Load data generation method and device, electronic equipment and computer storage medium

Abstract

The present application relates to new energy technology, and provide a kind of load data generation method, device, electronic equipment and computer storage medium, the method comprises: receiving natural language description for generating load data generation requirement;Utilize load generation large language model, based on generation requirement to carry out question and answer interaction, obtain requirement parameter, load generation large language model is pre-trained to large language model using target corpus, and target corpus is generated based on new energy project sample data and load generation rule;Demand parameter is used as input parameter, and target load data is generated by load generation large language model and based on target load data generation natural language description generation result.The present application can generate load data by question and answer interaction, improve efficiency.

Description

Technical Field

[0001] This invention relates to the field of new energy technology, and more specifically, to a method, apparatus, electronic device, and computer storage medium for generating load data. Background Technology

[0002] Dual carbon emissions have become a global trend, and new energy sources are developing rapidly worldwide, revolutionizing energy production, consumption, and the environment. Under this trend, numerous new energy projects are being planned and constructed at an accelerated pace. In the planning process of these projects, load data generation is a crucial and necessary step. Good load data can effectively improve the credibility, detail, and accuracy of economic assessments of new energy project planning schemes. Current load data generation algorithms are often complex, requiring numerous user-input parameters and involving complex adjustable influencing factors, increasing the difficulty and reducing the efficiency of load data generation. Summary of the Invention

[0003] The purpose of this invention is to provide a method, apparatus, electronic device, and computer storage medium for generating load data, which utilizes a large model to reduce the difficulty of generating load data and improve the efficiency of load data generation.

[0004] The embodiments of the present invention can be implemented as follows:

[0005] In a first aspect, the present invention provides a method for generating load data, the method comprising:

[0006] Receive the generation requirements for generating load data, described in natural language.

[0007] Using a load generation language model, question-and-answer interaction is performed based on the generated requirements to obtain the requirements parameters. The load generation language model is pre-trained using a target corpus, which is generated based on sample data from new energy projects and load generation rules.

[0008] Using the demand parameters as input parameters, the target load data is generated through the load generation big language model, and a natural language description is generated based on the target load data.

[0009] In an optional implementation, the method further includes:

[0010] Obtain sample data and load generation rules for new energy projects;

[0011] Based on the sample data and load generation rules, the original corpus is transformed into the target corpus;

[0012] Using the target corpus as input, the large language model is trained a second time to obtain the trained large model.

[0013] Establish the calling relationship between the trained large model and the preset load data generation model to obtain the load generation large language model.

[0014] In an optional implementation, the step of converting the original corpus into a target corpus based on the sample data and the load generation rules includes:

[0015] The sample data and load generation rules are processed in a structured manner to obtain a structured database. The structured database includes the input data, processing method, and output data of the load generation rules. The input data is the data to be processed in the sample data, the processing method is the generation method corresponding to the load generation rule, and the output data is the load data generated by the load generation rule.

[0016] The input data, processing method, and output data are used as keywords to determine the topic boundaries from the original corpus;

[0017] Construct a conversational corpus based on the original corpus located within the boundaries of the aforementioned topics;

[0018] The conversational corpus is structured and annotated to obtain the target corpus. The target corpus includes the conversational corpus and corresponding annotation elements. The annotation elements include entity elements, relation elements, and intent elements. The entity elements are determined based on the input data and / or output data. The relation elements are determined based on the relationship between the input data and / or output data and the processing method. The intent elements are determined based on the preset dialogue requirements of the conversational corpus.

[0019] In an optional implementation, the load generation language model and the preset load data generation model are pre-established with a calling relationship;

[0020] The step of using the demand parameters as input parameters, generating target load data through the load generation big language model, and generating a natural language description based on the target load data includes:

[0021] Using the demand parameters as input parameters, the target load data is generated by calling the preset load data generation model using the load generation big language model.

[0022] The target load data is converted into a natural language description using the load generation large language model.

[0023] In an optional implementation, the input parameters include the annual total electricity consumption of the target area, time step, daily variability, and monthly variability. The step of using the demand parameters as input parameters and calling the preset load data generation model using the load generation big data language model to generate target load data includes:

[0024] Based on the total annual electricity consumption and the preset typical daily load, a basic hourly load matrix is ​​generated;

[0025] The disturbance factor matrix is ​​determined based on the time step, the daily variability, and the monthly variability;

[0026] The target load data is obtained based on the basic hourly load matrix and the disturbance factor matrix.

[0027] In an optional implementation, the demand parameters include the target climate region, and prior to the step of determining the perturbation factor matrix based on the time step, the daily variability, and the monthly variability, the following steps are included:

[0028] Based on the target climate region, target meteorological data corresponding to the target climate region is determined from a preset meteorological list, which includes multiple climate regions and their corresponding meteorological data.

[0029] The target meteorological data is used as the inter-monthly variability.

[0030] In an optional implementation, the inter-monthly variability includes the daily average temperature of the target climate in the target area, and the step of determining the perturbation factor matrix based on the time step, the daily variability, and the inter-monthly variability includes:

[0031] Based on the time step, hourly time perturbation values ​​for each day within a year are randomly extracted from a preset normal distribution, and all the time perturbation values ​​are combined into an intraday perturbation factor matrix.

[0032] Based on the daily variability, daily perturbation values ​​for each day within a year are randomly selected from the preset normal distribution, and all the daily perturbation values ​​are combined into a daytime perturbation factor matrix;

[0033] Calculate the inter-monthly perturbation factor matrix based on all the stated daily average temperatures;

[0034] The disturbance factor matrix is ​​generated based on the intraday disturbance factor matrix, the inter-day disturbance factor matrix, and the inter-month disturbance factor matrix.

[0035] In a second aspect, the present invention provides a load data generation apparatus, the apparatus comprising:

[0036] The receiving module is used to receive generation requests for generating payload data, described in natural language.

[0037] The acquisition module is used to generate a large language model using load, perform question-and-answer interaction based on the generation requirements, and obtain the requirements parameters. The large language model for load generation is pre-trained using a target corpus, which is generated based on sample data from new energy projects and load generation rules.

[0038] The generation module is used to take the demand parameters as input parameters, generate target load data through the load generation big language model, and generate a natural language description based on the target load data.

[0039] Thirdly, the present invention provides an electronic device including a processor and a memory, the memory being used to store a program, and the processor being used to implement the load data generation method as described in any of the foregoing embodiments when executing the program.

[0040] Fourthly, the present invention provides a computer storage medium having a computer program stored thereon, which, when executed by a processor, implements the load data generation method as described in any of the foregoing embodiments.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] This invention utilizes a large-scale load generation language model. Based on the generation requirements of natural language descriptions, a question-and-answer interaction is performed to obtain the requirement parameters. These requirement parameters are then used as input parameters to obtain the target load data through the large-scale load generation language model and generate the natural language description results. Since the target corpus is generated based on sample data from new energy projects and load generation rules, and the large-scale load generation language model is trained using the target corpus, it achieves the goal of obtaining requirement parameters through question-and-answer interaction, reducing the difficulty of load data generation and improving the efficiency of load data generation. Attached Figure Description

[0043] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is an overall framework diagram provided for this embodiment.

[0045] Figure 2A flowchart of the load data generation method provided in this embodiment Figure 1 .

[0046] Figure 3 A flowchart of the load data generation method provided in this embodiment Figure 2 .

[0047] Figure 4 This is a block diagram of the load data generation device provided in this embodiment.

[0048] Figure 5 This is a block diagram of the electronic device provided in this embodiment.

[0049] Icons: 10-Electronic device; 11-Processor; 12-Memory; 13-Bus; 100-Load data generation device; 110-Receiving module; 120-Acquisition module; 130-Generation module; 140-Training module. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0051] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0052] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0053] In the description of this invention, it should be noted that if terms such as "upper," "lower," "inner," or "outer" are used to indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of this invention is usually placed, they are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0054] Furthermore, the terms "first" and "second" are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0055] It should be noted that, where there is no conflict, the features in the embodiments of the present invention can be combined with each other.

[0056] Please refer to Figure 1 , Figure 1 This is the overall framework diagram provided in this embodiment. Figure 1 For the process of generating load data, please refer to [link / reference]. Figure 1 As shown by the blue arrows: Users interact with the load generation big language model, providing the load generation big language model with natural language descriptions of their generation requirements for generating load data. Based on these requirements, the load generation big language model guides users to provide the required parameters through a question-and-answer format. The load generation big language model uses the required parameters as input parameters to generate target load data, and then generates load data with natural language descriptions based on the target load data.

[0057] In order to enable the large language model for load generation to have interactive and generative capabilities for generating load data, Figure 1 Its construction process is also shown:

[0058] First, collect load data from new energy projects; to achieve better generation results, collect as much load data from new energy projects as possible.

[0059] Secondly, organize the load generation rules and combine them with the load data to form a large database stored in a structured form;

[0060] Third, a corpus for generating load data of new energy projects will be constructed based on a large database.

[0061] Fourth, secondary training is performed based on the large language model, and the pre-built load data generation algorithm model is called during secondary training;

[0062] Fifth, construct a large language model within the context of payload data generation, so that the final, retrained payload generation large language model can guide and restrict dialogue within the specialized field of payload data generation, analyze the results of payload data generation, provide evaluations of rationality, scientific validity, and credibility, and offer suggestions for improvement.

[0063] Furthermore, the pre-built load data generation algorithm model can be trained using data from a large database, or obtained by fitting formulas based on data from a large database. The load data generation algorithm model also provides an access interface that can be called by the large language model, enabling the final load generation large language model to take demand parameters as input parameters, call the load data generation algorithm model to generate load data, and then convert the load data returned by the load data generation algorithm model into a natural language description of the generated result.

[0064] The following will be based on Figure 1 The overall architecture is described in detail, including its workflow and build process.

[0065] Please refer to Figure 2 , Figure 2 A flowchart of the load data generation method provided in this embodiment Figure 1 The method includes the following steps:

[0066] Step S101: Receive the generation request for generating load data described in natural language.

[0067] In this embodiment, the generated request can be a load generation task for a specific new energy project expressed by the user in natural language. The generated request can be a complete parameter description or a vague or incomplete request. Its purpose is to initiate the entire load generation process and provide initial input for subsequent question-and-answer interactions. For example, when a user needs load data for an industrial park, they can input the generated request "Please help me generate load data for an industrial park".

[0068] Step S102: Using the load generation large language model, question-and-answer interaction is performed based on the generation requirements to obtain the requirement parameters. The load generation large language model is pre-trained using a target corpus, which is generated based on sample data from new energy projects and load generation rules.

[0069] In this embodiment, the load generation large language model is a double-trained large language model capable of understanding problems in the field of new energy projects. It can proactively guide users to provide the required parameters for generating target load data. This guidance can be provided through a user-friendly dialogue, guiding users to input the algorithm parameters needed to generate load data, or by guiding users to upload an input file readable by the algorithm. Its training corpus originates from a target corpus, which is a structured collection of data formed after systematic processing of sample data from new energy projects and load generation rules. New energy projects broadly refer to renewable energy power generation scenarios such as wind power and photovoltaics. Sample data consists of various load data and related background information collected from actual or simulated new energy projects. Load generation rules refer to mathematical models and engineering empirical formulas used to simulate load change trends in specific application scenarios.

[0070] It should be noted that the question-and-answer interaction can be one round or multiple rounds, depending on whether the load generation big data model can generate load data based on the obtained demand parameters. Through one or more rounds of question-and-answer interaction, the load generation big data model can gradually clarify the key parameters required by the user, such as load type, geographical area, building area, and energy consumption characteristics, forming a complete set of demand parameters.

[0071] Step S103: Using the demand parameters as input parameters, the target load data is generated through the load generation large language model, and a natural language description is generated based on the target load data.

[0072] In this embodiment, the load generation language model can obtain target load data through a preset load data generation algorithm. Furthermore, it is responsible for converting the target load data output by the algorithm into a natural language description format that is easy to understand and use. For example, the generated result could be a textual feedback of the result such as "Based on the information you provided, the estimated average daily electricity consumption of this park is 1200 kWh."

[0073] The method provided in this embodiment generates a target corpus based on sample data and load generation rules of new energy projects, and then uses the target corpus to train a large language model to obtain a load generation large language model. Through the load generation large language model, demand parameters can be obtained through question-and-answer interaction, which reduces the difficulty of generating load data, improves the efficiency of load data generation, and significantly enhances usability and user experience.

[0074] The above is about Figure 1 The overall architecture and workflow have been introduced; the following will discuss... Figure 1 The process of constructing a large language model with medium load is described in detail. Please refer to [link / reference]. Figure 3 , Figure 3 A flowchart of the load data generation method provided in this embodiment Figure 2 The method includes the following steps:

[0075] Step S201: Obtain sample data and load generation rules for new energy projects.

[0076] In this embodiment, the sample data for new energy projects includes, but is not limited to, fine-grained load data (covering multiple energy types such as electricity, gas, coal, and steam), business registration information of industrial parks or enterprises, geographic information data (such as latitude and longitude, climate characteristics), and building parameters (such as area, production capacity, and energy consumption per unit area), among other dimensions. Load generation rules are used to guide the generation of load data. These rules can be determined based on historical data, for example, by fitting historical data into a formula and using the fitted formula to generate load data; or they can be determined by analogy with similar objects, referencing the load data generation methods of similar objects and adjusting the generation method to obtain load generation rules capable of generating the target load data; or they can be determined based on machine learning-based generation rules.

[0077] Step S202: Based on the sample data and the load generation rules, the original corpus is transformed into the target corpus.

[0078] In this embodiment, the original corpus can be a collection of language materials from general or professional fields, possibly including unstructured or semi-structured information such as dialogue records, technical documents, and expert experience texts. By structuring the sample data and load generation rules, a target corpus with a clear functional orientation of generating load data for new energy projects can be constructed. This embodiment provides a specific implementation method for transforming the original corpus into the target corpus:

[0079] First, the sample data and load generation rules are processed in a structured manner to obtain a structured database. The structured database includes the input data, processing method and output data of the load generation rules. The input data is the data to be processed in the sample data, the processing method is the generation method corresponding to the load generation rule, and the output data is the load data generated by the load generation rule.

[0080] In this embodiment, the input data may be park area, climate characteristics, etc., the processing method may be fitting formula, machine learning model, etc., and the output data may be load curve, annual electricity consumption, etc.

[0081] Secondly, the input data, processing methods, and output data are used as keywords to determine the topic boundaries from the original corpus;

[0082] In this embodiment, keyword matching can be used to identify corpus content related to the new energy load data generation task from the original corpus. Specifically, input data, processing methods, and output data from the structured database are used as keywords to filter out semantically relevant and thematically consistent original corpus, and the thematic boundaries of the corpus are defined accordingly. In this way, general corpus unrelated to load data generation can be effectively excluded, ensuring the professionalism and focus of subsequent corpus construction.

[0083] Third, construct a conversational corpus based on the original corpus within the topic boundaries;

[0084] In this embodiment, based on the original corpus within the topic boundaries, a question-and-answer interaction scenario between the user and the system is simulated to construct a conversational corpus with practical interactive significance. For example, a conversational corpus similar to the following can be generated: "Human: Please help me generate electrical load and heat load data for a park. Assistant: Okay, please tell me the relevant parameters. Let's start with the electrical load, including the park type, park area, city where the park is located, annual electricity consumption per unit area of ​​the park, or total expected total consumption, etc. Human: Located in city A, covering an area of ​​10,000 square meters, a typical office park, I don't know the expected consumption. Assistant: Okay, I will process it according to the average level of typical office parks in city A. Next, please tell me the heat load characteristics, such as building insulation type, whether it includes industrial heat or computer room heat, etc." The conversational corpus can reflect the interaction logic in real application scenarios, providing structured, task-oriented input samples for subsequent secondary training.

[0085] Finally, the conversational corpus is structured and annotated to obtain the target corpus. The target corpus includes the conversational corpus and the corresponding annotation elements. The annotation elements include entity elements, relation elements, and intent elements. Entity elements are determined based on the input data and / or output data. Relation elements are determined based on the relationship between the input data and / or output data and the processing method. Intent elements are determined based on the preset dialogue requirements of the conversational corpus.

[0086] In this embodiment, entity elements refer to specific parameters or objects involved in the conversational corpus, such as park type, building area, climate zone, etc., which are usually derived from input data and output data; relational elements describe the logical relationship between input, output and processing method, such as the positive correlation between "park area" and "annual electricity consumption"; intent elements identify the functional purpose carried by the corpus, such as "requesting load generation", "guiding user input" or "feedback generation results".

[0087] Through the above transformation process, the original scattered and unstructured raw data can be transformed into a target corpus with clear semantic structure and interaction logic, providing high-quality and task-oriented data support for the subsequent secondary training of large language models.

[0088] Step S203: Using the target corpus as input, perform secondary training on the large language model to obtain the trained large model.

[0089] In this embodiment, the large language model refers to an initial language model built on an open-source architecture, such as LLaMA, ChatGLM, or other models adapted to the Chinese environment. Since the original large language model lacks expertise in the field of new energy load generation, it needs to be transformed into a specialized model with vertical domain capabilities through secondary training. Specific training methods can be flexibly selected based on resource conditions, including but not limited to pre-training, fine-tuning, or knowledge-based integration. By using the target corpus as training input, the trained large model acquires the language understanding capabilities, parameter recognition logic, and interactive guidance strategies required for the load generation task.

[0090] Step S204: Establish the calling relationship between the trained large model and the preset load data generation model to obtain the load generation large language model.

[0091] In this embodiment, the preset load data generation model is a pre-built load forecasting algorithm module, for example, Figure 1 The load data generation algorithm model can be a traditional algorithm based on mathematical formulas or a trained machine learning model. The large-scale load generation language model provides a calling interface through an encapsulated interface mechanism, enabling automatic invocation of the load data generation algorithm. This calling interface can be an HTTP interface based on the HTTP mechanism, an interface implemented based on a DLL (Dynamic Link Library) component, or an API (Application Programming Interface). Specifically, after establishing the calling relationship, the large-scale load generation language model can determine how to pass input parameters and how to parse the returned results. The establishment of the calling relationship not only enables the large-scale language model to undertake natural language interaction tasks but also endows it with the ability to coordinate computational logic and integrate analysis results, thus forming a complete closed-loop system for load data generation.

[0092] It should be noted that, in order to further improve the interaction experience and effectiveness between the large-scale load generation language model and the user, a context for load data generation can be constructed for the large-scale load generation language model: one implementation method is as follows:

[0093] The above describes the construction of a large language model for load generation, which aims to limit question-and-answer interaction to the domain of load data generation.

[0094] The following text is constructed for the large language model of load generation, so that the large language model of load generation can evaluate and analyze the generation results and output the evaluation results and analysis results.

[0095] For example, the preceding text could be "Please act as an energy expert and guide and limit the conversation to the specialized area of ​​load generation." The following text could be "Evaluate the parameters of load generation and analyze potential optimizable items," or "Analyze the results of load generation, provide evaluations of its rationality, scientific validity, and credibility, and offer suggestions for improvement."

[0096] It should also be noted that, to improve the ease of use of the large language model generated by the workload, the large language model can be practically encapsulated. The encapsulated model can take the form of, but is not limited to, web pages, software, apps, and mini-programs. The deployment of the server running the encapsulated model can be carried out as needed, including local deployment, cloud deployment, centralized deployment, and distributed deployment.

[0097] Given that the large language model for load generation and the preset load data generation model have a pre-established calling relationship, this embodiment also provides an implementation method for obtaining the generation result based on the requirement parameters;

[0098] First, the demand parameters are used as input parameters, and the target load data is generated by calling the preset load data generation model using the load generation big language model.

[0099] Secondly, the target load data is converted into a natural language description using a load generation large language model.

[0100] In this embodiment, the form of the generated result can be pre-specified in the context of generating a large language model, including, but not limited to, web page chart presentation, report, report, graphic and text display, etc.

[0101] This embodiment takes the annual total electricity consumption of the target area, time step, daily variability, and monthly variability as input parameters as an example to provide an implementation method for generating target load data. This implementation method includes the following processing:

[0102] (1) Generate a basic hourly load matrix based on the total annual electricity consumption and the preset typical daily load;

[0103] (2) Determine the disturbance factor matrix based on the time step, daily variability and monthly variability;

[0104] In this embodiment, the total annual electricity consumption refers to the estimated or historically accumulated electricity consumption within the user-specified area throughout the year. The preset typical daily load is the basic reference data for generating the target load data. The preset typical daily load can be the daily load utilization rate of a park of similar size and type to the user-specified area. The preset typical daily load is usually derived from historical data statistical analysis or industry standard settings, used to describe the basic trend of load data changing over time within a 24-hour period. The basic hourly load matrix is ​​a standard load distribution formed by extending the typical daily load curve to the entire year without random disturbances. It is constructed by proportionally allocating the total annual electricity consumption to each hour throughout the year and combining it with the typical daily load. The basic hourly load matrix includes load data for each of the 365 days of the year, 24 hours a day.

[0105] In this embodiment, the time step defines the temporal resolution of the load data, such as hourly or daily; daily variability represents the degree of fluctuation in load data between different dates, which is usually related to factors such as user behavior patterns and holiday arrangements; while monthly variability reflects the differences in load data between different months, which is usually affected by seasonal changes and climatic conditions. The disturbance factor matrix refers to a set of weighted coefficients used to simulate the actual fluctuation of load, and its composition covers multiple dimensions such as intraday disturbance, interday disturbance, and intermonthly disturbance.

[0106] In this embodiment, since inter-monthly perturbations and climate change are closely related, the required parameter of inter-monthly variability can be provided directly by the user or indirectly obtained through a target climate region provided by the user. When the user provides a target climate region, inter-monthly variability can be obtained based on meteorological data of the target climate region. One method of obtaining this parameter is as follows:

[0107] Based on the target climate region, target meteorological data corresponding to the target climate region is determined from the preset meteorological list. The preset meteorological list includes multiple climate regions and their corresponding meteorological data.

[0108] The target meteorological data is used as an inter-monthly variability.

[0109] In this embodiment, the preset weather list is a pre-built climate information database containing multiple climate regions and their corresponding meteorological data, such as the annual average temperature, monthly average temperature, daily average temperature variation curve, annual sunshine hours, and annual precipitation for each region. Target meteorological data is obtained by matching the user-provided target climate region against this list.

[0110] Since the disturbance factors affecting the generation of target load data are multi-dimensional, in order to consider the impact of multiple disturbance factors and to reasonably determine each disturbance factor, this embodiment quantifies intraday fluctuations, interday differences, and intermonthly variations separately, and integrates them into a complete disturbance factor matrix. One implementation method is as follows:

[0111] First, based on the time step, randomly extract hourly time perturbation values ​​for each day within a year from a preset normal distribution and form an intraday perturbation factor matrix from all time perturbation values;

[0112] In this embodiment, the intraday disturbance factor matrix is ​​used to simulate the set of disturbance coefficients for load fluctuations at different times within a day. The preset normal distribution can be a normal distribution with a mean of 0 and a standard deviation of 0.05. The intraday disturbance factor matrix is ​​a matrix form in which the extracted time disturbance values ​​are organized in a 365-day × 24-hour format.

[0113] Second, based on constant variability, daily disturbance values ​​for each day within a year are randomly selected from a preset normal distribution, and all daily disturbance values ​​are combined into a daytime disturbance factor matrix;

[0114] In this embodiment, the inter-day disturbance factor matrix is ​​used to describe the load difference between different dates. Its generation method is similar to that of the intra-day disturbance factor matrix, but its time unit is "daily", that is, each element represents the disturbance coefficient of a day. The inter-day disturbance factor matrix is ​​also generated based on a preset normal distribution. The preset normal distribution used by the inter-day disturbance factor matrix and the preset normal distribution used by the intra-day disturbance factor matrix can be the same or different.

[0115] Third, calculate the inter-monthly perturbation factor matrix based on all daily average temperatures;

[0116] In this embodiment, the inter-monthly variability is indirectly obtained from the target meteorological data, and the inter-monthly variability is calculated using the daily average temperature in the target meteorological data as an example. The inter-monthly perturbation factor matrix reflects the seasonal variation of load data between different months.

[0117] One approach is to establish a mapping function based on the empirical relationship between daily average temperature and load data (e.g., the positive correlation between temperature and air conditioning load data, and the negative correlation with heating load data), and then calculate the disturbance coefficient for each month accordingly. For example, in the hot summer months, increased air conditioning load leads to increased electricity load, so the disturbance coefficient for that month will be greater than 1; while in spring or autumn, the load is relatively stable, and the disturbance coefficient is close to 1. Finally, the disturbance coefficients are organized into an inter-monthly disturbance factor matrix.

[0118] Fourth, generate the disturbance factor matrix based on the intraday disturbance factor matrix, the inter-day disturbance factor matrix, and the inter-month disturbance factor matrix.

[0119] In this embodiment, the intraday disturbance factor matrix, the inter-day disturbance factor matrix, and the inter-month disturbance factor matrix are combined element by element, for example by multiplication or weighted superposition, to generate the disturbance factor matrix.

[0120] As a specific implementation method, each element in the perturbation factor matrix can be calculated using the following formula, taking the target element as any element in the perturbation factor matrix as an example:

[0121] The target element in the disturbance factor = 1 + the element corresponding to the target element in the intraday disturbance factor matrix + the element corresponding to the target element in the inter-day disturbance factor matrix + the element corresponding to the target element in the inter-month disturbance factor matrix.

[0122] (3) Based on the basic hourly load matrix and the disturbance factor matrix, the target load data is obtained.

[0123] It should be noted that the execution Figure 2 Electronic devices and execution in each step of the process Figure 3 The electronic devices used for each step and its sub-steps can be the same or different. That is, a large language model for load generation can be built on one electronic device and then used on another electronic device. In addition, a preset load data generation model can be built independently, and then the preset load data generation model and the large language model can be fused together to perform a second training on the large language model to obtain the large language model for load generation.

[0124] To perform the corresponding steps in the above embodiments and various possible implementations, an implementation of the load data generation device 100 is given below. Please refer to... Figure 4 , Figure 4 This is a block diagram of the load data generation device provided in this embodiment. It should be noted that the basic principle and technical effects of the load data generation device 100 provided by the present invention are the same as those of the corresponding embodiments described above. For the sake of brevity, some parts of this embodiment are not mentioned.

[0125] The load data generation device 100 includes a receiving module 110, an acquisition module 120, and a generation module 130.

[0126] The receiving module 110 is used to receive a generation request for generating payload data described in natural language;

[0127] The acquisition module 120 is used to generate a large language model using load, conduct question-and-answer interaction based on the generation requirements, and obtain the requirements parameters. The large language model for load generation is pre-trained using a target corpus, which is generated based on sample data from new energy projects and load generation rules.

[0128] The generation module 130 is used to take the demand parameters as input parameters, generate target load data through a load generation large language model, and generate a natural language description based on the target load data.

[0129] In an optional embodiment, the load data generation device 100 further includes a training module 140, which is specifically used for:

[0130] Obtain sample data and load generation rules for new energy projects;

[0131] Based on the sample data and the load generation rules, the original corpus is transformed into the target corpus;

[0132] The target corpus is used as input to perform secondary training on the large language model, resulting in a trained large model.

[0133] Establish the calling relationship between the trained large model and the preset load data generation model to obtain the load generation large language model.

[0134] In an optional implementation, the training module 140 is specifically used to transform the original corpus into a target corpus based on sample data and payload generation rules:

[0135] The sample data and load generation rules are processed in a structured manner to obtain a structured database. The structured database includes the input data, processing method and output data of the load generation rules. The input data is the data to be processed in the sample data, the processing method is the generation method corresponding to the load generation rule, and the output data is the load data generated by the load generation rule.

[0136] The input data, processing method, and output data are used as keywords to determine the topic boundaries from the original corpus;

[0137] Construct a conversational corpus based on the original corpus within the topic boundaries;

[0138] The conversational corpus is structured and annotated to obtain the target corpus. The target corpus includes the conversational corpus and the corresponding annotation elements. The annotation elements include entity elements, relation elements and intent elements. Entity elements are determined based on the input data and / or output data. Relation elements are determined based on the relationship between the input data and / or output data and the processing method. Intent elements are determined based on the preset dialogue requirements of the conversational corpus.

[0139] In an optional implementation, the load generation large language model and the preset load data generation model are pre-created with a calling relationship, and the generation module 130 is specifically used for;

[0140] Using demand parameters as input parameters, the target load data is generated by calling a preset load data generation model using a large language model for load generation.

[0141] The target load data is converted into a natural language description by using a load generation large language model.

[0142] In an optional implementation, the input parameters include the annual total electricity consumption of the target area, time step, daily variability, and monthly variability. The generation module 130, when using the demand parameters as input parameters and utilizing the load generation big data model to call a preset load data generation model to generate target load data, specifically performs the following:

[0143] A basic hourly load matrix is ​​generated based on the total annual electricity consumption and the preset typical daily load.

[0144] The disturbance factor matrix is ​​determined based on the time step, daily variability, and monthly variability.

[0145] The target load data is obtained based on the basic hourly load matrix and the disturbance factor matrix.

[0146] In an optional implementation, the requirement parameters include the target climate region, and the generation module 130 is further configured to:

[0147] Based on the target climate region, target meteorological data corresponding to the target climate region is determined from the preset meteorological list. The preset meteorological list includes multiple climate regions and their corresponding meteorological data.

[0148] The target meteorological data is used as an inter-monthly variability.

[0149] In an optional implementation, the inter-monthly variability includes the daily average temperature of the target climate in the target area. The generation module 130, when determining the perturbation factor matrix based on the time step, daily variability, and inter-monthly variability, specifically uses:

[0150] Based on the time step, hourly time perturbation values ​​for each day within a year are randomly selected from a preset normal distribution, and all time perturbation values ​​are combined into an intraday perturbation factor matrix.

[0151] Based on daily variability, daily disturbance values ​​are randomly selected from a pre-defined normal distribution for each day within a year, and all daily disturbance values ​​are combined to form an inter-day disturbance factor matrix;

[0152] Calculate the inter-monthly perturbation factor matrix based on all daily average temperatures;

[0153] The disturbance factor matrix is ​​generated based on the intraday disturbance factor matrix, the inter-day disturbance factor matrix, and the inter-month disturbance factor matrix.

[0154] This invention also provides a block diagram of an electronic device 10, which implements the load data generation method of the aforementioned embodiments. Please refer to... Figure 5 , Figure 5 This is a block diagram of the electronic device 10 provided in this embodiment. The electronic device 10 includes a processor 11, a memory 12 and a bus 13. The processor 11 and the memory 12 are connected through the bus 13.

[0155] The processor 11 can be an integrated circuit chip with signal processing capabilities. In implementation, each step of the load data generation method in the above embodiments can be completed by the integrated logic circuits in the hardware of the processor 11 or by software instructions. The processor 11 can be a general-purpose processor, including a CPU (Central Processing Unit), an NP (Network Processor), etc.; it can also be a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Logic Gate Array), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0156] The memory 12 is used to store the program that implements the load data generation method. The program may be a software function module stored in the memory 12 in the form of software or firmware or embedded in the OS (Operating System) of the electronic device 10.

[0157] After receiving the execution instruction, the processor 11 executes the program to implement the load data generation method of the aforementioned embodiment.

[0158] This embodiment provides a computer storage medium storing a computer program, which, when executed by a processor, implements the load data generation method as described in the foregoing embodiments.

[0159] In summary, embodiments of the present invention provide a method, apparatus, electronic device, and computer storage medium for generating load data. The method includes: receiving a generation request for generating load data described in natural language; using a large language model for load generation, performing question-and-answer interaction based on the generation request to obtain request parameters, wherein the large language model for load generation is pre-trained using a target corpus, which is generated based on sample data from new energy projects and load generation rules; using the request parameters as input parameters, generating target load data through the large language model for load generation, and generating a natural language description based on the target load data. Compared with the prior art, this embodiment has at least the following advantages: (1) A large language model is trained based on the sample data of new energy projects and the target corpus generated by the load generation rules to obtain a load generation large language model, which enables the load generation large language model to obtain the demand parameters through question and answer interaction, reducing the difficulty of load data generation and improving the efficiency of load data generation; (2) A large language access interface for the load data generation model is constructed to ensure that the load generation large language model can correctly call the load data generation model, pass in the input parameters and obtain the load data; (3) A large language model is constructed in the context of load generation, so that the constructed load generation large language model can focus on serving the vertical field of load generation of new energy projects, and automatically generate evaluation and improvement suggestions based on the load generation large language model for the input parameters and the returned load data.

[0160] The above descriptions are merely various embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for generating load data, characterized in that, The method includes: Receive the generation requirements for generating load data, described in natural language. Using a load generation language model, question-and-answer interaction is performed based on the generated requirements to obtain the requirements parameters. The load generation language model is pre-trained using a target corpus, which is generated based on sample data from new energy projects and load generation rules. The method further includes: using the demand parameters as input parameters, generating target load data through the load generation large language model, and generating a natural language description based on the target load data; the method also includes: Obtain sample data and load generation rules for new energy projects; Based on the sample data and load generation rules, the original corpus is transformed into the target corpus; Using the target corpus as input, the large language model is trained a second time to obtain the trained large model. Establishing the calling relationship between the trained large model and the preset load data generation model to obtain the load generation large language model; the step of converting the original corpus into the target corpus according to the sample data and load generation rules includes: The sample data and load generation rules are processed in a structured manner to obtain a structured database. The structured database includes the input data, processing method, and output data of the load generation rules. The input data is the data to be processed in the sample data, the processing method is the generation method corresponding to the load generation rule, and the output data is the load data generated by the load generation rule. The input data, processing method, and output data are used as keywords to determine the topic boundaries from the original corpus; Construct a conversational corpus based on the original corpus located within the boundaries of the aforementioned topics; The conversational corpus is structured and annotated to obtain the target corpus. The target corpus includes the conversational corpus and corresponding annotation elements. The annotation elements include entity elements, relation elements, and intent elements. The entity elements are determined based on the input data and / or output data. The relation elements are determined based on the relationship between the input data and / or output data and the processing method. The intent elements are determined based on the preset dialogue requirements of the conversational corpus.

2. The method according to claim 1, characterized in that, The load generation language model and the preset load data generation model have a pre-created calling relationship; The step of using the demand parameters as input parameters, generating target load data through the load generation big language model, and generating a natural language description based on the target load data includes: Using the demand parameters as input parameters, the target load data is generated by calling the preset load data generation model using the load generation big language model. The target load data is converted into a natural language description using the load generation large language model.

3. The method according to claim 2, characterized in that, The input parameters include the annual total electricity consumption of the target area, time step, daily variability, and monthly variability. The step of using the demand parameters as input parameters and calling the preset load data generation model using the load generation big data language model to generate target load data includes: Based on the total annual electricity consumption and the preset typical daily load, a basic hourly load matrix is ​​generated; The disturbance factor matrix is ​​determined based on the time step, the daily variability, and the monthly variability; The target load data is obtained based on the basic hourly load matrix and the disturbance factor matrix.

4. The method according to claim 3, characterized in that, The demand parameters include the target climate region. Prior to the step of determining the perturbation factor matrix based on the time step, the daily variability, and the monthly variability, the following steps are included: Based on the target climate region, target meteorological data corresponding to the target climate region is determined from a preset meteorological list, which includes multiple climate regions and their corresponding meteorological data. The target meteorological data is used as the inter-monthly variability.

5. The method according to claim 3, characterized in that, The inter-monthly variability includes the daily average temperature of the target climate in the target area, and the step of determining the perturbation factor matrix based on the time step, the daily variability, and the inter-monthly variability includes: Based on the time step, hourly time perturbation values ​​for each day within a year are randomly extracted from a preset normal distribution, and all the time perturbation values ​​are combined into an intraday perturbation factor matrix. Based on the daily variability, daily perturbation values ​​for each day within a year are randomly selected from the preset normal distribution, and all the daily perturbation values ​​are combined into a daytime perturbation factor matrix; Calculate the inter-monthly perturbation factor matrix based on all the stated daily average temperatures; The disturbance factor matrix is ​​generated based on the intraday disturbance factor matrix, the inter-day disturbance factor matrix, and the inter-month disturbance factor matrix.

6. A load data generation device, characterized in that, The device includes: The receiving module is used to receive generation requests for generating payload data, described in natural language. The acquisition module is used to generate a large language model using load, perform question-and-answer interaction based on the generation requirements, and obtain the requirements parameters. The large language model for load generation is pre-trained using a target corpus, which is generated based on sample data from new energy projects and load generation rules. The generation module is used to take the demand parameters as input parameters, generate target load data through the load generation big language model, and generate a natural language description based on the target load data. The training module is used for: acquiring sample data and load generation rules of new energy projects; converting the original corpus into a target corpus based on the sample data and load generation rules; using the target corpus as input to perform secondary training on the large language model to obtain the trained large model; and establishing a calling relationship between the trained large model and the preset load data generation model to obtain the load generation large language model. The training module, when used to transform the original corpus into a target corpus based on the sample data and load generation rules, specifically performs the following: It performs structured processing on the sample data and load generation rules to obtain a structured database. The structured database includes input data, processing methods, and output data of the load generation rules. The input data is the data to be processed in the sample data, the processing method is the generation method corresponding to the load generation rule, and the output data is the load data generated by the load generation rule. It uses the input data, processing methods, and output data as keywords to determine topic boundaries from the original corpus. It constructs a conversational corpus based on the original corpus within the topic boundaries. It performs structured annotation on the conversational corpus to obtain the target corpus. The target corpus includes the conversational corpus and corresponding annotation elements. The annotation elements include entity elements, relation elements, and intent elements. The entity elements are determined based on the input data and / or output data. The relation elements are determined based on the relationship between the input data and / or output data and the processing method. The intent elements are determined based on the preset dialogue requirements of the conversational corpus.

7. An electronic device, characterized in that, It includes a processor and a memory, the memory being used to store a program, and the processor being used to implement the load data generation method as described in any one of claims 1-5 when executing the program.

8. A computer storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the load data generation method as described in any one of claims 1-5.

Images