Energy carbon data analysis method and system, electronic equipment and computer storage medium

Abstract

The invention discloses an energy carbon data analysis method and system, electronic equipment and a computer storage medium, and belongs to the technical field of large models. The method comprises the following steps: receiving an energy carbon data analysis request of a user side; analyzing the energy-carbon data analysis request through a pre-constructed large language model to obtain an overall task corresponding to the energy-carbon data analysis request; decomposing the overall task into a plurality of sub-tasks, wherein each sub-task processes data analysis in a specific field; calling a corresponding sub-model for each sub-task, and obtaining an intermediate result output by the sub-model; and fusing the intermediate results to obtain an energy-carbon data analysis result corresponding to the energy-carbon data analysis request. Through the method, the technical problems that a traditional decision analysis method is single in data source and limited in dimension are solved.

Description

Technical Field

[0001] The present application relates to the field of large model technology, and in particular to an energy-carbon data analysis method, system, electronic device and computer storage medium. Background Art

[0002] Energy-carbon data refers to energy and carbon emission data. The energy dual-carbon decision-making system is an integrated platform or tool designed to help enterprises, governments and other organizations make decisions about energy management and carbon emissions. Among them, "dual carbon" is the abbreviation of carbon peak and carbon neutrality.

[0003] In the process of pursuing the "dual carbon" goals, traditional decision-making analysis methods have a single data source and limited dimensions. Specifically: traditional decision-making analysis methods mainly rely on energy data or carbon emission data in a single field, ignoring the comprehensive consideration of related factors such as transportation and meteorology, making it difficult to comprehensively assess the impact, resulting in a lack of integrity in dual carbon decisions.

[0004] With respect to the above-mentioned problems existing in the related technologies, no effective solutions have been found so far. Summary of the invention

[0005] The present application provides an energy carbon data analysis method, system, electronic device and computer storage medium to solve the above-mentioned technical problems existing in the related technologies.

[0006] According to an embodiment of the present application, a method for analyzing energy carbon data is provided, comprising: receiving an energy carbon data analysis request from a user terminal; parsing the energy carbon data analysis request through a pre-built large language model to obtain an overall task corresponding to the energy carbon data analysis request; decomposing the overall task into multiple subtasks, each subtask processing data analysis within a specific field; calling the corresponding sub-model for each subtask to obtain an intermediate result output by the sub-model; and fusing the intermediate results to obtain an energy carbon data analysis result corresponding to the energy carbon data analysis request.

[0007] According to another embodiment of the present application, an energy carbon data analysis system is provided, including: a receiving module for receiving an energy carbon data analysis request from a user terminal; a parsing module for parsing the energy carbon data analysis request through a pre-built large language model to obtain an overall task corresponding to the energy carbon data analysis request; a decomposition module for decomposing the overall task into multiple subtasks, each subtask processing data analysis within a specific field; a scheduling module for calling the corresponding sub-model for each subtask to obtain an intermediate result output by the sub-model; and a fusion module for performing weighted fusion on the intermediate results to obtain an energy carbon analysis result corresponding to the energy carbon data analysis request.

[0008] According to another embodiment of the present application, a computer storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above-mentioned device embodiments when running.

[0009] According to another embodiment of the present application, there is also provided an electronic device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; wherein: the memory is used to store computer programs; the processor is used to execute the steps in the above method by running the program stored in the memory.

[0010] According to yet another embodiment of the present application, a computer program product comprising instructions is also provided, which enables the computer to execute the steps in the above method when the computer program product is executed on the computer.

[0011] The above technical solution provided by the embodiment of the present application has the following advantages compared with the prior art:

[0012] (1) Fusion of multi-source heterogeneous data and comprehensive analysis of influencing factors. It is able to collect multi-dimensional data such as electricity, carbon emissions, meteorology, and transportation across fields, and comprehensively analyze various influencing factors, greatly improving the integrity and foresight of decision-making, and avoiding the one-sidedness of decision-making caused by the single data source of traditional methods.

[0013] (2) Multi-scale intelligent modeling to improve computational complexity processing capabilities. The integration of large models and specialized sub-models can efficiently decompose complex decision-making problems, divide and conquer distributed modeling and calculation, and use large models for intelligent scheduling and result fusion, which has obvious advantages in dealing with high-complexity problems.

[0014] (3) Closed-loop optimization feedback to achieve continuous improvement in decision-making quality. By tracking the actual decision-making execution results and feeding them back to model training, a self-iterative, continuously optimized closed-loop mechanism is formed to ensure that the system decision quality can continue to improve over time and have a lasting competitive advantage.

[0015] (4) Human-machine collaborative decision-making improves the transparency and explainability of decision-making. It supports natural language interaction with users. Users can actively modify constraints, query details, etc. The system decision-making process has strong explainability and transparency, which is conducive to improving the trust of decision makers. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0017] Figure 1 is a hardware structure block diagram of a computer according to an embodiment of the present application;

[0018] Figure 2 is a flow chart of an energy carbon data analysis method according to an embodiment of the present application;

[0019] Figure 3 It is a schematic diagram of the implementation process of energy dual-carbon decision-making in an embodiment of the present application;

[0020] Figure 4 It is a structural block diagram of an energy carbon data analysis device according to an embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only embodiments of a part of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present application. It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0022] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0023] Example 1

[0024] The method embodiment provided in the first embodiment of the present application can be executed in a mobile phone, a computer, a tablet or a similar computing device. Taking running on a computer as an example, Figure 1 is a hardware structure block diagram of a computer in an embodiment of the present application. Figure 1 As shown, the computer may include one or more ( Figure 1Only one is shown in the figure) a processor 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA) and a memory 104 for storing data. Optionally, the computer may also include a transmission device 106 and an input / output device 108 for communication functions. It can be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above-mentioned computer. Figure 1 More or fewer components as shown, or with Figure 1 Different configurations are shown.

[0025] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as a computer program corresponding to an energy carbon data analysis method in an embodiment of the present application. The processor 102 executes various functional applications and data processing by running the computer program stored in the memory 104, that is, to implement the above method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 104 may further include a memory remotely arranged relative to the processor 102, and these remote memories may be connected to the computer via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0026] The transmission device 106 is used to receive or send data via a network. The specific example of the above network may include a wireless network provided by a communication provider of the computer. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, referred to as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0027] In this embodiment, a method for analyzing energy-carbon data is provided. Figure 2 is a flow chart of an energy carbon data analysis method according to an embodiment of the present application, such as Figure 2 As shown, the process includes the following steps:

[0028] Step S10, receiving an energy carbon data analysis request from a user terminal;

[0029] Step S20, parsing the energy carbon data analysis request through a pre-built large language model to obtain an overall task corresponding to the energy carbon data analysis request;

[0030] In this embodiment, energy-carbon data includes energy data and carbon emission data, and energy-carbon data analysis requests include but are not limited to analysis and prediction of energy data and carbon emission data. After receiving the energy-carbon data analysis request from the user, the large language model (hereinafter referred to as the large model) performs semantic parsing and intent recognition on the energy-carbon data analysis request. The large model uses its language understanding ability learned in the pre-training stage to map the user's natural language request to a standardized task description. In one example, the user enters a natural language query statement "Predict the power generation in the next five years". The large model performs semantic understanding of the query statement and identifies the overall task as "long-term power generation forecast."

[0031] Step S30, decomposing the overall task into multiple subtasks, each subtask processing data analysis in a specific field;

[0032] Step S40, calling the corresponding sub-model for each sub-task to obtain the intermediate result output by the sub-model;

[0033] Step S50, integrating the intermediate results to obtain the energy-carbon data analysis result corresponding to the energy-carbon data analysis request.

[0034] According to the standardized task description, the big model decomposes the overall task into multiple subtasks, each of which processes data analysis in a specific field. For example, the prediction task "long-term power generation forecast" in the above example is decomposed into future power load forecast, power investment planning optimization, climate scenario simulation, power generation carbon emission estimation and other subtasks. The big model dispatches the corresponding submodels to perform these subtasks. For example, the load forecast submodel loads historical load data, meteorological data, population and economic data, and predicts the power load curve for the next five years; the power planning submodel loads power plant investment costs, equipment parameters, policy constraints and other data to optimize the power supply investment plan; the climate submodel loads climate change scenario data and satellite remote sensing images to simulate the meteorological conditions for the next five years; the carbon emission submodel loads fossil fuel process parameters and operation data to estimate the carbon emissions of power generation in the next year. After each submodel completes the analysis and calculation, the results are output to the big model. The big model uses the attention fusion mechanism to cross-reason the prediction results of multiple dimensions such as power load, power planning, climate scenarios, and carbon emissions to obtain a comprehensive forecast value of total power generation in the next five years.

[0035] Through this embodiment, a comprehensive analysis is conducted by comprehensively considering multiple fields and dimensions, which improves the integrity and foresight of decision-making. The large model and sub-models are integrated to realize multi-scale modeling, decompose complex problems, and coordinate decision-making by scheduling multiple specialized sub-models through the large model, thereby improving analysis capabilities and decision-making quality.

[0036] In one implementation of this embodiment, decomposing the overall task into multiple subtasks includes:

[0037] S31, calling a pre-built task decomposition knowledge base, wherein the task decomposition knowledge base stores the hierarchical structure of tasks and the correlation and dependency between tasks;

[0038] S32, based on the hierarchical structure of the tasks and the relevance and dependency between the tasks, decompose the overall task into a plurality of subtasks.

[0039] The overall task of energy dual-carbon decision-making involves many considerations. In this implementation, the overall task is decomposed into multiple sub-tasks based on the hierarchical structure, relevance and dependency between tasks, effectively mining the deep correlations in massive and complex data, and improving analysis capabilities and decision-making quality.

[0040] In this embodiment, a multi-knowledge source fusion framework for task decomposition is pre-built: the hierarchical structure and logical constraints of the task are established using the expert knowledge graph, and the correlation and dependency between tasks are discovered through data mining to form a complementary task decomposition knowledge base. Based on the task decomposition knowledge base, the optimal decomposition scheme for the overall task is automatically inferred through heuristic search and constraint satisfaction. Among them, heuristic search helps to find the optimal solution among many possible decomposition schemes, and constraint satisfaction ensures that the decomposition scheme meets all logical and practical constraints. In this embodiment, the decomposition scheme can be adjusted in real time according to the update of the knowledge base. With the continuous update of the knowledge base, such as the addition of new data, changes in expert opinions, etc., the decomposition scheme can be adjusted in real time to maintain its optimal state.

[0041] In another implementation of this embodiment, calling a corresponding sub-model for each sub-task includes:

[0042] S41, analyzing the dependency relationship and data flow between the subtasks to form a task flow of a directed acyclic graph;

[0043] S421, for subtasks with dependency relationships, distribute the preceding subtask to the corresponding first submodel according to the data flow direction of the task flow, obtain the intermediate result output by the first submodel, and use the intermediate result as the input of the second submodel corresponding to the subsequent subtask;

[0044] S422: For multiple independent subtasks, schedule the third sub-models corresponding to the independent subtasks in parallel.

[0045] In this embodiment, the sub-model is adaptively constructed, and the appropriate model type, structure and parameters are automatically selected according to the decomposed sub-tasks. Through meta-learning, transfer learning, etc., based on the general features learned from the large model, a sub-model that adapts to the specific task is quickly constructed to improve the efficiency and flexibility of sub-model construction. Optionally, this embodiment can also construct a closed-loop update mechanism for models, data and knowledge. The operating data of the sub-model is continuously tracked, and new task patterns and associations are mined to update the task decomposition knowledge base. The expert's optimization feedback on the task tree is used as new structured knowledge and supplemented to the knowledge graph. Through the cyclic iteration of models, data and knowledge, the continuous evolution of task decomposition and sub-model construction is achieved.

[0046] The big model analyzes the dependencies and data flows between subtasks to form a directed acyclic graph (DAG) task flow. Each subtask node corresponds to a submodel to be called. Optionally, in actual applications, the big model inserts necessary data format conversion operations into the task flow according to the input and output interface definitions of the submodel to ensure data consistency and compatibility.

[0047] The big model distributes subtasks to the corresponding submodels. After the submodel completes the calculation, it returns the intermediate results to the big model. For subtasks with dependencies, the intermediate results of the previous subtask are passed to the submodel corresponding to the next subsequent subtask. For independent subtasks, the big model can use parallel scheduling to improve efficiency.

[0048] In order to improve data transmission efficiency, in this embodiment, a data cache is maintained in the memory or disk of the large model to store frequently accessed or reused intermediate results. The large model dynamically adjusts the cache strategy to balance the timeliness of data and storage overhead.

[0049] In this implementation, after calling the corresponding sub-model for each sub-task, the method further includes:

[0050] S43, monitoring the operating status of each sub-model in real time;

[0051] S44: if it is detected that the target sub-model is abnormal or exceeds the limit, the sub-task corresponding to the target sub-model is rescheduled, or the backup sub-model of the target sub-model is called.

[0052] The large model monitors the running status of each sub-model in real time. If a sub-model is found to be abnormal or timed out, the sub-task will be rescheduled in time, or the backup sub-model will be called according to the redundancy strategy to ensure the fault tolerance and robustness of the entire task flow.

[0053] This implementation method realizes task decomposition and sub-model construction through expert knowledge graph and data-driven, decomposing specialized tasks such as power load forecasting, energy planning optimization, carbon emission estimation, etc. for different decision-making goals and links; building and training corresponding sub-models, such as time series forecasting model, optimization solution model, carbon emission model, etc. Sub-models can be flexibly combined and reused while maintaining professional division of labor, with stronger modeling and prediction capabilities.

[0054] In another implementation of this embodiment, the intermediate results are merged to obtain the energy-carbon analysis result corresponding to the energy-carbon data analysis request, including:

[0055] S51, obtaining weights corresponding to the intermediate results, and performing weighted summation on the intermediate results based on the weights to obtain a fusion result;

[0056] S52, generating an explanatory text of the fusion result through the large language model.

[0057] In this embodiment, after all sub-models complete the calculation, the large model collects the intermediate results output by the sub-models. The intermediate results output by each sub-model may include data in different forms such as predicted values, confidence levels, and feature vectors. Therefore, the large model formats these heterogeneous data in a unified manner and constructs a result matrix. The result matrix is ​​subjected to correlation analysis, and indicators such as correlation coefficients or mutual information between the intermediate results output by different sub-models are calculated. These indicators reflect the similarity and redundancy of the sub-model outputs. Using the results of the correlation analysis, the large model can identify highly correlated sub-models to avoid excessive inclination toward these correlated outputs during fusion.

[0058] Optionally, a weighted fusion model can be constructed through the large model to learn the weights of each sub-model output. The weights reflect the contribution of different sub-models to the final decision. The large model trains the weighted fusion model based on historical decision data, and its optimization goal is to minimize the difference between the fusion result and the actual optimal decision. Commonly used weight learning algorithms include linear regression, logistic regression, support vector machine, etc. In the actual decision-making process, the large model dynamically adjusts the weights of the sub-model outputs.

[0059] The big model performs weighted summation of the outputs of the sub-models according to the learned weights to obtain the fusion result. Then, the big model calibrates the fusion result. The calibration process refers to domain knowledge and constraints to make necessary corrections and normalization to the fusion result. For example, for energy scheduling decisions, the fusion result needs to meet physical constraints such as power balance and upper and lower limits of unit output.

[0060] In this implementation, generating the explanatory text of the fusion result by the large language model includes:

[0061] S521, extracting key information from each of the intermediate results;

[0062] S522, analyzing the logical association between the key information;

[0063] S523, converting the key information and the logical associations between the key information into natural language expressions;

[0064] S524: Use the natural language expression as an explanatory text of the fusion result.

[0065] In this embodiment, the explanatory text is used to explain the contribution of each sub-model and the key influencing factors of the fusion result. This interpretability analysis helps decision makers understand the decision-making process and improve their trust and acceptance of the decision results. Extract key information from the output of each sub-model, for example, extract the installed capacity change trend in the next five years from the power planning model, extract the growth rate of electricity consumption from the load forecasting model, and extract the change trend of total power generation from the power generation simulation model. These key information serve as the basic material for explanatory explanations. Then use the big model to perform causal reasoning on the extracted information. The big model analyzes the logical association between different factors based on the energy field knowledge learned in the pre-training stage. For example, the big model can infer the causal relationship between "decline in coal-fired power installed capacity" and "tightening of coal policies", or the causal relationship between "increase in total power generation" and "increase in electricity demand". This step gives the extracted information a logical link, making it a coherent explanatory explanation. After obtaining the key information and its causal logic, the dialogue model is used to convert it into a natural language expression. The dialogue model takes information and logic as input to generate explanatory text that is easy for humans to understand. During the generation process, the dialogue model can refer to the corpus and writing samples in the energy field, use professional terms and writing styles, and improve the readability and professionalism of the explanatory description. The system automatically generates explanatory descriptions of the forecast results through the dialogue model, such as "According to analysis, in the next five years, coal-fired power generation will continue to decline slowly, but total power generation will still increase slightly", and intuitively presents the predicted power generation trend in the form of a curve chart.

[0066] In this implementation, users can continue to query details, such as "Why will coal-fired power generation decline?" and "Which new energy sources will increase significantly?" The dialogue model will give an explanatory response based on the calculation results of each sub-model. Users can also modify certain constraints, such as the growth rate of civil electricity, policy red lines, etc., and the system will recalculate and update the forecast results.

[0067] At the same time, the system will record user feedback and perform error analysis and parameter optimization on the model after obtaining actual power generation data to improve the accuracy of the next round of predictions. The edge nodes deployed on site will present the above prediction process and key results in real time on the monitoring screen to assist decision makers; at the same time, the process data will be uploaded to the cloud in real time for incremental model learning.

[0068] In one implementation of this embodiment, before parsing the data analysis request using the pre-built large language model, the method further includes:

[0069] Step A1, collecting multi-source heterogeneous data;

[0070] In this embodiment, multi-source heterogeneous data refers to data from multiple fields, such as electricity (such as grid load, power generation), carbon emissions, meteorology, and transportation. Optionally, the multi-source heterogeneous data can be collected through IoT sensing devices and external data interfaces. To facilitate subsequent data analysis, the collected raw data can also be subjected to data cleaning preprocessing such as denoising, missing fill, and format normalization.

[0071] Step A2, based on the multi-source heterogeneous data, performing self-supervisory training on the initial model to obtain an intermediate model;

[0072] Step A3, summarizing each domain-specific data set in the multi-source heterogeneous data, and using the domain-specific data set to perform domain-incremental fine-tuning on the intermediate model to obtain a large language model.

[0073] Based on massive historical multi-source heterogeneous data, the initial model (the initial model is a large-scale pre-trained model) is trained by self-supervised learning to obtain an intermediate model, so that the intermediate model can obtain a general pattern representation capability. The data sets in specific fields such as power grid load and carbon emission greenhouse gas concentration are aggregated, and supervised fine-tuning and transfer learning are performed on the data sets in specific fields to form a special large language model in the field of energy dual carbon.

[0074] Optionally, the specific method of constructing the large language model includes:

[0075] First, the energy-carbon knowledge graph is integrated with multi-source heterogeneous data to build a domain-enhanced pre-training dataset. The fusion process includes: converting structured information such as entities and relationships in the energy-carbon knowledge graph into text sequences, and combining text sequences with unstructured data from multi-source heterogeneous data (such as power generation operation logs, meteorological data, etc.) to form a multimodal pre-training corpus.

[0076] Secondly, pre-design self-supervision tasks with dual-carbon energy characteristics, such as power load curve generation tasks, abnormal operating condition identification tasks, etc. Use these self-supervision tasks to train the pre-training model, so that the model can contact and learn typical patterns and physical laws in the field during the pre-training stage, and improve the model's adaptability to dual-carbon energy scenarios.

[0077] Thirdly, a progressive domain fine-tuning strategy is adopted. Various domain-specific data sets in multi-source heterogeneous data are aggregated. In one example, the progressive domain-specific data sets include general domain corpus, energy domain corpus, and vertical domain corpus (such as power generation operation, carbon emissions, etc.). First, the general domain corpus is used to perform preliminary fine-tuning on the intermediate model, then the energy domain corpus is used for intermediate fine-tuning, and finally the vertical domain data is used for fine fine-tuning. By gradually adapting to domain knowledge and data distribution, a specialized energy dual-carbon large language model is obtained.

[0078] Finally, we continue to collect new data in the field of energy and carbon emissions, and conduct incremental training and online updates on the large language model. At the same time, we monitor the performance indicators of the large language model on key tasks to identify whether the large language model needs to be retrained or parameter adjustment. Through continuous learning and monitoring feedback, we can maintain the leading and robustness of the large language model.

[0079] The solution of this embodiment provides an energy carbon data analysis method. Figure 3 It is an implementation flow chart of the embodiment of the present application, including:

[0080] The first step is to collect and clean multi-dimensional heterogeneous data such as electricity, carbon emissions, meteorology, and transportation.

[0081] The second step is to build a large model dedicated to the energy and dual carbon field based on the field data set.

[0082] The third step is to decompose dedicated tasks and build and train domain sub-models based on expert knowledge graphs and data-driven methods for different decision-making objectives.

[0083] The fourth step is to build a multi-model collaborative decision-making framework based on the big model. The big model acts as the master control scheduler to assign tasks to each sub-model for distributed modeling and calculation. The large model performs weighted fusion on the sub-model outputs to obtain the final overall decision result.

[0084] This can be followed by:

[0085] Step 5: Simulation verification and online deployment: Based on typical business scenarios, build a simulation test environment to verify the overall performance of the framework; deploy the offline trained model in the online system, and continuously track the actual decision-making effects to obtain real data feedback.

[0086] The sixth step is human-computer collaborative optimization: Based on the dialogue model, realize user natural language interaction, understand the decision-making background, and determine the optimization goals and constraints; provide a visual display of the decision results and an explanation of the reasons; record human-computer interaction and the execution of decision-making plans as the data basis for future model optimization and iteration.

[0087] The seventh step is continuous online learning and updating: establish a computing architecture based on cloud-edge collaboration, perform computing-intensive training and model updates in the cloud; send the optimized model increments to the edge nodes to achieve online updates of decisions; the edge nodes process online data and continuously provide feedback to form a closed-loop optimized online learning mechanism. Among them, the model deployment method includes: the model is encapsulated in the form of a container (such as Docker) to decouple it from the operating environment. The cloud sends the containerized model image to the edge node. The edge node manages and schedules the model container through a container orchestration tool (such as Kubernetes). Containerized deployment allows the model to run flexibly across platforms and environments, while facilitating version management and rapid updates.

[0088] The cloud-edge collaboration mechanism includes: the edge node regularly initiates model update requests to the cloud to obtain the latest model version. At the same time, after the model update is completed, the cloud actively pushes update notifications and issues instructions to the edge node. This two-way trigger mechanism ensures the timeliness of model updates. The edge node continuously collects local data and regularly transmits the data back to the cloud. The cloud aggregates data from different edge nodes and retrains and optimizes the model based on incremental learning. Incremental learning enables the model to continuously learn from new data and continuously improve performance. At the same time, this also reduces the pressure on the cloud to process large quantities of data at one time. The cloud monitors the workload and response time of each edge node. Based on real-time performance indicators, the cloud dynamically adjusts the scheduling strategy of the task, assigns the task to the node with lighter load, and achieves load balancing. In addition, the scheduling strategy will also consider factors such as the geographical location of the node and network latency, and try to assign the task to the optimal node for execution.

[0089] This embodiment scheme implements the concepts of multi-source data drive, multi-granularity model fusion, and human-machine collaborative optimization, and can provide an end-to-end intelligent solution for energy dual-carbon decision-making.

[0090] The energy-carbon data analysis method proposed in this embodiment can be widely used in the following scenarios: (1) Energy system optimization and scheduling. As a core application scenario, it can be used for intelligent production scheduling decisions of traditional and new energy systems such as thermal power plants, wind farms, and photovoltaic power stations. By comprehensively analyzing multi-dimensional influencing factors, it can optimize energy utilization efficiency and reduce carbon emission intensity. (2) Urban dual-carbon planning decisions. It can provide decision-making support for dual-carbon planning in urban energy, transportation, environment and other fields, assist in the formulation of scientific low-carbon city development plans, and guide urban energy structure transformation and energy conservation and emission reduction. (3) Industrial park dual-carbon management and control. It is suitable for high-emission industrial parks such as petrochemicals, metallurgy, and building materials. It integrates data from multiple industries and enterprises to provide overall low-carbon transformation and energy-saving diagnosis decisions at the park level. (4) Enterprise dual-carbon asset management. Enterprises can use this method to evaluate the carbon footprint of their own production and operations, reasonably formulate energy-saving and emission reduction plans, and manage greenhouse gas emissions rights and interests.

[0091] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0092] Example 2

[0093] In this embodiment, an energy carbon data analysis system is also provided to implement the above embodiments and preferred implementations, which have been described and will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0094] Figure 4 is a structural block diagram of an energy carbon data analysis system according to an embodiment of the present application, such as Figure 4 As shown, the system includes:

[0095] The receiving module 40 is used to receive an energy and carbon data analysis request from a user terminal;

[0096] A parsing module 41 is used to parse the energy-carbon data analysis request through a pre-built large language model to obtain an overall task corresponding to the energy-carbon data analysis request;

[0097] A decomposition module 42, for decomposing the overall task into a plurality of subtasks, each subtask processing data analysis in a specific field;

[0098] The scheduling module 43 is used to call the corresponding sub-model for each sub-task and obtain the intermediate result output by the sub-model;

[0099] The fusion module 44 is used to perform weighted fusion on the intermediate results to obtain the energy-carbon analysis result corresponding to the energy-carbon data analysis request.

[0100] It should be noted that the above modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to: the above modules are all located in the same processor; or the above modules are located in different processors in any combination.

[0101] Example 3

[0102] An embodiment of the present application further provides a computer storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0103] Optionally, in this embodiment, the computer storage medium may be configured to store a computer program for performing the following steps:

[0104] S1, receiving the energy and carbon data analysis request from the user;

[0105] S2, parsing the energy carbon data analysis request through a pre-built large language model to obtain an overall task corresponding to the energy carbon data analysis request;

[0106] S3, decomposing the overall task into multiple subtasks, each subtask processing data analysis in a specific field;

[0107] S4, calling the corresponding sub-model for each sub-task to obtain the intermediate result output by the sub-model;

[0108] S5, integrating the intermediate results to obtain the energy-carbon data analysis result corresponding to the energy-carbon data analysis request.

[0109] Optionally, in this embodiment, the above-mentioned computer storage medium may include but is not limited to: a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and other media that can store computer programs.

[0110] An embodiment of the present application further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0111] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0112] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments and optional implementation modes, and this embodiment will not be described in detail here.

[0113] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0114] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0115] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0116] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0117] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0118] If the integrated unit is implemented in the form of a software functional unit 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 application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a computer storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned computer storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, disk or CD-ROM and other media that can store program codes.

[0119] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for analyzing energy and carbon data, It is characterized in that The method comprises: Receive energy and carbon data analysis requests from the user; Parsing the energy-carbon data analysis request through a pre-built large language model to obtain an overall task corresponding to the energy-carbon data analysis request; Decomposing the overall task into a plurality of subtasks, each subtask dealing with data analysis within a specific domain; For each subtask, the corresponding sub-model is called to obtain the intermediate result output by the sub-model; The intermediate results are integrated to obtain the energy-carbon data analysis result corresponding to the energy-carbon data analysis request.

2. The method according to claim 1, It is characterized in that Decomposing the overall task into multiple subtasks includes: Calling a pre-built task decomposition knowledge base, wherein the task decomposition knowledge base stores the hierarchical structure of tasks and the correlation and dependency between tasks; Based on the hierarchical structure of the tasks and the relevance and dependency between the tasks, the overall task is decomposed into a plurality of subtasks.

3. The method according to claim 1, It is characterized in that The corresponding sub-models for each sub-task are: Analyze the dependencies and data flows between the subtasks to form a task flow of a directed acyclic graph; For subtasks with dependencies, distribute the preceding subtask to the corresponding first submodel according to the data flow of the task flow, obtain the intermediate result output by the first submodel, and use the intermediate result as the input of the second submodel corresponding to the subsequent subtask; For multiple independent subtasks, the third sub-models corresponding to the independent subtasks are scheduled in parallel.

4. The method according to claim 1, It is characterized in that After calling the corresponding sub-model for each sub-task, the method further includes: Monitor the operating status of each sub-model in real time; If it is detected that the target sub-model is abnormal or has timed out, the sub-task corresponding to the target sub-model is rescheduled, or the backup sub-model of the target sub-model is called.

5. The method according to claim 1, It is characterized in that The intermediate results are merged to obtain the energy-carbon analysis result corresponding to the energy-carbon data analysis request, including: Obtaining weights corresponding to the intermediate results, and performing weighted summation on the intermediate results based on the weights to obtain a fusion result; An explanatory text of the fusion result is generated through the large language model, and the explanatory text is used as the energy-carbon analysis result corresponding to the energy-carbon data analysis request.

6. The method according to claim 5, It is characterized in that Generating the explanatory text of the fusion result by the large language model includes: Extract key information from each of the intermediate results; Analyzing the logical relationship between the key information; Converting the key information and the logical relationships between the key information into natural language expressions; The natural language expression is used as an explanatory text of the fusion result.

7. The method according to claim 1, It is characterized in that Before parsing the energy and carbon data analysis request through the pre-built large language model, the method further includes: Collect multi-source heterogeneous data; Based on the multi-source heterogeneous data, the initial model is self-supervised trained to obtain an intermediate model; The specific domain data sets in the multi-source heterogeneous data are aggregated, and the specific domain data sets are used to perform domain progressive fine-tuning on the intermediate model to obtain a large language model.

8. An energy carbon data analysis system, It is characterized in that include: A receiving module, used for receiving an energy and carbon data analysis request from a user; A parsing module, used to parse the energy-carbon data analysis request through a pre-built large language model to obtain an overall task corresponding to the energy-carbon data analysis request; A decomposition module, used to decompose the overall task into multiple subtasks, each subtask processing data analysis in a specific field; A scheduling module, used to call the corresponding sub-model for each sub-task and obtain the intermediate results output by the sub-model; The fusion module is used to perform weighted fusion on the intermediate results to obtain the energy-carbon analysis result corresponding to the energy-carbon data analysis request.

9. An electronic device, It is characterized in that It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; wherein: Memory, used to store computer programs; A processor, configured to execute the data analysis method according to any one of claims 1 to 7 by running a program stored in a memory.

10. A computer storage medium, It is characterized in that The computer storage medium includes a stored program, wherein the program executes the data analysis method according to any one of claims 1 to 7 when executed.

Images