A method and system for restating log reports based on carbon emission monitoring

Abstract

This invention provides a method and system for restating log reports based on carbon emission monitoring. First, it acquires raw carbon emission monitoring log data consisting of multiple monitoring entries and timestamp sequences. Each entry includes fields such as emission source identifiers. Next, it extracts features from the raw carbon emission monitoring log data to obtain a structured log feature group containing core emission parameter features and cross-entry temporal correlation features. Then, it calls a pre-built model to semantically reconstruct the feature group, generating a log restatement logical framework containing causal relationship chains of emission events and key information distribution maps. Based on this log restatement logical framework, it generates an initial restatement log report. Finally, it performs semantic consistency and structural integrity checks on the initial restatement log report to generate a final restatement log report. This method can efficiently process carbon emission monitoring log data and generate accurate, clear, and logically sound reports.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and more specifically, to a method and system for restating log reports based on carbon emission monitoring. Background Technology

[0002] In today's context of actively addressing climate change, carbon emission monitoring and management have become crucial. Enterprises, institutions, and other carbon emitters need to accurately and clearly record and report their carbon emissions to meet regulatory requirements, optimize their own carbon emission management, and demonstrate environmental responsibility to the public.

[0003] Currently, carbon emission monitoring typically generates a large amount of raw log data, which exists in the form of multiple continuously collected monitoring entries. Each entry includes information such as emission source identification, monitoring parameter descriptions, and status change records. However, existing log processing methods often simply store and perform basic queries on this raw carbon emission monitoring log data, lacking effective integration and analysis tools. Due to the large volume and complex format of raw log data, manual reading and analysis of this data is not only inefficient but also prone to errors, making it difficult to quickly and accurately grasp the overall situation, causal relationships, and key information of carbon emission events. Furthermore, the lack of scientific and reasonable methods in generating log reports leads to potential problems such as incomplete information and unclear logic in the report content. Summary of the Invention

[0004] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a log report restatement method based on carbon emission monitoring, the method comprising:

[0005] Obtain raw carbon emission monitoring log data containing timestamp sequences. The raw carbon emission monitoring log data consists of multiple continuously collected monitoring entries. Each monitoring entry includes an emission source identification field, a monitoring parameter description field, and a status change record field.

[0006] Feature extraction is performed on the original carbon emission monitoring log data to obtain a structured log feature group, which includes the core emission parameter features of each monitoring item and cross-item time-series correlation features;

[0007] The pre-built log report paraphrase logic model is invoked to perform semantic reconstruction processing on the structured log feature group, generating a log paraphrase logic framework, which includes a causal relationship chain of emission events and a distribution map of key information.

[0008] Based on the log paraphrasing logic framework, a text generation operation is performed to generate an initial paraphrased log report, which includes a description of emission events organized according to a causal chain.

[0009] The initial paraphrase log report is subjected to two-dimensional verification processing to generate a final paraphrase log report that passes the verification. The two-dimensional verification processing includes semantic consistency verification and structural integrity verification.

[0010] In another aspect, embodiments of the present invention also provide a log report recap system based on carbon emission monitoring, including a processor and a machine-readable storage medium connected to the processor. The machine-readable storage medium is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the machine-readable storage medium to implement the above-described method.

[0011] Based on the above, this embodiment of the invention acquires raw carbon emission monitoring log data containing timestamp sequences and extracts features from it to obtain a structured log feature group containing core emission parameter features and cross-entry time-series correlation features. This achieves effective integration and in-depth mining of the raw log data. A pre-built log report paraphrasing logic model is then invoked to perform semantic reconstruction processing on the structured log feature group, generating a log paraphrasing logic framework containing causal relationship chains of emission events and key information distribution maps. This clearly presents the logical relationships and key information between carbon emission events. An initial paraphrased log report is generated based on this log paraphrasing logic framework, and the emission event descriptions are organized according to causal relationship chains, greatly improving the readability and logic of the report. Finally, the initial paraphrased log report undergoes dual-dimensional verification processing to ensure semantic consistency and structural integrity, ultimately generating an accurate, clear, and reliable final paraphrased log report. Attached Figure Description

[0012] Figure 1 This is a schematic diagram of the execution flow of the log report restatement method based on carbon emission monitoring provided in an embodiment of the present invention.

[0013] Figure 2 This is a schematic diagram of exemplary hardware and software components of the log report recap system based on carbon emission monitoring provided in an embodiment of the present invention. Detailed Implementation

[0014] The present invention will now be described in detail with reference to the accompanying drawings. Figure 1 This is a flowchart illustrating a log report restatement method based on carbon emission monitoring, provided in one embodiment of the present invention. The following is a detailed description of this log report restatement method based on carbon emission monitoring.

[0015] Step S110: Obtain raw carbon emission monitoring log data containing timestamp sequences. The raw carbon emission monitoring log data consists of multiple continuously collected monitoring entries. Each monitoring entry includes an emission source identification field, a monitoring parameter description field, and a status change record field.

[0016] In a comprehensive energy park, there are various types of carbon emission sources, such as power plants, factory production equipment, and transport vehicles within the park. To comprehensively and accurately grasp the carbon emission situation within the park, multiple monitoring devices need to be deployed. These devices continuously collect carbon emission-related data, thereby forming raw carbon emission monitoring log data.

[0017] In this embodiment, the timestamp sequence records the collection time of each monitoring data point. The timestamps can be accurate to a specific moment, reflecting the chronological order of carbon emission data.

[0018] Raw carbon emission monitoring log data consists of multiple continuously collected monitoring entries. Each entry contains specific information, including an emission source identification field that identifies the specific emission source from which the data originates. For example, different generator sets at a power plant can be identified by specific numbers; for transport vehicles, license plate numbers or vehicle identification numbers can be used for differentiation. The monitoring parameter description field details various parameters related to carbon emissions, covering the composition of emitted gases, such as the content of carbon dioxide and methane, as well as emission flow rate and pressure. The status change record field records the status changes of the emission source at the time of data collection, such as whether the generator sets at a power plant are in a startup, running, or shutdown state, or whether the factory production equipment is operating normally or experiencing a malfunction.

[0019] The monitoring equipment collects data at fixed time intervals and transmits the collected data to the data center for storage and processing in a timely manner. The data center performs preliminary sorting and classification of this data to ensure the integrity and accuracy of the original carbon emission monitoring log data.

[0020] Step S120: Extract features from the original carbon emission monitoring log data to obtain a structured log feature group. The structured log feature group includes the core emission parameter features of each monitoring item and cross-item time-series correlation features.

[0021] After obtaining the raw carbon emission monitoring log data, due to its complex data structure and the presence of a large amount of redundant information, feature extraction is required to obtain a structured log feature set for better subsequent analysis and processing. The structured log feature set comprises two important parts: the core emission parameter features for each monitoring entry and the cross-entry time-series correlation features. The core emission parameter features reflect the key information directly related to carbon emissions in each monitoring entry, while the cross-entry time-series correlation features demonstrate the temporal relationships between different monitoring entries.

[0022] Step S121: Perform field parsing processing on the original carbon emission monitoring log data, and extract the emission source identifier field value, monitoring parameter description field value, and status change record field value for each monitoring entry.

[0023] This embodiment requires field parsing processing of the raw carbon emission monitoring log data. Since the raw carbon emission monitoring log data is stored in a relatively complex format, and each field contains different types of information, it is necessary to extract the emission source identifier field value, monitoring parameter description field value, and status change record field value for each monitoring entry.

[0024] For example, the raw carbon emission monitoring log data can be scanned and analyzed line by line based on its storage format and rules. For the emission source identification field, specific identifiers or coding rules can be identified to accurately extract the specific identifier of the emission source. For instance, if the emission source identification uses a numbering system, the corresponding number can be extracted by finding the defined character position or coding rule. For the monitoring parameter description field, various parameter information can be extracted according to a predefined parameter description format. For instance, if the parameter description uses the format "parameter name: parameter value", the parameter name and corresponding parameter value can be extracted by separating them with a colon (:). For the status change record field, the status change information of the emission source can be extracted based on the defined vocabulary and syntax rules of the status description. For instance, if the status description uses the format "status name + change description", accurate status change information can be extracted by identifying keywords in the status name and change description.

[0025] The extracted emission source identifier field value, monitoring parameter description field value, and status change record field value will be stored in a temporary data structure.

[0026] Step S122: Perform keyword filtering on the monitoring parameter description field value to select parameter words directly related to carbon emissions as core parameter candidate words. The core parameter candidate words include emission type limiting words, measurement unit words, and threshold descriptive words.

[0027] After extracting the values ​​from the monitoring parameter description field, it may contain a large number of terms that are both related to and unrelated to carbon emissions. To filter out the parameter terms that are truly directly related to carbon emissions, keyword filtering of the monitoring parameter description field values ​​is necessary.

[0028] Keyword filtering uses a predefined list of keywords containing various parameter terms directly related to carbon emissions. These parameter terms can be categorized into emission type qualifiers, measurement unit terms, and threshold descriptors. Emission type qualifiers specify the exact type of emission, such as carbon dioxide emissions or methane emissions; measurement unit terms indicate the unit of measurement, such as cubic meters or kilograms; and threshold descriptors describe the threshold range of the parameter, such as the upper and lower emission thresholds.

[0029] Filtering can examine each word in the monitoring parameter description field value one by one to determine if it is in the keyword list. If it is in the keyword list, it is retained as a core parameter candidate word; otherwise, it is filtered out. For example, for the monitoring parameter description field value "Equipment temperature: 25℃, carbon dioxide emission flow rate: 10 cubic meters / hour", keyword filtering can retain "carbon dioxide" and "cubic meters / hour" as core parameter candidate words, while filtering out "equipment temperature" and "25℃".

[0030] The selected core parameter candidate words will be compiled into a new list for further analysis and processing.

[0031] Step S123: Perform semantic strength analysis on the state change record field value, calculate the degree adverb weight value of the state change description, sort the core parameter candidate words by importance based on the degree adverb weight value, and generate core emission parameter features containing weight information.

[0032] After obtaining the core parameter candidate words, semantic strength analysis is needed to further determine the importance of these parameters on the state change record field values. The state change record field values ​​may contain some degree adverbs, such as "significant," "slight," and "drastic," which can reflect the intensity of the state change.

[0033] Semantic strength analysis uses a predefined weight table for degree adverbs, which assigns a corresponding weight value to each degree adverb. The magnitude of the weight value is positively correlated with the significance of the state change; that is, the more significant the state change, the larger the corresponding degree adverb weight value.

[0034] For each degree adverb in the state change record field value, its corresponding weight value can be found in the degree adverb weight table. Then, the core parameter candidate words are sorted by importance based on these weight values. The principle of sorting is that the more closely related the core parameter candidate word is to the degree adverb, the higher its importance. For example, if the state change record field value is "Equipment operating status changed significantly, carbon dioxide emissions increased sharply", where "significantly" and "sharply" have high weight values, then the core parameter candidate word "carbon dioxide emissions" associated with these two degree adverbs will be relatively more important.

[0035] By ranking based on importance, a corresponding weight value can be assigned to each candidate term for core parameters, thereby generating core emission parameter features that include weighted information. This weighted information reflects the importance of each core parameter in carbon emission monitoring.

[0036] Step S1231: Establish a vocabulary library of degree adverbs describing the state of carbon emissions, wherein the vocabulary library contains a set of words representing the magnitude of change.

[0037] To perform semantic intensity analysis, a lexicon of adverbs describing the degree of carbon emission status needs to be established. This lexicon is a set of words that represent the magnitude of change, accurately describing the intensity of changes in the emission source status.

[0038] When building a database of degree adverbs, a large amount of text data related to carbon emissions can be analyzed and organized to collect degree adverbs that indicate the magnitude of change. These degree adverbs can be divided into different levels, such as "slight" and "absolutely" to indicate slight changes, "obvious" and "significant" to indicate moderate changes, and "drastic" and "significant" to indicate drastic changes.

[0039] The degree adverb library is stored in a database or data file for later querying and use. During semantic strength analysis, the corresponding degree adverb can be retrieved by searching for it in the library.

[0040] Step S1232: Perform adverb recognition processing on the value of the state change record field, and extract adverb words that match the degree adverb library as state degree adverbs.

[0041] After establishing the degree adverb library, adverb recognition processing is required on the state change record field values. The purpose of adverb recognition processing is to extract adverb words that match the degree adverb library from the state change record field values. These adverb words will be used as state degree adverbs for subsequent semantic strength analysis.

[0042] Each word in the status change record field value can be checked individually to determine if it is in the degree adverb library. If it is in the degree adverb library, it will be extracted as a status degree adverb. For example, for the status change record field value "equipment operating status has changed significantly", the word "significantly" is in the degree adverb library, so it will be extracted as a status degree adverb.

[0043] The extracted state degree adverbs will be stored in a temporary data structure for further processing.

[0044] Step S1233: Assign a preset weight coefficient value to each state degree adverb, wherein the weight coefficient value is positively correlated with the significance of the state change.

[0045] After extracting the adverbs of state degree, each adverb of state degree needs to be assigned a preset weight coefficient value. These weight coefficient values ​​are preset based on the significance of the state change, and the magnitude of the weight coefficient value is positively correlated with the significance of the state change.

[0046] The preset weight coefficient values ​​are stored in a weight table, which corresponds to the degree adverb library. For each degree adverb, its corresponding weight coefficient value can be found in the weight table. For example, the degree adverb "significant" may have a relatively high weight coefficient value, while the degree adverb "slight" may have a relatively low weight coefficient value.

[0047] By assigning weight coefficients to adverbs of state, the significance of state changes can be quantified.

[0048] Step S1234: Count the number of co-occurrences of each core parameter candidate word and the adverb of state degree, and calculate the co-occurrence frequency value of each core parameter candidate word.

[0049] After assigning weight coefficients to the adverbs of state degree, it is necessary to count the number of times each core parameter candidate word and the adverb of state degree co-occur. The number of co-occurrences refers to the number of times a core parameter candidate word and a certain adverb of state degree appear simultaneously in the state change record field value.

[0050] The method for counting co-occurrences involves scanning each row of the state change record field values. For each core parameter candidate word, it is checked whether it co-occurs with a certain degree adverb. If they co-occur, the corresponding co-occurrence count is incremented by 1.

[0051] After counting co-occurrences, the co-occurrence frequency (COF) of each core parameter candidate word can be calculated. The COF is the ratio of the number of co-occurrences to the total number of occurrences of the core parameter candidate word. The COF reflects the degree of association between the core parameter candidate word and the adverb of state / degree; a higher COF indicates a stronger association between the core parameter candidate word and the state change.

[0052] Step S1235: Multiply the co-occurrence frequency value by the weight coefficient value of the corresponding state degree adverb to obtain the importance score value of each core parameter candidate word.

[0053] After obtaining the co-occurrence frequency value and the weight coefficient value of the state degree adverb, the importance score value of each core parameter candidate word is obtained by multiplying the co-occurrence frequency value with the weight coefficient value of the corresponding state degree adverb.

[0054] The importance score comprehensively considers the closeness of the association between core parameter candidate words and state degree adverbs, as well as the significance of state changes. By multiplying the co-occurrence frequency value by the weight coefficient value, the importance of core parameter candidate words that are closely associated with state changes and whose state changes are significant can be highlighted.

[0055] Step S1236: Sort the candidate words of the core parameters in descending order according to the importance score to generate a core parameter sequence sorted by importance.

[0056] After obtaining the importance score for each core parameter candidate word, the core parameter candidate words need to be sorted in descending order based on these scores. The purpose of descending order is to prioritize the core parameter candidate words with higher importance for subsequent analysis and processing.

[0057] Candidate words for core parameters can be sorted according to their importance scores, with higher-scoring candidates appearing first and lower-scoring candidates appearing last. By sorting in descending order, a sequence of core parameters ordered by importance can be generated.

[0058] Step S1237: Assign a corresponding importance score value to each parameter in the core parameter sequence to generate core emission parameter features containing weight information.

[0059] After generating a sequence of core parameters ordered by importance, a corresponding importance score is assigned to each parameter, thereby generating core emission parameter features containing weight information.

[0060] This weighting information accurately reflects the importance of each core parameter in carbon emission monitoring. For example, when making carbon emission forecasts, this weighting information can be used to focus on the more important core parameters, thereby improving the accuracy of the forecasts.

[0061] Step S124: Perform interval calculation processing on the timestamp sequences of adjacent monitoring items to obtain the time interval value of adjacent monitoring items; perform similarity calculation processing on the core emission parameter features of adjacent monitoring items to obtain the parameter change similarity value of adjacent monitoring items; construct a cross-item time-series association model based on the time interval value and the parameter change similarity value, calculate the time-series association strength value of adjacent monitoring items through the cross-item time-series association model, and generate cross-item time-series association features containing association strength information.

[0062] After obtaining the core emission parameter characteristics containing weighted information, in order to analyze the temporal correlation between different monitoring items, it is necessary to perform interval calculation processing on the timestamp sequences of adjacent monitoring items. The timestamp sequence records the collection time of each monitoring item. By calculating the timestamp difference between adjacent monitoring items, the time interval value between adjacent monitoring items can be obtained. The time interval value can reflect the time span of adjacent monitoring data collection.

[0063] Simultaneously, it is necessary to perform similarity calculations on the core emission parameter characteristics of adjacent monitoring items. This similarity calculation compares the degree of similarity in the core emission parameter characteristics of adjacent monitoring items to obtain a similarity value for parameter changes. This parameter change similarity value reflects the changes in core emission parameters within adjacent monitoring data.

[0064] Based on time interval values ​​and parameter change similarity values, a cross-item time-series association model can be constructed. This model comprehensively considers the factors of time interval and parameter change to calculate the time-series association strength value of adjacent monitoring items. The time-series association strength value reflects the degree of association between adjacent monitoring items; the higher the strength value, the stronger the association between adjacent monitoring items.

[0065] By calculating the time-series correlation strength values, cross-entry time-series correlation features containing correlation strength information can be generated. These features can help analyze the changing trends and correlations of carbon emission data over time.

[0066] Step S1241: Normalize the time interval value to obtain a time interval normalized value, and normalize the parameter change similarity value to obtain a parameter similarity normalized value.

[0067] After obtaining the time interval value and parameter change similarity value, since the range of these two values ​​may be different, they need to be normalized in order to achieve unified processing in the cross-item time series association model.

[0068] Normalization maps time interval values ​​and parameter change similarity values ​​to a set range, typically [0, 1]. For time interval values, normalization converts each time interval value into a relative value based on its maximum and minimum values; this relative value is the time interval normalized value. Similarly, parameter change similarity values ​​are normalized based on their maximum and minimum values ​​to obtain the parameter similarity normalized value.

[0069] Normalization can eliminate the influence of different parameter value ranges, making the time interval value and parameter change similarity value equally important and comparable in cross-item time series association models.

[0070] Step S1242: Calculate the temporal association strength value of adjacent monitoring items using the temporal association strength calculation formula, wherein the temporal association strength calculation formula is: temporal association strength value = reciprocal of the time interval normalization value × parameter similarity normalization value.

[0071] After obtaining the time interval normalized value and the parameter similarity normalized value, the time series association strength value of adjacent monitoring items can be calculated using the time series association strength calculation formula. The time series association strength calculation formula is designed based on the characteristics of the time interval normalized value and the parameter similarity normalized value, and it comprehensively considers the factors of time interval and parameter variation.

[0072] Specifically, the temporal correlation strength value is equal to the reciprocal of the time interval normalized value multiplied by the parameter similarity normalized value. The reciprocal of the time interval normalized value indicates that the shorter the time interval, the greater its contribution to the temporal correlation strength; the parameter similarity normalized value indicates that the more similar the parameter changes, the greater its contribution to the temporal correlation strength. Using this method, the temporal correlation strength value of adjacent monitoring items can be accurately calculated.

[0073] Step S1243: Perform a sliding window averaging process on the temporal correlation strength values ​​of K consecutive monitoring items to obtain the long-range temporal correlation strength value.

[0074] To further analyze the temporal correlations of carbon emission data over a longer period, a sliding window averaging process is needed for the temporal correlation strength values ​​of K consecutive monitoring items. The sliding window averaging process uses a fixed-size window that slides across the sequence of temporal correlation strength values, calculating the average value within the window to obtain the long-range temporal correlation strength value.

[0075] The size K of the sliding window is preset and can be adjusted according to actual needs. By using sliding window averaging, fluctuations in time-series correlation strength values ​​can be smoothed out, highlighting the changing trends and correlations of carbon emission data over a longer time period.

[0076] Step S1244: Use the temporal correlation strength value and the long-range temporal correlation strength value as components of the cross-entry temporal correlation feature; assign different feature dimensions to the temporal correlation strength value and the long-range temporal correlation strength value to generate a cross-entry temporal correlation feature containing correlation strength information.

[0077] After obtaining the temporal correlation strength value and the long-range temporal correlation strength value, they can be used as components of the cross-entry temporal correlation feature. To distinguish between these two different strength values, different feature dimensions can be assigned to them.

[0078] By assigning different feature dimensions to temporal correlation strength values ​​and long-range temporal correlation strength values, they can be combined into a multidimensional cross-entry temporal correlation feature. This cross-entry temporal correlation feature contains short-term and long-term temporal correlation information between adjacent monitoring entries.

[0079] Step S130: Call the pre-built log report paraphrase logic model to perform semantic reconstruction processing on the structured log feature group, and generate a log paraphrase logic framework, which includes the causal relationship chain of emission events and the distribution map of key information.

[0080] After obtaining the structured log feature set, in order to transform this feature information into a log reporting framework with logical structure and semantic relationships, it is necessary to call a pre-built log report paraphrasing logic model for semantic reconstruction processing. This model is trained on a large amount of carbon emission monitoring log data and can effectively identify the semantic relationships and logical connections between monitoring items.

[0081] In this embodiment, the log report paraphrasing logic model has a specific structure and functional modules, including a feature input layer, a causal relationship extraction module, and a key information detection module. First, the structured log feature set is input into the feature input layer of the log report paraphrasing logic model. The function of the feature input layer is to adjust the feature dimensions of the input structured log feature set to match the input dimensions of the log report paraphrasing logic model. Because different models have specific requirements for the dimensions of input data, if the dimensions of the input data do not match, the model may not function correctly or may produce inaccurate results. For example, the features in the structured log feature set may have different dimensions and formats. The feature input layer will uniformly adjust these features, converting them into a form suitable for model processing, resulting in the adjusted feature set.

[0082] Next, the causal relationship extraction module of the log report recap logic model performs relational reasoning processing on the adjusted feature groups. The main task of this module is to identify the types of causal relationships between monitoring items, specifically including state change relationships caused by parameter changes and trend continuation relationships caused by temporal continuity. In actual carbon emission monitoring, complex causal relationships may exist between different monitoring items. For example, when a parameter of an emission source changes, it may cause a corresponding change in the state of that emission source; or due to temporal continuity, certain trends of the emission source may continue. The causal relationship extraction module will conduct in-depth analysis and identification of these relationships.

[0083] Step S131: Input the structured log feature group into the feature input layer of the log report paraphrase logic model, perform feature dimension adjustment processing, and obtain the adjusted feature group that matches the input dimension of the log report paraphrase logic model.

[0084] After the structured log feature set is input into the feature input layer of the log report paraphrasing logic model, the feature input layer processes the features according to specific rules. First, each feature in the structured log feature set is examined to determine its dimensions and format. Then, based on the input requirements of the log report paraphrasing logic model, the features are adjusted accordingly. For example, if the model requires a fixed value for the input feature dimension, and a certain feature dimension in the structured log feature set does not meet this requirement, the feature input layer will adjust it using various methods. This might involve feature expansion, adding new feature dimensions to the existing features to meet the model's requirements; or feature compression, removing unnecessary feature dimensions to match the model's input dimensions. The adjusted feature set is stored in a unified format and dimension for subsequent processing by the causal relationship extraction module and the key information detection module.

[0085] Step S132: The adjusted feature group is subjected to relational reasoning processing by the causal relationship extraction module of the log report restatement logic model to identify the causal relationship types between monitoring items. The causal relationship types include state change relationships caused by parameter changes and trend continuation relationships caused by time continuity.

[0086] After obtaining the adjusted feature set, the causal relationship extraction module begins relational reasoning processing. First, it extracts the core emission parameter features of adjacent monitoring items from the adjusted feature set, obtaining the parameter features of the preceding and succeeding items. These two features represent key parameter information directly related to carbon emissions in two adjacent monitoring items. Then, a difference analysis is performed on the parameter features of the preceding and succeeding items. This analysis compares the differences between these two features to obtain information on the direction and magnitude of parameter changes. The direction of parameter change can be categorized as increase, decrease, or no change, while the magnitude of change indicates the degree of parameter variation.

[0087] Simultaneously, the causal relationship extraction module also extracts cross-item temporal correlation features between adjacent monitoring items in the adjusted feature group, obtaining a temporal correlation strength value. The temporal correlation strength value reflects the degree of correlation between adjacent monitoring items in the time dimension.

[0088] To accurately identify the types of causal relationships between monitoring items, causal relationship determination rules can be established. These rules determine the type of causal relationship based on the direction of parameter change, the direction of state change, and the strength of time-series correlation. When the direction of parameter change is consistent with the direction of state change and the strength of time-series correlation exceeds a preset threshold, it is determined to be a state change relationship caused by parameter change. For example, if the emission flow parameter of an emission source increases, and its state changes from low emission to high emission, and the strength of time-series correlation between adjacent monitoring items exceeds a preset threshold, it can be determined to be a state change relationship caused by parameter change. When the direction of parameter change is inconsistent with the direction of state change, but the strength of time-series correlation exceeds a preset threshold, it is determined to be a trend continuation relationship caused by time continuity. For example, although the emission flow parameter of an emission source does not change significantly, its emission state continues to develop according to the previous trend, and the strength of time-series correlation between adjacent monitoring items exceeds a preset threshold, it can be determined to be a trend continuation relationship caused by time continuity.

[0089] Using this causal relationship determination rule, the direct causal relationship type is obtained by determining the relationship between adjacent monitoring items. For non-adjacent monitoring items, a skip-step relationship reasoning process can be performed. Based on the causal relationship type of intermediate monitoring items, the indirect causal relationship type of non-adjacent monitoring items is derived. For example, if monitoring item A and monitoring item C are separated by monitoring item B, and the causal relationship types between monitoring item A and monitoring item B, and between monitoring item B and monitoring item C are known, the indirect causal relationship type between monitoring item A and monitoring item C can be derived through these intermediate relationships. Finally, the direct and indirect causal relationship types are used as the causal relationship types between monitoring items.

[0090] Step S1321: Extract the core emission parameter features of adjacent monitoring items in the adjusted feature group to obtain the parameter features of the preceding item and the parameter features of the succeeding item.

[0091] When processing the adjusted feature group in the causal relationship extraction module, adjacent monitoring entries can be located first. Adjacent monitoring entries refer to two monitoring entries that are sequentially adjacent, recording relevant information about the emission source at adjacent times. Then, core emission parameter features are extracted from these two adjacent monitoring entries. Core emission parameter features are key parameter information directly related to carbon emissions, such as the composition of emitted gases and the emission flow rate. The core emission parameter features corresponding to the preceding monitoring entry are defined as the precursor entry parameter features, and the core emission parameter features corresponding to the following monitoring entry are defined as the successor entry parameter features. These two features contain key carbon emission parameter information of the emission source at adjacent times.

[0092] Step S1322: Perform difference analysis on the parameter features of the predecessor entry and the parameter features of the successor entry to obtain information on the direction and magnitude of parameter changes.

[0093] After obtaining the parameter characteristics of the predecessor and successor entries, a difference analysis can be performed on them. This analysis compares each parameter in both the predecessor and successor entry parameter characteristics one by one. For each parameter, the difference between the parameter value in the successor entry parameter characteristics and the parameter value in the predecessor entry parameter characteristics can be calculated. The sign of the difference determines the direction of parameter change: a positive difference indicates an increase in the parameter value; a negative difference indicates a decrease in the parameter value; and a zero difference indicates no change in the parameter value. Furthermore, the magnitude of the parameter change can be determined based on the absolute value of the difference; a larger absolute value indicates a larger magnitude of parameter change. Through this method, the direction and magnitude of parameter change can be accurately obtained.

[0094] Step S1323: Extract the cross-item temporal correlation features of adjacent monitoring items in the adjusted feature group to obtain the temporal correlation strength value.

[0095] While performing difference analysis on the core emission parameter characteristics of adjacent monitoring items, the causal relationship extraction module also extracts cross-item time-series correlation features between adjacent monitoring items. These cross-item time-series correlation features are obtained in previous steps by processing the timestamp sequences and core emission parameter characteristics of adjacent monitoring items, reflecting the temporal correlation between them. The module finds the cross-item time-series correlation features of adjacent monitoring items from the adjusted feature set and extracts the time-series correlation strength value. The time-series correlation strength value is an indicator that comprehensively considers the time interval and the similarity of parameter changes, accurately reflecting the degree of correlation between adjacent monitoring items.

[0096] Step S1324: Establish causal relationship determination rules. The causal relationship determination rules are as follows: when the direction of parameter change is consistent with the direction of state change and the temporal correlation strength value exceeds a preset threshold, it is determined to be a state change relationship caused by parameter change; when the direction of parameter change is inconsistent with the direction of state change but the temporal correlation strength value exceeds a preset threshold, it is determined to be a trend continuation relationship caused by time continuity.

[0097] To accurately identify the types of causal relationships between monitoring items, a clear causal relationship determination rule needs to be established. This rule is based on the analysis and summarization of a large amount of carbon emission monitoring data. First, the consistency between the direction of parameter change and the direction of state change can be considered. In carbon emission monitoring, parameter changes often affect the state of emission sources. When the direction of parameter change is consistent with the direction of state change, it indicates that the parameter change is likely the cause of the state change. Simultaneously, the time-series correlation strength value is also considered. The time-series correlation strength value reflects the degree of temporal correlation between adjacent monitoring items. If the time-series correlation strength value exceeds a preset threshold, it indicates a strong correlation between adjacent monitoring items. When the direction of parameter change is consistent with the direction of state change and the time-series correlation strength value exceeds the preset threshold, it can be determined that the state change relationship is caused by the parameter change. For example, when the emission flow parameter of an emission source increases, and its emission state changes from low emission to high emission, and the time-series correlation strength value between adjacent monitoring items exceeds the preset threshold, it can be determined that the state change was caused by the parameter change.

[0098] When the direction of parameter change is inconsistent with the direction of state change, it indicates that the parameter change may not be the direct cause of the state change. However, if the time-series correlation strength value exceeds a preset threshold, it indicates a strong temporal continuity between adjacent monitoring items, and the state of the emission source may continue to develop according to the previous trend. In the above case, it is determined to be a trend continuation relationship caused by time continuity. For example, although the emission flow parameter of the emission source has not changed significantly, its emission state continues to increase according to the previous upward trend, and the time-series correlation strength value between adjacent monitoring items exceeds a preset threshold, which can be determined to be a trend continuation relationship caused by time continuity.

[0099] Step S1325: Perform relationship determination processing on adjacent monitoring items using the causal relationship determination rules to obtain the direct causal relationship type.

[0100] After establishing the causal relationship determination rules, the causal relationship extraction module uses these rules to determine the relationship between adjacent monitoring items. For each pair of adjacent monitoring items, their parameter change direction, state change direction, and temporal correlation strength value are first extracted. Then, this information is substituted into the causal relationship determination rules for judgment. According to the rules, if the parameter change direction is consistent with the state change direction and the temporal correlation strength value exceeds a preset threshold, the causal relationship between these adjacent monitoring items is determined to be a state change relationship caused by parameter change; if the parameter change direction is inconsistent with the state change direction but the temporal correlation strength value exceeds the preset threshold, it is determined to be a trend continuation relationship caused by temporal continuity. By performing the above determination process on all adjacent monitoring items, the direct causal relationship type between each pair of adjacent monitoring items can be obtained.

[0101] Step S1326: Perform skip-step relationship reasoning on non-adjacent monitoring items, and deduce the indirect causal relationship type of non-adjacent monitoring items based on the causal relationship type of intermediate monitoring items.

[0102] After determining the direct causal relationship types between adjacent monitoring items, it is necessary to analyze the causal relationships between non-adjacent monitoring items. For non-adjacent monitoring items, a step-by-step relationship reasoning process can be performed. Assume there are monitoring items A, B, and C, where A and C are non-adjacent monitoring items, and B is an intermediate monitoring item between them. Knowing the causal relationship types between A and B, and between B and C, the indirect causal relationship type between A and C can be deduced through the analysis of these intermediate relationships. If the relationship between A and B is a state change relationship caused by parameter changes, and the relationship between B and C is also a state change relationship caused by parameter changes, and these changes are transitive, then it can be inferred that a state change relationship caused by parameter changes may also exist between A and C. Through this step-by-step relationship reasoning process, a comprehensive analysis of the causal relationships between monitoring items can be performed, yielding more accurate causal relationship information.

[0103] Step S1327: The direct causal relationship type and the indirect causal relationship type are used as the causal relationship types between monitoring items.

[0104] After completing the causal relationship analysis of adjacent and non-adjacent monitoring items, the direct causal relationship types and indirect causal relationship types are merged into a complete causal relationship type between monitoring items. This comprehensively reflects the causal connections between monitoring items.

[0105] Step S133: Construct a causal relationship chain based on the causal relationship type. The causal relationship chain consists of nodes and edges representing causal relationships. Nodes represent monitoring items, and edges represent causal relationship types.

[0106] After determining the causal relationship types between monitoring items, causal relationship chains can be constructed based on these relationship types. A causal relationship chain is a graphical representation consisting of nodes and edges. Nodes represent monitoring items, with each node representing a specific monitoring item record; edges represent causal relationship types, with different types of edges representing different causal relationships. For example, solid edges represent state changes caused by parameter changes, while dashed edges represent trend continuation caused by time continuity. Based on the causal relationships between monitoring items, the corresponding nodes are connected with edges to form a complete causal relationship chain. This causal relationship chain can intuitively demonstrate the causal connections between monitoring items.

[0107] Step S134: The key information detection module of the log report retelling logic model performs information density calculation on the adjusted feature group and extracts monitoring items with information density exceeding a preset threshold as key information nodes.

[0108] The key information detection module in the log report recap logic model performs information density calculations on the adjusted feature groups. Information density refers to the richness of important carbon emission-related information contained in a monitoring item. The key information detection module analyzes each monitoring item in the adjusted feature group and calculates its information density. The specific calculation method may comprehensively consider factors such as the core emission parameter characteristics and state change records in the monitoring item. For example, if a monitoring item contains multiple key emission parameter information and obvious state change information, then its information density will be relatively high.

[0109] After calculating the information density of each monitoring item, the information density can be compared with a preset threshold. If the information density of a monitoring item exceeds the preset threshold, it is extracted as a key information node. These key information nodes contain a significant amount of important information and play a crucial role in understanding the key aspects of carbon emission events.

[0110] Step S135: Perform weighted summation on the core emission parameter features of the key information nodes to obtain the comprehensive weight value of the key information nodes.

[0111] After identifying key information nodes, their core emission parameter features can be weighted and accumulated. Each key information node's core emission parameter features contain multiple parameters, each with a corresponding weight value. These weight values ​​were assigned based on the importance of the parameters during the previous feature extraction steps. The weight values ​​of each parameter in the key information node's core emission parameter features are then summed to obtain the node's overall weight value. This overall weight value reflects the importance of the key information node within the overall carbon emission monitoring data; a higher weight value indicates more important information contained within the node.

[0112] Step S136: Perform weight labeling on the nodes in the causal relationship chain based on the comprehensive weight value to generate a key information distribution map containing node weight information.

[0113] After obtaining the comprehensive weight values ​​of key information nodes, weight labeling can be performed on the nodes in the causal relationship chain based on these weight values. The comprehensive weight values ​​of key information nodes are labeled onto the corresponding nodes in the causal relationship chain. Meanwhile, non-key information nodes can be assigned a relatively low weight value according to predefined rules. Through this method, a key information distribution map containing node weight information is generated. This key information distribution map can intuitively display the importance of each node in the causal relationship chain.

[0114] Step S137: Combine the causal relationship chain and the key information distribution map to generate a log retelling logic framework.

[0115] After constructing the causal relationship chain and generating the key information distribution map, they are combined. This combination can be done by overlaying the key information distribution map onto the causal relationship chain, or by associating them in a predetermined manner. Through this combination process, a complete log recap logic framework is generated. This framework includes the causal relationship chain of emission events and the key information distribution map.

[0116] Step S140: Perform a text generation operation based on the log paraphrasing logic framework to generate an initial paraphrased log report, which includes emission event descriptions organized according to causal chains.

[0117] After generating the log retelling logic framework, the next step is text generation, which aims to transform the logic framework into a natural, fluent, and organized initial log retelling report.

[0118] Step S141: Extract the causal relationship chain in the log paraphrasing logic framework and determine the description order of the emission events, wherein the description order is arranged according to the node order of the causal relationship chain.

[0119] In this step, the causal chain is extracted from the log recap logic framework. The causal chain presents the causal relationships between monitoring items, with nodes representing monitoring items and edges representing causal relationship types. The order of nodes in the causal chain clarifies the order in which emission events are described. This is done because describing emission events in the order of causal relationships allows the report's readers to clearly understand the causal connections between each emission event, giving the report a clear logical structure. For example, if node A in the causal chain leads to node B, and node B in turn triggers node C, then when describing emission events, the emission events related to node A will be described first, followed by those related to node B, and finally those related to node C.

[0120] Step S142: Extract the key information distribution map in the log paraphrasing logic framework, determine the level of detail in the description of each node, and the level of detail in the description is positively correlated with the comprehensive weight value of the node.

[0121] Next, the key information distribution map of the log recap logic framework is extracted. This key information distribution map contains the weight information of each node in the causal chain. This weight information is obtained by weighting and accumulating the core emission parameter characteristics of the key information nodes. Based on the key information distribution map, the level of detail in the description of each node is determined. Since the level of detail is positively correlated with the node's overall weight value, nodes with higher overall weight values ​​mean that the information they contain is more important and will be described in detail in the report; while nodes with lower overall weight values ​​will be described more briefly. For example, a key information node with a high overall weight value might describe in detail the specific situation of its emission source, the changes in core emission parameters, and the specific manifestations of state changes; while a node with a low overall weight value might only briefly mention the general situation of the emission event.

[0122] Step S143: Perform text template matching processing on each node in the causal relationship chain, wherein the text template contains a statement structure corresponding to the causal relationship type.

[0123] After determining the order and level of detail in the descriptions, text template matching is performed on each node in the causal chain. The text templates are pre-designed and contain sentence structures corresponding to different types of causal relationships. These sentence structures help to accurately express the emission events represented by the nodes in natural language.

[0124] Step S1431: Establish a mapping table between causal relationship types and text templates. In the mapping table, the state change relationship caused by parameter changes corresponds to the first type of text template, and the trend continuation relationship caused by time continuity corresponds to the second type of text template.

[0125] First, a mapping table is established between causal relationship types and text templates. This mapping table clarifies the text templates corresponding to different causal relationship types. Specifically, state change relationships caused by parameter changes correspond to the first type of text template, which typically emphasizes how parameter changes lead to changes in the emission source's state; trend continuation relationships caused by temporal continuity correspond to the second type of text template, which focuses on describing the trend of the emission event over time. For example, the first type of text template might be "Due to a [change] in [precursor parameters], the emission source's state changed from [precursor state] to [successor state]"; the second type of text template might be "As time goes on, the emission source's [state] continues to [change] according to the previous trend."

[0126] Step S1432: Identify the causal relationship type corresponding to each node in the causal relationship chain.

[0127] Then, each node in the causal chain is analyzed to identify its corresponding causal relationship type. This is based on the causal relationship type results obtained in the causal relationship extraction module. By examining the relevant characteristics of the node, including core emission parameter characteristics and cross-entry time-series correlation characteristics, it is determined whether the node corresponds to a state change relationship caused by parameter changes or a trend continuation relationship caused by time continuity.

[0128] Step S1433: Match the corresponding text template from the mapping relationship table according to the causal relationship type.

[0129] Based on the identified causal relationship type, the corresponding text template is searched in the mapping table. If the node corresponds to a state change relationship caused by parameter changes, the first type of text template is selected from the mapping table; if it is a trend continuation relationship caused by continuous time, the second type of text template is selected.

[0130] Step S1434: Perform parameter placeholder marking processing on the text template. The parameter placeholders include predecessor parameter placeholders, successor state placeholders, time interval placeholders, and parameter trend placeholders.

[0131] After selecting a text template, parameter placeholders are added. Parameter placeholders are reserved positions within the text template used to fill in specific information. Common parameter placeholders include: predecessor parameter placeholders, used to fill in core emission parameter information from the previous monitoring item; successor status placeholders, used to fill in the status information of the emission source in the next monitoring item; time interval placeholders, used to fill in the time interval information between two monitoring items; and parameter trend placeholders, used to fill in the parameter's changing trend information, such as increase or decrease.

[0132] Step S1435: Extract the core emission parameter features of the precursor monitoring entries and successor monitoring entries corresponding to the node to obtain the content that needs to be filled into the precursor parameter placeholders and successor status placeholders.

[0133] Next, the core emission parameter features of the preceding and subsequent monitoring entries corresponding to the node are extracted. These features, obtained in the previous feature extraction steps, contain key parameter information directly related to carbon emissions. Relevant information from the core emission parameter features of the preceding monitoring entries is filled into the preceding parameter placeholders in the text template, and state-related information from the core emission parameter features of the subsequent monitoring entries is filled into the subsequent state placeholders. For example, if the preceding parameter placeholder is "[Preceding Emission Flow]", emission flow information is extracted from the core emission parameter features of the preceding monitoring entries and filled in; if the subsequent state placeholder is "[Subsequent Emission Status]", emission status information is extracted from the core emission parameter features of the subsequent monitoring entries and filled in.

[0134] Step S1436: Extract the timestamp information corresponding to the node, calculate the time interval value, and obtain the content to be filled into the time interval placeholder.

[0135] Extract the timestamp information corresponding to the node, and calculate the time interval value by calculating the timestamp difference between adjacent monitoring items. This time interval value is the content to be filled into the time interval placeholder in the text template. For example, if the time interval placeholder is "[time interval]", the calculated time interval value will be filled into this placeholder.

[0136] Step S1437: Extract the parameter change trend information corresponding to the node to obtain the target content that needs to be filled into the parameter trend placeholder.

[0137] Extract the parameter change trend information corresponding to the node. This is obtained by performing a difference analysis on the core emission parameter characteristics of the preceding and subsequent monitoring items, including information such as whether the parameter is increasing, decreasing, or remaining unchanged. Use this parameter change trend information as target content to populate the parameter trend placeholders in the text template. For example, if the parameter trend placeholder is "[parameter trend]", enter "increase" if the parameter is increasing, and "decrease" if the parameter is decreasing.

[0138] Step S1438: Perform one-to-one correspondence processing between the target content and the parameter placeholders to complete the text template matching process.

[0139] Finally, the extracted target content is matched one-to-one with the parameter placeholders in the text template. The corresponding content is accurately filled into the appropriate parameter placeholders, thus completing the text template matching process. After this series of operations, each node has a corresponding text template filled with specific information.

[0140] Step S144: Fill the parameter placeholders in the text template with content according to the level of detail in the description. The content filling process includes filling in the specific description of the core emission parameter features.

[0141] After completing the text template matching process, the parameter placeholders in the text template are further populated with content based on the previously determined level of detail. For nodes with higher level of detail, in addition to filling in basic parameter information, more detailed descriptions of core emission parameter characteristics are provided. For example, not only are the specific values ​​of the emission parameters filled in, but also the changing trends of the emission parameters over a set time period and their correlation with other parameters are described. For nodes with lower level of detail, only basic parameter information is filled in. Through this method, the text template is made to more richly and accurately reflect the emission event represented by the node.

[0142] Step S145: Adjust the sentence coherence of the filled text template by adding connecting adverbs and time adverbs to enhance the logical connection between sentences.

[0143] A pre-filled text template may lack coherence, thus requiring adjustments. Adding conjunctions and adverbs of time can enhance the logical flow between sentences. Conjunctions such as "therefore," "so," and "however" clearly express causal or adversative relationships; adverbs of time such as "subsequently," "following," and "after a period of time" specify the chronological order of events. For example, when describing two adjacent emission events, adding "Subsequently, due to [parameter changes], the state of the emission source underwent [state change]" makes the report's language more fluent and the logic clearer.

[0144] Step S146: The adjusted text template is spliced ​​together according to the described order to generate an initial restatement log report containing emission event descriptions organized according to causal chains.

[0145] Finally, the text templates, after being processed for sentence coherence adjustment, are concatenated according to the previously determined descriptive order. The text templates corresponding to each node are connected sequentially to form a complete text. This text is the initial restatement log report containing descriptions of emission events organized according to causal chains. This initial restatement log report presents the causal relationships and key information in the carbon emission monitoring data in natural language.

[0146] Step S150: Perform a two-dimensional verification process on the initial paraphrase log report to generate a final paraphrase log report that passes the verification. The two-dimensional verification process includes semantic consistency verification and structural integrity verification.

[0147] After generating the initial restatement log report, in order to ensure the quality and accuracy of the report, it is necessary to perform two-dimensional verification, including semantic consistency verification and structural integrity verification.

[0148] Step S151: Input the initial restatement log report and the original carbon emission monitoring log data into a pre-trained semantic similarity model, and calculate the semantic similarity score between the two; extract the core emission parameter description content from the initial restatement log report, and perform matching degree verification processing with the core emission parameter features in the original carbon emission monitoring log data to obtain the parameter matching degree score; perform weighted average processing on the semantic similarity score and the parameter matching degree score to obtain the semantic consistency verification score.

[0149] First, semantic consistency is verified. The initial restatement log report and the original carbon emission monitoring log data are input into a pre-trained semantic similarity model. This semantic similarity model, trained on a large amount of data, is capable of accurately calculating the semantic similarity between two texts. The model calculates a semantic similarity score between the initial restatement log report and the original carbon emission monitoring log data; this score reflects the degree of semantic similarity between the two.

[0150] Simultaneously, the core emission parameter descriptions in the initial recap log report are extracted and compared with the core emission parameter features in the original carbon emission monitoring log data. By comparing the parameter information in both reports, including parameter names, values, and changes, a parameter matching score is obtained. The parameter matching score reflects the degree of agreement between the core emission parameter descriptions in the initial recap log report and the original carbon emission monitoring log data.

[0151] Then, the semantic similarity score and parameter matching score are weighted and averaged. Different weights are assigned to the semantic similarity score and parameter matching score according to their importance, and the weighted average of the two is calculated to obtain the semantic consistency verification score. This semantic consistency verification score comprehensively reflects the degree of semantic consistency between the initial restated log report and the original carbon emission monitoring log data.

[0152] Step S152: Extract the causal relationship chain description from the initial paraphrase log report, and perform node count verification with the causal relationship chain in the log paraphrase logic framework to obtain a node integrity score; extract the key information description from the initial paraphrase log report, and perform weight matching verification with the key information distribution map in the log paraphrase logic framework to obtain a key integrity score; perform weighted average processing on the node integrity score and the key integrity score to obtain a structural integrity verification score.

[0153] Next, structural integrity verification is performed. The causal chain descriptions in the initial retelling log report are extracted and compared with the causal chain in the log retelling logic framework by checking the number of nodes. If the number of nodes matches, the causal chain is structurally complete; if they do not match, there are missing or redundant nodes. Based on the verification results, a node integrity score is obtained. The node integrity score reflects the structural integrity of the causal chain in the initial retelling log report.

[0154] Simultaneously, key information descriptions are extracted from the initial log recap report and compared with the key information distribution map in the log recap logic framework using weighted matching. For example, the specific steps are as follows:

[0155] Step S1521: Perform text parsing processing on the initial restatement log report and extract descriptive statements containing degree adverbs as key information description content.

[0156] The initial recap log report is parsed, and degree adverbs such as "significant," "slight," and "drastic" are identified in the text. Descriptive statements containing these degree adverbs are extracted as key information. Because degree adverbs are typically used to emphasize key information, descriptive statements containing degree adverbs often contain important information.

[0157] Step S1522: Extract keywords from the key information description to obtain a set of key parameter words.

[0158] Keyword extraction is performed on the extracted key information descriptions. By analyzing the semantics of the descriptive statements, keywords related to the core emission parameters are extracted, forming a key parameter vocabulary set. These keywords represent the key information in the key information descriptions.

[0159] Step S1523: Extract the key information distribution map in the log paraphrasing logic framework to obtain a list of key parameter words and their corresponding comprehensive weight values.

[0160] The log retelling logic framework extracts a distribution map of key information, from which a list of key parameter terms and a comprehensive weight value for each key parameter are obtained. The list of key parameter terms contains parameter terms considered important in the distribution map, and the comprehensive weight value reflects the importance of these parameters.

[0161] Step S1524: Calculate the number of intersections between the set of key parameter words and the list of key parameter words to obtain the number of word matches.

[0162] Calculate the number of intersections between the set of key parameter words and the list of key parameter words. The number of intersections represents the degree of word matching between the key information descriptions in the initial log retelling report and the key information distribution map in the log retelling logical framework. By comparing the words in the two sets and counting the number of identical words, the word matching count is obtained.

[0163] Step S1525: Perform statistical processing on the description length of each word in the key parameter vocabulary set to obtain the description length value of each word.

[0164] The description length of each word in the key information description content within the key parameter vocabulary set is statistically processed. Description length can be the number of characters or words in the sentence containing that word. By calculating the description length of each word, a description length value for each word is obtained.

[0165] Step S1526: Normalize the comprehensive weight value of each word in the key information distribution map to obtain the normalized weight value.

[0166] The comprehensive weight value of each word in the key information distribution map is normalized. The purpose of normalization is to map the comprehensive weight value to a set range for subsequent calculations. Through normalization, the normalized weight value of each word is obtained.

[0167] Step S1527: Calculate the Pearson correlation coefficient between the description length value and the normalized weight value to obtain the weight matching correlation value.

[0168] Calculate the Pearson correlation coefficient between the description length of each word in the key parameter vocabulary set and the normalized weight of the corresponding word in the key information distribution map. The Pearson correlation coefficient reflects the degree of linear correlation between the two variables. By calculating this correlation coefficient, the weight-matched correlation value is obtained. The weight-matched correlation value reflects the degree of matching between the description length of key information in the initial recap log report and the parameter weights in the key information distribution map.

[0169] Step S1528: Multiply the number of matched words by the weighted matching relevance value to obtain the key integrity score.

[0170] Finally, the number of word matches is multiplied by the weighted relevance value to obtain the key information completeness score. This score comprehensively considers both word matching and weighted matching, reflecting the degree of matching between the key information descriptions in the initial log retelling report and the key information distribution map in the log retelling logical framework.

[0171] The node integrity score and the key integrity score are weighted and averaged. Different weights are assigned to the node integrity score and the key integrity score based on their importance, and the weighted average of the two is then calculated to obtain the structural integrity verification score. This structural integrity verification score comprehensively reflects the degree of structural integrity of the initial restatement log report.

[0172] Step S153: Perform verification and judgment processing on the initial paraphrased log report according to the verification pass rules. If the judgment passes, output the initial paraphrased log report as the final paraphrased log report. If the judgment fails, return to the log paraphrasing logic framework to perform the text generation operation correction step. The verification pass rules are that the semantic consistency verification score exceeds the first preset threshold and the structural integrity verification score exceeds the second preset threshold.

[0173] Finally, the initial restated log report is validated according to the validation rules. The validation rules are that the semantic consistency score exceeds a first preset threshold and the structural integrity score exceeds a second preset threshold. If both the semantic consistency and structural integrity scores of the initial restated log report meet these rules, it is considered passed, and the initial restated log report is output as the final restated log report. If the rules are not met, it is considered failed, and the log restated logic framework needs to be returned to perform a text generation correction step. In the correction step, text generation parameters may be adjusted, such as the level of detail in the description and the selection of the text template. The initial restated log report is regenerated and then subjected to two-dimensional validation again until the validation rules are met.

[0174] Figure 2 The diagram illustrates exemplary hardware and software components of a carbon emission monitoring-based log report recap system 100, which implements the inventive concept, according to some embodiments of the present invention. For example, a processor 120 may be used in the carbon emission monitoring-based log report recap system 100 and to perform the functions described in the present invention.

[0175] The log report recap system 100 based on carbon emission monitoring can be a general-purpose server or a special-purpose server; both can be used to implement the log report recap method based on carbon emission monitoring of this invention. Although only one server is shown in this invention, for convenience, the functions described in this invention can be implemented in a distributed manner on multiple similar platforms to balance the load.

[0176] For example, a carbon emission monitoring-based log report recap system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the carbon emission monitoring-based log report recap system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present invention can be implemented according to these program instructions. The carbon emission monitoring-based log report recap system 100 also includes an I / O interface 150 between the computer and other input / output devices.

[0177] For ease of explanation, only one processor is described in the carbon emission monitoring-based log report recap system 100. However, it should be noted that the carbon emission monitoring-based log report recap system 100 of this invention may also include multiple processors, and therefore the steps performed by one processor described in this invention may also be performed jointly by multiple processors or individually. For example, if the processor of the carbon emission monitoring-based log report recap system 100 performs steps A and B, it should be understood that steps A and B may also be performed jointly by two different processors or individually by one processor. For example, the first processor performs step A, the second processor performs step B, or the first processor and the second processor jointly perform steps A and B.

[0178] Furthermore, this embodiment of the invention also provides a readable storage medium, wherein computer-executable instructions are preset in the readable storage medium, and when the processor executes the computer-executable instructions, the above-mentioned log report restatement method based on carbon emission monitoring is implemented.

[0179] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A method for restating log reports based on carbon emission monitoring, characterized in that, The method includes: Obtain raw carbon emission monitoring log data containing timestamp sequences. The raw carbon emission monitoring log data consists of multiple continuously collected monitoring entries. Each monitoring entry includes an emission source identification field, a monitoring parameter description field, and a status change record field. Feature extraction is performed on the raw carbon emission monitoring log data to obtain a structured log feature group, which includes the core emission parameter features of each monitoring item and cross-item time-series correlation features; The pre-built log report paraphrase logic model is invoked to perform semantic reconstruction processing on the structured log feature group, generating a log paraphrase logic framework, which includes a causal relationship chain of emission events and a key information distribution map; Based on the log paraphrasing logic framework, a text generation operation is performed to generate an initial paraphrased log report, which includes a description of emission events organized according to a causal chain. The initial paraphrase log report is subjected to two-dimensional verification processing to generate a final paraphrase log report that passes the verification. The two-dimensional verification processing includes semantic consistency verification and structural integrity verification. The method of invoking a pre-built log report paraphrasing logic model to perform semantic reconstruction processing on the structured log feature groups, generating a log paraphrasing logic framework, includes: The structured log feature set is input into the feature input layer of the log report paraphrase logic model, and the feature dimension is adjusted to obtain the adjusted feature set that matches the input dimension of the log report paraphrase logic model. The causal relationship extraction module of the log report restatement logic model performs relational reasoning processing on the adjusted feature group to identify the causal relationship types between monitoring items. The causal relationship types include state change relationships caused by parameter changes and trend continuation relationships caused by time continuity. A causal relationship chain is constructed based on the causal relationship type. The causal relationship chain consists of nodes and edges representing causal relationships. Nodes represent monitoring items, and edges represent causal relationship types. The key information detection module of the log report paraphrase logic model performs information density calculation on the adjusted feature group and extracts monitoring items whose information density exceeds a preset threshold as key information nodes. The core emission parameter features of the key information nodes are weighted and accumulated to obtain the comprehensive weight value of the key information nodes; Based on the comprehensive weight value, the nodes in the causal relationship chain are weighted and labeled to generate a key information distribution map containing node weight information. The causal relationship chain and the key information distribution map are combined and processed to generate a log paraphrase logic framework; The causal relationship extraction module of the log report restatement logic model performs relational reasoning processing on the adjusted feature group to identify the causal relationship type between monitoring items, including: Extract the core emission parameter features of adjacent monitoring items in the adjusted feature group to obtain the parameter features of the preceding item and the parameter features of the succeeding item. The parameter characteristics of the predecessor entry and the parameter characteristics of the successor entry are analyzed for differences to obtain information on the direction and magnitude of parameter changes. Extract the cross-item temporal correlation features of adjacent monitoring items in the adjusted feature group to obtain the temporal correlation strength value; A causal relationship determination rule is established, wherein the causal relationship determination rule is as follows: when the direction of parameter change is consistent with the direction of state change and the temporal correlation strength value exceeds a preset threshold, it is determined to be a state change relationship caused by parameter change; when the direction of parameter change is inconsistent with the direction of state change but the temporal correlation strength value exceeds a preset threshold, it is determined to be a trend continuation relationship caused by time continuity. The relationship determination rules are used to determine the direct causal relationship type by processing the relationship between adjacent monitoring items. Skip-step relationship reasoning is performed on non-adjacent monitoring items, and the indirect causal relationship type of non-adjacent monitoring items is derived based on the causal relationship type of intermediate monitoring items; The direct causal relationship type and the indirect causal relationship type are used as the causal relationship types between monitoring items; The step of performing text generation operations based on the log paraphrasing logic framework to generate an initial paraphrased log report includes: Extract the causal relationship chain from the log paraphrasing logic framework to determine the description order of emission events, wherein the description order is arranged according to the node order of the causal relationship chain; Extract the key information distribution map from the log retelling logic framework, determine the level of detail in the description of each node, and the level of detail in the description is positively correlated with the comprehensive weight value of the node; Each node in the causal relationship chain is subjected to text template matching processing, wherein the text template contains a statement structure corresponding to the causal relationship type; The parameter placeholders in the text template are filled with content according to the level of detail in the description. The content filling process includes filling in the specific descriptive content of the core emission parameter features. The text template after filling is processed to improve sentence coherence by adding connecting adverbs and time adverbs to enhance the logical connection between sentences; The adjusted text templates are concatenated in the order described to generate an initial restatement log report containing descriptions of emission events organized according to causal chains.

2. The log report restatement method based on carbon emission monitoring according to claim 1, characterized in that, The process of extracting features from the original carbon emission monitoring log data to obtain structured log feature groups includes: The original carbon emission monitoring log data is parsed to extract the emission source identifier field value, monitoring parameter description field value, and status change record field value for each monitoring entry; The monitoring parameter description field values ​​are subjected to keyword filtering to select parameter words directly related to carbon emissions as core parameter candidate words. The core parameter candidate words include emission type limiting words, measurement unit words, and threshold description words. Semantic strength analysis is performed on the state change record field values ​​to calculate the degree adverb weight values ​​describing the state change. Based on the degree adverb weight values, the core parameter candidate words are sorted by importance to generate core emission parameter features containing weight information. The time interval between adjacent monitoring items is calculated by performing interval calculation on the timestamp sequences of adjacent monitoring items; The similarity of the core emission parameter characteristics of adjacent monitoring items is calculated to obtain the parameter change similarity value of adjacent monitoring items; A cross-item temporal association model is constructed based on the time interval value and the parameter change similarity value. The temporal association strength value of adjacent monitoring items is calculated through the cross-item temporal association model, and cross-item temporal association features containing association strength information are generated. The core emission parameter features and the cross-entry time-series correlation features are processed by dimensionality standardization to generate structured log feature groups.

3. The log report restatement method based on carbon emission monitoring according to claim 2, characterized in that, The semantic strength analysis of the state change record field values ​​is performed to calculate the degree adverb weight values ​​describing the state change. Based on the degree adverb weight values, the importance of the core parameter candidate words is ranked to generate core emission parameter features containing weight information, including: Establish a lexicon of adverbs describing the degree of carbon emission status, wherein the lexicon contains a set of words representing the magnitude of change; The value of the state change record field is processed by adverb recognition, and adverb words that match the degree adverb library are extracted as state degree adverbs; Each adverb of state degree is assigned a preset weight coefficient value, and the weight coefficient value is positively correlated with the significance of the state change; Count the number of times each core parameter candidate word co-occurs with state degree adverbs, and calculate the co-occurrence frequency value of each core parameter candidate word; Multiply the co-occurrence frequency value by the weight coefficient value of the corresponding state degree adverb to obtain the importance score value of each core parameter candidate word; The candidate words for core parameters are sorted in descending order based on the importance score to generate a core parameter sequence sorted by importance. Assign a corresponding importance score to each parameter in the core parameter sequence to generate core emission parameter features containing weight information.

4. The log report restatement method based on carbon emission monitoring according to claim 2, characterized in that, The method involves constructing a cross-item temporal association model based on the time interval value and the parameter change similarity value, calculating the temporal association strength value of adjacent monitoring items through the cross-item temporal association model, and generating cross-item temporal association features containing association strength information, including: The time interval value is normalized to obtain a time interval normalized value, and the parameter change similarity value is normalized to obtain a parameter similarity normalized value. The temporal association strength value of adjacent monitoring items is calculated using the temporal association strength calculation formula, which is: temporal association strength value = reciprocal of the time interval normalization value × parameter similarity normalization value; The long-range time-series correlation strength value is obtained by averaging the time-series correlation strength values ​​of K consecutive monitoring items using a sliding window method. The temporal correlation strength value and the long-range temporal correlation strength value are used as components of the cross-entry temporal correlation feature; Different feature dimensions are assigned to the temporal correlation strength value and the long-range temporal correlation strength value to generate cross-entry temporal correlation features containing correlation strength information.

5. The log report restatement method based on carbon emission monitoring according to claim 1, characterized in that, The text template matching process is performed on each node in the causal relationship chain. The text template contains a statement structure corresponding to the causal relationship type, including: Establish a mapping table between causal relationship types and text templates. In the mapping table, the state change relationship caused by parameter changes corresponds to the first type of text template, and the trend continuation relationship caused by time continuity corresponds to the second type of text template. Identify the causal relationship type corresponding to each node in the causal relationship chain; Match the corresponding text template from the mapping table according to the causal relationship type; The text template is marked with parameter placeholders, which include predecessor parameter placeholders, successor state placeholders, time interval placeholders, and parameter trend placeholders. Extract the core emission parameter features of the preceding and following monitoring entries corresponding to the node to obtain the content that needs to be filled into the preceding parameter placeholder and the following state placeholder; Extract the timestamp information corresponding to the node, calculate the time interval value, and obtain the content that needs to be filled into the time interval placeholder; Extract the parameter change trend information corresponding to the node to obtain the target content that needs to be filled into the parameter trend placeholder; The target content is matched one-to-one with the parameter placeholders to complete the text template matching process.

6. The log report restatement method based on carbon emission monitoring according to claim 1, characterized in that, The step of performing a two-dimensional verification process on the initial paraphrase log report to generate a final paraphrase log report that passes the verification includes: The initial restatement log report and the original carbon emission monitoring log data are input into a pre-trained semantic similarity model to calculate their semantic similarity scores. The core emission parameter descriptions in the initial restatement log report are extracted and compared with the core emission parameter features in the original carbon emission monitoring log data to obtain a parameter matching score. The semantic similarity score and the parameter matching score are then weighted and averaged to obtain a semantic consistency verification score. The causal relationship chain description content is extracted from the initial paraphrase log report, and the node number is checked against the causal relationship chain in the log paraphrase logic framework to obtain a node integrity score. The key information description content is extracted from the initial paraphrase log report, and the key information distribution map in the log paraphrase logic framework is checked against a weighted match to obtain a key information integrity score. The node integrity score and the key information integrity score are weighted and averaged to obtain a structural integrity verification score. The initial paraphrased log report is validated according to the validation rules. If the validation passes, the initial paraphrased log report is output as the final paraphrased log report. If the validation fails, the log paraphrasing logic framework is returned to perform the text generation operation correction step. The validation rules are that the semantic consistency validation score exceeds the first preset threshold and the structural integrity validation score exceeds the second preset threshold.

7. A log report summarization system based on carbon emission monitoring, characterized in that, The system includes a processor and a memory, the memory being connected to the processor. The memory is used to store programs, instructions, or code, and the processor is used to execute the programs, instructions, or code in the memory to implement the log report restatement method based on carbon emission monitoring as described in any one of claims 1-6.

Images