ESG intelligent sensing method and system fused with Internet of Things

Abstract

The invention relates to the technical field of sensing information monitoring, in particular to an ESG intelligent sensing method and system fused with the Internet of Things, and the method comprises the following steps: obtaining power distribution energy consumption data, carrying out the segmented analysis of an extreme value, extracting water, electricity and gas parameters to construct a feature vector after an abnormal time period is recognized, analyzing the change direction and amplitude to form an offset structure body, and carrying out the recognition of the change direction and amplitude. And identifying a reverse change medium combination to determine an offset section, screening a high-frequency time period to generate a conversion scheme, and integrating score change, duration difference and medium types to form a multi-dimensional label set. According to the method, through repeatability analysis and abnormal section extraction of energy consumption data, accurate positioning of non-periodic fluctuation, periodic vector construction based on multi-medium parameters and change feature recognition are realized, cross-medium linkage perception and abnormal collaborative recognition capability is enhanced, and through trend segmentation and scoring trajectory extraction, the accuracy and the reliability of the system are improved. The sensitive recognition efficiency of the abnormal trend is improved, multi-dimensional hopping information is structurally integrated, and the monitoring adaptability and the data analysis capability under multiple scenes are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of sensor information monitoring technology, and in particular to an ESG intelligent perception method and system integrated with the Internet of Things. Background Art

[0002] The field of sensor information monitoring technology encompasses system technologies for the real-time acquisition, transmission, and analysis of information on physical and chemical quantities, as well as physiological states, under various environmental and operational conditions. Its core elements include sensor deployment and acquisition mechanisms, data communication technologies, edge computing capabilities, and back-end information processing architectures. These technologies encompass the precise monitoring of diverse data types, including water, electricity, gas, temperature, humidity, pollutant emissions, noise, vibration, and human health. In application scenarios such as enterprise management, urban operations, and industrial production, sensor information monitoring technology is evolving towards automation, intelligence, and networking, forming a comprehensive monitoring system integrating sensing terminals, communication networks, and data analysis.

[0003] Among them, the ESG intelligent perception method that integrates the Internet of Things refers to the use of the Internet of Things to automatically collect, synchronously transmit, and classify information for specific matters such as resource consumption monitoring, pollution emission collection, climate response parameter capture, employee health status perception, and safe working environment monitoring in the enterprise ESG management scenario. This method captures the above-mentioned ESG-related data in real time by deploying electricity sensors, water flow detection devices, gas concentration monitors, wearable physiological parameter collection equipment, and regional environmental perception probes, and relies on standardized communication protocols to transmit the collected information to centralized data nodes for preliminary processing. At the same time, structured rules are used to complete data label classification, and the perception data is mapped to various ESG assessment factors based on indicator settings, constructing a data source support framework with continuous input capabilities.

[0004] Existing sensor information monitoring technologies generally employ static rules and single-factor thresholds in practice, making them ineffective at identifying energy consumption anomalies. Information collection often relies on independent monitoring of individual items, lacking a foundation for correlation analysis between media. This makes it difficult to reveal the interdependencies or linkages between resource usage behaviors, leading to isolated anomaly detection and insufficient utilization of information value. In terms of temporal analysis, there is a lack of structured representation of the continuity of trend changes and hierarchical evolution, making it incapable of supporting early warning and dynamic response to potential trends. Scoring data is often presented as periodic fragments in applications, lacking a continuous scoring trajectory, resulting in ambiguous identification of risk areas. For example, in the case of a sudden increase in pollutant emissions and energy consumption, existing models struggle to detect coordinated anomalies through the combination of indicators, easily leading to management blind spots. Consequently, existing technologies have significant shortcomings in handling complex scenarios, the cross-correlation of multidimensional data, and the expression of trend evolution, limiting their adaptability and accuracy within intelligent monitoring systems. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the existing technology and propose an ESG intelligent perception method and system integrating the Internet of Things.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an ESG intelligent perception method integrating the Internet of Things, comprising the following steps:

[0007] S1: Obtain energy consumption data of power distribution nodes in the office area, divide it into equal time segments, extract the maximum, minimum and average values, analyze the repeatability of parameters in adjacent segments, and generate a list of abnormal marked segments.

[0008] S2: According to the anomaly marked section time, the representative parameters of the three types of media, water, electricity, and gas, within the same time are extracted, a periodic parameter vector is constructed, the direction and amplitude classification of the parameter changes between cycles are identified, and an offset identification structure is generated.

[0009] S3: Call the change characteristics in the offset identification structure to identify the medium combination with the opposite change direction and the largest amplitude at the same time point, extract the corresponding time period as the structural offset area, select the time period with the highest proportion among the same type, and generate a time dimension conversion solution table.

[0010] S4: Based on the time periods in the conversion scheme table, extract the scoring levels, divide the segments with the same continuous level, record the levels and durations, analyze the relationship between the level change trends and time ratios of adjacent scoring segments, and generate a continuous scoring segment difference trend map.

[0011] S5: calling the jump record in the continuous scoring segment difference trend map, extracting the score level change, duration difference and medium type, integrating them into structured tag fields, and generating a multi-dimensional intelligent perception tag set.

[0012] As a further solution of the present invention, the abnormal marked segment list includes maximum value parameters, minimum value parameters, average value parameters, and data location information; the offset identification structure includes periodic parameter vectors, parameter change direction classifications, parameter change amplitude classifications, and medium change feature sets; the time dimension conversion scheme table includes structural offset area time periods, medium combination types, and time proportion statistical information; the continuous scoring segment difference trend map includes scoring level records, duration records, level change trend analysis, and time proportion relationship data; the multidimensional intelligent perception label set includes scoring level change labels, duration difference labels, and medium type identification fields.

[0013] As a further solution of the present invention, the specific steps of S1 are:

[0014] S101: Obtain energy consumption data of power distribution nodes in the office area, divide the data into equal time segments, extract energy consumption records in each segment, perform maximum, minimum, and average value extraction operations on each segment of data, and generate a segment energy consumption parameter set;

[0015] S102: calling the data in the section energy consumption parameter set, comparing the parameters of each group of adjacent time sections item by item, determining the repetition state according to the error tolerance threshold, and generating a section parameter repetition mark sequence;

[0016] S103: extracting time segments with consecutive non-repeated markers according to the segment parameter repeat mark sequence, screening position numbers that meet the segment length requirements, integrating them into segment start and end ranges, and generating a list of abnormally marked segments.

[0017] As a further solution of the present invention, the specific steps of S2 are:

[0018] S201: Obtain the time range in the abnormal marked segment list, extract the continuous data of the water meter, electricity meter and gas meter within the time range, divide the periodic segments by time, arrange the periodic segment data of the medium in sequence, and generate a periodic data vector sequence;

[0019] S202: Based on the periodic data vector sequence, differential identification is performed on the data changes of the medium in the continuous periodic segments, classification is performed based on the change amplitude and direction, periodic change characteristics are extracted, and periodic change trends of the medium are obtained;

[0020] S203: Based on the periodic change trend of the medium, identifying the combination characteristics of the medium change within the period, collecting the segment information with the same change trend in consecutive periods, defining the trend attribution category, and generating an offset identification structure.

[0021] As a further solution of the present invention, the specific steps of S3 are:

[0022] S301: Calling the single-point response parameters and time-series change parameters in the offset recognition structure, detecting the response trend direction of the medium combination at the same time point, extracting the combination with the opposite direction and the maximum displacement amplitude, establishing the corresponding relationship between the combination and the time point, and obtaining the direction conflict maximum value matrix;

[0023] S302: Extracting the structural response signal value within the corresponding time period based on the time point sequence in the directional conflict maximum value matrix, selecting a set of time segments with continuously increasing amplitudes and the largest coverage, converting them into structural offset area time segments, and obtaining the structural offset area time interval value;

[0024] S303: Call the media combination types covered by the time interval value of the structure offset area, count the proportion of time periods in the same range, select the time period with the largest proportion and perform number matching, build a time dimension index table, and obtain a time dimension conversion solution table.

[0025] As a further solution of the present invention, the specific calculation formula for the proportion of the number of the same type in the statistical time period is:

[0026]

[0027] Calculate the quantity ratio parameter, select the time period with the largest ratio and match the numbers, build a time dimension index table, and obtain the time dimension conversion solution table;

[0028] Among them, H is the quantity ratio parameter, τ represents the time interval number, n represents the total number of time periods, ω τ Represents the weight coefficient of the medium combination type in the τth time period, μ τ represents the number of medium combinations covered in the τth time period, γ represents the structural offset zone number, m represents the total number of offset zones, and λ γ represents the normalized time span of the γth offset zone, α τ represents the starting time offset of the τth time period, β τ represents the end time offset of the τth time period, σ τ represents the benchmark time threshold of the τth time period, θ τ Represents the actual time fluctuation in the τth time period.

[0029] As a further solution of the present invention, the specific steps of S4 are:

[0030] S401: Based on the time period in the time dimension conversion solution form, extract the rating level information in the corresponding record, identify the segments that continuously maintain the same level in chronological order, determine the start and end time of each segment, divide the segment range, and generate the level duration interval value;

[0031] S402: calling the level duration interval value, comparing the score level differences between adjacent segments, identifying the direction characteristics of the level change, and extracting the duration ratio between the segments to generate the level change trend and time ratio coefficient;

[0032] S403: Constructing a time sequence graphic identifier of the scoring segment according to the level duration interval value, the level change trend and the time ratio coefficient, marking the level change trend and the time ratio change form, and generating a continuous scoring segment difference trend map.

[0033] As a further solution of the present invention, the specific steps of S5 are:

[0034] S501: Calling the jump record in the continuous scoring segment difference trend map, extracting the scoring level and scoring time information before and after the jump, combining the medium type identifier, constructing the corresponding relationship between the score level change, duration and medium type of the jump segment, calculating the jump correlation feature value, and generating a level jump correlation value set;

[0035] S502: Forming a rating level change rate for each segment based on the rating level change and duration information in the level jump associated value set, merging the rating level change rate with the medium type identifier, and obtaining a rating level change rate set;

[0036] S503: Based on the rating level change rate set, the rating level change rates under different media types are classified, and the rating level changes and media type identifiers are integrated to generate a multi-dimensional intelligent perception tag set.

[0037] As a further solution of the present invention, the specific calculation formula for calculating the jump correlation characteristic value is:

[0038]

[0039] Among them, A is the jump correlation characteristic value, ΔR represents the absolute difference between the rating levels before and after the jump, T represents the duration of the jump segment, M represents the weight coefficient corresponding to the medium type identifier, α represents the normalization adjustment factor calculated based on the standard deviation of historical data, and Δt k Represents the time offset of the kth sampling point in the jump segment, and n represents the total number of sampling points in the jump segment.

[0040] An ESG intelligent perception system integrated with the Internet of Things, including:

[0041] The energy consumption duplication identification module obtains hourly energy consumption data of power distribution nodes in the office area, and extracts the maximum power, minimum power, and average power of each segment based on the equal-length time segment division method. Based on the numerical relationship between the parameter vector of the current time segment and the parameter vector of the previous time segment, it determines whether the parameters simultaneously meet the same conditions. If not, the corresponding time segment is recorded and aggregated to generate a list of abnormally marked segments.

[0042] The medium feature collection module calls the time information in the abnormal annotated segment list to extract representative parameters of water meters, electricity meters, and gas meters in the same time period, constructs parameter vectors according to the period, classifies the change direction and amplitude characteristics of the parameters between consecutive periods, and generates an offset identification structure based on the change attributes of the differential medium collection;

[0043] The structural offset identification module uses the change characteristics in the offset identification structure to identify the water, electricity, and gas medium combinations with opposite directions and the highest amplitude levels at the same time point, extracts the corresponding time periods and performs quantity statistics, selects the time periods with the highest proportion of quantities in the office area, and generates a time dimension conversion plan table;

[0044] The scoring trend analysis module calls the time period information in the time dimension conversion solution table, extracts the scoring level of the corresponding segment, divides the time period that continuously maintains the same level, records the level and duration, analyzes the level change trend and duration ratio between adjacent scoring segments, and generates a trend map of differences between consecutive scoring segments;

[0045] The label set generation module calls the jump records in the continuous scoring segment difference trend map, extracts the score level changes, duration differences and associated medium types, integrates them into multi-dimensional field information, constructs a label structure associated with water, electricity and gas media, and generates a multi-dimensional intelligent perception label set.

[0046] Compared with the prior art, the advantages and positive effects of the present invention are:

[0047] In the present invention, the repeatability analysis and abnormal segment extraction of energy consumption data are used to achieve accurate positioning of non-periodic fluctuations. The periodic vector construction and change feature identification based on multi-media parameters enhance the cross-media linkage perception and abnormal collaborative identification capabilities. Through trend segmentation and scoring trajectory extraction, the evolutionary expression in the time dimension is constructed, and the sensitive identification efficiency of abnormal trends is improved. Finally, the multi-dimensional jump information is structured and integrated to form a high-precision composite label system, which enhances the monitoring adaptability and data analysis capabilities in multiple scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 Schematic diagram of the steps of the present invention;

[0049] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0051] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0052] See also Figure 1 , an ESG intelligent perception method integrating the Internet of Things, comprising the following steps:

[0053] S1: Obtain energy consumption data of power distribution nodes in the office area, divide it into equal time segments, extract maximum, minimum, and average parameters, analyze the parameter repeatability of adjacent time segments, extract data locations with consecutive lack of duplicate items, and generate a list of abnormal marked segments;

[0054] S2: Based on the time information in the anomaly annotated section list, the representative parameters of the three media types of water, electricity, and gas at the same time are extracted, and a periodic parameter vector is constructed. The direction and amplitude of parameter changes between periods are identified, and the change characteristics of the media are summarized to generate an offset identification structure.

[0055] S3: Call the offset recognition structure to identify the change characteristics in the structure, identify the medium combination with the opposite change direction and the maximum amplitude at the same time point, extract the time period as the structural offset area, select the time period with the highest proportion in the same range, and generate a time dimension conversion plan table;

[0056] S4: Based on the time period in the time dimension conversion plan form, extract the rating level information, divide the time period that maintains the same level continuously, record the level and duration of the rating segment, analyze the level change trend and time ratio relationship between adjacent rating segments, and generate a trend map of the difference between consecutive rating segments;

[0057] S5: Call the jump records in the continuous scoring segment difference trend map, extract the score level changes, duration differences and corresponding media types, integrate them into structured label fields, and generate a multi-dimensional intelligent perception label set.

[0058] The list of abnormal marked sections includes maximum value parameters, minimum value parameters, average value parameters, and data location information. The offset identification structure includes periodic parameter vectors, parameter change direction classification, parameter change amplitude classification, and medium change feature set. The time dimension conversion scheme table includes structural offset area time periods, medium combination types, and time proportion statistics. The continuous scoring section difference trend map includes scoring level records, duration records, level change trend analysis, and time proportion relationship data. The multi-dimensional intelligent perception label set includes scoring level change labels, duration difference labels, and medium type identification fields.

[0059] See also Figure 1 , the specific steps of S1 are:

[0060] S101: Obtain energy consumption data of power distribution nodes in the office area, divide the data into equal time segments, extract energy consumption records in each segment, perform maximum, minimum, and average value extraction operations on each segment of data, and generate a segment energy consumption parameter set;

[0061] To obtain energy consumption information for each node in the office area power distribution system, each node connected to a smart meter must be numbered and located. Each meter uploads its current power value, node ID, and time information to a database once a minute, forming a continuous energy consumption data set. This data is then divided into 60-minute time periods, such as 8:00 to 9:00 a.m., 9:00 to 10:00 a.m., and so on. Each period contains 60 power records. After extracting these records, energy consumption statistics are processed within each period to obtain the maximum and minimum power values, as well as the average value. For example, in one period, the power values ​​range from 2.3 kW to 3.1 kW. The maximum value is 3.1 kW, the minimum is 2.3 kW, and the average value is 2.65 kW. The statistical results for each period are recorded sequentially to form a set of segment energy consumption parameters for subsequent judgment and identification operations.

[0062] S102: Calling the data in the section energy consumption parameter set, comparing the parameters of each group of adjacent time sections item by item, determining the repetition status according to the error tolerance threshold, and generating a section parameter repetition mark sequence;

[0063] Within the segment energy consumption parameter set, each set of adjacent time periods is compared, and the differences between the maximum, minimum, and average values ​​of each pair of segments are examined. If the difference in maximum and minimum values ​​between adjacent segments within a set of segments does not exceed 0.1kW, and the difference in average values ​​is within 0.05kW, the data for that set is considered duplicated. Otherwise, the data is considered discrepant. For example, if the maximum value of one segment is 3.1kW, it differs by 0.1kW from the maximum value of the next segment, 3.2kW, while the minimum value changes from 2.3kW to 2.1kW, a difference of 0.2kW, and the average value drops from 2.65kW to 2.60kW, a difference of 0.05kW. Because the difference in minimum values ​​exceeds the set range, the two segments are considered non-duplicate. This continuous comparison of all segments generates a sequence representing duplicates, with duplicates marked as 1 and non-duplicates marked as 0. For example, the sequence might be "0, 1, 0, 1, 1, 0."

[0064] S103: Extracting time segments with consecutive non-repeated markers based on the segment parameter repeat mark sequence, selecting position numbers that meet the segment length requirements, integrating them into segment start and end ranges, and generating a list of abnormally marked segments;

[0065] Through the above-mentioned marking sequence, find the part with continuous markings of 0, that is, the time range of continuous non-repeating segments, and set the judgment standard to at least two consecutive segments to be considered an abnormal segment. If the 3rd to 4th bits in the marking sequence are both 0, it means that there are continuous differences in the parameters of the two segments. Combined with the time sequence of the segments, the specific time range can be located. For example, if the 3rd segment is from 9:00 to 10:00 and the 4th segment is from 10:00 to 11:00, then the abnormal segment time is determined to be from 9:00 to 11:00. Further summarize the energy consumption parameters of all nodes in this time period. For example, the maximum power of node B in this segment is 3.5kW, the minimum power is 2.0kW, and the average power is 2.7kW. Finally, a list of abnormal segments is formed, including the start time, end time, node number, various power values, and status identification of whether it is an abnormal segment. This list can be used as the input basis for subsequent processing links.

[0066] See also Figure 1 , the specific steps of S2 are:

[0067] S201: Obtain the time range in the abnormal marked segment list, extract the continuous data of the water meter, electricity meter and gas meter within the time range, divide the periodic segments by time, arrange the periodic segment data of the medium in sequence, and generate a periodic data vector sequence;

[0068] When obtaining the time range of the abnormal marked segment, the system should first mark and filter the original data in the water meter, electricity meter and gas meter, and filter out the abnormal segments according to abnormal rules such as sudden changes, long-term constant values, data interruptions or illogical values, and extract the corresponding start and end times to form a time range set. For example, if the water meter data marked as from May 1 to May 7, 2019 is an abnormal segment, the continuous data of the water meter numbered W001 within this time period must be extracted. The same is true for electricity and gas meters. If there is a missing data record on a certain day, it should be supplemented by interpolation or forward filling in combination with the sampling frequency to form a complete time period data set. The time period is then divided into periodic segments, such as 1 day, 6 hours, or 1 hour. The period length is selected according to the actual sampling frequency. Continuous data from the water meter, electricity meter, and gas meter are extracted in each periodic segment to form a medium data vector. If the period is 1 hour, each day is divided into 24 periodic segments, and each data vector is arranged in sequence. For example, from 8:00 to 9:00 on May 1, 2019, the water meter record is 3.5, the electricity meter record is 1.2, and the gas meter record is 0.8, forming a periodic data vector of [3.5, 1.2, 0.8]. Multiple periodic vectors are then arranged in chronological order to generate a complete periodic data vector sequence for subsequent processing.

[0069] S202: Based on the periodic data vector sequence, the data changes of the medium in the continuous periodic segments are differentially identified, classified according to the change amplitude and direction, and the periodic change characteristics are extracted to obtain the periodic change trend of the medium;

[0070] After the periodic data vector sequence is constructed, the changes in the water, electricity, and gas meter data need to be identified period by period. The data vectors of each two adjacent period segments are compared, and the difference of each data item is extracted. For example, if the previous period is [3.5, 1.2, 0.8] and the next period is [3.6, 1.1, 0.9], the water meter data increases by 0.1, the electricity meter data decreases by 0.1, and the gas meter data increases by 0.1. These differences represent the magnitude and direction of the change. The direction of change is divided into three types: rising, falling, and stable. The magnitude is determined by a set interval: less than 0.05 is stable, between 0.05 and 0.2 is slight, and greater than 0.2 is significant. This classification method is applied uniformly to all media. The changes in each medium within each period are labeled as rising, falling, or stable. Continuous tracking of multi-period change types forms a trend sequence. For example, the classification of 10 period segments may be [rising, falling, rising, stable, falling, rising, rising, stable, falling, falling]. This sequence describes the state transition and periodic changes of the medium in the time dimension and serves as a data source for trend features.

[0071] S203: Based on the periodic change trend of the medium, identify the combined characteristics of the medium's changes within the period, collect information on segments with consistent change trends in consecutive periods, define trend categories, and generate an offset identification structure;

[0072] According to the extracted media periodic change trend sequence, it is necessary to identify its combination characteristics within the period. A fixed-length sliding window method is used to segment and combine trend data for identification. For example, if the window length is 3, the combination of every 3 trends is extracted in sequence. The extracted sequence is [rise, rise, fall], [rise, fall, stable], etc. The recurring change pattern is identified according to the arrangement of the trend sequence. The trend combination is represented by abbreviated symbols, rising as U, falling as D, and stable as S, forming patterns such as UUD, DDS, SUD, etc. The frequency of combination occurrence is counted, the main change combination characteristics are screened, and the continuity and repeatability of the change direction are combined to identify If the same combination appears more than three times in a continuous period, it is defined as a consistent trend segment, and the medium type, equipment number, trend type and start and end time are recorded. For example, if the trend combination of a water meter from 8:00 on May 1, 2019 to 8:00 on May 4, 2019 is all declining, it is marked as a "continuous decline" trend segment. This information is classified into the trend attribution category, and a structure containing fields such as type, number, trend, start and end period, and combination code is generated for classification and summary subsequent processing. For example, water, number W001, trend is declining, start and end period is from 8:00 on May 1, 2019 to 8:00 on May 4, 2019, combination code DDD.

[0073] See also Figure 1 , the specific steps of S3 are:

[0074] S301: Calling the single-point response parameters and time-series change parameters in the offset recognition structure to detect the response trend direction of the medium combination at the same time point, extracting the combination with the opposite direction and the maximum displacement amplitude, establishing the corresponding relationship between the combination and the time point, and obtaining the direction conflict maximum value matrix;

[0075] First, the response records of each measuring point at different time points within the structure must be read. For example, in a bridge monitoring scenario, each measuring point, such as P1, P2, and P3, records displacement changes over a time interval. The response values ​​of these nodes at specific time points are then extracted sequentially to analyze their changing trends. Extracting time-series variation parameters relies on a comprehensive analysis of the displacement differences between each measuring point at adjacent time points. This can be determined by observing an upward or downward displacement trend. For example, if P1's displacement records are 4.3mm, 4.7mm, and 5.1mm at time points 120s, 121s, and 122s, respectively, this can be preliminarily determined to be an upward trend. Next, for each time point, the response trend direction of multiple media combinations is extracted, such as a concrete-steel plate and concrete-composite material combination. By comparing the trend directions, it is determined whether there are opposite directions. Opposite directions can be simply understood as one group showing an upward displacement and the other showing a downward displacement. Among the combinations with opposite trends, we further screen out those with maximum displacement amplitudes. For example, at a certain time point, the displacement amplitudes of two combinations were 12.3mm and 13.1mm, respectively. If the maximum level is set to greater than 10mm, both combinations meet the requirements. These qualified combinations are paired with the corresponding time points and recorded to create an association list. For example, a certain time point is paired with the concrete-steel plate and concrete-composite materials combinations. Finally, all qualified combinations and time point correspondences are formed into a matrix for subsequent analysis and processing. The matrix uses rows to represent the time point sequence number and columns to represent the combination number. The cell values ​​can be conflict amplitudes or displacement differences, which is convenient for subsequent calls.

[0076] S302: Extracting the structural response signal value within the corresponding time period based on the time point sequence in the directional conflict maximum value matrix, selecting a set of time segments with continuously increasing amplitudes and the largest coverage, converting them into structural offset time segments, and obtaining the structural offset time interval value;

[0077] Based on the sequence of time points extracted from the matrix, the structural response signal values ​​corresponding to these time points are extracted one by one. After arranging them in chronological order, the response amplitudes of every two adjacent time points are compared. Only time segments with a clear increasing amplitude trend are retained. For example, if the response amplitudes of four consecutive time points are 5.1 mm, 6.0 mm, 6.8 mm, and 7.6 mm, this sequence is considered a valid sequence with increasing amplitudes. This is determined by the fact that the subsequent value is greater than the previous value by more than 0.5 mm. This judgment process is repeated for all time points to obtain several sets of time segments with increasing amplitudes. For each set, the coverage of the corresponding measurement point range is calculated. If it covers more than 80% of the sensor or medium combination, it is marked as a valid time segment. Among all valid time segments, the segment with the largest coverage is selected. For example, if a time segment covers 12 measurement points, accounting for 85% of the total number of measurement points, this time segment is selected as the first choice and converted into a structural offset zone time segment. Its start and end time points are recorded to form a clear time interval for subsequent calculations and indexing.

[0078] S303: Retrieving the media combination types covered by the time interval value of the structure offset area, counting the proportion of time periods in the same range, selecting the time period with the largest proportion and matching the numbers, constructing a time dimension index table, and obtaining a time dimension conversion solution table;

[0079] The specific calculation formula for the proportion of quantity in the same category during the statistical time period is:

[0080]

[0081] Calculate the quantity ratio parameter, select the time period with the largest ratio and match the numbers, build a time dimension index table, and obtain the time dimension conversion solution table;

[0082] Among them, H is the quantity ratio parameter, τ represents the time interval number, n represents the total number of time periods, ω τ Represents the weight coefficient of the medium combination type in the τth time period, μ τ represents the number of medium combinations covered in the τth time period, γ represents the structural offset zone number, m represents the total number of offset zones, and λ γ represents the normalized time span of the γth offset zone, α τ represents the starting time offset of the τth time period, β τ represents the end time offset of the τth time period, σ τ represents the benchmark time threshold of the τth time period, θ τ represents the actual time fluctuation in the τth time period;

[0083] Set the total number of time periods n = 5, the total number of offset zones m = 3, and the time interval number τ = 2;

[0084] The number of medium combinations μ2 is obtained through sensor collection, and the measured value is 12;

[0085] The weight coefficient ω2 is determined according to the time sensitivity classification table in the GB / T20234-2023 standard. When the medium combination type belongs to Class II, ω2 = 0.85;

[0086] Normalized time span λ γ Calculated from the monitoring data of the structural offset zone, when γ = 1 in the offset zone, λ1 = 8.3 hours, when γ = 2, λ2 = 6.7 hours, and when γ = 3, λ3 = 7.9 hours;

[0087] The start time offset α2 is obtained from the timestamp recording system as +0.45 hours, and the end time offset β2 is -0.23 hours;

[0088] The benchmark time threshold σ2 is set to 24.0 hours according to the ISO8601 standard, and the actual time fluctuation θ2 is calculated to be 23.6 hours through time series analysis;

[0089] Numerator calculation:

[0090] |0.85×12|=10.22;

[0091] Denominator calculation:

[0092]

[0093] Time offset calculation: (0.45 + (-0.23)) / |24.0 - 23.6| = 0.22 / 0.4 = 0.554;

[0094] Comprehensive calculation: H = (10.2 / 13.27) × 0.55 ≈ 0.768 × 0.55 ≈ 0.422;

[0095] The result shows that the parameter H for the proportion of the number of time interval τ=2 is 42.2%. This value directly corresponds to the "proportion of the number of statistical time periods in the same category" in the step result. The larger the H value, the higher the proportion of the time period in the same category. When the H value exceeds the preset threshold of 0.4, the number matching mechanism is triggered and the time dimension index table construction process begins.

[0096] See also Figure 1 , the specific steps of S4 are:

[0097] S401: Based on the time period in the time dimension conversion plan form, extract the rating level information in the corresponding record, identify the segments that continuously maintain the same level in chronological order, determine the start and end time of each segment, divide the segment range, and generate the level duration interval value;

[0098] First, the format of each record time field in the plan form is standardized. Information such as year, month, day, hour, and minute is converted to a unified time format and expressed as a numerical value, such as a timestamp in seconds or minutes. The processed records are sorted chronologically to form a time series list. Next, the rating information within each record is extracted, such as a category labeled A, B, or C. The rating value of each record is read chronologically, and the rating of each record is continuously checked to see if the current record's rating matches the previous one. If so, the current record is grouped into the same segment, and the current segment's end time is continuously updated. If not, the current segment is terminated and the segment's rating type, start time, end time, and total duration are recorded. The recording process for a new rating segment is then started, and this process is repeated until all data records have been traversed. For example, if the rating from March 1st to March 5th is B, and the rating changes to A on March 6th, the B-grade segment will last for 5 days from March 1st to March 5th, and the A-grade segment will begin on March 6th. If the grade remains A from March 7 to March 10, the duration of the current A-grade segment can be extended to March 10, forming a continuous grade segment. Ultimately, each grade segment contains information such as grade, start and end times, and duration, forming a complete grade segment sequence for the next stage of change analysis.

[0099] S402: Calling the grade duration interval value, comparing the grade differences between adjacent segments, identifying the directional characteristics of grade changes, and extracting the duration ratio between segments to generate the grade change trend and time ratio coefficient;

[0100] By continuously reading the generated grade duration segment sequence, the rating grade marks of each two adjacent segments are compared in chronological order, and numerical labels are assigned according to the grade type. For example, A is assigned a value of 1, B is 2, C is 3, etc. The direction of change can be determined by subtracting the grade value of the previous segment from the grade value of the latter segment. If the result is positive, it means the grade has increased, a negative value means it has decreased, and zero means it remains the same. At the same time, the duration of each of the current two segments is extracted. For example, if the first segment is 4 days and the latter segment is 6 days, the total duration is 10 days, and the latter segment accounts for 60%. This ratio reflects the relative importance of the duration of the latter grade in the grade change. The entire segment sequence is processed in this way, and a set of grade change trend information is obtained, including the direction of each change, the numerical size, and the proportion of the duration in the entire change, which is used to characterize the change trajectory and rhythm strength of the overall score grade.

[0101] S403: constructing a time sequence graphic identifier for the scoring segment based on the grade duration interval value, the grade change trend, and the time ratio coefficient, marking the grade change trend and the time ratio change form, and generating a continuous scoring segment difference trend map;

[0102] A grade rating segment graphic is constructed with time as the horizontal axis and the grade numerical identifier as the vertical axis. From left to right, the rating segments are plotted as rectangles or bars. The length of each segment corresponds to the duration of the grade, and the height indicates the grade level. Adjacent segments are connected by broken lines to form a trend line that indicates the direction of grade change. The trend line is accompanied by a numerical description of the change, such as +1 for an increase of one level and -1 for a decrease of one level. The duration ratio between each pair of segments is also marked in the graph, such as 40% or 70%, reflecting the weight of the time occupied by the grade after the change. During the graphic construction process, the order of the segments is maintained to ensure continuity, and the different grades are distinguished by color or fill to enhance the graphic's recognizability. The overall graphic structure clearly shows the duration and direction of change of different rating grades over time, providing a visual reference for the grade change process.

[0103] See also Figure 1 , the specific steps of S5 are:

[0104] S501: Calling the jump record in the continuous scoring segment difference trend map, extracting the scoring level and scoring time information before and after the jump, combining it with the medium type identifier, constructing the corresponding relationship between the score level change, duration and medium type of the jump segment, calculating the jump correlation feature value, and generating a level jump correlation value set;

[0105] The specific calculation formula for calculating the jump correlation characteristic value is:

[0106]

[0107] Among them, A is the jump correlation characteristic value, ΔR represents the absolute difference of the score level before and after the jump (ΔR=|R 后 -R 前 |), T represents the duration of the jump segment (unit: seconds), M represents the weight coefficient corresponding to the medium type identifier (obtained by looking up the table), and α represents the normalization adjustment factor calculated based on the standard deviation of historical data (α=σ 历史 / σ 当前 ), Δtk represents the time offset of the kth sampling point in the jump segment, and n represents the total number of sampling points in the jump segment;

[0108] ΔR is monitored by the scoring system to obtain the grade data before and after the jump. In a certain jump record, R 后 =Level 5, R 前 =3 level, ΔR=5-3=2;

[0109] T is calculated by the timestamp difference, and the jump start time t 始 =2025-05-2710:00:00, end time t 终 =2025-05-27 10:30:00, T=1800 seconds;

[0110] M is based on the media type quantization table. When the media type is video stream, M = 1.2 (weight table definition: video stream 1.2, audio stream 1.0, text stream 0.8);

[0111] α is calculated by the standard deviation of historical data σ 历史 =0.5 and the current data standard deviation σ 当前 =0.4 calculation, α

[0112] =0.5 / 0.4=1.25;

[0113] Δtk is monitored by sampling point time offset. The offsets of the three sampling points in the jump segment are Δt1 = 5 seconds, Δt2 = 10 seconds, Δt3 = 15 seconds, and n = 3.

[0114] Formula calculation process:

[0115] First calculation:

[0116]

[0117] The second summation is calculated as:

[0118]

[0119] Final result:

[0120] A=70.71+0.0208=70.7308;

[0121] Parameter setting basis:

[0122] M=1.2 is based on the media type quantization table, which defines the video stream as having the highest weight on the score jump based on historical data analysis; α=1.25 reflects that the fluctuation amplitude of historical data is 25% higher than that of current data, which is used to balance the data distribution difference; Δt_k=5 / 10 / 15 seconds comes from the timestamp recording accuracy of the actual monitoring equipment (±0.1 second).

[0123] Explanation of numerical results:

[0124] A=70.7308 is the jump correlation characteristic value, which is obtained by quantifying the comprehensive effect of the rating level change amplitude, time duration and medium weight, and superimposing the normalized cumulative effect of the time offset. The final result is included in the level jump correlation value set as one-dimensional data in the multidimensional characteristic vector.

[0125] S502: Based on the rating level change and duration information in the level jump association value set, a rating level change rate for each segment is generated, and the rating level change rate and the medium type identifier are merged to obtain a rating level change rate set;

[0126] Based on the constructed jump data entries, the score level change rate is calculated by the ratio between the score level change value and the time difference. First, the start and end values ​​of the score level in each record are extracted, and then the score duration is obtained. The score rate is obtained by dividing the score level difference by the score duration. The rate result unit is level per minute. After the rate calculation is completed, it is combined with the medium type information to form a score rate information entry. If there are multiple score rate entries under the same type of medium, these entries are classified and summarized. The summary method is to take the average of all score rates under this type of medium to form an overall reference value representing the score level change rate of the medium. For example, there are 5 score jump records under the metal coating medium, and the corresponding rates are 0.2, 0.25, 0.3, 0.35 and 0.4 levels per minute respectively. The average score rate of the medium is the sum of the 5 values ​​divided by 5, which is 0.3 levels per minute. Finally, a collection of different medium types and corresponding score change rates is formed, laying a data foundation for trend analysis.

[0127] S503: Based on the rating level change rate set, classify the rating level change rates under different media types, integrate the rating level changes and the media type identifier, and generate a multi-dimensional intelligent perception label set;

[0128] After the score change rate set is established, the score rates need to be grouped and analyzed by medium type. The concentration of the rate distribution is calculated for each data set. The judgment is based on the deviation range of each rate value relative to the average. Distributions with a deviation less than 0.05 levels per minute are considered concentrated and marked as stable. Distributions with a deviation greater than 0.2 levels per minute are considered dispersed and marked as fluctuating. At the same time, trends are divided according to the absolute value of the score rate itself. Rates less than 0.1 are classified as slowly varying, between 0.1 and 0.3 as moderately varying, and greater than or equal to 0.3 as rapidly varying. The final label for the medium type score change trend is composed of the concentration label and the rate range label. For example, if the average score rate of a polymer layer material is 0.28 levels per minute and the data deviation is 0.03 levels per minute, the label is moderately varying-stable. All medium types are classified according to this rule, and the results are unified to form a trend dispersion label dataset.

[0129] See also Figure 2 , an ESG intelligent perception system integrated with the Internet of Things, including:

[0130] The energy consumption duplication identification module obtains hourly energy consumption data of power distribution nodes in the office area, and extracts the maximum power, minimum power, and average power of each segment based on the equal-length time segment division method. Based on the numerical relationship between the parameter vector of the current time segment and the parameter vector of the previous time segment, it determines whether the parameters simultaneously meet the same conditions. If not, the corresponding time segment is recorded and aggregated to generate a list of abnormally marked segments.

[0131] The medium feature collection module uses the time information in the abnormal annotated segment list to extract representative parameters of water, electricity, and gas meters within the same time period. It constructs parameter vectors based on the period, classifies the direction and amplitude characteristics of parameter changes between consecutive periods, and generates an offset identification structure based on the change attributes of the differential medium.

[0132] The structural offset identification module uses the change characteristics in the offset identification structure to identify the water, electricity, and gas medium combinations with opposite directions and the highest amplitude levels at the same time point. It then extracts the corresponding time periods and performs quantity statistics, screening the time periods with the highest proportion of quantities in the office area and generating a time dimension conversion plan table.

[0133] The scoring trend analysis module calls the time period information in the time dimension conversion plan table, extracts the scoring level of the corresponding segment, divides the time period that maintains the same level continuously, records the level and duration, analyzes the level change trend and duration ratio between adjacent scoring segments, and generates a trend map of the difference between consecutive scoring segments;

[0134] The label set generation module calls the jump records in the continuous scoring segment difference trend map, extracts the score level changes, duration differences and related medium types, integrates them into multi-dimensional field information, constructs the label structure associated with water, electricity and gas media, and generates a multi-dimensional intelligent perception label set.

[0135] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. An ESG intelligent perception method integrating the Internet of Things, characterized by: The following steps are involved: S1: Obtain energy consumption data of power distribution nodes in the office area, divide it into equal time segments, extract the maximum, minimum and average values, analyze the repeatability of parameters in adjacent segments, and generate a list of abnormal marked segments. S2: According to the anomaly marked section time, the representative parameters of the three types of media, water, electricity, and gas, within the same time are extracted, a periodic parameter vector is constructed, the direction and amplitude classification of the parameter changes between cycles are identified, and an offset identification structure is generated. S3: Call the change characteristics in the offset identification structure to identify the medium combination with the opposite change direction and the largest amplitude at the same time point, extract the corresponding time period as the structural offset area, select the time period with the highest proportion among the same type, and generate a time dimension conversion solution table. S4: Based on the time periods in the conversion scheme table, extract the scoring levels, divide the segments with the same continuous level, record the levels and durations, analyze the relationship between the level change trends and time ratios of adjacent scoring segments, and generate a continuous scoring segment difference trend map. S5: calling the jump record in the continuous scoring segment difference trend map, extracting the score level change, duration difference and medium type, integrating them into structured tag fields, and generating a multi-dimensional intelligent perception tag set.

2. The ESG intelligent perception method for integrating the Internet of Things according to claim 1 is characterized in that: The abnormal marked segment list includes maximum value parameters, minimum value parameters, average value parameters, and data location information; the offset identification structure includes periodic parameter vectors, parameter change direction classifications, parameter change amplitude classifications, and medium change feature sets; the time dimension conversion scheme table includes structural offset area time periods, medium combination types, and time proportion statistical information; the continuous scoring segment difference trend map includes scoring level records, duration records, level change trend analysis, and time proportion relationship data; the multidimensional intelligent perception tag set includes scoring level change tags, duration difference tags, and medium type identification fields.

3. The ESG intelligent perception method for integrating the Internet of Things according to claim 1 is characterized in that: The specific steps of S1 are: S101: Obtain energy consumption data of power distribution nodes in the office area, divide the data into equal time segments, extract energy consumption records in each segment, perform maximum, minimum, and average value extraction operations on each segment of data, and generate a segment energy consumption parameter set; S102: calling the data in the section energy consumption parameter set, comparing the parameters of each group of adjacent time sections item by item, determining the repetition state according to the error tolerance threshold, and generating a section parameter repetition mark sequence; S103: extracting time segments with consecutive non-repeated markers according to the segment parameter repeat mark sequence, screening position numbers that meet the segment length requirements, integrating them into segment start and end ranges, and generating a list of abnormally marked segments.

4. The ESG intelligent perception method for integrating the Internet of Things according to claim 3 is characterized in that: The specific steps of S2 are: S201: Obtain the time range in the abnormal marked segment list, extract the continuous data of the water meter, electricity meter and gas meter within the time range, divide the periodic segments by time, arrange the periodic segment data of the medium in sequence, and generate a periodic data vector sequence; S202: Based on the periodic data vector sequence, differential identification is performed on the data changes of the medium in the continuous periodic segments, classification is performed based on the change amplitude and direction, periodic change characteristics are extracted, and periodic change trends of the medium are obtained; S203: Based on the periodic change trend of the medium, identifying the combination characteristics of the medium change within the period, collecting the segment information with the same change trend in consecutive periods, defining the trend attribution category, and generating an offset identification structure.

5. The ESG intelligent perception method for integrating the Internet of Things according to claim 4 is characterized in that: The specific steps of S3 are: S301: Calling the single-point response parameters and time-series change parameters in the offset recognition structure, detecting the response trend direction of the medium combination at the same time point, extracting the combination with the opposite direction and the maximum displacement amplitude, establishing the corresponding relationship between the combination and the time point, and obtaining the direction conflict maximum value matrix; S302: Extracting the structural response signal value within the corresponding time period based on the time point sequence in the directional conflict maximum value matrix, selecting a set of time segments with continuously increasing amplitudes and the largest coverage, converting them into structural offset area time segments, and obtaining the structural offset area time interval value; S303: Call the media combination types covered by the time interval value of the structure offset area, count the proportion of time periods in the same range, select the time period with the largest proportion and perform number matching, build a time dimension index table, and obtain a time dimension conversion solution table.

6. The ESG intelligent perception method for integrating the Internet of Things according to claim 5 is characterized in that: The specific calculation formula for the proportion of the number of similar items in the statistical time period is: Calculate the quantity ratio parameter, select the time period with the largest ratio and match the numbers, build a time dimension index table, and obtain the time dimension conversion solution table; Among them, H is the quantity ratio parameter, τ represents the time interval number, n represents the total number of time periods, ω τ Represents the weight coefficient of the medium combination type in the τth time period, μ τ represents the number of medium combinations covered in the τth time period, γ represents the structural offset zone number, m represents the total number of offset zones, and λ γ represents the normalized time span of the γth offset zone, α τ represents the starting time offset of the τth time period, β τ represents the end time offset of the τth time period, σ τ represents the benchmark time threshold of the τth time period, θ τ Represents the actual time fluctuation in the τth time period.

7. The ESG intelligent perception method for integrating the Internet of Things according to claim 5 is characterized in that: The specific steps of S4 are: S401: Based on the time period in the time dimension conversion solution form, extract the rating level information in the corresponding record, identify the segments that continuously maintain the same level in chronological order, determine the start and end time of each segment, divide the segment range, and generate the level duration interval value; S402: calling the level duration interval value, comparing the score level differences between adjacent segments, identifying the direction characteristics of the level change, and extracting the duration ratio between the segments to generate the level change trend and time ratio coefficient; S403: Constructing a time sequence graphic identifier of the scoring segment according to the level duration interval value, the level change trend and the time ratio coefficient, marking the level change trend and the time ratio change form, and generating a continuous scoring segment difference trend map.

8. The ESG intelligent perception method for integrating the Internet of Things according to claim 7 is characterized in that: The specific steps of S5 are: S501: Calling the jump record in the continuous scoring segment difference trend map, extracting the scoring level and scoring time information before and after the jump, combining the medium type identifier, constructing the corresponding relationship between the score level change, duration and medium type of the jump segment, calculating the jump correlation feature value, and generating a level jump correlation value set; S502: Forming a rating level change rate for each segment based on the rating level change and duration information in the level jump associated value set, merging the rating level change rate with the medium type identifier, and obtaining a rating level change rate set; S503: Based on the rating level change rate set, the rating level change rates under different media types are classified, and the rating level changes and media type identifiers are integrated to generate a multi-dimensional intelligent perception tag set.

9. The ESG intelligent perception method for integrating the Internet of Things according to claim 8, characterized in that: The specific calculation formula for calculating the jump correlation characteristic value is: Where A is the jump-associated characteristic value, ΔR represents the absolute difference in the rating levels before and after the jump, T represents the duration of the jump segment, M represents the weight coefficient corresponding to the medium type identifier, α represents the normalization adjustment factor calculated based on the standard deviation of historical data, Δtk represents the time offset of the kth sampling point in the jump segment, and n represents the total number of sampling points in the jump segment.

10. An ESG intelligent perception system integrated with the Internet of Things, characterized by: According to the ESG intelligent perception method integrated with the Internet of Things according to any one of claims 1 to 9, the system comprises: The energy consumption duplication identification module obtains hourly energy consumption data of power distribution nodes in the office area, and extracts the maximum power, minimum power, and average power of each segment based on the equal-length time segment division method. Based on the numerical relationship between the parameter vector of the current time segment and the parameter vector of the previous time segment, it determines whether the parameters simultaneously meet the same conditions. If not, the corresponding time segment is recorded and aggregated to generate a list of abnormally marked segments. The medium feature collection module calls the time information in the abnormal annotated segment list to extract representative parameters of water meters, electricity meters, and gas meters in the same time period, constructs parameter vectors according to the period, classifies the change direction and amplitude characteristics of the parameters between consecutive periods, and generates an offset identification structure based on the change attributes of the differential medium collection; The structural offset identification module uses the change characteristics in the offset identification structure to identify the water, electricity, and gas medium combinations with opposite directions and the highest amplitude levels at the same time point, extracts the corresponding time periods and performs quantity statistics, selects the time periods with the highest proportion of quantities in the office area, and generates a time dimension conversion plan table; The scoring trend analysis module calls the time period information in the time dimension conversion solution table, extracts the scoring level of the corresponding segment, divides the time period that continuously maintains the same level, records the level and duration, analyzes the level change trend and duration ratio between adjacent scoring segments, and generates a trend map of differences between consecutive scoring segments; The label set generation module calls the jump records in the continuous scoring segment difference trend map, extracts the score level changes, duration differences and associated medium types, integrates them into multi-dimensional field information, constructs a label structure associated with water, electricity and gas media, and generates a multi-dimensional intelligent perception label set.

Images