Intelligent energy management system based on power distribution room interaction robot monitoring

Abstract

The invention discloses a smart energy management system based on power distribution room interaction robot monitoring. According to the invention, a pre-training module, a model fine tuning module, a knowledge base module, a data training module and an optimization algorithm module are arranged in an electric power large model module, the pre-training module trains models on a large amount of power distribution room related data and documents through unsupervised learning, and learns universal language representation; the model has the semantic understanding capability in the power distribution field, and a basis is provided for subsequent model fine tuning. And the model fine tuning module performs fine tuning on the model by using a supervised data set of a specific task on the basis of the pre-trained model, so that the model can adapt to the application in the field of the power distribution room, and the accuracy and adaptability of the model are improved. The knowledge base module provides professional knowledge, equipment information, operation and maintenance regulations and the like in the power distribution field for the model, helps the model to better understand and answer problems related to the power distribution room, improves the intelligent degree of the model, and improves the training efficiency of the model.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent management of power distribution rooms, and specifically is an intelligent energy management system based on interactive robot monitoring of power distribution rooms. Background Art

[0002] The distribution room robot is mainly used in the power distribution room. It follows the safety and regulatory management requirements of the distribution room and the industry regulations, realizes intelligent interaction, intelligent identification, intelligent inspection, accurate diagnosis, safety protection functions, and fixed robots are used in combination to complete the 7X24-hour uninterrupted intelligent guarding and duty functions of the distribution room, and solve the problem of power safety in the distribution room. The distribution room robot is a high-tech product designed for power distribution rooms, aiming to provide intelligent operation and maintenance solutions for the power industry. This robot has a variety of functions, including intelligent voice interaction, fault warning, fault diagnosis, intelligent inspection, etc., and can guard the distribution room 24 hours a day to ensure the safety of electricity use. The distribution room robot is equipped with a variety of sensors and equipment, including visible light cameras, thermal imaging infrared cameras, microphone matrices, speakers, etc., which can interact with the environment and personnel. Through the SLAM navigation system and differential steering mechanism, the robot can move autonomously and perform inspection tasks. In addition, the robot also has an automatic charging system to ensure long-term operation.

[0003] However, common interactive robots in distribution rooms cannot provide rich training samples during use, which prevents the model from being exposed to common problems and scenarios in distribution room operation and maintenance, thus affecting the generalization ability of the robot model. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent energy management system based on interactive robot monitoring in a distribution room in order to solve the above-mentioned problems.

[0005] The technical solution adopted by the present invention is as follows: A smart energy management system based on interactive robot monitoring in a distribution room, comprising a power supply module, an interactive robot module, a large power model module, a model training management module, a data interface module, a business application module, a pre-training module, a model fine-tuning module, a knowledge base module, a data training module and an optimization algorithm module, wherein the output end of the power supply module is connected to the input end of the interactive robot module, the output end of the interactive robot module is connected to the input end of the large power model module, the output end of the large power model module is connected to the input end of the model training management module, the output end of the model training management module is connected to the input end of the data interface module, and the output end of the data interface module is connected to the input end of the business application module.

[0006] In a preferred embodiment, the large electric power model module is internally provided with a pre-training module, a model fine-tuning module, a knowledge base module, a data training module and an optimization algorithm module, and the overall output end of the pre-training module, the model fine-tuning module, the knowledge base module, the data training module and the optimization algorithm module is connected to the input end of the large electric power model module.

[0007] In a preferred embodiment, the interactive robot module realizes intelligent voice interaction, provides intelligent question and answer, fault warning, fault diagnosis and other functions. The robot is equipped with hardware such as visible light camera, thermal imaging infrared camera, microphone matrix, speaker, etc. to realize interaction with the environment and people. In addition, the robot realizes autonomous movement through SLAM navigation system and differential steering mechanism, and is equipped with automatic charging system.

[0008] In a preferred embodiment, the pre-training module uses deep learning technology to build a model with a multi-layer neural network structure. The pre-training module is based on the Transformer architecture, can process long-distance dependency information, and is suitable for tasks such as speech recognition and semantic understanding;

[0009] The calculation formula of the forward propagation of the pre-training module during use is y=f(wx+b), where f is the activation function, w is the weight, b is the bias, and x is the input; the pre-training module uses an unsupervised learning method to train the model on a large amount of distribution room related data and documents to learn a common language representation. The basic calculation formula for the training of autoencoders or variational autoencoders is to maximize the likelihood function, that is, to maximize p(x|z).

[0010] In a preferred embodiment, the model fine-tuning module uses a supervised dataset of a specific task to fine-tune the model based on the pre-trained model, so that the model is suitable for application in the field of power distribution rooms;

[0011] The model fine-tuning module involves a back-propagation algorithm, and its basic calculation formula is: Where η is the learning rate, C(w) is the loss function, and w is the weight.

[0012] In a preferred embodiment, the knowledge base module includes:

[0013] Professional knowledge of power distribution room: including basic knowledge of electrical engineering, power system, power automation and other related fields. Working principle, operating parameters, maintenance methods of power equipment, etc. Operating mechanism and mutual relationship of various parts of power system.

[0014] Equipment information: detailed information on the models, specifications, parameters, etc. of various equipment in the power distribution room. Normal operating range, alarm threshold, failure mode, etc. of the equipment.

[0015] Operation and maintenance procedures: Rules and regulations for the operation and maintenance of the distribution room, safety operating procedures, emergency plans, etc. Regular inspection and maintenance procedures for equipment. Fault handling procedures and methods for handling abnormal situations.

[0016] Distribution room environmental parameters: standard ranges of distribution room environmental parameters such as temperature, humidity, gas composition, etc. Important environmental factors related to power system operation.

[0017] Typical failure cases: common failure cases of distribution room equipment, failure cause analysis, and failure handling methods. Historical failure handling records and case analysis.

[0018] Professional knowledge Q&A: Prepare professional knowledge Q&A collection for problems that may be encountered during the operation and maintenance of the power distribution room, including common questions, difficult questions and answers.

[0019] In a preferred embodiment, the data training module includes:

[0020] Automatically generated data: Using text generation models, such as GPT, a large amount of question-answering dialogue data is automatically generated based on the relevant scenarios and knowledge of the distribution room. This data can cover common problems and scenarios in the operation and maintenance of the distribution room and serve as the basic data for model training.

[0021] Manually labeled data: Question-answering dialogue data from actual operation and maintenance scenarios are collected and manually labeled by power industry experts. These data are of high quality and can reflect the actual situation of distribution room operation and maintenance, which is crucial for model training.

[0022] Transfer learning technology: Use models pre-trained in general fields, such as BERT, to fine-tune them in the distribution room scenario. Through transfer learning, the model can quickly adapt to specific applications in the distribution room field.

[0023] Data fusion strategy: automatically generated data and manually annotated data are integrated to form a training data set. Automatically generated data provides a large number of samples, and manually annotated data provides high-quality samples. The combination of the two improves the model training effect.

[0024] Data enhancement technology: Expand the scale of training data through data enhancement technology, such as back translation and synonym replacement. Data enhancement can improve the generalization ability of the model.

[0025] Negative sample generation: Generate negative samples that are different from positive samples to train the model's ability to distinguish between positive and negative samples. Negative sample generation can improve the robustness of the model.

[0026] In a preferred embodiment, the optimization algorithm module uses an optimization algorithm such as Adam to adjust the model parameters during the model training process to minimize the loss function. The loss function measures the gap between the model prediction value and the true value. Minimizing the loss function means improving the accuracy of the model. The optimization algorithm module updates the model parameters through iteration, calculates the gradient of the loss function each iteration, and uses the gradient information to update the parameters;

[0027] The calculation formula of the optimization algorithm module is:

[0028] m_t=β_1*m_{t-1}+(1-β_1)*g_t;

[0029] v_t=β_2*v_{t-1}+(1-β_2)*g_t^2; where m_t is the momentum at the tth iteration, v_t is the variance at the tth iteration, g_t is the gradient at the tth iteration, and β_1 and β_2 are hyperparameters;

[0030] The parameter update formula is:

[0031] w_t=w_{t-1}-η*m_t / (sqrt(v_t)+ε); where w_t is the parameter of the tth iteration, η is the learning rate, and ε is the smoothing term.

[0032] In a preferred embodiment, the model training management module is internally configured with:

[0033] Large model training management platform: provides the computing resources, storage resources and management functions required for model training, and supports model training, deployment, evaluation and iterative optimization.

[0034] Knowledge base management component: manages and maintains the professional knowledge base related to the distribution room, including functions such as knowledge storage, query, update and deletion.

[0035] Intelligent dialogue management component: responsible for managing the interaction process between the robot and the user, including dialogue state tracking, multi-round dialogue management, and dialogue strategy selection.

[0036] Robot management component: realizes remote monitoring and control of robots, including robot status monitoring, task scheduling, and exception handling;

[0037] The data interface module is internally provided with:

[0038] Real-time data interface: responsible for obtaining real-time power data and environmental parameters from enterprise business systems, distribution room monitoring systems, etc., providing real-time data support for model training and robot interaction.

[0039] Historical data interface: responsible for obtaining historical power data and environmental parameters from the enterprise data center for model training and verification.

[0040] Data preprocessing module: preprocess the received data, including data cleaning, data conversion and data enhancement, to ensure data quality.

[0041] In a preferred embodiment, the business application module is internally provided with:

[0042] Intelligent question-and-answer system: Based on large models and robots, it provides intelligent question-and-answer services to answer users' questions about the operation and maintenance of the distribution room.

[0043] Fault warning system: Use real-time data and model analysis to provide early warning of potential faults in the distribution room.

[0044] Fault diagnosis system: Use the model to analyze the fault data in the distribution room to diagnose and locate the cause of the fault.

[0045] Intelligent inspection system: Use robots to conduct periodic inspections and combine model analysis to monitor the status of equipment in the distribution room.

[0046] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0047] 1. In the present invention, the electric power model module is internally provided with a pre-training module, a model fine-tuning module, a knowledge base module, a data training module and an optimization algorithm module. The pre-training module trains the model on a large amount of distribution room related data and documents through unsupervised learning, learns common language representation, and enables the model to have semantic understanding ability in the field of distribution, providing a basis for subsequent model fine-tuning.

[0048] Based on the pre-trained model, the model fine-tuning module uses a supervised dataset for a specific task to fine-tune the model so that the model can adapt to the application in the distribution room field and improve the accuracy and adaptability of the model.

[0049] The knowledge base module provides the model with professional knowledge, equipment information, operation and maintenance procedures in the field of power distribution, helping the model to better understand and answer questions related to the power distribution room and improve the intelligence of the model.

[0050] The data training module performs model training through automatically generated data, manually annotated data, and transfer learning technology, providing rich training samples so that the model can be exposed to common problems and scenarios in the operation and maintenance of the distribution room, thereby improving the generalization ability of the model.

[0051] The optimization algorithm module adjusts model parameters by using optimization algorithms such as Adam to minimize the loss function and improve the accuracy of the model. At the same time, momentum and adaptive learning rate are used to improve the training efficiency of the model.

[0052] 2. In the present invention, the operation and maintenance efficiency is improved: the robot can move and perform inspection tasks autonomously, reduce the labor intensity of manual inspection, and improve the operation and maintenance efficiency. Enhanced safety: the robot can realize 24-hour uninterrupted monitoring, detect abnormal situations in time, realize fault warning and fault diagnosis, and improve the safety of the distribution room. Improved intelligence: the large model provides a powerful "brain" for the robot, has professional knowledge in the field of power distribution, and can realize advanced functions such as intelligent question and answer, fault warning, and fault diagnosis. Reduced operation and maintenance costs: the robot can replace some manual inspections and operation and maintenance work, reducing operation and maintenance costs. Realize intelligent operation and maintenance: the system can realize intelligent functions such as intelligent question and answer, fault warning, and fault diagnosis, and promote the intelligent operation and maintenance management of the distribution room. Provide decision support: the system can collect and analyze various data of the distribution room to provide data support for operation and maintenance decisions. Improve user experience: the robot can interact with users through voice, provide intelligent services, and improve user experience. Continuous learning and optimization: the system has the ability to continuously learn and optimize, and can continuously improve performance and adapt to changes in the distribution room environment. Reduce manual workload: Robots can perform highly repetitive inspection tasks, reducing the workload of maintenance personnel. Provide all-round monitoring: The system can achieve all-round monitoring of the distribution room environment and promptly identify potential safety hazards. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 is the overall system block diagram of the present invention;

[0054] Figure 2 This is a system block diagram of the power large model module in the present invention;

[0055] Figure 3 This is a basic flow chart of the interactive scenario in the present invention.

[0056] Markings in the figure: 1-power supply module, 2-interactive robot module, 3-power large model module, 4-model training management module, 5-data interface module, 6-business application module, 7-pre-training module, 8-model fine-tuning module, 9-knowledge base module, 10-data training module, 11-optimization algorithm module. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0058] Reference Figure 1-3 ,

[0059] A smart energy management system based on interactive robot monitoring in a power distribution room, comprising a power supply module 1, an interactive robot module 2, a large power model module 3, a model training management module 4, a data interface module 5, a business application module 6, a pre-training module 7, a model fine-tuning module 8, a knowledge base module 9, a data training module 10 and an optimization algorithm module 11, wherein the output end of the power supply module 1 is connected to the input end of the interactive robot module 2, the output end of the interactive robot module 2 is connected to the input end of the large power model module 3, the output end of the large power model module 3 is connected to the input end of the model training management module 4, the output end of the model training management module 4 is connected to the input end of the data interface module 5, and the output end of the data interface module 5 is connected to the input end of the business application module 6.

[0060] The large power model module 3 is internally provided with a pre-training module 7, a model fine-tuning module 8, a knowledge base module 9, a data training module 10 and an optimization algorithm module 11, and the overall output end of the pre-training module 7, the model fine-tuning module 8, the knowledge base module 9, the data training module 10 and the optimization algorithm module 11 is connected to the input end of the large power model module 3.

[0061] Interactive robot module 2 realizes intelligent voice interaction and provides functions such as intelligent question and answer, fault warning, and fault diagnosis. The robot is equipped with hardware such as visible light camera, thermal imaging infrared camera, microphone matrix, and speaker to realize interaction with the environment and people. In addition, the robot realizes autonomous movement through SLAM navigation system and differential steering mechanism, and is equipped with automatic charging system.

[0062] Pre-training module 7 uses deep learning technology to build a model with a multi-layer neural network structure. Pre-training module 7 is based on the Transformer architecture, which can process long-distance dependency information and is suitable for tasks such as speech recognition and semantic understanding;

[0063] The calculation formula of the forward propagation of the pre-training module 7 during use is y=f(wx+b), where f is the activation function, w is the weight, b is the bias, and x is the input; the pre-training module 7 uses an unsupervised learning method to train the model on a large amount of distribution room related data and documents to learn a common language representation. The basic calculation formula for the training of the autoencoder or variational autoencoder is to maximize the likelihood function, that is, to maximize p(x|z).

[0064] The model fine-tuning module 8 fine-tunes the model based on the pre-trained model using a supervised dataset of a specific task to adapt the model to the application in the field of power distribution rooms;

[0065] Model fine-tuning module 8 involves the back propagation algorithm, and its basic calculation formula is Where η is the learning rate, C(w) is the loss function, and w is the weight

[0066] Knowledge base module 9 includes:

[0067] Professional knowledge of power distribution room: including basic knowledge of electrical engineering, power system, power automation and other related fields. Working principle, operating parameters, maintenance methods of power equipment, etc. Operating mechanism and mutual relationship of various parts of power system.

[0068] Equipment information: detailed information on the models, specifications, parameters, etc. of various equipment in the power distribution room. Normal operating range, alarm threshold, failure mode, etc. of the equipment.

[0069] Operation and maintenance procedures: Rules and regulations for the operation and maintenance of the distribution room, safety operating procedures, emergency plans, etc. Regular inspection and maintenance procedures for equipment. Fault handling procedures and methods for handling abnormal situations.

[0070] Distribution room environmental parameters: standard ranges of distribution room environmental parameters such as temperature, humidity, gas composition, etc. Important environmental factors related to power system operation.

[0071] Typical failure cases: common failure cases of distribution room equipment, failure cause analysis, and failure handling methods. Historical failure handling records and case analysis.

[0072] Professional knowledge Q&A: Prepare professional knowledge Q&A collection for problems that may be encountered during the operation and maintenance of the power distribution room, including common questions, difficult questions and answers.

[0073] The data training module 10 includes:

[0074] Automatically generated data: Using text generation models, such as GPT, a large amount of question-answering dialogue data is automatically generated based on the relevant scenarios and knowledge of the distribution room. This data can cover common problems and scenarios in the operation and maintenance of the distribution room and serve as the basic data for model training.

[0075] Manually labeled data: Question-answering dialogue data from actual operation and maintenance scenarios are collected and manually labeled by power industry experts. These data are of high quality and can reflect the actual situation of distribution room operation and maintenance, which is crucial for model training.

[0076] Transfer learning technology: Use models pre-trained in general fields, such as BERT, to fine-tune them in the distribution room scenario. Through transfer learning, the model can quickly adapt to specific applications in the distribution room field.

[0077] Data fusion strategy: automatically generated data and manually annotated data are integrated to form a training data set. Automatically generated data provides a large number of samples, and manually annotated data provides high-quality samples. The combination of the two improves the model training effect.

[0078] Data enhancement technology: Expand the scale of training data through data enhancement technology, such as back translation and synonym replacement. Data enhancement can improve the generalization ability of the model.

[0079] Negative sample generation: Generate negative samples that are different from positive samples to train the model's ability to distinguish between positive and negative samples. Negative sample generation can improve the robustness of the model.

[0080] During the model training process, the optimization algorithm module 11 uses an optimization algorithm such as Adam to adjust the model parameters to minimize the loss function. The loss function measures the gap between the model prediction value and the true value. Minimizing the loss function means improving the accuracy of the model. The optimization algorithm module 11 updates the model parameters through iterations, calculates the gradient of the loss function each iteration, and uses the gradient information to update the parameters;

[0081] The calculation formula of the optimization algorithm module 11 is:

[0082] m_t=β_1*m_{t-1}+(1-β_1)*g_t;

[0083] v_t=β_2*v_{t-1}+(1-β_2)*g_t^2; where m_t is the momentum at the tth iteration, v_t is the variance at the tth iteration, g_t is the gradient at the tth iteration, and β_1 and β_2 are hyperparameters;

[0084] The parameter update formula is:

[0085] w_t=w_{t-1}-η*m_t / (sqrt(v_t)+ε); where w_t is the parameter of the tth iteration, η is the learning rate, and ε is the smoothing term.

[0086] The internal settings of the model training management module 4 are:

[0087] Large model training management platform: provides the computing resources, storage resources and management functions required for model training, and supports model training, deployment, evaluation and iterative optimization.

[0088] Knowledge base management component: manages and maintains the professional knowledge base related to the distribution room, including functions such as knowledge storage, query, update and deletion.

[0089] Intelligent dialogue management component: responsible for managing the interaction process between the robot and the user, including dialogue state tracking, multi-round dialogue management, and dialogue strategy selection.

[0090] Robot management component: realizes remote monitoring and control of robots, including robot status monitoring, task scheduling, and exception handling;

[0091] The data interface module 5 is internally provided with:

[0092] Real-time data interface: responsible for obtaining real-time power data and environmental parameters from enterprise business systems, distribution room monitoring systems, etc., providing real-time data support for model training and robot interaction.

[0093] Historical data interface: responsible for obtaining historical power data and environmental parameters from the enterprise data center for model training and verification.

[0094] Data preprocessing module: preprocess the received data, including data cleaning, data conversion and data enhancement, to ensure data quality.

[0095] The internal configuration of the business application module 6 includes:

[0096] Intelligent question-and-answer system: Based on large models and robots, it provides intelligent question-and-answer services to answer users' questions about the operation and maintenance of the distribution room.

[0097] Fault warning system: Use real-time data and model analysis to provide early warning of potential faults in the distribution room.

[0098] Fault diagnosis system: Use the model to analyze the fault data in the distribution room to diagnose and locate the cause of the fault.

[0099] Intelligent inspection system: Use robots to conduct periodic inspections and combine them with model analysis to monitor the status of equipment in the distribution room.

[0100] In the present invention, the large power model module 3 is internally provided with a pre-training module 7, a model fine-tuning module 8, a knowledge base module 9, a data training module 10 and an optimization algorithm module 11. The pre-training module trains the model on a large amount of distribution room related data and documents through unsupervised learning, learns common language representation, and enables the model to have semantic understanding ability in the field of distribution, providing a basis for subsequent model fine-tuning.

[0101] Based on the pre-trained model, the model fine-tuning module uses a supervised dataset for a specific task to fine-tune the model so that the model can adapt to the application in the distribution room field and improve the accuracy and adaptability of the model.

[0102] The knowledge base module provides the model with professional knowledge, equipment information, operation and maintenance procedures in the field of power distribution, helping the model to better understand and answer questions related to the distribution room and improve the intelligence of the model.

[0103] The data training module performs model training through automatically generated data, manually annotated data, and transfer learning technology, providing rich training samples so that the model can be exposed to common problems and scenarios in the operation and maintenance of the distribution room, thereby improving the generalization ability of the model.

[0104] The optimization algorithm module adjusts model parameters by using optimization algorithms such as Adam to minimize the loss function and improve the accuracy of the model. At the same time, momentum and adaptive learning rate are used to improve the training efficiency of the model.

[0105] In the present invention, the operation and maintenance efficiency is improved: the robot can move and perform inspection tasks autonomously, reduce the labor intensity of manual inspection, and improve the operation and maintenance efficiency. Enhanced safety: the robot can realize 24-hour uninterrupted monitoring, detect abnormal situations in time, realize fault warning and fault diagnosis, and improve the safety of the distribution room. Improved intelligence: the large model provides a powerful "brain" for the robot, has professional knowledge in the field of power distribution, and can realize advanced functions such as intelligent question and answer, fault warning, and fault diagnosis. Reduced operation and maintenance costs: the robot can replace part of the manual inspection and operation and maintenance work, reducing the operation and maintenance costs. Realize intelligent operation and maintenance: the system can realize intelligent functions such as intelligent question and answer, fault warning, and fault diagnosis, and promote the intelligent operation and maintenance management of the distribution room. Provide decision support: the system can collect and analyze various data of the distribution room to provide data support for operation and maintenance decisions. Improve user experience: the robot can interact with users through voice, provide intelligent services, and improve user experience. Continuous learning and optimization: the system has the ability to continuously learn and optimize, and can continuously improve performance and adapt to changes in the distribution room environment. Reduce manual workload: Robots can perform highly repetitive inspection tasks, reducing the workload of maintenance personnel. Provide all-round monitoring: The system can achieve all-round monitoring of the distribution room environment and promptly identify potential safety hazards.

[0106] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprises" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprises a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0107] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A smart energy management system based on interactive robot monitoring in a power distribution room, comprising a power supply module (1), an interactive robot module (2), a large power model module (3), a model training management module (4), a data interface module (5), a business application module (6), a pre-training module (7), a model fine-tuning module (8), a knowledge base module (9), a data training module (10) and an optimization algorithm module (11), characterized in that: The output end of the power supply module (1) is connected to the input end of the interactive robot module (2), the output end of the interactive robot module (2) is connected to the input end of the large power model module (3), the output end of the large power model module (3) is connected to the input end of the model training management module (4), the output end of the model training management module (4) is connected to the input end of the data interface module (5), and the output end of the data interface module (5) is connected to the input end of the business application module (6).

2. The intelligent energy management system based on interactive robot monitoring in power distribution room as claimed in claim 1, characterized in that: The large electric power model module (3) is internally provided with a pre-training module (7), a model fine-tuning module (8), a knowledge base module (9), a data training module (10) and an optimization algorithm module (11); the overall output end of the pre-training module (7), the model fine-tuning module (8), the knowledge base module (9), the data training module (10) and the optimization algorithm module (11) is connected to the input end of the large electric power model module (3).

3. The intelligent energy management system based on interactive robot monitoring in power distribution room as claimed in claim 1, characterized in that: The interactive robot module (2) realizes intelligent voice interaction and provides intelligent question and answer, fault warning and fault diagnosis functions; the robot is equipped with a visible light camera, a thermal imaging infrared camera, a microphone matrix and speaker hardware to realize interaction with the environment and people; in addition, the robot realizes autonomous movement through a SLAM navigation system and a differential steering mechanism, and is equipped with an automatic charging system.

4. The intelligent energy management system based on interactive robot monitoring in power distribution room as claimed in claim 1, characterized in that: The pre-training module (7) uses deep learning technology to construct a model with a multi-layer neural network structure; the pre-training module (7) is based on the Transformer architecture, can process long-distance dependency information, and is suitable for speech recognition and semantic understanding tasks; The forward propagation calculation formula of the pre-training module (7) during use is y=f(wx+b), wherein f is the activation function, w is the weight, b is the bias, and x is the input; the pre-training module (7) uses an unsupervised learning method to train the model on a large amount of distribution room related data and documents to learn a common language representation; the training of an autoencoder or a variational autoencoder is involved, and its basic calculation formula is to maximize the likelihood function, that is, to maximize p(x|z).

5. The intelligent energy management system based on interactive robot monitoring in power distribution room as claimed in claim 1, characterized in that: The model fine-tuning module (8) fine-tunes the model based on the pre-trained model using a supervised data set of a specific task, so as to adapt the model to the application in the field of power distribution rooms; The model fine-tuning module (8) involves a back propagation algorithm, and its basic calculation formula is Δ Where η is the learning rate, C(w) is the loss function, and w is the weight.

6. The intelligent energy management system based on interactive robot monitoring in power distribution room as claimed in claim 1, characterized in that: The knowledge base module (9) comprises: Professional knowledge of power distribution room: including basic knowledge in electrical engineering, power system, power automation and related fields; working principle, operating parameters and maintenance methods of power equipment; operating mechanism and mutual relationship of various parts of power system; Equipment information: detailed information on the models, specifications, and parameters of various equipment in the power distribution room; normal working range, alarm threshold, and failure mode of the equipment; Operation and maintenance procedures: rules and regulations, safety operating procedures, and emergency plans for the operation and maintenance of the distribution room; regular inspections and maintenance procedures for equipment; fault handling procedures and methods for handling abnormal situations; Environmental parameters of the power distribution room: standard ranges of temperature, humidity, and gas composition environmental parameters of the power distribution room; important environmental factors related to the operation of the power system; Typical failure cases: common failure cases of distribution room equipment, failure cause analysis, and failure handling methods; historical failure handling records and case analysis; Professional knowledge Q&A: Prepare professional knowledge Q&A sets for problems that may be encountered during the operation and maintenance of the distribution room, including common questions, difficult questions and answers.

7. The intelligent energy management system based on interactive robot monitoring in power distribution room as claimed in claim 1, characterized in that: The data training module (10) comprises: Automatically generated data: Using text generation models, including GPT, a large amount of question-answering dialogue data is automatically generated based on the relevant scenarios and knowledge of the distribution room. This data can cover common problems and scenarios in the operation and maintenance of the distribution room and serve as the basic data for model training. Manually labeled data: Question-and-answer dialogue data from actual operation and maintenance scenarios are collected and manually labeled by power industry experts. This data is of high quality and can reflect the actual situation of distribution room operation and maintenance, which is crucial for model training. Transfer learning technology: Use models pre-trained in general fields, including BERT, to fine-tune them in the distribution room scenario; through transfer learning, the model can quickly adapt to specific applications in the distribution room field; Fusion data strategy: Fusion of automatically generated data and manually annotated data to form a training data set; automatically generated data provides a large number of samples, and manually annotated data provides high-quality samples. The combination of the two improves the model training effect; Data enhancement technology: Through data enhancement technology, including back translation and synonym replacement, the scale of training data can be expanded; data enhancement can improve the generalization ability of the model; Negative sample generation: Generate negative samples that are different from positive samples to train the model's ability to distinguish between positive and negative samples; negative sample generation can improve the robustness of the model.

8. The intelligent energy management system based on interactive robot monitoring in power distribution room as claimed in claim 1, characterized in that: The optimization algorithm module (11) uses the Adam optimization algorithm to adjust the model parameters during the model training process to minimize the loss function; the loss function measures the gap between the model prediction value and the true value, and minimizing the loss function means improving the accuracy of the model; the optimization algorithm module (11) updates the model parameters through iteration, calculates the gradient of the loss function in each iteration, and uses the gradient information to update the parameters; The calculation formula of the optimization algorithm module (11) is: m_t=β_1*m_{t-1}+(1-β_1)*g_t; v_t=β_2*v_{t-1}+(1-β_2)*g_t^2; where m_t is the momentum at the tth iteration, v_t is the variance at the tth iteration, g_t is the gradient at the tth iteration, and β_1 and β_2 are hyperparameters; The parameter update formula is: w_t=w_{t-1}-η*m_t / (sqrt(v_t)+ε); where w_t is the parameter of the tth iteration, η is the learning rate, and ε is the smoothing term.

9. The intelligent energy management system based on interactive robot monitoring in power distribution room as claimed in claim 1, characterized in that: The model training management module (4) is internally configured with: Large model training management platform: provides computing resources, storage resources and management functions required for model training, and supports model training, deployment, evaluation and iterative optimization; Knowledge base management component: manages and maintains the professional knowledge base related to the power distribution room, including knowledge storage, query, update and deletion functions; Intelligent dialogue management component: responsible for managing the interaction process between the robot and the user, including dialogue state tracking, multi-round dialogue management, and dialogue strategy selection; Robot management component: realizes remote monitoring and control of robots, including robot status monitoring, task scheduling and exception handling functions; The data interface module (5) is internally provided with: Real-time data interface: responsible for obtaining power data and environmental parameters from the enterprise business system and the distribution room monitoring system in real time, providing real-time data support for model training and robot interaction; Historical data interface: responsible for obtaining historical power data and environmental parameters from the enterprise data center for model training and verification; Data preprocessing module: preprocess the received data, including data cleaning, data conversion and data enhancement, to ensure data quality.

10. The intelligent energy management system based on interactive robot monitoring in power distribution room as claimed in claim 1, characterized in that: The business application module (6) is internally configured with: Intelligent question-answering system: Based on large models and robots, it provides intelligent question-answering services to answer users’ questions about the operation and maintenance of the power distribution room; Fault early warning system: Use real-time data and model analysis to provide early warning of potential faults in the power distribution room; Fault diagnosis system: Use the model to analyze the fault data in the distribution room to diagnose and locate the cause of the fault; Intelligent inspection system: Use robots to conduct periodic inspections and combine them with model analysis to monitor the status of equipment in the distribution room.

Images