Real-time intelligent monitoring method and system for park carbon emission based on big data collection

Abstract

The present application relates to the technical field of park carbon monitoring, and discloses a park carbon emission real-time intelligent monitoring method and system based on big data collection. The method comprises continuously capturing dynamic emission data streams of various carbon emission sources in the park through a distributed sensing node group, synchronously collecting production activity and timing data and environmental state timing data; performing instantaneous feature enhancement processing on the dynamic emission data streams, extracting a multi-dimensional carbon emission intensity sequence containing timing correlation, and generating a park carbon emission dynamic portrait; based on the above sequence and production activity timing data, constructing a multi-level carbon emission quantification framework containing different carbon emission aggregation granularity levels, calculating emission anomaly concentration indexes of each level; generating a current carbon emission state summary and mapping it to the corresponding level, calculating the distribution similarity to obtain a real-time carbon emission online deviation index, generating a hierarchical early warning signal and a regulation strategy sequence accordingly, and realizing real-time and accurate monitoring and effective regulation of park carbon emission.

Description

Technical Field

[0001] This invention relates to the field of carbon monitoring technology in industrial parks, specifically to a real-time intelligent monitoring method and system for carbon emissions in industrial parks based on big data collection. Background Technology

[0002] With increasing global attention to climate change, industrial parks, as key areas for industrial production, commercial operations, and residential life, require precise control of their carbon emissions as a crucial step in achieving dual-carbon goals. Currently, traditional carbon emission monitoring in industrial parks relies heavily on periodic manual sampling and laboratory analysis. This approach not only suffers from long data collection cycles, making it difficult to reflect the dynamic changes in carbon emissions in real time, but is also susceptible to human influence during data collection, compromising the accuracy and reliability of the data.

[0003] While some industrial parks have introduced sensing devices for carbon emission data collection, most use single-type sensor nodes, which can only acquire carbon emission data from a specific source or a specific dimension. This fails to comprehensively capture emission information from all types of carbon sources within the park, resulting in a lack of comprehensiveness and systematicity in subsequent carbon emission analysis and assessment based on this data. Furthermore, current technologies often only perform simple filtering and statistical analysis on the collected carbon emission data, failing to fully explore the temporal correlations and multi-dimensional characteristics inherent in the data. This makes it difficult to generate a dynamic profile that accurately reflects the actual carbon emission situation in the park, leaving park managers unable to clearly and intuitively understand the changing trends and key influencing factors of carbon emissions.

[0004] In terms of carbon emission quantification, existing technologies lack a scientific and reasonable multi-level quantification framework. They cannot perform quantitative analysis of carbon emissions at different granularities based on varying management needs and application scenarios, resulting in weak practicality and specificity of the quantification results. Furthermore, existing technologies are significantly inadequate in monitoring and early warning of carbon emission anomalies. Most can only conduct post-event analysis and summarization after anomalies occur, failing to identify anomalies in real time and issue timely warnings. They also struggle to formulate scientific and effective control strategies based on anomalies, leading to a passive approach to carbon emission management in industrial parks and hindering timely measures to control total carbon emissions. This is detrimental to the low-carbon operation and sustainable development of industrial parks. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for real-time intelligent monitoring of carbon emissions in industrial parks based on big data collection, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, this invention provides a real-time intelligent monitoring method for carbon emissions in industrial parks based on big data collection, the method comprising:

[0007] The system continuously captures dynamic emission data streams from various carbon emission sources within the park through a distributed sensor node cluster, and simultaneously collects corresponding production activity time-series data and environmental status time-series data.

[0008] Instantaneous feature enhancement processing is performed on the dynamic emission data stream to extract a multi-dimensional carbon emission intensity sequence with temporal correlation, and a dynamic carbon emission profile of the park is generated.

[0009] Based on multi-dimensional carbon emission intensity sequences and production activity time series data, a multi-level carbon emission quantification framework reflecting the park's operational status is constructed, with each level corresponding to a different carbon emission aggregation granularity.

[0010] Within the multi-level carbon emission quantification framework, based on the data distribution patterns at the time of historical carbon emission events, the emission anomaly concentration index at each level is calculated.

[0011] A summary of the current carbon emission status is generated based on the real-time collected dynamic emission data stream, and then mapped to the corresponding level within the multi-level carbon emission quantification framework.

[0012] Calculate the distribution similarity between the current carbon emission status summary and the historical abnormal emission concentration index of this level, and generate a real-time online carbon emission deviation index;

[0013] Based on real-time online deviation indicators of carbon emissions, a tiered early warning signal and a sequence of control strategies are generated.

[0014] Preferably, the step of generating a tiered early warning signal and control strategy sequence based on real-time online carbon emission deviation indicators includes:

[0015] An adaptive quantitative mapping model is constructed by integrating real-time online carbon emission deviation indicators, multi-dimensional carbon emission intensity sequences, and environmental status time series data.

[0016] The short-term dynamic trajectory of carbon emissions in the park is extrapolated by an adaptive quantization mapping model, and the evolution path of carbon emission intensity within the future time window is output.

[0017] Based on the matching results between the carbon emission intensity evolution path and the preset carbon emission control boundary conditions, a tiered early warning signal and control strategy sequence are generated.

[0018] Preferably, a distributed sensor node cluster continuously captures dynamic emission data streams from various carbon emission sources within the park, and simultaneously collects corresponding production activity time-series data and environmental status time-series data, specifically including:

[0019] Multiple types of sensor groups are configured and distributed in different areas of the park to collect carbon emission concentration data, energy consumption data and material flow data at a fixed sampling frequency;

[0020] The raw data streams collected by multiple types of sensor groups are aggregated through IoT key points, and the data streams are timestamped and bound to source identifiers.

[0021] Synchronously integrate equipment operation status sequences, work plan logs, and capacity output indicators from the production management system to form time-series data of production activities;

[0022] Simultaneously, environmental sensor data on temperature, humidity, wind speed, and air pressure are collected to construct time-series data on environmental conditions.

[0023] Preferably, the dynamic emission data stream undergoes instantaneous feature enhancement processing to extract a multi-dimensional carbon emission intensity sequence with temporal correlation, and a dynamic carbon emission profile of the park is generated, specifically including:

[0024] A sliding time window algorithm is used to enhance the local features of dynamic emission data streams, highlighting transient emission peaks and continuous emission baselines in the data streams.

[0025] Independent intensity sequences corresponding to different carbon emission sources were separated from the enhanced data stream, including direct emission intensity sequences and indirect emission intensity sequences;

[0026] The independent intensity sequences were integrated along the time axis into a multi-dimensional carbon emission intensity sequence with spatial identifiers.

[0027] Based on the spatial distribution and temporal fluctuation characteristics of multi-dimensional carbon emission intensity sequences, a dynamic profile of carbon emissions in the park is generated.

[0028] Preferably, based on multi-dimensional carbon emission intensity sequences and production activity time-series data, a multi-level carbon emission quantification framework reflecting the park's operational status is constructed, where each level corresponds to a different carbon emission aggregation granularity, specifically including:

[0029] Based on the physical distribution and logical connections of carbon emission sources within the park, a carbon emission aggregation hierarchy is defined from the equipment level, workshop level to the park level.

[0030] For each aggregation level, extract the corresponding carbon emission intensity subsequence from the multi-dimensional carbon emission intensity sequence;

[0031] Extract a series of production efficiency indicators that match each aggregation level from the time-series data of production activities;

[0032] By integrating the carbon emission intensity subsequence and the production efficiency indicator sequence, a carbon emission operation status vector is formed under each aggregation level;

[0033] A multi-level carbon emission quantification framework is constructed based on the carbon emission operational status vectors at all levels.

[0034] Preferably, within the multi-level carbon emission quantification framework, based on the data distribution patterns at the time of historical carbon emission events, the emission anomaly concentration index at each level is calculated, specifically including:

[0035] Retrieve historical records of abnormal carbon emission events and multi-dimensional carbon emission intensity sequences at the time of the events;

[0036] Extract the set of carbon emission operation status vectors corresponding to each aggregation level when an abnormal carbon emission event occurs;

[0037] Calculate the distribution density and distribution offset of the carbon emission operating state vector in this set;

[0038] The calculation rules for the emission anomaly concentration index at each level are determined based on the distribution density and distribution offset.

[0039] The historical emission anomaly concentration index for each level is obtained by applying the calculation rules described above.

[0040] Preferably, a summary of the current carbon emission status is generated based on the real-time collected dynamic emission data stream, and mapped to the corresponding level within the multi-level carbon emission quantification framework, specifically including:

[0041] The current dynamic emission data stream is subjected to the same instantaneous feature enhancement processing as historical data to obtain the current multidimensional carbon emission intensity sequence.

[0042] Extract the current carbon emission operational status vector from the current multi-dimensional carbon emission intensity sequence;

[0043] Identify the hierarchical affiliation of the current carbon emission operational status vector within a multi-level carbon emission quantification framework;

[0044] The current carbon emission operating status vector is matched with the set of historical carbon emission operating status vectors of the level to which it belongs.

[0045] Preferably, the distribution similarity between the current carbon emission status summary and the historical emission anomaly concentration index of that level is calculated to generate a real-time online carbon emission deviation index, specifically including:

[0046] Obtain the historical abnormal emission concentration index and the corresponding historical carbon emission operational status vector distribution for the current tier.

[0047] Calculate the core density estimate of the current carbon emission operating state vector and the distribution of the historical carbon emission operating state vector;

[0048] Based on the core density estimate, the similarity of the current carbon emission status with the distribution of historical anomalies is derived;

[0049] Distribution similarity is converted into a real-time online carbon emission deviation index, and the value of the real-time online carbon emission deviation index is negatively correlated with distribution similarity.

[0050] Preferably, an adaptive quantitative mapping model is constructed by integrating real-time online carbon emission deviation indicators, multi-dimensional carbon emission intensity sequences, and environmental status time-series data, specifically including:

[0051] Real-time carbon emission deviation indicators are used as dynamic weighting factors;

[0052] Weighted fusion of recent data segments from multi-dimensional carbon emission intensity sequences and corresponding data segments from environmental status time series data;

[0053] A temporal neural network architecture is used to process the weighted and fused data segments, and a mapping relationship from the current state to the future state is established.

[0054] The parameters of the time-series neural network are adjusted by training with historical data to form an adaptive quantization mapping model;

[0055] By using an adaptive quantization mapping model to extrapolate the short-term dynamic trajectory of carbon emissions in the park, the evolution path of carbon emission intensity within future time windows is output, specifically including:

[0056] Input the current multidimensional carbon emission intensity sequence and real-time environmental status data into the adaptive quantization mapping model;

[0057] The model recursively extrapolates the carbon emission intensity values ​​at multiple future time points, forming a carbon emission intensity evolution sequence;

[0058] Based on the park's production plan data, adjust the carbon emission intensity values ​​corresponding to planned activities in the evolution sequence;

[0059] Output the carbon emission intensity evolution path within the adjusted future time window;

[0060] Based on the matching results between the carbon emission intensity evolution path and the preset carbon emission control boundary conditions, a tiered early warning signal and control strategy sequence are generated, specifically including:

[0061] Load the multidimensional boundary condition set from the park's carbon emission control rule base;

[0062] The evolution path of carbon emission intensity is compared point by point with the set of multidimensional boundary conditions;

[0063] Identify the time periods and magnitudes of exceeding boundary conditions in the evolutionary path;

[0064] Different levels of graded warning signals are generated based on the magnitude of the exceedance.

[0065] For each warning signal, the corresponding control measures are retrieved from the control strategy library and sorted by priority to form a control strategy sequence.

[0066] Preferably, the present invention also includes a real-time intelligent monitoring system for carbon emissions in industrial parks based on big data collection. The system includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of the real-time intelligent monitoring method for carbon emissions in industrial parks based on big data collection as described above.

[0067] Compared with the prior art, the beneficial effects of the present invention are:

[0068] By continuously capturing dynamic emission data streams from various carbon emission sources within the park through a distributed sensor node cluster, and simultaneously collecting corresponding production activity time-series data and environmental status time-series data, it is possible to comprehensively and accurately obtain various data information related to carbon emissions in the park. This breaks through the limitations of traditional single sensor node data collection, providing a more comprehensive and reliable data foundation for subsequent carbon emission analysis and assessment, and helping to more accurately grasp the overall situation of carbon emissions in the park.

[0069] When processing dynamic emission data streams, this method uses instantaneous feature enhancement to extract multi-dimensional carbon emission intensity sequences with temporal correlations and generate dynamic carbon emission profiles for the park. This fully explores the potential information contained in the carbon emission data, enabling park managers to clearly understand the changing characteristics and patterns of carbon emissions in different time periods and dimensions, intuitively grasp the dynamic trends of carbon emissions in the park, and provide more valuable reference for subsequent carbon emission management decisions.

[0070] A multi-level carbon emission quantification framework, constructed based on multi-dimensional carbon emission intensity sequences and production activity time-series data, provides a framework for carbon emission quantification. Each level corresponds to a different granularity of carbon emission aggregation, which can meet the needs of industrial parks for carbon emission quantification analysis at different management levels and in different application scenarios. Whether it is to control the total amount of carbon emissions at the overall park level or to conduct micro-level carbon emission detail analysis at the level of specific production links or specific carbon emission sources, this framework can provide accurate quantification results, thereby improving the practicality and pertinence of carbon emission quantification work.

[0071] Within the multi-level carbon emission quantification framework, the abnormal emission concentration index at each level is calculated based on the data distribution patterns at the time of historical carbon emission events. Combined with the current carbon emission status summary generated from real-time data collection, the real-time online carbon emission deviation index is generated by calculating the distribution similarity. This enables real-time identification and monitoring of abnormal carbon emissions in the park, changing the passive situation of traditional post-event analysis and enabling timely detection of problems when abnormal signs of carbon emissions appear.

[0072] Based on real-time online carbon emission deviation indicators, this method generates tiered early warning signals and control strategy sequences. Different levels of warning signals are issued according to the severity of carbon emission anomalies, allowing park managers to quickly understand the urgency of the situation and take timely and effective control measures based on the corresponding control strategy sequences. This helps control carbon emission anomalies in the shortest possible time, preventing significant exceedances of total carbon emissions, promoting low-carbon operation in the park, and helping the park better address the challenges of climate change while ensuring normal production and operation, thus promoting sustainable development. Furthermore, this method enhances the intelligence level of carbon emission management in the park, reduces manual intervention, lowers management costs, and improves management efficiency, providing a more scientific and efficient solution for carbon emission management in the park. Attached Figure Description

[0073] Figure 1 This is a schematic diagram illustrating the working principle of the real-time intelligent monitoring method for carbon emissions in industrial parks based on big data collection, as described in this invention.

[0074] Figure 2 This is a flowchart of the data acquisition and synchronization process;

[0075] Figure 3 A flowchart for constructing a multi-level carbon emission quantification framework. Detailed Implementation

[0076] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0077] Please see Figure 1This invention provides a real-time intelligent monitoring method for carbon emissions in industrial parks based on big data collection. The method includes: integrating distributed sensing technology, time-series data analysis, and machine learning models to achieve dynamic monitoring, anomaly detection, and intelligent early warning of carbon emissions in industrial parks. The specific embodiments of this invention are described in detail below with reference to the accompanying drawings. The overall implementation scheme is based on the core steps of the method, including data acquisition, feature processing, framework construction, anomaly calculation, state mapping, deviation indicator generation, and early warning strategy output. A distributed cluster of sensor nodes is deployed at key locations within the park to continuously capture dynamic emission data streams from carbon emission sources, while simultaneously collecting time-series data on production activities and environmental conditions. The dynamic emission data streams undergo instantaneous feature enhancement processing to extract multi-dimensional carbon emission intensity sequences, thereby generating a dynamic profile reflecting the park's real-time emission status. Based on these multi-dimensional sequences and production data, a multi-level carbon emission quantification framework is constructed, with each level corresponding to a different aggregation granularity, such as equipment level, workshop level, or park level. Within this multi-level framework, based on the data distribution patterns of historical carbon emission events, anomaly concentration indices for each level are calculated. Real-time data streams are used to generate a summary of the current carbon emission status and map it to the corresponding level. By calculating the distribution similarity between the current status and historical anomaly indices, real-time online carbon emission deviation indicators are generated. Finally, based on these deviation indicators, tiered early warning signals and control strategy sequences are generated to achieve proactive carbon emission management.

[0078] Example 1: See Figure 2 The deployment of distributed sensor node clusters is the physical foundation for carbon emission monitoring in industrial parks. Its configuration strategy directly determines the comprehensiveness and accuracy of data collection. When deploying multiple types of sensor groups within the park, the distribution density of emission sources, process characteristics, and spatial geometry must be comprehensively considered. Sensor group types include carbon concentration sensors for monitoring flues and exhaust outlets, smart meters for measuring electricity, gas, and steam consumption, and RFID or flow metering devices for tracking the flow of raw materials and semi-finished products. These sensors operate continuously at a pre-set fixed sampling frequency, collecting raw carbon emission concentration pulses, cumulative energy consumption values, and instantaneous material flow rates. IoT nodes, acting as regional data aggregation centers, are strategically positioned within the communication range of the sensor clusters. Their primary function is to receive and aggregate raw data streams from sensors of different physical locations and protocol types. During aggregation, the gateway adds a high-precision timestamp to each data packet and binds it to the source device identifier. This process eliminates data timing errors caused by network transmission delays or device clock drift, ensuring that subsequent analysis is based on a strictly consistent time reference.

[0079] Synchronous integration of data from the production management system is a crucial step in understanding the background of carbon emissions. The formation of production activity time-series data relies on interfaces with the Manufacturing Execution System (MES) and Enterprise Resource Planning (ERP) systems. Real-time extraction of equipment operating status sequences, such as machine tool speed, reactor temperature, and production line cycle time, is used to obtain detailed work plan logs including product batches, formula numbers, and planned output. Actual capacity output indicators, such as the number of qualified products and energy consumption per unit of product, are also recorded. This production data is correlated with sensor data streams through a unified transaction clock, thus precisely linking intangible carbon emissions with tangible production activities. Simultaneously, environmental status time-series data is also collected concurrently. Weather stations distributed across open areas of the park and building rooftops monitor parameters such as temperature, humidity, wind speed, and air pressure. These environmental factors not only affect the dispersion rate of emissions but may also have potential correlations with the emission intensity of certain processes.

[0080] After entering the processing stage, dynamic emission data streams undergo transient feature enhancement to extract valuable information. A sliding time window algorithm is used to process the continuous dynamic emission data stream. This algorithm slides sequentially across the data stream in fixed time window units, resampling and filtering the data within each window to highlight transient emission peaks that rise sharply in a short period and relatively stable continuous emission baselines. The feature-enhanced data stream is then separated according to source identifiers and merged into their respective corresponding carbon emission sources, forming independent intensity sequences. For example, carbon concentration data attributed to boiler combustion is combined with gas consumption data to calculate a direct emission intensity sequence, and data attributed to purchased electricity consumption is converted into an indirect emission intensity sequence based on emission factors. These time-aligned independent intensity sequences are further integrated, with the spatial coordinates or regional codes of their collection points added, to construct a multi-dimensional carbon emission intensity sequence with spatiotemporal identification. This sequence not only records "when and how much was emitted" but also clearly reflects "where the emission occurred."

[0081] Generating a dynamic carbon emission profile for a park based on a multi-dimensional carbon emission intensity sequence is a data-to-information transformation process. The generation of this dynamic profile relies on in-depth analysis of the spatial distribution and temporal fluctuation characteristics implicit in the sequence. Spatial distribution characteristics are visualized by using heatmaps or gradient maps to show emission hotspots at different locations within the same time slice. Temporal fluctuation characteristics are characterized by analyzing the periodicity, trend, and suddenness of the sequence. For example, identifying typical emission intensity changes during shift changes or detecting deviations in emission distribution under abnormal weather conditions. The resulting dynamic profile can be a set of feature vectors integrating spatiotemporal information, or a graphical dashboard rendered based on these vectors, providing operators with a real-time and intuitive overview of the overall carbon emission situation in the park. The entire implementation process emphasizes the continuity of data flow and the progressiveness of processing. From physical signal acquisition to multi-source data fusion, and then to feature enhancement and information extraction, each step provides carefully processed input for subsequent quantitative framework construction and anomaly detection. This end-to-end processing chain ensures that the monitoring results have both microscopic precision and reflect macroscopic trends.

[0082] Example 2: See Figure 3 The construction of a multi-level carbon emission quantification framework begins with an in-depth analysis of the physical distribution and logical relationships of carbon emission sources within the industrial park. Physical distribution refers to the actual geographical location of emission sources on the park map, such as the specific coordinates of facilities like boiler rooms, painting workshops, and final assembly lines. Logical relationships reflect the energy and material transfer relationships in the production process, such as a continuous heat treatment furnace forming an energy utilization unit with its upstream waste heat boiler and downstream drying system. Based on this analysis and classification, carbon emission aggregation levels with practical management significance are defined, typically using a division from micro to macro. The equipment level focuses on a single energy-consuming or polluting piece of equipment, such as an air compressor or smelting furnace; the workshop level covers the collection of all emission sources within an independent plant or a complete production line; and the park level encompasses the total emissions of the entire industrial park. Each level corresponds to a different data aggregation granularity: the equipment level retains the most detailed instantaneous fluctuations, the workshop level performs a certain degree of smoothing and summing, and the park level presents macro trends.

[0083] For each clearly defined aggregation level, it is necessary to extract the corresponding carbon emission intensity subsequence from the generated multi-dimensional carbon emission intensity sequence. This work relies on the source identifier and spatial identifier attached to the data points. For the equipment level, the system scans the multi-dimensional sequence, filters out all data points whose identifiers completely match the unique code of the equipment, and arranges them in chronological order to form the intensity subsequence of that equipment. The extraction of subsequences at the workshop level is more complex. It is necessary to first group all equipment identifiers within the workshop's jurisdiction into one group, and then extract the data points of all equipment under that group from the total sequence, perform time alignment, and then accumulate or weighted average them to form a subsequence representing the overall emission level of the workshop. The subsequence at the park level is the sum of all equipment-level subsequences, reflecting the overall scale of carbon emissions in the park. In order for the quantitative framework to not only reflect emission levels but also reveal the intrinsic relationship between emissions and production activities, it is necessary to extract a series of production efficiency indicators that match each aggregation level from the time-series data of production activities. The selection of production efficiency indicators is closely related to the characteristics of the level. Equipment-level metrics may correspond to indicators such as unit time output, operating load rate, and OEE (Overall Equipment Efficiency) of a single device. Workshop-level metrics focus on total output, overall energy intensity, and working hour utilization. Park-level metrics use macro-level indicators such as total output value and total energy consumption. The extraction process also relies on timestamps and device or department identifiers to align and extract the corresponding production efficiency indicators from the massive production activity time series data, forming a series of efficiency indicators that correspond one-to-one with the carbon emission intensity subsequence on the time axis.

[0084] Fusion of the carbon emission intensity subsequence and the production efficiency index sequence is a key step in forming the carbon emission operational state vector. Fusion is not a simple data splicing, but aims to create a multi-dimensional data point that can comprehensively reflect the relationship between "emission intensity" and "production intensity" at a specific level at a specific time. For a certain level at a certain sampling time i, its carbon emission operational state vector Vi can be represented as (Ei, Pi, Ti), where Ei represents the value of the carbon emission intensity subsequence at that time, Pi represents the efficiency value extracted from the production efficiency index sequence at the corresponding time, and Ti is the timestamp. The vector may contain more dimensions; for example, when considering environmental factors, temperature or humidity data can be incorporated. Each dimension in the vector has been standardized to eliminate dimensional differences, allowing data of different scales to be compared and calculated in the same vector space.

[0085] Based on carbon emission operational status vectors at all levels, a multi-level carbon emission quantification framework is constructed. Internally, this framework is represented as a structured database or a set of related data tables. The top level is the park-level vector set, the middle level is the vector sets for each workshop level, and the bottom level is the vector set for all equipment levels. Each level's vector set is indexed chronologically, allowing for rapid retrieval by time range. The framework also maintains the relationships between levels; for example, a workshop-level vector can be used to quickly locate all subordinate equipment-level vectors. This tree-like or graph-like structure allows analysts to drill down from top to bottom, tracing anomalies from the overall park to specific workshops or equipment, or to aggregate from bottom to top, assessing the impact of emissions changes from individual equipment on overall workshop and even park indicators. After constructing the multi-level carbon emission quantification framework, the next step is to calculate the emission anomaly concentration index for each level. This index aims to quantify the density of the corresponding level's operational status vector clustering in the vector space when a known carbon emission anomaly event occurs within a historical period. The calculation process begins with a search of the historical database, which contains logs of abnormal carbon emission events recorded over a period of time, such as "on a certain day, a workshop experienced a momentary exceedance of carbon emission standards due to equipment failure." The database also archives corresponding multi-dimensional carbon emission intensity sequences and production activity time-series data from the past.

[0086] Extract the set of carbon emission operation status vectors corresponding to each aggregation level when an abnormal carbon emission event occurs. For an abnormal event, first determine the scope of its directly affected levels. Assuming the event record is "Equipment B in Workshop A experienced an anomaly near time T", then it is necessary to extract the carbon emission operation status vectors of Equipment B (equipment level), Workshop A (workshop level), and the entire park (park level) within a time window before and after time T (e.g., 1 hour before the event to 1 hour after the event). Vectors extracted from all historical abnormal events belonging to the same level constitute the set of historical abnormal vectors for that level. Calculate the distribution density and distribution offset of the carbon emission operation status vectors in this abnormal vector set. Distribution density describes the tightness of the clustering of vector points representing abnormal states in the vector space. If, during multiple historical anomalies, the operation status vectors of this level all appear in a relatively small area of ​​the vector space, then the distribution density of that area is considered high. Distribution offset measures the degree of deviation of this abnormal vector set from the overall distribution (including normal and abnormal) of all historical vectors at that level. A large offset indicates a more obvious distinction between abnormal and normal states. The specific calculation rules for the abnormal emission concentration index at each level are determined based on the calculated distribution density and distribution offset. One feasible rule is to perform a combination operation on the distribution density and distribution offset, such as a weighted product. Regions with high density and large offset also have high abnormal concentration indices, indicating that these regions are typical "patterns" of historical abnormal events and require high vigilance. After the rule is determined, it is applied to traverse possible positions in the vector space, or to perform a global calculation on the set of historical abnormal vectors, thereby obtaining the historical abnormal emission concentration index for each level. This index can be a numerical value or a function (such as a probability density function), which characterizes which state regions in the vector space of the corresponding level are historically highly correlated with abnormal events.

[0087] Example 3: Generating a summary of the current carbon emission status is a core step in the real-time monitoring process. This process begins with the immediate processing of the continuously flowing dynamic emission data stream. The processing method is strictly consistent with the instantaneous feature enhancement method used when building the historical database to maintain the consistency of the analysis benchmark. The newly acquired raw data stream first undergoes local feature enhancement using the same sliding time window algorithm. This algorithm slides across the real-time data stream with a window width identical to the historical processing parameters, filtering and smoothing the data points within the window. The aim is to suppress measurement noise while highlighting meaningful transient emission peaks and continuous emission baselines that characterize stable operating conditions. The enhanced data stream then enters the sequence separation stage. Based on the source identifier embedded in each data packet, the system decomposes the mixed data stream and assigns it to its specific carbon emission source. For example, the data is routed to independent logical channels such as "Boiler No. 1" and "Painting Production Line," and an independent direct or indirect carbon emission intensity sequence is calculated for each source. These independent intensity sequences, which are strictly synchronized in time, are further assigned spatial coordinate information of their collection points, and are finally integrated to generate a multi-dimensional carbon emission intensity sequence at the current time point. This sequence constitutes the underlying data foundation for the current carbon emission status summary.

[0088] Extracting the current carbon emission operational status vector from the current multi-dimensional carbon emission intensity sequence is an information condensation step that fully follows the vector definition rules preset in the multi-level carbon emission quantification framework. The extraction of the operational status vector is performed separately for each aggregation level defined in the framework. For the equipment level, the system retrieves the latest carbon emission intensity data point corresponding to that equipment from the current multi-dimensional sequence, and simultaneously obtains the equipment's production efficiency indicators (such as instantaneous power and output rate) from the real-time production data stream. These two values ​​are combined with a timestamp to form the equipment's operational status vector at the current moment. For the workshop level, it is necessary to aggregate the carbon emission intensity of all equipment under that workshop at the current moment, calculate their sum or average as the workshop's total emission intensity, and simultaneously obtain indicators representing the overall efficiency of the workshop, thus forming the workshop-level operational status vector. The generation of the park-level vector adopts a similar aggregation logic, but the data scope is extended to the entire park.

[0089] Identifying the level to which the current carbon emission operating state vector belongs within the multi-level carbon emission quantification framework is a logical judgment process. This process is not simply assigning the vector to its source level, but rather determining which level's "typical state" in the framework best matches the state represented by the current vector. The system performs a preliminary similarity comparison between the newly generated current operating state vector and the historical operating state vector set across all levels in the quantification framework, with the comparison conducted separately within the vector space of each level. For example, a device-level vector extracted from "No. 2 smelting furnace" not only necessarily belongs to the device level, but the system also calculates its "distance" from the center of the workshop-level historical vectors to determine whether its abnormal pattern is more consistent with a certain overall fluctuation characteristic of the workshop level. This cross-level correlation analysis helps identify events originating from a single device but representing broader systemic problems, thus achieving a more accurate attribution judgment. The current vector is marked as belonging to the level of the historical state vector most similar to it, serving as the primary target for subsequent in-depth comparisons. Matching the current carbon emission operating state vector with the historical carbon emission operating state vector set of its assigned level is a preparatory step for anomaly detection; the matching operation aims to examine real-time data within a historical context. The system retrieves all operational state vectors recorded within a historical period (e.g., the past three months) from the database of the quantization framework, forming a historical vector set. This set contains various state samples of that level under normal and known abnormal operating conditions. The matching process does not search for a completely identical vector, but rather evaluates the position of the current vector within the vector space spanned by the historical vector set. It observes whether the vector falls into a dense region where historical normal state vectors are clustered, a sparse region where historical abnormal state vectors are located, or a new region not covered by historical data. This positional relationship provides the initial context for subsequent distributional similarity calculations.

[0090] Calculating the distribution similarity between the current carbon emission status summary and the historical emission anomaly concentration index at that level is a core algorithmic step in quantifying real-time anomaly risk. This calculation relies on kernel density estimation methods in nonparametric statistics. Within this framework, the historical emission anomaly concentration index is essentially represented as a probability density function estimated based on a set of historical anomaly event vectors. Let this density function represent the density of a point in the state vector space. The probability density of historical anomalous states occurring nearby. Currently, the evaluation needs to be based on the real-time acquired runtime state vector. With this historical anomalous distribution The degree of similarity.

[0091] Distribution similarity The similarity can be quantified by calculating the kernel density estimate of the current vector under historical anomalous distributions; specifically, it is the similarity. Proportional to Its mathematical expression is as follows:

[0092]

[0093] in: This represents the calculated distribution similarity; it is a scalar value. This represents the selected kernel function, such as the Gaussian kernel, which is used to smoothly measure how close the current vector is to historical outlier vectors. This represents a vector representing the current carbon emission operating status extracted from real-time data. This represents the set of all historical abnormal carbon emission operation status vectors retrieved from the historical database that correspond to the hierarchical level to which this real-time vector belongs. This represents a bandwidth parameter matrix that controls the smoothness of the kernel function, affecting the accuracy and bias of the density estimation. It is calculated as follows: After setting the value, the quantification of distribution similarity is complete. The larger the value, the stronger the current state vector. In the set of historical abnormal state vectors The higher the probability of occurrence in the resulting distribution, that is, the more similar the current state is to historical anomaly patterns.

[0094] Converting distribution similarity into a real-time online carbon emission deviation index is the final step in generating the ultimate actionable signal. The deviation index is designed to be similar to the distribution similarity. A negative correlation, meaning a higher similarity and a lower deviation from the indicator value, indicates that the current state is closer to historical anomaly patterns, and the risk level is higher. This negative correlation can be achieved using a monotonically decreasing function, such as a linear transformation. (Where A and B are constants used to calibrate the index range), or an exponential decay function can be used. (C and D are parameters) to amplify the sensitivity of high-similarity regions. The converted real-time carbon emissions deviate from the online index. It is a standardized value, usually normalized to the range of [0,1] or [0,100], for easy understanding by monitoring personnel. The lower the value, the greater the deviation of current carbon emission behavior from the expected trajectory, and the higher the probability of triggering an early warning. This indicator provides a quantitative basis for subsequent tiered early warning and control strategies.

[0095] Example 4: The integration of real-time online carbon emission deviation indicators, multi-dimensional carbon emission intensity sequences, and environmental state time-series data to construct an adaptive quantitative mapping model is a dynamic and continuously self-adjusting data-driven process. The model's construction begins with the preprocessing and weighted integration of the input data. The real-time online carbon emission deviation indicator plays the role of a dynamic weighting factor, reflecting in real-time the similarity between the current system state and historical anomaly patterns; a lower value indicates a higher anomaly risk. During the weighted fusion process, the system selects data segments within the most recent time window from the multi-dimensional carbon emission intensity sequence, such as park-level carbon emission intensity data per minute over the past two hours, while simultaneously extracting environmental state time-series data from the same period, such as temperature and wind speed readings. Weighted fusion is not a simple data splicing but rather a dynamic adjustment of the influence of the two types of data in the model based on the magnitude of the deviation indicator. When a lower deviation indicator suggests an increased anomaly risk, the model assigns higher weights to recent carbon emission intensity data while appropriately reducing the weights of environmental data, assuming that current emission dynamics are primarily driven by internal factors. Conversely, when the deviation indicator is within the normal range, the weights of environmental data such as wind speed and humidity on emission diffusion are correspondingly increased.

[0096] The core of the model lies in employing a temporal neural network architecture to process the weighted and fused data segments. This network structure typically includes long short-term memory units or gated recurrent units to capture the temporal dependencies in the data. The network input is a weighted, chronologically ordered sequence of fused data points, each containing a weighted carbon emission intensity and weighted environmental parameters. The output is a predicted carbon emission intensity for a single future time point. Training and adjusting the temporal neural network parameters using historical data is crucial for the model's predictive ability. The training process utilizes a large amount of historical data; the input is weighted and fused data from a window prior to a specific historical time point, and the expected output is the actual observed carbon emission intensity at a specific time interval after that point. Through algorithms such as backpropagation, the weights and biases of the network's internal connections are continuously adjusted, making the network's predicted output increasingly closer to the true value, ultimately forming an adaptive quantization mapping model capable of extrapolating the near future based on recent data and the current state.

[0097] Extrapolating the short-term dynamic trajectory of carbon emissions in a park using an adaptive quantization mapping model is a recursive, multi-step prediction process. The starting point is to input the latest multi-dimensional carbon emission intensity sequence and real-time environmental status data, after undergoing the same weighted fusion process, into a pre-trained adaptive quantization mapping model. The model first outputs the predicted carbon emission intensity for the next time point. This predicted value is then used as part of a new input sequence, combined with possible environmental data for future times (such as weather forecasts or persistent predictions), and input again to obtain the predicted value for the next time point. This process is repeated recursively to generate a carbon emission intensity evolution sequence consisting of multiple future time points, for example, extrapolating the intensity trajectory for the next 6 hours, with each point occurring every 15 minutes.

[0098] Adjusting the evolution sequence by incorporating industrial park production plan data is a crucial step in improving forecast accuracy. This is because the carbon emission intensity evolution path is not only influenced by inertia but also closely related to planned production activities. The system incorporates future production plans, such as "Workshop 1 will start a large heat treatment furnace in three hours, expected to last four hours." For such known major energy consumption events, the model will, within the corresponding time period, make local upward adjustments to the carbon emission intensity evolution sequence based purely on data extrapolation, according to the historical emission efficiency data of the equipment. This makes the predicted path more accurately reflect anticipated operational changes. The adjusted carbon emission intensity evolution path within the future time window is output. This path is a time-emission intensity curve, depicting the most likely trend of carbon emission intensity change over a future period, serving as a direct basis for early warning decisions.

[0099] Generating tiered early warning signals and control strategy sequences based on the matching results between the carbon emission intensity evolution path and preset carbon emission control boundary conditions is a rule-based decision-making process. The decision-making process begins by loading a multi-dimensional set of boundary conditions from the park's carbon emission control rule base. These boundary conditions are typically not single values ​​but a complex set of rules, which may include: the instantaneous upper limit of carbon emission intensity at the park level, the hourly cumulative allowable emissions of carbon emission intensity at the workshop level, and the carbon emission intensity limits for specific processes under specific products, forming a multi-dimensional, multi-threshold boundary condition system. Comparing the carbon emission intensity evolution path with the multi-dimensional boundary condition set point by point requires discretizing the continuous prediction path to specific time points. The system scans each prediction time point on the evolution path, checking whether the predicted carbon emission intensity value at that point exceeds any boundary condition in the rule base applicable to that time point and level. Identifying the time periods and magnitudes of exceeding boundary conditions in the evolutionary path is key to quantifying risk. The system not only records exceedances at individual time points but also focuses on continuous exceedance events. For example, it was found that the predicted path showed that between the 3rd and 3.5th hour in the future, the predicted value of carbon emission intensity at the park level would be continuously higher than the set emergency threshold, and the maximum exceedance would reach 15% of the threshold level.

[0100] Different levels of tiered warning signals are generated based on the magnitude of the exceedance. The warning level is usually linked to the severity and duration of the exceedance. For example, a momentary slight exceedance may trigger a "blue" level warning, a sustained exceedance between 5% and 10% may trigger a "yellow" warning, and a sustained exceedance exceeding 10% may trigger an "orange" or even a "red" level warning. For each generated warning signal, the system retrieves corresponding control measures from a pre-defined control strategy library. The measures in the strategy library are associated with the warning level and the type of the exceedance source. For example, for an orange warning of total exceedance at the park level, the retrieved measures may include "suggesting adjusting the power consumption sequence of non-critical production lines" and "activating backup clean energy." These measures are sorted according to predefined priorities, implementation costs, and scope of impact, ultimately forming an ordered and executable sequence of control strategies for operators to use as a reference for decision-making.

[0101] Table 1: Results of matching carbon emission intensity evolution path with boundary conditions

[0102]

[0103] Example 5: The generation of tiered early warning signals and control strategy sequences based on real-time online carbon emission deviation indicators constitutes the decision output terminal of the entire monitoring method. This process begins with a comprehensive assessment of the current system state. The real-time online carbon emission deviation indicator serves as a quantitative risk benchmark, dynamically reflecting the similarity between the current carbon emission pattern and historical anomaly patterns. When the indicator value falls below a preset threshold, the early warning generation process is triggered. The system first reads the current value of the indicator and its recent trend. For example, if the indicator rapidly drops from the normal range of 75 units to 30 units within ten minutes, this rapid decline itself is a risk signal that requires attention. Integrating real-time online carbon emission deviation indicators, multi-dimensional carbon emission intensity sequences, and environmental state time-series data to construct an adaptive quantitative mapping model is a key step in forward-looking prediction. In this stage, the real-time deviation indicator not only serves as a trigger condition but also as a dynamic weighting factor integrated into the model input. The system assigns higher weight to recent carbon emission intensity data, especially emission source data from identified potential anomaly areas. Simultaneously, current environmental parameters such as wind speed and inversion layer information are also considered to assess the impact of atmospheric diffusion conditions on future emission concentrations. The constructed adaptive quantization mapping model is usually a time-series prediction model trained with a large amount of historical data. It can understand the typical evolution of the park's carbon emission system under different initial states and external conditions.

[0104] An adaptive quantization mapping model is used to extrapolate the short-term dynamic trajectory of carbon emissions in the industrial park, providing a lead time for early warning. The model receives the current fused data state and simulates the system's behavior over a future period. The extrapolation process is iterative. The model outputs a predicted carbon emission intensity value for the first future time point. This predicted value is then used as part of a new initial state, combined with predicted future environmental conditions (such as short-term wind speed changes from weather forecasts), and the model is run again to obtain the prediction for the next time point. This cycle is repeated to generate a carbon emission intensity evolution path within a future time window. This path depicts the most likely trajectory of carbon emission intensity change. For example, the model might predict that due to the continuation of the current trend, the total emission intensity of the industrial park may reach the warning line in 40 minutes. Based on the matching results of the carbon emission intensity evolution path and the preset carbon emission control boundary conditions, a tiered early warning signal and control strategy sequence are generated, which is the core of translating prediction into action. The system compares the extrapolated future emission path with pre-set multi-level control boundaries, which typically include attention thresholds, warning thresholds, and severity thresholds. The matching process not only checks whether the path crosses the boundary, but also analyzes the timing of the crossing, the duration, and the extent of the exceedance. For example, the evolution path shows that the emission intensity will exceed the warning threshold after 25 minutes, and will continue to rise after exceeding the threshold, and may reach the severe threshold after 50 minutes.

[0105] Based on the matching results, a tiered early warning signal is generated. The warning level is directly related to the severity of the boundary crossing. A slight, momentary exceedance of the attention threshold may only trigger a low-level "attention" signal, while a predicted sustained exceedance of the severity threshold will trigger the highest-level "emergency" warning signal. Each warning signal is associated with a series of control measures in the contingency plan library. The system retrieves the corresponding control actions from the strategy library based on the warning level, the type of emission source involved, and the predicted duration, and sorts them according to preset priority, response speed, and implementation cost to form an ordered sequence of control strategies. For example, for a predicted total emission exceedance at the park level, the generated strategy sequence might be: "send energy-saving initiatives to each workshop," "automatically adjust central air conditioning settings," and "negotiate with the power grid to adjust non-productive electricity consumption."

[0106] The memory stores massive amounts of historical sensor data, model parameters, early warning rule bases, and control strategy bases. The processor executes instructions from the computer program, coordinating the data flow and computational tasks of the entire system. The system continuously collects dynamic emission data streams, production activity time-series data, and environmental status time-series data through a distributed cluster of sensor nodes deployed throughout the park. This data is aggregated by an IoT gateway and transmitted to the system's data processing core. When executing the computer program, the processor progressively implements each step of the aforementioned monitoring method. It performs instantaneous feature enhancement processing on the received raw data stream, extracts multi-dimensional carbon emission intensity sequences, and generates dynamic profiles. A multi-level carbon emission quantification framework reflecting the park's operational status is built and maintained in the system memory. The processor uses historical data to calculate the emission anomaly concentration index for each level and maps the real-time generated current carbon emission status summary to the corresponding level within the framework. By calculating distribution similarity, the processor updates the online carbon emission deviation index in real time and drives the adaptive quantification mapping model to perform short-term trajectory extrapolation. Finally, the system automatically generates tiered early warning signals and control strategy sequences based on the prediction results and provides them to park operation and management personnel through a human-machine interface or API interface, forming a closed loop from data perception to decision support. The system's operation can be illustrated through a specific scenario. Suppose that on a weekday afternoon, the system, through real-time calculations, detects a sharp increase in the similarity between the carbon emission operating status vector of the foundry workshop and the historical abnormal concentration index distribution, causing the real-time online carbon emission deviation index to rapidly decrease to a low level. This change triggers an early warning process. Based on the current high emission intensity of the foundry workshop, the low wind speed conditions in the surrounding area, and this index, the adaptive quantification mapping model deduces that the workshop's emission intensity is likely to exceed the warning threshold within the next hour. Based on this, the system generates an "Orange Warning for the Foundry Workshop" signal and automatically generates a sequence of control strategies. The suggested actions in the sequence might include "checking the cupola furnace combustion efficiency" and "recommending that some non-emergency pouring operations be postponed until the evening when wind speeds are higher."

[0107] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for real-time intelligent monitoring of carbon emissions in industrial parks based on big data collection, characterized in that, include: The system continuously captures dynamic emission data streams from various carbon emission sources within the park through a distributed sensor node cluster, and simultaneously collects corresponding production activity time-series data and environmental status time-series data. Instantaneous feature enhancement processing is performed on the dynamic emission data stream to extract multi-dimensional carbon emission intensity sequences with temporal correlation, and a dynamic carbon emission profile of the park is generated, specifically including: A sliding time window algorithm is used to enhance the local features of dynamic emission data streams, highlighting transient emission peaks and continuous emission baselines in the data streams. Independent intensity sequences corresponding to different carbon emission sources were separated from the enhanced data stream, including direct emission intensity sequences and indirect emission intensity sequences; The independent intensity sequences were integrated along the time axis into a multi-dimensional carbon emission intensity sequence with spatial identifiers. Based on the spatial distribution and temporal fluctuation characteristics of multi-dimensional carbon emission intensity sequences, a dynamic profile of carbon emissions in the park is generated. Based on multi-dimensional carbon emission intensity sequences and production activity time series data, a multi-level carbon emission quantification framework reflecting the park's operational status is constructed, with each level corresponding to a different carbon emission aggregation granularity. Within the multi-level carbon emission quantification framework, based on the data distribution patterns at the time of historical carbon emission events, the emission anomaly concentration index at each level is calculated, specifically including: Retrieve historical records of abnormal carbon emission events and multi-dimensional carbon emission intensity sequences at the time of the events; Extract the set of carbon emission operation status vectors corresponding to each aggregation level when an abnormal carbon emission event occurs; Calculate the distribution density and distribution offset of the carbon emission operating state vector in this set; The calculation rules for the emission anomaly concentration index at each level are determined based on the distribution density and distribution offset. The historical emission anomaly concentration index for each level is obtained by applying the aforementioned calculation rules; A summary of the current carbon emission status is generated based on the real-time collected dynamic emission data stream, and then mapped to the corresponding level within the multi-level carbon emission quantification framework. Calculate the distribution similarity between the current carbon emission status summary and the historical emission anomaly concentration index at this level to generate a real-time online carbon emission deviation index, specifically including: Obtain the historical abnormal emission concentration index and the corresponding historical carbon emission operational status vector distribution for the current tier. Calculate the core density estimate of the current carbon emission operating state vector and the distribution of the historical carbon emission operating state vector; Based on the core density estimate, the similarity of the current carbon emission status with the distribution of historical anomalies is derived; The distribution similarity is converted into a real-time online carbon emission deviation index, and the value of the real-time online carbon emission deviation index is negatively correlated with the distribution similarity. Based on real-time online deviation indicators of carbon emissions, a tiered early warning signal and a sequence of control strategies are generated.

2. The method for real-time intelligent monitoring of carbon emissions in industrial parks based on big data collection as described in claim 1, characterized in that, The process of generating tiered early warning signals and control strategy sequences based on real-time online carbon emission deviation indicators includes: An adaptive quantitative mapping model is constructed by integrating real-time online carbon emission deviation indicators, multi-dimensional carbon emission intensity sequences, and environmental status time series data. The short-term dynamic trajectory of carbon emissions in the park is extrapolated by an adaptive quantization mapping model, and the evolution path of carbon emission intensity within the future time window is output. Based on the matching results between the carbon emission intensity evolution path and the preset carbon emission control boundary conditions, a tiered early warning signal and control strategy sequence are generated.

3. The method for real-time intelligent monitoring of carbon emissions in industrial parks based on big data collection as described in claim 2, characterized in that, The system continuously captures dynamic emission data streams from various carbon emission sources within the park through a distributed sensor node cluster, and simultaneously collects corresponding time-series data on production activities and environmental status, specifically including: Multiple types of sensor groups are configured and distributed in different areas of the park to collect carbon emission concentration data, energy consumption data and material flow data at a fixed sampling frequency; The raw data streams collected by multiple types of sensor groups are aggregated through IoT key points, and the data streams are timestamped and bound to source identifiers. Synchronously integrate equipment operation status sequences, work plan logs, and capacity output indicators from the production management system to form time-series data of production activities; Simultaneously, environmental sensor data on temperature, humidity, wind speed, and air pressure are collected to construct time-series data on environmental conditions.

4. The method for real-time intelligent monitoring of carbon emissions in industrial parks based on big data collection according to claim 3, characterized in that, Based on multi-dimensional carbon emission intensity sequences and production activity time-series data, a multi-level carbon emission quantification framework reflecting the park's operational status is constructed, where each level corresponds to a different carbon emission aggregation granularity, specifically including: Based on the physical distribution and logical connections of carbon emission sources within the park, a carbon emission aggregation hierarchy is defined from the equipment level, workshop level to the park level. For each aggregation level, extract the corresponding carbon emission intensity subsequence from the multi-dimensional carbon emission intensity sequence; Extract a series of production efficiency indicators that match each aggregation level from the time-series data of production activities; By integrating the carbon emission intensity subsequence and the production efficiency indicator sequence, a carbon emission operation status vector is formed under each aggregation level; A multi-level carbon emission quantification framework is constructed based on the carbon emission operational status vectors at all levels.

5. The method for real-time intelligent monitoring of carbon emissions in industrial parks based on big data collection according to claim 4, characterized in that, A summary of the current carbon emission status is generated based on the real-time collected dynamic emission data stream, and then mapped to the corresponding level within a multi-level carbon emission quantification framework, specifically including: The current dynamic emission data stream is subjected to the same instantaneous feature enhancement processing as historical data to obtain the current multidimensional carbon emission intensity sequence. Extract the current carbon emission operational status vector from the current multi-dimensional carbon emission intensity sequence; Identify the hierarchical affiliation of the current carbon emission operational status vector within a multi-level carbon emission quantification framework; The current carbon emission operating status vector is matched with the set of historical carbon emission operating status vectors of the level to which it belongs.

6. The method for real-time intelligent monitoring of carbon emissions in industrial parks based on big data collection as described in claim 5, characterized in that, By integrating real-time online carbon emission deviation indicators, multi-dimensional carbon emission intensity sequences, and environmental status time-series data, an adaptive quantitative mapping model is constructed, which specifically includes: Real-time carbon emission deviation indicators are used as dynamic weighting factors; Weighted fusion of recent data segments from multi-dimensional carbon emission intensity sequences and corresponding data segments from environmental status time series data; A temporal neural network architecture is used to process the weighted and fused data segments, and a mapping relationship from the current state to the future state is established. The parameters of the time-series neural network are adjusted by training with historical data to form an adaptive quantization mapping model; By using an adaptive quantization mapping model to extrapolate the short-term dynamic trajectory of carbon emissions in the park, the evolution path of carbon emission intensity within future time windows is output, specifically including: Input the current multidimensional carbon emission intensity sequence and real-time environmental status data into the adaptive quantization mapping model; The model recursively extrapolates the carbon emission intensity values ​​at multiple future time points, forming a carbon emission intensity evolution sequence; Based on the park's production plan data, adjust the carbon emission intensity values ​​corresponding to planned activities in the evolution sequence; Output the carbon emission intensity evolution path within the adjusted future time window; Based on the matching results between the carbon emission intensity evolution path and the preset carbon emission control boundary conditions, a tiered early warning signal and control strategy sequence are generated, specifically including: Load the multidimensional boundary condition set from the park's carbon emission control rule base; The evolution path of carbon emission intensity is compared point by point with the set of multidimensional boundary conditions; Identify the time periods and magnitudes of exceeding boundary conditions in the evolutionary path; Different levels of graded warning signals are generated based on the magnitude of the exceedance. For each warning signal, the corresponding control measures are retrieved from the control strategy library and sorted by priority to form a control strategy sequence.

7. A real-time intelligent monitoring system for carbon emissions in a park based on big data acquisition, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the real-time intelligent monitoring method for carbon emissions in industrial parks based on big data collection as described in any one of claims 1 to 6.

Images