Electric power knowledge enhancement generative model optimization method and system based on knowledge reasoning

Abstract

The invention belongs to the field of machine learning, and particularly relates to an electric power knowledge enhancement generative model optimization method and system based on knowledge reasoning, which can more comprehensively capture trends and changes in an electric power system by combining historical electric power data and current information data, and avoid depending on a single data source. Historical data can provide a long-term trend of system operation, and current information data can reflect real-time conditions. The combination of the double information can effectively enhance the inference ability of the model for implicit association and improve the model precision. According to the electric power knowledge base, a large amount of actual experience and historical data in an electric power system are accumulated, so that potential modes and laws which are difficult to see through direct data analysis can be mined. According to the method, knowledge features in the power knowledge base are mined, implicit associations between power systems can be revealed, and the generative model can learn the associations, so that the accuracy of the model facing an unknown or changing environment is improved.

Description

Technical Field

[0001] The present invention belongs to the field of machine learning, and particularly relates to an optimization method and system for a power knowledge enhanced generative model based on knowledge reasoning. Background Art

[0002] The current power system faces increasingly complex management and maintenance challenges, especially in aspects such as operation prediction, load scheduling, and fault warning. With the expansion of the power grid scale and the growth of power demand, a large amount of historical operation data has been accumulated in the power system, and these data carry the operation rules and change trends of the power system. However, due to the complexity and multidimensionality of power data, the implicit correlation relationships and logical rules in these data are difficult to extract and utilize through traditional statistical or conventional data analysis means, resulting in the failure to fully exert their potential value. At the same time, the power system is affected by multiple external factors during the actual operation process, and there are obvious uncertainties in aspects such as load fluctuations, fault occurrences, and dynamic environment adaptation. The existing technologies of the power system often have difficulty in effectively integrating massive historical data and cannot deeply mine and utilize the implicit knowledge in the power field when facing such complex operation scenarios. Therefore, in the process of the intelligent development of the power system, how to extract and enhance knowledge from the huge historical data, how to more accurately understand the implicit correlations between data, and use them to guide actual operation and maintenance has become an urgent problem to be solved. Summary of the Invention

[0003] The purpose of the present invention is to overcome the problem of insufficient accuracy of the implicit correlations between the above data, and provide an optimization method and system for a power knowledge enhanced generative model based on knowledge reasoning.

[0004] To achieve the above purpose, the present invention adopts the following technical solutions: In the first aspect, the present invention provides an optimization method for a power knowledge enhanced generative model based on knowledge reasoning, including the following steps: Obtain the current information data of the power environment; Extract the knowledge features in the power knowledge base, and perform knowledge mining on the power knowledge base according to the knowledge features and the required knowledge needs to obtain a knowledge mining data set; Perform data reasoning on the knowledge mining data set to obtain a knowledge reasoning result; Analyze the generative model according to the knowledge reasoning result to obtain the learning rate of the generative model, and enhance the generative model according to the learning rate of the generative model to obtain a knowledge parameter group; Perform collaborative analysis on the knowledge parameter group to obtain a first decision-making suggestion, and perform execution feedback on the first decision-making suggestion in the enhanced generative model to obtain feedback response data; Dynamically optimize the first decision-making suggestion based on the feedback response data combined with the current information data, generate the second decision-making suggestion, and re-perform the execution feedback in the enhanced generative model.

[0005] A further improvement of the present invention lies in that the power knowledge base is constructed based on historical power data, and the specific method for constructing the power knowledge base is as follows: Obtain the expert experience data in this field, perform structured analysis on the expert experience data in this field, and obtain a number of knowledge entries; Classify the historical power data based on a number of knowledge entries to obtain a number of data classes; Perform logical analysis on the expert experience data in this field to obtain the logical rules between all knowledge entries, and formulate knowledge integration rules according to the logical rules between all knowledge entries; Integrate the knowledge of all data classes according to the knowledge integration rules to obtain a number of knowledge levels; Perform data consistency verification on all knowledge levels, and form a power knowledge base with the knowledge levels that pass the data consistency verification.

[0006] A further improvement of the present invention lies in extracting the knowledge features in the power knowledge base, and performing knowledge mining on the power knowledge base according to the knowledge features and the required knowledge needs. The specific method for obtaining the knowledge mining data set is as follows: Extract the knowledge entries corresponding to the knowledge levels in the power knowledge base, perform feature analysis on the power knowledge base based on the knowledge entries, and obtain a number of knowledge features; Traverse and screen the power database based on all knowledge features to obtain a number of screened data, and perform explicit and implicit analysis on the knowledge entries corresponding to the screened data to obtain an implicit knowledge set and an explicit knowledge set; Perform association analysis on the implicit knowledge set to obtain a number of association support degrees, and perform knowledge mining on the implicit knowledge set according to the number of association support degrees to determine the frequent association patterns; Perform reasoning analysis on the explicit knowledge set to obtain a number of processing confidence degrees, and perform knowledge mining on the explicit knowledge set according to the number of processing confidence degrees to determine the effective processing patterns; Integrate the frequent association patterns and the effective processing patterns to obtain a knowledge mining data set.

[0007] A further improvement of the present invention lies in performing data reasoning on the knowledge mining data set to obtain the knowledge reasoning result. The specific method is as follows: Perform correlation coefficient analysis on all knowledge features to obtain a number of feature scores; Arrange all feature scores in descending order to obtain a feature score sequence, and screen all knowledge features according to the feature score sequence to determine a number of key features; Based on the frequent association patterns in the knowledge mining dataset and combining all key features, the implicit knowledge set is labeled to obtain the first label class; Based on the effective processing patterns in the knowledge mining dataset and combining all key features, the explicit knowledge set is labeled to obtain the second label class; Based on the first label class and the second label class, a data conversion table is constructed based on the implicit knowledge set and the explicit knowledge set; According to the preset inference conditions according to the requirements, the data conversion table is traversed according to the inference conditions for logical reasoning to obtain the knowledge inference result.

[0008] A further improvement of the present invention is that the specific method for constructing a data conversion table based on the implicit knowledge set and the explicit knowledge set according to the first label class and the second label class is as follows: Based on the first label class, a label index of the implicit knowledge set is constructed as the first label index; Based on the second label class, a label index of the explicit knowledge set is constructed as the second label index; Traverse the implicit knowledge set according to the first label index for matching to obtain the first matching result, and generate the first conversion relationship according to the first matching result; Traverse the explicit knowledge set according to the second label index for matching to obtain the second matching result, and generate the second conversion relationship according to the second matching result; Fill the implicit knowledge set with data according to the first conversion relationship in combination with the first label index to obtain the first data filling result; Fill the explicit knowledge set with data according to the second conversion relationship in combination with the second label index to obtain the second data filling result; Combine the first data filling result and the second data filling result to obtain the data conversion table.

[0009] A further improvement of the present invention is that the specific method for traversing the data conversion table according to the preset inference conditions according to the requirements for logical reasoning to obtain the knowledge inference result is as follows: Generate a number of power demand information according to the requirements, generate a number of inference scenario information according to all the power demand information, and perform trigger analysis on all the inference scenario information to obtain the inference conditions; Traverse the data conversion table based on the inference conditions to determine the implicit knowledge entry information and the explicit knowledge entry information; Perform logical reasoning on the implicit knowledge set based on the implicit knowledge entry information to obtain the first inference result; Perform logical reasoning on the explicit knowledge set based on the explicit knowledge entry information to obtain the second inference result; Deduplicate and merge the first inference result and the second inference result to obtain the knowledge inference result.

[0010] A further improvement of the present invention lies in analyzing the generative model according to the knowledge reasoning result to obtain the learning rate of the generative model, and enhancing the generative model according to the learning rate of the generative model. The specific method for obtaining the knowledge parameter set is as follows: Perform a structured classification on the knowledge reasoning result to obtain several knowledge structure categories; Combine all knowledge structure categories with all knowledge features for normalization processing to obtain a normalization processing result, and set a feature window according to the normalization processing result; Analyze the generative model based on the feature window, perform performance evaluation according to the analysis result, and obtain the performance score of the generative model; Formulate an incremental learning strategy for the generative model according to the performance score of the generative model; Conduct an optimization analysis on the incremental learning strategy formulated for the generative model to obtain the learning rate of the generative model, enhance the generative model according to the learning rate of the generative model, output the enhanced model parameter set of the generative model, and combine the enhanced model parameter set with the learning rate of the generative model to obtain the knowledge parameter set.

[0011] A further improvement of the present invention lies in conducting a collaborative analysis on the knowledge parameter set to obtain the first decision-making suggestion. The specific method is as follows: Perform parameter traversal on the knowledge parameter set using the learning rate of the generative model, conduct a combined deviation evaluation of the knowledge parameter set according to the traversal result, and obtain a parameter combined deviation evaluation value; Use the parameter combined deviation evaluation value and the corresponding knowledge parameter set as incremental data for deviation correction to obtain a parameter correction result; Conduct a collaborative analysis on the knowledge parameter set using the parameter correction result, set a collaborative priority, and perform weighted scoring on the knowledge parameter set according to the collaborative priority to obtain the first decision-making suggestion.

[0012] In a second aspect, the present invention provides an optimized system for a power knowledge enhanced generative model based on knowledge reasoning, including: A data acquisition unit for obtaining current information data of the power environment; A data conversion unit for extracting knowledge features from the power knowledge base, performing knowledge mining on the power knowledge base according to the knowledge features and the required knowledge needs, and obtaining a knowledge mining data set; A data reasoning unit for performing data reasoning on the knowledge mining data set to obtain a knowledge reasoning result; An optimization and enhancement unit for analyzing the generative model according to the knowledge reasoning result to obtain the learning rate of the generative model, and enhancing the generative model according to the learning rate of the generative model to obtain a knowledge parameter set; An execution feedback unit is used to perform collaborative analysis on the knowledge parameter group to obtain a first decision recommendation, and perform execution feedback on the first decision recommendation in the enhanced generative model to obtain feedback response data; A dynamic optimization unit is used to dynamically optimize the first decision recommendation according to the feedback response data combined with the current information data to generate a second decision recommendation, and re-perform execution feedback in the enhanced generative model.

[0013] A further improvement of the present invention is that the power knowledge base in the data conversion unit is constructed based on historical power data, and the specific process of constructing the power knowledge base is as follows: Obtain the expert experience data in the field, perform structured analysis on the expert experience data in the field to obtain a number of knowledge entries; Classify the historical power data based on a number of knowledge entries to obtain a number of data classes; Perform logical analysis on the expert experience data in the field to obtain the logical rules between all knowledge entries, and formulate knowledge integration rules according to the logical rules between all knowledge entries; Perform knowledge integration on all data classes according to the knowledge integration rules to obtain a number of knowledge levels; Perform data consistency verification on all knowledge levels, and form a power knowledge base with the knowledge levels that pass the data consistency verification.

[0014] A further improvement of the present invention is that the function of the data conversion unit is realized by the following method: Extract the knowledge entries corresponding to the knowledge levels in the power knowledge base, perform feature analysis on the power knowledge base based on the knowledge entries to obtain a number of knowledge features; Traverse and screen the power database based on all knowledge features to obtain a number of screened data, perform explicit and implicit analysis on the knowledge entries corresponding to the screened data to obtain an implicit knowledge set and an explicit knowledge set; Perform correlation analysis on the implicit knowledge set to obtain a number of correlation support degrees, and perform knowledge mining on the implicit knowledge set according to the number of correlation support degrees to determine frequent association patterns; Perform reasoning analysis on the explicit knowledge set to obtain a number of processing confidence degrees, and perform knowledge mining on the explicit knowledge set according to the number of processing confidence degrees to determine effective processing patterns; Integrate the frequent association patterns and the effective processing patterns to obtain a knowledge mining data set.

[0015] A further improvement of the present invention is that the function of the data reasoning unit is realized by the following method: Perform correlation coefficient analysis on all knowledge features to obtain a number of feature scores; Sort all feature scores in descending order to obtain a feature score sequence, and screen all knowledge features according to the feature score sequence to determine a number of key features; According to the frequent association patterns in the knowledge mining dataset, combined with all key features, perform tagging on the tacit knowledge set to obtain the first tag class; According to the effective processing patterns in the knowledge mining dataset, combined with all key features, perform tagging on the explicit knowledge set to obtain the second tag class; Based on the first tag class and the second tag class, construct a data conversion table based on the tacit knowledge set and the explicit knowledge set; Preset inference conditions according to requirements, and traverse the data conversion table according to the inference conditions for logical reasoning to obtain knowledge inference results.

[0016] A further improvement of the present invention is that in the data inference unit, the function of constructing a data conversion table based on the tacit knowledge set and the explicit knowledge set according to the first tag class and the second tag class is realized by the following method; Construct a tag index for the tacit knowledge set based on the first tag class as the first tag index; Construct a tag index for the explicit knowledge set based on the second tag class as the second tag index; Traverse the tacit knowledge set according to the first tag index for matching to obtain the first matching result, and generate the first conversion relationship according to the first matching result; Traverse the explicit knowledge set according to the second tag index for matching to obtain the second matching result, and generate the second conversion relationship according to the second matching result; Fill in the data of the tacit knowledge set according to the first conversion relationship combined with the first tag index to obtain the first data filling result; Fill in the data of the explicit knowledge set according to the second conversion relationship combined with the second tag index to obtain the second data filling result; Combine the first data filling result and the second data filling result to obtain a data conversion table.

[0017] A further improvement of the present invention is that in the data inference unit, the function of presetting inference conditions according to requirements, traversing the data conversion table according to the inference conditions for logical reasoning, and obtaining knowledge inference results is realized by the following method; Generate a number of power demand information according to requirements, generate a number of inference scenario information according to all power demand information, perform trigger analysis on all inference scenario information to obtain inference conditions; Traverse the data conversion table based on the inference conditions to determine the tacit knowledge entry information and the explicit knowledge entry information; Perform logical reasoning on the tacit knowledge set based on the tacit knowledge entry information to obtain the first inference result; Perform logical reasoning on the explicit knowledge set based on the explicit knowledge entry information to obtain a second reasoning result; Deduplicate and merge the first reasoning result and the second reasoning result to obtain a knowledge reasoning result.

[0018] A further improvement of the present invention lies in that the function of the optimization enhancement unit is realized by the following method: Perform structured classification on the knowledge reasoning result to obtain several knowledge structure categories; Combine all knowledge structure categories with all knowledge features for normalization processing to obtain a normalization processing result, and set a feature window according to the normalization processing result; Analyze the generative model based on the feature window, perform performance evaluation according to the analysis result, and obtain the performance score of the generative model; Formulate an incremental learning strategy for the generative model according to the performance score of the generative model; Perform optimization analysis on the incremental learning strategy formulated for the generative model to obtain the learning rate of the generative model, enhance the generative model according to the learning rate of the generative model, output the enhanced model parameter set of the generative model, and combine the enhanced model parameter set with the learning rate of the generative model to obtain a knowledge parameter group.

[0019] A further improvement of the present invention lies in that in the execution feedback unit, the function of performing collaborative analysis on the knowledge parameter group to obtain a first decision suggestion is realized by the following method: Perform parameter traversal on the knowledge parameter group using the learning rate of the generative model, perform a combined deviation evaluation of the knowledge parameter group according to the traversal result, and obtain a parameter combined deviation evaluation value; Use the parameter combined deviation evaluation value and the corresponding knowledge parameter group as incremental data for deviation correction to obtain a parameter correction result; Perform collaborative analysis on the knowledge parameter group using the parameter correction result, set a collaborative priority, and perform weighted scoring on the knowledge parameter group according to the collaborative priority to obtain a first decision suggestion.

[0020] In a third aspect, the present invention provides an electronic device, including a memory and a processor, where the memory stores a computer program, and when the processor executes the computer program, it implements the steps of the method for optimizing a power knowledge enhanced generative model based on knowledge reasoning.

[0021] In a fourth aspect, the present invention provides a storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the method for optimizing a power knowledge enhanced generative model based on knowledge reasoning.

[0022] Compared with the prior art, the present invention has the following beneficial effects: By combining historical power data and current information data, the present invention can capture trends and changes in the power system more comprehensively and avoid relying solely on a single data source. Historical data can provide long-term trends in system operation, while current information data can reflect real-time situations. The combination of this dual information can effectively enhance the model's reasoning ability for implicit associations and improve model accuracy. The power knowledge base of the present invention can help uncover potential patterns and regularities that are not easily seen through direct data analysis by accumulating a large amount of practical experience and historical data in the power system. Mining the knowledge features in the power knowledge base can reveal the implicit associations between power systems and enable the generative model to learn these associations, thereby improving the model's accuracy when facing unknown or changing environments. The present invention can provide more accurate input parameters for the generative model through knowledge reasoning results, making the model's output more in line with the actual situation. By analyzing and adjusting the learning rate of the generative model, overfitting or underfitting can be avoided, and the model's adaptability can be improved. In addition, implementing a feedback mechanism (feeding decision suggestions back into the model for correction) can further optimize the model output, ensure dynamic adjustment during model operation, and gradually overcome the deficiencies of implicit associations. By providing feedback after model execution and dynamically optimizing in combination with current information data, the present invention can timely discover the deficiencies of the model in dealing with implicit associations, thereby adjusting decision suggestions. This feedback-based optimization method can continuously improve the model's accuracy and reliability during operation, making it more flexible in responding to changing power environments. In summary, by combining historical data with real-time data, strengthening the application of the power knowledge base, and utilizing reasoning and feedback mechanisms, the present invention can not only effectively mine and process implicit associations, but also ensure continuous optimization of the model, adapt to changes in the power environment, and thus improve the accuracy and quality of decision support. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flowchart of Embodiment 1; Figure 2 is a system diagram of Embodiment 2; Figure 3 is a system diagram of Embodiment 11. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0024] To further understand the content of the present invention, the following provides a detailed description of the present invention in combination with the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely for explaining the present invention rather than limiting it.

[0025] Embodiment 1: See Figure 1 , a method for optimizing a power knowledge enhanced generative model based on knowledge reasoning, includes the following steps: S1, obtain the current information data of the power environment.

[0026] S2. Extract the knowledge features from the power knowledge base, perform knowledge mining on the power knowledge base according to the knowledge features and the required knowledge needs, and obtain a knowledge mining data set.

[0027] S3. Perform data reasoning on the knowledge mining data set to obtain a knowledge reasoning result.

[0028] S4. Analyze the generative model based on the knowledge reasoning result to obtain the learning rate of the generative model, and enhance the generative model according to the learning rate of the generative model to obtain a knowledge parameter group.

[0029] S5. Conduct collaborative analysis on the knowledge parameter group to obtain a first decision-making suggestion, and perform execution feedback on the first decision-making suggestion in the enhanced generative model to obtain feedback response data.

[0030] S6. Dynamically optimize the first decision-making suggestion according to the feedback response data combined with the current information data to generate a second decision-making suggestion, and re-perform execution feedback in the enhanced generative model.

[0031] In this embodiment, by combining historical power data and current information data, the model can comprehensively capture the long-term trends and real-time changes in the power system, avoid relying solely on a single data source, enhance the reasoning ability for implicit associations, and improve the prediction accuracy. By constructing a power knowledge base, potential patterns and rules in the power system are extracted, providing rich knowledge support for the generative model. These mined implicit knowledge features can help the model identify and accurately reflect the complex associations in the power system. The analysis of the knowledge reasoning result and the adjustment of the learning rate of the generative model in this embodiment make the model output more in line with the actual situation. Through the execution feedback mechanism, the decision-making suggestions are continuously optimized to ensure that the model processes implicit associations more precisely. After the model is executed, through feedback response and dynamic optimization, the decision-making suggestions are adjusted in real time, gradually improving the adaptability and accuracy of the model in different power environments.

[0032] Embodiment 2: See Figure 2 , a power knowledge enhanced generative model optimization system based on knowledge reasoning, including: A data acquisition unit for obtaining the current information data of the power environment; A data conversion unit for extracting the knowledge features from the power knowledge base, performing knowledge mining on the power knowledge base according to the knowledge features and the required knowledge needs, and obtaining a knowledge mining data set; A data reasoning unit for performing data reasoning on the knowledge mining data set to obtain a knowledge reasoning result; An optimization and enhancement unit for analyzing the generative model according to the knowledge reasoning result to obtain the learning rate of the generative model, and enhancing the generative model according to the learning rate of the generative model to obtain a knowledge parameter group; An execution feedback unit for collaboratively analyzing a knowledge parameter group to obtain a first decision-making suggestion, and performing execution feedback on the first decision-making suggestion in the enhanced generative model to obtain feedback response data; A dynamic optimization unit for dynamically optimizing the first decision-making suggestion according to the feedback response data in combination with the current information data to generate a second decision-making suggestion, and re-performing execution feedback in the enhanced generative model.

[0033] Embodiment 3: This embodiment further defines the method for constructing a power knowledge base, specifically as follows: Step 1: Obtain the expert experience data in the field, and perform structured analysis on the expert experience data in the field to obtain a number of knowledge entries.

[0034] Step 2: Classify the historical power data based on a number of knowledge entries to obtain a number of data classes.

[0035] Step 3: Perform logical analysis on the expert experience data in the field to obtain the logical rules between all knowledge entries, and formulate knowledge integration rules according to the logical rules between all knowledge entries.

[0036] Step 4: Perform knowledge integration on all data classes according to the knowledge integration rules to obtain a number of knowledge levels.

[0037] Step 5: Perform data consistency verification on all knowledge levels, and form a power knowledge base with the knowledge levels that pass the data consistency verification.

[0038] During use, the system terminal calls the knowledge base construction unit through a pre-established communication connection to perform adaptive integration on the historical power data, collecting and integrating various historical data such as the operation data, equipment status, and fault records of the previous power system. Through the knowledge base construction unit, the historical power data will be automatically screened, cleaned, and classified to ensure the accuracy and consistency of the historical power data. At the same time, it will also integrate the experience and rules of experts in the field, and structurally embed this implicit knowledge into the power knowledge base, so that the power knowledge base not only contains direct numerical data, but also covers the professional knowledge and logical relationships related to the data. During the adaptive integration process, the knowledge base construction unit will perform intelligent analysis and automatic classification on the data according to the characteristics of different types of data. For example, the data will be hierarchically stored according to the fault category, equipment type, or load change situation, and the potential associations between the data will be extracted through causal analysis or statistical laws, thereby constructing a power knowledge base. The power knowledge base integrates various historical information in an efficient and available manner, becoming a dynamic and highly relevant power knowledge resource, providing rich and multi-dimensional support for subsequent data mining and model optimization.

[0039] This embodiment can retrieve the expert experience data in the field for data integration to obtain the expert experience records in the field, perform structured storage according to the expert experience records in the field to obtain multiple knowledge items; classify the historical power data according to the multiple knowledge items to determine multiple data categories; perform causal analysis based on the expert experience records in the field to obtain knowledge logic rules, and set knowledge integration rules according to the knowledge logic rules; according to the knowledge integration rules, perform knowledge integration on the historical power data according to the multiple data categories to determine multiple knowledge levels; perform data consistency verification on the multiple knowledge levels, and construct the power knowledge base according to the verification results.

[0040] In a preferred embodiment, during the construction of the knowledge base, the system terminal first retrieves the knowledge data of experts in the field, integrates various expert experiences and analysis records to form a structured record of expert experiences in the field. Then, the record of expert experiences in the field is processed by category and organized into multiple knowledge entries, and each knowledge entry corresponds to a certain specific aspect of experience or rule provided by the expert. These entries include not only quantitative data but also qualitative descriptions, such as the cause of failure, the law of equipment state change, etc., in order to form a comprehensive knowledge framework. Subsequently, the historical power data is classified according to these knowledge entries, and the data is divided into multiple data categories. For example, equipment knowledge category, load knowledge category, fault prevention category, etc. Such classification facilitates the system terminal to more efficiently locate and call specific categories of data, and at the same time provides a convenient reference for subsequent causal analysis and knowledge integration. After that, the key content is extracted from the record of expert experiences in the field, mainly the records related to the state of power equipment, load changes, fault occurrence, etc. The system terminal will scan these records through natural language processing algorithms and identify the parts containing cause and effect expressions. For example, under high load operation, the equipment temperature rises rapidly, which is likely to cause a failure, or when the environmental humidity increases, the insulation of the equipment decreases and the probability of failure increases. Then, the extracted causal information is labeled, and the cause and effect are marked separately. For example, high load operation is marked as the cause, and equipment temperature rise and failure are marked as the effects. After obtaining this causal information, the labeled causal information is sorted in the form of a chain to form a complete causal chain. For example, the chain of high load → temperature rise → failure. Then, frequency analysis is performed on the extracted causal chain, and the frequency of each causal chain appearing in the expert record is calculated, and then the reliability of the causal relationship is evaluated. For causal relationships with higher frequencies, higher weights will be assigned to give priority consideration when generating knowledge logic rules. Based on the causal chain, the system terminal generates knowledge logic rules. For example, for the causal chain of high load operation causing temperature rise and then leading to failure, corresponding logic rules are generated. For example, when the equipment load exceeds a specific threshold, the temperature rises to a specific range and the failure risk increases. Similar logic rules can be reused multiple times and conditionally extended to adapt to different application scenarios. According to the knowledge logic rules, the system terminal sets knowledge integration rules and structurally integrates the historical power data according to the rules. For example, if the knowledge logic rule indicates that high load operation affects the equipment state, the system terminal will integrate the data related to high load operation into the load knowledge layer, and the equipment state data into the equipment knowledge layer, and link the two knowledge layers through logical association. This process ensures the hierarchical classification of data and the manifestation of logical relationships, thus forming a knowledge system with a hierarchical structure, that is, multiple knowledge levels. Finally, in order to ensure the consistency and relevance of data between these knowledge levels, the system terminal verifies the consistency of the content of each level.In this process, the system terminal identifies the data content of each level from the constructed knowledge hierarchy. For example, the device knowledge level contains data such as device operation status and maintenance records; the load knowledge level includes information such as load history and fluctuation data; the fault prevention level contains data such as fault records and fault modes. The data content of each level will be sorted according to predefined standards to ensure the clarity of the data in terms of content and structure. After identifying the data content of each level, consistency verification rules are set according to the requirements of the power knowledge base and data standards. The verification rules include data format consistency, data timestamp alignment, data integrity, and logical relationship rationality. For example, the timestamp in the load data needs to be aligned with the timestamp of the device status data; the chronological order of the fault records should conform to the logic of actual fault occurrence. Then, the data format in each knowledge level is checked to ensure that the data is recorded in a unified format. For example, numerical data such as current, voltage, and load should uniformly use the same unit (such as amperes, volts), and the date and time information format should be consistent (such as YYYY-MM-DD hh:mm) for subsequent analysis and processing. After the consistency verification is completed, the system terminal will record all verification results, including the list of qualified data and abnormal data, and mark the data entries that do not meet the consistency standards. For the data that passes the verification, it is marked as verified and officially entered into the power knowledge base, thus completing the construction of the power knowledge base and making it a knowledge resource covering multiple aspects such as device operation, load prediction, and fault prevention. For the data that fails to pass the verification, it is marked as verification failed and will be repaired, supplemented, or excluded as appropriate.

[0041] Embodiment 4: This embodiment further defines the specific steps of S2 in Embodiment 1 and the functional implementation method of the data conversion unit in Embodiment 2, as follows: S21. Extract the knowledge entries corresponding to the knowledge levels in the power knowledge base, and perform feature analysis on the power knowledge base based on the knowledge entries to obtain several knowledge features.

[0042] S22. Traverse and filter the power database based on all the knowledge features to obtain several filtered data, and perform explicit and implicit analysis on the knowledge entries corresponding to the filtered data to obtain an implicit knowledge set and an explicit knowledge set.

[0043] S23. Perform association analysis on the implicit knowledge set to obtain several association support degrees, and perform knowledge mining on the implicit knowledge set according to the several association support degrees to determine frequent association patterns.

[0044] S24. Perform reasoning analysis on the explicit knowledge set to obtain several processing confidence degrees, and perform knowledge mining on the explicit knowledge set according to the several processing confidence degrees to determine effective processing patterns.

[0045] S25. Integrate the frequently associated patterns with the effective processing patterns to obtain a knowledge mining data set.

[0046] During use, the system terminal synchronizes the power knowledge base to the data conversion unit for in-depth knowledge mining. After receiving the power knowledge base, the data conversion unit analyzes the data at each level stored in the power knowledge base to determine the implicit knowledge set and the explicit knowledge set, and then conducts knowledge mining on these two knowledge sets respectively to form a knowledge mining data set. According to the characteristics of the knowledge mining data set, the data entries are converted into structured information suitable for reasoning analysis.

[0047] In this embodiment, the data entries are converted into structured information suitable for reasoning analysis. This conversion includes steps such as correlation coefficient analysis, tagging processing, and logical relationship reasoning to ensure that the mined knowledge can match the model, thereby improving the accuracy and efficiency of the reasoning process. Through data conversion, the knowledge mining data set is further refined to form a structured knowledge reasoning result for use by the optimization and enhancement unit, ensuring that the model can more accurately identify key influencing factors and make a more intelligent response in practical applications.

[0048] In this embodiment, the power knowledge base is synchronized to the data reasoning unit to extract the multiple knowledge entries, and the power knowledge base is analyzed for its characteristics according to the multiple knowledge entries to obtain multiple knowledge features; based on the multiple knowledge features, the power knowledge base is traversed and screened, and the implicit knowledge set and the explicit knowledge set are determined by analyzing the multiple screening data in combination with the multiple knowledge entries.

[0049] In this embodiment, the system terminal synchronizes the data in the power knowledge base to the data inference unit for subsequent analysis and processing. The data in the power knowledge base includes information such as equipment operation status, historical load data, and fault records. The system terminal will orderly import this data into the data conversion unit. The data conversion unit identifies and extracts the core knowledge entries related to the power system from the power knowledge base. For example, the temperature change of the equipment during the peak load period, the relationship between the fault frequency and the humidity change, etc. The system terminal automatically filters out all knowledge entries with representativeness and research value according to specific rules or screening conditions (thresholds of key parameters, specific time ranges, such as several hours before and after a fault). Once the knowledge entries are extracted, the system terminal conducts feature analysis on each knowledge entry to identify the data features with significant significance in the knowledge entry, such as numerical values, time stamps, load levels, etc. For example, the features of the load entry include the load fluctuation amplitude, the load change speed, etc., and the features of the fault entry include the fault type, the ambient temperature, etc. Through feature analysis, multiple knowledge features are gradually constructed to better understand and distinguish the important information in the knowledge entries. After obtaining multiple knowledge features, the system terminal traverses and filters the power knowledge base, and uses these knowledge features to filter and group the data in the power knowledge base. During the screening process, it will search for data that matches the knowledge features and determine whether it meets the preset screening conditions. For example, if a certain feature is that the load peak reaches more than 80%, the system terminal will traverse all the load data in the power knowledge base and filter out the data items that meet this condition. After completing the traversal and screening, all the data items that meet the knowledge feature conditions are summarized to generate a screened data set. This data set covers the data entries that meet the knowledge features and provides an accurate data basis for subsequent association analysis. Based on the screened data set, the system terminal analyzes these data entries to find potential rules. For example, analyze the co-occurrence relationship between high-load data and fault data to determine whether there is a trend that the load peak causes equipment failures; or compare the temperature fluctuation data with the fault frequency to find the impact of temperature changes on the occurrence of faults. During the analysis process, deep-seated associations that are difficult to directly observe will be extracted to construct an implicit knowledge set. These implicit knowledge usually contains multiple feature associations. For example, when the load exceeds 80% and the temperature is higher than the normal value, the equipment failure probability increases significantly. These rules often need to be discovered through multi-dimensional data comparison, so they are classified into the implicit knowledge set. For those knowledge entries with intuitive features and clear rules, they are classified into the explicit knowledge set. For example, when the equipment temperature rises under a specific load or the humidity exceeds a certain limit, the failure rate increases, etc. These explicit knowledge is relatively easy to directly understand and can be directly classified into the explicit knowledge set for convenient subsequent quick call.

[0050] Perform association analysis based on the implicit knowledge set to generate multiple association supports, perform knowledge mining on the implicit knowledge set according to the multiple association supports to determine frequent association patterns; perform reasoning analysis based on the explicit knowledge set to generate multiple processing confidences, perform knowledge mining on the explicit knowledge set according to the multiple processing confidences to determine effective processing patterns; integrate the frequent association patterns and the effective processing patterns to determine the knowledge mining data set.

[0051] The system terminal extracts various data relationships and features from the implicit knowledge set, such as the relationship between peak load and equipment temperature rise, or the potential correlation between equipment aging and failure frequency. These data are typically implicit associations discovered through multi-dimensional data analysis. Based on the implicit knowledge set, the system terminal performs multi-level association analysis on the data to explore potential associations between data items. Specifically, it uses association rule mining algorithms (such as the Apriori algorithm or the FP-Growth algorithm) to gradually identify frequently co-occurring combinations between data items. For each combination analyzed, an association support is generated, which is the frequency of occurrence of the combination in the implicit knowledge set. This is calculated by comparing the number of times the combination appears to the total number of occurrences. The higher the association support, the stronger the association. The system terminal then screens association patterns within the implicit knowledge set based on the calculated association support. The system terminal sets a support threshold (such as 0.5) and selects association combinations with support above the threshold, identifying these combinations as frequent association patterns. Frequent association patterns reflect key relationships within the implicit knowledge set. For example, when the load exceeds 80% and the temperature exceeds 40°C, the probability of equipment failure increases. Furthermore, the system terminal processes the explicit knowledge set, which contains relatively intuitive and directly observable knowledge items. For example, temperatures exceeding 50°C trigger equipment alarms, and humidity exceeding 90% increases the failure rate. These items represent explicit, easily observable characteristic relationships. The system terminal uses inference analysis within the explicit knowledge set to calculate confidence levels to determine the reliability of each knowledge item in addressing practical problems. Confidence indicates the reliability of solving or explaining common problems when a certain condition is met. It is calculated by comparing the number of times the condition and outcome occur together with the number of times the condition occurs. The system terminal calculates confidence for each condition-outcome pair in the explicit knowledge set (e.g., "Humidity exceeds 90% → Failure occurs"). Condition-outcome pairs with higher confidence levels indicate that the inference rule is more effective in solving such problems, thereby identifying valuable knowledge patterns. Based on the multiple confidence levels calculated, knowledge mining is performed on the explicit knowledge set. The system terminal sets a confidence threshold (e.g., 0.6) and selects inference rules with confidence levels above the threshold, identifying these rules as valid processing patterns. For example, a processing pattern that shows a significant increase in equipment failure rate when humidity exceeds 90% indicates a high processing confidence level for this condition and can be used for real-time fault warning. Finally, the system terminal integrates the identified frequent association patterns and valid processing patterns to form a knowledge mining dataset. This dataset not only contains information about individual patterns but also includes the interactions between patterns, providing comprehensive knowledge support for subsequent reasoning and decision-making.

[0052] Example 5: This embodiment further defines the specific steps of S3 in Embodiment 1 and the functional implementation method of the data reasoning unit in Embodiment 2, which are as follows: S31. Perform correlation coefficient analysis on all knowledge features to obtain a number of feature scores.

[0053] S32. Arrange all the feature scores in descending order to obtain a feature score sequence, and screen all the knowledge features according to the feature score sequence to determine a number of key features.

[0054] S33. According to the frequent association patterns in the knowledge mining dataset, combined with all the key features, perform tagging processing on the implicit knowledge set to obtain the first tag class.

[0055] S34. According to the effective processing patterns in the knowledge mining dataset, combined with all the key features, perform tagging processing on the explicit knowledge set to obtain the second tag class.

[0056] S35. According to the first tag class and the second tag class, construct a data conversion table based on the implicit knowledge set and the explicit knowledge set.

[0057] S36. Preset inference conditions according to requirements, and traverse the data conversion table according to the inference conditions to perform logical reasoning to obtain a knowledge reasoning result.

[0058] This embodiment uses a feature selection algorithm to perform correlation coefficient analysis on the multiple knowledge features to obtain multiple feature scores; serializes the multiple feature scores in descending order to obtain a feature score sequence, and screens the multiple knowledge features according to the feature score sequence to determine multiple key features.

[0059] The system terminal applies feature selection algorithms, such as Pearson correlation coefficient analysis, chi-square test, etc., to multiple knowledge features in the power knowledge base. These algorithms are used to evaluate the contribution of each feature to the target variable (such as fault occurrence, load overload, etc.). The correlation coefficient reflects the linear relationship between the feature and the target variable. The higher the value, the greater the impact of the feature on the target. In the analysis process, taking Pearson correlation coefficient analysis as an example, the system terminal calculates the average value of each feature and the target variable to measure their overall level, compares the difference between each data point and the average value, calculates the covariance between the feature and the target variable (i.e., the synchronous change trend of the two), and then standardizes the covariance by measuring the dispersion degree (standard deviation) of the data of the feature and the target respectively, so as to obtain a correlation coefficient value. This coefficient ranges from -1 to 1, which is used to represent the association strength and direction between the feature and the target, and the absolute value of the coefficient is used as the score of the feature. The higher the score, the stronger the correlation between the feature and the target variable. Subsequently, the system terminal arranges the scores of all knowledge features in descending order to generate a feature score sequence. This sequence arranges the influence of each feature on the target variable from high to low. The score sequence can help the system terminal identify which features are more representative in analysis and decision-making, and help reduce the interference of unimportant features on the analysis results. According to the generated feature score sequence, the system terminal screens multiple knowledge features to determine the most influential features. This screening process is essentially to evaluate the contribution of each feature to the target. The system terminal will set a screening threshold (such as 0.7), and retain the features with scores higher than the screening threshold as multiple key features. In this way, the system terminal can focus on processing the features that have the greatest influence on the target variable, so as to be more accurate in subsequent analysis.

[0060] Based on the frequent association pattern and the multiple key features, the implicit knowledge set is labeled to obtain the first label class; based on the effective processing pattern and the multiple key features, the explicit knowledge set is labeled to obtain the second label class.

[0061] The system terminal combines multiple key features with frequent association patterns and effective processing patterns. For example, if the key features include load fluctuations and equipment temperature, the roles of these features in the frequent association patterns and effective processing patterns will be examined to ensure that label processing focuses more on the features that have the greatest impact on the target. The system terminal will mark the combination of the frequent association pattern and the key feature on each data item in the implicit knowledge set. For example, if the combination of a load fluctuation exceeding 80% and an equipment temperature higher than 40°C frequently causes equipment failures, then these conditions will be used as labels for the relevant data items. In this way, the system terminal checks each data item one by one to see if it meets a certain combination of a frequent association pattern and a key feature. If it does, the corresponding label is assigned to the data item, thus marking all data items that meet the combination of the frequent association pattern and the key feature as the first label class, forming a label set that can reflect implicit associations. In addition, the system terminal will perform a similar operation on the explicit knowledge set, but using the effective processing pattern and the combination of key features. For example, if a humidity exceeding 90% is directly related to the failure risk, then these conditions will be used as labels for the conforming explicit knowledge data items. The system terminal traverses each data item in the explicit knowledge set to find the data items that meet the combination of the effective processing pattern and the key feature, and marks these data items as the second label class. In this way, the data items that meet the conditions in the explicit knowledge set are given clear and intuitive labels to reflect the feature patterns that have a direct impact on the target variable.

[0062] Map the implicit knowledge set and the explicit knowledge set to the data inference unit for data invocation according to the first label class and the second label class, and construct a data conversion table; set inference conditions, and traverse the data conversion table according to the inference conditions for logical inference to generate the knowledge inference result.

[0063] The system terminal maps the data entries in the implicit knowledge set and the explicit knowledge set to the data inference unit according to the first label class and the second label class. Through the first label class, the frequent association patterns in the implicit knowledge set are classified and mapped according to the label class. The implicit knowledge items carry complex association patterns and multi-dimensional features, and these data will be marked by the first label class during mapping so that they can be called during the inference process. Through the second label class, the knowledge entries in the explicit knowledge set that conform to the effective processing pattern are mapped to the data inference unit and marked with the second label class. After the mapping is completed, it is systematically sorted according to the transformation relationship within the data inference unit to construct a data transformation table. The data transformation table arranges the knowledge entries in an orderly manner, clarifies the relationship between each entry and the label class and features, and provides an efficient structured entry for subsequent calls. After determining the data transformation table, the system terminal sets inference conditions according to the target requirements. Each inference condition corresponds to specific knowledge features or association patterns to ensure that the inference process focuses on the key factors that may cause abnormalities in the power system. Then, it traverses the data transformation table one by one according to the set inference conditions, and conducts logical inferences on each knowledge entry to generate corresponding inference results. Subsequently, all the generated inference results are deduplicated and merged to obtain the knowledge inference results. These results include possible risk predictions, event trigger conditions, associated fault modes, etc., and finally form a complete inference output covering implicit and explicit knowledge inferences to support subsequent decision-making and responses.

[0064] Embodiment 6: This embodiment further defines the specific method for constructing a data transformation table based on the implicit knowledge set and the explicit knowledge set according to the first label class and the second label class in Embodiment 3, as follows: S351, construct a label index for the implicit knowledge set based on the first label class as the first label index.

[0065] S352, construct a label index for the explicit knowledge set based on the second label class as the second label index.

[0066] S353, traverse the implicit knowledge set for matching according to the first label index to obtain the first matching result, and generate the first transformation relationship according to the first matching result.

[0067] S354, traverse the explicit knowledge set for matching according to the second label index to obtain the second matching result, and generate the second transformation relationship according to the second matching result.

[0068] S355, fill the implicit knowledge set with data according to the first transformation relationship in combination with the first label index to obtain the first data filling result.

[0069] S356, fill the explicit knowledge set with data according to the second transformation relationship in combination with the second label index to obtain the second data filling result.

[0070] S357. Combine the first data filling result and the second data filling result to obtain a data conversion table.

[0071] Map the implicit knowledge set to the data inference unit for response based on the first tag class to construct a first tag index; map the explicit knowledge set to the data inference unit for response based on the second tag class to construct a second tag index; traverse the implicit knowledge set according to the first tag index for matching to generate a first matching result, make a call based on the first matching result, and set a first conversion relationship; traverse the explicit knowledge set according to the second tag index for matching to generate a second matching result, make a call based on the second matching result, and set a second conversion relationship; perform data filling on the implicit knowledge set according to the first conversion relationship in combination with the first tag index to obtain a first data filling result; perform data filling on the explicit knowledge set according to the second conversion relationship in combination with the second tag index to obtain a second data filling result; add the first data filling result and the second data filling result to the data conversion table.

[0072] Based on the specific patterns and characteristics in the first tag class, the system terminal maps the implicit knowledge set to the data reasoning unit. Each mapped knowledge entry is tagged according to its associated frequent patterns and key characteristics for quick identification and invocation. In this process, the system terminal creates a first tag index, classifying and arranging all entries in the implicit knowledge set according to tags, enabling the data reasoning unit to efficiently access the implicit knowledge items. Similarly, based on the second tag class, the explicit knowledge set is mapped to the data reasoning unit. Explicit knowledge entries are intuitive and easily observable combinations of patterns. When mapping, the system terminal tags each explicit knowledge entry with the corresponding tag and creates a second tag index for these entries according to the second tag class. This index helps quickly locate explicit knowledge items for a fast response during the reasoning process. Subsequently, according to the first tag index, each data entry in the implicit knowledge set is traversed to check if it meets the set pattern and feature conditions. The system terminal marks the data entries that meet the conditions as the first matching results. For example, if an entry in the implicit knowledge set meets the frequent pattern of peak load period and rising equipment temperature, it will be marked as a first matching result. After obtaining the first matching results, the system terminal sets the first transformation relationship. This transformation relationship is used to maintain the data logic and feature associations in the implicit knowledge set during subsequent data filling. For example, according to the peak load - rising temperature pattern in the first matching results, the transformation relationship of the implicit data is defined as the association pattern of peak load (greater than 80%) and rising temperature (greater than 40°C), and all implicit knowledge items that meet this pattern are mapped to a unified transformation structure. According to the second tag index, the explicit knowledge set is traversed to find the knowledge entries that meet the conditions of the second tag class and generate the second matching results. For example, if an entry in the explicit knowledge set meets the pattern that the failure rate increases when the humidity exceeds 90%, it will be marked as a second matching result. After obtaining the second matching results, the second transformation relationship is set based on this result. This relationship applies to the data filling process of the explicit knowledge set to ensure the logical consistency and pattern applicability of the explicit knowledge items. For example, for the pattern that the failure rate increases when the humidity exceeds 90%, all explicit knowledge items that meet this pattern are mapped to the same type of transformation relationship. Then, according to the first transformation relationship and in combination with the first tag index, data filling processing is performed on the implicit knowledge set. In this process, the system terminal filters out the knowledge entries that meet the first transformation relationship from the implicit knowledge set according to the first tag index. For example, in the peak load - rising temperature pattern, all entries containing a load exceeding 80% and a temperature exceeding 40°C will be found and used as candidates for data filling. Then, according to the key features (such as load and temperature) in the first transformation relationship, the standard fields for data filling are set. At this time, the characteristic values of load and temperature are uniformly filled into the corresponding fields, such as the peak load status field and the temperature rising status field, to ensure the unified structure of the filled entries.After the above filling step, the system terminal converts all implicit knowledge entries that conform to the first conversion relationship into standardized filling results. The filled data entries have consistent fields, unified feature values, and clear causal relationship labels, forming the first data filling result. For example, the implicit knowledge item of peak load - temperature rise is filled into a consistent data structure to ensure that the data inference unit can uniformly process this type of data during analysis. The second conversion relationship, combined with the second label index, fills the explicit knowledge set. At this time, all explicit knowledge items that conform to patterns such as humidity exceeding 90% will be filled into a standardized structure, forming the second data filling result, thus providing consistent explicit knowledge data for the data inference unit. Then, the system terminal integrates the first data filling result and the second data filling result and adds them to the data conversion table. The data conversion table records all filled knowledge entries according to label categories and features, providing an ordered and standardized data structure for subsequent logical reasoning and correlation analysis.

[0073] Based on the power knowledge base, perform demand analysis to generate multiple power demand information, collect multiple inference scenario information according to the multiple power demand information, perform trigger analysis according to the multiple inference scenario information, and set the inference conditions; based on the inference conditions, trigger the data inference unit to traverse the data conversion table to determine implicit knowledge entry information and explicit knowledge entry information; perform logical reasoning on the implicit knowledge set according to the implicit knowledge entry information to generate a first inference result; perform logical reasoning on the explicit knowledge set according to the explicit knowledge entry information to generate a second inference result; de-duplicate and merge the first inference result and the second inference result to obtain the knowledge inference result.

[0074] Embodiment 7: This embodiment further limits the specific method of presetting inference conditions according to requirements and traversing the data conversion table for logical reasoning to obtain the knowledge inference result in Embodiment 5, as follows: S361, generate several power demand information according to requirements, generate several inference scenario information according to all power demand information, perform trigger analysis on all inference scenario information, and obtain inference conditions.

[0075] S362, based on the inference conditions, traverse the data conversion table to determine implicit knowledge entry information and explicit knowledge entry information.

[0076] S363, perform logical reasoning on the implicit knowledge set based on the implicit knowledge entry information to obtain a first inference result.

[0077] S364, perform logical reasoning on the explicit knowledge set based on the explicit knowledge entry information to obtain a second inference result.

[0078] S365, deduplicate and merge the first inference result and the second inference result to obtain the knowledge inference result.

[0079] The system terminal conducts a comprehensive requirements analysis of the power knowledge base to identify and extract the current requirement elements of the power system. For example, power demand information includes requirements for load distribution optimization, fault warning, equipment maintenance planning, etc. By analyzing the data patterns and historical records in the power knowledge base, multiple pieces of power demand information are generated, and these demand information reflect the core issues that need attention and the requirements to be solved during the operation of the power system. Subsequently, according to the power demand information, the inference scenario information related to these requirements is collected. The inference scenario information is specific operating or status data used to describe the actual requirements of the power system under different circumstances. For example, for the requirement of load distribution optimization, the inference scenario information includes real-time load data, historical load fluctuation patterns, etc.; for the requirement of fault warning, the inference scenario information includes equipment fault history, current operating status, etc. The system terminal compares and analyzes multiple pieces of inference scenario information with the requirements to identify which conditions trigger specific inference processes. For example, if the real-time load exceeds a set threshold (such as 80%), the system terminal takes this as a trigger condition and starts to infer the load distribution optimization plan. Similarly, for the scenario where the equipment temperature reaches the warning value, a fault warning trigger condition will be set. These trigger conditions are set as inference conditions for starting the analysis of the data conversion table by the data inference unit. Then, the system terminal traverses the data conversion table according to the inference conditions to identify the implicit knowledge entries and explicit knowledge entries that meet the conditions. The implicit knowledge entries contain deep-seated association relationships and implicit patterns, while the explicit knowledge entries contain directly observable patterns and processing rules. The system terminal extracts the data that meet the inference conditions item by item and classifies them according to the implicit and explicit knowledge entry information for subsequent logical inference. Among the extracted implicit knowledge entries, the system terminal focuses on the key association patterns and causal chains. By analyzing these relationships, the potential causal relationships between the data are understood. For example, in the pattern of peak load → temperature rise → fault, the indirect impact of the load level change on temperature and fault occurrence can be identified. Then, an inference path is established for each implicit knowledge entry, that is, based on the causal chain of each entry, its impact on the current power demand is analyzed. For example, in the load fluctuation pattern, analyze how the high-load condition gradually affects the temperature rise and ultimately increases the probability of equipment failure. The inference path shows the causal chain of events, enabling the system to gradually evaluate the impact of different factors in complex associations. The purpose of this step is to analyze the chain reaction of events along the inference path to more accurately identify the risk points. After determining the inference path, the system terminal applies the preset inference rules to conduct specific inferences on the association patterns. The inference rules are usually set according to the characteristics of the implicit knowledge set, including probabilistic inference, causal inference, etc. The system verifies the causal relationships in the inference path through these rules, judges whether the association pattern meets the inference conditions, and thus draws further conclusions. The system terminal integrates the above inference path, association pattern, and the results of the applied rules into the first inference result.This result is an accurate description of implicit associations and potential risks, facilitating subsequent decision-making and responses. Similarly, the system terminal conducts logical reasoning on the explicit knowledge set based on explicit knowledge entry information. The reasoning process of the explicit knowledge set focuses on direct causal and processing relationships. For example, when the humidity exceeds 90%, the failure probability increases. The system terminal uses the processing patterns in the explicit knowledge entries and combines them in a similar manner as above to form a second reasoning result. Finally, the first and second reasoning results are de-duplicated to remove duplicate or similar conclusions and information. The combined reasoning result contains the valid information from both the implicit and explicit knowledge sets, forming the final knowledge reasoning result. This result is a comprehensive analysis output that covers deep implicit associations and surface-level explicit processing suggestions, providing comprehensive support for intelligent decision-making and responses in the power system.

[0080] Example 8: This embodiment further defines the specific steps of S4 in Embodiment 1 and the functional implementation method of the optimization and enhancement unit in Embodiment 2, as follows: S41, Structurally classify the knowledge reasoning result to obtain several knowledge structure categories.

[0081] S42, Combine all knowledge structure categories with all knowledge features for normalization processing to obtain a normalization processing result, and set a feature window according to the normalization processing result.

[0082] S43, Analyze the generative model based on the feature window, conduct performance evaluation according to the analysis result, and obtain the performance score of the generative model.

[0083] S44, Develop an incremental learning strategy for the generative model according to the performance score of the generative model.

[0084] S45, Conduct an optimization analysis on the incremental learning strategy developed for the generative model to obtain the learning rate of the generative model, enhance the generative model according to the learning rate of the generative model, output the enhanced model parameter set of the generative model, and combine the enhanced model parameter set with the learning rate of the generative model to obtain a knowledge parameter group.

[0085] In the optimization and enhancement unit, the system uses the obtained knowledge reasoning result and inputs the knowledge mining data set into the optimization and enhancement unit for further analysis. The purpose of this step is to enable the model to identify deep patterns and key features in the data based on the reasoning result, thereby improving the model's recognition ability and prediction accuracy. The optimization and enhancement unit will conduct performance evaluation based on the various data and features of the knowledge mining data set, obtain multiple model performance scores, and perform incremental learning on the generative model based on these scores. By adjusting the parameter configuration of the model, it can be optimized to more accurately respond to complex power system scenarios.

[0086] Import the knowledge mining data set into the optimization and enhancement unit for structured classification based on the knowledge reasoning result to determine multiple knowledge structure categories; perform normalization processing in combination with the multiple knowledge features according to the multiple knowledge structure categories to generate a normalization processing result, and set a feature window according to the normalization processing result; analyze the generative model through the feature window to obtain the model analysis result, perform performance evaluation according to the model analysis result to obtain multiple model performance scores; perform incremental learning on the generative model according to the multiple model performance scores to formulate a learning strategy; perform optimization analysis on the generative model by executing the learning strategy to obtain a model learning rate, enhance the generative model according to the model learning rate, output an enhanced model parameter set, and determine the knowledge parameter group according to the enhanced model parameter set in combination with the model learning rate.

[0087] The system terminal imports the knowledge mining dataset into the optimization and enhancement unit, and structurally classifies the data based on the knowledge reasoning results, identifying multiple knowledge structure categories. For example, the dataset is divided into categories such as load management, equipment failure, and environmental factors. Each category contains associated data entries, facilitating more targeted analysis by the model. Under each knowledge structure category, the system terminal normalizes the relevant knowledge features (such as load, temperature, humidity, etc.) through the maximum-minimum method. Normalization is to adjust the data values of different features to the same scale range (for example, between 0 and 1) to ensure that the influence of each feature will not be unbalanced due to numerical differences during model training. The normalization result ensures the consistency of each knowledge feature in the analysis, laying a foundation for feature comparison and pattern analysis. Based on the normalized data, the system terminal sets a feature window, that is, selects a set of key features as the analysis input of the model. The feature window can serve as the perspective for the model to observe, enabling the model to focus on the feature combinations that have a significant impact on the prediction results during training. The system terminal will select important features such as load and temperature to enter the feature window based on the normalization result to optimize the feature input structure of the model. Through the feature combination input through the feature window, the generative model is analyzed to understand the prediction performance of the model under different feature combinations. The model analysis results include the performance differences of the model under different feature windows. The system terminal will evaluate the performance of the model based on these results, calculate the performance scores of each combination, for example, the accuracy of the model, the loss value of the loss function, the error rate, etc. Multiple model performance scores provide the degree of influence of each feature combination on the overall performance of the model. According to the performance scores, the system terminal uses the feature combinations with higher scores for incremental learning of the generative model. Incremental learning is to gradually add new learning data based on the existing knowledge of the model to continuously optimize the model performance. The system terminal formulates a learning strategy according to the scoring results to determine which feature combinations to prioritize learning and updating during training. This learning strategy enables the model to gradually accumulate the influence of important features and improve the prediction effect. On the basis of incremental learning, further optimization analysis is carried out to calculate the learning rate of the model (that is, the speed at which the model adapts to new knowledge). Specifically, the system terminal first sets an initial learning rate to control the step size of model parameter updates, enabling the model to gradually adapt to new feature data, and then monitors the change trend of the error during the incremental learning process, and judges whether the current learning rate is appropriate by calculating the change rate of the error. If the error change of the model exceeds the error threshold, it means that its learning speed for new data needs to be accelerated, and the system terminal will accordingly increase the learning rate; when the error change tends to be stable, the learning rate will gradually decrease to ensure that the model can capture subtle features more accurately. In this way, by dynamically adjusting the learning rate, the system can find an optimal balance point, thereby determining the model learning rate, enabling the model to efficiently absorb new knowledge without ignoring details due to overly large learning steps.Finally, by setting the model learning rate, the system terminal first allows the model to learn important feature combinations and outputs an enhanced model parameter set. This enhanced parameter set fully integrates the influence of the optimal feature combination, and then combines these enhanced model parameters based on the model learning rate to form a more adaptable and accurate prediction knowledge parameter group, providing reliable support for the intelligent analysis and decision-making of the power system.

[0088] Embodiment 9: This embodiment further defines the specific steps of the collaborative analysis of the knowledge parameter group in S5 of Embodiment 1 to obtain the first decision-making suggestion and the corresponding functional implementation method in the execution feedback unit of Embodiment 2, as follows: Step 1: Use the learning rate of the generative model to perform parameter traversal on the knowledge parameter group, and evaluate the combination deviation of the knowledge parameter group according to the traversal result to obtain a parameter combination deviation evaluation value.

[0089] Step 2: Use the parameter combination deviation evaluation value and the corresponding knowledge parameter group as incremental data for deviation correction to obtain a parameter correction result. <{

[0090] Step 3: Use the parameter correction result to perform collaborative analysis on the knowledge parameter group, set the collaborative priority, and perform weighted scoring on the knowledge parameter group according to the collaborative priority to obtain the first decision-making suggestion.

[0091] In the execution feedback unit, the system terminal traverses the previously obtained knowledge parameter group in combination with the model learning rate through the execution feedback unit to analyze the performance of different parameter combinations and conduct a combined deviation evaluation. This evaluation process generates a parameter combination deviation value, reflecting the adaptability differences of each parameter combination under different conditions. The system terminal inputs the deviation evaluation value and the corresponding knowledge parameter group into the execution feedback unit as incremental data to carry out deviation correction to ensure that the parameter group better meets the actual requirements. After completing the deviation correction, a collaborative analysis is further performed on the corrected knowledge parameter group, and the relative importance of different parameters is sorted by setting the collaborative priority. The system terminal weights and scores each knowledge parameter group according to the priority and generates the first decision recommendation in combination with the actual situation. This recommendation undergoes multi-dimensional evaluation to ensure its efficiency and applicability under the current conditions, providing reliable support for the operation of the power system, such as optimizing load distribution, adjusting equipment operation parameters, or issuing maintenance warnings. These recommendations aim to improve the stability and efficiency of the power system and help the system respond more effectively to the current operating conditions. Subsequently, the system terminal performs simulation operations or adjustments based on the decision recommendation, and the feedback after the execution of the decision is recorded to form feedback response data. The feedback response data includes the execution result and specific information on the response of the power system. For example, whether the decision achieves the expected effect, whether a new operating state or problem has occurred, etc. The system terminal organizes these feedback data to provide a reference for future decisions, thereby continuously optimizing the intelligence level of the execution feedback unit.

[0092] Perform parameter traversal on the knowledge parameter group according to the model learning rate, conduct a combined deviation evaluation of the knowledge parameter group based on the traversal result, and obtain a parameter combination deviation evaluation value; input the parameter combination deviation evaluation value and the corresponding knowledge parameter group into the execution feedback unit as incremental data for deviation correction to obtain a parameter correction result; perform a collaborative analysis on the knowledge parameter group through the parameter correction result, set the collaborative priority, and weight and score the knowledge parameter group according to the collaborative priority to formulate the first decision recommendation.

[0093] The system terminal conducts a traversal test on each parameter combination in the knowledge parameter group based on the model learning rate to analyze the performance of different combinations under different conditions. A deviation will be generated for each combination during the traversal process, that is, the gap between the actual performance and the expected target. The system terminal records this difference as the combined deviation evaluation value to quantify the adaptability and reliability of each combination. The smaller the deviation evaluation value, the more the performance of the combination meets the expected requirements. The system terminal combines all the obtained deviation evaluation values and the corresponding knowledge parameters into incremental data and inputs it into the execution feedback unit. The execution feedback unit performs a weighted average on the deviation evaluation values in these incremental data through the deviation correction algorithm to quantify the overall parameter deviation. Then, according to the direction and magnitude of the deviation, the parameter values are gradually adjusted. For parameters with large deviations (such as key parameters of load or equipment status), these values will be adjusted preferentially to make them gradually tend to the actual requirements. For example, if the load parameter deviation is large, the parameter value will be reduced to a range more in line with the current load level to avoid excessive load setting causing stress on the equipment. During the adjustment process, the system terminal sets a convergence criterion (such as the allowable error range), that is, when the corrected parameter deviation drops to a certain range, the system terminal will stop adjusting and determine that the parameter has reached the optimal corrected value. After each adjustment, the overall deviation of all parameter groups is rechecked to ensure that each parameter is balanced with each other on the basis of global optimization, so as to obtain the final parameter correction result. After completing the deviation correction, the system terminal conducts a collaborative analysis based on the corrected knowledge parameter group to evaluate the interaction and overall effect of each parameter combination. The system terminal will comprehensively consider the performance of the parameter combination in combination with the load status and equipment status, and set a collaborative priority for each group of parameters, that is, the priority is set according to the corrected deviation evaluation value in the parameter correction result. The larger the corrected deviation evaluation value, the lower the priority. The collaborative priority indicates the relative importance of each parameter combination in achieving the overall optimization goal, enabling the system terminal to focus on the parameter groups with more decision-making influence. According to the collaborative priority, the system terminal performs a weighted scoring on each knowledge parameter group. This scoring process combines the performance values of the load status and equipment status, and screens out the combination with the highest score, so as to formulate the optimal regulation strategy. For example, during the current peak load period, the adjustment strategy suitable for load peak shaving in the parameter combination will be preferentially recommended, that is, the optimal regulation decision is screened out by combining the weighted scores of the load status and equipment status to reduce the equipment stress and improve the operation stability. Finally, the system terminal generates the first decision recommendation, providing an optimal solution based on the actual conditions and priority analysis for the operation regulation of the power system.

[0094] Example 10: In the dynamic optimization unit, the system terminal introduces a power environment information set (such as real-time weather, power demand fluctuations, equipment aging status, etc.), combines it with the previous feedback response data, and further dynamically optimizes the first decision-making suggestion. This optimization aims to enable the system terminal to adjust the original decision-making plan considering the external environmental impact, making it more suitable for the current power system state. During this process, the system terminal analyzes the key factors in the power environment information set. For example, if high temperature or strong wind weather is shown, it will prioritize increasing the operation of cooling equipment or adjusting the load distribution to reduce the pressure on key equipment. At the same time, it will combine the actual execution effects of the first decision-making suggestion recorded in the feedback response data to determine which aspects of the adjustment can most effectively improve system stability. Finally, a second decision-making suggestion is generated based on the first decision, making the new decision more dynamically and comprehensively adapt to the current power demand and environmental conditions, providing a more reliable basis for the optimized management of the power system.

[0095] Embodiment 11: Please refer to Figure 3 As shown, the present invention also provides an electronic device 100 for an optimization method of a power knowledge enhanced generative model based on knowledge reasoning; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0096] The memory 101 can be used to store the computer program 103. The processor 102 realizes the steps of the optimization method of the power knowledge enhanced generative model based on knowledge reasoning described in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the electronic device 100 (such as audio data, etc.). In addition, the memory 101 can include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0097] The at least one processor 102 may be a Central Processing Unit (CPU), or may also be other general-purpose processors, Digital Signal Processors (DSPs), Application Specific Integrated Circuits (ASICs), Field-Programmable Gate Arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or the processor 102 may also be any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and connects various parts of the entire electronic device 100 through various interfaces and lines.

[0098] The memory 101 in the electronic device 100 stores multiple instructions to implement an optimization method for a power knowledge enhanced generative model based on knowledge reasoning. The processor 102 can execute the multiple instructions to implement: Obtain the current information data of the power environment; Extract the knowledge features in the power knowledge base, perform knowledge mining on the power knowledge base according to the knowledge features and the required knowledge requirements, and obtain a knowledge mining data set; Perform data reasoning on the knowledge mining data set to obtain a knowledge reasoning result; Analyze the generative model according to the knowledge reasoning result to obtain the learning rate of the generative model, and enhance the generative model according to the learning rate of the generative model to obtain a knowledge parameter group; Perform collaborative analysis on the knowledge parameter group to obtain a first decision recommendation, and perform execution feedback on the first decision recommendation in the enhanced generative model to obtain feedback response data; Dynamically optimize the first decision recommendation according to the feedback response data combined with the current information data, generate a second decision recommendation, and perform execution feedback again in the enhanced generative model.

[0099] Embodiment 12: If the modules / units integrated in the electronic device 100 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, and read-only memory (ROM, Read-Only Memory).

[0100] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, system, or computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0101] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of processes and / or blocks in the flowchart and / or block diagram can also be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0102] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements the specified functions in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0103] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions for implementing the steps of the process Figure 1 one process or a plurality of processes and / or blocks Figure 1 steps for the functions specified in one block or a plurality of blocks.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications or equivalent replacements can still be made to the specific embodiments of the present invention. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. An optimization method for an electricity knowledge enhanced generative model based on knowledge reasoning, characterized in that It includes the following steps: Obtain the current information data of the power environment; Extract the knowledge features from the power knowledge base, perform knowledge mining on the power knowledge base according to the knowledge features and the required knowledge needs, and obtain a knowledge mining data set; Perform data reasoning on the knowledge mining data set to obtain a knowledge reasoning result; Analyze the generative model according to the knowledge reasoning result to obtain the learning rate of the generative model, and enhance the generative model according to the learning rate of the generative model to obtain a knowledge parameter group; Perform collaborative analysis on the knowledge parameter group to obtain a first decision-making suggestion, and perform execution feedback on the first decision-making suggestion in the enhanced generative model to obtain feedback response data; Dynamically optimize the first decision-making suggestion according to the feedback response data combined with the current information data to generate a second decision-making suggestion, and perform execution feedback again in the enhanced generative model.

2. The optimization method of the power knowledge enhanced generative model based on knowledge reasoning according to claim 1, wherein The power knowledge base is constructed based on historical power data. The specific method for constructing the power knowledge base is as follows: Obtain the expert experience data in this field, perform structured analysis on the expert experience data in this field, and obtain a number of knowledge entries; Classify the historical power data based on a number of knowledge entries to obtain a number of data classes; Perform logical analysis on the expert experience data in this field to obtain the logical rules between all knowledge entries, and formulate knowledge integration rules according to the logical rules between all knowledge entries; Perform knowledge integration on all data classes according to the knowledge integration rules to obtain a number of knowledge levels; Perform data consistency verification on all knowledge levels, and form a power knowledge base with the knowledge levels that pass the data consistency verification.

3. The optimization method of the power knowledge enhanced generative model based on knowledge reasoning according to claim 1, characterized in that The specific method for extracting the knowledge features from the power knowledge base and performing knowledge mining on the power knowledge base according to the knowledge features and the required knowledge needs to obtain a knowledge mining data set is as follows: Extract the knowledge entries corresponding to the knowledge levels in the power knowledge base, perform feature analysis on the power knowledge base based on the knowledge entries, and obtain a number of knowledge features; Traverse and screen the power database based on all knowledge features to obtain a number of screened data, and perform explicit and implicit analysis on the knowledge entries corresponding to the screened data to obtain an implicit knowledge set and an explicit knowledge set; Perform association analysis on the implicit knowledge set to obtain a number of association support degrees, and perform knowledge mining on the implicit knowledge set according to the number of association support degrees to determine frequent association patterns; Perform reasoning analysis on the explicit knowledge set to obtain a number of processing confidence degrees, and perform knowledge mining on the explicit knowledge set according to the number of processing confidence degrees to determine effective processing patterns; Integrate the frequent association patterns and the effective processing patterns to obtain a knowledge mining data set.

4. The optimization method of the power knowledge enhanced generative model based on knowledge reasoning according to claim 1, characterized in that, The specific method for performing data reasoning on the knowledge mining data set to obtain a knowledge reasoning result is as follows: Perform correlation coefficient analysis on all knowledge features to obtain a number of feature scores; Arrange all feature scores in descending order to obtain a feature score sequence, and screen all knowledge features according to the feature score sequence to determine a number of key features; According to the frequent association patterns in the knowledge mining data set, combined with all key features, perform tagging processing on the implicit knowledge set to obtain a first tag class; According to the effective processing mode in the knowledge mining dataset, combined with all key features, the explicit knowledge set is labeled to obtain the second label class; Based on the first label class and the second label class, a data conversion table is constructed based on the implicit knowledge set and the explicit knowledge set; According to the preset inference conditions according to the requirements, traverse the data conversion table according to the inference conditions for logical reasoning to obtain the knowledge reasoning result.

5. The optimization method of the power knowledge enhanced generative model based on knowledge reasoning according to claim 4, characterized in that The specific method for constructing a data conversion table based on the first label class and the second label class, based on the implicit knowledge set and the explicit knowledge set, is as follows: Construct a label index for the implicit knowledge set based on the first label class as the first label index; Construct a label index for the explicit knowledge set based on the second label class as the second label index; Traverse the implicit knowledge set according to the first label index for matching to obtain the first matching result, and generate the first conversion relationship according to the first matching result; Traverse the explicit knowledge set according to the second label index for matching to obtain the second matching result, and generate the second conversion relationship according to the second matching result; Fill in the data of the implicit knowledge set according to the first conversion relationship combined with the first label index to obtain the first data filling result; Fill in the data of the explicit knowledge set according to the second conversion relationship combined with the second label index to obtain the second data filling result; Combine the first data filling result with the second data filling result to obtain the data conversion table.

6. The optimization method of the power knowledge enhanced generative model based on knowledge reasoning according to claim 4, wherein The specific method for presetting inference conditions according to the requirements, traversing the data conversion table according to the inference conditions for logical reasoning to obtain the knowledge reasoning result is as follows: Generate a number of power demand information according to the requirements, generate a number of inference scenario information according to all power demand information, and perform trigger analysis on all inference scenario information to obtain the inference conditions; Traverse the data conversion table based on the inference conditions to determine the implicit knowledge entry information and the explicit knowledge entry information; Perform logical reasoning on the implicit knowledge set based on the implicit knowledge entry information to obtain the first reasoning result; Perform logical reasoning on the explicit knowledge set based on the explicit knowledge entry information to obtain the second reasoning result; Deduplicate and merge the first reasoning result and the second reasoning result to obtain the knowledge reasoning result.

7. The optimization method of the power knowledge enhanced generative model based on knowledge reasoning according to claim 1, wherein Analyze the generative model according to the knowledge reasoning result to obtain the learning rate of the generative model, and enhance the generative model according to the learning rate of the generative model. The specific method for obtaining the knowledge parameter group is as follows: Classify the knowledge reasoning result structurally to obtain a number of knowledge structure categories; Normalize all knowledge structure categories combined with all knowledge features to obtain the normalization result, and set the feature window according to the normalization result; Analyze the generative model based on the feature window, and perform performance evaluation according to the analysis result to obtain the performance score of the generative model; Formulate an incremental learning strategy for the generative model according to the performance score of the generative model; Optimize and analyze the incremental learning strategy formulated for the generative model to obtain the learning rate of the generative model, enhance the generative model according to the learning rate of the generative model, output the enhanced model parameter set of the generative model, and combine the enhanced model parameter set with the learning rate of the generative model to obtain the knowledge parameter group.

8. The optimization method of the power knowledge enhanced generative model based on knowledge reasoning according to claim 1, wherein The specific method for collaborative analysis of the knowledge parameter group to obtain the first decision-making suggestion is as follows: Use the learning rate of the generative model to perform parameter traversal on the knowledge parameter group, and evaluate the combined deviation of the knowledge parameter group according to the traversal results to obtain the combined deviation evaluation value of the parameters; Use the combined deviation evaluation value of the parameters and the corresponding knowledge parameter group as incremental data for deviation correction to obtain the parameter correction result; Use the parameter correction result to perform collaborative analysis on the knowledge parameter group, set the collaborative priority, and perform weighted scoring on the knowledge parameter group according to the collaborative priority to obtain the first decision-making suggestion.

9. The power knowledge enhanced generative model optimization system based on knowledge reasoning is characterized in that Including: A data acquisition unit for obtaining the current information data of the power environment; A data conversion unit for extracting the knowledge features in the power knowledge base, performing knowledge mining on the power knowledge base according to the knowledge features and the required knowledge requirements to obtain a knowledge mining data set; A data reasoning unit for performing data reasoning on the knowledge mining data set to obtain a knowledge reasoning result; An optimization enhancement unit for analyzing the generative model according to the knowledge reasoning result to obtain the learning rate of the generative model, and enhancing the generative model according to the learning rate of the generative model to obtain the knowledge parameter group; An execution feedback unit for performing collaborative analysis on the knowledge parameter group to obtain the first decision-making suggestion, and performing execution feedback on the first decision-making suggestion in the enhanced generative model to obtain feedback response data; A dynamic optimization unit for dynamically optimizing the first decision-making suggestion according to the feedback response data combined with the current information data to generate a second decision-making suggestion, and performing execution feedback again in the enhanced generative model.

10. The power knowledge enhanced generative model optimization system based on knowledge reasoning according to claim 9, wherein In the data conversion unit, the power knowledge base is constructed according to historical power data. The process of constructing the power knowledge base is as follows: Obtain the expert experience data in this field, perform structured analysis on the expert experience data in this field to obtain a number of knowledge entries; Classify the historical power data based on a number of knowledge entries to obtain a number of data classes; Perform logical analysis on the expert experience data in this field to obtain the logical rules between all knowledge entries, and formulate knowledge integration rules according to the logical rules between all knowledge entries; Perform knowledge integration on all data classes according to the knowledge integration rules to obtain a number of knowledge levels; Perform data consistency verification on all knowledge levels, and form a power knowledge base with the knowledge levels that pass the data consistency verification.

11. The power knowledge enhanced generative model optimization system based on knowledge reasoning according to claim 9, wherein The function of the data conversion unit is realized by the following method: Extract the knowledge entries corresponding to the knowledge levels in the power knowledge base, perform feature analysis on the power knowledge base based on the knowledge entries to obtain a number of knowledge features; Traverse and screen the power database based on all knowledge features to obtain a number of screened data, and perform explicit and implicit analysis on the knowledge entries corresponding to the screened data to obtain an implicit knowledge set and an explicit knowledge set; Perform association analysis on the implicit knowledge set to obtain a number of association support degrees, and perform knowledge mining on the implicit knowledge set according to the number of association support degrees to determine the frequent association patterns; Perform reasoning analysis on the explicit knowledge set to obtain a number of processing confidence levels, and perform knowledge mining on the explicit knowledge set according to the number of processing confidence levels to determine the effective processing patterns; Integrate frequent association patterns with effective processing patterns to obtain a knowledge mining dataset.

12. The power knowledge enhanced generative model optimization system based on knowledge reasoning according to claim 9, wherein, The functions of the data reasoning unit are implemented by the following methods: Perform correlation coefficient analysis on all knowledge features to obtain several feature scores; Sort all feature scores in descending order to obtain a feature score sequence, and screen all knowledge features according to the feature score sequence to determine several key features; According to the frequent association patterns in the knowledge mining dataset, combined with all key features, perform tagging on the implicit knowledge set to obtain the first tag class; According to the effective processing patterns in the knowledge mining dataset, combined with all key features, perform tagging on the explicit knowledge set to obtain the second tag class; Based on the first tag class and the second tag class, construct a data conversion table based on the implicit knowledge set and the explicit knowledge set; Preset inference conditions according to requirements, and traverse the data conversion table according to the inference conditions for logical reasoning to obtain knowledge inference results.

13. The power knowledge enhanced generative model optimization system based on knowledge reasoning according to claim 12, characterized in that, In the data reasoning unit, the function of constructing a data conversion table based on the implicit knowledge set and the explicit knowledge set according to the first tag class and the second tag class is implemented by the following method; Construct a tag index for the implicit knowledge set based on the first tag class as the first tag index; Construct a tag index for the explicit knowledge set based on the second tag class as the second tag index; Traverse the implicit knowledge set for matching according to the first tag index to obtain the first matching result, and generate the first conversion relationship according to the first matching result; Traverse the explicit knowledge set for matching according to the second tag index to obtain the second matching result, and generate the second conversion relationship according to the second matching result; Fill the implicit knowledge set with data according to the first conversion relationship combined with the first tag index to obtain the first data filling result; Fill the explicit knowledge set with data according to the second conversion relationship combined with the second tag index to obtain the second data filling result; Combine the first data filling result with the second data filling result to obtain a data conversion table.

14. The power knowledge enhanced generative model optimization system based on knowledge reasoning according to claim 12, characterized in that, In the data reasoning unit, the function of presetting inference conditions according to requirements, traversing the data conversion table according to the inference conditions for logical reasoning, and obtaining knowledge inference results is implemented by the following method; Generate several power demand information according to requirements, generate several inference scenario information according to all power demand information, perform trigger analysis on all inference scenario information to obtain inference conditions; Traverse the data conversion table based on the inference conditions to determine implicit knowledge entry information and explicit knowledge entry information; Perform logical reasoning on the implicit knowledge set based on the implicit knowledge entry information to obtain the first inference result; Perform logical reasoning on the explicit knowledge set based on the explicit knowledge entry information to obtain the second inference result; Deduplicate and merge the first inference result and the second inference result to obtain knowledge inference results.

15. The power knowledge enhanced generative model optimization system based on knowledge reasoning according to claim 9, wherein The functions of the optimization and enhancement unit are implemented by the following methods: Classify the knowledge inference results structurally to obtain several knowledge structure categories; Perform normalization processing on all knowledge structure categories combined with all knowledge features to obtain a normalization processing result, and set a feature window according to the normalization processing result; Analyze the generative model based on the feature window, and perform performance evaluation according to the analysis result to obtain the performance score of the generative model; Formulate an incremental learning strategy for the generative model according to the performance score of the generative model; Conduct an optimization analysis on the incremental learning strategy formulated for the generative model to obtain the learning rate of the generative model. Enhance the generative model according to the learning rate of the generative model, output the enhanced model parameter set of the generative model, and combine the enhanced model parameter set with the learning rate of the generative model to obtain a knowledge parameter group.

16. The power knowledge enhanced generative model optimization system based on knowledge reasoning according to claim 9, characterized in that, In the execution feedback unit, the function of obtaining the first decision suggestion through collaborative analysis of the knowledge parameter group is realized by the following method: Use the learning rate of the generative model to perform parameter traversal on the knowledge parameter group, conduct a combined deviation evaluation of the knowledge parameter group according to the traversal results, and obtain a parameter combined deviation evaluation value; Use the parameter combined deviation evaluation value and the corresponding knowledge parameter group as incremental data for deviation correction to obtain a parameter correction result; Use the parameter correction result to conduct collaborative analysis on the knowledge parameter group, set the collaborative priority, and perform weighted scoring on the knowledge parameter group according to the collaborative priority to obtain the first decision suggestion.

17. An electronic device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the method for optimizing the power knowledge enhanced generative model based on knowledge reasoning according to any one of claims 1 to 8.

18. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the method for optimizing the power knowledge enhanced generative model based on knowledge reasoning according to any one of claims 1 to 8.

Images