchange station would reduce leakage, it could also cause a drop in heating temperatures in surrounding areas not directly affected, potentially expanding the impact. It recommended closely monitoring the heating temperatures in non-affected areas while reducing the load and making timely adjustments. Regarding Plan III, Agent I considered the idea of deploying heating resources from other areas feasible, but the two-hour timeframe was too long. It recommended optimizing the scheduling algorithm to shorten the deployment time.

[0113] Agent II's judgment on other agents' plans:

[0114] Agent II analyzed Plan I, arguing that closing valves and mobilizing emergency repair teams was appropriate. However, the backup heat source took a long time to activate, suggesting that preheating the backup heat source be done in advance to shorten the startup time. Regarding Plan III, Agent II noted that the work arrangements for inspecting branch lines were reasonable, but the plan to distribute temporary heating equipment might not meet the needs of all residents due to limited availability. It recommended coordinating the opening of public heating facilities in surrounding communities.

[0115] Agent III's judgment on other agents' plans:

[0116] When evaluating Plan I, Agent III determined that the repair team was taking too long to arrive at the scene and recommended optimizing dispatch routes to improve traffic efficiency. Regarding Plan II, Agent III suggested diversifying user notification methods. Besides direct notification, Agent III also suggested expanding the reach of users through channels like social media and the heating company's app.

[0117] 3) Emergency response collaborative agent determines the optimal solution

[0118] Comprehensive assessment:

[0119] The emergency response collaborative agent collects the emergency plans of Agents I, II, and III, as well as information about their interactions. Regarding the effectiveness of restoring the hydraulic balance of the heating network, Plan I's valve closure and activation of backup heat sources, and Plan III's allocation of heating resources, have a positive effect on restoring hydraulic balance. Regarding the reduction of heat losses, Plan II's reduction of heat exchange station loads and leakage, along with measures taken by Plans I and III to ensure partial heat supply, can reduce heating fee losses. Regarding implementation costs, Plan I's mobilization of repair teams and equipment, and Plan III's resource allocation, involve human and material costs. Regarding the impact on user heating experience, Plan III's provision of temporary heating equipment and Plan II's notifications have also played a role.

[0120] Weight setting:

[0121] Taking into account the current serious pipeline rupture and large heat loss, the weight of the heat loss reduction degree is set to 0.4; the weight of the heat network hydraulic balance restoration effect is set to 0.3; the weight of the implementation cost is set to 0.2; and the weight of the impact on the user's heating experience is set to 0.1.

[0122] Solution options:

[0123] The emergency response collaborative agent assigned a quantitative score to each plan. Plan I received an overall score of 80, Plan II received a score of 75, and Plan III received a score of 82. Ultimately, the optimal plan, combining Plans I and III, was determined. The plan closed the appropriate valves, activated backup heat sources, deployed heating resources, and mobilized emergency repair teams and equipment for expedited repairs. This plan was then provided to heating service personnel. Based on the optimal plan, heating service personnel quickly implemented it, mobilizing heating resources, inspecting branch lines, and, where financially feasible, providing residents with temporary heating equipment or necessary financial compensation. This effectively addressed the pipeline rupture emergency, minimized heat losses, and ensured the stable operation of the heating system and the basic heating needs of users.

[0124] In this embodiment, the emergency response scheduling decision task is expressed as: ;

[0125] D To generate emergency response scheduling plan; A large-scale model for emergency response and dispatching of heating systems; To quickly reconstruct the hydraulic balance of the heating network and reduce heat loss; Provides current heating system operation information; Provide emergency response and dispatch knowledge for heating systems.

[0126] In this embodiment, each emergency response scheduling agent is driven by a pre-selected large model of emergency response scheduling for different heating systems.

[0127] In the several embodiments provided in this application, it should be understood that the disclosed systems and methods can also be implemented in other ways. The system embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architectures, functions, and operations of the systems, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment, or a portion of code, and the module, program segment, or a portion of code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the boxes can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes 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 and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified functions or actions, or can be implemented using a combination of dedicated hardware and computer instructions.

[0128] In addition, the functional modules in each embodiment of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part. If the function is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method of each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk, or an optical disk.

[0129] With the above-described preferred embodiments of the present invention as a guide, and with reference to the above description, relevant personnel are fully capable of making various changes and modifications without departing from the technical scope of this invention. The technical scope of this invention is not limited to the contents of the specification and must be determined according to the scope of the claims.

Claims

1. A multi-stage emergency response scheduling method for a heating system with a large model, characterized in that: include: Emergency prevention and preparation stage, emergency response stage, emergency disposal and aftermath stage; The emergency preparedness phase includes: A multi-type emergency event question-and-answer database for the heating system is constructed based on the large model. The database generates targeted questions and answers by extracting key event information from the heating system's emergency response and scheduling data. The targeted questions and answers include intelligent questions and answers about emergency knowledge. The database provides emergency event cause analysis, risk and hidden danger query, and recommendation of emergency response and scheduling measures. The emergency response phase includes: Use prompt word engineering to guide the large model to output the optimal feature combination to characterize various types of emergency events, and identify various types of emergency events; Evaluate the heat loss caused by the identified emergency events, and set emergency response scheduling decision-making tasks based on the goals of quickly reconstructing the hydraulic balance of the heating network and reducing heat loss; Multiple emergency response scheduling agents respectively output emergency response scheduling plans based on the emergency response scheduling decision-making tasks; Each agent judges the emergency response and dispatch plans output by other agents; Obtain the optimal emergency response scheduling plan based on the judgment and analysis of the emergency response collaborative agent and each agent; The emergency response and aftermath phase includes: The emergency response and dispatch process is analyzed and summarized using a large model, and emergency response and dispatch plan simulations are performed using intelligent agents to strengthen the emergency response and dispatch process, including: Set up an intelligent agent to simulate the operating environment of the heating system, generate an emergency event scenario and execute the emergency response dispatch plan; Based on the potential vulnerabilities identified by the intelligent agent simulation, improvement suggestions are generated using the large model and fed back into the emergency response and scheduling plan library to update and improve the corresponding plans. At the same time, the intelligent agent simulation is re-run regularly to verify whether the improved plans are effective.

2. The multi-stage emergency response scheduling method for a heating system with a large model according to claim 1 is characterized in that: During the emergency response phase, a feature combination workflow is set up to output various types of emergency events. The feature combination workflow includes uploading an emergency response scheduling dataset, describing the emergency event, describing the task requirements for feature extraction using machine learning, selecting a feature combination mechanism, and outputting the optimal feature combination. Use prompt word engineering to guide the large model to sequentially implement the workflow of output feature combination, and output the optimal feature combination to represent various types of emergency events; The optimal feature combination of each type of emergency event is input into the preset machine learning model for training and learning, and a heating emergency event recognition model is established to identify the types of emergency events occurring in the heating system.

3. The multi-stage emergency response scheduling method for a heating system with a large model according to claim 2 is characterized in that: Establishing a heating emergency event recognition model includes: setting up a two-layer machine learning model to extract features from the emergency event dataset, where the input end of the first layer of the machine learning model is the original emergency response scheduling dataset, and the features of the original emergency response scheduling dataset are abstracted through unsupervised learning; The second-layer machine learning model uses local feature extraction capabilities to perform deep extraction of various data features; The cross-validation method is used to evaluate different feature combinations, and the feature combinations are optimized and iterated based on the evaluation results to output the optimal feature combination for identifying emergency events in the heating system.

4. The multi-stage emergency response scheduling method for a heating system with a large model according to any one of claims 1 to 3, characterized in that: The heat loss assessment step includes: constructing an economic loss calculation model, and calculating the economic losses caused by the emergency event based on the data identified in the emergency event.

5. The multi-stage emergency response scheduling method for a heating system with a large model according to claim 4 is characterized in that: The data in the emergency event include the number of affected heat users, heating area, missing value of heat load, and emergency repair time.

6. The multi-stage emergency response scheduling method for a heating system with a large model according to claim 5 is characterized in that: The method for constructing an economic loss calculation model includes, based on the economic loss data of historical emergency events and the currently determined loss factors, taking the number of affected heat users, the duration of the heating interruption, and the unit price of the heating fee as independent variables, and the economic loss value as the dependent variable, training the model to learn the relationship between the loss factors and the economic losses, and constructing an economic loss calculation model.

7. The multi-stage emergency response scheduling method for a heating system with a large model according to claim 1 is characterized in that: The method for rapidly reconstructing the hydraulic balance of the heating network includes constructing a hydraulic model of the heating network based on the topological structure data of the heating network, the principles of fluid mechanics and historical operation data, and calculating the impact of emergency events on the hydraulic balance of the heating network.

8. A multi-stage emergency response scheduling method for a heating system with a large model according to any one of claims 1-3 or 5-7, characterized in that: A multi-level emergency response and scheduling strategy is set up: in the first layer, multiple emergency response and scheduling intelligent agents are set up to analyze and reason about the emergency event information occurring in the heating system, and output their own emergency response and scheduling plans; in the second layer, each emergency response and scheduling intelligent agent exchanges information, and judges, discusses and puts forward opinions on the emergency response and scheduling plans output by other emergency response and scheduling intelligent agents; in the third layer, the set emergency response collaborative intelligent agent integrates the current emergency event problem and the inference of each emergency response and scheduling intelligent agent for judgment and analysis, and obtains the optimal emergency response and scheduling plan.

9. The multi-stage emergency response scheduling method for a heating system according to claim 8, characterized in that: described The emergency response dispatch decision-making task is expressed as: ; Where: D To generate emergency response scheduling plan; A large-scale model for emergency response and dispatching of heating systems; To quickly reconstruct the hydraulic balance of the heating network and reduce heat loss; Provides current heating system operation information; Provide emergency response and dispatch knowledge for heating systems.

10. The multi-stage emergency response scheduling method for a heating system with a large model according to claim 1 is characterized in that: It also includes extracting key event information from the heating system's emergency response and scheduling data, generating targeted questions and answers, and using large models to automatically build a question-and-answer database for multiple types of emergency events in the heating system.

Images