Comprehensive energy coordination optimization method and system based on language model analysis

Abstract

The present invention discloses a comprehensive energy coordination optimization method and system based on language model analysis, which relates to the field of energy management and control technology. A large language model is used to analyze historical equipment load, meteorological, operating conditions and economic indicator data to form a historical valid information segment group describing the energy system; a historical information feature parameter library is constructed; through primary and secondary equivalent classification, a historical information feature parameter subset is obtained; a reference feature parameter equivalent range interval library is constructed; the library is used to retrieve and analyze the real-time information feature parameter group, match the corresponding interval and allocate the energy system according to the disposal strategy; the above technical scheme of the present invention realizes the comprehensive integration, intelligent analysis and real-time optimization of energy system data, and effectively improves the intelligence and refinement level of energy management.

Description

Technical Field

[0001] The present invention relates to the field of energy management and control technology, and in particular to a comprehensive energy coordination optimization method and system based on language model analysis. Background Art

[0002] In the current energy management field, as energy systems become increasingly complex and diversified, how to achieve efficient and intelligent comprehensive energy coordination and optimization has become a key issue that needs to be solved urgently. Traditional energy management methods often rely on manual experience and simple data analysis, and it is difficult to effectively integrate and process massive multi-source heterogeneous data in energy systems, resulting in low energy utilization efficiency, high costs, and unstable system operation. In particular, in the face of real-time changes in energy systems, traditional methods are even more powerless and unable to make accurate decisions and deployments in a timely manner. Summary of the invention

[0003] The purpose of the present invention is to provide a comprehensive energy coordination optimization method and system based on language model analysis for realizing rapid energy system deployment.

[0004] The present invention discloses a comprehensive energy coordination optimization method based on language model analysis, comprising:

[0005] Step S100, for each of the historical equipment load data, the historical meteorological data, the historical equipment operation status data and the historical economic indicator data, an identification keyword table is set, a number of identification keywords are set in the identification keyword table, the historical equipment load data, the historical meteorological data, the historical equipment operation status data and the historical economic indicator data are analyzed by using a large language model, and based on the identification keyword table, a number of historical valid information segments to be intercepted are determined, and the several historical valid information segments are mutually associated to obtain a historical valid information segment group for describing the energy system;

[0006] Step S200, performing feature analysis on the historical valid information segment group, and constructing a historical information feature parameter group based on the analysis result, and recording a combination of several historical information feature parameter groups as a historical information feature parameter library;

[0007] Step S300, configuring a treatment strategy for each historical information feature parameter group in the historical information feature parameter library, performing a primary equivalence classification on the historical information feature parameter groups in the historical information feature parameter library to obtain a number of historical information feature parameter sets, and performing a secondary equivalence classification on the historical information feature parameter groups in the historical information feature parameter sets based on the equivalence of the treatment strategies to obtain a historical information feature parameter sub-set;

[0008] Step S400, comparing the historical information feature parameter groups in the historical information feature parameter sub-set, determining the initial range interval of each historical information feature parameter, and determining the equivalent range interval of the historical information feature parameter in the historical information feature parameter group based on the length of the initial range interval, and recording the combination of the equivalent range intervals of the historical information feature parameters as a reference feature parameter equivalent range interval group, and constructing a plurality of reference feature parameter equivalent range interval groups into a reference feature parameter equivalent range interval library;

[0009] Step S500, using the reference characteristic parameter equivalent range interval library, retrieves and analyzes the real-time information characteristic parameter group generated in real time, determines the matching reference characteristic parameter equivalent range interval, and allocates the energy system based on the treatment strategy corresponding to the reference characteristic parameter equivalent range interval.

[0010] In some embodiments disclosed in the present invention, the method for performing feature analysis on a valid information segment group includes:

[0011] Step S201, determining the identification keyword corresponding to each historical valid information segment in the valid information segment group, and determining the degree description paragraph or degree description parameter related to the identification keyword in the valid information segment;

[0012] Step S202, the corresponding recognition keywords, relevant degree description paragraphs and degree description parameters are combined to obtain an information feature group, the information feature group is compared and analyzed with the preset information feature identification and judgment standard, the information feature type in the information feature group and the information feature parameters corresponding to the information feature type are determined, and the information feature parameter group is constructed in a one-to-one correspondence between the information feature type and the information feature parameter.

[0013] In some embodiments disclosed in the present invention, a method for performing an equivalent classification on information feature parameter groups includes:

[0014] Step S301, establishing a parameter mapping axis for each information feature type in the information feature parameter group, and mapping information feature parameters corresponding to a number of information feature types to the parameter mapping axis in the form of information feature parameter mapping points;

[0015] Step S302, analyzing the clustering characteristics of the information feature parameter mapping points on the parameter mapping axis, and based on the analysis results, determining a number of mapping point clustering blocks on the information feature parameter mapping points, calculating the average information feature parameter mapping point of the information feature parameter mapping in the mapping point clustering blocks, the calculation method of the average information feature parameter mapping point includes calculating the average value of the information feature parameters corresponding to the information feature parameter mapping points to obtain the average information feature parameters, and recording the combination of the average information feature parameters as an average information feature parameter group, and identifying the mapping point of the average information feature parameter on the mapping point clustering block as the average information feature parameter mapping point;

[0016] Step S303, comparing the information feature parameter group with different average information feature parameter groups, determining the most consistent average information feature parameter group, and classifying it under the average information feature parameter group;

[0017] Step S304, analyzing the mapping point difference vector of each information feature parameter mapping point in the information feature parameter group relative to the average information feature parameter mapping point to obtain a mapping point difference vector group;

[0018] Step S305, calibrate feature weight coefficients for different information feature types in the information feature parameter group, combine the mapping point difference vector group of the information feature parameter group to obtain a mapping point difference vector difference group, and calibrate feature weight coefficients for different information feature types to determine the reduced dimension feature parameters of the information feature parameter group, and based on the reduced dimension feature parameters, reclassify the information feature parameter group classified under the average information feature parameter group to obtain a historical information feature parameter set.

[0019] In some embodiments disclosed in the present invention, a method for analyzing the aggregation characteristics of information feature parameter mapping points on a parameter mapping axis includes:

[0020] Step S3021, a number of mapping segments are set on the parameter mapping axis, and the number of mapping points of the information feature parameter mapping points in each mapping segment is determined, the difference in the number of mapping points between adjacent mapping segments is compared, and the ratio of the difference in the number of mapping points to the difference in the number of mapping points is calculated. If the ratio of the difference in the number of mapping points is less than or equal to a preset value, the adjacent mapping segments are successively combined to form an aggregated mapping segment, wherein if the ratio of the difference in the number of mapping points between adjacent mapping segments is greater than a preset value, the adjacent mapping segments are determined as the boundary of the aggregated mapping segment.

[0021] In some embodiments disclosed in the present invention, a method for comparing an information feature parameter group with different average information feature parameter groups is as follows:

[0022] Step S3031, comparing the feature parameter difference of each relative information feature parameter, if the feature parameter difference is less than or equal to a preset value, then determining the corresponding information feature type as a matching information feature type;

[0023] Step S3032, based on the proportion of the matching information feature types to all the matching information feature types, determine the degree of matching between the information feature parameter group and the average information feature parameter group, and based on the degree of matching, select the most matching average information feature parameter group.

[0024] In some embodiments disclosed in the present invention, a method for determining a dimension reduction feature parameter of an information feature parameter group includes:

[0025] Step S3051, analyzing the mapping point difference vector difference group, respectively calculating the positive sub-dimension reduction feature parameters of the positive difference vector difference and the negative sub-dimension reduction feature parameters of the negative difference vector difference, and recording the positive sub-dimension reduction feature parameters and the negative sub-dimension reduction feature parameters as the dimension reduction feature parameters;

[0026] Among them, the expression for calculating the positive sub-dimensionality reduction feature parameter or the negative sub-dimensionality reduction feature parameter is: ;

[0027] in, is the positive sub-dimensionality reduction feature parameter or the negative sub-dimensionality reduction feature parameter, For the The feature weight coefficient corresponding to the difference of the difference vector of the positive or negative mapping points, is the total number of positive or negative mapping point difference vector differences, For the The difference vector of the mapping points in the positive or negative direction, is the adjustment coefficient of the difference effect of the mapping point difference vector, The difference vector difference effect adjustment constant for the mapping point.

[0028] In some embodiments disclosed in the present invention, the method for determining the initial range interval of each information feature parameter includes:

[0029] Step S401, analyzing each feature parameter group in the information feature parameter subset, determining the minimum information feature parameter and the maximum information feature parameter among the information feature parameters of the same information feature type, and identifying the interval between the minimum information feature parameter and the maximum information feature parameter as the initial range interval.

[0030] In some embodiments disclosed in the present invention, based on the length of the initial range interval, a method for determining an equivalent range interval of information feature parameters in an information feature parameter group includes:

[0031] Step S402, a range interval length correspondence table is set for the initial range interval length, and the range interval length correspondence expression includes several initial range interval length-equivalent range interval length correspondence groups. According to the mapping of the initial range length of the information feature parameter in the range interval length correspondence table, the equivalent range interval length corresponding to the information feature parameter is determined, and the equivalent range interval of the information feature parameter is determined based on the equivalent range interval length and the information feature parameter as the center.

[0032] In some embodiments disclosed in the present invention, there is also disclosed a comprehensive energy coordination optimization system based on language model analysis, including:

[0033] The first module is used to set a recognition keyword table for each of the historical equipment load data, the historical meteorological data, the historical equipment operation status data and the historical economic indicator data, wherein a number of recognition keywords are set in the recognition keyword table, and the historical equipment load data, the historical meteorological data, the historical equipment operation status data and the historical economic indicator data are analyzed by using a large language model, and based on the recognition keyword table, a number of historical valid information segments to be intercepted are determined, and the several historical valid information segments are correlated with each other to obtain a historical valid information segment group for describing the energy system;

[0034] The second module is used to perform feature analysis on the historical valid information segment group, and based on the analysis result, construct a historical information feature parameter group, and record the combination of several historical information feature parameter groups as a historical information feature parameter library;

[0035] The third module is used to configure a disposal strategy for each historical information feature parameter group in the historical information feature parameter library, perform a primary equivalence classification on the historical information feature parameter groups in the historical information feature parameter library to obtain a number of historical information feature parameter sets, and perform a secondary equivalence classification on the historical information feature parameter groups in the historical information feature parameter sets based on the equivalence of the disposal strategies to obtain a historical information feature parameter sub-set;

[0036] The fourth module is used to compare the historical information feature parameter groups in the historical information feature parameter subconcentration, determine the initial range interval of each historical information feature parameter, and determine the equivalent range interval of the historical information feature parameter in the historical information feature parameter group based on the length of the initial range interval, and record the combination of the equivalent range intervals of the historical information feature parameters as a reference feature parameter equivalent range interval group, and construct a plurality of reference feature parameter equivalent range interval groups into a reference feature parameter equivalent range interval library;

[0037] The fifth module is used to use the reference characteristic parameter equivalent range interval library to search and analyze the real-time information characteristic parameter group generated in real time, determine the matching reference characteristic parameter equivalent range interval, and allocate the energy system based on the treatment strategy corresponding to the reference characteristic parameter equivalent range interval.

[0038] The present invention discloses a comprehensive energy coordination optimization method and system based on language model analysis, which relates to the field of energy management and control technology. A large language model is used to analyze historical equipment load, meteorological, operating conditions and economic indicator data to form a historical valid information segment group describing the energy system; a historical information feature parameter library is constructed; through primary and secondary equivalent classification, a historical information feature parameter subset is obtained; a reference feature parameter equivalent range interval library is constructed; this library is used to retrieve and analyze real-time information feature parameter groups, match corresponding intervals and allocate energy systems according to disposal strategies; the above technical solution of the present invention realizes comprehensive integration, intelligent analysis and real-time optimization of energy system data, and effectively improves the intelligence and refinement level of energy management.

[0039] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a method step diagram of the comprehensive energy coordination optimization method based on language model analysis disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] The technical solution of the present invention is further described below through the accompanying drawings and embodiments.

[0042] The following will be combined with the accompanying drawings and specific embodiments to clearly and completely describe the technical solution of the present invention. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention and cannot be understood as limiting the scope of protection of the present invention. Those skilled in the art in this field can make some non-essential improvements and adjustments based on the content of the present invention described below. In the present invention, unless otherwise clearly specified and limited, the technical terms used in the present invention should be the common meanings understood by the technical personnel described in the present invention.

[0043] Example:

[0044] The present invention discloses a comprehensive energy coordination optimization method based on language model analysis, see Figure 1 ,include:

[0045] Step S100, an identification keyword table is set for each of the historical equipment load data, the historical meteorological data, the historical equipment operating status data and the historical economic indicator data, a number of identification keywords are set in the identification keyword table, the historical equipment load data, the historical meteorological data, the historical equipment operating status data and the historical economic indicator data are analyzed using a large language model, and based on the identification keyword table, a number of historical valid information segments to be intercepted are determined, and the several historical valid information segments are correlated to obtain a historical valid information segment group for describing the energy system.

[0046] This step is the key link of data preprocessing. First, for multiple data sources such as historical equipment load, weather, operating conditions, and economic indicators, identification keyword tables are set up respectively. These keyword tables are the "sieves" for subsequent analysis, used to filter out information closely related to the energy system from massive data. Using advanced large language models, we can deeply understand text data and accurately intercept historical valid information segments based on keyword tables. These intercepted information segments do not exist in isolation, but through mutual correlation, they form a group of valid information segments that comprehensively describe the historical operating status of the energy system, providing a solid foundation for subsequent feature analysis and optimization.

[0047] Step S200, performing feature analysis on the historical valid information segment group, and constructing a historical information feature parameter group based on the analysis result, and recording a combination of several historical information feature parameter groups as a historical information feature parameter library.

[0048] The historical valid information segment group is exemplified as follows:

[0049] Equipment load data: During the period from January 1 to January 7, 2023, the power load during the daily evening peak period (18:00-20:00) exceeded 90% of the rated capacity.

[0050] Meteorological data: During the same period, the average temperature in the region dropped by 3 degrees Celsius compared with the same period last year, and there was continuous light rain.

[0051] Equipment operation status data: A key equipment had a brief fault alarm on January 5, but then automatically resumed normal operation.

[0052] Economic indicator data: Industrial electricity consumption increased by 5% month-on-month during the week, reflecting the active economic activities.

[0053] After obtaining the historical valid information segment group, the next task is to conduct in-depth mining of this information and extract key features that can reflect the characteristics of the energy system. By performing feature analysis on the historical valid information segment group, parameters or indicators that have a significant impact on the operation of the energy system can be identified. These features are organized into historical information feature parameter groups and further integrated into a historical information feature parameter library. This library is like a "feature warehouse" that stores the "fingerprint" of the historical operating status of the energy system, providing strong data support for subsequent strategy configuration and decision support.

[0054] In some embodiments disclosed in the present invention, the method for performing feature analysis on a valid information segment group includes:

[0055] Step S201, determining the identification keyword corresponding to each historical valid information segment in the valid information segment group, and determining the degree description paragraph or degree description parameter related to the identification keyword in the valid information segment.

[0056] Step S202, the corresponding recognition keywords, relevant degree description paragraphs and degree description parameters are combined to obtain an information feature group, the information feature group is compared and analyzed with the preset information feature identification and judgment standard, the information feature type in the information feature group and the information feature parameters corresponding to the information feature type are determined, and the information feature parameter group is constructed in a one-to-one correspondence between the information feature type and the information feature parameter.

[0057] Step S300, configure a disposal strategy for each historical information feature parameter group in the historical information feature parameter library, perform an equivalent classification on the historical information feature parameter groups in the historical information feature parameter library, obtain a number of historical information feature parameter sets, and based on the equality of the disposal strategies, perform a secondary equivalent classification on the historical information feature parameter groups in the historical information feature parameter sets to obtain a historical information feature parameter sub-set.

[0058] After having the historical information feature parameter library, the next step is to configure corresponding disposal strategies for these feature parameters. This step is the key to connecting data analysis with actual operations. By configuring disposal strategies for each historical information feature parameter group, it can be clear how to deal with specific features when they appear. At the same time, in order to simplify management and improve efficiency, it is also necessary to classify feature parameter groups equally. Through primary and secondary equal classification, a more refined subset of historical information feature parameters can be obtained. This subset not only retains key features, but also realizes centralized management of strategies through the equality of disposal strategies, which facilitates subsequent real-time decision-making.

[0059] In some embodiments disclosed in the present invention, a method for performing an equivalent classification on information feature parameter groups includes:

[0060] Step S301: establishing a parameter mapping axis for each information feature type in the information feature parameter group, and mapping information feature parameters corresponding to a number of information feature types onto the parameter mapping axis in the form of information feature parameter mapping points.

[0061] Step S302, analyzing the clustering characteristics of the information feature parameter mapping points on the parameter mapping axis, and based on the analysis results, determining a number of mapping point clustering blocks on the information feature parameter mapping points, calculating the average information feature parameter mapping point of the information feature parameter mapping in the mapping point clustering blocks, the calculation method of the average information feature parameter mapping point includes calculating the average value of the information feature parameters corresponding to the information feature parameter mapping point to obtain the average information feature parameter, and recording the combination of the average information feature parameters as an average information feature parameter group, and identifying the mapping point of the average information feature parameter on the mapping point clustering block as the average information feature parameter mapping point.

[0062] Step S303, compare the information feature parameter group with different average information feature parameter groups, determine the most consistent average information feature parameter group, and classify it under the average information feature parameter group.

[0063] Step S304: Analyze the mapping point difference vector of each information feature parameter mapping point in the information feature parameter group relative to the average information feature parameter mapping point to obtain a mapping point difference vector group.

[0064] Step S305, calibrate feature weight coefficients for different information feature types in the information feature parameter group, combine the mapping point difference vector group of the information feature parameter group to obtain a mapping point difference vector difference group, and calibrate feature weight coefficients for different information feature types to determine the reduced dimension feature parameters of the information feature parameter group, and based on the reduced dimension feature parameters, reclassify the information feature parameter group classified under the average information feature parameter group to obtain a historical information feature parameter set.

[0065] In some embodiments disclosed in the present invention, a method for analyzing the aggregation characteristics of information feature parameter mapping points on a parameter mapping axis includes:

[0066] Step S3021, a number of mapping segments are set on the parameter mapping axis, and the number of mapping points of the information feature parameter mapping points in each mapping segment is determined, the difference in the number of mapping points between adjacent mapping segments is compared, and the ratio of the difference in the number of mapping points to the difference in the number of mapping points is calculated. If the ratio of the difference in the number of mapping points is less than or equal to a preset value, the adjacent mapping segments are successively combined to form an aggregated mapping segment, wherein if the ratio of the difference in the number of mapping points between adjacent mapping segments is greater than a preset value, the adjacent mapping segments are determined as the boundary of the aggregated mapping segment.

[0067] In some embodiments disclosed in the present invention, a method for comparing an information feature parameter group with different average information feature parameter groups is as follows:

[0068] Step S3031, comparing the feature parameter difference of each relative information feature parameter, if the feature parameter difference is less than or equal to a preset value, the corresponding information feature type is determined to be a matching information feature type.

[0069] Step S3032, based on the proportion of the matching information feature types to all the matching information feature types, determine the degree of matching between the information feature parameter group and the average information feature parameter group, and based on the degree of matching, select the most matching average information feature parameter group.

[0070] In some embodiments disclosed in the present invention, a method for determining a dimension reduction feature parameter of an information feature parameter group includes:

[0071] Step S3051, analyze the mapping point difference vector difference group, calculate the positive sub-dimension reduction feature parameters of the positive difference vector difference, and the negative sub-dimension reduction feature parameters of the negative difference vector difference, and record the positive sub-dimension reduction feature parameters and the negative sub-dimension reduction feature parameters as dimensionality reduction feature parameters.

[0072] Among them, the expression for calculating the positive sub-dimensionality reduction feature parameter or the negative sub-dimensionality reduction feature parameter is: .

[0073] in, is the positive sub-dimensionality reduction feature parameter or the negative sub-dimensionality reduction feature parameter, For the The feature weight coefficient corresponding to the difference of the difference vector of the positive or negative mapping points, is the total number of positive or negative mapping point difference vector differences, For the The difference vector of the mapping points in the positive or negative direction, is the adjustment coefficient of the difference effect of the mapping point difference vector, The difference vector difference effect adjustment constant for the mapping point.

[0074] Step S400, compare the historical information feature parameter groups in the historical information feature parameter subconcentration, determine the initial range interval of each historical information feature parameter, and based on the length of the initial range interval, determine the equivalent range interval of the historical information feature parameters in the historical information feature parameter group, and record the combination of the equivalent range intervals of the historical information feature parameters as a reference feature parameter equivalent range interval group, and construct several reference feature parameter equivalent range interval groups into a reference feature parameter equivalent range interval library.

[0075] After obtaining the historical information feature parameter subset, it is necessary to further determine the value range of each feature parameter. This step is the basis for constructing a reference feature parameter equivalent range interval library. By comparing the feature parameters in the historical information feature parameter subset, the initial range interval of each feature parameter can be determined. Then, based on the length of the initial range interval, the equivalent range interval in the historical information feature parameter group is further determined. These equivalent range intervals are combined into a reference feature parameter equivalent range interval group, and finally a reference feature parameter equivalent range interval library is constructed. This library is like a "range guide", which provides a clear reference for the subsequent matching of real-time information feature parameter groups.

[0076] In some embodiments disclosed in the present invention, the method for determining the initial range interval of each information feature parameter includes:

[0077] Step S401, analyzing each feature parameter group in the information feature parameter subset, determining the minimum information feature parameter and the maximum information feature parameter among the information feature parameters of the same information feature type, and identifying the interval between the minimum information feature parameter and the maximum information feature parameter as the initial range interval.

[0078] In some embodiments disclosed in the present invention, based on the length of the initial range interval, a method for determining an equivalent range interval of information feature parameters in an information feature parameter group includes:

[0079] Step S402, a range interval length correspondence table is set for the initial range interval length, and the range interval length correspondence expression includes several initial range interval length-equivalent range interval length correspondence groups. According to the mapping of the initial range length of the information feature parameter in the range interval length correspondence table, the equivalent range interval length corresponding to the information feature parameter is determined, and the equivalent range interval of the information feature parameter is determined based on the equivalent range interval length and the information feature parameter as the center.

[0080] Step S500, using the reference characteristic parameter equivalent range interval library, retrieves and analyzes the real-time information characteristic parameter group generated in real time, determines the matching reference characteristic parameter equivalent range interval, and allocates the energy system based on the treatment strategy corresponding to the reference characteristic parameter equivalent range interval.

[0081] The last step is real-time decision-making and deployment. Using the reference feature parameter equivalent range interval library built earlier, the real-time information feature parameter group generated in real time can be retrieved and analyzed. By matching the reference feature parameter equivalent range interval, historical situations similar to the current operating status can be quickly located, and the energy system can be deployed based on the corresponding disposal strategy. This step realizes a seamless connection from data analysis to actual operation, ensuring that the energy system can make timely and accurate responses according to actual conditions, thereby achieving efficient, intelligent and coordinated optimization of comprehensive energy.

[0082] In some embodiments disclosed in the present invention, there is also disclosed a comprehensive energy coordination optimization system based on language model analysis, including:

[0083] The first module is used to set up an identification keyword table for each of the historical equipment load data, the historical meteorological data, the historical equipment operating status data and the historical economic indicator data, wherein a number of identification keywords are set in the identification keyword table, and the historical equipment load data, the historical meteorological data, the historical equipment operating status data and the historical economic indicator data are analyzed using a large language model, and based on the identification keyword table, a number of historical valid information segments to be intercepted are determined, and the several historical valid information segments are correlated to obtain a historical valid information segment group for describing the energy system.

[0084] The second module is used to perform feature analysis on the historical valid information segment group, and based on the analysis results, construct a historical information feature parameter group, and record the combination of several historical information feature parameter groups as a historical information feature parameter library.

[0085] The third module is used to configure a disposal strategy for each historical information feature parameter group in the historical information feature parameter library, perform an equivalent classification on the historical information feature parameter groups in the historical information feature parameter library, and obtain a number of historical information feature parameter sets; based on the equality of the disposal strategies, perform a secondary equivalent classification on the historical information feature parameter groups in the historical information feature parameter sets, and obtain a sub-set of historical information feature parameters.

[0086] The fourth module is used to compare the historical information feature parameter groups in the historical information feature parameter subconcentration, determine the initial range interval of each historical information feature parameter, and based on the length of the initial range interval, determine the equivalent range interval of the historical information feature parameters in the historical information feature parameter group, and record the combination of the equivalent range intervals of the historical information feature parameters as a reference feature parameter equivalent range interval group, and construct several reference feature parameter equivalent range interval groups into a reference feature parameter equivalent range interval library.

[0087] The fifth module is used to use the reference characteristic parameter equivalent range interval library to search and analyze the real-time information characteristic parameter group generated in real time, determine the matching reference characteristic parameter equivalent range interval, and allocate the energy system based on the treatment strategy corresponding to the reference characteristic parameter equivalent range interval.

[0088] The present invention discloses a comprehensive energy coordination optimization method and system based on language model analysis, which relates to the field of energy management and control technology. A large language model is used to analyze historical equipment load, meteorological, operating conditions and economic indicator data to form a historical valid information segment group describing the energy system; a historical information feature parameter library is constructed; through primary and secondary equivalent classification, a historical information feature parameter subset is obtained; a reference feature parameter equivalent range interval library is constructed; this library is used to retrieve and analyze real-time information feature parameter groups, match corresponding intervals and allocate energy systems according to disposal strategies; the above technical solution of the present invention realizes comprehensive integration, intelligent analysis and real-time optimization of energy system data, and effectively improves the intelligence and refinement level of energy management.

[0089] Through the description of the above implementation methods, those skilled in the art can clearly understand that the present invention can be implemented by hardware, or by software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each implementation scenario of the present invention.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A comprehensive energy coordination optimization method based on language model analysis, characterized in that: include: For historical equipment load data, historical meteorological data, historical equipment operation status data and historical economic indicator data, an identification keyword table is set respectively, and a number of identification keywords are set in the identification keyword table. The historical equipment load data, historical meteorological data, historical equipment operation status data and historical economic indicator data are analyzed by using a large language model, and based on the identification keyword table, a number of historical valid information segments to be intercepted are determined, and the several historical valid information segments are mutually associated to obtain a historical valid information segment group for describing the energy system; Performing feature analysis on the historical valid information segment group, and constructing a historical information feature parameter group based on the analysis result, and recording a combination of several historical information feature parameter groups as a historical information feature parameter library; A treatment strategy is configured for each historical information feature parameter group in the historical information feature parameter library, and the historical information feature parameter groups in the historical information feature parameter library are once equally classified to obtain a number of historical information feature parameter sets, and based on the equality of the treatment strategies, the historical information feature parameter groups in the historical information feature parameter sets are twice equally classified to obtain a historical information feature parameter sub-set; Compare the historical information characteristic parameter groups in the historical information characteristic parameter subconcentration, determine the initial range interval of each historical information characteristic parameter, and determine the equivalent range interval of the historical information characteristic parameter in the historical information characteristic parameter group based on the length of the initial range interval, and record the combination of the equivalent range intervals of the historical information characteristic parameters as a reference characteristic parameter equivalent range interval group, and construct a plurality of reference characteristic parameter equivalent range interval groups into a reference characteristic parameter equivalent range interval library; Using the reference characteristic parameter equivalent range interval library, the real-time information characteristic parameter group generated in real time is retrieved and analyzed to determine the matching reference characteristic parameter equivalent range interval, and the energy system is deployed based on the treatment strategy corresponding to the reference characteristic parameter equivalent range interval; The method of classifying the information characteristic parameter groups equally includes: Establishing a parameter mapping axis for each information feature type in the information feature parameter group, and mapping information feature parameters corresponding to a number of information feature types to the parameter mapping axis in the form of information feature parameter mapping points; Analyze the clustering characteristics of the information feature parameter mapping points on the parameter mapping axis, and based on the analysis results, determine a number of mapping point clustering blocks on the information feature parameter mapping points, calculate the average information feature parameter mapping point of the information feature parameter mapping in the mapping point clustering blocks, and calculate the average information feature parameter mapping point in a manner including calculating the average value of the information feature parameters corresponding to the information feature parameter mapping points to obtain the average information feature parameters, and record the combination of the average information feature parameters as an average information feature parameter group, and identify the mapping point of the average information feature parameter on the mapping point clustering block as the average information feature parameter mapping point; Comparing the information characteristic parameter group with different average information characteristic parameter groups, determining the most consistent average information characteristic parameter group, and classifying it under the average information characteristic parameter group; Analyze the mapping point difference vector of each information feature parameter mapping point in the information feature parameter group relative to the average information feature parameter mapping point to obtain a mapping point difference vector group; The feature weight coefficients are calibrated for different information feature types in the information feature parameter group, and the mapping point difference vector group of the information feature parameter group is combined to obtain the mapping point difference vector difference group. The feature weight coefficients are calibrated in combination with different information feature types to determine the reduced dimension feature parameters of the information feature parameter group. Based on the reduced dimension feature parameters, the information feature parameter group classified under the average information feature parameter group is reclassified to obtain the historical information feature parameter set.

2. The comprehensive energy coordination optimization method based on language model analysis according to claim 1 is characterized in that: The method for performing feature analysis on the valid information segment group includes: Determine the identification keyword corresponding to each historical valid information segment in the valid information segment group, and determine the degree description paragraph or degree description parameter related to the identification keyword in the valid information segment; The corresponding recognition keywords, relevant degree description paragraphs and degree description parameters are combined to obtain an information feature group, and the information feature group is compared and analyzed with the preset information feature identification and judgment standards to determine the information feature type in the information feature group and the information feature parameters corresponding to the information feature type, and the information feature parameter group is constructed in a one-to-one correspondence between the information feature type and the information feature parameter.

3. The comprehensive energy coordination optimization method based on language model analysis according to claim 1 is characterized in that: The method for analyzing the aggregation characteristics of information feature parameter mapping points on the parameter mapping axis includes: A number of mapping segments are set on the parameter mapping axis, and the number of mapping points of the information feature parameter mapping points in each mapping segment is determined, the difference in the number of mapping points between adjacent mapping segments is compared, and the ratio of the difference in the number of mapping points to the difference in the number of mapping points is calculated. If the ratio of the difference in the number of mapping points is less than or equal to a preset value, the adjacent mapping segments are successively combined to form an aggregated mapping segment, wherein if the ratio of the difference in the number of mapping points between adjacent mapping segments is greater than a preset value, the adjacent mapping segments are determined as the boundary of the aggregated mapping segment.

4. The comprehensive energy coordination optimization method based on language model analysis according to claim 1 is characterized in that: Method for comparing information feature parameter groups with different average information feature parameter groups: Compare the feature parameter difference of each relative information feature parameter, if the feature parameter difference is less than or equal to a preset value, then determine that the corresponding information feature type is a matching information feature type; Based on the proportion of the matching information feature types to all the matching information feature types, the degree of matching between the information feature parameter group and the average information feature parameter group is determined, and based on the degree of matching, the most matching average information feature parameter group is selected.

5. The comprehensive energy coordination optimization method based on language model analysis according to claim 1 is characterized in that: The method for determining the dimension reduction feature parameters of the information feature parameter group includes: Analyze the difference vector difference group of the mapping point, calculate the positive sub-dimension reduction feature parameters of the positive difference vector difference and the negative sub-dimension reduction feature parameters of the negative difference vector difference, and record the positive sub-dimension reduction feature parameters and the negative sub-dimension reduction feature parameters as the dimension reduction feature parameters; Among them, the expression for calculating the positive sub-dimensionality reduction feature parameter or the negative sub-dimensionality reduction feature parameter is: ； in, is the positive sub-dimensionality reduction feature parameter or the negative sub-dimensionality reduction feature parameter, For the The feature weight coefficient corresponding to the difference of the difference vector of the positive or negative mapping points, is the total number of positive or negative mapping point difference vector differences, For the The difference vector of the mapping points in the positive or negative direction, is the adjustment coefficient of the difference effect of the mapping point difference vector, The difference vector difference effect adjustment constant for the mapping point.

6. The comprehensive energy coordination optimization method based on language model analysis according to claim 1 is characterized in that: The method for determining the initial range interval of each information characteristic parameter includes: Each characteristic parameter group in the information characteristic parameter sub-set is analyzed to determine the minimum information characteristic parameter and the maximum information characteristic parameter among the information characteristic parameters of the same information characteristic type, and the interval between the minimum information characteristic parameter and the maximum information characteristic parameter is identified as the initial range interval.

7. The comprehensive energy coordination optimization method based on language model analysis according to claim 1 is characterized in that: The method of determining the equivalent range interval of the information characteristic parameters in the information characteristic parameter group based on the length of the initial range interval includes: A range interval length correspondence table is set for the initial range interval length. The range interval length correspondence expression includes several initial range interval length-equivalent range interval length correspondence groups. The equivalent range interval length corresponding to the information feature parameter is determined according to the mapping of the initial range length of the information feature parameter in the range interval length correspondence table, and the equivalent range interval of the information feature parameter is determined based on the equivalent range interval length and the information feature parameter as the center.

8. Comprehensive energy coordination and optimization system based on language model analysis, characterized by: The comprehensive energy coordination optimization method for executing any one of claims 1 to 7 comprises: The first module is used to set a recognition keyword table for each of the historical equipment load data, the historical meteorological data, the historical equipment operation status data and the historical economic indicator data, wherein a number of recognition keywords are set in the recognition keyword table, and the historical equipment load data, the historical meteorological data, the historical equipment operation status data and the historical economic indicator data are analyzed by using a large language model, and based on the recognition keyword table, a number of historical valid information segments to be intercepted are determined, and the several historical valid information segments are correlated with each other to obtain a historical valid information segment group for describing the energy system; The second module is used to perform feature analysis on the historical valid information segment group, and based on the analysis result, construct a historical information feature parameter group, and record the combination of several historical information feature parameter groups as a historical information feature parameter library; The third module is used to configure a disposal strategy for each historical information feature parameter group in the historical information feature parameter library, perform a primary equivalence classification on the historical information feature parameter groups in the historical information feature parameter library to obtain a number of historical information feature parameter sets, and perform a secondary equivalence classification on the historical information feature parameter groups in the historical information feature parameter sets based on the equivalence of the disposal strategies to obtain a historical information feature parameter sub-set; The fourth module is used to compare the historical information feature parameter groups in the historical information feature parameter subconcentration, determine the initial range interval of each historical information feature parameter, and determine the equivalent range interval of the historical information feature parameter in the historical information feature parameter group based on the length of the initial range interval, and record the combination of the equivalent range intervals of the historical information feature parameters as a reference feature parameter equivalent range interval group, and construct a plurality of reference feature parameter equivalent range interval groups into a reference feature parameter equivalent range interval library; The fifth module is used to use the reference characteristic parameter equivalent range interval library to search and analyze the real-time information characteristic parameter group generated in real time, determine the matching reference characteristic parameter equivalent range interval, and allocate the energy system based on the treatment strategy corresponding to the reference characteristic parameter equivalent range interval.

Images