A method for generating a gas turbine operation and maintenance report based on an intelligent agent

Abstract

The present application relates to the technical field of data processing, in particular to a gas turbine operation and maintenance report generation method based on intelligent agent, comprising the following steps: obtaining a gas turbine temperature and pressure generated working condition characteristic sequence, matching placeholders with labels to generate field mapping, combining temperature and pressure with flow conversion standard flow, comparing settlement readings to extract error alarm and cumulative flow, and filling data into node output operation and maintenance report according to the mapping. In the present application, by extracting real-time temperature and pressure parameters of the gas turbine to generate working condition characteristic sequence, dynamically establishing mapping association between report placeholders and parameter labels, combining natural gas standard state conversion and consumption deviation evaluation to extract error alarm features and automatically fill document nodes, etc. Technical means, the automatic and accurate accounting of multi-source operation and maintenance data and the intelligent matching generation of report content are realized, which gets rid of the dependence on manual reporting, greatly improves the real-time and accuracy of report output, flexibly adapts to different template formats, and enhances the abnormal warning capability of equipment metering.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for generating gas turbine operation and maintenance reports based on intelligent agents. Background Technology

[0002] Data processing technology primarily involves a series of operations such as collecting, organizing, calculating, and analyzing various raw information. Its main purpose is to transform complex underlying information into structured results with application value through specific logic, supporting business decisions and process management. It is widely used in various industries such as industrial manufacturing and the Internet. Among these, the traditional gas turbine operation and maintenance report generation method refers to the process of summarizing and organizing various monitoring indicators and work records generated during the daily operation and maintenance of gas turbines, and finally forming a fixed-format document. This document is mainly used to completely record the real-time operating status of the gas turbine, the troubleshooting steps, and maintenance details, so that managers can intuitively grasp the overall health status and work cycle of the equipment. It is a core link in ensuring equipment safety and optimizing operating efficiency.

[0003] However, traditional gas turbine operation and maintenance mainly relies on manual extraction of data from scattered sensors and log files, followed by manual input into preset templates. This approach suffers from low efficiency, high error rates, and delayed response times. Especially in complex scenarios involving multiple units and multiple data sources, manual processing struggles to meet real-time requirements, and report content often lacks consistency due to operational variations, impacting the accuracy of fault diagnosis and decision-making. Furthermore, existing automation tools often employ fixed rule-based statistics, failing to flexibly adapt to the indicator requirements of different operation and maintenance scenarios, thus limiting the practicality of reports. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method for generating gas turbine operation and maintenance reports based on intelligent agents.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for generating gas turbine operation and maintenance reports based on intelligent agents, comprising the following steps: S1: During the first sampling time period of gas turbine operation, acquire the real-time temperature parameters and real-time pressure parameters of the gas turbine at the gas turbine inlet, extract the state deviation features, and generate an effective gas turbine operating condition feature sequence. S2: Extract the set of text placeholder identifiers from the gas turbine operation and maintenance report template document and the parameter name tags from the effective gas turbine operating condition feature sequence, determine the maximum matching association relationship, and generate intelligent proxy report field mapping association data; S3: Collect the volumetric flow rate read from the instrument of the gas turbine monitoring equipment, extract the actual operating pressure parameters and actual operating temperature parameters at the corresponding time in the effective gas turbine operating condition characteristic sequence, perform natural gas state conversion to obtain standard state flow characteristics, and generate standard state natural gas flow data; S4: During the second sampling time period of gas turbine operation and maintenance, obtain the natural gas settlement form reading, extract the cumulative time value of the standard state natural gas flow data during the second sampling time period, extract the gas turbine intelligent agent metering error alarm feature, and concatenate it with the cumulative time integral value to generate gas turbine intelligent agent error alarm and cumulative flow data; S5: Extract the target location index coordinate information based on the data mapping association of the intelligent agent report fields, fill the gas turbine intelligent agent error alarm and cumulative flow data and the effective gas turbine operating condition feature sequence into the preset document node structure based on the target location index coordinate information, splice the document object sequence, and output the gas turbine operation and maintenance report data.

[0006] As a further aspect of the present invention, the effective gas turbine operating condition feature sequence includes an operating condition baseline vector, an effective state identifier matrix, and an anomaly filtering flag; the intelligent agent report field mapping association data includes a field matching weight dictionary, an agent binding index table, and a parsing rule encoding set; the standard state natural gas flow data includes standard condition instantaneous volume, gaseous compensation coefficient, and reference condition correction value; the gas turbine intelligent agent error alarm and cumulative flow data includes out-of-tolerance threshold trigger records, gas consumption loss measurement values, and periodic settlement gas volume details; the gas turbine operation and maintenance report data includes an operation and maintenance summary summary, energy efficiency calculation charts, and an appendix of agent diagnostic conclusions.

[0007] As a further aspect of the present invention, in the process of determining the maximum matching association, the set of text placeholder identifiers is semantically similar to the parameter name labels to determine the label relationship with the greatest semantic similarity.

[0008] As a further aspect of the present invention, during the process of extracting the intelligent agent metering error alarm features of the gas turbine, a consumption deviation assessment is performed based on the reading of the natural gas settlement form and the cumulative value of time integration.

[0009] As a further aspect of the present invention, the specific steps for obtaining the effective gas turbine operating condition characteristic sequence are as follows: S111: Collect the real-time temperature parameters and real-time pressure parameters of the gas turbine at the gas turbine inlet during the first sampling time period, extract the actual observed pressure value and the prior estimate of pressure from the real-time pressure parameters of the gas turbine, extract the actual observed temperature value and the prior estimate of temperature from the real-time temperature parameters of the gas turbine and perform difference calculation to generate the gas turbine operating state temperature deviation. S112: Based on the temperature deviation of the gas turbine operating state, perform state difference calculation on the actual observed pressure value and the prior estimated pressure value to obtain the pressure deviation value, and perform dot multiplication calculation with the preset state prediction covariance matrix to obtain the pressure deviation characteristics of the gas turbine operating state. S113: Combine the temperature deviation of the gas turbine operating state and the pressure deviation of the gas turbine operating state to perform a numerical update operation of the covariance matrix. Then, perform numerical traversal and serial sorting and recombination operations on the updated state prediction covariance matrix to establish an effective gas turbine operating condition feature sequence.

[0010] As a further aspect of the present invention, the specific steps for obtaining the intelligent agent report field mapping associated data are as follows: S211: Extract the set of text placeholder identifiers in the template document and the parameter name tags in the effective gas turbine operating condition feature sequence; perform natural language processing multidimensional vector transformation operation on the parameter name tags to obtain multidimensional word vector codes; extract the baseline semantic vector coordinates in the set of text placeholder identifiers; extract vector coordinates for the multidimensional word vector codes; and establish feature semantic vector coordinates. S212: Prioritize the reference semantic vector coordinates, perform a spatial dot product and modulus product division operation on the reference semantic vector coordinates and the feature semantic vector coordinates to obtain all mapping space cosine similarity values, and traverse and filter them in descending order to obtain the maximum value of mapping space cosine similarity. S213: Determine the semantic matching degree between the maximum value of the cosine similarity of the mapping space and the elements of the text placeholder identifier set, and map and bind the placeholder identifier corresponding to the maximum value of the cosine similarity of the mapping space with the maximum value to the parameter name label to generate intelligent proxy report field mapping association data.

[0011] As a further aspect of the present invention, the specific steps for obtaining the standard state natural gas flow data are as follows: S311: During the data collection and verification period, the instrument reads the volumetric flow rate, extracts the actual operating pressure value and actual operating temperature value from the effective gas turbine operating condition feature sequence, obtains the standard state absolute pressure constant and the standard state absolute temperature constant, and divides the actual operating pressure value by the standard state absolute pressure constant to obtain the gas turbine intake absolute pressure ratio. S312: Divide the absolute temperature constant under standard conditions with the actual operating temperature value to obtain the ratio of the absolute temperature of the gas turbine intake air, and multiply the ratio of the absolute pressure of the gas turbine intake air with the ratio of the absolute temperature of the gas turbine intake air to obtain the state transition multiplication factor. S313: Based on the state transition multiplication factor and the volumetric flow rate read by the instrument, perform a numerical product conversion operation to obtain the standard state natural gas flow rate product value. Then, serialize and bind the standard state natural gas flow rate product value with the data verification time period to establish standard state natural gas flow rate data.

[0012] As a further aspect of the present invention, the specific steps for obtaining the gas turbine intelligent agent error alarm and cumulative flow data are as follows: S411: Collect the readings of the natural gas settlement form issued by the upper-level circulation gateway within the second sampling time period, perform definite integral surface accumulation and addition operations on the natural gas flow data under the standard state within the time node of the second sampling time period, calculate the total sum of the overall flow consumption within the time interval, and establish the time integral cumulative value; S412: Based on the cumulative value of time integration and the reading of the natural gas settlement form, perform difference subtraction to obtain the difference in natural gas consumption settlement volume, divide the difference in natural gas consumption settlement volume by the reading of the natural gas settlement form to obtain the percentage deviation rate of natural gas consumption; S413: Based on the natural gas consumption percentage deviation rate and the set allowable error constant, extract the gas turbine intelligent agent metering error alarm features that exceed the set allowable error constant range, and combine them with the time integral cumulative value to generate gas turbine intelligent agent error alarm and cumulative flow data.

[0013] As a further aspect of the present invention, the specific steps for obtaining the gas turbine operation and maintenance report data are as follows: S511: Extract the field mapping associated data of the intelligent agent report, perform node traversal search operation, parse and read the internal structure hierarchical attribute information of the data, obtain the target position index positioning point in the corresponding text paragraph interval based on the parsing result, and establish a set of document target position index coordinate information. S512: Obtain the document node structure, provide positioning parameters according to the document target location index coordinate information set, perform node text insertion operation on the gas turbine intelligent agent error alarm and cumulative flow data according to the positioning parameters, fill in the blank node positions of the document node structure, and obtain the document object sequence set; S513: Perform page sequence connection and text paragraph concatenation for each independent document object element in the document object sequence set, merge the text content of all document object elements, construct a global file output format object, and generate gas turbine operation and maintenance report data.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, operating condition feature sequences are generated by extracting real-time temperature and pressure parameters of the gas turbine. Semantic similarity matching is used to dynamically establish a mapping relationship between report placeholders and parameter labels. Combined with natural gas standard state conversion and consumption deviation assessment, error alarm features are extracted and document nodes are automatically filled. This achieves automated and accurate accounting of multi-source operation and maintenance data and intelligent matching and generation of report content. It eliminates the dependence on manual filling, greatly improves the real-time performance and accuracy of report output, flexibly adapts to different template formats, and enhances the early warning capability of equipment metering anomalies. Attached Figure Description

[0015] Figure 1 This is a flowchart of the main steps of the present invention; Figure 2 This is a flowchart of the effective gas turbine operating condition characteristic sequence acquisition process of the present invention; Figure 3 This is a flowchart illustrating the process of obtaining associated data through intelligent agent report field mapping in this invention. Figure 4 This is a flowchart of the standard state natural gas flow data acquisition process of the present invention; Figure 5 This is a flowchart of the intelligent agent error alarm and cumulative flow data acquisition process for gas turbines according to the present invention; Figure 6 This is a flowchart of the gas turbine operation and maintenance report data acquisition process of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0017] Please see Figure 1 A method for generating gas turbine operation and maintenance reports based on intelligent agents includes the following steps: S1: During the first sampling time period of gas turbine operation, acquire the real-time temperature parameters and real-time pressure parameters of the gas turbine at the gas turbine inlet. Extract state deviation features based on the real-time temperature parameters and real-time pressure parameters of the gas turbine to generate an effective gas turbine operating condition feature sequence. S2: Extract the set of text placeholder identifiers and parameter name tags from the gas turbine operation and maintenance report template document and the effective gas turbine operating condition feature sequence. Perform semantic similarity matching between the set of text placeholder identifiers and the parameter name tags to determine the maximum matching association relationship and generate intelligent proxy report field mapping association data. S3: Collect the volumetric flow rate read from the gas turbine monitoring equipment, extract the actual operating pressure and temperature parameters at the corresponding time from the effective gas turbine operating condition characteristic sequence, and perform natural gas state conversion based on the volumetric flow rate read from the instruments, the actual operating pressure value, and the actual operating temperature value to obtain the standard state flow characteristics and generate standard state natural gas flow data. S4: During the second sampling time period of gas turbine operation and maintenance, obtain the natural gas settlement form reading, extract the time cumulative value of standard state natural gas flow data during the second sampling time period, evaluate the consumption deviation based on the natural gas settlement form reading and the time integral cumulative value, extract the gas turbine intelligent agent metering error alarm features, and concatenate the gas turbine intelligent agent metering error alarm features with the time integral cumulative value to generate gas turbine intelligent agent error alarm and cumulative flow data; S5: Extract the target location index coordinate information based on the data mapping of the intelligent agent report fields. Fill the gas turbine intelligent agent error alarm and cumulative flow data and the effective gas turbine operating condition characteristic sequence into the preset document node structure based on the target location index coordinate information. Concatenate the document object sequence and output the gas turbine operation and maintenance report data.

[0018] The effective gas turbine operating condition characteristic sequence includes the operating condition baseline vector, effective state identifier matrix, and anomaly filtering flag; the intelligent agent report field mapping association data includes the field matching weight dictionary, agent binding index table, and parsing rule encoding set; the standard state natural gas flow data includes the standard condition instantaneous volume, gaseous compensation coefficient, and reference condition correction value; the gas turbine intelligent agent error alarm and cumulative flow data includes the out-of-tolerance threshold trigger record, gas consumption loss measurement value, and periodic settlement gas volume details; the gas turbine operation and maintenance report data includes the operation and maintenance summary, energy efficiency accounting chart, and agent diagnostic conclusion appendix.

[0019] Please see Figure 2 The specific steps of S1 are as follows: S111: Collect the real-time temperature parameters and real-time pressure parameters of the gas turbine at the gas turbine inlet during the first sampling time period, extract the actual observed pressure value and the prior estimate of pressure from the real-time pressure parameters of the gas turbine, extract the actual observed temperature value and the prior estimate of temperature from the real-time temperature parameters of the gas turbine and perform difference calculation to generate the gas turbine operating state temperature deviation. During the first sampling period, real-time temperature and pressure parameters of the gas turbine at the inlet are continuously collected. The specific boundary value of the first sampling period is strictly determined based on the average time interval between two adjacent peak airflow pulsations contained in the continuous full-load operation records of similar units. The actual observed pressure value, representing the absolute result of the current measurement, is completely extracted from the collected real-time pressure parameters. Simultaneously, a priori pressure estimate is retrieved from the underlying data table based on dynamic evolution calculations from multiple time-series nodes. Next, the actual observed temperature value is extracted from the real-time temperature parameters, and the corresponding priori temperature estimate is obtained by extrapolating from historical temperature variation rates. An independent difference operation is immediately performed on the actual observed temperature value and the priori temperature estimate. The specific logic of the difference operation is to use the actual observed temperature value as the subtrahend and the priori temperature estimate as the minuend, and then subtract the priori temperature estimate from the actual observed temperature value. If the final difference obtained by subtraction is within the positive range (greater than zero), it means the intake air carries more heat than the standard setting; if the final difference is within the negative range (less than zero), it means the intake air heat does not meet the standard setting; if the difference is exactly zero, it means the standard setting is perfectly met. Finally, the corresponding difference results are saved to generate the gas turbine operating temperature deviation.

[0020] S112: Based on the temperature deviation of the gas turbine operating state, perform state difference calculation on the actual observed pressure value and the prior estimated pressure value to obtain the pressure deviation value, and perform dot multiplication calculation with the preset state prediction covariance matrix to obtain the pressure deviation characteristics of the gas turbine operating state. After the gas turbine operating temperature deviation is extracted, a similar state difference operation is performed in parallel on the previously extracted actual pressure observation value and the similarly retrieved prior pressure estimate value to obtain the pressure deviation value. The state difference operation process is as follows: the actual pressure observation value is defined as the internal benchmark subtracted value in the difference operation, and the prior pressure estimate value is defined as the internal direct subtracted value in the difference operation. The pressure deviation value representing the pipeline pressure variation offset component is obtained by subtracting the pressure prior estimate value from the actual pressure observation value. Subsequently, the pressure deviation value is multiplied by a specific dimension with the preset state prediction covariance matrix written into the operation cache set. The preset state prediction covariance matrix contains a fixed covariance constant coefficient, which is specifically limited to a value of 0.05. The fixed covariance constant coefficient of 0.05 is obtained by extracting the original dataset transmitted from all measurement front-ends under past fault-free operating conditions and solving the mean of the Gaussian distribution variance of the original dataset. The logic of the dot product calculation is to directly extract the fixed covariance constant coefficients from the diagonal of the preset state prediction covariance matrix, and then directly multiply the obtained pressure deviation value by these fixed covariance constant coefficients. This achieves scaling and adjustment of the statistical uncertainty of the physical dimensions of the pressure deviation value. After the multiplication operation and adjustment, the final product value is directly identified as a specific indicator representing the current comprehensive risk exposure level. Finally, the calculation and deduction are completed to obtain the pressure deviation characteristics of the gas turbine operating state that reflect the abnormal fluctuation characteristics of the pipeline.

[0021] S113: Combine the temperature deviation and pressure deviation characteristics of the gas turbine operating status to perform numerical update operation of the covariance matrix. Then, perform numerical traversal and serial sorting and recombination operations on the updated state prediction covariance matrix to establish an effective gas turbine operating condition feature sequence. The generated gas turbine operating state temperature deviation and pressure deviation features are extracted and then combined to perform a covariance matrix numerical update operation. The covariance matrix numerical update operation involves multiplying the gas turbine operating state temperature deviation by the gas turbine operating state pressure deviation features to obtain a cross-influence product term. This cross-influence product term is then added to the original values ​​of the corresponding cross-dimensional coordinate points within the existing preset state prediction covariance matrix, completing the replacement and correction operation of the elements within the old matrix. The updated state prediction covariance matrix, obtained after the correction operation, is immediately subjected to numerical traversal and serialized sorting and recombination operations. Specifically, the numerical traversal involves reading the floating-point values ​​of all coordinate points within the updated state prediction covariance matrix, starting from the first row and first column, row by row and column by column. After reading, a descending serialized sorting and recombination operation is performed based on the absolute values ​​of the floating-point values. If multiple floating-point values ​​have completely identical absolute values, then the original positive and negative signs of the floating-point values ​​are compared, with positive numbers placed before negative numbers. If the positive and negative signs are also completely identical, then they are arranged in ascending order according to their original row coordinates. After detailed comparison, screening, selection, and sorting, the continuously arranged one-dimensional numerical results are fixed and saved, thus thoroughly establishing an effective gas turbine operating condition characteristic sequence with significant dynamic time-series tracking capabilities.

[0022] Please see Figure 3 The specific steps of S2 are as follows: S211: Extract the set of text placeholder identifiers and parameter name labels from the effective gas turbine operating condition feature sequence within the template document; perform natural language processing multidimensional vector transformation operation on the parameter name labels to obtain multidimensional word vector codes; extract the baseline semantic vector coordinates from the set of text placeholder identifiers; extract vector coordinates for the multidimensional word vector codes; and establish feature semantic vector coordinates. The process involves extracting a set of text placeholder identifiers from a pre-defined standardized template document, and simultaneously extracting parameter name labels from the effective gas turbine operating condition feature sequence constructed in the previous steps. Natural Language Processing (NLP) multidimensional vector transformation operations are then performed on the parameter name labels to obtain high-dimensional multidimensional word vector codes. This NLP multidimensional vector transformation operation primarily utilizes a lexical embedding processing network, which includes a dedicated receiver layer for receiving parameter name labels, three fully connected hidden layers, and an output feedback layer responsible for outputting the multidimensional word vector codes. The neuron execution process involves the receiver layer decomposing the Chinese characters of the parameter name labels into basic numeric codes, which are then fed into the hidden layers. The hidden layers perform nonlinear combination mapping operations on different text features using a linear rectified function. The output feedback layer ultimately outputs a one-dimensional multidimensional word vector code composed of sixty-four floating-point numbers. Subsequently, the baseline semantic vector coordinates are extracted from the set of text placeholder identifiers. After extracting the baseline semantic vector coordinates, a deep internal vector coordinate extraction operation is performed on the fresh multidimensional word vector codes. The core process of vector coordinate extraction is to remove the format control symbols at the beginning and end of the multidimensional word vector encoding, extract and retain the continuous floating-point number sequence in the middle of the multidimensional word vector encoding that bears the core expressive function, and use the continuous floating-point number sequence directly as the newly established position anchor point in the mapping space to establish semantic vector coordinates that represent the accurate meaning of the real-time changing label.

[0023] S212: Pre-determine the baseline semantic vector coordinates, perform a spatial dot product and modulus product division operation on the baseline semantic vector coordinates and the feature semantic vector coordinates to obtain all mapping space cosine similarity values, and traverse and filter them in descending order to obtain the maximum value of mapping space cosine similarity. The system reads the coordinates of the prepared baseline semantic vector in advance. Simultaneously, it performs a spatial dot product and a modulus product division operation on the baseline semantic vector coordinates and the extracted feature semantic vector coordinates. The specific execution logic of the spatial dot product and modulus product division operation is divided into three independent operation levels. The first step involves multiplying each dimension value of the baseline semantic vector coordinates with the corresponding dimension values ​​of the feature semantic vector coordinates, and summing all the multiplication results to obtain the final spatial dot product sum. The second step involves squaring and summing the values ​​of each dimension of the baseline semantic vector coordinates, taking the square root to obtain the first modulus value, and similarly squaring and summing the values ​​of each dimension of the feature semantic vector coordinates to obtain the second modulus value. The first modulus value is multiplied by the second modulus value to obtain the overall modulus product value. The third step uses the final spatial dot product sum as the dividend and the overall modulus product value as the divisor, performing a division operation to obtain the quotient, which is then identified as the cosine similarity value of the mapping space. The cosine similarity scores of all matching pairs in the mapping space are aggregated and sorted in descending order of numerical value. The results are then iterated and compared sequentially. The value with the largest quantified numerical result is extracted from the first and last result of the comparison process, representing the maximum cosine similarity score between the parameter name label and the text placeholder identifier, indicating the highest potential match in the mapping space.

[0024] S213: Determine the semantic matching degree between the maximum value of the cosine similarity of the mapping space and the elements of the text placeholder identifier set, and map and bind the placeholder identifier corresponding to the maximum value of the cosine similarity of the mapping space with the parameter name label to generate intelligent agent report field mapping association data. The process involves extracting the maximum cosine similarity value of the previously filtered mapping space and determining its true potential semantic match with each element within the predefined set of text placeholder identifiers. This determination requires a pre-established benchmark threshold, fixed at 0.85. This threshold is based on a manually reviewed and verified analysis of over 10,000 historical similarity matching data sets, ensuring that a similarity value exceeding 0.85 results in a mismatch rate of less than one in a thousand in actual business operations. The maximum cosine similarity value is then compared to the benchmark threshold of 0.85. If the maximum cosine similarity value is greater than the threshold, a definite semantic match is determined. If the maximum value is less than or equal to the threshold, automatic matching is not possible, and an anomaly warning signal is sent to the manual review queue. Under the condition that the semantic match is fully established, the actual text placeholder identifier corresponding to the maximum final value of the mapping space cosine similarity value is written into a low-level relational data table for mapping and binding operations with the provided content parameter name label. The mapping and binding association operation execution logic is to create a new record with an independent primary key data mapping relationship in the data table structure, and write the unique resource locator link of the two into the same data slot. This is then persisted for later filling processes to directly call the intelligent proxy report field mapping association data.

[0025] Please see Figure 4 The specific steps of S3 are as follows: S311: During the data collection and verification period, the instrument reads the volumetric flow rate, extracts the actual operating pressure value and actual operating temperature value from the effective gas turbine operating condition feature sequence, obtains the standard state absolute pressure constant and the standard state absolute temperature constant, and divides the actual operating pressure value by the standard state absolute pressure constant to obtain the gas turbine intake absolute pressure ratio. According to the data acquisition node plan, within the data verification time, the absolute volumetric flow rate representing the natural gas flow is read from the on-site measurement record list. Simultaneously, the actual operating pressure value and the actual operating temperature value representing the current thermodynamic environment are extracted from the previously established effective gas turbine operating condition characteristic sequence. A request is initiated to the preset configuration set to obtain the legally mandated standard state absolute pressure constant and the legally mandated standard state absolute temperature constant. Both the standard state absolute pressure constant and the standard state absolute temperature constant are fixed reference benchmark values ​​explicitly mandated by the National Energy Measurement Verification Regulations. After preparing all the parameters involved in the calculation, a division operation is directly performed between the actual operating pressure value and the standard state absolute pressure constant. The specific execution logic of the division operation stipulates that the actual operating pressure value extracted from the effective gas turbine operating condition characteristic sequence is set as the value to be divided, representing the actual measurement result; the extracted standard state absolute pressure constant is set as the value to be divided, representing the baseline reference value; and the actual operating pressure value is divided by the standard state absolute pressure constant to obtain a pure numerical result of an absolute ratio stripped of the original pressure physical dimension. If the result of the division operation is greater than one, it means that the actual gas flow pressure in the pipeline is in a positive deviation from the expansion state; if the result is less than one, it means that the actual gas flow pressure in the pipeline is in a negative deviation from the contraction state. The final result of the division operation is extracted and saved, and the absolute pressure ratio of the gas turbine intake is directly defined.

[0026] S312: Divide the absolute temperature constant under standard conditions with the actual operating temperature value to obtain the ratio of the absolute temperature of the gas turbine intake air, and multiply the ratio of the absolute pressure of the gas turbine intake air with the ratio of the absolute temperature of the gas turbine intake air to obtain the state transition multiplication factor. The process involves a second proportional conversion and division operation between the standard state absolute temperature constant obtained from the preset configuration set and the previously extracted actual operating temperature value. The specific logic of this second division operation is completely opposite to that of the pressure dimension division operation. The second round of division requires explicitly using the standard state absolute temperature constant, representing the legally mandated reference baseline, as the core value to be divided, and explicitly using the actual operating temperature value, reflecting the true thermodynamic environment on site, as the core value to be divided. The standard state absolute temperature constant is divided by the actual operating temperature value to obtain the gas turbine intake absolute temperature ratio specifically used to correct for physical deviations in thermal expansion and contraction. The gas turbine intake absolute pressure ratio, obtained just from the pressure division operation, is extracted and combined with the newly calculated gas turbine intake absolute temperature ratio for direct numerical multiplication. Specifically, the gas turbine intake absolute pressure ratio is multiplied by the gas turbine intake absolute temperature ratio, and the pressure and temperature influence correction coefficients, which are currently independent, are merged into a unified comprehensive multiplier to adjust the absolute factor through a weighted multiplication method. By multiplying the two ratios, the relative offset of the two isolated dimensions is transformed into a unified scale absolute conversion scalar with global adjustment capabilities. This multiplication deduction yields the state transformation multiplication factor, a core indicator to be applied to the subsequent original fluid volume correction calculation.

[0027] S313: Based on the state transition multiplication factor and the volumetric flow rate read by the instrument, perform a numerical product conversion operation to obtain the standard state natural gas flow rate product value. Then, serialize and bind the standard state natural gas flow rate product value with the data verification time period to establish standard state natural gas flow rate data. Based on the state transition multiplication factor calculated using both absolute temperature and absolute pressure environmental factors, and the volumetric flow rate value obtained in real-time from the on-site measurement record list, a product conversion calculation to restore the true flow scale of the fluid is immediately executed. The product conversion calculation process requires that the state transition multiplication factor, including the basic environmental deviation ratio, be used as an independent multiplier, directly multiplied by the volumetric flow rate value representing the on-site apparent measurement scale. By performing the corresponding multiplication calculation, all original on-site volume readings that are easily affected by external random physical environmental factors and exhibit expansion and contraction deviations are forcibly converted and folded back to the legally mandated ideal physical environment conditions. After rigorous volume dimension correction and conversion, a standard state natural gas flow rate product value that objectively reflects the true material flow scale of natural gas is obtained. The stamp code responsible for identifying the specific time node of the data verification period is extracted, and the newly calculated standard state natural gas flow rate product value is linked to the time identification data verification period using a low-level serialization binding operation. The serialization binding and association operation involves merging and compressing the product value of the standard state natural gas flow rate along with the data verification time period to generate an independent message data packet with time traceability and verification capabilities. This independent message data packet is then pushed to the time-series data table for persistent disk storage. After disk storage is completed, a time-series state data set of standard state natural gas flow rate data with verification and benchmarking value is established.

[0028] There are issues with multi-point statistics for natural gas volume. The accuracy of statistical instruments varies due to differences in flow meter readings at different metering points (such as inlet, outlet, and combustion chamber), cumulative metering instrument errors, and the influence of environmental factors. Additional calculations are required every second, as shown in the formula below.

[0029]

[0030] Volume under standard conditions (101.325 kPa, 20℃); The instrument reads the volumetric flow rate; Real-time absolute pressure (unit: kPa); Real-time absolute temperature (unit: K, i.e., °C + 273.15); 101.325 kPa; 293.15K (i.e., 20 degrees Celsius); The natural gas meter readings are periodically calibrated against the calculated results (daily), and the data is entered into the system. The percentage difference is calculated and used as an adjustment factor. On days when the difference exceeds 2% (the set value), the reporting AI will remind maintenance personnel to conduct a manual review in the report analysis results.

[0031] Please see Figure 5 The specific steps of S4 are as follows: S411: Collect the readings of the natural gas settlement form issued by the upper-level circulation gateway within the second sampling time period, perform definite integral surface accumulation and addition operations on the natural gas flow data under standard conditions within the time nodes of the second sampling time period, calculate the total sum of overall flow consumption within the time interval, and establish the time integral cumulative value; The system collects and receives natural gas settlement form readings from the upper-level transfer gateway via a dedicated data link within the second sampling period. It retrieves all discrete standard-state natural gas flow data within the second sampling period from the corresponding time-series data table. For each subdivided time interval within the second sampling period, it performs a definite integral surface accumulation and addition operation, akin to calculus. This operation involves directly adding the standard-state natural gas flow data corresponding to each of two consecutive time nodes and dividing by two to obtain the absolute value of the average instantaneous flow. Then, it multiplies this absolute value by the absolute time interval between the two consecutive time nodes to obtain a micro-element of natural gas consumption within a local, minute time period. Finally, it sequentially adds these micro-element values ​​for each of the minute time intervals covered by the second sampling period. Finally, it sums up the consumption of all these micro-segments to calculate the total estimated flow consumption within the entire second sampling period. The total traffic consumption figure is calculated and stored in a memory cache variable that is responsible for storing cumulative measurement results. This is done by directly assigning values ​​to variables. This establishes a core calculation reference value based on the cumulative time integral, which is used to cross-reference and verify the data with external forms.

[0032] S412: Based on the cumulative value of time integration and the reading of the natural gas settlement form, perform difference subtraction to obtain the difference in natural gas consumption settlement volume. Divide the difference in natural gas consumption settlement volume by the reading of the natural gas settlement form to obtain the percentage deviation rate of natural gas consumption. The system extracts the cumulative time integral value derived from tiny slices of time-series integration and the reading from the natural gas settlement form directly received from the external upper-level gateway. Immediately, it performs a simple difference subtraction operation on these two physically similar volume indicators from drastically different data sources. The specific steps of this difference subtraction operation are as follows: using the natural gas settlement form reading, representing the authoritative basis for trade settlement, as the base value for the subtraction operation, and using the cumulative time integral value derived autonomously by the underlying computing power as the specific subtraction value, subtracting the cumulative time integral value from the natural gas settlement form reading. If the resulting difference is not equal to zero, whether in a positive or negative range, the system directly obtains the absolute deviation gap in natural gas consumption settlement volume between the official summary bill and the calculated result. Subsequently, a relative proportional conversion division operation is performed independently on the natural gas consumption settlement volume difference. Specifically, the absolute quantity difference in natural gas consumption settlement volume difference is directly divided by the absolute scale of the natural gas settlement form reading representing the official benchmark. Dividing the absolute deviation gap by the overall statistical scale transforms the absolute error, which has physical measurement dimensions, into a pure relative ratio without any specific physical measurement dimensions, and calculates the percentage result. After converting the pure relative ratio percentage result obtained by division into a standardized percentage display format, the unified standard metric for measuring the percentage deviation rate of natural gas consumption is obtained.

[0033] S413: Based on the natural gas consumption percentage deviation rate and the set allowable error constant, extract the gas turbine intelligent agent metering error alarm features that exceed the set allowable error constant range, and combine them with the time integral cumulative value to generate gas turbine intelligent agent error alarm and cumulative flow data; The calculated percentage deviation rate of natural gas consumption is compared with the pre-set allowable error constant within the safety risk control rules to determine if it exceeds the limit. The allowable error constant is specifically and explicitly limited to a fixed value of 0.5%. This 0.5% value is determined by extracting and analyzing the extreme values ​​of historical routine operation tolerance baseline data from all qualified trade settlement metering instruments certified by national specialized verification institutions over the past three consecutive years. The resulting percentage deviation rate of natural gas consumption is then compared with the fixed allowable error constant of 0.5%. If the percentage deviation rate exceeds the absolute limit of 0.5%, it indicates a very high risk of data distortion in the current billing process, immediately triggering the underlying preset safety warning rules to extract alarm features specifically for gas turbine intelligent agent metering error alerts that exceed the set allowable error constant range. The extracted intelligent agent metering error alarm feature's unique string code is then directly concatenated and packaged with the accumulated value of the actual consumption time obtained from previous calculus deductions. The connection and combination operation steps are as follows: the exclusive alarm character information prefix is ​​concatenated to the first part of the time integral cumulative value number string, and then combined and encapsulated into a comprehensive alarm communication message data object, generating dual key prevention information including abnormal qualitative judgment and specific loss absolute value, gas turbine intelligent agent error alarm and cumulative flow data.

[0034] Please see Figure 6 The specific steps of S5 are as follows: S511: Extract the data associated with the field mapping of the intelligent agent report, perform node traversal search operation, parse and read the internal structure hierarchical attribute information of the data, obtain the target position index positioning point within the corresponding text paragraph interval based on the parsing result, and establish a set of document target position index coordinate information. Upon entering the report generation and presentation stage, the system extracts the intelligent proxy report field mapping and related data that was properly written and saved in the data persistence collection during the previous information processing phase. A full business node traversal and search operation is then performed on the intelligent proxy report field mapping and related data. The specific derivation process of the node traversal and search operation involves reading the mapping identity binding information record by record, starting from the first row and first column of the relational data table. The system then calls the document outline's specific hierarchical parsing model to deeply analyze and extract the internal structural hierarchical attribute information of the data. This internal structural hierarchical attribute information covers key page display identifiers such as the hierarchical relationship between main and subheadings, the paragraph layout order, and even the table's vertical and horizontal layout. Based on the extracted outline parsing results, the system further obtains the target position index point with absolute page layout coordinates within the corresponding text's specific presentation paragraph range within the electronic document layout logic architecture. The target position index point precisely represents the actual coordinate position of the absolute intersection and overlap of a row and column within the document display page. The search process retrieves the true coordinates of all landing points, which are then collected, archived, and stored in a single in-memory data dictionary. This creates a comprehensive set of index coordinates for the target locations in the navigation map document, covering all pre-filled slots. This set of index coordinates provides the typesetting and form-filling program with unbiased coordinate guidance.

[0035] S512: Obtain the document node structure, provide positioning parameters based on the document target location index coordinate information set, perform node text insertion operation on the gas turbine intelligent agent error alarm and cumulative flow data according to the positioning parameters, fill in the blank node positions of the document node structure, and obtain the document object sequence set; The system extracts a panoramic tree diagram of the blank document node structure, based on the standard specifications for data entry and processing. It aggregates all data from the target location index coordinate information set gathered in the previous step, providing strictly defined horizontal and vertical coordinate positioning parameters. The core data to be submitted, including gas turbine intelligent agent error alarms and cumulative flow data, is broken down into discrete, independent text fragments. Following the precise direction guided by the extracted positioning parameters, the system performs targeted and accurate node text insertion. The core process of node text insertion involves sending text fragments representing alarm / warning characters to the blank areas of the text box specifically designated by the positioning parameters, and sending text fragments representing the absolute scale of flow volume to the corresponding blank areas of the table's rows and columns. Once all extracted independent text fragments are accurately placed into their designated blank node positions within the document node structure, all previously empty placeholders awaiting data entry are completely filled and closed by real, operational business data, resulting in a richer and more complete layout. After completing all batch loop insertion, filling, and generation actions, each data object element carrying a specific entity is relocked, encapsulated, packaged, and merged to finally obtain a collection of document object sequences representing specific formatted paragraphs that have completed real data updates and upgrades.

[0036] S513: Perform page order connection and text paragraph concatenation for each independent document object element in the document object sequence set, merge the text content of all document object elements, construct a global file output format object, and generate gas turbine operation and maintenance report data; For each independent document object element in the generated document object sequence set that has not yet been connected, the system immediately performs standardized page order connection and text paragraph concatenation operations. The page order connection operation reads the document object elements carrying their original outline and hierarchical numbering order, placing the document object element representing the cover title at the beginning of the layout memory display, and placing the document object element representing core data charts immediately after the cover title document object element in the main body position. The text paragraph concatenation operation forcibly writes page breaks, line breaks, and indentation control characters at the beginning and end of each independent document object element, guiding the seamless merging and interweaving of all scattered and separate document object element text content into a tightly structured and coherent overall reading text flow. After all text content is concatenated and aggregated, a dedicated digital file encoding rendering output generation algorithm is invoked. This algorithm compiles the underlying instructions to uniformly format and reconstruct the coherent overall reading text flow according to internationally accepted electronic portable document fixed typesetting standards, successfully constructing a global file output format object that fully meets the hard requirements of external reading environments. Completing the global format object construction means that the layout file is written to the storage medium and finally successfully generates a complete aggregated on-site business status gas turbine operation and maintenance report data.

[0037] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for generating gas turbine operation and maintenance reports based on intelligent agents, characterized in that, Includes the following steps: S1: During the first sampling time period of gas turbine operation, acquire the real-time temperature parameters and real-time pressure parameters of the gas turbine at the gas turbine inlet, extract the state deviation features, and generate an effective gas turbine operating condition feature sequence. S2: Extract the set of text placeholder identifiers from the gas turbine operation and maintenance report template document and the parameter name tags from the effective gas turbine operating condition feature sequence, determine the maximum matching association relationship, and generate intelligent proxy report field mapping association data; S3: Collect the volumetric flow rate read from the instrument of the gas turbine monitoring equipment, extract the actual operating pressure parameters and actual operating temperature parameters at the corresponding time in the effective gas turbine operating condition characteristic sequence, perform natural gas state conversion to obtain standard state flow characteristics, and generate standard state natural gas flow data; S4: During the second sampling time period of gas turbine operation and maintenance, obtain the natural gas settlement form reading, extract the cumulative time value of the standard state natural gas flow data during the second sampling time period, extract the gas turbine intelligent agent metering error alarm feature, and concatenate it with the cumulative time integral value to generate gas turbine intelligent agent error alarm and cumulative flow data; S5: Extract the target location index coordinate information based on the data mapping association of the intelligent agent report fields, fill the gas turbine intelligent agent error alarm and cumulative flow data and the effective gas turbine operating condition feature sequence into the preset document node structure based on the target location index coordinate information, splice the document object sequence, and output the gas turbine operation and maintenance report data.

2. The method for generating gas turbine operation and maintenance reports based on intelligent agents according to claim 1, characterized in that: The effective gas turbine operating condition feature sequence includes an operating condition baseline vector, an effective state identifier matrix, and an anomaly filtering flag; the intelligent agent report field mapping association data includes a field matching weight dictionary, an agent binding index table, and a parsing rule encoding set; the standard state natural gas flow data includes standard condition instantaneous volume, gaseous compensation coefficient, and reference condition correction value; the gas turbine intelligent agent error alarm and cumulative flow data includes out-of-tolerance threshold trigger records, gas consumption loss measurement values, and periodic settlement gas volume details; the gas turbine operation and maintenance report data includes an operation and maintenance summary summary, energy efficiency calculation charts, and an appendix of agent diagnostic conclusions.

3. The method for generating gas turbine operation and maintenance reports based on intelligent agents according to claim 1, characterized in that: In the process of determining the maximum matching association, the set of text placeholder identifiers is semantically similar to the parameter name labels to determine the label relationship with the greatest semantic similarity.

4. The method for generating gas turbine operation and maintenance reports based on intelligent agents according to claim 1, characterized in that: During the process of extracting the intelligent agent metering error alarm features of the gas turbine, the consumption deviation is assessed based on the reading of the natural gas settlement form and the cumulative value of time integration.

5. The method for generating gas turbine operation and maintenance reports based on intelligent agents according to claim 1, characterized in that, The specific steps for obtaining the effective gas turbine operating condition feature sequence are as follows: S111: Collect the real-time temperature parameters and real-time pressure parameters of the gas turbine at the gas turbine inlet during the first sampling time period, extract the actual observed pressure value and the prior estimate of pressure from the real-time pressure parameters of the gas turbine, extract the actual observed temperature value and the prior estimate of temperature from the real-time temperature parameters of the gas turbine and perform difference calculation to generate the gas turbine operating state temperature deviation. S112: Based on the temperature deviation of the gas turbine operating state, perform state difference calculation on the actual observed pressure value and the prior estimated pressure value to obtain the pressure deviation value, and perform dot product calculation with the preset state prediction covariance matrix to obtain the pressure deviation characteristics of the gas turbine operating state. S113: Combine the temperature deviation of the gas turbine operating state and the pressure deviation of the gas turbine operating state to perform a numerical update operation of the covariance matrix. Then, perform numerical traversal and serial sorting and recombination operations on the updated state prediction covariance matrix to establish an effective gas turbine operating condition feature sequence.

6. The method for generating gas turbine operation and maintenance reports based on intelligent agents according to claim 1, characterized in that, The specific steps for obtaining the data related to the field mapping of the intelligent agent report are as follows: S211: Extract the set of text placeholder identifiers in the template document and the parameter name labels in the effective gas turbine operating condition feature sequence, obtain the multi-dimensional word vector encoding, extract the baseline semantic vector coordinates in the set of text placeholder identifiers, extract the vector coordinates for the multi-dimensional word vector encoding, and establish the feature semantic vector coordinates. S212: Prioritize the reference semantic vector coordinates, perform a spatial dot product and modulus product division operation on the reference semantic vector coordinates and the feature semantic vector coordinates to obtain all mapping space cosine similarity values, and traverse and filter them in descending order to obtain the maximum value of mapping space cosine similarity. S213: Determine the semantic matching degree between the maximum value of the cosine similarity of the mapping space and the elements of the text placeholder identifier set, and map and bind the placeholder identifier corresponding to the maximum value of the cosine similarity of the mapping space with the maximum value to the parameter name label to generate intelligent proxy report field mapping association data.

7. The method for generating gas turbine operation and maintenance reports based on intelligent agents according to claim 1, characterized in that, The specific steps for obtaining the standard state natural gas flow data are as follows: S311: During the data collection and verification period, the instrument reads the volumetric flow rate, extracts the actual operating pressure value and actual operating temperature value from the effective gas turbine operating condition feature sequence, obtains the standard state absolute pressure constant and the standard state absolute temperature constant, and divides the actual operating pressure value by the standard state absolute pressure constant to obtain the gas turbine intake absolute pressure ratio. S312: Divide the absolute temperature constant under standard conditions with the actual operating temperature value to obtain the ratio of the absolute temperature of the gas turbine intake air, and multiply the ratio of the absolute pressure of the gas turbine intake air with the ratio of the absolute temperature of the gas turbine intake air to obtain the state transition multiplication factor. S313: Based on the state transition multiplication factor and the volumetric flow rate read by the instrument, perform a numerical product conversion operation to obtain the standard state natural gas flow rate product value. Then, serialize and bind the standard state natural gas flow rate product value with the data verification time period to establish standard state natural gas flow rate data.

8. The method for generating gas turbine operation and maintenance reports based on intelligent agents according to claim 1, characterized in that, The specific steps for obtaining the gas turbine intelligent agent error alarm and cumulative flow data are as follows: S411: Collect the readings of the natural gas settlement form issued by the upper-level circulation gateway within the second sampling time period, perform definite integral surface accumulation and addition operations on the natural gas flow data under the standard state within the time node of the second sampling time period, calculate the total sum of the overall flow consumption within the time interval, and establish the time integral cumulative value; S412: Based on the cumulative value of time integration and the reading of the natural gas settlement form, perform difference subtraction to obtain the difference in natural gas consumption settlement volume, divide the difference in natural gas consumption settlement volume by the reading of the natural gas settlement form to obtain the percentage deviation rate of natural gas consumption; S413: Based on the natural gas consumption percentage deviation rate and the set allowable error constant, extract the gas turbine intelligent agent metering error alarm features that exceed the set allowable error constant range, and combine them with the time integral cumulative value to generate gas turbine intelligent agent error alarm and cumulative flow data.

9. The method for generating gas turbine operation and maintenance reports based on intelligent agents according to claim 1, characterized in that, The specific steps for obtaining the gas turbine operation and maintenance report data are as follows: S511: Extract the field mapping associated data of the intelligent agent report, perform node traversal search operation, parse and read the internal structure hierarchical attribute information of the data, obtain the target position index positioning point in the corresponding text paragraph interval based on the parsing result, and establish a set of document target position index coordinate information. S512: Obtain the document node structure, provide positioning parameters according to the document target location index coordinate information set, perform node text insertion operation on the gas turbine intelligent agent error alarm and cumulative flow data according to the positioning parameters, fill in the blank node positions of the document node structure, and obtain the document object sequence set; S513: Perform page sequence connection and text paragraph concatenation for each independent document object element in the document object sequence set, merge the text content of all document object elements, construct a global file output format object, and generate gas turbine operation and maintenance report data.

10. The method for generating gas turbine operation and maintenance reports based on intelligent agents according to claim 6, characterized in that, When obtaining the multidimensional word vector encoding, natural language processing multidimensional vector transformation operation is performed on the parameter name label.

Images