Heat supply energy consumption diagnosis and analysis system fusing big language model and cloud and mist edge collaboration

Abstract

The invention discloses a heat supply energy consumption diagnosis and analysis system fusing a big language model and cloud-mist edge collaboration, and the system comprises an edge processing layer which is used for monitoring the energy consumption related data and heat supply operation data of a heat supply system diagnosis object; the fog processing layer is used for extracting strong correlation data characteristics influencing energy consumption and establishing an energy consumption prediction model of a heat supply system diagnosis object; the analysis module is also used for establishing an energy consumption anomaly analysis knowledge graph according to the heat supply operation data and energy consumption anomaly analysis mechanism knowledge; the cloud processing layer is used for training a heat supply system energy consumption abnormity diagnosis large model according to the energy consumption predicted value of the heat supply system diagnosis object in combination with a preset energy consumption diagnosis standard value; and the server is also used for obtaining an energy consumption abnormality diagnosis type and an abnormality diagnosis object according to the heat supply system energy consumption abnormality diagnosis large model, performing question and answer intention recognition and constructing an energy consumption abnormality analysis question and answer agent based on the large language model in combination with the energy consumption abnormality analysis knowledge graph, and generating an energy consumption abnormality analysis and energy-saving recommendation scheme answer.

Description

Technical Field

[0001] The present invention belongs to the technical field of heating energy consumption diagnosis, and particularly relates to a heating energy consumption diagnosis and analysis system integrating large language models and cloud-edge-fog collaboration. Background Art

[0002] With the gradual increase in the scale of centralized heating, heating enterprises have built and introduced monitoring systems one after another to ensure energy management and service quality, hoping to use modern means to master the pipe network conditions and achieve energy conservation and consumption reduction. As heating enterprises, effectively controlling various heating indicators, completing normal heating work with the best energy efficiency ratio, and doing a good job in energy consumption indicators are more conducive to the economic and stable operation and development of the heating system.

[0003] The energy consumption indicators of heating work mainly include three major items: water consumption, electricity consumption, and heat consumption. During the operation of the heating system, in order to improve the operation efficiency of the heating system, it is necessary to monitor the water consumption, electricity consumption, and heat consumption of the system in real time, judge whether the energy consumption is abnormal under the current operating conditions, analyze the reasons for energy consumption anomalies, and give energy-saving solutions. At present, the traditional methods for heating system energy consumption diagnosis and analysis have weak intelligence levels, insufficient decision-making analysis, and poor accuracy of energy consumption diagnosis and analysis. In addition, with the continuous expansion of the scale of the heating system, the amount of data in heat substations and heat networks has also increased, resulting in large data transmission and calculation pressures and slow responses.

[0004] Based on the above technical problems, it is necessary to design a new heating energy consumption diagnosis and analysis system integrating large language models and cloud-edge-fog collaboration. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide a heating energy consumption diagnosis and analysis system integrating large language models and cloud-edge-fog collaboration. Through the multi-level architecture of cloud-edge-fog collaborative computing, it analyzes the processing of heating energy consumption diagnosis and energy consumption anomaly analysis under cloud-edge-fog collaboration, effectively integrates the advantages of cloud, fog, and edge computing, improves the data processing ability and response speed of heating energy consumption diagnosis and analysis, and integrates and utilizes technologies such as large language models, knowledge graphs, and agents, which can better adapt to complex heating environments, improve the accuracy and efficiency of heating energy consumption diagnosis and analysis services, and propose interpretable energy consumption diagnosis and analysis decisions by understanding and analyzing the current energy consumption diagnosis and analysis environment and related energy consumption problems to help heating operation and maintenance personnel achieve scientific and reasonable final energy consumption diagnosis and analysis.

[0006] To solve the above technical problems, the technical solution of the present invention is:

[0007] The present invention provides a heating energy consumption diagnosis and analysis system integrating large language models and cloud-edge-fog collaboration, which includes:

[0008] The edge processing layer is used to monitor the energy consumption - related data and heating operation data of the diagnosis objects in the heating system;

[0009] The fog processing layer is used to extract the strongly - related data features affecting energy consumption, establish an energy consumption prediction model for the diagnosis objects in the heating system; it is also used to establish a knowledge graph for energy consumption anomaly analysis based on the heating operation data and the knowledge of energy consumption anomaly analysis mechanism;

[0010] The cloud processing layer is used to train a large - model for energy consumption anomaly diagnosis in the heating system according to the predicted energy consumption values of the diagnosis objects in the heating system, combined with the preset energy consumption diagnosis standard values; it is also used to obtain the energy consumption anomaly diagnosis types and anomaly diagnosis objects according to the large - model for energy consumption anomaly diagnosis in the heating system, combined with the knowledge graph for energy consumption anomaly analysis, conduct question - answering intention recognition and construct an energy consumption anomaly analysis question - answering intelligent agent based on the large language model, and generate answers for energy consumption anomaly analysis and energy - saving recommendation solutions.

[0011] Furthermore, the edge processing layer, which is used to monitor the energy consumption - related data and heating operation data of the diagnosis objects in the heating system, includes:

[0012] In the edge processing layer, through the set edge processing devices, monitor the energy consumption - related data and heating operation data of the heat networks and heat substations in different regions of the heating system, including the water consumption data, heat consumption data, and power consumption data of the heat networks and heat substations in different regions, as well as the supply - return water temperature, circulating water flow rate, make - up water volume, pump operation frequency, outdoor temperature, indoor temperature, building envelope insulation performance, pipeline network insulation parameters, and heating area of the heat networks and heat substations;

[0013] After pre - processing, data encapsulation, and parsing of the energy consumption - related data and heating operation data of the heat networks and heat substations in different regions, transmit them to the fog processing layer.

[0014] Furthermore, the edge processing layer, which is connected to the fog processing layer and the cloud processing layer, is used to obtain the energy consumption anomaly diagnosis types and anomaly diagnosis objects transmitted by the cloud processing layer and display them to the heating operation and maintenance personnel;

[0015] It is also used to transmit the energy consumption anomaly problems input by the heating operation and maintenance personnel to the cloud processing layer, and obtain the answers for energy consumption anomaly analysis and energy - saving recommendation solutions transmitted by the cloud processing layer and display them to the heating operation and maintenance personnel;

[0016] And, it is used to generate historical energy consumption curves in three time dimensions of daily, monthly, and heating season based on the historical energy consumption data of the heat networks and heat substations in different regions.

[0017] Furthermore, the fog processing layer, which is used to extract the relevant data features affecting energy consumption and establish an energy consumption prediction model for the diagnosis objects in the heating system, includes:

[0018] In the fog processing layer, for the energy consumption-related data and heating operation data of the heating system diagnosis object transmitted by the edge processing layer, after using a deep convolutional neural network to extract the water consumption, heat consumption, and power consumption characteristics of different regional heat networks and heat substations, the self-attention mechanism in the Transformer model is used to capture the global correlations between different time points and different energy consumption types. Then, through a fully connected layer, the mapping relationship between the energy consumption feature vector and the energy consumption value is learned, and an energy consumption prediction model for the heating system diagnosis object is established to output the predicted values of water consumption, heat consumption, and power consumption of different regional heat networks and heat substations in the future period.

[0019] Among them, the fog processing layer deploys the established energy consumption prediction model of the heating system diagnosis object to the server of the fog processing layer or distributes it to each edge processing device corresponding to the edge processing layer for real-time energy consumption prediction.

[0020] Furthermore, the fog processing layer is also used to establish an energy consumption anomaly analysis knowledge graph based on the heating operation data and the knowledge of the energy consumption anomaly analysis mechanism, including:

[0021] In the fog processing layer, obtain the heating operation data of different regional heat networks and heat substations in the heating system in the historical time period, reflecting the operation status under different operating conditions of the system;

[0022] Obtain the theoretical knowledge about water consumption anomaly, heat consumption anomaly, and power consumption anomaly analysis from papers, academic reports, and industry standards, including energy consumption anomaly classification, anomaly causes, and energy-saving solutions; the energy-saving solutions include equipment repair and transformation, heating dispatch strategy optimization, and equipment operation parameter adjustment;

[0023] Communicate with experts in the heating field to obtain the empirical knowledge about energy consumption anomaly analysis in actual work, including the energy consumption anomaly manifestations caused by different faults and judging energy consumption anomalies through changes in operation parameters;

[0024] Set the energy consumption anomaly analysis logic knowledge, including: the analysis logic knowledge when water consumption diagnosis is abnormal, heat consumption diagnosis is abnormal, and power consumption diagnosis is abnormal;

[0025] Identify various equipment entities, operation parameter entities, energy consumption anomaly type entities, anomaly cause entities, and energy-saving solution entities from the heating operation data and the obtained theoretical knowledge, empirical knowledge, and logic knowledge about water consumption anomaly, heat consumption anomaly, and power consumption anomaly analysis;

[0026] Determine the relationship between equipment and operation parameters, the correlation between changes in operation parameters and energy consumption anomalies, the causal relationship between energy consumption anomaly types and the causes leading to such anomalies, and the relationship between energy consumption anomaly causes and solutions;

[0027] After knowledge fusion based on the recognized entities and determined relationships, the extracted and fused entities are added as nodes to the knowledge graph, and corresponding edges are created according to the relationships between the entities to establish an energy consumption anomaly analysis knowledge graph.

[0028] Further, the setting of the energy consumption anomaly analysis logic knowledge includes:

[0029] When the water consumption diagnosis is abnormal, judge whether the room temperature compliance rate of the heat users is abnormal. If the room temperature compliance rate is abnormal, it indicates poor heating quality, and the heat users may have behaviors such as forced circulation by draining water or there is a risk of heat network leakage, and heat network leakage diagnosis and location investigation are required; if the room temperature compliance rate is normal, there is a risk of heat network leakage, and heat network leakage diagnosis and location investigation are required;

[0030] When the heat consumption diagnosis is abnormal, judge whether the transmission efficiency of the primary pipe network is abnormal. If the transmission efficiency of the primary pipe network is abnormal, it indicates that the heat loss through the pipeline is large, and there is a risk of old pipelines or damaged pipeline insulation layers; otherwise, there is hydraulic imbalance, resulting in thermal imbalance, or there are old residential areas, and the building envelope structures of the old residential areas need to be renovated;

[0031] When the power consumption diagnosis is abnormal, judge whether the water consumption is abnormal and whether the energy efficiency of the circulating water pump is abnormal. If the water consumption is abnormal and the energy efficiency of the circulating water pump is abnormal, it indicates excessive water loss, frequent start and stop of the makeup water pump, and unreasonable selection of the circulating water pump; if the water consumption is abnormal and the energy efficiency of the circulating water pump is normal, it indicates more water loss, frequent start and stop of the makeup water pump, and there is a risk of unreasonable selection of the circulating water pump; if the water consumption is normal and the energy efficiency of the circulating water pump is abnormal, the selection of the circulating water pump is unreasonable; if the water consumption is normal and the energy efficiency of the circulating water pump is normal, there is a risk of unreasonable selection of the circulating water pump.

[0032] Further, the cloud processing layer is used to train a large model for energy consumption anomaly diagnosis of the heating system according to the energy consumption prediction value of the diagnosis object of the heating system, in combination with the preset energy consumption diagnosis standard value, including:

[0033] In the cloud processing layer, obtain the energy consumption prediction value of the diagnosis object of the heating system from the fog processing layer;

[0034] Determine the energy consumption diagnosis standard value under different working conditions according to the design parameters, historical operation data and industry standards of the heating system;

[0035] Compare the energy consumption prediction value with the preset energy consumption diagnosis standard value to judge whether there is energy consumption anomaly and perform data annotation;

[0036] Integrate the energy consumption prediction value, the energy consumption diagnosis standard value under different working conditions, and the annotation data into a data set, and input it into the pre-trained large model, and use the method of combining the large model and prompt words to train a large model for energy consumption anomaly diagnosis of the heating system.

[0037] Furthermore, the setting of the prompt words includes:

[0038] Role setting: Clearly define that the large model is an expert in heating energy consumption anomaly diagnosis during this conversation process, and guide the large model to output the types of energy consumption anomaly diagnosis and the objects of anomaly diagnosis;

[0039] Requirement description: Describe the task requirements completed by the large model, including: obtaining energy consumption prediction values, energy consumption diagnosis standard values, and energy consumption anomaly diagnosis.

[0040] Furthermore, the cloud processing layer is used to obtain the types of energy consumption anomaly diagnosis and the objects of anomaly diagnosis according to the large model of heating system energy consumption anomaly diagnosis, combine with the knowledge graph of energy consumption anomaly analysis, conduct question-and-answer intention recognition and construct an energy consumption anomaly analysis question-and-answer intelligent agent based on the large language model, and generate answers to energy consumption anomaly analysis and energy-saving recommendation solutions, including:

[0041] Define the types of energy consumption anomaly diagnosis, the objects of anomaly diagnosis, historical energy consumption anomaly analysis questions and answers, question-and-answer intentions, and the knowledge graph of energy consumption anomaly analysis output by the large model of heating system energy consumption anomaly diagnosis as the environment, where changes in each variable will lead to different current energy consumption anomaly analysis environments;

[0042] Input the environmental information and the preset energy consumption anomaly analysis prompt words into the large language model. The large language model uses its own language understanding and generation capabilities to construct an energy consumption anomaly analysis large model and integrate it into the intelligent agent;

[0043] After the intelligent agent obtains historical energy consumption anomaly analysis questions and answers, the types of energy consumption anomaly diagnosis, the objects of anomaly diagnosis, the knowledge of energy consumption anomaly analysis extracted from the knowledge graph of energy consumption anomaly analysis, and the recognized question-and-answer intentions from the current energy consumption anomaly analysis environment, it conducts analysis and reasoning according to the current environmental state to generate the actions of the current environment, then verifies the generated actions, generates answers to energy consumption anomaly analysis and energy-saving recommendation solutions, and finally updates the current energy consumption anomaly analysis environment to achieve the iteration of multi-round questions and answers.

[0044] Furthermore, the generation of the actions of the current environment after analysis and reasoning includes:

[0045] Assume that each energy consumption anomaly analysis question and answer contains corresponding question-and-answer intentions, including energy consumption anomaly symptoms, energy consumption anomaly causes, and energy-saving solutions;

[0046] After extracting the context semantic features in the energy consumption anomaly analysis question-and-answer sentences, predict the question-and-answer behavior and attribute information to obtain the overall question-and-answer intention;

[0047] According to the overall intention of the question and answer, extract the energy consumption anomaly diagnosis type and the entity of the anomaly diagnosis object from the question and answer context, align and calculate the similarity with the entities in the energy consumption anomaly analysis knowledge graph, and identify the list of energy consumption anomaly analysis entities with the highest similarity;

[0048] According to the identified list of energy consumption anomaly analysis entities, extract the corresponding entity and relationship information from the energy consumption anomaly analysis knowledge graph, and perform pruning of the knowledge subgraph. After retaining the subgraph information with strong relevance, it is transformed into the corresponding actions in the current environment.

[0049] The beneficial effects of the present invention are:

[0050] (1) Through the multi-level architecture of cloud-edge-fog collaborative computing, the present invention analyzes the processing of heating energy consumption diagnosis and energy consumption anomaly analysis under cloud-edge-fog collaboration, effectively integrates the advantages of cloud, fog, and edge computing, improves the data processing ability and response speed of heating energy consumption diagnosis and analysis, and combines technologies such as large language models, knowledge graphs, and agents, enabling better adaptation to complex heating environments, improving the accuracy and efficiency of heating energy consumption diagnosis and analysis services, and by understanding and analyzing the current energy consumption diagnosis and analysis environment and related energy consumption problems, providing interpretable energy consumption diagnosis and analysis decisions to help heating operation and maintenance personnel achieve scientific and reasonable final energy consumption diagnosis and analysis;

[0051] (2) The present invention monitors the energy consumption-related data and heating operation data of the diagnosis object of the heating system through the edge processing layer, ensures the acquisition of accurate and timely raw data, performs data collection and preliminary processing at the edge, and transmits the processed data to the fog processing layer and the cloud processing layer, reducing the network transmission volume and the network bandwidth pressure;

[0052] (3) The present invention extracts strongly relevant data features affecting energy consumption through the fog processing layer and establishes an energy consumption prediction model for the diagnosis object of the heating system; and, according to the heating operation data and the knowledge of the energy consumption anomaly analysis mechanism, establishes an energy consumption anomaly analysis knowledge graph; which helps to understand the heating energy consumption trend in advance, facilitates subsequent energy consumption anomaly diagnosis, and can structurally represent complex energy consumption anomaly analysis knowledge, helps to quickly locate and analyze energy consumption anomaly problems, provides rich knowledge support for energy consumption anomaly analysis questions and answers, and improves the heating intelligent level;

[0053] (4) The present invention trains a large model for diagnosing abnormal energy consumption in a heating system through the cloud processing layer based on the predicted energy consumption value of the diagnosis object in the heating system and in combination with a preset standard value for energy consumption diagnosis; and, obtains the type of abnormal energy consumption diagnosis and the abnormal diagnosis object according to the large model for diagnosing abnormal energy consumption in the heating system, and in combination with the knowledge graph for analyzing abnormal energy consumption, conducts question-and-answer intention recognition and constructs an intelligent agent for analyzing abnormal energy consumption based on a large language model to generate answers for analyzing abnormal energy consumption and recommending energy-saving solutions; by introducing a large language model and an intelligent agent, in the form of conversational instructions and question-and-answer, interacts naturally and smoothly with the large model and the intelligent agent, accurately identifies abnormal energy consumption situations, discovers potential problems, helps heating operation and maintenance personnel quickly understand abnormal energy consumption problems, and timely take measures for energy consumption adjustment. At the same time, the user experience is greatly improved and the usage threshold is reduced.

[0054] Other features and advantages will be described in the following specification, and in part will be obvious from the specification, or will be understood by implementing the present invention. The objectives and other advantages of the present invention are achieved and obtained by the structures specifically pointed out in the specification and the drawings.

[0055] To make the above objectives, features, and advantages of the present invention more obvious and understandable, the following specific preferred embodiments are given, and in conjunction with the accompanying drawings, the detailed description is as follows. Brief Description of the Drawings

[0056] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0057] Figure 1 It is a block diagram of the principle of a heating energy consumption diagnosis and analysis system that integrates a large language model and cloud-edge-fog collaboration of the present invention;

[0058] Figure 2 It is a flowchart of the method for establishing a knowledge graph for analyzing abnormal energy consumption of the present invention;

[0059] Figure 3 It is a flowchart of the method for generating answers for analyzing abnormal energy consumption and recommending energy-saving solutions of the present invention. Detailed Description of the Specific Embodiments

[0060] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0061] As Figure 1 shown, this embodiment provides a heating energy consumption diagnosis and analysis system that integrates large language models and cloud-edge-fog collaboration, which includes:

[0062] An edge processing layer for monitoring energy consumption-related data and heating operation data of the diagnosis object of the heating system;

[0063] A fog processing layer for extracting strongly correlated data features affecting energy consumption and establishing an energy consumption prediction model for the diagnosis object of the heating system; it is also used to establish an energy consumption anomaly analysis knowledge graph based on heating operation data and energy consumption anomaly analysis mechanism knowledge;

[0064] A cloud processing layer for training a large model for diagnosing heating system energy consumption anomalies according to the predicted energy consumption value of the diagnosis object of the heating system in combination with a preset energy consumption diagnosis standard value; it is also used to obtain the energy consumption anomaly diagnosis type and anomaly diagnosis object according to the large model for diagnosing heating system energy consumption anomalies, and in combination with the energy consumption anomaly analysis knowledge graph, perform question-and-answer intent recognition and construct an energy consumption anomaly analysis question-and-answer intelligent agent based on the large language model to generate answers for energy consumption anomaly analysis and energy-saving recommendation solutions.

[0065] It should be noted that as a computing service model between the edge processing layer and the cloud processing layer, the fog processing layer deploys computing resources close to the terminal device to meet the requirements of latency sensitivity and avoid the limitations of cloud computing and edge computing; the cloud processing layer is responsible for the training of the large model for abnormal diagnosis of the energy consumption of the heating system and the training and construction of the intelligent agent for answering questions about abnormal energy consumption analysis based on the large language model, and transmits the abnormal diagnosis results of the energy consumption of the heating system, the abnormal energy consumption analysis, and the answers to the energy-saving recommendation solutions to the edge processing layer through the fog processing layer so that the heating operation and maintenance personnel can know in time; the fog processing layer is responsible for the establishment of the energy consumption prediction model and the knowledge graph for abnormal energy consumption analysis, and can also receive the lightweight large model for abnormal diagnosis of the energy consumption of the heating system and the intelligent agent for answering questions about abnormal energy consumption analysis sent by the cloud processing layer. After pruning, decomposing, and compressing the large model and the intelligent agent, it conducts regional model training; the edge processing layer is responsible for data collection and can also receive the energy consumption prediction model sent by the fog processing layer to predict the energy consumption of the corresponding heat network area and heat substation. Among them, there are multiple edge processing devices in the edge processing layer; through the multi-level architecture of cloud-edge-fog collaborative computing, it analyzes the processing of heating energy consumption diagnosis and abnormal energy consumption analysis under cloud-edge-fog collaboration, effectively integrating the advantages of cloud, fog, and edge computing, improving the data processing ability and response speed of heating energy consumption diagnosis and analysis, and by integrating and utilizing technologies such as large language models, knowledge graphs, and intelligent agents, it can better adapt to complex heating environments, improve the accuracy and efficiency of heating energy consumption diagnosis and analysis services, and through understanding and analyzing the current energy consumption diagnosis and analysis environment and related energy consumption problems, it proposes interpretable energy consumption diagnosis and analysis decisions to help heating operation and maintenance personnel achieve scientific and reasonable final energy consumption diagnosis and analysis.

[0066] In this embodiment, the edge processing layer is used to monitor the energy consumption-related data and heating operation data of the diagnosis object of the heating system, including:

[0067] In the edge processing layer, the edge processing devices are set to monitor the energy consumption-related data and heating operation data of different areas of the heat network and heat substations in the heating system, including the water consumption data, heat consumption data, and power consumption data of different areas of the heat network and heat substations, as well as the supply and return water temperatures, circulating water flow rate, make-up water volume, pump operation frequency, outdoor temperature, indoor temperature, building envelope insulation performance, pipeline network insulation parameters, and heating area of the heat network and heat substations;

[0068] After preprocessing, data encapsulation, and parsing of the energy consumption-related data and heating operation data of different areas of the heat network and heat substations, they are transmitted to the fog processing layer.

[0069] In practical applications, the energy consumption diagnosis objects include district heating networks and heat substations in different regions. When diagnosing the energy consumption of a heating network, key factors such as the electricity consumption of circulating pumps, the insulation effect of pipelines, the length and diameter of pipelines, and the water flow velocity are focused on. For example, long-distance, small-diameter pipelines or too fast water flow velocity will increase resistance and the electricity consumption of circulating pumps; poor pipeline insulation will increase heat loss. A heat substation is an intermediate link connecting the heating network and users in the heating system, responsible for heat exchange, regulation, and distribution of the hot water in the heating network. When diagnosing the energy consumption of a heat substation, the heat transfer efficiency and heat transfer volume of heat exchangers, the electricity consumption of in-station circulating pumps and make-up pumps, and the operating conditions of the control system are key factors.

[0070] Factors affecting the heat consumption of the heating system include building age, geographical location, insulation performance, building type, operation regulation strategy, make-up water volume, etc.; factors affecting the water consumption of the heating system include equipment information (running, leaking, dripping, and overflowing caused by aging, rust, pipeline corrosion, and disrepair), user behavior (users privately draining water), etc.; factors affecting the electricity consumption of the heating system include the operating parameters of circulating pumps, hydraulic imbalance parameters (heat transfer and transportation efficiency of the pipe network, flow ratio of the pipe network, hydraulic balance degree), temperature difference and flow rate (the method of large temperature difference and small flow rate can reduce the electricity consumption of heat substations).

[0071] In this embodiment, the edge processing layer is connected to the fog processing layer and the cloud processing layer, and is used to obtain the energy consumption anomaly diagnosis types and anomaly diagnosis objects transmitted by the cloud processing layer and display them to the heating operation and maintenance personnel;

[0072] It is also used to transmit the energy consumption anomaly problems input by the heating operation and maintenance personnel to the cloud processing layer, and obtain the energy consumption anomaly analysis and energy-saving recommendation solution answers transmitted by the cloud processing layer and display them to the heating operation and maintenance personnel;

[0073] And it is used to generate historical energy consumption curves in three time dimensions of daily, monthly, and heating season based on the historical energy consumption data of district heating networks and heat substations in different regions.

[0074] In this embodiment, the fog processing layer is used to extract relevant data features affecting energy consumption and establish an energy consumption prediction model for the diagnosis objects of the heating system, including:

[0075] In the fog processing layer, for the energy consumption-related data and heating operation data of the heating system diagnosis objects transmitted by the edge processing layer, after using a deep convolutional neural network to extract the water consumption, heat consumption, and electricity consumption features of district heating networks and heat substations in different regions, the self-attention mechanism in the Transformer model is used to capture the global correlation between different time points and different energy consumption types, and then the mapping relationship between the energy consumption feature vector and the energy consumption value is learned through a fully connected layer to establish an energy consumption prediction model for the heating system diagnosis objects, and output the predicted values of water consumption, heat consumption, and electricity consumption of district heating networks and heat substations in the future time period;

[0076] Among them, the fog processing layer deploys the energy consumption prediction model of the established heating system diagnosis object to the server of the fog processing layer or distributes it to each edge processing device corresponding to the edge processing layer for real-time energy consumption prediction.

[0077] In actual applications, the energy consumption prediction model comprehensively uses a deep convolutional neural network, a Transformer model, and a fully connected layer. The deep convolutional neural network has a powerful local feature extraction ability and can extract the local features of water consumption, heat consumption, and power consumption of different regional heat networks and heat substations from the input energy consumption-related data and heating operation data. The deep convolutional neural network includes a convolutional layer, a pooling layer, and an activation function layer. For example, multiple convolutional layers can be stacked, and each convolutional layer is followed by a pooling layer for downsampling to reduce the data dimension. The activation function can be selected as the ReLU function, which has the advantages of simple calculation and fast convergence speed. The preprocessed data is input into the deep convolutional neural network, and the local features of water consumption, heat consumption, and power consumption of different regional heat networks and heat substations are extracted through convolutional operations. The convolutional kernel slides on the data to extract features from the local area, and each convolutional kernel can learn different feature patterns. The pooling layer can reduce the dimension of the extracted features, reduce the amount of calculation, and enhance the robustness of the features.

[0078] The self-attention mechanism of the Transformer model can capture the global correlations between these features at different time points and different energy consumption types, and mine the long-range dependencies in the data. Embedding layer: The feature vectors extracted by the deep convolutional neural network are input into the embedding layer of the Transformer model to convert the feature vectors into embedding representations suitable for Transformer processing. Self-attention mechanism: The core of the Transformer model is the self-attention mechanism, which can calculate the correlation between each element in the input sequence and other elements, so as to capture the global correlations between different time points and different energy consumption types. Multi-head attention mechanism: In order to capture different types of correlation information, the multi-head attention mechanism can be used to execute the self-attention mechanism multiple times in parallel and then splice the results together. Feed-forward neural network: After the multi-head attention mechanism, a feed-forward neural network is used to further process the output to enhance the expression ability of the model.

[0079] Finally, the fully connected layer maps the learned energy consumption feature vectors to specific energy consumption values, thereby realizing the prediction of water consumption, heat consumption, and power consumption in future time periods. Network structure: The feature vectors output by the Transformer model are input into the fully connected layer. The fully connected layer consists of multiple neurons, and each neuron is connected to all neurons in the previous layer. Mapping learning: Through the linear transformation and activation function of the fully connected layer, the mapping relationship between the energy consumption feature vectors and the energy consumption values is learned. The output of the last layer is the predicted values of water consumption, heat consumption, and power consumption of the heat networks and heat substations in different regions in future time periods.

[0080] As Figure 2 shown, in this embodiment, the fog processing layer is further configured to establish a knowledge graph for energy consumption anomaly analysis according to the heat supply operation data and the knowledge of the energy consumption anomaly analysis mechanism, including:

[0081] In the fog processing layer, obtain the heat supply operation data of the heat networks and heat substations in different regions of the heat supply system in the historical time period, which reflects the operation status under different operating conditions of the system;

[0082] Obtain the theoretical knowledge about water consumption anomaly, heat consumption anomaly, and power consumption anomaly analysis from papers, academic reports, and industry standards, including energy consumption anomaly classification, anomaly causes, and energy-saving solutions; the energy-saving solutions include equipment repair and transformation, optimization of heat supply dispatching strategies, and adjustment of equipment operation parameters;

[0083] Communicate with experts in the heat supply field to obtain the empirical knowledge about energy consumption anomaly analysis in actual work, including the energy consumption anomaly manifestations caused by different faults and the judgment of energy consumption anomalies through changes in operation parameters;

[0084] Set the logical knowledge of energy consumption anomaly analysis, including: the logical knowledge of analysis when water consumption diagnosis is abnormal, heat consumption diagnosis is abnormal, and power consumption diagnosis is abnormal;

[0085] Identify various equipment entities, operation parameter entities, energy consumption anomaly type entities, anomaly cause entities, and energy-saving solution entities from the heat supply operation data and the obtained theoretical knowledge, empirical knowledge, and logical knowledge about water consumption anomaly, heat consumption anomaly, and power consumption anomaly analysis;

[0086] Determine the relationship between equipment and operation parameters, the association between changes in operation parameters and energy consumption anomalies, the causal relationship between energy consumption anomaly types and the causes leading to such anomalies, and the relationship between energy consumption anomaly causes and solutions;

[0087] After knowledge fusion according to the identified entities and determined relationships, add the extracted and fused entities as nodes to the knowledge graph, and create corresponding edges according to the relationships between the entities to establish a knowledge graph for energy consumption anomaly analysis.

[0088] It should be noted that the solutions for abnormal energy consumption include:

[0089] 1) Optimization of the pipe network system

[0090] Balance the hydraulic pressure of the pipe network: Through the installation of hydraulic balance devices, such as static balance valves, dynamic flow balance valves, etc., conduct hydraulic balance debugging on the heating pipe network to ensure reasonable flow distribution in each branch pipeline, avoid the situation of overheating in some areas and overcooling in some areas, and thus reduce the overall energy consumption.

[0091] Reduce heat loss of the pipe network: Strengthen the insulation measures of the heating pipe network, use high-quality insulation materials, such as polyurethane foam insulation pipes, etc., to reduce the loss of heat during transmission. At the same time, regularly inspect and repair the insulation layer of the pipe network, promptly discover and handle damaged parts, and reduce heat loss.

[0092] 2) Energy-saving control of the heat substation

[0093] Install a climate compensator: Automatically adjust the supply water temperature and flow of the heat substation according to the change of outdoor temperature to achieve heating on demand. For example, when the weather is warmer, reduce the supply water temperature and flow to avoid overheating and thus save energy.

[0094] Adopt variable frequency speed regulation technology: Adopt variable frequency speed regulation control for the circulating water pump and make-up water pump of the heat substation, automatically adjust the speed of the water pump according to the change of the actual heating load, and reduce the power consumption of the water pump. Generally speaking, adopting variable frequency speed regulation technology can save 30%-50% energy for the water pump.

[0095] 3) Energy-saving measures at the user end

[0096] Promote intelligent temperature control devices: Install intelligent thermostats in users' rooms. Users can set the indoor temperature according to their own needs. When the indoor temperature reaches the set value, the thermostat automatically closes the valve and stops heating to avoid energy waste caused by too high indoor temperature. At the same time, the intelligent thermostat can also achieve time-sharing and room-by-room control, improving the comfort and energy-saving effect of heating.

[0097] Strengthen the publicity of users' energy-saving awareness: Through publicity and education, improve users' energy-saving awareness and let users develop good heating habits. For example, reasonably set the indoor temperature, avoid setting it too high; close the valve in time when heating is not needed; regularly clean the radiator to improve the heat dissipation effect, etc.

[0098] In this embodiment, the setting of the abnormal energy consumption analysis logic knowledge includes:

[0099] When the water consumption diagnosis is abnormal, judge whether the room temperature compliance rate of heat users is abnormal. If the room temperature compliance rate is abnormal, it indicates poor heating quality, and there may be behaviors such as water discharge and forced circulation by heat users, or there is a risk of heat network leakage, and heat network leakage diagnosis and positioning investigation are required; if the room temperature compliance rate is normal, there is a risk of heat network leakage, and heat network leakage diagnosis and positioning investigation are required;

[0100] When the heat consumption diagnosis is abnormal, judge whether the transmission efficiency of the primary pipe network is abnormal. If the transmission efficiency of the primary pipe network is abnormal, it indicates that the heat loss due to pipe heat dissipation is large, and there is a risk of old pipes or damaged pipe insulation layers; otherwise, there is hydraulic imbalance, resulting in thermal imbalance, or there are old residential areas, and the building envelope structure of old residential areas needs to be renovated;

[0101] When the power consumption diagnosis is abnormal, judge whether the water consumption is abnormal and whether the energy efficiency of the circulating pump is abnormal. If the water consumption is abnormal and the energy efficiency of the circulating pump is abnormal, it indicates excessive water loss, frequent start and stop of the makeup pump, and unreasonable selection of the circulating pump; if the water consumption is abnormal and the energy efficiency of the circulating pump is normal, it indicates more water loss, frequent start and stop of the makeup pump, and there is a risk of unreasonable selection of the circulating pump; if the water consumption is normal and the energy efficiency of the circulating pump is abnormal, the circulating pump is unreasonably selected; if the water consumption is normal and the energy efficiency of the circulating pump is normal, there is a risk of unreasonable selection of the circulating pump.

[0102] In this embodiment, the cloud processing layer is used to train a large model for abnormal energy consumption diagnosis of the heating system according to the predicted energy consumption value of the diagnosis object of the heating system and in combination with the preset energy consumption diagnosis standard value, including:

[0103] In the cloud processing layer, obtain the predicted energy consumption value of the diagnosis object of the heating system from the fog processing layer;

[0104] According to the design parameters, historical operation data and industry standards of the heating system, determine the energy consumption diagnosis standard values under different working conditions;

[0105] Compare the predicted energy consumption value with the preset energy consumption diagnosis standard value, judge whether there is abnormal energy consumption, and perform data annotation;

[0106] Integrate the predicted energy consumption value, the energy consumption diagnosis standard values under different working conditions, and the annotation data into a data set, and input it into a pre-trained large model, and use the method of combining the large model and prompt words to train a large model for abnormal energy consumption diagnosis of the heating system.

[0107] It should be noted that by analyzing historical operation data and classifying and sorting the data according to different working conditions, such as different outdoor temperatures, heating loads, etc., combined with design parameters and industry standards, a reasonable energy consumption range of the heating system under various working conditions is determined, which is used as the standard value for energy consumption diagnosis. For example, according to historical data statistics, within a certain outdoor temperature range, the normal heat consumption, power consumption, and water consumption ranges corresponding to different heating areas can be obtained.

[0108] Select a pre-trained large model: According to actual needs and computing resources, select a suitable pre-trained large model, such as some neural network models based on deep learning, such as recurrent neural network (RNN), long short-term memory network (LSTM), or Transformer, etc. These models have good performance in processing sequence data and complex non-linear relationships.

[0109] Design prompt words: According to the characteristics of the heating system and the requirements of energy consumption anomaly diagnosis, design a series of prompt words. The prompt words can include information such as the description of the working conditions of the heating system and the characteristics of the energy consumption data. By such prompt words, the model is guided to focus on key information and improve the accuracy of the model for energy consumption anomaly diagnosis.

[0110] Model training: Input the integrated dataset into the pre-trained large model, combined with the designed prompt words, and use optimization methods to train the model. During the training process, continuously adjust the parameters of the large model so that the large model can accurately output the diagnosis results of energy consumption anomalies, such as the types of energy consumption anomalies and the objects of anomaly diagnosis, according to data such as the predicted energy consumption values and the standard values for energy consumption diagnosis input. At the same time, evaluate and adjust the output of the large model according to the labeled data to improve the generalization ability and accuracy of the model. After multiple iterative trainings until the large model reaches satisfactory performance indicators.

[0111] In this embodiment, the setting of the prompt words includes:

[0112] Role setting: Clearly define that the large model is an expert in heating energy consumption anomaly diagnosis during this conversation process, and guide the large model to output the types of heating energy consumption anomalies and the objects of anomaly diagnosis; [[ID=##]] [[ID=##]]

[0113] Requirement description: Describe the task requirements completed by the large model, including: obtaining predicted energy consumption values, standard values for energy consumption diagnosis, and energy consumption anomaly diagnosis.

[0114] As Figure 3 shown, in this embodiment, the cloud processing layer is used to obtain the types of heating energy consumption anomalies and the objects of anomaly diagnosis according to the large model for heating energy consumption anomaly diagnosis, combine the knowledge graph of heating energy consumption anomaly analysis, perform question-and-answer intention recognition and construct an intelligent agent for heating energy consumption anomaly analysis based on a large language model, and generate answers for heating energy consumption anomaly analysis and energy-saving recommendation solutions, including:

[0115] Define the energy consumption anomaly diagnosis type, anomaly diagnosis object, historical energy consumption anomaly analysis Q&A, Q&A intention, and energy consumption anomaly analysis knowledge graph output by the large model for energy consumption anomaly diagnosis of the heating system as the environment. Changes in each variable will result in a different current energy consumption anomaly analysis environment;

[0116] Input the environmental information and preset energy consumption anomaly analysis prompt words into the large language model. The large language model uses its own language understanding and generation capabilities to construct a large model for energy consumption anomaly analysis and integrate it into the intelligent agent;

[0117] After the intelligent agent obtains historical energy consumption anomaly analysis Q&A, energy consumption anomaly diagnosis type, anomaly diagnosis object, energy consumption anomaly analysis knowledge extracted from the energy consumption anomaly analysis knowledge graph, and identified Q&A intention from the current energy consumption anomaly analysis environment, it performs analysis and reasoning according to the current environmental state to generate the action of the current environment, then verifies the generated action, generates the answers for energy consumption anomaly analysis and energy-saving recommendation solutions, and finally updates the current energy consumption anomaly analysis environment to achieve the iteration of multi-round Q&A.

[0118] It should be noted that the basic components of the intelligent agent include a sensor, a brain, and actions. The sensor mainly collects environmental information, the brain integrates and processes this information, then executes actions that match the current environment, and further affects the environment. Using an intelligent agent based on a large language model, the large model is used as the brain and has different states when the current Q&A intention is different. Finally, it analyzes and reasons the information of the current environment to make a scientific and reasonable decision.

[0119] In this embodiment, the generation of the action of the current environment after analysis and reasoning includes:

[0120] Assume that each energy consumption anomaly analysis Q&A contains corresponding Q&A intentions, including energy consumption anomaly symptoms, energy consumption anomaly causes, and energy-saving solutions;

[0121] After extracting the context semantic features in the energy consumption anomaly analysis Q&A sentence, predict the Q&A behavior and attribute information to obtain the overall Q&A intention;

[0122] According to the overall Q&A intention, extract the energy consumption anomaly diagnosis type and anomaly diagnosis object entities from the Q&A context, align them with the entities in the energy consumption anomaly analysis knowledge graph, and calculate the similarity to identify the list of energy consumption anomaly analysis entities with the highest similarity;

[0123] According to the identified list of energy consumption anomaly analysis entities, extract the corresponding entity and relationship information from the energy consumption anomaly analysis knowledge graph, and perform pruning of the knowledge subgraph. After retaining the subgraph information with strong relevance, transform it into the corresponding action of the current environment.

[0124] It should be noted that deep learning models, such as recurrent neural networks (RNNs), long short-term memory networks (LSTMs), or Transformers, etc., are used to extract context semantic features from Q&A sentences. These models can capture the semantic relationships and sequential information between words in a sentence. For example, the self-attention mechanism of Transformer can calculate the correlation between each word in a sentence and other words, thereby obtaining rich context semantic representations. The extracted context semantic features are input into a classifier or regression model to predict Q&A behaviors (such as informing, questioning, diagnosing, etc.) and attribute information (such as the type of energy consumption anomaly informed and the diagnostic object, the reasons for the energy consumption anomaly and the energy-saving solutions questioned, the reasons for the diagnosed anomaly, etc.).

[0125] Entity extraction: According to the overall intention of the Q&A, extract the energy consumption anomaly diagnosis types (such as heat consumption anomaly, power consumption anomaly, etc.) and the entity of the anomaly diagnosis object (such as a specific heat station, heat network, etc.) from the Q&A context. Named entity recognition technology can be used, combined with the professional knowledge and dictionaries in the heating field, to accurately identify these entities.

[0126] Alignment with knowledge graph entities: Align the extracted entities with the entities in the energy consumption anomaly analysis knowledge graph. By calculating the similarity in aspects such as entity names and attributes, find the entity in the knowledge graph that best matches the extracted entity. For example, string matching algorithms (such as edit distance) or semantic similarity calculation methods (such as cosine similarity based on word vectors) can be used to measure the similarity between entities.

[0127] Identify the list of entities with the highest similarity: Sort the similarities of the entities with the entities in the knowledge graph, and select several entities with the highest similarity to form a list of energy consumption anomaly analysis entities. These entities will serve as the basis for subsequent knowledge extraction.

[0128] Knowledge extraction: According to the identified list of energy consumption anomaly analysis entities, extract the corresponding entity and relationship information from the energy consumption anomaly analysis knowledge graph. This information may include the attributes of entities, the causal relationships between entities, and solutions, etc.

[0129] Knowledge subgraph pruning: Screen and prune the extracted knowledge, and retain the subgraph information with strong relevance to the Q&A intention. It can be judged according to factors such as the relationship weights between entities and the importance of entities in the knowledge graph. For example, if the Q&A intention is to explore the reasons for excessive heat consumption, then only retain the entity related to the reasons for excessive heat consumption and its relationship with the heat consumption anomaly entity.

[0130] Transformation into actions: Transform the pruned knowledge subgraph information into corresponding actions in the current environment. For example, if the knowledge subgraph shows that the reason for excessive heat consumption is fouling of the heat exchanger, the corresponding action could be to arrange for cleaning and maintenance of the heat exchanger; if the abnormal energy consumption is caused by unreasonable operating parameters, the action could be to adjust the relevant operating parameters.

[0131] In several embodiments provided by the present 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 systems, methods, and computer program products according to multiple embodiments of the present invention. In this regard, each block in the flowchart or block diagram may represent a module, a program segment, or a part of code, and the module, program segment, or part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than marked in the accompanying drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, as well as the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system for performing the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

[0132] In addition, each functional module in various embodiments of the present invention can be integrated together to form an independent part, or each module can exist alone, 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, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories, random access memories, magnetic disks, or optical discs that can store program codes.

[0133] Based on the above inspiration from the ideal embodiments of the present invention, through the above description, relevant staff can make various changes and modifications completely within the scope without departing from the technical idea of this invention. The technical scope of this invention is not limited to the content in the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A heating energy consumption diagnosis and analysis system integrating large language models and cloud-edge-fog collaboration, characterized in that, It includes: An edge processing layer for monitoring energy consumption - related data and heating operation data of the diagnosis object of the heating system; A fog processing layer for extracting strongly - related data features affecting energy consumption and establishing an energy consumption prediction model for the diagnosis object of the heating system; it is also used to establish a knowledge graph for energy consumption anomaly analysis based on heating operation data and energy consumption anomaly analysis mechanism knowledge; A cloud processing layer for training a large - model for heating system energy consumption anomaly diagnosis according to the predicted energy consumption value of the diagnosis object of the heating system and in combination with a preset energy consumption diagnosis standard value; it is also used to obtain the energy consumption anomaly diagnosis type and anomaly diagnosis object according to the large - model for heating system energy consumption anomaly diagnosis, and in combination with the knowledge graph for energy consumption anomaly analysis, to identify the question - answering intention and construct an energy consumption anomaly analysis question - answering intelligent agent based on a large - language model, and generate answers for energy consumption anomaly analysis and energy - saving recommendation solutions.

2. The heating energy consumption diagnosis and analysis system according to claim 1, wherein The edge processing layer for monitoring energy consumption - related data and heating operation data of the diagnosis object of the heating system includes: In the edge processing layer, through the set edge processing devices, monitor the energy consumption - related data and heating operation data of the heat networks and heat stations in different regions of the heating system, including the water consumption data, heat consumption data, and power consumption data of the heat networks and heat stations in different regions, as well as the supply - return water temperature, circulating water flow rate, make - up water volume, pump operation frequency, outdoor temperature, indoor temperature, building envelope insulation performance, pipe network insulation parameters, and heating area of the heat networks and heat stations; After pre - processing, data encapsulation, and parsing of the energy consumption - related data and heating operation data of the heat networks and heat stations in different regions, transmit them to the fog processing layer.

3. The heating energy consumption diagnosis and analysis system according to claim 1, wherein The edge processing layer, connected to the fog processing layer and the cloud processing layer, is used to obtain the energy consumption anomaly diagnosis type and anomaly diagnosis object transmitted by the cloud processing layer and display them to the heating operation and maintenance personnel; It is also used to transmit the energy consumption anomaly problems input by the heating operation and maintenance personnel to the cloud processing layer and obtain the answers for energy consumption anomaly analysis and energy - saving recommendation solutions transmitted by the cloud processing layer and display them to the heating operation and maintenance personnel; And, it is used to generate historical energy consumption curves in three time dimensions of daily, monthly, and heating season according to the historical energy consumption data of the heat networks and heat stations in different regions.

4. The heating energy consumption diagnosis and analysis system according to claim 1, wherein The fog processing layer for extracting relevant data features affecting energy consumption and establishing an energy consumption prediction model for the diagnosis object of the heating system includes: In the fog processing layer, for the energy consumption - related data and heating operation data of the diagnosis object of the heating system transmitted by the edge processing layer, use a deep convolutional neural network to extract the water consumption, heat consumption, and power consumption features of the heat networks and heat stations in different regions, then capture the global correlation between different time points and different energy consumption types through the self - attention mechanism in the Transformer model, and then learn the mapping relationship between the energy consumption feature vector and the energy consumption value through a fully - connected layer to establish an energy consumption prediction model for the diagnosis object of the heating system, and output the predicted values of water consumption, heat consumption, and power consumption of the heat networks and heat stations in different regions in the future time period; Among them, the fog processing layer deploys the established energy consumption prediction model for the diagnosis object of the heating system to the server of the fog processing layer or distributes it to the corresponding edge processing devices of the edge processing layer for real - time energy consumption prediction.

5. The heating energy consumption diagnosis and analysis system according to claim 1, wherein The fog processing layer is also used to establish a knowledge graph for abnormal energy consumption analysis based on heating operation data and the knowledge of the abnormal energy consumption analysis mechanism, including: In the fog processing layer, obtain the heating operation data of the heat supply network and heat substations in different regions of the heating system during a historical period, which reflects the operating status under different operating conditions of the system; Obtain theoretical knowledge on abnormal water consumption, abnormal heat consumption, and abnormal power consumption analysis from papers, academic reports, and industry standards, including the classification of abnormal energy consumption, the causes of abnormalities, and energy-saving solutions; the energy-saving solutions include equipment repair and transformation, optimization of heating dispatching strategies, and adjustment of equipment operating parameters; Communicate with experts in the heating field to obtain empirical knowledge on abnormal energy consumption analysis in actual work, including the manifestations of abnormal energy consumption caused by different faults and the judgment of abnormal energy consumption through changes in operating parameters; Set the logical knowledge of abnormal energy consumption analysis, including: the logical knowledge of analysis when water consumption diagnosis is abnormal, heat consumption diagnosis is abnormal, and power consumption diagnosis is abnormal; Identify various equipment entity, operating parameter entity, abnormal energy consumption type entity, cause of abnormality entity, and energy-saving solution entity from the heating operation data and the obtained theoretical knowledge, empirical knowledge, and logical knowledge on abnormal water consumption, abnormal heat consumption, and abnormal power consumption analysis; Determine the relationship between equipment and operating parameters, the correlation between changes in operating parameters and abnormal energy consumption, the causal relationship between abnormal energy consumption types and the causes of such abnormalities, and the relationship between the causes of abnormal energy consumption and solutions; After knowledge fusion based on the identified entities and determined relationships, add the extracted and fused entities as nodes to the knowledge graph, and create corresponding edges according to the relationships between the entities to establish a knowledge graph for abnormal energy consumption analysis.

6. The heating energy consumption diagnosis and analysis system according to claim 5, characterized in that, The setting of the logical knowledge of abnormal energy consumption analysis includes: When the water consumption diagnosis is abnormal, judge whether the room temperature compliance rate of heat users is abnormal. If the room temperature compliance rate is abnormal, it indicates poor heating quality, and heat users may have behaviors such as forced circulation by draining water or there is a risk of heat network leakage, and heat network leakage diagnosis and location investigation are required; if the room temperature compliance rate is normal, there is a risk of heat network leakage, and heat network leakage diagnosis and location investigation are required; When the heat consumption diagnosis is abnormal, judge whether the transmission efficiency of the primary pipeline network is abnormal. If the transmission efficiency of the primary pipeline network is abnormal, it indicates that the heat loss through the pipeline is large, and there is a risk of old pipelines or damaged pipeline insulation layers; otherwise, there is hydraulic imbalance, resulting in thermal imbalance, or there are old residential areas, and the building envelope structure of old residential areas needs to be renovated; When the power consumption diagnosis is abnormal, judge whether the water consumption is abnormal and whether the energy efficiency of the circulating water pump is abnormal. If the water consumption is abnormal and the energy efficiency of the circulating water pump is abnormal, it indicates excessive water loss, frequent start and stop of the make-up water pump, and unreasonable selection of the circulating water pump; if the water consumption is abnormal and the energy efficiency of the circulating water pump is normal, it indicates more water loss, frequent start and stop of the make-up water pump, and there is a risk of unreasonable selection of the circulating water pump; if the water consumption is normal and the energy efficiency of the circulating water pump is abnormal, the circulating water pump is unreasonably selected; if the water consumption is normal and the energy efficiency of the circulating water pump is normal, there is a risk of unreasonable selection of the circulating water pump.

7. The heating energy consumption diagnosis and analysis system according to claim 1, characterized in that The cloud processing layer is used to train a large model for diagnosing abnormal energy consumption in the heating system based on the predicted energy consumption value of the diagnosis object in the heating system and in combination with a preset standard value for energy consumption diagnosis, including: In the cloud processing layer, obtain the predicted energy consumption value of the diagnosis object in the heating system from the fog processing layer; Determine the standard values for energy consumption diagnosis under different working conditions according to the design parameters, historical operation data, and industry standards of the heating system; Compare the predicted energy consumption value with the preset standard value for energy consumption diagnosis, determine whether there is abnormal energy consumption, and perform data annotation; Integrate the predicted energy consumption value, the standard values for energy consumption diagnosis under different working conditions, and the annotation data into a data set, and input it into a pre-trained large model. Use a method combining the large model and prompt words to train a large model for diagnosing abnormal energy consumption in the heating system.

8. The heating energy consumption diagnosis and analysis system according to claim 7, characterized in that The setting of the prompt words includes: Role setting: Clearly define that the large model is an expert in diagnosing abnormal heating energy consumption during this conversation process, and guide the large model to output the type of abnormal energy consumption diagnosis and the object of abnormal diagnosis; Requirement description: Describe the task requirements completed by the large model, including: obtaining the predicted energy consumption value, the standard value for energy consumption diagnosis, and the diagnosis of abnormal energy consumption.

9. The heating energy consumption diagnosis and analysis system according to claim 1, characterized in that The cloud processing layer is used to obtain the type of abnormal energy consumption diagnosis and the object of abnormal diagnosis based on the large model for diagnosing abnormal energy consumption in the heating system, and in combination with the knowledge graph of abnormal energy consumption analysis, perform question-and-answer intention recognition and construct an intelligent agent for abnormal energy consumption analysis based on the large language model, and generate answers for abnormal energy consumption analysis and energy-saving recommendation solutions, including: Define the type of abnormal energy consumption diagnosis, the object of abnormal diagnosis, historical abnormal energy consumption analysis questions and answers, question-and-answer intention, and the knowledge graph of abnormal energy consumption analysis output by the large model for diagnosing abnormal energy consumption in the heating system as the environment, where the change of each variable will result in a different current abnormal energy consumption analysis environment; Input the environmental information and the preset prompt words for abnormal energy consumption analysis into the large language model. The large language model uses its own language understanding and generation capabilities to construct a large model for abnormal energy consumption analysis and integrate it into the intelligent agent; After the intelligent agent obtains historical abnormal energy consumption analysis questions and answers, the type of abnormal energy consumption diagnosis, the object of abnormal diagnosis, the knowledge of abnormal energy consumption analysis extracted from the knowledge graph of abnormal energy consumption analysis, and the recognized question-and-answer intention from the current abnormal energy consumption analysis environment, it performs analysis and reasoning according to the current environmental state to generate the action of the current environment, then verifies the generated action, generates answers for abnormal energy consumption analysis and energy-saving recommendation solutions, and finally updates the current abnormal energy consumption analysis environment to achieve iterative multi-round question and answer.

10. The heating energy consumption diagnosis and analysis system according to claim 9, characterized in that The generation of the action of the current environment after performing analysis and reasoning includes: Assume that each abnormal energy consumption analysis question and answer contains corresponding question-and-answer intentions, including abnormal energy consumption symptoms, reasons for abnormal energy consumption, and energy-saving solutions; After extracting the context semantic features in the abnormal energy consumption analysis question-and-answer sentence, predict the question-and-answer behavior and attribute information to obtain the overall question-and-answer intention; According to the overall question-and-answer intention, extract the entities of the type of abnormal energy consumption diagnosis and the object of abnormal diagnosis from the question-and-answer context, align them with the entities in the knowledge graph of abnormal energy consumption analysis, and calculate the similarity to identify the list of entities of abnormal energy consumption analysis with the highest similarity; According to the list of entities analyzed for abnormal energy consumption identified, extract the corresponding entity and relationship information from the knowledge graph of abnormal energy consumption analysis, and perform pruning of the knowledge subgraph. After retaining the subgraph information with strong relevance, convert it into actions corresponding to the current environment.

Images