Intelligent question-answering and energy efficiency optimization system for refrigerating room based on RAG architecture and Dify platform

Abstract

The invention provides an intelligent question-answering and energy efficiency optimization system for a refrigerating machine room based on an RAG architecture and a Dify platform, and the system achieves the efficient detection and energy-saving optimization of an abnormal value of the operation data of the refrigerating machine room through the steps of data preprocessing, feature selection, knowledge base retrieval, optimization suggestion generation and the like. Meanwhile, a domain knowledge base, a real-time data interface and a professional computing tool chain are packaged into a Dify tool, a question and answer Agent with an expert-level decision-making ability is constructed, and in combination with an ECharts visualization engine, image-text interactive intelligent question and answer services of equipment state monitoring, operation health diagnosis and energy efficiency optimization are provided.

Description

Technical Field

[0001] The present application relates to the field of intelligent industrial refrigeration system and energy consumption management, and in particular to an intelligent question-answering and energy efficiency optimization system for refrigeration rooms based on the RAG architecture and the Dify platform. Background Art

[0002] With the continuous expansion of industrial production and increasing automation, industrial refrigeration systems play a vital role in many fields. Their stable, efficient, and energy-saving operation is directly related to the smooth progress of the production process and the assurance of product quality. However, current industrial refrigeration systems still have significant deficiencies in intelligent assistance, especially in the application of intelligent question-answering systems.

[0003] Currently, most question-and-answer systems for industrial refrigeration systems are still in their early stages, offering only simple search functionality and extracting small amounts of parameter information from a pre-set database through keyword matching. These systems lack deep interaction capabilities, are unable to conduct multi-round conversations based on complex user questions, and are unable to accurately answer questions based on real-time data and dynamic operating conditions. Consequently, existing question-and-answer systems struggle to effectively address critical scenarios such as routine maintenance and training, operational health diagnosis and emergency response, energy-saving optimization, and operational evaluation in refrigeration rooms, and fail to meet real-world needs.

[0004] Taking the daily maintenance of the refrigeration room as an example, maintenance personnel need to keep abreast of the equipment's operating status, historical fault records, and relevant operating specifications. However, the existing question-and-answer system can only provide isolated parameter information and is unable to conduct comprehensive analysis and guidance based on the equipment's actual operating data and historical maintenance records. In terms of operational health diagnosis and emergency response, when an abnormality occurs in the refrigeration system, maintenance personnel need to quickly and accurately locate the cause of the fault and take appropriate measures. However, the existing system is unable to perform dynamic analysis based on real-time monitoring data and fault phenomena, and is unable to provide targeted solutions. In terms of energy-saving optimization and operation evaluation, the existing system is unable to make dynamic optimization recommendations based on real-time operating conditions and energy consumption data, making it difficult to help users achieve efficient and energy-saving operation of the system.

[0005] In addition, the application of existing retrieval-augmented generation (RAG) technology in industrial refrigeration systems also has limitations. Traditional RAG retrieval is mainly based on static files, such as equipment specifications, operating manuals, and other document contents. These static files cannot reflect the dynamic operating conditions of the refrigeration system in real time, such as the real-time operating parameters of the equipment, ambient temperature changes, load fluctuations, etc. The industrial refrigeration system is a complex dynamic system, and its operating status will be affected by multiple factors in real time. Therefore, the retrieval method based on static files cannot meet the real-time and dynamic requirements of the refrigeration system, and it is difficult to directly apply it to the intelligent question-answering scenario of the refrigeration system.

[0006] Therefore, in the field of intelligent industrial refrigeration system and energy consumption management, there is an urgent need for an intelligent question-and-answer system that can combine real-time dynamic data and has deep interactive capabilities, so as to improve the operating efficiency and management level of industrial refrigeration systems. Summary of the Invention

[0007] In view of the above-mentioned shortcomings of the prior art, the purpose of this application is to provide a technical solution for solving the technical problems of the existing industrial refrigeration system question-answering system in terms of interactivity, dynamics, real-time performance, etc.

[0008] To achieve the above and other related objectives, the first aspect of the present application provides a refrigeration room intelligent question-answering and energy efficiency optimization system based on the RAG architecture and the Dify platform, comprising:

[0009] The data preprocessing and feature selection module is used to preprocess the raw time-series data collected from the refrigeration room and extract several refrigeration room operating features that are most relevant to energy-saving optimization. The optimization suggestion generation module is used to mine frequent operating patterns based on the selected feature set and form high-energy consumption trigger conditions. The high-energy consumption trigger conditions are input into the pre-trained large model for classification and identification of abnormal energy consumption fluctuation scenarios. The fault modes and fault factors corresponding to abnormal energy consumption fluctuation scenarios are mapped to a predefined suggestion library to automatically generate executable energy-saving optimization solutions.

[0010] A knowledge base construction module is used to construct a static knowledge base based on the RAG architecture and a dynamic database based on the operating characteristics of the refrigeration room; the static knowledge base, the dynamic knowledge base and the predefined suggestion library are integrated to generate RAG search results that integrate static documents, dynamic operating conditions and energy-saving optimization solutions in response to user queries;

[0011] The tool chain encapsulation and agent integration module is used to encapsulate the agent and preset calculation modules related to the energy efficiency of the refrigeration room in the Dify development environment and publish them as schedulable tools; the agent recognizes user intentions through prompt words and RAG search results, and schedules the corresponding tools to perform energy efficiency-related calculation tasks and generate energy efficiency calculation results.

[0012] In some embodiments of the first aspect of the present application, the intelligent question-answering and energy efficiency optimization system for the refrigeration room based on the RAG architecture and the Dify platform also includes a visualization module; the visualization module is integrated with a visualization engine, which is used to generate and output one or more types of charts based on the energy efficiency calculation results and energy-saving optimization plans.

[0013] In some embodiments of the first aspect of the present application, the method of constructing the dynamic database includes: using a clustering algorithm to fill and / or eliminate data from the collected original time series data; using an adaptive equal-frequency discretization algorithm to convert continuous variables into discrete intervals; pre-screening variables that have a high correlation with the operation of the refrigeration room, and driving the large model through structured prompt words to extract several refrigeration room operation characteristics that are most relevant to energy-saving optimization from the conversation context information and user concerns.

[0014] In some embodiments of the first aspect of the present application, the optimization suggestion generation module mines frequent operating patterns on a selected feature set and forms high-energy consumption trigger conditions, including: mining frequent operating patterns from the selected feature set through the FP-Max algorithm, and filtering out short-term transient patterns by setting support and confidence thresholds, and forming typical high-energy consumption trigger conditions.

[0015] In some embodiments of the first aspect of the present application, the method of constructing a static knowledge base based on the RAG architecture includes: vectorizing multimodal documents related to the refrigeration room and storing them in the static knowledge base, and regularly updating them in synchronization with industry public information.

[0016] In some embodiments of the first aspect of the present application, the computing module integrated in the tool chain encapsulation and intelligent agent integration module includes any one or more combinations of the following: a data cleaning module, a feature engineering module, an EER energy efficiency calculation module, and a thermal balance rate analysis module.

[0017] In some embodiments of the first aspect of the present application, the predefined suggestion library includes known failure modes, failure factors and their corresponding energy-saving optimization solutions, and each energy-saving optimization solution is accompanied by a corresponding threshold and verification algorithm.

[0018] To achieve the above-mentioned purpose and other related purposes, the second aspect of the present application provides a method for intelligent question-answering and energy efficiency optimization of a refrigeration room based on the RAG architecture and the Dify platform, comprising: pre-processing the original time series data collected from the refrigeration room, and extracting several refrigeration room operation features that are most relevant to energy-saving optimization to build a dynamic database; mining frequent operation modes on the selected feature set and forming high-energy consumption trigger conditions; inputting the high-energy consumption trigger conditions into the pre-trained large model for classification and identifying abnormal energy consumption fluctuation scenarios; mapping the fault modes and fault factors corresponding to the abnormal energy consumption fluctuation scenarios to the pre-trained large model; Define a suggestion library to automatically generate executable energy-saving optimization solutions; build a static knowledge base based on the RAG architecture, and fuse the static knowledge base, dynamic knowledge base and predefined suggestion library to generate RAG search results that integrate static documents, dynamic working conditions and energy-saving optimization solutions in response to user queries; encapsulate the intelligent agent and the preset calculation modules related to the energy efficiency of the refrigeration room in the Dify development environment and publish them as schedulable tools; the intelligent agent recognizes the user's intention through prompt words and RAG search results, and schedules the corresponding tools to perform energy efficiency-related calculation tasks and generate energy efficiency calculation results.

[0019] To achieve the above-mentioned purpose and other related purposes, the third aspect of the present application provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements the intelligent question-answering and energy efficiency optimization method for the refrigeration room based on the RAG architecture and the Dify platform.

[0020] To achieve the above-mentioned objectives and other related objectives, the fifth aspect of the present application provides a computer device, including a memory, a processor, and a computer program stored on the memory; the processor executes the computer program to implement the intelligent question-answering and energy efficiency optimization method for the refrigeration room based on the RAG architecture and the Dify platform.

[0021] As described above, the intelligent question-answering and energy efficiency optimization system for refrigeration rooms based on the RAG architecture and the Dify platform of this application has the following beneficial effects:

[0022] (1) The fusion of multimodal static knowledge base and real-time operation data not only covers historical information such as drawings, logs, and process flows, but also accesses the dynamic parameters of PLCs and sensors to achieve a comprehensive assessment of the operating status of the refrigeration room throughout its life cycle.

[0023] (2) Through clustering algorithms, outlier filling, equal-frequency discretization, and automatic feature selection driven by structured prompts, as well as high-frequency pattern mining and professional calculation module verification based on FP-Max, energy consumption anomaly detection and optimization suggestions are ensured to be highly consistent with actual working conditions and physical laws.

[0024] (3) Real-time data collection and RAG retrieval can respond to user queries in seconds. Dify Agent is an instant scheduling calculation tool that quickly outputs verified energy-saving optimization plans, significantly shortening the time from data acquisition to decision-making recommendations.

[0025] (4) Integrate ECharts interactive charts, introduce multiple visualization forms such as line charts, heat maps, and dashboards, support custom filtering and export, and help operation and maintenance personnel intuitively understand the relationship between energy consumption trends and faults, thereby improving decision-making efficiency.

[0026] (5) The modular architecture is deeply integrated with the Dify platform, supporting multi-model parallelism and flexible expansion of the tool chain. It can be smoothly migrated to high-precision manufacturing scenarios such as air compressor stations and semiconductor production lines, meeting the intelligent question-answering and energy-saving optimization needs of different industrial systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Shown is a module diagram of a refrigeration room energy-saving optimization system in one embodiment of the present application.

[0028] Figure 2 Shown is a module operation diagram of a refrigeration room energy-saving optimization system in one embodiment of the present application.

[0029] Figure 3 Shown is a flow chart of a method for energy-saving optimization of a refrigeration room in one embodiment of the present application.

[0030] Figure 4 Shown is a structural diagram of a computer device in one embodiment of the present application. DETAILED DESCRIPTION

[0031] The following describes the embodiments of the present application through specific examples. Those skilled in the art can easily understand the other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied through other different specific embodiments. The details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that the following embodiments and features in the embodiments can be combined with each other unless they conflict.

[0032] Before further explaining the present invention in detail, the nouns and terms involved in the embodiments of the present invention are explained. The nouns and terms involved in the embodiments of the present invention are subject to the following interpretations:

[0033] <1> The RAG architecture (Retrieval-Augmented Generation) combines information retrieval technology with large generative language models. By integrating external knowledge bases in real time, it improves the model's performance on knowledge-intensive tasks. The RAG architecture consists of two core modules: the Retriever module and the Generator module. The Retriever module retrieves information relevant to the input query from a large-scale knowledge base, while the Generator module generates the final answer based on the retrieved documents and the input query.

[0034] <2> The Dify platform, an open-source Large Language Model (LLM) application development platform, integrates the concepts of Backend as a Service (BaaS) and Large Language Model Operations (LLMOps). Designed to simplify and accelerate the creation and deployment of generative AI applications, it provides developers with a user-friendly interface and a suite of powerful tools. The Dify platform features a high-quality RAG engine, an intuitive prompt orchestration interface, a robust agent framework, and supports a variety of mainstream large language models.

[0035] The embodiment of the present application aims to provide a technical solution that can perform energy consumption data mining and energy-saving optimization of refrigeration rooms based on a large language model, as well as a combined application with an intelligent question-and-answer system for refrigeration rooms that is deeply integrated with the RAG architecture and the Dify platform. The embodiment of the present application achieves efficient detection and energy-saving optimization of abnormal values ​​in refrigeration room operating data through steps such as data preprocessing, feature selection, knowledge discovery, and generation of optimization suggestions. At the same time, the domain knowledge base, real-time data interface, and professional computing tool chain are encapsulated as Dify tools to build a question-and-answer agent with expert-level decision-making capabilities, and combined with the ECharts visualization engine to provide interactive intelligent question-and-answer services with graphics for equipment status monitoring, operational health diagnosis, and energy efficiency optimization.

[0036] Combine Figure 1 Detailed explanation, Figure 1 This diagram shows the structure of a refrigeration room intelligent question-answering and energy efficiency optimization system based on the RAG architecture and the Dify platform, according to an embodiment of the present invention. The refrigeration room energy efficiency optimization system in this embodiment primarily includes the following modules: data preprocessing and feature selection module 101, optimization suggestion generation module 102, knowledge base construction module 103, tool chain encapsulation and agent integration module 104, and visualization module 105.

[0037] For the convenience of understanding by those skilled in the art, Figure 1 The module diagram of the energy-saving optimization system for the refrigeration room is shown, and Figure 2 The operation flow diagram is shown in detail.

[0038] The data preprocessing and feature selection module 101 is used to preprocess the original time series data collected from the refrigeration room and extract several refrigeration room operation features that are most relevant to energy-saving optimization.

[0039] Furthermore, the construction method of the dynamic database includes: first, using a clustering algorithm to fill and / or eliminate data from the collected original time series data; then using an adaptive equal-frequency discretization algorithm to convert continuous variables into discrete intervals; finally, pre-screening variables with a high correlation with the operation of the refrigeration room, and driving the large model through structured prompt words to extract several refrigeration room operation characteristics that are most relevant to energy-saving optimization from the conversation context information and user concerns.

[0040] Specifically, in refrigeration rooms, advanced monitoring systems typically collect a large amount of critical operational data. This data primarily includes raw time-series data such as temperature, pressure, and power from programmable logic controllers (PLCs), sensors, and historical logs. As the core of the automated control system, PLCs record equipment operating instructions and status feedback in real time. Sensors are distributed throughout key locations in the refrigeration room, accurately measuring ambient and equipment parameters such as temperature and pressure. This data, recorded in a time-series format, provides a comprehensive overview of refrigeration operating conditions. For example, temperature data reflects the refrigerant temperature, ambient temperature, and equipment surface temperature in the refrigeration system, helping monitoring personnel promptly detect abnormal temperature increases or insufficient cooling. Pressure data captures the high and low pressure states of the refrigerant, ensuring the proper operation of the refrigeration cycle. Power data reflects the energy consumption of the refrigeration equipment, providing a basis for energy-saving optimization. This raw time-series data not only provides real-time data for daily equipment monitoring but also provides rich information support for subsequent data analysis, operational health diagnostics, and preventive maintenance.

[0041] Clustering algorithm is a kind of unsupervised learning method, and its main function is to divide the data points in the data set into several clusters (or groups) so that the data points in the same cluster are as similar as possible, while the data points between different clusters are as different as possible. Specifically, algorithms such as K-Means clustering, DBSCAN clustering (Density-Based Spatial Clustering of Applications with Noise), hierarchical clustering, GMM clustering (Gaussian Mixture Models) can be used to identify and fill abnormal data, eliminate missing data or noise data. Of course, other clustering algorithms outside these clustering algorithms can also be used. In fact, algorithms that can fill abnormal data, eliminate missing data or noise data can be applied to the embodiments of the present application.

[0042] The adaptive equal-frequency discretization algorithm is a method for converting continuous data into discrete data. It can dynamically adjust the discretization interval based on the actual data distribution. The core concept is to divide the data into several intervals so that the number of data points in each interval is roughly equal. This method ensures that the importance of each interval is relatively balanced, avoiding the situation where some intervals are too sparse or too dense due to uneven data distribution. It should be understood that in data analysis and machine learning, data often exists in the form of continuous variables, while many data mining algorithms are more suitable for processing discrete variables. Discrete variables can simplify model complexity, improve algorithm efficiency and interpretability, reduce noise, and enhance model robustness. Taking the key parameters in the refrigeration room as an example, such as the cold water outlet temperature and evaporation pressure, these parameters are usually continuous numerical variables. Through adaptive equal-frequency discretization, they can be divided into several discrete intervals. For example: the original data range of the cold water outlet temperature: 5℃-15℃; the intervals after discretization: [5℃-7℃], (7℃-9℃], (9℃-11℃], (11℃-13℃], (13℃-15℃]. For another example, the original data range of the evaporation pressure: 0.5Mpa-1.5Mpa, the intervals after discretization: [0.5MPa-0.7MPa], (0.7MPa-0.9MPa], (0.9MPa-1.1MPa], (1.1MPa-1.3MPa], (1.3MPa-1.5MPa]. Each continuous variable is converted into a discrete interval to facilitate subsequent data mining and analysis. It will be understood that the above case is only used as an example and is not used to limit the discretization method in the embodiments of the present application.

[0043] Pre-screening variables with a high correlation to refrigeration room operations can be accomplished by leveraging domain expertise. Pre-screening based on empirical values ​​can more accurately locate target variables. Understandably, in data analysis and machine learning projects, the data is often very high-dimensional, containing a large number of variables that may be unrelated to the target. Directly analyzing this data can introduce noise, reducing model performance and efficiency. Leveraging refrigeration expert knowledge, variables closely related to refrigeration room operations can be screened from a wide range of variables, such as chilled water outlet temperature, evaporation pressure, condensing pressure, and unit start-stop sequence. The chilled water outlet temperature directly affects cooling performance and energy consumption, the evaporation pressure reflects the operating status of the refrigeration system, and the unit start-stop sequence directly determines the energy consumption pattern of the refrigeration system.

[0044] However, even if the knowledge of domain experts is used to pre-screen variables that are closely related to the operation of the refrigeration room, there may still be a large amount of redundant information or hidden associations in the data. Therefore, a structured prompt can be used to clearly tell the model that the goal of the task is to extract features related to energy-saving optimization. The prompt includes the operating background of the refrigeration room, the key variables that have been screened, and specific issues that the user is concerned about (such as energy consumption optimization). Prompts are instructions input by users to the large language model to guide the model to generate specific outputs, and structured prompts refer to prompts with a clear structure and format that can effectively guide the model to complete specific tasks. For example, structured prompts include the following structural elements: clarifying the task goal (telling the model that we need to extract high-value features related to energy-saving optimization), providing contextual information (helping the model understand the background of the data and the focus of the user), and guiding model output (ensuring that the model can accurately extract the most valuable information). Taking a refrigeration room as an example, a structured prompt might read, "The following is the operating data of a refrigeration room over the past month, including unit start and stop times, tower temperature, energy consumption data, etc. Please extract high-value features related to energy-saving optimization from these data, focusing on the unit start and stop sequence, tower temperature to energy consumption ratio, etc., and sort them by importance." Conversation context information refers to the content of the conversation between the user and the large model, while user focus refers to features that the user is particularly concerned about, such as "unit start and stop sequence," "tower temperature to energy consumption ratio," and so on. Through structured prompts, the model can automatically extract the most valuable features from context information and user focus for the user's subsequent energy-saving optimization analysis.

[0045] The optimization suggestion generation module 102 is used to mine frequent operating patterns on the selected feature set and form high-energy consumption trigger conditions; input the high-energy consumption trigger conditions into the pre-trained large model for classification and identify abnormal energy consumption fluctuation scenarios; map the fault modes and fault factors corresponding to the abnormal energy consumption fluctuation scenarios to the predefined suggestion library to automatically generate executable energy-saving optimization plans.

[0046] Furthermore, the FP-Max algorithm can be used to mine frequent operating patterns from the selected feature set. By setting support and confidence thresholds, short-term transient patterns can be filtered out, thereby forming typical trigger conditions for high energy consumption. It should be understood that the FP-Max algorithm is based on a frequent pattern tree (FP-Tree) for mining frequent patterns. The FP-Tree construction process involves converting the data in the selected feature set into a transaction dataset; each transaction represents a record of a device operation and contains a set of feature values; and then constructing an FP-Tree to store frequent patterns in a tree format for efficient mining. The process of mining frequent patterns requires setting two key parameters: support and confidence. Support indicates the frequency of a pattern in the dataset. The support threshold is used to filter out infrequent patterns and retain only those with higher frequency. For example, if the support threshold is set to 10%, only patterns with a frequency exceeding 10% will be retained. Confidence indicates the reliability of a pattern and is used to filter out unreliable patterns and retain only those with higher confidence. For example, if the confidence threshold is set to 80%, only patterns with a confidence exceeding 80% will be retained. By setting support and confidence thresholds, we can effectively filter out short-term transient patterns. These patterns are often caused by short-term fluctuations or noise in the equipment and have low reference value for energy optimization. As mentioned earlier, the frequently mined operating patterns have been rigorously screened and have high confidence and support. By analyzing these mined frequent operating patterns, we can identify pattern characteristics associated with high energy consumption. For example, if a frequent pattern indicates a significant increase in energy consumption under a specific combination of cold water outlet temperature and evaporation pressure, these patterns can be defined as high-energy consumption trigger conditions.

[0047] Preferably, the refrigeration room operating data has the characteristics of large and complex data volume, high pattern repeatability, and high data real-time requirements. Specifically, the refrigeration room operating data includes a large amount of sensor data, equipment operating parameters (such as temperature, pressure, flow, power, etc.), operation logs, etc. These data are highly dynamic and complex in time series; the operating mode of the refrigeration room may show similar patterns in the operating parameters of the equipment in different time periods (such as daytime and nighttime), different seasons or different load conditions. These repetitive patterns appear in the data as frequently occurring transaction combinations; the operating status of the refrigeration room needs to be monitored and analyzed in real time, and potential energy consumption problems or equipment failures need to be discovered in time, so the frequent pattern mining algorithm needs to process real-time data efficiently.

[0048] Therefore, the embodiment of the present application introduces a dynamic node insertion scheme to optimize the FP-Tree construction process. By checking whether the items in the current transaction already exist in the tree, it avoids repeated generation of nodes, thereby reducing the width and depth of the tree, which can effectively cope with the characteristics of the refrigeration room operation data.

[0049] The specific implementation steps for dynamic node insertion include: first, creating an empty FP-Tree containing a root node (usually marked as "null"); and initializing a global frequent item header table to record all frequent items and their corresponding linked list head pointers. Second, sorting the items in each transaction according to the order in the global frequent item header table; the frequent item header table is sorted in descending order of support to ensure path uniqueness. Third, performing the following steps for each transaction: starting from the root node of the FP-Tree, processing the items in the transaction one by one; checking whether the path exists, including: starting from the current node, traversing down the tree, checking whether the items in the current transaction already exist on the path; if the current item already exists on the path, directly updating the support of the node (support += 1); if the current item does not exist on the path, creating a new node and inserting it into the path; if a new node is created, updating the child node pointer of its parent node and adding the new node to the child node linked list; and finally, updating the linked list pointer of the corresponding item in the global frequent item header table and adding the new node to the linked list. Fourth, repeat the above insertion process for all transactions in the dataset until all transactions are inserted into the FP-Tree.

[0050] Furthermore, a pre-trained large model (such as Deepseek) can be used to classify and judge the frequent operating modes mined, and identify abnormal energy consumption fluctuation scenarios. In multi-classification tasks, the operating modes can be divided into normal energy consumption modes and abnormal energy consumption modes. The characteristics of the frequent operating modes (such as cold water outlet temperature, evaporation pressure, etc.) are input into the large model. The large model outputs the classification results corresponding to each mode, and accurately identifies abnormal energy consumption fluctuation scenarios. For example, the input of the large model is "cold water outlet temperature is 10℃~12℃ and evaporation pressure is 0.6MPa-0.8Mpa", and the classification result output by the large model is to judge that the mode is "abnormal start and stop of cooling water pump" scenario.

[0051] Furthermore, the predefined suggestion library contains known failure modes, failure factors and their corresponding energy-saving optimization solutions, and each energy-saving optimization solution is accompanied by a corresponding threshold and verification algorithm.

[0052] The mapping process involves first matching the identified abnormal energy consumption fluctuation scenario with the fault patterns in the predefined suggestion library to find the fault pattern with the highest similarity. Second, based on the matching results, the corresponding fault factors are identified. Finally, solutions related to the fault factors are extracted from the library. A fault mode refers to a possible type of fault in a device or system, such as "reduced cooling tower efficiency" or "faulty cooling unit valve." A fault factor refers to the specific cause of the fault, such as "insufficient number of cooling towers" or "improper cooling unit valve opening." Once a fault mode or fault factor is detected, a corresponding solution is searched from the predefined suggestion library and an executable energy-saving optimization plan is generated. For example, if the fault mode matched in the predefined suggestion library is excessively high chilled water outlet temperature and significantly increased energy consumption, and the identified fault factor is insufficient chiller cooling water flow, the solution extracted from the library for this fault factor is to check the operating status of the cooling water pump, clean the cooling water pipes, and adjust the cooling water flow. Each energy-saving optimization plan is accompanied by a threshold and a verification algorithm. The threshold serves as the standard for determining whether a specific action is required, while the verification algorithm verifies that the action was executed correctly. For example, when the efficiency of a cooling tower falls below a certain threshold (e.g., 80%), it is recommended to increase the number of cooling towers. After increasing the number of cooling towers, a formula (e.g., efficiency improvement percentage) is used to verify whether the expected effect has been achieved.

[0053] The knowledge base construction module 103 is used to construct a static knowledge base based on the RAG architecture and to construct a dynamic database based on the operating characteristics of the refrigeration room; the static knowledge base, the dynamic knowledge base and the predefined suggestion library are integrated to form a retrieval result that integrates static documents, dynamic working conditions and energy-saving optimization solutions in response to user queries.

[0054] Furthermore, the methods of constructing a static knowledge base based on the RAG architecture include: vectorizing the multimodal documents related to the refrigeration room and storing them in the static knowledge base, and regularly updating them in sync with industry public information. Multimodal documents related to the refrigeration room include different types of documents such as bidding documents, equipment maintenance systems, process flow charts, etc. These documents not only contain text information, but may also contain non-text content such as images and tables. In the process of vectorization, text vectorization can use natural language processing technology (such as BERT, GPT, etc.) to convert text content into high-dimensional vectors, image vectorization can use computer vision technology (such as ResNet, YOLO, etc.) to convert image content into vectors, and table vectorization can extract key information after structured processing of table data and convert it into vectors. The vectorized documents are stored in a static knowledge base for subsequent retrieval and query. At the same time, in order to ensure the timeliness and accuracy of the knowledge base content, it needs to be regularly updated with industry public information. This process includes: regularly collecting the latest information from industry websites, public databases, academic papers and other sources; preprocessing the collected information, such as text cleaning, image recognition, table parsing, etc.; vectorizing the processed data and storing it in the knowledge base, updating the index; versioning the knowledge base, recording the content and time of each update, for easy traceability and auditing.

[0055] It's worth noting that knowledge bases based on non-RAG architectures rely on manual operations to update knowledge, making it difficult to reflect the latest information in real time. They typically only contain structured data (such as tables in a database) but have limited support for unstructured data (such as documents and images). Knowledge bases based on RAG architectures can integrate multimodal data and dynamically update content. They leverage natural language processing (NLP) and computer vision (CV) technologies to automatically extract and integrate knowledge, reducing manual intervention. At the same time, existing retrieval-augmented generation (RAG) technology has limitations in its application in industrial refrigeration systems. Traditional RAG retrieval is primarily based on static documents, such as equipment specifications and operating manuals. These static documents cannot reflect the dynamic operating conditions of the refrigeration system in real time, such as the equipment's real-time operating parameters, ambient temperature changes, and load fluctuations.

[0056] Furthermore, the latest operating data is injected into the static knowledge base in real time, aiming to collect the latest operating data from the field through scheduled tasks or on-demand calls, and convert it into text or vector form, and inject it into the knowledge retrieval layer in real time to ensure that in the question-answering system, both the design parameters can be referenced and the current operating conditions can be verified, providing more accurate and targeted information support. Traditional text retrieval algorithms (such as TF-IDF, BM25, etc.) can be used to retrieve static documents, and retrieval algorithms based on vector similarity (such as pre-similarity, Euclidean distance, etc.) can be used to retrieve dynamic vector data. Combine the results of text retrieval and vector retrieval to generate comprehensive retrieval results.

[0057] The latest operating data can be collected from field equipment such as equipment sensors, control systems (such as PLCs), and monitoring systems. These data include real-time parameters such as temperature, pressure, power, and equipment status. The knowledge retrieval layer is a retrieval system that integrates static documents and dynamic data. Static documents include pre-stored design parameters, maintenance logs, process flow charts, etc., while dynamic data are operating data collected in real time. The retrieval system can retrieve relevant design parameters from the static knowledge base based on user questions, and can also verify the current operating conditions in combination with real-time injected dynamic data. For example, when a user asks "Is the current equipment operating normally?", the retrieval system can combine real-time data (such as temperature, pressure, etc.) and design parameters to determine whether the equipment is operating within the normal range and give a corresponding answer.

[0058] It's understandable that this approach, combining a static knowledge base with dynamic data, has a wide range of applications, including but not limited to: equipment operation monitoring scenarios, where users can query the equipment's operating status in real time through a question-and-answer system, providing accurate answers based on design parameters and real-time data; operational health diagnostics scenarios, where users inquire about equipment anomalies, and the system quickly locates the problem and provides solutions based on real-time data and historical maintenance logs; and process optimization scenarios, where users inquire about how to optimize the process, and the system provides optimization suggestions based on design documents and real-time operational data. Of course, the application scenarios far exceed those listed above, and since they are inexhaustible, I will not elaborate on them here.

[0059] The tool chain encapsulation and agent integration module 104 is used to encapsulate the agent and the preset calculation module related to the energy efficiency of the refrigeration room in the Dify development environment and publish them as a schedulable tool; the agent recognizes the user's intention through prompt words and RAG search results, and schedules the corresponding tool to perform energy efficiency-related calculation tasks accordingly.

[0060] Furthermore, the functional modules encapsulated in the Dify development environment include, but are not limited to, a data cleaning module, a feature engineering module, an EER energy efficiency calculation module, a thermal balance rate analysis module, an energy consumption data mining module, an expert diagnosis module, a data verification module, and a chart generation module. These modules are encapsulated as independent functions or services, making them easier for agents to call in subsequent processes. Of course, the modules encapsulated in the Dify development environment are not limited to those listed above; corresponding calculation modules can be encapsulated as needed, and this is not a limitation in the present embodiment.

[0061] It should be understood that the data cleaning module is used to remove noise data, fill missing values, and standardize data formats. The feature engineering module is used to extract and transform data features, such as calculating new feature variables and normalizing data. The energy efficiency calculation module is used to calculate the energy efficiency ratio (EER) based on equipment operating data and evaluate the equipment's energy efficiency. The thermal balance analysis module is used to analyze the system's thermal balance and evaluate the equipment's thermal efficiency. The energy consumption data mining module is used to extract valuable information and patterns from large amounts of energy consumption data. By analyzing patterns and trends in the data, it identifies high-energy consumption links and potential energy-saving points. The expert diagnosis module evaluates the operating status and operational health of the refrigeration room based on expert experience and rules, providing accurate fault location and optimization recommendations. The data verification module verifies the accuracy and completeness of data, ensuring that the collected data meets preset quality standards and preventing erroneous data from affecting subsequent analysis and decision-making. The chart generation module automatically generates charts based on the results generated by the data verification module.

[0062] Furthermore, the Agent automatically dispatches the appropriate tool based on the prompt at runtime. For ease of description, "Agent" will refer to the Agent, and "Prompt" will refer to the prompt. A prompt is a structured instruction that defines the task's objectives and context. Based on the prompt's content, the Agent automatically selects and invokes the packaged tool to complete data processing, energy efficiency calculation (EER) and thermal balance analysis, and generate answers based on the calculation results that comply with physical and industry standards.

[0063] Furthermore, during the debugging phase, the Agent supports running multiple models in parallel, such as Deepseek and GPT-4. The Agent dynamically selects the optimal model based on the task type (e.g., health diagnosis or energy efficiency assessment). For example, for health diagnosis, a model that excels at logical reasoning and fault pattern recognition can be selected. For energy efficiency assessment, a model that excels at data analysis and prediction can be selected.

[0064] Take a specific operational health diagnosis scenario as an example: the prompt for operational health diagnosis is set to "Check whether the device has any abnormalities." The agent's operating steps are as follows: first, call the data cleaning module to process the device operation data, then call the feature engineering module to extract key features, then call the operational health diagnosis model (such as Deepseek) to analyze the data and generate diagnostic results, and finally output a diagnostic report and indicate whether the device is abnormal and the possible causes of the abnormality. Take another specific energy-saving assessment scenario as an example: the prompt for energy-saving assessment is set to "Evaluate the energy efficiency level of the device." The agent's operating steps are as follows: first, call the data cleaning module to process the device operation data, then call the EER energy efficiency calculation module to calculate the energy efficiency ratio of the device, then call the energy-saving assessment model (such as GPT-4) to analyze the energy efficiency level and propose optimization suggestions, and finally output an energy-saving assessment report and propose the current energy efficiency level and corresponding optimization suggestions.

[0065] Therefore, the toolchain encapsulation and agent integration module automates data processing and computational verification by encapsulating specialized computational modules within the Dify development environment and automatically scheduling these tools at runtime through the Agent. During the debugging phase, the Agent supports the parallel execution of multiple models and dynamically selects the optimal model for each task.

[0066] The visualization module 105 is used to render and generate different types of charts and output reports based on the generated analysis results and optimization suggestions.

[0067] Furthermore, the visualization module 105 integrates the Echarts visualization engine to present the analysis results and optimization suggestions in the form of intuitive charts. It should be understood that the Echarts visualization engine is an open-source, powerful visualization library that supports the generation of various types of charts. For example, if the analysis result is the change in device energy consumption over time, a line chart is generated; if the analysis result is the distribution of device energy consumption in different time periods, a heat map is generated; if the analysis result is the current energy consumption level of the device, a dashboard is generated.

[0068] For example, a specific integration method can embed the Echarts visualization engine into the front-end page, and use the API interface provided by Echarts to dynamically generate charts based on the analysis results generated by the system. The front-end page supports users to customize filtering based on time windows, equipment types, energy consumption levels and other conditions. The time window can select a specific time range, such as the past 24 hours, the past week, etc.; the equipment type can select a specific type of equipment, such as chillers, cooling towers, etc.; the energy consumption level can select a specific energy consumption level, such as high energy consumption, medium energy consumption, low energy consumption, etc. The filtering function can provide a drop-down menu or input box for users to select conditions, and then dynamically update the charts and analysis results based on the conditions selected by the user.

[0069] Furthermore, users can export the charts and analysis results on the current page into a report with both pictures and text in one click. The report format can be PDF, Word, or other common document formats. In addition, in the Q&A interface, each user's query not only receives a text answer, but also a corresponding dynamic visualization of the results. For example, if a user asks "What has been the change in the energy consumption of the device over the past 24 hours?", the system generates a line chart and displays it in the Q&A interface, while providing a text answer "The energy consumption of the device has been on an upward trend over the past 24 hours, with an average energy consumption of 100KW."

[0070] For example, consider a specific operational health diagnosis scenario: a user queries whether a device is experiencing anomalies, and the system generates a heat map showing the device's energy consumption distribution over different time periods. A user queries the cause of a device failure, and the system generates a chart and provides detailed text analysis. Another example involves an energy-saving assessment scenario: a user queries the device's energy-saving potential, and the system generates a chart comparing energy consumption before and after optimization. A user queries for energy-saving suggestions, and the system generates a chart and provides specific optimization measures.

[0071] Above, the intelligent question-and-answer and energy efficiency optimization system for the refrigeration room based on the RAG architecture and the Dify platform provided in the embodiment of the present application has been explained and illustrated in detail. In fact, the energy-saving optimization system for the refrigeration room of the embodiment of the present application can be applied to a variety of scenarios, including but not limited to: daily maintenance and training of the refrigeration room, operational health diagnosis and emergency response, energy-saving optimization and operational evaluation, and cross-domain intelligent question-and-answer, etc. In the daily maintenance and training process of the refrigeration room, new employees use the intelligent question-and-answer interface, combined with the static knowledge base and real-time working condition data, to quickly understand the system structure, the meaning of key parameters and common fault handling processes, and can call the calculation module in the Dify platform to verify the rationality of the parameters, significantly shortening the training cycle. In the process of operational health diagnosis and emergency response, the system integrates historical fault records and PLC / sensor data in real time, automatically generates diagnostic ideas and troubleshooting suggestions based on expert rules, and intuitively presents abnormal trends through a visual dashboard to help operation and maintenance personnel quickly locate faults. During energy-saving optimization and operational evaluation, the energy consumption data mining module can be used to compare real-time and historical energy consumption patterns. Combining data mining results with EER calculations, it automatically outputs dynamic energy consumption reports and optimization strategies. ECharts also generates trend and comparison charts to support management in making scientific energy-saving decisions. During cross-domain intelligent question-and-answer sessions, the platform architecture can be smoothly migrated to high-precision manufacturing scenarios such as high-efficiency air compressor stations and semiconductor production lines. By loading the corresponding domain knowledge base and parameter models, it provides integrated services for multi-scenario equipment status monitoring, operational health diagnosis, and energy-saving consulting.

[0072] In order to facilitate those skilled in the art to better understand the technical solutions of the embodiments of the present application, a specific embodiment is now provided to facilitate understanding.

[0073] First, build a knowledge base: re-extract computer room drawings, construction plans, and historical fault logs from the project management system, pre-process unstructured documents using OCR and image recognition tools, and then store them in a vectorized knowledge base using domain-defined text and image embedding models.

[0074] Secondly, access real-time data: deploy lightweight Python plug-ins on the PLC and field sensors to regularly collect parameters such as temperature, pressure, and power. After K-Means cleaning and equal-frequency discretization, the data is uploaded to the data fusion platform through the Dify workflow.

[0075] Furthermore, specialized tools are integrated: a kernel density estimation algorithm, an energy consumption data mining module, and an EER and thermal balance rate calculation module are registered on the Dify platform and encapsulated as callable interfaces. When a user asks, "Is the current return water temperature abnormal?", the Agent automatically dispatches the corresponding tool to complete the numerical verification and feeds the results back to the LLM.

[0076] Finally, the order of the question-answer response process is as follows: (1) the user initiates a query; (2) the RAG engine retrieves static knowledge and the latest working conditions; (3) the agent relies on the Dify tool chain to perform calculations and verifications; (4) the LLM combines contextual information with calculation results to generate answers with pictures and texts; (5) the front end renders dynamic trend charts / dashboards through ECharts; (6) the user feedback quality assessment triggers the iteration of the knowledge base and tool rules.

[0077] Therefore, this invention deeply integrates the energy consumption data mining capabilities of a large language model with the professional computing tools of the Dify platform to create a closed loop of "integrated knowledge base - real-time operating conditions - intelligent question and answer - visual output." This not only enables efficient detection of refrigeration room operating data anomalies and energy-saving optimization, but also provides interactive graphical decision support. This solution significantly improves operational efficiency and decision-making accuracy, reduces training and troubleshooting costs, and is expected to be expanded to more high-precision manufacturing fields in the future, providing solid technical support for the transformation of industrial intelligence.

[0078] Figure 3 A flowchart of a method for intelligent question-answering and energy efficiency optimization in a refrigeration room based on the RAG architecture and the Dify platform is shown in an embodiment of the present invention. The method for intelligent question-answering and energy efficiency optimization in a refrigeration room based on the RAG architecture and the Dify platform includes:

[0079] Step S31: pre-processing the original time series data collected from the refrigeration room, and extracting several refrigeration room operation characteristics that are most relevant to energy-saving optimization to build a dynamic database.

[0080] Step S32: mine frequent operating patterns on the selected feature set and form high-energy consumption trigger conditions; input the high-energy consumption trigger conditions into the pre-trained large model for classification and identify abnormal energy consumption fluctuation scenarios; map the fault modes and fault factors corresponding to the abnormal energy consumption fluctuation scenarios to a predefined suggestion library to automatically generate an executable energy-saving optimization plan.

[0081] Step S33: construct a static knowledge base based on the RAG architecture, and fuse the static knowledge base, dynamic knowledge base and predefined suggestion base to form a RAG search result that integrates static documents, dynamic working conditions and energy-saving optimization solutions in response to user queries.

[0082] Step S34: Encapsulate the intelligent agent and the preset calculation module related to the energy efficiency of the refrigeration room in the Dify development environment and publish them as a schedulable tool; the intelligent agent recognizes the user's intention through the prompt word and RAG search results, and schedules the corresponding tool to perform energy efficiency-related calculation tasks and generate energy efficiency calculation results.

[0083] It should be noted that the implementation process and principles of the embodiments of the present application are similar to the intelligent question-and-answer and energy efficiency optimization system for refrigeration rooms based on the RAG architecture and Dify platform mentioned above, and will not be repeated here.

[0084] In the embodiments of the present application, terms such as "first" and "second" are used to distinguish between identical or similar items with substantially the same functions and effects. For example, the first XX and the second XX are merely used to distinguish between different XXs and do not limit their order. Those skilled in the art will understand that terms such as "first" and "second" do not limit the quantity or order of execution, and that terms such as "first" and "second" do not necessarily mean that they are different.

[0085] It should be noted that in the embodiments of this application, words such as "exemplary" or "for example" represent examples, illustrations, or descriptions. Any embodiment or design described in this application as "exemplary" or "for example" should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0086] In the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can represent: a, b, c, ab, ac, bc or abc, where a, b, c can be single or multiple.

[0087] Figure 4 This is a schematic block diagram of a computer device / equipment / system provided in an embodiment of the present application. Figure 4 As shown, the computer device includes: at least one processor 401, memory 402, at least one network interface 403 and a user interface 405. The various components in the device are coupled together via a bus system 404. It is understood that the bus system 404 is used to achieve connection and communication between these components. In addition to including a data bus, the bus system 404 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, the bus system 404 is not described in detail. Figure 4 In the text, various buses are labeled as bus systems.

[0088] The user interface 405 may include a display, a keyboard, a mouse, a trackball, a click gun, keys, buttons, a touch pad or a touch screen.

[0089] It will be appreciated that the memory 402 may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM) or a programmable read-only memory (PROM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memory described in the embodiments of the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0090] The memory 402 in the embodiment of the present invention is used to store various categories of data to support the operation of the electronic terminal 400. Examples of such data include: any executable program for operating on the electronic terminal 400, such as an operating system 4021 and an application 4022; the operating system 4021 includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application 4022 can include various applications, such as a media player (Media Player), a browser (Browser), etc., for implementing various application services. The implementation of the XX method provided in the embodiment of the present invention can be included in the application 4022.

[0091] The methods disclosed in the above embodiments of the present invention can be applied to or implemented by processor 401. Processor 401 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in processor 401 or by software instructions. The above processor 401 may be a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 401 can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor 401 may be a microprocessor or any conventional processor. The steps of the accessory optimization method provided in conjunction with the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium located in a memory. The processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0092] In an exemplary embodiment, the electronic terminal 400 may be configured to execute the aforementioned method using one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), or complex programmable logic devices (CPLDs).

[0093] According to the method provided in the embodiment of the present application, the present application also provides a computer-readable storage medium, which stores a program code. When the program code is run on a computer, the computer executes the above method.

[0094] As used in this specification, the terms "component," "module," "system," and the like are used to represent computer-related entities, hardware, firmware, a combination of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, and / or a computer. By way of illustration, both an application running on a computing device and a computing device can be a component. One or more components can reside in a process and / or an execution thread, and a component can be located on a computer and / or distributed between two or more computers. In addition, these components can be executed from various computer-readable media having various data structures stored thereon. Components can communicate, for example, via local and / or remote processes based on signals having one or more data packets (e.g., data from two components interacting with another component on a local system, a distributed system, and / or a network, such as the Internet interacting with other systems via signals).

[0095] Those skilled in the art will appreciate that the various illustrative logical blocks and steps described in conjunction with the embodiments disclosed herein can be implemented using electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0096] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0097] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0098] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0099] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0100] In the above embodiments, the functions of each functional unit can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions (programs). When the computer program instructions (program) are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media integrated therein. Available media may be magnetic media (e.g., floppy disks, hard disks, tapes), optical media (e.g., high-density digital video discs (DVDs), or semiconductor media (e.g., solid state disks (SSDs)).

[0101] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk, and other media that can store program codes.

[0102] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

[0103] In summary, this application provides a refrigeration room intelligent question-answering and energy efficiency optimization system based on the RAG architecture and the Dify platform. Through steps such as data preprocessing, feature selection, knowledge discovery, and optimization suggestion generation, it achieves efficient detection of abnormal values ​​in refrigeration room operating data and energy-saving optimization. At the same time, it encapsulates the domain knowledge base, real-time data interface, and professional computing tool chain into the Dify tool, constructs a question-answering agent with expert-level decision-making capabilities, and combines it with the ECharts visualization engine to provide interactive intelligent question-answering services with graphics and text for equipment status monitoring, operational health diagnosis, and energy efficiency optimization. Therefore, this application effectively overcomes the various shortcomings of the existing technology and has high industrial application value.

[0104] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Anyone skilled in the art may modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by one of ordinary skill in the art without departing from the spirit and technical concepts disclosed in this application shall be covered by the claims of this application.

Claims

1. An intelligent question-answering and energy efficiency optimization system for refrigeration rooms based on the RAG architecture and the Dify platform, characterized by: include: The data preprocessing and feature selection module is used to preprocess the raw time series data collected from the refrigeration room and extract several refrigeration room operation features that are most relevant to energy-saving optimization; An optimization suggestion generation module is used to mine frequent operating patterns based on a selected feature set and generate high-energy consumption trigger conditions. The high-energy consumption trigger conditions are input into a pre-trained large model for classification and identification of abnormal energy consumption fluctuation scenarios. The fault modes and fault factors corresponding to these abnormal energy consumption fluctuation scenarios are mapped to a pre-defined suggestion library to automatically generate executable energy-saving optimization solutions. A knowledge base construction module is used to construct a static knowledge base based on the RAG architecture and to construct a dynamic database based on the operating characteristics of the refrigeration room; fusing the static knowledge base, the dynamic knowledge base, and the predefined suggestion base to generate RAG search results that integrate static documents, dynamic working conditions, and energy-saving optimization solutions in response to user queries; The tool chain encapsulation and agent integration module is used to encapsulate the agent and preset calculation modules related to the energy efficiency of the refrigeration room in the Dify development environment and publish them as schedulable tools; the agent recognizes user intentions through prompt words and RAG search results, and schedules the corresponding tools to perform energy efficiency-related calculation tasks and generate energy efficiency calculation results.

2. The refrigeration room energy-saving optimization system according to claim 1, characterized in that: Also includes: The visualization module is integrated with a visualization engine and is used to generate and output one or more types of charts based on the energy efficiency calculation results and energy-saving optimization solutions.

3. The refrigeration room energy-saving optimization system according to claim 1, characterized in that: The dynamic database is constructed in the following manner: Use clustering algorithms to fill and / or remove data from the collected original time series data; Adaptive equal-frequency discretization algorithm is used to convert continuous variables into discrete intervals; Variables with high correlation with refrigeration room operation are pre-screened, and a large model is driven by structured prompt words to extract several refrigeration room operation features that are most relevant to energy-saving optimization from the conversation context information and user concerns.

4. The refrigeration room energy-saving optimization system according to claim 1, characterized in that: The optimization suggestion generation module mines frequent operation patterns on the selected feature set and forms high energy consumption trigger conditions in the following manner: mining frequent operation patterns from the selected feature set through the FP-Max algorithm, filtering out short-term transient patterns by setting support and confidence thresholds, and forming typical high energy consumption trigger conditions.

5. The refrigeration room energy-saving optimization system according to claim 1, characterized in that: The methods for building a static knowledge base based on the RAG architecture include: vectorizing multimodal documents related to refrigeration rooms and storing them in the static knowledge base, and regularly updating them in sync with industry public information.

6. The refrigeration room energy-saving optimization system according to claim 1, characterized in that: The computing modules integrated in the tool chain encapsulation and intelligent agent integration module include any one or more combinations of the following: data cleaning module, feature engineering module, EER energy efficiency calculation module, and thermal balance rate analysis module.

7. The refrigeration room energy-saving optimization system according to claim 1, characterized in that: The predefined suggestion library contains known failure modes, failure factors and their corresponding energy-saving optimization solutions, and each energy-saving optimization solution is accompanied by a corresponding threshold and verification algorithm.

8. A method for intelligent question-answering and energy efficiency optimization of a refrigeration room based on the RAG architecture and the Dify platform, characterized in that: include: Preprocess the raw time series data collected from the refrigeration room and extract several refrigeration room operation characteristics most relevant to energy-saving optimization to build a dynamic database; Mining frequent operation patterns on the selected feature set and forming high energy consumption trigger conditions; High energy consumption trigger conditions are input into a pre-trained large model for classification and identification of abnormal energy consumption fluctuation scenarios. The fault modes and failure factors corresponding to abnormal energy consumption fluctuation scenarios are mapped to a pre-defined suggestion library to automatically generate executable energy-saving optimization solutions. Constructing a static knowledge base based on the RAG architecture, and fusing the static knowledge base, dynamic knowledge base, and predefined suggestion base to generate RAG search results that integrate static documents, dynamic working conditions, and energy-saving optimization solutions in response to user queries; An intelligent agent and preset calculation modules related to the energy efficiency of refrigeration rooms are encapsulated in the Dify development environment and published as a schedulable tool. The intelligent agent identifies user intentions through prompt words and RAG search results, and schedules corresponding tools to perform energy efficiency-related calculation tasks and generate energy efficiency calculation results.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the intelligent question-answering and energy efficiency optimization method for the refrigeration room based on the RAG architecture and the Dify platform as described in claim 8 is implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory, wherein: The processor executes the computer program to implement the intelligent question-answering and energy efficiency optimization method for a refrigeration room based on the RAG architecture and the Dify platform as described in claim 8.

Images